View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Call for Papers

Bank

Federal
of Chicago

2012 Conference on
Bank Structure and
Competition

Third Quarter and Fourth Quarter 2011

Economic.

perspectives

Third Quarter
82

Evaluating the role of labor market mismatch
in rising unemployment
Gadi Barlevy

97

Emergence of immediate funds transfer as a generalpurpose means of payment
Bruce J. Summers and Kirstin E. Wells

Fourth Quarter
113

How do benefit adjustments for government transfer
programs compare with their participants’ inflation
experiences?
Leslie McGranahan and Anna L. Paulson

137

Clearing over-the-counter derivatives
Ed Nosal

147

Worker flows and matching efficiency
Marcelo Veracierto

Economic.

perspectives
President
Charles L. Evans

Executive Vice President and Director of Research
Daniel G. Sullivan
Senior Vice President and Economic Advisor
Spencer Krane
Senior Vice President, Financial Markets Group
David Marshall
Microeconomic Policy Research
Daniel Aaronson, Vice President

Macroeconomic Policy Research
Jonas D. M. Fisher, Vice President
Markets Team
Richard Heckinger, Vice President
Finance Team
Anna L. Paulson, Vice President

Regional Programs
William A. Testa, Vice President
Economics Editor
Richard D. Porter, Vice President

Editors
Helen O’D. Koshy
Han Y. Choi
Graphics and Layout
Rita Molloy

Proofreading
Sheila A. Mangier
Production
Julia Baker
Economic Perspectives is published by the Economic
Research Department of the Federal Reserve Bank of Chicago.
The views expressed are the authors’ and do not necessarily
reflect the views of the Federal Reserve Bank of Chicago
or the Federal Reserve System.

© 2011 Federal Reserve Bank of Chicago
Economic Perspectives articles may be reproduced
in whole or in part, provided the articles are not reproduced
or distributed for commercial gain and provided the source
is appropriately credited. Prior written permission must
be obtained for any other reproduction, distribution,
republication, or creation of derivative works of Economic
Perspectives articles. To request permission, please
contact Helen Koshy, senior editor, at 312-322-5830 or email
Helen.Koshy@chi.frb.org.
Economic Perspectives and other Bank
publications are available at www.chicagofed.org.

a chicagofed. org
ISSN 0164-0682

Announcement
Final Print Edition of Economic Perspectives

To our subscribers,
As part of our ongoing efforts to reduce costs and protect the environment, we have decided
to discontinue printing issues of Economic Perspectives. We will continue to publish the

journal online on a quarterly basis.

We encourage you to visit our website, www.chicagofed.org, to sign up for an online
subscription to Economic Perspectives. Our email alert service will notify you whenever
we publish a new article online.
Our website also features a complete archive of past issues of Economic Perspectives,
as well as other Federal Reserve Bank of Chicago publications.

We thank you most sincerely for your interest in our work.
Regards,

Helen Koshy,
Managing Editor
Economic Research Publications
Federal Reserve Bank of Chicago

& chicagofed. org

Contents
Third and Fourth Quarters 2011, Volume XXXV, Issues 3 and 4

Third Quarter
82

Evaluating the role of labor market mismatch in rising unemployment
Gadi Barlevy
This article shows how much a shock to the ability of firms to hire qualified workers accounts for the
rise in unemployment. Using a matching function approach, the author finds that such a shock implies
an unemployment rate of no more than 7.1 percent, much below the actual unemployment rate during
the past two years. Hence, the recent patterns in unemployment and vacancy data do not necessarily
rule out a role for monetary policy, which works by encouraging firms to hire more.

97

Emergence of immediate funds transfer as a general-purpose means
of payment
Bruce J. Summers and Kirstin E. Wells
Immediate funds transfer (IFT) is a highly convenient, certain, secure, and economical means of
payment using bank money. IFT is not available in the U.S. banking system, except for large-value
business payments, interbank transfers, and specialized financial market transactions. This article
examines the successful experience with IFT in Mexico, South Africa, Switzerland, and the UK
and concludes that payment system governance is the principal barrier to IFT innovation in the U.S.

Fourth Quarter
113

How do benefit adjustments for government transfer programs
compare with their participants’ inflation experiences?
Leslie McGranahan and Anna L. Paulson
The authors measure the inflation experienced by demographic groups that likely received benefits
from major government transfer programs during the period 1980-2010. They then compare the
group-specific inflation measures with the transfer programs’ benefit adjustments, which are typically
based on aggregate inflation. The extent to which the program benefits keep up with group inflation
differs across the programs and their targeted groups, depending on both the ways in which the
benefits are adjusted for price changes and the spending patterns of the various groups.

137

Clearing over-the-counter derivatives
Ed Nosal
Prior to the financial crisis of 2008, the over-the-counter derivatives market was not required to
“clear” transactions. This changed with the Dodd-Frank Act of 2010. This article explains how
clearing works and how it could improve liquidity and competition for over-the-counter derivatives.

147

Worker flows and matching efficiency
Marcelo Veracierto
Although job vacancies have increased quite significantly since mid-2009, the unemployment rate
has not declined significantly. This article analyzes the matching efficiency of jobs and workers in
U.S. labor markets and its impact on the behavior of the unemployment rate and other labor market
outcomes since the start of the latest recession.

170

Conference on Bank Structure and Competition: Call for papers

Evaluating the role of labor market mismatch
in rising unemployment
Gadi Barlevy

Introduction and summary
From the second half of 2009 through the end of 2010,
the U.S. labor market witnessed a systematic increase
in the rate ofjob openings while the unemployment rate
remained essentially unchanged. Some have argued
that, evidently, the problem in the labor market during
this period was not that firms were reluctant to hire
additional workers, but that, for whatever reason, firms
seemed unable to find suitable workers to staff the
positions they were trying to fill. By this logic, using
monetary policy to encourage further hiring by firms
would have been unlikely to drive down unemployment:
If firms were already trying to hire but could not, why
should policy actions that mainly serve to encourage
even more hiring have any impact on unemployment?
The unemployment rate did finally register a decline
in late 2010 and early 2011—a development that may
eventually render less acute the debate about the need
for monetary policy to address the problem of high
unemployment. Still, constructing a framework for
interpreting such labor market patterns and their policy
implications remains an important goal. This is espe­
cially true given that there have been other periods in
which job vacancy rates seemed to rise without a commensurately large fall in unemployment, although those
episodes were not as dramatic nor as long as the most
recent one.
In this article, I show how the labor market matching
function approach developed by Pissarides (1985) and
Mortensen and Pissarides (1994) can be used to assess
the validity of the proposition that recent trends in
vacancies and unemployment necessarily point to a
diminished role for monetary policy. More specifically,
I show that this framework indeed suggests that an
increase in unemployment without a commensurate
decline in vacancies can be indicative of a labor market
shock that monetary policy cannot offset. However,
this framework can also be used to derive a bound on

82

how much a shock of this type can affect unemployment.
Applying these insights to the period of the Great
Recession reveals that this type of shock by itself
would lead to an unemployment rate of 7.1 percent,
considerably lower than the unemployment rate during
most of this period. The higher actual unemployment
rate suggests that other types of shocks, which mone­
tary policy may be able to address, must also be oper­
ating. Hence, the recent patterns in unemployment and
vacancy data do not necessarily rule out an important
role for monetary policy. Whether more expansionary
monetary policy would have been beneficial is a ques­
tion that is beyond the scope of this article. Neverthe­
less, the matching function approach frames this question
in a potentially useful way—that is, as a question of why
the value of taking on additional workers appears to be
so much lower now than in normal economic times.
My article is organized as follows. I begin by
describing the matching function approach, and then
I show that the shocks that affect unemployment in
this framework can be decomposed into two groups—
those that affect the ability of firms to find and hire
qualified workers and those that affect the value to a
firm of taking on an additional worker. I next explain
how the model can be used to predict how a shock to
the ability of firms to hire, calibrated to match the facts
on unemployment and vacancies during the Great
Recession, affects unemployment. Using this result,
I argue that the increase in unemployment due to this
shock is much smaller than the actual increase in un­
employment during this period, so a shock to the abil­
ity of firms to hire cannot by itself account for the

Gadi Barlevy is a senior economist and research advisor in
the Economic Research Department at the Federal Reserve
Bank of Chicago. The author thanks Dan Aaronson, Dale
Mortensen, Ay§egiil §ahin, Dan Sullivan, and Marcelo
Veracierto for their helpful comments.

3Q/2011, Economic Perspectives

rise in unemployment during this time. I conclude
with a discussion about how measurement issues
are likely to affect these conclusions.

The matching function approach
In this section, I lay out the key features of the labor
market matching framework developed in Pissarides
(1985) and Mortensen and Pissarides (1994). This
framework rests on two key assumptions.
The first key assumption is that the total number
of new hires h in any given period can be expressed
as a function of the number of workers who are un­
employed during that period, u, and the number of
vacancies firms post over that same period, v:
1)

/? = zz; (z/,v).

This assumption is similar to the assumption in­
voked by macroeconomists that one can use an aggre­
gate production function to express total output produced
in a given period as a function of the total number of
hours and the aggregate capital stock for that period.
That is, the process by which unemployed workers
looking for jobs and employers with vacancies looking
for workers form new hires is assumed to operate
with such regularity that one can reliably predict the
number of new hires per period by using only data on
the number of unemployed workers and the number of
vacancies firms post. Empirical analysis supports the
idea that the number of new hires can be related to the
number of unemployed and vacant positions in a fairly
predictable way. Much of this evidence is summarized
in the survey by Petrongolo and Pissarides (2001).1
The function m is referred to as a matching func­
tion. A common restriction on the matching function is
that the number of new hires h falls short of both the
number of unemployed u and the number of vacant
positions v. That is, some unemployed workers and
some positions will remain unmatched by the end of
the period. This is meant to capture various frictions
in the process of filling new jobs from the ranks of
the unemployed—such as a lack of coordination that
leads multiple workers to apply to the same vacancies
while other vacancies remain unfilled, or the fact that
workers and firms do not initially know how well suited
they are for each other and figuring this out can be timeconsuming. Studies that explore these frictions in detail
reveal that they do not always give rise to empirically
plausible matching functions, and in some cases they
suggest different interpretations for why hiring, vacan­
cies, and unemployment are related. Moreover, these
frictions sometimes imply that the number of new
hires should depend on other variables besides just

Federal Reserve Bank of Chicago

the number of unemployed workers and the number of
vacant positions.2 However, Petrongolo and Pissarides
(2001) argue that the matching function approach per­
forms quite well empirically and is suitable for ana­
lyzing certain questions concerning the labor market,
just as assuming an aggregate production function is
often useful for analyzing macroeconomic questions.
That is, many macroeconomists are willing to posit
an aggregate production function that is invariant to
various shocks that affect the economy, even though
the conditions under which one can ignore the decisions
of individual firms in different sectors and express
aggregate output as a function of aggregate inputs are
quite stringent.3 In defense of this assumption, these
macroeconomists would argue that the aggregate pro­
duction function performs well empirically, so it is likely
to be useful in predicting how the economy would re­
spond to shocks that only affect the aggregate capital
stock and labor hours—for example, a change in in­
come taxes that affects how much labor is supplied
but does not affect the technology available for pro­
ducing goods and services. Likewise, advocates of
the matching function approach view the close empir­
ical relationship between aggregate hiring and aggre­
gate unemployment and vacancies as justification for
ignoring the decisions of workers and employers that
underlie the process of job creation. These advocates
proceed as if hiring can be summarized with a mapping
of aggregate vacancies and unemployment to aggre­
gate new hires that is “structural,” meaning that the
mapping is invariant to shocks that affect unemploy­
ment and vacancies but not the frictions inherent in
the matching process.
Petrongolo and Pissarides (2001) argue that the
matching function is particularly well approximated
by a Cobb-Douglas specification, that is,

2)

zz/(z/,v) =/h/'v1 “,

where A is a scale parameter that determines the pro­
ductivity of the matching process and ex reflects the
sensitivity of the number of new hires to the number
of unemployed workers. That is, this specification
will produce reasonably good predictions for the
actual hiring rate given the unemployment rate u
and vacancy rate v for coefficients a, and A that
remain stable over relatively long periods and that
can be estimated from historical data.
The second key assumption of the labor market
matching framework is that firms post vacancies as
long as doing so remains profitable, implying that the
expected discounted profits to a firm from posting a
vacancy should be zero in equilibrium. Let / denote

83

the value of a filled job to the employer who creates it,
and let k denote the cost of posting (and maintaining)
the vacancy, including screening and interviewing po­
tential candidates. Then the assumption that employers
are free to enter the labor market and attempt to hire
workers as long as it remains profitable to do so implies
that the value of a filled job times the probability of
filling it should equal the cost of posting the vacancy,
ensuring expected profits are equal to zero. Pissarides
(1985) and Mortensen and Pissarides (1994) assume
that each posted vacancy is equally likely to be matched,
so with zz?(zz,v) new hires, the probability of filling a
job is equal to m(u,v)/v. In this case, the implications
of free entry can be summarized as follows:
3)

=

v

Not all models that give rise to a matching function
representation as in equation 1 imply that the probability
that any given firm expects its vacancy to be filled with­
in the relevant period corresponds to m(u,v)/v. For ex­
ample, even when there are underlying differences in
firms, such as differences in the costs of processing ap­
plicants, we might still observe a stable relationship be­
tween aggregate hiring, unemployment, and vacancies.
But different firms will assign different probabilities to
filling their positions within the relevant period. Still,
as long as m(u,v)/v reasonably captures the probability
of filling a position for the marginal firm at any point in
time, proceeding with this assumption will be appropriate.
Substituting the Cobb-Douglas specification for
m(u,v) from equation 2 into the free-entry condition
as given by equation 3 reveals that the free-entry con­
dition can be expressed solely in terms of the ratio v/zz.4
This ratio is known in the literature as market tightness,
since it reflects how many vacant positions are competing
for each unemployed worker. In particular, the free-entry
condition as given by equation 3 can be written as

4,

^p = t.

As long as the parameters a and A remain con­
stant over time, the free-entry condition as given by
equation 4 tells us that if the value of a filled job rela­
tive to the cost of posting a vacancy, J/k, varies over
time for any reason (such as a change in aggregate
demand or changes in aggregate productivity), the
market tightness ratio, v/zz, would have to change as
well. The fact that v/zz changes with J/k ensures that
firms continue to expect zero cumulative discounted
profits from posting additional vacancies. Hence, if

84

we knew how a particular macroeconomic event
affected the ratio J/k, we could use the free-entry con­
dition as given by equation 4 to deduce how this event
should change the ratio of the vacancy rate to the un­
employment rate we observe in the labor market. The
two assumptions—the existence of a matching function
and free entry into the labor market—thus impose a lot
of structure on how various shocks affect labor market
tightness as reflected in the ratio v/zz.5

The Beveridge curve
With the introduction of one additional assumption,
the labor market matching framework can be used to
predict not only how v/zz changes with J/k, but also how
zz and v change individually. In particular, suppose that
the rate at which employed workers separate from jobs
into unemployment is constant over time. This assump­
tion may seem implausible at first, especially given the
incidence of mass layoffs during recessions. However,
the job separation rate that I need to assume is constant
does not involve one-off spikes of job destruction that
reflect immediate adjustment by firms to changes in
economic conditions. Rather, the relevant rate is the
one that corresponds to what happens in a recession
once all bursts of job destruction are done.6 Shimer
(2005a) and Hall (2005) argue that fluctuations in this
separation rate contribute little to overall changes in
unemployment and can be ignored. In subsequent work,
others argue that the separation rate appears to be quite
cyclically sensitive, and find the separation rate makes
an important but still relatively small contribution to
overall fluctuations in unemployment.7 However, their
papers all look at the role of flows of workers from
employment and unemployment without accounting
for spikes of job destruction. Flows into unemployment
that include bursts of job destruction may account for
fluctuations in total unemployment, even if the separation
rate that is relevant for my analysis is fairly stable. Later,
I argue that data on unemployment and vacancies from
three distinct episodes of high unemployment support
the claim that the relevant separation rate, s, does not
rise much during recessions. Moreover, if the separa­
tion rate were in fact higher during recessions, my cal­
culation would only exaggerate the role of labor market
mismatch, and the bound I derive for the effect of a
shock to the ability of firms to hire would be too high.
To see what the model predicts for the behavior
of v and zz as opposed to their ratio v/zz, consider what
happens if J/k varies over time. Conditional on a given
value of J/k, the free-entry condition as given by equa­
tion 4 tells us that the vacancy-to-unemployment ratio,
v/zz, must remain constant as long as J/k is constant.
However, zz and v could themselves change even while

3Q/2011, Economic Perspectives

Jlk remains fixed, as long as they change in the right
proportion. Still, one can show that as long as Jlk remains
fixed, u and v will converge to some steady-state values
that depend on Jlk and, moreover, that this convergence
will be rapid. This quick pace of convergence is not
just a theoretical result; rather, it has been confirmed
empirically.8 This finding may seem odd at first, since
time-series data suggest unemployment is fairly per­
sistent over time. However, it is important to note that
I am referring to conditional (as opposed to uncondi­
tional) convergence in u and v. In other words, given
a value of Jlk, both u and v converge quickly to the
steady-state values associated with this particular value
of Jlk. But if J/k follows a persistent process, unemploy­
ment will still appear to change slowly over time. Rapid
conditional convergence is thus fully consistent with
unemployment appearing to be a slow-moving process.
Given that convergence to a steady state for a given
Jlk is quick, it follows that whatever the value of Jlk
happens to be at any point in time, the values of u and
r we would observe should roughly coincide with the
steady-state levels of these variables for that J/k.
To compute the conditional steady-state unemploy­
ment for a given Jlk, note that the flow into unemploy­
ment is equal to <1 -u), where .v denotes the separation
rate into unemployment, while the flow out of unem­
ployment is equal to the number of new hires, AiC v1_“.9
Since flows into and out of unemployment are equal
in steady state, I can use this equality to arrive at an
implicit fonnula for the conditional steady-state unem­
ployment rate associated with a particular v/u ratio,
which is associated with a particular value of J/k.

s + A(vlufa'

Rearranging equation 5 allows me to express the va­
cancy rate v implied by the model for a given unem­
ployment rate u as follows:

1

As long as the separation rate into unemployment
v is constant, equation 6 implies a negative relationship
between u and v. This relationship, when displayed
graphically as a plot of the vacancy rate against the
unemployment rate, is known as a Beveridge curve after
the British economist William Beveridge, who first docu­
mented the negative relationship between the two series.
The negatively sloped Beveridge curve can be seen
by plotting out the u and v implied by the model for

Federal Reserve Bank of Chicago

different values of J/k. hi particular, according to
equation 4, changes in Jlk will force changes in the
ratio v/u. Intuitively, as jobs become more valuable,
the probability of filling a job must fall to ensure firms
still expect to earn zero profits. One can then deduce,
from equation 5, that higher values of the ratio v/u
imply lower values of u and, from equation 6, that
lower values of u imply higher values of v.
Indeed, the only thing that induces a movement
along a Beveridge curve as defined by equation 5 is a
change in Jlk. This result holds because the Beveridge
curve in equation 5 is defined as the relationship between
u and v for fixed values ofA, a, and s. When these values
are fixed, it is apparent from equation 5 that the unem­
ployment rate u only changes if the ratio v/u changes.
But when A and a are fixed, the free-entry condition
as given by equation 4 tells us that the ratio vlu is en­
tirely determined by Jlk. Thus, a movement along the
Beveridge curve occurs if and only if the value of taking
on an additional worker relative to the cost of posting
a vacancy changes. Various events can shift this value,
including a change in worker productivity, a change in
the bargaining power of workers, a change in aggregate
demand, and a change in the employer’s operating cost
(such as a change to the cost of borrowing). But, for
our purposes, all of these events can be grouped into
a catchall category of shocks that affect the net value
of a filled job or, alternatively, shocks that move the
economy along a stable Beveridge curve.
The natural counterpart to shocks that induce a
movement along a Beveridge curve are shocks that shift
the Beveridge curve itself. As evident from equation 5,
which defines the Beveridge curve, as long as .v is fixed,
the only way for the Beveridge curve to shift is if the
matching function m(u,v) itself somehow changes.
A shift in the Beveridge curve thus corresponds to a
shock that changes the way in which workers and em­
ployers come together to form new hires. One example
of such a shock is a disruption that gives rise to greater
mismatch between the skills employers require to fill
their positions and the skills that unemployed workers
currently possess—such as a shift in demand away from
products the labor force is already skilled at making.
Such a shock would presumably result in fewer positions
being filled given the same number of unemployed
workers and vacant positions, and thus, the productivity
term A in the matching function would decline. The
model thus delivers a clean dichotomy: Shifts of the
Beveridge curve correspond to shocks to the ability
of firms to hire (that is, changes in A), while movements
along a fixed Beveridge curve correspond to changes
in the incentives for firms to hire (that is, changes in Jlk).

85

We can now recast the debate on the role of mon­
etary policy in the face of high unemployment, using
the terminology of the matching function approach.
The observation that vacancies rose while unemploy­
ment was virtually unchanged implies the Beveridge
curve must have shifted, that is, the hiring process be­
came less efficient. There is arguably little monetary
policy can do to affect the process by which firms and
unemployed workers match up to generate new hires.
However, whether there is any role for monetary policy
depends on whether a shock to match productivity,
A, is the only shock responsible for high unemploy­
ment. If the increase in unemployment is also due to
a change in the relative value of a filled job, J/k, there
may be some scope for monetary policy after all. So,
for example, if the lower J/k reflects weak aggregate
demand due to some underlying frictions, then mone­
tary policy would have a role in addressing this. In
the remainder of this article, I infer the decline in A
from the shift in the Beveridge curve, and then I use
the matching function approach to gauge how much
a shock to A of this magnitude should have raised un­
employment if the parameters that govern the value
of a filled job remain equal to their levels during nor­
mal economic times (that is, to pre-recession levels).
Since the implied unemployment rate falls far short
of the actual unemployment rate that prevailed during
this time, the high unemployment rate suggests that
the relative value of a filled job, J/k, must have been
lower during this period than during normal econom­
ic times. Whether this finding admits a role for mone­
tary policy depends on why the value of a filled job is
lower. Still, the calculation suggests that data on un­
employment and vacancies do not rule out a role for
policy per se and that high unemployment is due not
only to an inability to hire among employers but also
to a reduced willingness to hire.

Empirical Beveridge curves and
estimating the matching function
The first step in my analysis involves inferring
the reduction in match productivity A over this period
from shifts in the Beveridge curve. For this, I must
begin with a benchmark value for A in normal times.
I can do this by fitting the Beveridge curve relation­
ship in equation 6 to data from before the recent crisis.
That is, I estimate the parameters ex and ,4 of the matching
function, using data only for the period before unem­
ployment began to take off, and then I look at how
this relationship holds up in predicting vacancy rates
for observed unemployment rates once the unemploy­
ment rate begins to climb. To do this, I use data from
the U.S. Bureau of Labor Statistics’ Job Openings

86

and Labor Turnover Survey (JOLTS), which begins
in December 2000. To estimate the Beveridge curve,
I use data through August 2008, just before the big
run-up in unemployment that started a few months
after the official start date of the recession according
to the National Bureau of Economic Research (NBER).
To estimate ex and A, I follow Shimer (2005b)—who
estimates the job separation rate at a monthly frequency—
and set s = 0.03. However, the choice ofs is essentially
a normalization.10 To infer A and ex, I set out to match
two specific aspects of the data. First, for each month,
I use equation 6 to predict a vacancy rate v given an
unemployment rate u in that month. The parameters
A and ex were chosen to ensure that the average predicted
vacancy rate over all these months was equal to the
actual average vacancy rate over the same period,
namely, 0.029. Second, I chose the parameters to
ensure that the difference in v between the start date
and end date of my series, which is 0.013 in the data,
matched the difference in the predicted v at these two
dates. Matching the model and the data this way yielded
values A = 0.75 and ex = 0.46. The implied (fitted)
Beveridge curve corresponds to the dark gray line in
panel A of figure 1, which is shown together with data
on unemployment and vacancies. The points in black
correspond to the data through August 2008 that were
used to estimate the curve, while the points in red cor­
respond to observations from September 2008 onward
that were not used in estimating the curve. To help
illustrate how u and v evolved from September 2008
onward, consecutive months are connected.
As evident in panel A of figure 1, the Beveridge
curve implied by the model does a reasonable job
initially of predicting the vacancy rate at each unem­
ployment rate beyond the period it was estimated to
match—in fact, beyond the historical range of both
the unemployment and vacancy series used to estimate
the curve. The forecast only starts to break down around
August 2009, suggesting a change in the matching
function. The fact that the curve fits well throughout
the official recession as determined by the NBER—
that is, from December 2007 through June 2009—and
only breaks down afterward provides some reassurance
that the separation rate, s, did not appear to rise sig­
nificantly while the economy was contracting.
As a further check on how well the matching
function approach fits the data, I went back and re­
peated the same exercise for two other periods with
similarly high unemployment—namely, for November
1973-March 1975 and for January-July 1980 and
July 1981-November 1982. Since JOLTS only begins
in December 2000,1 use the Conference Board’s
Help-Wanted Advertising Index for my measure of

3Q/2011, Economic Perspectives

FIGURE 1

Actual and predicted job vacancy rates for given unemployment rates during selected episodes
A. December 2000-February 2011
vacancy rate

B. July 1978-May 1984

unemployment rate
•

—•—

Data through August 2008
Data from September 2008 onward

•

Data through December 1979

—Data from January 1980 onward

Fitted Beveridge curve

Fitted Beveridge curve

Fitted mismatch curve

Fitted mismatch curve

C. May 1972-September 1976
vacancy rate

unemployment rate
•

—•—

Data through October 1973
Data from November 1973 onward
Fitted Beveridge curve
Fitted mismatch curve

Notes: All curves are fitted only on data indicated by the black points. The fitted mismatch curves are based on Shimer (2007).
See the text for further details.
Sources: Author’s calculations based on data from the U.S. Bureau of Labor Statistics, Job Openings and Labor Turnover Survey
and civilian unemployment rate series; and Conference Board, Help-Wanted Advertising Index, from Haver Analytics.

vacancies. This index is constructed using the number
of newspaper advertisements for vacant positions. To
transform this index into a vacancy rate, I normalized
the series to coincide with the JOLTS vacancy rate for
the period in which they overlap. For each recession,
I followed a similar approach to estimating the matching
function—that is, by taking data from a period prior
to the recession to estimate the function and then see­
ing how the implied Beveridge curve does during the
recession. However, since the Conference Board’s HelpWanted Advertising Index may be unreliable over long

Federal Reserve Bank of Chicago

periods (given various gradual changes in the tenden­
cy of employers to rely on newspaper advertising for
recruiting), I restrict attention to shorter periods for my
estimation. For the 1973-75 recession, I look at the
18-month period before the NBER peak date, that is,
May 1972 through October 1973. For the 1980 and
1981-82 recessions, I look at the 18-month period be­
fore the NBER peak date for the 1980 recession, that is,
July 1978 through December 1979. In both cases, I esti­
mate A and a in the same manner as for the data between
December 2000 and August 2008—that is, I choose

87

these parameters so that the average predicted vacancy
rate over the period is equal to the actual average and
the difference in vacancy rates between the start and
end dates is the same in the predicted series as in the
actual series. The estimated coefficients are reported
in table 1, and the implied (fitted) Beveridge curves
are illustrated as dark gray lines in the panels of fig­
ure 1 (the light gray lines are explained in the next
paragraph). For the 1973-75 recession (panel C) and
the 1980 and 1981-82 recessions (panel B), the data
points used to estimate the curves are depicted in
black, while the remaining data points are depicted in
red. Note that in both panels B and C of figure 1, the
original Beveridge curves are estimated from a period
with little variation in the data, especially in the case
of the 1980 and 1981-82 recessions. Still, the approach
to using data before the recession(s) to estimate a curve
performs well. For all of the recessionary periods I con­
sider, the vacancy rate predicted for a given unemploy­
ment rate remains close to the actual vacancy rate
once unemployment begins to rise. In both panels B
and C of figure 1, the Beveridge curves do eventually
appear to shift, although the shifts are much smaller
than in the most recent episode, shown in panel A of
figure 1. As evident in table 1, the coefficient a is es­
timated to be essentially the same in all three periods.
This is consistent with my maintained approach of as­
suming the parameter a is fixed and that any changes
in the matching function must therefore be attributed
to A, the match productivity parameter.
For comparison, I also considered an alternative
explanation for the matching function based on the
notion of mismatch advanced by Shimer (2007), which
was used by Kocherlakota (2010) to analyze the same
labor market trends I consider here. The Shimer (2007)
mismatch model offers a different interpretation for
the relationship between new hires and unemployment
and vacancies, and leads to a different zero-profit con­
dition from equation 3.11 Shimer’s (2007) model also
involves two parameters, which he denotes m and «.
As I did earlier, I use the period before the recent run­
up in unemployment (and the 18-month period before
the start of the NBER recession for the two earlier epi­
sodes) to estimate these parameters and then consider
how the model performs when unemployment rises.
Following Kocherlakota (2010), in each case I choose
these parameters to match the average values of u and
v in the earlier period that is meant to reflect normal
economic times. The estimates for the two parameters
in each of the three episodes are summarized in table 2,
and the implied curves relating unemployment and
vacancies are shown in the respective panels in figure 1
in light gray. In all three episodes, Shimer’s (2007)

88

TABLE 1

Estimated parameters for a Cobb-Douglas
matching function
A
May 1972-October 1973

0.68

0.42

July 1978-December 1979

0.56

0.43

December 2000-August 2008

0.75

0.46

Note: See the text for further details.
Sources: Author’s calculations based on data from the U.S.
Bureau of Labor Statistics, Job Openings and Labor Turnover
Survey and civilian unemployment rate series; and Conference
Board, Help-Wanted Advertising Index, from Haver Analytics.

TABLE 2

Estimated parameters for the Shimer (2007)
mismatch model
m

n
165.8

May 1972-October 1973

168.7

July 1978-December 1979

119.7

118.0

December 2000-August 2008

210.5

205.5

Note: See the text for further details.
Sources: Author’s calculations based on data from the U.S.
Bureau of Labor Statistics, Job Openings and Labor Turnover
Survey and civilian unemployment rate series; and Conference
Board, Help-Wanted Advertising Index, from Haver Analytics.

mismatch model predicts that the vacancy rate should
decline more rapidly with unemployment than either
what my estimated Beveridge curves predict or what
we actually observe in the data. Since a shift in the
curve in Shimer’s model can be thought of as a shock
to the ability of unemployed workers and job vacancies
to match up with one another, this would suggest that
all three recessionary periods and their subsequent re­
coveries were associated with significant rises in mis­
match. While this reading of the data is certainly possible,
it is striking that much of the discussion of the role of
labor market mismatch during the Great Recession has
tended to treat this phenomenon as exceptional; many
of the explanations for the rise in mismatch in the labor
market over the course of the Great Recession have em­
phasized features that are unique to this episode, such
as the unprecedented collapse in house prices. Such
views seem at odds with a specification that implies all
three recessionary periods were associated with similarly
large increases in labor market mismatch. The matching
function approach is therefore more consistent with
the view that the most recent episode is exceptional.

3Q/2011, Economic Perspectives

curves for both values of Ao (the dark
gray line) and At (the light gray line).
Fitting Beveridge curves during the Great Recession
In principle, I can fit a new Beveridge
curve through any data point. The most
recent observation at the time of this
writing, for February 2011, lies on the
same Beveridge curve implied by the
data from December 2010, as evident in
figure 2. Moreover, this curve is close to
the highest curve one could fit through
any of the data points between September
2008 and the end of the JOLTS sample.
This leads me to focus on the curve that
runs though the data point corresponding
to December 2010 in measuring the de­
cline in match productivity A.
An alternative way to infer the change
in A over the course of the Great Recession
would be to bring in additional data on new
hires rather than only rely on the data for
unemployment rate
unemployment and vacancies. The idea is
•
Data, December 2000-August 2008
as follows. Since zm(m,v) corresponds to the
G
Data, September 2008-February 2011
number of new hires, which is measured
Beveridge curve, December 2010
in JOLTS, I can take the number of new
Beveridge curve, December 2000-August 2008
hires and divide by the expression u"v' ",
Note: See the text for further details.
using my previous estimate of a = 0.46.
Sources: Author’s calculations based on data from the U.S. Bureau of
In principle, this should give me a time
Labor Statistics, Job Openings and Labor Turnover Survey and civilian
unemployment rate series, from Haver Analytics.
series for match productivity A. This im­
plied time series is depicted in figure 3.
If I consider the period between August
Inferring the extent of mismatch
2008 and December 2010, the implied match produc­
tivity declined by about 20 percent—a little larger
After estimating the parameters associated with
than what I get without using hiring data and looking
the Beveridge curve for normal economic times, I next
only at the implied shift in the Beveridge curve. How­
turn to how the decline in match productivity A can
ever, as evident from figure 3, match productivity us­
be inferred from the apparent shift in the Beveridge
ing
data on new hires starts to fall around December
curve following the most recent recession. Using data
2007,
considerably before any indications of a shift in
through August 2008,1 know that the initial productivity
the
Beveridge
curve relating unemployment and vacan­
of the matching function is given by Ao = 0.75.1 can
cies.
The
decline
in match productivity between
deduce the value of At needed to match a given unem­
December
2007
and
December 2010 is thus much larger,
ployment and vacancy pair at any other point in time
on
the
order
of
25-30
percent. However, this decline
by using equation 6. For example, to match the data
is
sensitive
to
the
value
of a, and it corresponds to
for December 2010, when u = 0.094 and v = 0.022,
20
percent
if
I
set
a
=
0.40
instead of 0.46. But regard­
match productivity A j must solve
less of the precise value for a, data on hires suggest
the decline in match productivity begins much earlier
1
0.54
than the shift in the Beveridge curve. That said, both
0.03
7) 0.022 =
0.094
—0.094
the magnitude and timing of the decline of matching
A,
efficiency depend on the measure of new hires used.
Bamichon and Figura (2010) use the flows from un­
Solving for Aj yields At = 0.633, that is, by
employment to employment rather than all new hires,
December 2010 the productivity of the matching
and find that the decline in A in 2009 is sharp, and its
function declined 16 percent from its original level
magnitude is comparable to what I estimate using the
before the recession. Figure 2 shows the Beveridge
Beveridge curve.
FIGURE 2

Federal Reserve Bank of Chicago

89

Veracierto (2011) reviews several
different approaches to estimating the
productivity of the matching function,
based on shifts of the Beveridge curve.
These include accounting for flows into
and out of nonemployment (see note 1,
p. 94), measuring new hires based on
flows into employment from either just
unemployment or both unemployment
and nonemployment, and using either
shifts of the Beveridge curve or a com­
parison of changes in new hires to changes
in unemployment and vacancies to de­
duce a time series for A. His preferred
estimate suggests A had declined 15 per­
cent since December 2007, in line with
the estimate I infer from the shift in the
Beveridge curve.
Since my calculations rely on the
Beveridge curve specification in equation 5,
I will use the estimate for the change in
A based on how much the Beveridge curve
shifted during the Great Recession in
what follows.

The effects of mismatch
on unemployment
Once I determine that match productivity A de­
clined by 16 percent between the level I estimate for
normal economic times and the end of 2010,1 can
determine the effect of a shock of this size on the un­
employment rate. To do this, I start at a steady-state
unemployment rate of 5 percent, which roughly cor­
responds to the historical average of unemployment
for the period covered by JOLTS through August 2008.
From the Beveridge curve relationship implied by
equation 6,1 know the implied vacancy rate would
have to be

will, affect these values. However, for reasons I explain
in more detail later, changes in A are likely to move
Jlk in a particular direction, implying that the unem­
ployment rate holding Jlk fixed will correspond to an
upper bound on unemployment. Assuming Jlk is con­
stant thus offers a useful benchmark case.
Rearranging equation 4,1 get - =
v
Hence, given the estimated decline in the productivity
of matching, holding J and A-fixed, a decrease in A from
0.75 to 0.633 should lead the unemployment-to-vacancy

0

— (O.O5-046 — O.O5054
0.75y

0.03.

The implied ratio of ulv during these normal times

will therefore equal

= 1.67.

(

y/o.46

—:—

=1.45.

0.633
Given I needed the ratio ulv to equal 1.67 to support
a 5 percent unemployment rate under the original
Beveridge curve, I can deduce that the new equilibrium
ratio of ulv will equal

0.03

Next, I use the free-entry condition as given by
equation 4 to deduce how much a shock to A will affect
the ratio w/v. To do this, suppose the shock to A had no
effect on the ratio of the value of a filled job to the cost
of posting a vacancy, Jlk. In fact, J and k are determined
endogenously, and changes in A can, and in many cases

90

1.45 x 1.67 = 2.42.

Plugging in ulv = 2.42 into the Beveridge curve
relationship in equation 5 when At = 0.633 gives us
the implied unemployment rate that must prevail in
the new equilibrium:

3Q/2011, Economic Perspectives

0.03 + 0.633(2.42)-°54
Thus, a shock to A, calibrated to the magnitude
implied by the patterns observed in data on unemploy­
ment and vacancies alone, will raise the unemployment
rate to 7.1 percent as long as it leaves the value of a
filled job unchanged. Since 7.1 percent is much lower
than the actual unemployment rate, this value suggests
that shocks to the productivity of matching alone can­
not account for the high unemployment rate.
Figure 4 illustrates the same calculation graphically.
Each level of match productivity A is associated with
a distinct Beveridge curve and a distinct ratio w/v deter­
mined by the free-entry condition as given by equation 4,
which in the figure corresponds to the line emanating
from the origin. The original Beveridge curve and
free-entry condition associated with A = Ao are shown
in dark gray, while the new Beveridge curve and freeentry condition associated with A = Ax are shown in light
gray. A decline in A not only shifts the Beveridge curve
but also rotates the free-entry condition clockwise to

Federal Reserve Bank of Chicago

a degree that depends on the size of a.
Intuitively, if hiring becomes less effective,
firms will have an incentive to post fewer
vacancies per unemployed worker, ulti­
mately leaving more workers unemployed.
As I noted earlier, both k and J are in
fact determined endogenously and will
likely change when A does. For example,
the process of creating a vacancy requires
productive inputs such as labor, so the
cost k will depend on wages that are de­
termined endogenously. Since wages tend
to rise and fall with economic activity both
in the data and in the original Mortensen
and Pissarides (1994) model, I would ex­
pect the cost of posting a vacancy k to
fall as the unemployment rate rises. As
for the value of a filled job to an employer
J, there are various reasons to suspect it
will be higher when there is more unem­
ployment. Mortensen and Pissarides (1994)
posit that the value of a filled job is de­
termined as the result of Nash bargaining
between workers and firms over the sur­
plus from a match.12 But the surplus from
matching is higher when v/u is low, so a
fall in A will lead to a higher value for
J.'3 Intuitively, when it is easy to find a
match, matching immediately is only
slightly more valuable than separating and
letting the two parties search for new
matches, which they will likely find quickly. More gener­
ally, various realistic features that are absent in the bench­
mark model, such as curvature in the utility function
and diminishing returns to labor, would tend to make
a marginal job more valuable when fewer workers are
employed. Essentially, diminishing marginal utility or
diminishing marginal returns make another employed
worker more valuable when fewer workers are employed.
If both k falls and J rises at higher unemployment
rates, the effect of a shock to A on w/v would only be
smaller. As such, 7.1 percent should be viewed as an
upper bound rather than a point estimate.14 This result
only reinforces the point that the high unemployment
rate that was observed during this period should not
be blamed solely on a decline in the ability of firms to
fill their positions, but also on greater reluctance among
firms to hire as reflected in a lower J/k.

Measurement issues
The calculations presented in the preceding sec­
tion are based on the assumption that the inputs that
go into creating new matches—namely, unemployment

91

and vacancies—are measured accurately. However, there
are reasons to suspect both series may systematically
misrepresent the nature of inputs that enter the hiring
process while the empirical Beveridge curve shifted.
I now discuss each of these series in turn, as well as
the implications of mismeasurement for my analysis.
I will argue that in both cases, measurement issues
only strengthen the conclusion that the decline in the
ability of firms to hire cannot by itself account for the
bulk of the increase in unemployment during this period.
I first consider the unemployment series. One
distinguishing feature of the current episode of high
unemployment is the exceptionally long duration of
unemployment insurance (UI) benefits; in some U.S.
states, the unemployed can receive UI benefits for up
to 99 weeks. Indeed, several research papers have sought
to estimate the effect of these extensions on both the
unemployment rate and unemployment durations.15 The
extension of UI benefits can matter for my analysis in
several ways, including the method by which unemploy­
ment is measured. First, though, it will be useful to re­
view the various ways in which explicitly incorporating
UI benefits into the model can matter for unemployment.
One reason UI benefits can matter is that they
lower the cost of remaining unemployed, allowing
workers to be more selective about which job they
take. As a result, relative to the case in which UI ben­
efits remained unchanged, unemployed workers will
prefer to continue searching more often, and a smaller
fraction of the contacts between unemployed workers
and vacant positions will result in a match, that is, a
new hire. Indeed, this provides one potential explana­
tion for the apparent decline in match productivity.
Note that this effect is already taken into account in
the calculation I sketched out before; that calculation
tells us how much a decline in the ability of firms to
hire—for whatever reason—ought to affect unemploy­
ment. Indeed, all the papers estimating how much the
extension of UI benefits contributed to unemployment
find effects that are smaller than the bound I estimate.
Second, when a worker and an employer agree
to form a match, the extension of UI benefits may
require an employer to offer a worker higher wages
given that more generous UI benefits improve the
bargaining position of workers. This effect is empha­
sized in Kocherlakota (2011). Unlike the first effect
that appeared as a lower value for A, this effect would
show up directly as a lower value for J, the value of
filling a job to an employer. Indeed, this may be one
reason for why the value of a filled job to an employer
appears to be lower now than it is in normal times.
Neither of these two effects poses a problem for
determining whether the rise in unemployment can be

92

attributed solely to a decline in the ability of firms to
hire. Rather, they merely suggest potential interpreta­
tions for what might be driving shocks to A or J.
However, there is a third potential implication of ex­
tending UI benefits that may act to distort measured
unemployment and could pose a problem for my calcu­
lation. In particular, extended UI benefits may encour­
age disaffected workers who prefer to leave the labor
force to present themselves as nominally unemployed
in order to qualify for UI benefits. This will be the
case even if such workers are not actively looking for
a job beyond whatever token steps are needed to main­
tain their status. Such a phenomenon would make the
measured unemployment rate seem higher than its
true value. Formally, let u* denote the fraction of the
labor force that is actively looking for jobs, and let u
denote the fraction of the labor force that is not really
looking for a job but reports itself as being unemployed.
If the latter fraction literally takes no steps to search
for a job, the matching process will only partner up
the true unemployment and vacancy positions, and
the number of hires will be given by

8)

h = m (m’,v).

At the same time, the official unemployment series
will correspond to u = u + u0, leading us to expect
m(u' + u ,v) hires. Since the matching function is in­
creasing in both arguments, this will make the matching
process appear less efficient than it truly is: We would
observe surprisingly few hires given the seemingly large
number of unemployed. Hence, the decline in match
productivity A inferred from the shift in the Beveridge
curve would exceed the true decline in A that enters
the free-entry condition as given by equation 4. Since
my approach provides an upper bound on the effect
of a decline in the ability to hire on unemployment,
though, overstating the decline in match productivity
A will not overturn my results. If anything, it suggests
the unemployment rate that should be expected from
the decline in the ability of firms to hire is actually
smaller than 7.1 percent.
Next, I turn to the time series for vacancies. Recent
work by Davis, Faberman, and Haltiwanger (2010)
has called into question whether vacancies provide a
consistent measure of recruiting effort over time. In
particular, they show that the vacancy yield, or the
ratio of hires per vacancy, varies systematically across
employers. For example, growing firms seem to be
better at hiring, in the sense of being able to hire more
workers per each vacancy posted. Davis, Faberman,
and Haltiwanger (2010) argue that this pattern arises
because the process of hiring requires firms to invest

3Q/2011, Economic Perspectives

some effort into recruiting beyond posting the number
of vacant positions they are seeking to fill.16 They fur­
ther reason that the same pattern should also occur over
the business cycle: In recession times, when overall
hiring is low, firms are likely to put in less effort into
recruiting than in boom times. Thus, employers’ hiring
efforts would decline by more than would be reflected
in the time series for the number of vacancies posted.17
Davis, Faberman, and Haltiwanger (2010) formal­
ize these concerns as follows. Suppose that the effort
that firms invest in recruiting can be summarized by
the product of q and v, where q denotes recruiting
intensity and v denotes the vacancy rate. The total
number of hires is then given by

9)

/? = m (u,qv).

That is, matching depends not on the number of va­
cancies, but vacancies together with how much firms
invest in filling these vacancies. When recruiting in­
tensity q falls below its historical average, the time
series for vacancies v will fail to register this and will
therefore overstate the overall recruiting effort. Using
vacancies to proxy for recruiting efforts will then make
matching efficiency appear to fall more than it in fact
does. That is, we may wrongly conclude that firms
find it more difficult to hire when in fact they are vol­
untarily choosing to search in a way that reduces the
odds of hiring. Once again, this will cause us to over­
state the decline in match productivity ,4 from apparent
shifts in the Beveridge curve and, therefore, to overstate
the increase in the unemployment rate that can be attrib­
uted to less efficient matching now than in the past.
To provide a more quantitative illustration of this
result, I can use the suggestion in Davis, Faberman, and
Haltiwanger (2010) ofproxying for recruiting intensity
q by using the way in which the vacancy yield (hires per
vacancy) varies across firms with different hiring rates.
In particular, using variation in the vacancy yield across
firms, they conclude that the elasticity of q with respect
to overall firm hiring is given by 0.72. This implies set­
ting q = h012. Davis, Faberman, and Haltiwanger (2010)
provide some evidence that this modification improves
the time-series fit of the matching function. If this proxy
is accurate, I can simply repeat the calculation for how
much the apparent decline in match productivity should
have increased unemployment, but replace the vacancy
rate v in equations 1 through 6 with /?0 72v as the second
argument in the matching function.18 Fitting a Beveridge
curve to data on unemployment and this adjusted va­
cancy series through August 2008 yields Ag = 0.7 and
ex = 0.54. To match the data for December 2010 requires
= 0.605, which is a smaller decline of only about

Federal Reserve Bank of Chicago

14 percent. For this decline, the implied unemployment
rate due to just this shock to match productivity would
be at most 6.3 percent. Correcting for measurement
problems in vacancies can thus have a significant impact
on how much unemployment is attributed to reduced
effectiveness in hiring.

Conclusion
Recent labor market trends have raised concerns
that the unemployment rate is high not because employ­
ers are reluctant to hire but because they are unable to
hire—that is, for whatever reason, firms are unable to
find suitable workers to staff the positions they are trying
to fill. These concerns, if true, would cast doubt on using
monetary policy to stimulate the labor market, since it
works by encouraging firms to hire more. The matching
function approach pioneered by Pissarides (1985) and
Mortensen and Pissarides (1994) offers a framework for
analyzing these issues. In particular, that framework can
be used to separate the shocks that drive unemployment
into two groups: shocks that affect the probability of
finding a suitable worker and shocks to the value a
worker generates once hired. The same framework
allows us to estimate how much the probability of
finding a worker declined and to compute a bound on
how much this effect by itself would raise the unemploy­
ment rate. This bound as I have calculated it suggests
that a decline in the ability to hire accounts for less than
half of the total rise in unemployment during the Great
Recession and that part of this rise in unemployment
must be because firms find hiring less profitable.
While there is little monetary policy can do if firms
find it more difficult to find suitable workers, there
may be scope for monetary policy when firms find it
less profitable to hire workers than during normal times.
Whether such a role for monetary policy is warranted
depends on why the value of a filled job to an employer
is lower than in normal times. For example, if filled
jobs are less valuable because of a shock that makes
workers less productive, there is arguably little that
monetary policy should do in response. But if jobs are
less valuable because of insufficient aggregate demand
on account of some market friction, there may be a
role for monetary policy to stimulate demand. The key
question for policy, then, is not what unemployment
and vacancy data tell us about the possibility of mis­
match, but why firms seem to find hiring workers less
attractive than usual. Unfortunately, while the matching
function approach is useful in pointing out the value
of a filled job to an employer as an important variable,
it offers little direct guidance as to why this value is
so much lower now relative to normal times.

93

NOTES
’Of course, newly hired workers do not come only from the ranks
of the unemployed; some were employed elsewhere, while others
were not employed but did not report actively looking for a job
either (that is, they were classified as “not in the labor force” by
the U.S. Bureau of Labor Statistics, per the definition available at
www.bls.gov/cps/cps_htgm.htm#nilf). In practice, the hiring rate
can be accounted for quite well using data on unemployment, per­
haps because the number of hires from out of the labor force and
the number of hires of already employed workers move in opposite
directions over the business cycle and tend to offset one another.
One way to avoid the logical inconsistency of using data on unem­
ployment to explain all new hires regardless of whether the worker
was previously unemployed is to replace the number of new hires
in equation 1 with the flow of workers from unemployment to em­
ployment, as in Bamichon and Figura (2010) and Veracierto (2011).
While this approach restricts attention only to new hires who were
previously unemployed, it suffers from the problem that the total num­
ber of vacancies is an imperfect measure of firm inputs into hiring the
unemployed, since firms’ efforts to fill these vacancies are aimed at
hiring all workers and not just workers who are already unemployed.
2Petrongolo and Pissarides (2001) survey the micro foundations of
the matching function, although several important papers in this area
were published after their survey. The traditional model of coordi­
nation frictions, due to Butters (1977), assumes firms post vacancies,
workers submit a single application each to some vacancy chosen
at random, and each firm hires at random among the applications it
receives. Burdett, Shi, and Wright (2001) emphasize that this model
does not give rise to empirically plausible matching functions and
that the number of hires per period will depend on additional vari­
ables, such as the size distribution of firms. Albrecht, Gautier, and
Vroman (2003) assume workers can apply to multiple vacancies,
but this does not give rise to empirically plausible specifications
either. Lagos (2000) and Shimer (2007) model coordination frictions
by letting firms and workers end up at different locations; firms choose
locations at random and workers choose locations optimally to maxi­
mize their expected earnings (per Lagos) or at random (per Shimer).
There are no frictions at any given location, so whichever side (firms
or workers) arrives in smaller numbers winds up fully matched. Thus,
each location will remain with either unemployed workers or vacant
positions, but not both. Unemployed workers and vacancies are thus
not inputs into forming new hires as the matching function approach
implicitly assumes, but consequences of poor coordination between
employers and workers on where to locate. When workers choose
locations optimally, the matching function is not empirically plau­
sible. When workers instead choose locations at random, the matching
function matches the data well, at least for a certain range of unem­
ployment and vacancies rates. Stevens (2007) develops a different
theory of the matching function based on the notion that workers
take time to screen heterogeneous jobs, rather than on coordination
problems. She finds that the implied aggregate matching function
is approximately Cobb-Douglas, as in equation 2 (p. 83). Decreuse
(2010) develops a model where workers apply to jobs they do not
realize are already filled. He finds that the implied matching function
will depend on lagged variables beyond just the contemporaneous
numbers of unemployed workers and vacant positions.
3For a survey that criticizes the use of aggregate production func­
tions, see Felipe and Fisher (2003).
4The same is true more generally for any specification w(«,v) that
exhibits constant returns to scale.
5It should be noted that a recent body of literature, starting with
Shimer (2005b), argues that the matching function approach suffers
from serious shortcomings in its ability to match various labor mar­
ket facts over the business cycle. However, this critique concerns
whether the value of a filled job to the employer who creates it, J,

94

varies enough over the cycle in these models, not whether the match­
ing function can explain how new hires vary with unemployment
and vacancies. My calculation does not depend on how J varies
with aggregate conditions, nor does it impose much structure on
how / ought to change over the cycle; and hence, it is not subject
to this critique.

6More precisely, consider the Mortensen and Pissarides (1994) model
where the separation rate into unemployment is endogenous. That
model assumes jobs are hit with idiosyncratic shocks to the profit­
ability of any given job at a constant rate X per unit time. The shock
term g is drawn each time from some fixed distribution F. Firms
optimally choose to terminate a job and send the worker into un­
employment for severe enough shocks, that is, when s falls below
some critical level grf. Suppose that in a recession, firms become more
demanding and raise the critical level to some higher value g^.When
the shock associated with the recession first hits, the unemployment
rate will jump and the flow into unemployment will spike as all jobs
whose s lies between grf and zd will be terminated immediately. The
spike in the separation rate will appear large even when the regular
flow into unemployment XF(grf) changes only modestly. My assump­
tion that the separation rate is constant over time only requires that
XF(grf) is relatively stable, not that flow rates from employment to
unemployment (which will reflect spikes) be stable.
7Some examples are Mazumder (2007); Fujita and Ramey (2009);
and Elsby, Hobijn, and §ahin (2010).

8In particular, Shimer (2005a) shows that the steady-state unemploy­
ment level to which the economy should be converging at any point
in time can be readily computed from flows into and out of unemploy­
ment at that instant. He then shows that this steady state is nearly
always close to actual unemployment.
9Bamichon and Figura (2010) and Veracierto (2011) also take
into account flows between unemployment and not in the labor
force in computing steady-state unemployment, which I ignore.
Acknowledging that out of the labor force is a distinct labor market
state does not change my ultimate conclusion that steady-state un­
employment and vacancies will appear negatively related, although
it may affect the shape of the curve relating the two series and how
much we should conclude it may have shifted over time. I return to
these issues later.
10Formally, as evident in equation 6, the Beveridge curve only depends
on the ratio s/A. The levels of 5 and A depend on the frequency
used to measure flows between labor market states.

11 In particular, the probability of profitably hiring a worker in the
Shimer (2007) model will not equal w(«,v)/v. Instead, it corresponds
to the equilibrium fraction of locations with more workers than jobs.
Employers in locations with more jobs than workers may still hire,
but will earn zero profits. Although the probability of a profitable
hire differs from w(w,v)/v, this probability will still be negatively
related to v in equilibrium.
12Nash bargaining is one rule on how to divide a given amount of
resources between two parties. This particular rule for how to divide
resources was proposed by Nash (1950), who showed this rule had
various desirable properties. Since employers and workers must divide
the surplus that results from their joint production, Nash’s solution
has often been applied to determine the wage that workers receive.
13Kocherlakota (2011) shows that under Nash bargaining, / rises by
nearly as much as A falls, so labor market tightness v/u is essentially
the same regardless of/. In figure 4, keeping v/u unchanged but
shifting the Beveridge curve up to the value associated with .41 would

3Q/2011, Economic Perspectives

imply an unemployment rate of no more than 6 percent. But as
hinted at in note 5, Nash bargaining is a somewhat problematic
assumption, since for standard parameterizations it implies that
productivity shocks produce fluctuations in / that are too small
to explain business cycle volatility.

14In informal communication, Rob Shimer computed the effects
of a 16 percent drop in the productivity of the matching function
in a fully worked out equilibrium model with concave utility and
declining marginal product of labor. He found that the unemploy­
ment rate would rise from 5 percent to 5.8 percent. This suggests
my bound may be a substantial overestimate of the true effect.
15See, for example, Aaronson, Mazumder, and Schechter (2010);
Valletta and Kuang (2010); Fujita (2011); Mazumder (2011); and
Hu and Schechter (2011).

16More precisely, lower effort should be viewed as a change in
some unobserved determinant of hiring that results in lower hiring
rates for the same number of vacancies while holding unemployment
fixed. This change may reflect lower effort—for example, firms may
spend fewer resources on advertising a position or on screening
and interviewing potential candidates. But alternatively, recruiters
may raise the standards they expect from workers, which would
also lower vacancy yields without representing lower effort on
the part of the firms.

17Similarly, cyclical changes in the composition of which firms are
trying to hire may lead us to incorrectly infer a change in the ability
of the typical firm to hire. Suppose there was a rise in the share of
hiring by firms that tend to rely more heavily on posting vacancies.
In this case, measured vacancies would appear to rise more than over­
all hiring. Bamichon et al. (2010) provide evidence that the shift in
the Beveridge curve coincided with a change in composition across
industries toward industries that rely more heavily on posting vacan­
cies to hire workers.

18Davis, Faberman, and Haltiwanger (2010) note that it may be dif­
ficult to ensure that both recruiting intensity varies over time and
market tightness is determined by a free-entry condition such as
equation 4. For example, they cite a model from Pissarides (2000,
chapter 4) in which there is free entry into the labor market but
recruiting intensity is constant over time. However, it is possible
to get time-varying recruiting intensity in a model with free entry
if we impose that both effective vacancies and the cost of creating
effective vacancies to be homogeneous functions of q and v of the
same degree. For example, since the product qv in equation 9 is
homogeneous of degree 2 in q and v, the cost function for recruiting
effort and posting vacancies must also be homogeneous of degree 2
in q and v.

REFERENCES

Aaronson, Daniel, Bhashkar Mazumder, and
Shani Schechter, 2010, “What is behind the rise in
long-term unemployment?,” Economic Perspectives,
Federal Reserve Bank of Chicago, Vol. 34, Second
Quarter, pp. 28-51.
Albrecht, James, Pieter Gautier, and Susan Vroman,
2003, “Matching with multiple applications,” Economics
Letters, Vol. 78, No. 1, January, pp. 67-70.

Bamichon, Regis, Michael Elsby, Bart Hobijn, and
Aysegiil $ahin, 2010, “Which industries are shifting
the Beveridge curve?,” Federal Reserve Bank of San
Francisco, working paper, No. 2010-32, December.

Davis, Steven J., R. Jason Faberman, and John C.
Haltiwanger, 2010, “The establishment-level behavior
of vacancies and hiring,” National Bureau of Economic
Research, working paper, No. 16265, August.

Decreuse, Bruno, 2010, “Matching with phantoms,”
GREQAM (Groupement de Recherche en Economic
Quantitative d’Aix-Marseille), working paper,
No. 2010-07, March.
Elsby, Michael, Bart Hobijn, and Ay^egiil $ahin,
2010, “Disciplined estimates of gross flows among
labor market states,” presentation at Federal Reserve
Bank of Chicago, September 10, available at
www.econ.wisc.edu/workshop/Michigan.pdf.

Bamichon, Regis, and Andrew Figura, 2010, “What
drives movements in the unemployment rate? A decom­
position of the Beveridge curve,” Finance and Economics
Discussion Series, Board of Governors of the Federal
Reserve System, working paper, No. 2010-48.

Felipe, Jesus, and Franklin M. Fisher, 2003,
“Aggregation in production functions: What applied
economists should know,” Metroeconomica, Vol. 54,
No. 2-3, pp. 208-262.

Burdett, Kenneth, Shouyong Shi, and Randall
Wright, 2001, “Pricing and matching with frictions,”
Journal ofPolitical Economy, Vol. 109, No. 5,
October, pp. 1060-1085.

Fujita, Shigeru, 2011, “Effects of extended unem­
ployment insurance benefits: Evidence from the monthly
CPS,” Federal Reserve Bank of Philadelphia, working
paper, No. 10-35/R, January.

Butters, Gerard R., 1977, “Equilibrium distribution
of sales and advertising prices,” Review ofEconomic
Studies, Vol. 44, No. 3, October, pp. 465-491.

Fujita, Shigeru, and Garey Ramey, 2009, “The cycli­
cality of separation and job finding rates,” International
Economic Review, Vol. 50, No. 2, May, pp. 415-430.

Federal Reserve Bank of Chicago

95

Hall, Robert, 2005, “Job loss, job finding, and unem­
ployment in the U.S. economy over the past fifty years,”
NBER Macroeconomics Annual, Vol. 20, pp. 101-137.
Hu, Luojia, and Shani Schechter, 2011, “How much
of the decline in unemployment is due to the exhaustion
of unemployment benefits?,” Chicago Fed Letter,
Federal Reserve Bank of Chicago, No. 288, July.
Kocherlakota, Narayana, 2011, “Notes on ‘Labor
markets and monetary policy,’” supplementary material
for speech at St. Cloud State University, March 3,
available at www.minneapolisfed.org/news_events/
pres/kocherlakota_notes_March3_2011 .pdf.

___________ , 2010, “Inside the FOMC,” speech in
Marquette, MI, August 17.
Lagos, Ricardo, 2000, “An alternative approach to
search frictions,” Journal ofPolitical Economy,
Vol. 108, No. 5, October, pp. 851-873.
Mazumder, Bhashkar, 2011, “How did unemployment
insurance extensions affect the unemployment rate in
2008-10?,” Chicago Fed Letter, Federal Reserve
Bank of Chicago, No. 285, April.

__________ , 2007, “New evidence on labor market dy­
namics over the business cycle,” Economic Perspectives,
Federal Reserve Bank of Chicago, Vol. 31, First Quarter,
pp. 36-46.
Mortensen, Dale T., and Christopher A. Pissarides,
1994, “Job creation and job destruction in the theory of
unemployment,” Review ofEconomic Studies, Vol. 61,
No. 3, July, pp. 397-415.

Nash, John F., Jr., 1950, “The bargaining problem,”
Econometrica, Vol. 18, No. 2, April, pp. 155-162.

96

Petrongolo, Barbara, and Christopher A. Pissarides,
2001, “Looking into the black box: A survey of the
matching function,” Journal ofEconomic Literature,
Vol. 39, No. 2, June, pp. 390-431.

Pissarides, Christopher A., 2000, Equilibrium Unem­
ployment Theory, 2nd ed., Cambridge, MA: MIT Press.
___________ , 1985, “Short-run equilibrium dynamics
of unemployment vacancies, and real wages,” American
Economic Review, Vol. 75, No. 4, September,
pp. 676-690.

Shimer, Robert, 2007, “Mismatch,” 2007, American
Economic Review, Vol. 97, No. 4, September,
pp. 1074-1101.
___________ , 2005a, “The cyclicality of hires, separa­
tions, and job-to-job transitions,” Review, Federal
Reserve Bank of St. Louis, Vol. 87, No. 4, July/August,
pp. 493-507.
___________ , 2005b, “The cyclical behavior of equi­
librium unemployment and vacancies,” American
Economic Review, Vol. 95, No. 1, March, pp. 25-49.

Stevens, Margaret, 2007, “New microfoundations
for the aggregate matching function,” International
Economic Review, Vol. 48, No. 3, pp. 847-868.
Valletta, Rob, and Katherine Kuang, 2010, “Extended
unemployment and UI benefits,” FRBSF Economic
Letter, Federal Reserve Bank of San Francisco,
No. 2010-12, April 19.
Veracierto, Marcelo, 2011, “Worker flows and
matching efficiency,” Ecw?ow/T Perspective.?, Federal
Reserve Bank of Chicago, Vol. 35, Fourth Quarter,
pp. 147-169.

3Q/2011, Economic Perspectives

Emergence of immediate funds transfer as a general-purpose
means of payment
Bruce J. Summers and Kirstin E. Wells

Introduction and summary
In a modem economy, we pay for goods and services
and trade in financial markets by transferring money
held in accounts with banks. For the better part of the
last century in the United States, most noncash payments
were made with the paper check, a payment instrument
that met most needs for payment services. Since the
mid-1990s, use of the paper check has been in decline
(Gerdes, 2008), a development that reflects technolog­
ical advances and innovations by providers of payment
services in response to needs for new and different
payment instruments. Today, individuals, businesses,
and governments can choose from a variety of payment
instruments, each of which is designed to meet their
specific needs for attributes such as certainty, speed,
security, convenience, and cost (Foster et al., 2010).
The most advanced means of transferring money be­
tween bank accounts is immediate funds transfer (IFT),
which allows senders to pay receivers electronically
in a highly convenient, certain, and secure manner,
at low cost with no or minimal delay in the receivers’
receipt and use of funds.
Today in the United States, IFT payments made
through the banking system are mostly limited to large
business transactions, interbank transfers, and special­
ized financial market transactions involving purchases
of securities and the like. In total, these larger payments
account for a small proportion of the total number of
payments made throughout the economy. There is in­
creasing evidence that the popularity of IFT is growing
for everyday use, such as consumer purchases, payments
between individuals, and small business accounts pay­
able (Hough et al., 2010). To date, however, most gen­
eral-purpose IFT payments are made on systems operated
by nonbanks, the most familiar being PayPal.1 The
coverage of IFT systems supported by nonbank com­
panies is limited to their closed customer groups, and

Federal Reserve Bank of Chicago

transfers are made not in bank money but rather in
special units of account defined by the nonbanks.
A notable development in a number of countries
around the world is the everyday use of IFT for gen­
eral-purpose payments using money held in accounts
at banks. In these countries, banks have invested in
applied technologies that allow them to provide low­
cost IFT services to the general public, taking advan­
tage of established national clearing and settlement
arrangements that link all bank accounts together. As
IFT innovators, banks in other countries are working
together collectively and in cooperation with public
authorities, such as central banks, to provide national
clearing and settlement for the new IFT service.
This article examines the emergence of IFT as a
general-purpose means of payment in the U.S. and in
four other countries. We identify the public policy and
business issues that arise when a new means of payment
is introduced. We describe the attributes of payment
instruments that users find attractive and compare the
attribute profiles of different kinds of instruments, in­
cluding IFT. We examine demand for IFT in the U.S.
and present four international case studies of IFT. Finally,
we discuss barriers to adoption of IFT in the U.S.

Payment attributes
Payments are made to satisfy personal or com­
mercial obligations between and among individuals,
Bruce J. Summers is a consultant to the Financial Markets
Group and Kirstin E. Wells is a business economist in the
Financial Markets Group at the Federal Reserve Bank of
Chicago. The authors acknowledge and are grateful to the
followingfor their helpful ideas and comments on earlier
versions of this paper: Nancy Atkinson, Sujit Chakravorti,
Harry Leinonen, JeffMarquardt, Bruce Parker, Louise
Roseman, the participants in a March 16, 2010, seminar
at the Federal Reserve Bank of Chicago, and an anony­
mous reviewer

97

businesses (including nonprofits), and governments.
Cash is the most basic and widely used means of pay­
ment by individuals in industrialized countries for trans­
actions up to about $25 (Rysman, 2010; Smith, 2010).
Apart from small-value payments, however, cash is not
a preferred means of payment.2 Most money is held in
transaction accounts at depository institutions.3 Pay­
ment instruments that provide access to this “deposit
money,” such as checks and debit cards, are the primary
means of making payments (See box 1 for discussion
of the bank payment business). Payment instruments
are generally either credit transfers, whereby a payer
(sender) directly authorizes the movement of money,
or debit transfers, whereby a sender indirectly authorizes
the movement of money via the payee (receiver). Re­
gardless of payment type, the end result is the same;
deposit money is transferred from sender to receiver.4
In the U.S., various payment instruments, supported
by core processing systems in banks and interbank
clearing and settlement mechanisms, are used to transfer
deposit money. These include paper checks, payment
cards, electronic debits and credits, and wire transfers
(which, as we discuss later, are a specialized form of
IFT). Senders select a payment instrument based on
how well its attributes match the purpose of the pay­
ment (for example, point-of-sale transaction or trade
payment between businesses). Because payments are
two-sided transactions, the needs of both the sender
and receiver are relevant in selecting the payment
method to be used.5
The primary attributes considered by senders and
receivers when selecting a payment instrument are as
follows:
■

Certainty—assurance to the sender and receiver
that funds are transferred as ordered;

■

Speed—timeliness of funds transfer from sender
to receiver;

■

Security—assurance that payment is protected
against fraud and completed as ordered;

■

Control—the sender and receiver have good
information about and are able to control the
timing of payment;

■

Universal acceptance—the payment instrument
is broadly accepted;

■

Versatility—useful for a variety of personal and
business transactions, including the ability to
transmit remittance information; and

■

Low cost and transparent pricing—reasonable
cost relative to value; fees are clear to sender
and receiver.

98

BOX 1

Transaction accounts and the payment line
of business at banks

While considered part of the “payment business,”
bank transaction accounts offered to individuals
and business customers are estimated to account
for only a fraction of banks’ total payment busi­
ness revenue. Revenue from transactions accounts
is attributable to net interest income earned from
balances on deposit (typically the largest compo­
nent), transaction fees, penalty fees, and a variety
of other fees. The payment business also includes
issuing credit cards to consumers, which is the
largest piece of payment business revenue. Other
payment businesses include issuing commercial
cards, card services for merchants, money transfer
services, issuing prepaid cards, and other smaller
business lines (McKinsey & Company, 2009).
The fact that the majority of payment revenue
does not come from transaction accounts, which are
typically considered “core” banking services, can
be explained by banks’ ability to generate higher
marginal returns from credit-related services. The
transaction-account payment business has until
recently emphasized “free” account services pro­
vided at very low fees, perhaps even below cost,
as an inducement for customers to build accounts
and grow net interest income.

Providers of payment services attempt to deliver these
attributes in combinations that best meet the needs of
the customers they serve. Technology is a principal
catalyst leading to improvements in such services as
one or more attributes can be strengthened without
degrading other attributes.
A comparison of attributes across different pay­
ment instruments, including IFT, is shown in table 1,
along with some common examples. Here, we discuss
the attributes by type of payment instrument as sum­
marized in table 1.
Payment types—Debit transfer
Debit transfers support the movement of money
between accounts held with banks. Paper check and
direct electronic debit are the most common debit trans­
fer instruments. Historically, the paper check has been
the most widely used method for making debit trans­
fers. Paper checks have many attractive attributes, in­
cluding payer control over the timing of payment and
near-universal acceptance by payees. Checks are also
very versatile in that they can be used for most personal,
commercial, and government payments. Businesses in
particular are heavy users of checks due to established
back-office processes that link paper-based invoicing

3Q/2011, Economic Perspectives

Federa l Res erv e Bank of Chicag o

TABLE 1

Attributes and examples of payment instruments
Attribute

Check

Direct debit

Direct credit

Wire transfer

General-purpose IFP

Debit card

Certainty

Provisional
payment
to receiver

Provisional
payment
to receiver

Payment
guaranteed
to receiver

Payment guaranteed
to receiver with
immediate finality

Payment guaranteed
to receiver

Payment
guaranteed
to receiver

Speed

Minimum
one day

Minimum
one day

Minimum
one day

Real-time

Within minutes

Real-time
authorization
and guarantee;
funds transferred
end-of-day at the
earliest

Security

Checks may be
stolen and/or
forged

Bank account
and routing
information from
check can be
used to originate
debit transfer

Fraud is limited
because payer
directly sends
funds from
account

Fraud is limited
because payer
directly sends
funds from
account

Fraud is limited
because payer
directly sends
funds from
account

Card numbers
may be stolen;
use of PIN with
certain cards limits
unauthorized
transactions

Control of timing

Payer controls
instruction but
cedes control of
funds movement
to payee

Payer controls
instruction but
cedes control of
funds movement
to payee

Payer controls
transaction

Payer controls
transaction

Payer controls
transaction

Payer controls
transaction

Universal
acceptance

Yes

Sender and
receiver must
agree to use

Sender and
receiver must
agree to use

Yes

Closed system
with limited number
of usersb

Limited by
merchant
acceptance

Versatility

Most types
of payment
transactions

Bill payments,
business-tobusiness trade
payments (with
remittance
information)

Recurring payments,
business-to-business
payments
(with remittance
information)

Financial market
transactions (with
limited remittance
information)

Most types of
transactions but
limited POS

Point-of-sale
(POS) and
online only

Low cost and
transparent
pricing

Not transparent
to individuals;
per-transaction
fee to businesses

Not transparent
to individuals;
per-transaction
fee to businesses

Not transparent
to individuals;
per-transaction
fee to businesses

High cost for
sender and
receiver
(transaction fee)

Not transparent
to individuals;
ad valorem fee to
merchant (PayPal)

Not transparent
to individuals;
ad valorem
fee to merchant

Clearing &
settlement

National check
system

ACH system

ACH system

Accounts held
with Reserve Banks
or CHIPS

PayPal system

Card networks

Example

Business accounts
payable

Utility payment

Payroll deposit

Purchase and sale
of bank reserves

Purchase of goods
and services

Grocery payment

alnformation in this column is based on the features of PayPal, which is the nonbank IFT service most commonly used today by individuals (Shevlin, Fishman, and Bezard, 2010).
bAs we discuss in the text, IFT in other countries links all or most transaction accounts held at banks.

(O

«o

________________________________________________________________________________________________________________________________________________

and accounts payable systems to check-based payment
systems. In general, the need to link remittance infor­
mation with a payment is a key factor in a business’s
choice of a payment instrument and, historically, the
remittance process has been paper-based.
For individuals, the cost of a check payment is not
necessarily transparent because most banks bundle check
fees with other transaction account fees. Some banks
offer “free checking,” which does not reflect the true
cost. Businesses and governments are typically charged
explicit per-item transaction fees by their banks, which,
in combination with back-office processing costs, make
checks relatively more expensive than electronic sub­
stitutes (Wells, 1996). Despite higher costs, many busi­
ness users find established payment processes effective
and the cost of switching to an electronic workflow,
including persuading counterparties to accept electronic
payments, prohibitive (AFP, 2010).
Historically, the process of clearing checks, which
involves moving the check from sender, to receiver,
to receiving bank, to paying bank (possibly through
intermediary banks or a central clearinghouse), was
labor and capital intensive. Today, checks are converted
to digital images for electronic processing once they
enter the clearing process. This may happen at a mer­
chant location, even as early as the merchant’s point of
sale.6 Even though most checks are cleared electronically,
funds movement is still a relatively slow process.
Depending upon when checks are entered into the
collection process by the receiving bank, provisional
credit is available to a receiver either the same day or
the next day, and deposit money is transferred from
the sender’s bank within one or two days.
Another type of debit transfer is the electronic
equivalent of a check, called direct debit. Direct debits
are marketed to individuals as “autopay” or “direct
bill.” This instrument allows individuals to make pay­
ments directly from their bank accounts by supplying
their bank account and routing number to the payee.
The true cost of direct debit is hidden because it is
typically free, or bundled with account service fees.
Direct debits are used primarily to pay bills and, more
recently, for online purchases. Acceptance of direct
debit is limited because not all payees offer this option
to individual payers.
Businesses are heavy users of direct debits to make
and receive trade payments, because fees are lower than
for checks and because electronic payments support
greater back-office operating efficiency. Direct debits
are typically as versatile as checks because remittance
information may be included electronically with pay­
ments. Yet, acceptance is limited because both the sender
and receiver must agree to use electronic payments.

100

Direct debits are cleared and settled via the auto­
mated clearinghouse (ACH) network, to which payees
gain access through their account-holding banks. Pay­
ment transactions are sent in batch form to a central
operator for processing with settlement at pre-scheduled
times during the day. Sending and receiving banks sub­
sequently update the accounts of senders and receivers.
The ACH was designed as a batch system because checks
are processed in batch form, and this processing model
persists to this day. Because of batch processing, ACH
debit transfers are relatively slow—there is a one-day
gap between the time a payment is initiated and the time
deposit money is transferred. Thus, direct debits, though
electronic, are not necessarily quicker for end-users
than check payments.
As shown in table 1, checks and direct debits fall
short in terms of certainty, control, and security. Because
payees initiate the movement of funds from the accounts
of payers, payers are uncertain about the timing of the
movement of funds. The lack of certainty and control
for payers has a direct bearing on payment fraud, be­
cause someone who has obtained bank account and
routing information from a stolen check, for example,
may be able to initiate an account debit without a payer’s
knowledge by fraudulent means. Fraudulent payments,
once identified by the payer or the payer’s bank, may
be returned, but returned payments undermine certainty
and security.
Credit transfer
Credit transfer is accomplished in a variety of ways,
principally as electronic credit and IFT.7 Electronic
credit transfers are used by businesses and governments
to make recurring payments to individuals for obliga­
tions, such as payroll and social security payments. They
are also used for business trade payments. Recurring
payments are received by individuals as “direct deposit.”
Direct deposit is used for nearly all government-toindividual payments, but not all businesses have
adopted direct deposit. The cost of direct deposits is
not transparent to individuals because they are typi­
cally not charged to receive them, whereas business
users pay an explicit per-transaction fee.
Direct deposits and some other types of electronic
credit transfers are processed on the ACH network. As
in the case of debit transfers, ACH credit transfers are
relatively slow, with a one- or two-day lag between
the time the payment is initiated by the sender and the
time deposit money is transferred to the receiver. As
shown in table 1, electronic credits offer more certainty,
control, and security for senders, who directly autho­
rize the movement of money.

3Q/2011, Economic Perspectives

Immediate funds transfer is used today primarily
for large-value business and financial market transac­
tions, through bank wire transfer services. Wire trans­
fers constitute a small portion of the overall number
of payments and a large portion of the overall value
of payments; their daily value exceeds a trillion dollars.
Wire transfers are expensive, typically costing about
$25 to $35 per transaction, and are thus not widely used
by individuals. Wire transfers are not only immediate,
they are final. That is, wire transfers are irrevocable and
unconditional and offer the highest certainty of any
payment type. Wire transfers are accepted by most banks.
Clearing and settlement of wire transfers takes
place over one of two specialized systems: Fedwire,
which is operated by the Federal Reserve Banks, or
the Clearing House Interbank Payment System (CHIPS),
which is operated by The Clearing House Payments
Company L.L.C. In the case of Fedwire, banks trans­
fer balances directly between accounts they hold with
the Federal Reserve Banks. CHIPS is a closed net­
work whose members exchange payments, which are
settled by means of continuous multilateral netting.
As indicated in table 1, wire transfers are quick, certain,
and secure, and accordingly they are relied on in inter­
bank and financial markets worldwide and are often
made using real-time gross settlement (RTGS) systems
(World Bank Group, 2008). Virtually all RTGS systems,
including Fedwire, are operated by central banks, which
for these purposes are functioning as universal bankers’
banks. Wire transfers involve the transfer of deposit
money that banks hold in accounts with central banks
(sometimes referred to as “central bank money”). Public
oversight authorities have made the use of RTGS a
virtual requirement for systemically important payment
systems (BIS, 2001).
Much of the innovation in U.S. payment instru­
ments over the past decade has centered on generalpurpose IFT. Nonbanks have been at the forefront of
this innovation. The approach taken by nonbanks is
twofold: 1) offer payment services directly to end-users
that substitute for and compete with the services pro­
vided by banks; and 2) provide banks with the business
processes and technical capabilities that allow them to
offer IFT services to their account-holding customers.8
Under the first approach, nonbanks directly pro­
vide general-purpose IFT services to individuals and
small- to medium-sized businesses. A nonbank payment
provider must first establish a funding source for IFT
payments that are initiated by its customers, as it can­
not tap directly into the customers’ bank accounts. The
nonbank provider would typically do so by setting up
an omnibus account with its bank, to which its customers
make deposits. The customer funds pooled in the

Federal Reserve Bank of Chicago

omnibus account are then reflected in ledger accounts
set up by the nonbank on its computers that are denom­
inated not in commercial bank money, but in parallel
units of value identified with the nonbank provider
(for example, PayPal dollars). Collectively, these ledger
accounts constitute a closed, proprietary network that
supports transfers of value units among the users of
the nonbank providers’ services. Payments to receivers
outside the network are supported, but in this case a
conversion back to bank money is required. The con­
version back to bank money is accomplished by sending
deposits in the omnibus account back through the bank
payment network to the bank account of the receiver,
which is not part of the nonbank network. The nonbank
payment networks rely on modem, applied technologies
to support immediate funds transfers, and in-network
transfers occur virtually instantaneously. Out-of-network transfers that rely on the banking system may
take several days to complete.9
Under the second approach, banks use a technol­
ogy platform supplied by the nonbank company in
combination with their own in-house authorization
systems to provide IFT services to their account-holding
customers. Banks following this approach brand the
services as their own. Again, however, the resulting
network is closed, and proprietary, connecting accounts
at the limited number of banks that use a particular
nonbank vendor’s platform. So long as a payee and
payer hold accounts at banks that use the same non­
bank provider’s technology, they can transfer funds
directly to each other’s accounts.10 Out-of-network
transfers are possible, but again the transfer may take
several days to be completed.
Debit cards
Debit cards are a unique type of payment. While
payments made by debit card are cleared and settled
like debit transfers, they offer IFT-type attributes to
both cardholders and merchants, as shown in table 1.
In particular, debit card payments offer speed, certainty,
and control to both parties. Specialized authorization
systems instantaneously check, at the point of sale,
whether payers are able to fund purchases from their
bank accounts. Once a transaction is authorized, mer­
chants have the certainty of knowing that payment
will be received. Unlike IFT, however, funds are not
transferred from the individual’s to the merchant’s ac­
count until the end of the day at the earliest. Yet, the
pre-authorization makes the payment seem immediate
to cardholder and merchant.11
Debit cards offer limited versatility, as they are used
primarily at the merchant point-of-sale, with merchants
who have agreed to join a debit card network. The
cost of debit cards is not transparent to cardholders

101

(typically transactions are free), and merchants pay ad
valorem fees, which are a percentage of the transaction
amount.12 Debit cards are subject to unauthorized use
if stolen, and the card networks have security measures
in place to limit unauthorized transactions as well as
rules on limited liability for merchants. (Credit cards
are not taken up directly because, as described in box 2,
their principal purpose is to provide credit services.)
IFT innovation—General-purpose payments
The foregoing discussion of payment instruments
and their attributes shows that wire transfer and generalpurpose IFT offer attractive combinations of attributes
compared with other types of payment instruments,
especially certainty, speed, control, and versatility.
The average price of a wire transfer makes this payment
instrument unattractive for general-purpose use, and a
primary advantage of IFT is its low price. As we dis­
cuss in the next section, evidence of latent demand
and revealed preferences for certain combinations of
attributes support the view that there is an unmet need
for broadly available IFT in the U.S.

Demand for IR
Latent demand
Research conducted by the Federal Reserve System
on payment system user preferences provides evidence
that users desire a service with the attributes of IFT. In
a 2002 survey on the future of retail electronic payments
(Board of Governors of the Federal Reserve System,
2002), respondents appealed for the development of a
low-cost way for individuals and businesses to make
online real-time funds transfers.13 Survey participants
also noted the need for a new, uniform “deposit direc­
tory” of account numbers and account status, or some
other means of account verification, as well as a direc­
tory to route electronic payments more easily to recipi­
ents. Further, in a 2006 survey on barriers to innovation
in payments (Board of Governors of the Federal
Reserve System, 2006), payment industry respondents
indicated that wire transfers would be an effective mech­
anism for making smaller value payments at an accept­
ably low price (presumably the price would need to be
lower than the typical bank wire transfer fees) and
with remittance information easily linked to corporate
billing systems. These two surveys reveal a clear interest
in IFT, subject to the availability of directory and routing
information and responsiveness to specific user require­
ments, including low cost and improved support for
remittance information.14

Revealedpreferences
Other evidence to support the view that IFT may
be broadly desirable in the U.S. is the increased use

102

BOX 2

Credit cards

Credit cards are also commonly used by individuals
at merchant locations, yet credit card transactions
are not debit or credit transfers. Credit cards are a
means of providing access to short-term consumer
finance, whereby merchants receive funds from their
banks at the end of the day but cardholders do not
authorize the transfer of deposit money until they
pay their monthly credit card bill to the bank that
issues them the card. This bill is for the aggregate
amount owed to cover multiple transactions and is
not required to be paid in full. Thus, credit card
transactions, while often considered payment trans­
actions, do not fall under either the credit transfer
or debit transfer model. The distinction between a
credit card transaction and payment transaction
holds true even though an estimated 40 percent of
cardholders, so-called convenience users, do not rely
on short-term credit and pay their balance in full
each month (Herbst-Murphy, 2010). Convenience
users typically use credit cards for other reasons,
such as garnering reward points or simplifying
their cash management by accumulating payments
over a monthly grace period.

of payment instruments that offer attributes most closely
related to IFT. For example, the use of debit cards, which
offer more control, certainty, and speed than other pay­
ment instruments, has grown more rapidly than that
of any other means of payment for point-of-sale and
online purchases by individuals. In 2008, individuals
held more debit cards than credit cards and, on average,
used debit cards more often than cash, credit cards,
or checks individually (Foster et al., 2010). In 2008,
$1.00 of every $5.00 was spent with a debit card in
the U.S., up from $1.00 of every $14.00 in 2001
(Herbst-Murphy, 2010).
A portion of the increase in debit card usage can be
explained as a secular trend of growing familiarity with
electronic payments in general. As shown in figure 1,
the percentage of noncash payments made by electronic
methods has grown in the last ten years, which reflects
this trend. Other reasons cited for debit card preference
include increased convenience and speed of payment
(Rysman, 2009), which make debit cards more attrac­
tive than checks. Part of the growth in debit card usage
and decline in check usage shown in figure 1 can be
attributed to the substitution of debit cards for checks.
Business use of payment instruments with attri­
butes that closely resemble those of IFT has grown as
well. In 2010, one of the fastest-growing transactions
processed on the ACH network was direct credit for

3Q/2011, Economic Perspectives

general-purpose payments. These countries are China,
the Czech Republic, Serbia, the Slovak Republic,
Switzerland, Turkey, and Ukraine.
The banking systems of at least three other coun­
tries have created transaction processing infrastructures
specifically designed for IFT. These countries are Mexico,
South Africa, and the United Kingdom (UK). Conse­
quently, although their implementation approaches may
differ somewhat, the banking systems of at least ten
countries have taken coordinated steps to provide IFT
services. Here, we discuss the cases of Mexico, South
Africa, Switzerland, and the UK.15 These case studies
help us to identify several business and public policy
considerations that arise when a country seeks to es­
tablish a national network to support a new payment
instrument. A common consideration is reliance on
the national RTGS system to provide finality for IFT
payments, either directly by means of transaction pro­
cessing or indirectly by means of interbank settlement
of IFT obligations.

sending bills paid through online banking sites to biller
receivers (Digital Transactions, 2010). Direct credits
offer advantages over checks and direct debits for bill
payment in terms of certainty and security, much like
an IFT.
Experience with IFT in other countries provides
insights into the potential for this type of payment in
the U.S. In the next section, we present four interna­
tional case studies of the successfiil introduction of IFT.
In each case, IFT has been introduced as a universal or
near-universal payment instrument supported by clearing
and settlement mechanisms that connect virtually all
bank accounts within a given country. Universal sup­
port for IFT has been accomplished through industry­
wide cooperation, sometimes facilitated and
promoted by public authorities.

IFT case studies
As we noted earlier, wire transfer is a standard
means of payment worldwide and is most often sup­
ported by RTGS systems operated by the central banks.
These RTGS systems are capital intensive, benefit from
economies of scale, and in most cases are operating
well below efficient scale (Allsopp, Summers, and
Veale, 2009). The services provided by RTGS systems
in at least seven countries have been expanded to

Federal Reserve Bank of Chicago

Mexico
Immediate funds transfer was introduced in Mexico
in 2004, with the implementation of a new RTGS
system by Banco de Mexico. The new RTGS system,
known by the acronym SPEI, takes advantage of new
processing technologies that allow continuous upward
scaling of transaction processing volumes at low mar­
ginal cost, with strong security based on a public key
infrastructure (PKI). During the SPEI project, some
commercial banks indicated that they considered two
credit transfer systems (the other being the Mexican
ACH) to be wasteful. Accordingly, Banco de Mexico
designed SPEI to support a variety of credit payments
on one processing system, providing banks with a
choice between using the new RTGS and ACH. Banco
de Mexico has promoted the use of IFT through adver­
tisements in the mass media.
The central bank also provides payment services
to the Mexican government and had been using its old
RTGS for large government disbursements and the
ACH for smaller disbursements. It was clear that so
long as the Mexican government continued using the
ACH for any disbursements, commercial banks would
be forced to maintain their ACH systems. In 2008,
the government agreed to Banco de Mexico’s request
to use SPEI for all disbursements. Further, the govern­
ment decided to centralize its payroll processing and
use SPEI for government payrolls by the end of 2009.
To support government payments, Banco de Mexico
instituted an earlier opening time for SPEI in order to
allow commercial banks to maintain their established
processing schedules. The government and banks use

103

the straight-through processing capabilities that SPEI
offers, with the expectation that both efficiency and
service levels will increase throughout the payment
system.16 Most SPEI payments take less than a couple
of minutes to reach the beneficiary’s accounts. By
law, all SPEI payments are final, regardless of their
size or the beneficiary. Payments are final as soon as
the beneficiary’s bank receives a settlement notice.
Mexican commercial banks offer their customers
IFT payment services mainly online. The payer must
provide the bank routing and account numbers for the
payee. One-off payments are therefore difficult to make
because of the information that is needed on the payer
side. Point-of-sale transactions are not currently sup­
ported, in part because of stringent security requirements
established by the Mexican Banking Commission. Small
mobile payments are, however, now being supported
by new regulations and by a security agreement between
banks and the commission.
Banks follow a variety of practices for pricing IFT
payments. Large banks charge per-transaction fees of
up to $0.35 or bundle credit transfer services with their
Internet banking offerings for a fixed fee. The typical
fixed fee for Internet banking service in Mexico is around
$2.50. Prices for over-the-counter payments usually are
higher than for Internet banking transactions. Some
banks charge about half as much for ACH credit trans­
fers as for real-time credit transfers, whereas other
banks charge the same for both payment services.

South Africa
The introduction of IFT services for use by the gen­
eral public in South Africa is a direct result of a recent
initiative by commercial banks. The South African pay­
ment system has supported a number of general-purpose
payment options, including the paper check, the check
card (a means of initiating a credit transfer from a check­
ing account at the point of sale, upon authorization, and
usually available only to high-net-worth customers),
debit and credit cards, and ACH-type electronic funds
transfer (EFT) debit and credit payments. Access to
check payments would take from one to seven days;
and EFT and Internet payments would take on average
one day for the transfer of funds intrabank and three
days for the transfer of funds interbank.
Commercial banks in South Africa identified the
need for a payment instrument that would give the gen­
eral public the ability to transfer funds quickly and in
a manner that made funds available to the payee im­
mediately. Seven banks began collaborating in 2005
to develop a new clearing and settlement mechanism
called Real-Time Clearing (RTC), in cooperation with
the South African Reserve Bank, and the capability
was implemented in March 2007. The banks provide

104

services via Internet banking for consumers, online
initiation through corporate banking solutions for
businesses; and offline, over-the-counter initiation at
a bank branch or by telephone. In each of these cases,
the payer must follow an authentication procedure and
provide routing information (bank and account number)
for the payment. While no point-of-sale facilities are
currently available, mobile services over cell phones
are supported; and in theory, a merchant could be paid
by mobile IFT, although no confirming message would
be sent to the payee.
Immediate funds transfer payments made by the
RTC method are governed by rules established by the
Payment Clearing House (PCH), which banks are
bound to in bilateral agreements. In addition to rulemaking, the PCH functions as the system operator.
It clears RTC payment instructions and provides the
interface to the South African Reserve Bank RTGS
system, known by the acronym SAMOS, which clears
and settles the interbank obligations arising from RTC.
Once an RTC payment instruction is cleared by the PCH,
the receiving bank credits the beneficiary’s account
within 60 seconds. The interbank RTC clearing and
settlement obligations built up in the PCH are sent to
SAMOS on the hour every hour during the business
day, which significantly reduces the risks associated
with RTC payments.
Banks charge higher prices for IFT than for other
Internet banking and mobile payments. Pricing has two
parts, a per-transaction fee and a charge based on the
amount transacted for purchases, with a cap on the
maximum total cost of the payment. At about $1.00,
IFT per-transaction fees are about three times the pertransaction fees for regular Internet and mobile payments.
The charge based on the transaction amount is the same
across all three types of payments at approximately
$0.07 per $1.00. Finally, the cap on the total price per
payment is $5.00 for IFT payments, compared with
$1.40 for regular Internet and mobile payments. It should
be noted that IFT is differentiated from the pure RTGS
wire transfers, not only in terms of operational process
and timing (up to a one-hour delay for IFT compared
with real-time for RTGS) but also in pricing. In the
event that a bank client requests RTGS as the payment
method, an even higher premium is charged.
Switzerland
Credit transfers have a long history in Switzerland,
where the postal service has offered giro payments using
a national standard format for over 100 years. (The
credit transfer format known as Einzahlungsschein
[credit slip] dates to 1906 and prevails to this day in
a comparable form.) Traditionally, a credit slip has
been used to initiate recurring and one-off payments,

3Q/2011, Economic Perspectives

either over-the-counter at the post office or bank or,
more recently, through the mail. The payee company
would send a credit slip to the payer with pertinent
information filled out, including bank/post and personal
address; account number; and, if relevant, a reference
number to assist the payee company in processing the
payment. For payment purposes, account details are
typically not perceived as confidential information by
Swiss consumers and companies and are provided on
a need-to-know basis to facilitate payments.
Today, IFT is available to businesses and individ­
uals as an extension of the traditional credit slip. In
addition to the traditional paper method, IFT is avail­
able through Internet banking and ATMs.17 To illustrate
the payer experience with IFT, imagine a computer
terminal securely connected to a bank or PostFinance
(the Swiss Post’s financial institution) website. The
payer clicks on “making payments” and receives a
menu of choices among different types of credit slips,
for example, payments to accounts at the same bank,
at a different bank, payments with or without reference
numbers, and so on. When it is selected, a digital credit
slip opens and the payer fills out the necessary fields
using the information received from the payee company.
To reduce manual intervention, electronic payment-slip
readers can be used. When the payer completes the
instructions, the “electronic credit slip” is immediate­
ly verified by the system online and, assuming it is
complete and correct, delivered to the bank for pro­
cessing. The payer would typically not be aware of
the particular infrastructure used to settle payments.
Credit transfers are typically settled through the
Swiss RTGS system, called Swiss Interbank Clearing
(SIC). This system is overseen by the Swiss National
Bank (SNB) and operated by SIX Interbank Clearing Ltd.
on behalf of the SNB. Swiss Interbank Clearing is owned
by the Swiss commercial banks and PostFinance. Gen­
eral-purpose credit transfers have been more widely
settled in SIC since PostFinance became a participant
in 2001. The extension of SIC services beyond tradi­
tional large-value transfers is a cooperative development
involving the commercial banks, PostFinance, and
the central bank, and reflects their collective interest
in supporting more efficient credit transfers, in this
case making greater use of SIC and avoiding duplica­
tive infrastructure for processing small-value payments.
In this way, the banking system benefits from economies
of scale in operations and pooling of liquidity. In addition,
standards are followed to facilitate efficient processing
(for example, increasing use of the international bank
account number or IBAN) for routing information.
Pricing of IFT payments in Switzerland depends
on the bank providing the service and the customer

Federal Reserve Bank of Chicago

segment being served. Banks often include consumer
payments as a component part of their bundled account
service packages. Charges for account service packages
depend on the balance that is maintained. Domestic
payments would typically not carry a per-trans action
charge. An exception would be paper payments that
require manual processing steps for the banks or
PostFinance. These payments would typically carry
a surcharge as an incentive for the customer to use
online banking.
United Kingdom
Faster Payments is a new IFT service in the UK
that makes near-real-time and irrevocable credit trans­
fers available to all bank customers at nonpremium
prices. Introduced in May 2008, Faster Payments is
available across the banking industry and is supported
by common rules and a shared processing infrastruc­
ture. Faster Payments is a voluntary initiative of the
banking industry, agreed to by the Payment System
Task Force, which was organized and chaired by the
UK’s Office of Fair Trading (OFT). The OFT orga­
nized the task force in response to a mandate from
the Chancellor of the Exchequer. The official mandate
was reinforced by the threat of government-sponsored
legislation to remedy perceived inefficiencies in the pay­
ment system, resulting from insufficient competition
and overly slow cooperation among banks. Of princi­
pal concern to the government was a three-day delay
in the interbank clearing of electronic payments.
The Payment System Task Force told the payments
industry to devise a same-day service. The industry’s
response was to propose a near-real-time service, deliv­
ered through a special purpose infrastructure designed
and operated by VocaLink. The company that is respon­
sible for the Faster Payments Service (a name that is
acquiring a brand identity for purposes of marketing the
service to the public) is the CHAPS Clearing Company.
The company provides two main services: CHAPS
Sterling for systemically important payments and
Faster Payments for time-dependent payments.
The 13 banks that originally agreed to develop
the service now originate Faster Payments on behalf
of their customers, and approximately 68 credit insti­
tutions, representing an estimated 90 percent of all
transaction accounts in the UK, receive such payments.
Membership in the Faster Payments Service is open
to all credit institutions that have settlement accounts
with the Bank of England and can connect their networks
to the payment system infrastructure continuously,
24 hours a day, seven days a week. Indirect access is
also permitted, whereby an institution offers the Faster
Payments Service and settles through a member.

105

Customers can originate Faster Payments through
their banks either by phone or Internet connection
24 hours a day, seven days a week; it is estimated that
approximately two-thirds of all UK phone and Internet
payments are now made by this method. Support for
mobile Faster Payments is an important component
of the UK’s payment system strategy; it is seen by some
as a viable alternative to reliance on the paper check
(VocaLink and Price Wat erhouseCoopers, 2009). Oneoff payments are received by the beneficiary usually
within minutes, but always within two hours. These
one-off payments can be ordered on the payment date
or submitted as forward-dated payments to be made
on designated days in the future. Standing order pay­
ments are also possible, although these will be processed
for same-day settlement and then only on bank working
days. A direct corporate access feature has recently been
added that enables companies with large volumes of
payments to submit files directly to the Faster Payments
Service infrastructure, provided they are sponsored
by a member bank. This new feature is intended to
increase the attractiveness of the service for firms that
have a large number of expenses to pay, including pay­
rolls, and is analogous to the services provided to cor­
porate users of the ACH system in the UK.
A Faster Payment becomes final at the time the
sending bank submits the transaction to the processing
system; sending banks manage their risk by authenti­
cating the instruction received from the originator of
the payment and checking the customer’s account to
ensure that the balance is sufficient to fund the payment
order. The Faster Payments Service processing system
verifies that all of the required details are provided in
the proper format and forwards the payment to the re­
ceiving bank. The receiving bank verifies that the funds
are being directed to a valid account and sends a vali­
dation message back to the Faster Payments Service.
The receiving bank is then credited with the funds.
Confirmations of complete transactions are issued to
the sender and receiver.
The prices charged for Faster Payments are a frac­
tion of those charged for traditional CHAPS transfers,
which can cost up to $35.00 each. Marketing informa­
tion published by banks indicates that per transaction
prices are below $1.25, ranging downward to about
$0.50. Transactions for retail customers are typically
free. A size limit for transfers of GBP 100,000 has been
set as a risk-management measure; this may be raised
or eliminated in the future.

Summary
The four case studies are summarized in table 2.
For each country, the table identifies the catalyst behind
the introduction of the service, the delivery channels

106

through which the banks provide the services to their
customers, the back-end system for clearing and set­
tling payments, the routing number scheme, and the
prevailing fee structure. The four case studies illustrate
two general approaches to interbank IFT processing.
In Mexico and Switzerland, the national RTGS systems
are relied upon for interbank processing, extending
existing RTGS functionality to a broader set of under­
lying payments. In South Africa and the UK, the banks
have created new, shared utilities that handle all of
the interbank processing for the individual transactions
and, in turn, rely on the national RTGS for final inter­
bank settlement of netted IFT transfers periodically
throughout the day.
In two of the four cases (Mexico and the UK),
public authorities led in motivating a coordinated re­
sponse across the banking system. In Mexico, the central
bank served as catalyst and did so in part through its
operational role as a provider of RTGS services. In the
UK, the OFT, which shares responsibility for aspects
of payment system oversight with the central bank,
provided the motivation as a regulator concerned about
the quality of payment services available to the general
public. In contrast, in South Africa and Switzerland,
banks identified an unmet service need (and opportu­
nity) and took the lead, enlisting the central bank to
provide support where necessary.
Table 2 highlights the areas where banks cooper­
ate and compete in the provision of IFT services.
Cooperation in planning is necessary to support nation­
wide services. In South Africa and the UK, the opera­
tional cooperation extends to governance over creation
and enforcement of the rules that apply to the IFT
network, as well as sharing in the investment and
ongoing operating costs for the interbank processing
system. With regard to routing of payments, note that
only in Switzerland has the banking system adopted
a standard routing number scheme, which facilitates
processing for all parties to transactions and, further,
makes it easier for senders and receivers of payments
to manage the exchange of bank and account number
information that is needed to route the transactions effi­
ciently and accurately. As we describe later (in note 18),
in the UK the banking clearinghouse provides bank
routing information directly to the public.
The last column in the table summarizes the price
structures and prices that apply to general-purpose
IFT. In each case except South Africa, the price struc­
ture is essentially “cost-plus,” that is, fees are based
directly on the cost of production plus a markup re­
flecting service value and profit. In the case of South
Africa, the banks not only charge per-transaction fees,
but also an ad valorem fee component related to the

3Q/2011, Economic Perspectives

TABLE 2

International experience with immediate funds transfer
Clearing and
settlement

Routing

Fee

Online banking,
mobile

RTGS (SPEI)

BAN8

Fixed per transaction
(could be bundled),
$0.35-$2.50

Commercial
banks

Online banking,
mobile, over-thecounter

Real-Time
Clearing (RTC)

BAN

Fixed per transaction,
$1.00 + ad valorem,
$0.07/$1.00)

Switzerland

Majority of
banks and
central bank

Online banking,
ATM, over-thecounter

RTGS (SIC)

BAN, IBANb

Typically bundled with
account service fees

UK
(Faster
Payments)

Competition
authority

Online banking,
mobile, direct
corporate access

Faster Payments
Service (FPS)

BAN

Fixed per transaction,
typically free of explicit
charges for retail
customers, $0.50-$1.25

Country

Catalyst

Channel

Mexico

Central bank

South Africa

“Bank account number.
"International bank account number.

value of the transaction; this is similar to payment card
price structures. The two approaches to pricing high­
light an important two-part public policy question
concerning the optimal way to price payment network
services when credit risk is mitigated through the use
of the immediate funds transfer model. First, is par
clearing (receipt of the amount designated in the pay­
ment without deductions) a desirable goal? Second,
can and should prices charged to end-users be based
on production costs?

Issues with IFT implementation
What are the business and public policy issues that
would need to be considered prior to the national in­
troduction of IFT as a general-purpose means of pay­
ment in the U.S.? Three primary issues in addition to
pricing are network reach, payment routing, and gov­
ernance. Each of these issues has practical implications
for the feasibility of IFT as a new payment service and
each is important from a public policy perspective.
Network reach
IFT services are now available in the U.S., but
are limited to closed proprietary networks. The pro­
cess of clearing and settlement for these proprietary
networks works efficiently only for the members who
use a particular service provider’s technology. In the
case of a transfer destined for a receiver who is not a
member of the proprietary network, the transaction
must be routed through a bank payment system, such
as ACH, using the national banking network. From a
public policy perspective, the emergence of multiple,
incompatible, and proprietary payment networks is

Federal Reserve Bank of Chicago

not an efficient or effective way to provide IFT services.
A national clearing and settlement mechanism,
however, does not guarantee that the payment network
supporting an instrument such as IFT will connect all
bank deposit accounts. As illustrated by the case studies,
bankers may not be required to provide the service to
their customers by regulation or by the terms of their
clearinghouse memberships. An obvious practical
problem with voluntary network participation, well
illustrated in the case of Faster Payments in the UK,
is that senders need to know whether their intended
receivers hold accounts at a bank that can receive IFT
transfers. A national directory sponsored by the UK
clearinghouse is available online to help senders get
this information as efficiently as possible.18
While not the subject of this article, the chartering
and regulatory status of new, nonbank suppliers of
payment services also has a bearing on the network
reach issue. The innovators should not be prohibited
from joining and helping stimulate improvements in
the banking payment network by offering payment
accounts, so long as they can meet basic tests of sound­
ness and reliability, as do regular banks. As members
of banking clearinghouses and associations, the non­
bank innovators would contribute to the bank payment
network’s expansion. Moreover, to the extent that they
innovate through the use of “disruptive technologies,”
these nonbank companies would stimulate technological
innovation in services such as IFT. The U.S. financial
regulatory authorities should consider how payment
innovation can be encouraged by allowing nonbank
firms to offer deposit accounts on terms that are reason­
able and prudent.19

107

Payment routing
The principal operational advantage of payments
such as checks and electronic direct debits is that they
provide routing information that the payer would other­
wise have to request. On a paper check, for example,
the payer’s bank routing number and account number
are printed in magnetic ink at the bottom of the check.
Thus, the payment instruction automatically contains
the data needed by the payee’s bank to present the in­
strument for payment. Routing information is provided
with debit card payment instructions as well. For elec­
tronic credits and IFT, the payer needs to obtain payee
routing information and provide that information to
its bank. Acquisition of this information adds com­
plexity and cost, especially for transactions between
two parties that are not well known to one another.
Account numbers are sometimes considered to be
part of one’s “transactional identity,” which is sensi­
tive information that should be protected. Because of
this concern, receivers may be reluctant to give their
account number to a payer for an IFT payment. Such
concerns, however, should be reduced by the IFT pay­
ment flow and authorization model. First, IFT results
in money deposited to the receiver’s account, not with­
drawn from it. Second, bank controls are designed to
restrict the power to initiate transfers of funds to proper­
ly authenticated parties. Thus, there is limited oppor­
tunity for anyone to fraudulently order an IFT based on
knowledge of an account number and routing number.
As mentioned, paper checks contain complete rout­
ing information that is in plain view to anyone handling
the check. This is prima facie evidence that routing in­
formation is not unduly sensitive. It is not considered
so in the countries examined in connection with the four
case studies. Further, it is notable that the IFT payment
services provided by nonbanks often rely on widely
known and used “addresses” for routing and information
exchange over networks, including telephone numbers
and email addresses. The new approaches to routing
appear to point to the serviceability of highly public
addresses for transferring financial information, includ­
ing funds transfers, in a well-controlled environment
with strong information security protections.
A somewhat broader issue that arises when con­
sidering routing of payments and the use of account
numbers is that of standardization and portability of
financial addresses. If bank account numbers are not
standardized across the banking system and are not
portable, bank numbers change whenever an account
holder changes banks. Switching banks becomes more
complex because all established payment relationships
must be updated with new account information. Pro­
gressive banking practice and good public policy call

108

for both standardization and portability of bank account
numbers, both to increase the efficiency of the payment
system and to increase competition among banks by
making it harder to lock in customer relationships through
high switching costs. This is not an unreasonable expec­
tation in an information-intensive industry like banking.
Public policy that is concerned with the efficiency and
competitiveness of payment services could be informed
by practices and expectations in other information-intensive industries, for example, telecommunications.20
Payment system governance
Each of the four case studies discussed in this article
provides an example of payment system innovation co­
ordinated at the national level. The catalyst may be from
the public sector (central bank or other governmental
authority, such as the UK’s Office of Fair Trading) or
the private sector (groups of banks), but in each case
IFT innovation proved successful due to a national gov­
ernance approach. In addition, the governance approach
followed in the four countries recognizes the boundary
between cooperation and competition among banks.
This type of national, coordinated approach would
be difficult to achieve in the U.S. in light of its highly
decentralized payment system management, which is
reflected in part by the absence of a truly national
clearinghouse. Currently, multiple publicly and privately
operated payment systems operate in parallel in a com­
petitive environment. Sweeping national change in
the U.S. payment system in this century so far has
come about through legislation—the 2003 passage of
the Check Clearing for the 21st Century Act, which
facilitated electronic check clearing; and the 2010 Wall
Street Reform and Consumer Protection Act, which
mandated limits on fees that banks charge merchants
for debit card transactions. Without an explicit legis­
lative mandate or some other form of encouragement
from the government, it is unlikely that banks in the
U.S. will find a cooperative basis for IFT innovation.
In addition, because IFT may disrupt banks’ revenues
from high-priced wire transfer services, coordination
and cooperation may not be readily forthcoming.
Further, unless IFT clearing and settlement relies on
existing mechanisms (as in the cases of Mexico and
Switzerland), a national IFT system may have high
start-up costs that the industry might be unwilling to
bear. Overall, the complexity involved with implement­
ing a national IFT solution may be unworkable within
the existing U.S. banking structure.

Conclusion
General-purpose IFT is a means of payment that
offers attractive combinations of attributes to both

3Q/2011, Economic Perspectives

senders and receivers, such as certainty, speed, control,
and versatility, all at relatively low cost. There is evi­
dence of strong latent demand for IFT in the U.S. by
individuals, businesses, and governments, but to date
this demand is being met only to a limited extent and
principally by nonbank providers of payment services.
To satisfy the demand for IFT, it will be necessary to
provide access to money held in banks by linking all
bank deposit accounts through an immediate if not
real-time clearing and settlement system.
Within the last few years, IFT has become a fully
functional nationwide means of payment in a number
of countries, including four that we have examined in
detail in this article. International experience with IFT
shows that technology is a necessary but not sufficient
condition for innovation in payments and that enabling
real-time and universal access to deposit accounts at

banks is the key to meeting the public’s needs for more
certain, faster, and universal payment services. Perhaps
the most critical enabling factor is strong sponsorship
by a national body with the responsibility and motiva­
tion to stimulate continuous improvement in the nation­
al payment system. This body might be a consortium
of private banks collaborating through a national pay­
ment association, a public authority such as the central
bank, or a public-private partnership. It is not clear that
such sponsorship can be readily found in the U.S., at
least not at the present time, because there is no national
body that takes responsibility for the development of
the national payment system. As a consequence, IFT
and other national payment innovations are likely to
progress in a halting and incomplete manner and at a
pace that lags innovation that is observable in other
countries, such as those examined in this article.

NOTES
’See https://www.paypal.com.

2The exception to the norm is Japan, where cash is more widely
used than in any other industrialized country due to factors such
as relatively low crime rates, effective anti-counterfeiting measures,
and low-cost nationwide ATM networks (BIS, 2003).
3Depository institutions include banks, thrifts, and credit unions.
In this article, the term “bank” means all depository institutions.
4A full discussion of credit transfers and debit transfers is provided
in the appendix.
5Two-sided markets require the participation of two separate parties
in order to succeed (Rochet and Tirole, 2003). A sender and receiver
of a payment must use the same payment system in order to exchange
monetary value.
6Some checks are converted to electronic format at the point of
acceptance and are cleared through the automated clearinghouse
(ACH) network, as described later.

7A cash payment is also a credit transfer.
8As noted in the introduction, the most prominent example of the
first approach is PayPal. Examples of the second approach include
CashEdge (www.cashedge.com/) and Obopay (https://www.
obopay.com/consumer/welcome.shtml).

includes a provision to allow merchants to offer discounts for
customers who pay with cash or check. (Wall Street Reform and
Consumer Protection Act, §1075).

^Respondents included corporations, technology firms, banks,
payment processors, and infrastructure providers.

14In a joint April 26, 2010, press release, the Federal Reserve System
and The Clearing House Payments Company L.L.C. announced
plans to implement enhanced message formats to support extendedcharacter business remittance information for U.S. dollar wire
transfers on November 11, 2011.
15The findings in this section are based on correspondence with
central bankers and examination of the public websites of payment
services providers, including commercial and central banks and the
financial services arm of the post office. The authors acknowledge
and are grateful for the assistance provided by Ricardo Medina
(Banco de Mexico), Dave Mitchell and Mike Stocks (South African
Reserve Bank), Philipp Haene and Dave Maurer (Swiss National
Bank), and Paul Smee (UK Payments Council), none of whom
bear any responsibility for the descriptions, analysis, and conclusions
presented in this article.
16Straight-through processing (STP) is an operational design based
on standards that allow for fully automated processing of a payment
from its origination by the payer to its receipt by the payee.

9These closed proprietary networks were first described by Kuttner
and McAndrews (2001).

17Also, mobile payments for small accounts using cell phones have
been introduced by PostFinance for payments between PostFinance
account holders.

10The same description applies to transfers among accounts held at
the same bank, called intrabank or “on us” transfers.

18The directory can be found at www.ukpayments.org.uk/
sortcodechecker/.

11 Some cardholders are aware of the delay in the transfer of deposit
money and “play the float” with these transactions. For those card­
holders, debit card transactions are not perceived as immediate.

19One approach would be to charter so-called “narrow banks” that
specialize in payments. This approach has the advantage of encour­
aging innovation, while at the same time prudently extending the
public safety net of deposit insurance to new market entrants
(Litan, 1987).

12Debit card cost structure has become controversial to the point that
recent banking reform legislation directs the Board of Governors
of the Federal Reserve System to regulate merchant fees and

Federal Reserve Bank of Chicago

20Mobile phone numbers, for example, are portable from one carrier
to another.

109

APPENDIX: MODELS OF PAYMENT TRANSACTIONS
Two basic payment models frame the classification of all
types of payment transactions. These are 1) credit transfers
and 2) debit transfers. The end result of these transfers is
the same: Deposit money is transferred from payer to payee.
The process that results in the transfer of deposit money,
however, is quite different. In a credit transfer, deposit
money is moved directly from a payer’s or sender’s trans­
action account to a payee’s or receiver’s account. A credit
transfer is sometimes referred to as a “credit push” pay­
ment, meaning that money is delivered directly to the
receiver based on instructions made by the sender to the
sender’s bank. In a debit transfer, deposit money is moved
in a less direct manner and requires the receiver to request
a transfer from the sender’s bank, based on authorizing
instructions provided by the sender. A debit transfer is
sometimes referred to as a “debit pull” payment, meaning
that the receiver must present the sender’s instruction to
the sender’s bank before deposit money is transferred.
Operationally, payment transactions are more complex
than described in the foregoing paragraph. For purposes
of modeling, a generic payment transaction can be visu­
alized as consisting of two discrete infonnation flows in­
volving “instructions” and “funds movement,” which are
illustrated in figures A1 and A2 for credit and debit pay­
ments, respectively.1 Instructions are shown as solid
lines and funds movements are shown as dotted lines.
For credit transfers, as shown in figure Al, a sender
instructs his/her bank to deliver funds to a designated

110

receiver.2 These instructions result in a debit to the sender’s
transaction account and initiate movement of funds from
the sender’s bank to the receiver’s bank and credit to
the receiver’s account. For debit transfers, as shown in
figure A2, a sender does not directly instruct his/her bank
to transfer funds. Instead, payment instructions follow a
chain from sender to receiver, then from the receiver to
his/her bank, and finally from the receiver’s bank to the
sender’s bank to transfer money from the sender’s account.3
These instructions result in a credit to the receiver’s ac­
count; however, because the receiver’s bank is uncertain
at the time instructions are delivered to the sender’s bank
whether the sender’s bank will honor the instructions, final
credit to the receiver’s account is delayed by the time it
takes the sending bank to detennine whether it will honor
the payment. Accordingly, funds transferred by the debit
transfer method are typically made available to receivers
as provisional funds and are subject to reversal. If the
sender’s bank honors the instructions, then the sender’s
account is debited and provisional funds become final.
’Depending on the payment method and the system used, funds
movement may also include data related to the payment, such as
invoice or remittance infonnation and reference numbers.
2The discussion in this paragraph closely follows Geva (2009).
’For both credit and debit transfers, one or more intermediary
banks may stand between a sender's bank and a receiver's bank to
execute the transfer of deposit money. In addition, senders in both
models may use agents, such as a payroll processing company, to
initiate instructions on their behalf.

3Q/2011, Economic Perspectives

REFERENCES
Allsopp, P., B. Summers, and J. Veale, 2009, “The
evolution of real-time gross settlement: Access, liquidity
and credit, and pricing,” Financial Infrastructure Series:
Payment Systems Policy and Research, The World
Bank, February.

Association for Financial Professionals, 2010,
“2010 AFP electronic payments: Report of survey
results,” November.

Bank for International Settlements, 2003, Payment
and Settlement Systems in Selected Countries, Com­
mittee on Payment and Settlement Systems, CPSS
Publications No. 53, Basel, Switzerland, April.
________________ , 2001, “Core principles for systemically important payment systems,” report, Committee
on Payment and Settlement Systems, CPSS Publications
No. 43, Basel, Switzerland, January.

Board of Governors of the Federal Reserve System,
2010, The 2010 Federal Reserve Payments Study—
Noncash Payment Trends in the United States: 2006—
2009, research report, December 8.
________________ , 2007, The 2007 Federal Reserve
Payments Study—Noncash Payment Trends in the United
States: 2003-2006, research report, December 10.
________________ , 2006, “A summary of the round­
table discussion of the role of wire transfer in making
low-value payments,” Payments System Policy Advi­
sory Committee, report, Washington, DC, May, avail­
able at www.federalreserve.gov/paymentsystems/
lowvaluepay/lowvaluepayments.pdf.
________________ , 2004, The 2004 Federal Reserve
Payments Study—Analysis ofNoncash Payments Trends
in the United States: 2000-2003, research report,
December 15.
________________, 2002, “The future of retail electronic
payments systems: Industry interviews and analysis,”
staff study, No. 175, December.

Boland Hill Media, LLC, 2010, “Latest ACH stats
reflect consumer shift to electronic bill pay,” Digital
Transactions News, August 20, available at www.
digitaltransactions.net/newsstory.cfm?newsid=2616.

Federal Reserve Bank of Chicago

Foster, K., E. Meijer, S. Schuh, and M. Zabek,
2010, “The 2008 survey of consumer payment choice,”
Public Policy Discussion Papers, Federal Reserve
Bank of Boston, No. 09-10, April.
Gerdes, G.R., 2008, “Recent payment trends in the
United States,” Federal Reserve Bulletin, pp. A75A106, October.

Geva, B., 2008, “Payment finality and discharge in
funds transfers,” Chicago-Kent Law Review, Vol. 83,
No. 2, pp. 633-675.

Herbst-Murphy, S., 2010, “Trends and preferences
in consumer payments: Lessons from the VISA payment
panel study,” Federal Reserve Bank of Philadelphia,
discussion paper, May.
Hough, Daniel, George Simotas, Kim Allin, and
John Hansen, 2010, “Shifting to debit,” BAI Banking
Strategies, Bank Administration Institute, October 20.

Kuttner, K., and J. McAndrews, 2001, “Personal
on-line payments,” Economic Policy Review, Federal
Reserve Bank of New York, Vol. 7, No. 3, December,
pp. 35-50.
Litan, R., 1987, What Should Banks Do?, Washington,
DC: The Brookings Institution.

McKinsey & Company, 2009, “2009 U.S. payment
revenues,” unpublished PowerPoint slide.
Rochet, J-C., and J. Tirole, 2003, “Platform competition
in two-sided markets,” Journal of the European Eco­
nomics Association, Vol. 1, No. 4, June, pp. 990-1029.

Rysman, M., 2010, “Consumer payment choice:
Measurement topics,” in The Changing Retail Payments
Landscape: What Role for Central Banks?, Federal
Reserve Bank of Kansas City, pp. 61-81.
Shevlin, R., J. Fishman, and G. Bezard, 2010,
“Sizing person-to-person payments in the United
States, United Kingdom, and Australia,” Aite Group,
LLC, report, No. 2010110301, October.

Smith, K., 2010, “Consumer payment choice: Measure­
ment topics—Commentary,” in The Changing Retail
Payments Landscape: What Role for Central Banks?,
Federal Reserve Bank of Kansas City, pp. 83-92.

111

VocaLink and PriceWaterhouseCoopers, 2009,
“Tomorrow happened yesterday: How banks are
building a business case for Faster Payments,” report,
available at www.finextra.com/finextra-downloads/
newsdocs/tomorrow-happened-yesterday—how-banksare-building-a-business-case-for-faster-payments.pdf.

World Bank Group, The, 2008, Payments Systems
Worldwide: A Snapshot. Outcomes of the Global Pay­
ment Systems Survey 2008, Financial Infrastructure
Policy and Research Series.

Wells, Kirstin E., 1996, “Are checks overused?,”
Federal Reserve Bank ofMinneapolis Quarterly
Review, Vol. 20, No. 4, Fall, pp. 2-12.

112

3Q/2011, Economic Perspectives

How do benefit adjustments for government transfer programs
compare with their participants’ inflation experiences?
Leslie McGranahan and Anna L. Paulson

Introduction and summary
Millions of Americans rely on government transfer
programs as a way to make ends meet during a temporary
setback or as their main source of income during retire­
ment. Whether individuals qualify for unemployment
assistance, Temporary Assistance for Needy Families
(TANF), Social Security,1 or Supplemental Security
Income (SSI), the level of benefits they will receive is
affected by how the benefits are adjusted to deal with
inflation—the general increase in prices for goods and
services over time. Changes in benefit levels to address
inflation—that is, cost-of-living adjustments (COLAs)—
are determined by formulas that vary depending on the
program in question. Adjustments to some programs’
benefits are made automatically based on a government
inflation index, while adjustments to others require
legislative action.
COLAs can have a substantial impact on the welfare
of transfer program participants. Those who receive
benefits from a program for a long time are particularly
affected by the formulas determining COLAs. In addi­
tion, COLAs can have a large impact on transfer pro­
gram costs. For example, the bipartisan National
Commission on Fiscal Responsibility and Reform
(chaired by Democrat Erskine Bowles and Republican
Alan Simpson) recently proposed changing the way
COLAs are made for Social Security benefits. By making
Social Security COLAs using a chain-weighted Con­
sumer Price Index (C-CPI),2 as opposed to the current
method that relies on the Consumer Price Index for
Urban Wage Earners and Clerical Workers (CPI-W),
benefits are expected to increase by about 0.3 percentage
points less each year. This small change in the formula
for determining the Social Security COLAs would
significantly affect both the benefits received from the
program and the program’s costs. According to our
calculations, if inflation measured by the CPI-W

Federal Reserve Bank of Chicago

averaged 2.5 percent per year for the next 15 years, an
individual receiving $25,000 in Social Security payments
this year would receive 15 years from now an annual
payment of $36,207. Under the chain-weighted formula,
assuming inflation averaged 2.2 percent per year (0.3
percentage points less than under the current formula),
the same individual would receive 15 years from now
an annual payment of $34,650. According to the
National Commission on Fiscal Responsibility and
Reform (2010, pp. 54, 65), if the proposed change in
the method for calculating COLAs for Social Security
were enacted, this change would lead to savings of
$89 billion over the period 2012-20 and would reduce
the Social Security shortfall by 26 percent over 75 years.
Major U.S. transfer programs target individuals
with particular characteristics—for example, single
mothers and the elderly. These individuals are likely
to have different spending patterns than the average
individual. However, programmatic COLAs are typi­
cally based on aggregate inflation measures. Since dif­
ferent groups of individuals purchase different goods
and services, they may face a rise in their cost of living
that differs from that of the average household. For
example, the elderly spend more on health care than
the general population, and commuters spend more
on transportation. If health care costs increase more
rapidly than aggregate prices, then the inflation expe­
rienced by the elderly will be greater than general in­
flation. Similarly, if gas costs and therefore the costs
of transportation increase rapidly, then commuters will
face inflation that is higher than that of the general
Leslie McGranahan is a senior economist and Anna L. Paulson
is a vice president and director offinancial research in the
Economic Research Department at the Federal Reserve
Bank of Chicago. The authors thank Bruce Meyerfor help­
ful discussions and Daniel DiFranco, Lori Timmins, and
Nathan Marwell for excellent research assistance.

113

population. COLAs for major transfer programs do
not typically account for these differences in spend­
ing patterns.
In this article, we measure the inflation experienced
by different groups of people. We focus on groups that
are likely recipients of federal benefits: the elderly, single
mothers, individuals with less than a high school diploma,
the disabled, and the poor. We compute group-specific
inflation measures for the period 1980-2010, using data
on group spending patterns from the U.S. Bureau of
Labor Statistics’ (BLS) Consumer Expenditure Survey
in combination with item-specific inflation measures,
also from the BLS. We then compare our group-specific
inflation measures with the COLAs used for major
transfer programs to evaluate whether program benefits
that are adjusted using aggregate measures of inflation
or using other means “keep up” with the inflation ex­
perienced by a specific group.
COLAs can affect the welfare of transfer program
recipients in (at least) two ways: by determining the
initial level of benefits that they receive and by deter­
mining how benefit payments grow during program
participation. The latter is particularly important for
programs like Social Security that individuals often
participate in for a long time, and the former is a key
factor of programs like TANF that individuals usually
participate in for shorter periods.
We find that the elderly and, to a lesser extent, the
disabled, the poor, and those with less than a high school
diploma experienced higher inflation than the aggregate
population from 1980 through 2010. Because the Social
Security/SSI COLA is based on aggregate inflation,
Social Security/SSI COLAs have been less than the price
increases experienced by the elderly in most periods
since 1980. More specifically, in 2010, Social Security
benefits for an individual who had been on the program
since 1980 would be 265 percent of their nominal 1980
value, while the cost of the items purchased by the aver­
age elderly household was 270 percent of their nominal
1980 value. Inflation faced by the disabled, while above
aggregate inflation, has been slightly below the Social
Security/SSI COLA because of nuances in COLA deter­
mination. Single mothers experienced lower inflation
than the overall population during this period, but the
inflation they faced was larger than the increases in ben­
efits from the Aid to Families with Dependent Children
(AFDC) program and TANF. Increases in welfare ben­
efits from the AFDC and its successor program, TANF,
in most states have been substantially below both aggre­
gate inflation and the price increases faced by single
mothers over the period 1980-2010. In addition, we
find that the growth in benefits from the Supplemental
Nutrition Assistance Program, or SNAP (formerly

114

called the Food Stamp Program), has exceeded the infla­
tion faced by single mothers, the disabled, the poor, and
those with less than a high school diploma over the
period 1980-2010, largely because of increases in
benefit levels enacted as part of the American Recovery
and Reinvestment Act (ARRA) of 2009.
During the recent recession and subsequent recov­
ery, U.S. inflation has been atypically low. Also, during
this period, there have been somewhat unusual COLAs
for both Social Security/SSI benefits and SNAP bene­
fits. Because this period is unique from both an infla­
tion perspective and policy perspective, we break our
analysis into two periods: 1980-2008 and 2008-10.
The rest of our article is organized as follows. In
the next section, we discuss major U.S. transfer pro­
grams and report how benefits from these programs
are adjusted for inflation. Then, we describe the char­
acteristics of program recipients and compare them
with the overall U.S. population. These comparisons
are used to identify the groups whose inflation experi­
ences we would like to investigate. Next, we compare
the inflation experiences of these groups with the in­
flation experience of the aggregate U.S. population, and
discuss how these comparisons were generated. We
also compare group inflation experiences with program­
matic COLAs. We highlight four programs in our anal­
ysis of COLAs: Social Security, SSI, TANF, and SNAP.3
Finally, we review our conclusions and briefly discuss
them in the context of the policy debate concerning
COLAs, which has chiefly revolved around the Social
Security program.

COLAs for major transfer programs
The federal government transfers money to many
different recipient populations through a large variety
of targeted programs. Table 1 lists all of the federal gov­
ernment’s transfer programs with total direct payments
and indirect payments (which are largely payments made
via states) to individuals that exceeded $10 billion in
fiscal year 2010, according to the 2012 federal budget.
This table also lists the outlays on the program, the num­
ber of beneficiaries served, the way in which benefits
or expenditures are adjusted for changes in the price
level, and a brief description of the eligibility criteria.
There are 18 such programs, which served a total of
379 million recipients in 2010, indicating that the
average American is served by more than one of
these programs.
Combined, these programs cost the federal govern­
ment $2.2 trillion in fiscal year 2010 and made up over
95 percent of all federal payments to individuals. These
programmatic expenses represented approximately
60 percent of all federal government outlays in fiscal

4Q/2011, Economic Perspectives

Federa l Res erv e Bank of Chica go

TABLE 1

Government transfer programs with payments exceeding S10 billion, fiscal year 2010

Program

Outlays for payments

Beneficiaries

to individuals
(in millions of dollars)

(annual average
in thousands)

Inflation adjustment/
benefit determination

Eligibility

Beneficiaries receive low-cost or free

Children in school or child care from families with incomes

(not including Women, Infants,
and Children program and

nutritionally balanced breakfasts and
lunches. Payments to a state depend

at or below 130 percent of the FPL are eligible for free meals;
those from families with incomes between 130 percent and

Commodity Supplemental Food
Program) and Special Milk Program

on changes in the Food Away from
Home series in the CPI-U.

185 percent of the FPL are eligible for reduced price meals.

CPI-W (July-Sept.) determines cost-ofliving adjustment.

Retired federal government workers.

Thresholds and maximum credit are
adjusted by CPI-U (for the 12-month

Low-income working individuals and families. For single­
parent families with two children, annual income must be

period ending August 31).

below $40,363; for single individuals, annual income must
be below $13,460 (2010 tax year).

Child nutrition programs

16,430

34,889

Civil Service Retirement System

69,407

2,523

Earned income tax credit
(refundable portion)8

54,712

21,743

Hospital and medical care
for veterans'5

38,216

5,639

In kind. No maximum benefit level.

Individuals who actively served in the military (priority to
those with service-connected disabilities and low income).

Housing assistance'

49,959

3,200

In kind. For public housing rental payment
equal to 30 percent of monthly adjusted

Income below limits that are area specific (30 percent,
50 percent, and 80 percent of median).

income.

Tenant-based rental assistance/

17987

2,100

housing choice vouchers
(Section 8)c

Benefit amount equal to the difference

Income below some percentage (between 50 percent

between 30 percent of adjusted household
income and the public-housing-authority-

and 80 percent) of local area median. Housing authorities
can have additional criteria. At least 75 percent of new

determined payment standard (90 percent
to 110 percent of the fair market rent, or

households must have income at or below 30 percent
of the area median.

FMR). FMR is set by the federal government
as average gross rents (including utilities)

for medium-quality apartments.

Medicaid

272,771

59,339

In kind. No maximum benefit level.

States are required to serve pregnant women and children

under six years old with income below 133 percent of the
FPL, children aged six to 19 up to 100 percent of the FPL,
and Supplemental Security Income recipients. States can
include other groups.
Medicare: Hospital Insurance
(Part A)

250,223

46,906

In kind. No premium for most beneficiaries.

Free if aged 65 years or older and individual or spouse is
receiving Social Security or if under 65 and receiving Social

Security disability benefits. Nonqualifying individuals aged
65 or older can pay for Part A coverage.

11 5

11 6

TABLE 1 (CONTINUED)

Government transfer programs with payments exceeding S10 billion, fiscal year 2010

Program
Medicare: Supplemental Medical

Outlays for payments
to individuals

Beneficiaries
(annual average

Inflation adjustment/

(in millions of dollars)

in thousands)

benefit determination

Eligibility

268,945

43,569

Monthly premiums, deductibles, and

Anyone eligible for Medicare Part A can pay a premium

co-insurance amounts are adjusted by
the federal government, as determined

and enroll in Part B.

Insurance (Part B)

by formula. Premium is required to be the
amount needed to cover 25 percent of

estimated program costs for enrollees
aged 65 years and older.

Military retirement

Refundable (additional) child creditd

50,585

22,659

2,212

18,160

CPI-W (July-Sept.) determines cost-of-living

Retired members of the military (no specific age

adjustment.

requirement).

Child credit = $1,000 nominal; no automatic
adjustment. Set legislatively. Has been

Tax filers with children under 17 years old whose tax liability
is not large enough to fully discharge the $1,000 per child

increased on occasion—Economic Growth
and Tax Relief Reconciliation Act of 2001

credit (or the amount remaining after phaseouts) can get the
additional child credit. Child credit is reduced by 5 percent of

doubled the credit. It will return to $500
in 2013.

adjusted gross income (AGI) for married filers with AGI over
$110,000 and single parents with AGI over $75,000. The

maximum additional child tax credit is limited to 15 percent
of earnings above a threshold ($3,000 in 2010; historically
this threshold was indexed for inflation using CPI-U).

Social Security: Disability Insurance

123,507

9,822

CPI-W (July-Sept.) determines automatic

Individuals who worked in Social Security-covered

cost-of-living adjustment.

employment for a sufficient amount of time, are under
retirement age, and are unable to perform any substantial
work for at least one year.

Social Security: Old-Age and
Survivors Insurance

576,578

43,110

CPI-W (July-Sept.) determines automatic
cost-of-living adjustment.

Individuals who contributed to the program for 40 quarters
or more and who have reached the minimum retirement

age (62 years old). Surviving spouses and children of
contributors are also eligible.
Student assistance—U.S. Department

46,768

20,638

4Q /2 011, Econom ic Per spe ctiv es

of Education and other (primarily
Pell Grants and subsidized

Maximum loan amounts are nominally fixed,

Pell Grants for students with family annual income below

but periodically adjusted.

$45,000 (most with family annual income below $20,000).
Subsidized Stafford Loans based on “financial need.”

Stafford Loans)
Changes in the cost of items (using the

Household gross monthly income at or below 130 percent

Program (officially known as the
Food Stamp Program until

Consumer Price Index) in a “market basket"
based on the Thrifty Food Plan (TFP) from

of the FPL; net income (net of allowances) below 100 percent
of the FPL; and assets below $2,000 (or below $3,000 if

October 1,2008, and still commonly
referred to by this name)

June determine benefit levels starting in
October. Revised TFP originally relied

one person is aged 60 years or older or is disabled).

Supplemental Nutrition Assistance

70,492

40,302

on prices paid by low-income households.
The American Recovery and Reinvestment

Act of 2009 led to unusual adjustments.

Federa l Res erv e Bank of Chica go

TABLE 1 (CONTINUED)

Government transfer programs with payments exceeding $10 billion, fiscal year 2010

Program

Outlays for payments

Beneficiaries

to individuals
(in millions of dollars)

(annual average
in thousands)

Supplemental Security Income (SSI)

43,886

7522

Inflation adjustment/
benefit determination

Eligibility

CPI-W (July-Sept.) determines automatic
cost-of-living adjustment.

Aged (65 years old or older), blind, or disabled with assets
below $2,000 per individual or $3,000 per couple. Benefit

phases out with income. Benefit fully phased out for those
with monthly income (net of allowances) greater than $694
per individual and $1,031 per couple.

Temporary Assistance for Needy Families'-

Unemployment assistance

21,936

158,263

4,594

11,429

Maximum benefits have been fixed at

Families with dependent children and pregnant women

nominal levels since 1996 in many
jurisdictions. Legislated changes in others.

with income and assets below a state-determined threshold.
Phases out at about 75 percent of FPL in average state.

No inflation adjustment in grants to states.

Households subject to work requirements and time limits.

Benefits set at a fraction of weekly wage up
to a maximum. In 36 jurisdictions, maximum

Recently unemployed workers who are unemployed
through no fault of their own, earned qualifying wages,

benefit levels automatically adjust according
to weekly wages of covered employees. In

and are actively seeking work.

remainder, maximum benefits are set at a
fixed dollar amount that can be changed
by legislation.

Veterans service-connected compensation

43,377

3,498

CPI-W (July-Sept.) determines cost-of-living

Veterans who are at least 10 percent disabled as a

adjustment. Not automatic. Needs legislative
approval. Typically granted unanimously.

result of military service, and their survivors.

aThe number of earned income tax credit (EITC) beneficiaries is based on the number of tax returns with refundable EITC for tax year 2008 (see Internal Revenue Service, Statistics of Income, 2010, table 2.5).
bThe number of beneficiaries of hospital and medical care for veterans is based on the total number of patients from www.va.gov/vetdata/Utilization.asp.
cHousing assistance and tenant-based rental assistance beneficiary estimates are from the U.S. Department of Housing and Urban Development (2011). Tenant-based rental assistance/housing choice vouchers
(Section 8) is a subcategory of housing assistance.
dThe number of refundable (additional) child credit beneficiaries is based on number of tax returns with additional child tax credit for tax year 2008 (see Internal Revenue Service, Statistics of Income, 2010, table A).
eThe Temporary Assistance for Needy Families caseload data for 2010 are from www.acf.hhs.gov/programs/ofa/data-reports/caseload/caseload2010.htm. This program became the successor to the Aid to Families
with Dependent Children program in 1997.
Notes: CPI-U means Consumer Price Index for All Urban Consumers. CPI-W means Consumer Price Index for Urban Wage Earners and Clerical Workers. FPL means federal poverty line. See note 6 for the definition
of market basket. In some cases, there are minor differences between the coverage of the cost and beneficiary numbers.
Sources: Outlays data from White House, Office of Management and Budget (2011 b), table 11.3; beneficiaries data for most programs from White House, Office of Management and Budget (2011a), table 27-5;
and some beneficiaries data, inflation adjustment/benefit determination data, and eligibility data from various government sources.

11 7

year 2010.4 For some of these programs, in particular
Medicaid and unemployment assistance, state govern­
ments also expend significant sums of money The only
state dollars that are included in table 1 are those that
were funded by transfers from the federal government.
Inflation adjustment/benefit determination infor­
mation is presented in table 1 (pp. 115-117). The infla­
tion adjustment/benefit determination column explains
in detail how program benefits are adjusted for inflation
for an individual once he is already enrolled in the
program. For many programs such as TANF and SNAP,
the level of benefits upon initial enrollment is set using
the same formula. For other programs, initial benefits
are set using a ditferent formula. For example, the initial
Social Security benefit level depends on earnings over
the recipient’s working life.
The inflation adjustment/benefit determination
column in table 1 shows that there are four main types
of adjustments used by these programs. First, some
programs adjust benefit levels based on an aggregate
inflation index—either the Consumer Price Index for
All Urban Consumers (CPI-U) or CPI-W. The programs
in this category are Civil Service Retirement System,
earned income tax credit (EITC), military retirement,
Social Security (both Old-Age and Survivors Insurance
and Disability Insurance), SSI, and veterans serviceconnected compensation.5 These tend to be the large
income-transfer programs. The CPI-U and CPI-W are
the two aggregate indexes released by the BLS. They
are both consumer price indexes and as such represent
changes in the cost of “market baskets”6 consumed
by different demographic groups. The CPI-W is cal­
culated based on price increases for goods consumed
by households for which at least half of household in­
come comes from the earnings of workers in hourly
wage or clerical jobs. This index represents about
32 percent of the U.S. population. The CPI-U is based
on the market basket of all urban consumers; it repre­
sents 87 percent of the population. Second, some pro­
grams have benefits that are linked to price growth in
a particular category. The programs in this category
are child nutrition programs (in particular, the National
School Lunch and Breakfast Programs) and the Special
Milk Program; SNAP; and tenant-based rental assis­
tance (Section 8 housing assistance). These programs
are supporting consumption in a specific category, and
therefore, their benefits are linked to price growth in
that category. Third, some programs have no inflation
adjustment because benefits are in kind. These programs
are hospital and medical care for veterans, Medicare
(Parts A and B), Medicaid, and non-Section 8 housing
assistance. While there is no explicit benefit adjustment,
the value of the benefits increases as the cost of the

118

underlying good increases. A final set of programs
has benefits that are nominally fixed in value. Benefit
amounts can be changed through legislation. These
programs are student assistance, the refundable (addi­
tional) child tax credit, and TANF. Unemployment
assistance, for which the increases in benefits are based
on wage growth, does not fall into any of these cate­
gories. No programs are linked to the broad-based ex­
penditure needs of the program’s recipient population.
If we combine the program costs for the programs
that link to the CPI-U and CPI-W, we find that about
$960 billion in annual expenditures was linked to these
indexes, representing about 28 percent of total federal
expenditures for 2010.
In addition to indexing benefit levels to inflation,
the federal government indexes eligibility criteria for
many transfer programs to inflation. In many cases
eligibility is based on federal poverty guidelines, which
are indexed to the CPI-U. Also, many features of the
tax code, such as personal exemptions and tax brackets,
are indexed (Hanson and Andrews, 2008).7

Characteristics of program participants
Next, we are interested in finding out what percent­
age of the population benefits from these programs
and which demographic groups are especially depen­
dent on benefit payments. Benefit levels, and hence
COLAs, are especially important to households that
receive a large fraction of their income from federal
government transfer programs. We divide the popula­
tion in six different ways—by education, age, disabil­
ity status, family structure, veteran status, and poverty
status. We choose these six methods of segmenting the
population because they are in keeping with program
eligibility standards and because the groups based
on these different division criteria are some of the
groups that are highlighted in other research on trans­
fer program participation (see, for instance, Meyer
and Rosenbaum, 2001, and Haveman et al., 2003).
In addition, we are interested in groups whose recipient
status tends to be fairly persistent. Gaps between pro­
grammatic inflation adjustments and household expen­
diture growth will be more relevant if households benefit
from programs over long periods so that the gaps are
compounded over time. Because of this issue, we do
not look at population groups based on work status
because employment status has historically been fluid.
In box 1, we describe the criteria we use for the
inclusion of households in the groups listed there. As
delineated in box 1, our definition of the disabled only
includes those individuals who do not have a job rather
than all individuals with a disability. In table 2, we
present results showing what fraction of households

4Q/2011, Economic Perspectives

BOX 1

Demographic group variable descriptions

Variable

Description

Less than high school diploma

Neither the reference person nor spouse completed high school.

High school graduate, no college

Reference person or spouse obtained a high school diploma; neither
the reference person nor spouse pursued education beyond the high
school level.

Some college or more

Reference person or spouse pursued education beyond the high
school level.

Elderly

Reference person or spouse is at least 65 years old.

Disabled

Reference person or spouse is out of work because of a chronic health
condition or disability.

Single mother

Reference person is an unmarried female aged 18-64 years old; the
reference person’s child who is younger than 18 years old lives in the
household.

Other parent

Reference person aged 18-64 years old is either an unmarried male or
a married male or female; the reference person’s child who is younger
than 18 years old lives in the household.

Nonparent

Reference person is aged 18-64 years old; the reference person has
no children who are younger than 18 years old living with him or her
in the household.

Veteran

Reference person or spouse served on active duty in the U.S. Armed
Forces at some point in his or her lifetime (currently active members
of the Armed Forces are included).

Poor

Household’s income was below the poverty line (adjusted for household
size and composition) during the last month of reference period.

in these groups were recipients of benefits from the
different programs listed in table 1 (pp. 115-117). This
information is calculated from wave 4 of the 2004 panel
of the U.S. Census Bureau’s Survey ofIncome and
Program Participation (SIPP), corresponding to the
January-April period of 2005. The results displayed in
table 2 are consistent with the eligibility criteria outlined
in table 1. For example, 49 percent of families headed
by a single mother receive a benefit from the National
School Lunch and Breakfast Programs, while only
2 percent of households without children report receiving
a benefit from these programs. Similarly, the vast major­
ity of the elderly households receive both Medicare and
Social Security.
In table 3 (p. 122), we show the median and aver­
age numbers of transfer programs that members of these
different groups participated in. The median house­
hold of the overall sample receives a benefit from
one of these programs (table 3, final row). For many
groups, the median household participates in no benefit

Federal Reserve Bank of Chicago

programs—these groups are the households with some
college or more, nonelderly households, nondisabled
households, non-single-mother households, nonveteran
households, and the nonpoor. By contrast, among a num­
ber of groups, the median household benefits from two
programs—these groups are those with less than a high
school diploma or only a high school diploma, the elderly,
single-mother households, veteran households, and the
poor. The median disabled household receives benefits
from three programs. This suggests that program receipt
is fairly concentrated. Given the overlapping eligibility
criteria for many programs, this degree of concentration
is not surprising. The pattern for average benefit receipt
among the various demographic groups is quite simi­
lar to that for median benefit receipt. Our measure of
the average number of cash transfer programs excludes
those programs providing in-kind benefits. There is a
notable gap between the average number of transfer
programs and the average number of cash transfer
programs used by single mothers and the elderly.

119

120

0. 90
1.3 8
1.0 7

7.16
6.82

10 0. 00
All gr ou ps

9.8 2

11 .93
88 .0 7

Po or
Non po or

1.81

4. 23
0.45
5. 85
0. 78
0.11

18 .32
81 .6 8

Ve te ra n
Non ve te ra n

2. 02

7.23
25 .7 0
46 .4 4
Sing le m ot he r
O th er pa re nt
Non pa re nt

19.51

6.20
93 .8 0
Di sa ble d
Non disa bled

2. 04

28 .5 8
7.27

12 .06
6. 03

34 .0 8
3.52

0.91
1.0 9

0.38
1.0 2

0. 55
1.5 7
4. 77
0. 24
2.39
8.24
7.51

5.31

2. 69

3.49

5. 06
1.0 8
3.0 5
11 .33

5.22
0.29
0.58

6. 47
0. 46
1.1 2
0. 28
0. 88
1.0 3
31 .1 0
5.37

0. 10
0. 15
0. 94
49 .2 8

12.81

30 .4 9

5.00
0. 63

6. 68
1.03
1.8 6
1.0 2

32 .0 0
5.52

5. 64

0. 55
1.90
16 .59
9.3 7

6.91

0.87

1.4 7
11 .99

20 .6 4
79 .3 6
El de rly
N on el de rly

0.61

0.61

0.91

1.3 5
1.3 9

1.6 7
0.91

5.18
7.68

0. 62
8. 52
6. 45

24 .9 0
14 .18
6.7 4
7.65
22 .6 6
69 .69

0. 74
1.7 3
1.9 6

9.15
8. 40
5. 99

22 .7 2
11 .42
4.07

0. 63
1.1 3
1.10

3. 45
2. 16
0. 90

2.12
1.3 9

)

Le ss than hig h sc ho ol diplom a
Hig h sc ho ol gr ad ua te , no co lle ge
So me co lle ge or more

Fo od Sta m p
Prog ram
1
1

ta x cr ed it

( ....................................................................................................................................... p e rc e n t ...........................................................................................................................................

Hou si ng
as si st an ce

Te na nt -b as ed ren tal
as si st an ce /h ou si ng
ch oi ce vo uc he rs
(S ec tio n 8)
Ho sp ita l an d
m ed ic al ca re
fo r ve te ra ns
Ea rned
in co m e

Civil Ser vi ce
Ret ire m en t
Sy ste m

Nat io na l
Sc ho ol Lu nc h
an d Bre ak fa st
Pr og ra m s ’
Pe rcen t
of
sa m pl e

Percentage of households receiving government transfer program benefits, by demographic group

Next, we investigate what percent­
age of household income is received
from the transfer programs listed in
table 2. We do this in two steps. First,
in table 4, we show the average benefit
received from the different programs
by demographic group. These are aver­
age monthly benefit amounts among
all households in the group. In the final
column of table 4, we sum cash transfer
income across all the different programs.
The elderly receive the largest transfers
on average per month ($1,420), followed
by the disabled ($1,059) and veterans
($999). Second, in table 5 (p. 124), we
tabulate the percentage of total house­
hold income that is received from the
different transfer programs. In this
case, household income is defined as
the sum of cash transfers, the value of
Food Stamp Program (Supplemental
Nutrition Assistance Program) benefits,
and other income. While the average
U.S. household receives 11 percent of
its income from transfer programs
(table 5, final row), some groups of
households receive (on average) nearly
half of their income from these pro­
grams. In both tables 4 and 5, we are
not imputing values to in-kind assis­
tance, such as Medicare and housing
assistance, so these numbers under­
estimate true total transfer benefits.
We have looked at program partici­
pation, benefit levels, and income ratios.
By examining transfer programs and
their participants in this way, we find
that there are certain demographic groups
that are particularly dependent on trans­
fer income. We choose to further inves­
tigate those groups whose average
household (based on the data in table 3)
receives benefits from two or more
transfer programs (namely, those with
less than a high school diploma, the el­
derly, the disabled, single mothers, and
the poor) and also those groups whose
ratio of average monthly transfer in­
come to average total monthly income
(based on the data in table 5) exceeds
25 percent (namely, those with less
than a high school diploma, the elderly,
the disabled, and the poor). By using

4Q/2011, Economic Perspectives

Federa l Res erv e Bank of Chica go

TABLE 2 (CONTINUED)

Percentage of households receiving government transfer program benefits, by demographic group

Medicaid

Medicare

Military
retirement

Social
Security:

Social Security:
Old-Age and

Supplemental

Temporary
Assistance

Disability
Insurance

Survivors
Insurance

Security
Income

for Needy
Families

Veterans
serviceUnemployment
assistance

k...................

connected
compensation
............................. J

Less than high school diploma

47.34

41.71

0.18

10.35

36.21

16.08

4.01

2.37

1.20

High school graduate, no college
Some college or more

27.35
13.06

34.86
19.40

1.09
1.72

7.42
4.14

31.48
17.97

7.66
2.78

2.38
0.90

2.27
1.77

1.43
2.16

Elderly
Nonelderly

14.86
19.97

95.63
6.14

3.24
1.00

5.85
5.23

91.14
4.56

6.49
4.49

0.30
1.78

0.46
2.31

3.16
1.59

3.77

Disabled

56.10

49.84

1.24

48.69

15.62

33.22

6.31

1.33

Nondisabled

16.46

22.94

1.48

2.49

22.88

3.03

1.16

1.97

1.79

Single mother

54.47

3.49

0.04

3.20

2.04

8.56

9.11

1.86

0.36

Other parent
Nonparent

23.78
12.50

2.66
8.48

0.75
1.29

2.16
7.24

1.65
6.56

2.35
5.04

1.43
0.84

2.68
2.18

1.27
1.97

Veteran
Nonveteran

9.53
21.03

43.34
20.41

6.85
0.25

6.34
5.14

42.00
18.04

2.28
5.49

0.31
1.74

1.37
2.06

9.47
0.22

Poor
Nonpoor

50.86
14.59

18.72
25.41

0.06
1.65

7.37
5.08

12.71
23.74

15.18
3.51

7.09
0.72

2.66
1.83

0.68
2.08

All groups

18.92

24.61

1.46

5.36

22.43

4.90

1.48

1.93

1.92

aThe Survey of Income and Program Participation asks households specifically about their participation in the National School Lunch and Breakfast Programs, but federal budget sources used for table 1 cover
the broader category of child nutrition programs and the Special Milk Program.
bEarned income tax credit (EITC) data are missing for about 12 percent of the sample.
Notes: For descriptions of the demographic group variables, see box 1 on p. 119. The sample is limited to metropolitan-based households in which the head is at least 18 years old. There is insufficient detail in the
Survey of Income and Program Participation to measure the participation in or the benefit value from the refundable (additional) child credit and student assistance, which appear in table 1, as well as distinguish
between Medicare Parts A and B. The single mother, other parent, and nonparent households sum to less than 100 percent in the first column of data because the elderly are excluded from these demographic
groups (see box 1).
Source: Authors’ calculations based on data from wave 4 of the 2004 panel of the U.S. Census Bureau, Survey of Income and Program Participation, corresponding to the January-April period of 2005.

12 1

TABLE 3

Median and average government transfer program participation, by demographic group
Median
number of
programs

Average
number of
programs

Average
number of cash
transfer programs

Less than high school diploma
High school graduate, no college
Some college or more

2
2
0

2.33
1.67
0.94

1.05
0.77
0.45

Elderly
Nonelderly

2
0

2.40
0.91

1.23
0.39

Disabled
Nondisabled

3
0

2.93
1.11

1.53
0.51

Single mother
Other parent
Nonparent

2
0
0

2.31
0.97
0.63

0.87
0.32
0.35

Veteran
Nonveteran

2
0

1.48
1.17

0.81
0.52

Poor
Nonpoor

2
0

2.11
1.09

0.93
0.52

All groups

1

1.23

0.57

Notes: For descriptions of the demographic group variables, see box 1 on p. 119. For our measure of participation in cash transfer programs,
we exclude programs that supply in-kind benefits to recipients, such as Medicare and housing assistance. We include the Food Stamp
Program in our measure of cash transfer program participation.
Source: Authors’ calculations based on data from wave 4 of the 2004 panel of the U.S. Census Bureau, Survey of Income and Program
Participation, corresponding to the January-April period of 2005.

these two different criteria, we end up focusing on
the same groups, with the exception of single mothers,
who receive 13 percent of their monthly income from
transfer programs but are covered by 2.3 programs
on average. This discrepancy arises because many
single-mother households receive benefits from the
child nutrition programs and Medicaid, which are inkind programs and thus not included in our transfer
income calculations.8

Group expenditure patterns and
inflation rates
We next look at the expenditure patterns of house­
holds (or “consumer units”) in these five groups: those
with less than a high school diploma, the elderly, the
disabled, single mothers, and the poor. More specifi­
cally, we use data from the Consumer Expenditure
Survey to investigate whether their expenditure pat­
terns conform to those of the general population. We
measure these expenditure patterns by using merged
data from the Diary and Interview portions of the sur­
vey over the period 1980-2009.9 The unit of analysis
in the Consumer Expenditure Survey is the consumer
unit—a grouping defined as either a single individual
who makes independent consumption decisions, a group
of related individuals, or a group of individuals who
live together and make joint consumption decisions.10

122

We use the term “consumer unit” interchangeably with
“household” throughout this article. We define a house­
hold as having less than a high school diploma if both
the head and spouse have not graduated from high school.
We define a household as elderly if either the head or
spouse is aged 65 or over. We define a household as
headed by a single mother if the household contains
children younger than 18 and is headed by an unmarried
female aged 18-64. We define a household as disabled
if the head or spouse was not working during the past
12 months because he or she was “ill, disabled or un­
able to work,” as stated in the Consumer Expenditure
Survey's Codebook. Our definitions here are consistent
with the definitions presented in box 1 (p. 119) that
we used in our analysis of transfer program participa­
tion and the sources of transfer income in tables 2-5.
In table 6, panel A, we report 2009 expenditure
shares for our groups of interest, their complements,
and the overall population. The number 14.2 in the top
row of the column labeled “food” means that among
the entire population, 14.2 percent of all expenditures
is on food items. We refer to these expenditure shares
as the market baskets of households. For all expendi­
ture categories except for housing, these market baskets
are based on the out-of-pocket expenditures of house­
holds. For example, if a hospital visit was paid for
by Medicaid, it would not be included in household

4Q/2011, Economic Perspectives

Federa l Res erv e Bank of Chica go

TABLE 4

Average monthly income from government transfer programs, by demographic group
Social Security
(Old-Age,
Survivors,
and Disability
Insurance)

Supplemental
Security
Income
(federal and
state amounts)

Food
Stamp
Program

Military
retirement

Civil Service
Retirement
System

Temporary
Assistance
for Needy
Families (and
other welfare
amounts)

Unemployment
assistance
(state and
supplemental

unemployment
benefits)

Veterans
serviceconnected
compensation

Earned
income
tax credit3

---------------)

...................... dollars -

(-----------------------

Less than high school diploma

455.35
(611.98)

90.35
(258.84)

45.89
(111.76)

1.02
(28.04)

High school graduate,
no college

455.27
(732.86)

47.66
(202.83)

25.48
(92.61)

Some college or more

289.34
(643.91)

19.58
(160.29)

1,233.30
(780.44)

Nonelderly

Total
cash
transfer
income

15.90
(92.88)

22.23
(216.13)

9.02
(160.42)

15.24
(54.00)

672.86
(708.69)

12.22
(132.26)

6.56
(83.20)
25.18
(230.00)

16.68
(325.50)

22.60
(187.05)

8.14
(100.04)

13.09
(47.53)

632.78
(899.31)

8.80
(53.08)

28.51
(250.17)

40.93
(347.98)

4.80
(122.36)

17.91
(178.24)

12.89
(140.12)

9.09
(40.30)

432.53
(863.42)

32.37
(159.48)

5.56
(34.01)

47.97
(314.48)

116.48
(544.01)

1.51
(29.75)

4.58
(75.29)

20.36
(193.29)

0.74
(10.59)

1,419.88
(954.03)

107.23
(375.16)

31.09
(186.23)

17.98
(77.14)

16.15
(184.95)

13.47
(207.59)

10.11
(209.80)

23.13
(202.11)

9.22
(113.32)

13.24
(48.37)

248.08
(641.03)

Disabled

637.41
(831.69)

228.13
(479.31)

54.62
(121.48)

17.44
(174.55)

11.51
(244.42)

39.73
(487.77)

11.34
(118.46)

42.51
(306.90)

7.89
(38.15)

1,058.50
(1,100.43)

Nondisabled

319.95
(650.05)

18.35
(130.50)

12.83
(65.05)

23.07
(221.05)

36.27
(315.34)

6.26
(147.16)

19.83
(186.91)

9.47
(113.21)

10.74
(43.78)

460.97
(833.35)

Single mother

88.75
(329.08)

63.44
(267.67)

88.22
(155.40)

0.61
(32.04)

0.94
(29.57)

46.77
(457.54)

19.76
(166.08)

3.93
(86.11)

49.36
(84.32)

361.78
(732.37)

Other parent

51.62
(278.61)

18.98
(173.82)

16.29
(80.69)

11.89
(153.39)

3.22
(106.85)

7.54
(70.06)

27.08
(229.13)

6.60
(99.37)

20.48
(59.33)

175.15
(535.64)

Nonparent

140.88
(421.72)

32.76
(176.53)

7.99
(44.04)

20.93
(212.64)

21.10
(258.95)

5.83
(199.26)

21.47
(190.94)

11.49
(123.76)

3.55
(24.70)

263.42
(666.16)

Veteran

675.05
(910.42)

16.74
(184.27)

4.30
(36.93)

113.56
(484.52)

111.65
(571.63)

1.94
(34.82)

14.49
(160.84)

55.32
(284.90)

5.16
(31.09)

998.80
(1,237.90)

Nonveteran

264.40
(572.84)

17.91
(75.92)

2.34
(59.28)

17.48
(209.44)

9.77
(206.69)

20.38
(188.10)

1.69
(56.67)

11.66
(45.49)

397.04
(725.66)

Poor

120.97
(269.19)

34.63
(180.14)
74.81
(209.13)

82.97
(152.48)

0.35
(18.04)

25.27
(110.17)

18.22
(128.83)

2.39
(38.54)

19.76
(59.93)

350.06
(418.38)

6.05
(195.44)

19.45
(189.61)

12.75
(141.90)

9.16
(40.19)

522.60
(912.58)

8.34
(187.42)

19.30
(183.42)

11.52
(133.87)

10.55
(43.44)

499.98
(866.00)

Elderly

0.23

Nonpoor

369.25
(698.72)

25.47
(176.06)

6.27
(42.56)

25.75
(232.52)

(8.11)
39.41
(331.60)

All groups

339.63
(667.15)

31.35
(181.03)

15.42
(70.61)

22.72
(218.45)

34.73
(311.46)

aEarned income tax credit (EITC) data are missing for about 12 percent of the sample.
Notes: Standard deviations are in parentheses. For descriptions of the demographic group variables, see box 1 on p. 119. The sample is limited to metropolitan-based households in which the head is at least
18 years old. Total cash transfer income is based on about 88 percent of the households for which EITC data are available. Total cash transfer income includes the value of food stamp benefits. Values to in-kind
assistance, such as Medicare and housing assistance, are not imputed. There is insufficient detail in the Survey of Income and Program Participation to measure the participation in or the benefit value from the
refundable (additional) child credit and student assistance, which appear in table 1.
Source: Authors’ calculations based on data from wave 4 of the 2004 panel of the U.S. Census Bureau, Survey of Income and Program Participation, corresponding to the January-April period of 2005.

12 3

TABLE 5

Transfer income as a share of total income, by demographic group
Average total
monthly transfer income
(cash transfers and
food stamp benefits)

Average total
monthly income
(including cash
transfers and
food stamp benefits)

(......................................... dollars -

Less than high school diploma

High school graduate, no college
Some college or more

Elderly
Nonelderly

Disabled

Nondisabled
Single mother

Other parent

Nonparent
Veteran

Nonveteran

Poor
Nonpoor
All groups

672.86
(708.69)
632.78
(899.31)
432.53
(863.42)
1,419.88
(954.03)
248.08
(641.03)
1,058.50
(1,100.43)
460.97
(833.35)
361.78
(732.37)
175.15
(535.64)
263.42
(666.16)
998.80
(1,237.90)
397.04
(725.66)
350.06
(418.38)
522.60
(912.58)
499.98
(866.00)

)
2,150.75
(1,846.15)
3,070.68
(2,806.11)
5,435.51
(5,641.29)
3,061.06
(3,340.13)
5,023.68
(5,325.44)
2,384.28
(2,587.10)
4,756.69
(5,122.98)
2,785.72
(2,496.91)
6,363.32
(6,258.03)
4,767.55
(5,021.33)
5,118.47
(5,054.67)
4,495.14
(5,019.67)
737.42
(581.62)
5,184.78
(5,146.49)
4,601.77
(5,031.05)

Transfer
income as
a share of
total income
(percent)

31
21

8
46

5
44

10
13
3
6
20

9
47

10
11

Notes: Standard deviations are in parentheses. For descriptions of the demographic group variables, see box 1 on p. 119. The calculations in this
table are based on the results in table 4.
Source: Authors’ calculations based on data from wave 4 of the 2004 panel of the U.S. Census Bureau, Survey of Income and Program Participation,
corresponding to the January-April period of 2005.

expenditures, but if it was paid for directly by the house­
hold, it would be. Expenditure for owner-occupied
housing is set equal to the estimated rental value of
the property—in keeping with the methodology used
by the BLS in the creation of the Consumer Price
Index. Entries in panel A of table 6 are in bold if
expenditure shares in a given category for a group
differ from expenditure shares of the overall popula­
tion by more than 1 percentage point.
We want to highlight differences in spending in
three areas—food, transportation, and health. For food
expenditure (table 6, panel A, first column), those with
less than a high school diploma, the disabled, single
mothers, and the poor all concentrate a higher percent­
age of expenditures on food than the average consumer.
This is in keeping with other research that finds that

124

lower-income households spend a higher portion of
their expenditures on food and other necessities. For
transportation (fifth column), we see lower expenditure
than on average by those with less than a high school
diploma, the elderly, the disabled, and the poor. These
groups are less likely to have commuting expenses. We
find that both the elderly and the disabled spent more
on health than the average consumer (sixth column).
This pattern is consistent with the weakened health status
of these two demographic groups. The nonelderly,
single mothers, and the poor spend less on health than
the average consumer.
In table 6, panel B, we display annual 2009 expen­
diture levels by demographic group. We note that total
expenditures, as shown in the final column, are higher for
those groups that we would expect to have higher income

4Q/2011, Economic Perspectives

Federa l Res erv e Bank of Chica go

TABLE 6

Expenditure shares and annual expenditure levels, by demographic group, 2009

Food

Alcohol

Housing

Apparel

Transportation

Health

Entertainment

Personal
care
products

Personal
care
services

Reading
materials

Education

Tobacco

Miscellaneous

Total

(................................................................................................................................................... percent.........................................................................................................................................................)
A. Expenditure shares

All groups
Less than high school diploma
High school graduate or more
Elderly
Nonelderly
Disabled
Nondisabled
Single mother
Non-single-mother
Poor
Nonpoor

13.8

1.0
0.7
1.0
0.8
1.1
0.9
1.0
0.6
1.0
0.6
1.0

47.0
49.0
46.9
50.7
46.1
47.6
47.0
46.4
47.0
48.8
46.3

5,953.04
4,219.20

418.46
181.01

15,945.73
10,109.88

1,497.62
1,265.06

6,799.08
3,616.19

2,864.70
1,488.12

2,344.95
920.51

296.82
200.47

280.20
134.30

103.36
29.50

997.30
139.03

331.52
392.84

272.49
138.17

38,105.27
22,834.29

6,129.29
4,918.05
6,210.37
4,490.78
6,038.42
5,151.09
6,002.77
4,030.79
6,279.70

442.73
291.25
450.25
251.34
427.79
176.77
433.59
139.37
475.71

16,627.81
12,892.66
16,734.21
10,982.11
16,273.70
13,589.02
16,079.56
9,365.43
17,140.13

1,527.19
987.28
1,623.92
881.97
1,535.58
1,801.10
1,480.77
1,052.20
1,616.72

7,175.54
4,915.88
7,284.50
3,873.81
6,994.20
5,132.33
6,892.67
2,754.12
7,683.72

3,026.18
4,634.74
2,409.67
2,685.55
2,876.99
1,248.73
2,955.63
1,203.07
3,214.68

2,508.43
1,874.34
2,466.00
1,528.30
2,399.50
1,691.69
2,381.88
1,024.09
2,630.88

306.57
263.19
304.78
199.74
302.36
338.73
294.02
197.34
324.46

297.42
283.33
279.41
143.13
289.29
242.95
282.29
112.59
316.15

112.10
142.05
93.40
60.40
106.22
54.94
106.09
37.03
120.65

1,099.01
148.26
1,215.81
175.63
1,052.22
533.27
1,023.76
701.43
1,057.52

324.30
199.95
365.37
597.32
313.81
447.16
325.07
295.78
354.17

288.32
276.57
271.50
334.45
268.87
248.45
273.76
134.76
287.33

39,864.87
31,827.53
39,709.19
26,204.51
38,878.95
30,656.22
38,531.86
21,048.01
41,501.81

14.2
16.9
14.0
12.8
14.6

15.5
14.2
16.2
14.2

17.7

3.6
5.1
3.5
2.6
3.8
3.0
3.6
5.7
3.5
4.6
3.6

16.3
14.5
16.4
12.8
17.1
13.3
16.4
16.1
16.3

12.1
16.9

6.9
6.0
6.9

12.1
5.6
9.2
6.7
3.9
7.0
5.3
7.1

5.6
3.7
5.7
4.9
5.8
5.3
5.6
5.3
5.6
4.5
5.8

0.7
0.8
0.7
0.7
0.7
0.7
0.7
1.1
0.7
0.9
0.7

0.7
0.5
0.7
0.7
0.7
0.5
0.7
0.8
0.7
0.5
0.7

0.2
0.1
0.3
0.4
0.2
0.2
0.2
0.2
0.3
0.2
0.3

2.4
0.6
2.5

0.4
2.8
0.6
2.5
1.7
2.4
3.1
2.3

0.8
1.6
0.7
0.5
0.9

2.1
0.7
1.4
0.8
1.3
0.8

0.7
0.6
0.7
0.7
0.6
1.2
0.6
0.8
0.6
0.6
0.6

100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0

B. Annual expenditure levels

dollars.......................................................................................................................................................... )

(

All groups
Less than high school
diploma
High school graduate
or more
Elderly
Nonelderly
Disabled
Nondisabled
Single mother
Non-single-mother
Poor
Nonpoor

Notes: For descriptions of the demographic group variables, see box 1 on p. 119. Expenditure for owner-occupied housing is set equal to the estimated rental value of the property—in keeping with the methodology used by the U.S.
Bureau of Labor Statistics in the creation of the Consumer Price Index. Each row’s values may not sum to the value in the final column (labeled Total”) because of rounding. In panel A, the expenditure shares of specific groups are
in bold if they differ from the expenditure shares of the overall population by more than one percentage point.
Source: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Expenditure Survey.

12 5

126

5323
40

100

126

263

131

262
158

10

6 342 67 88
21 1 4 4 2 7 5 8

17

9563
62

505
227
159
140
484
104
1,133
256
212
209
823
138

10

786
249
177
178
715
110

282
175
138
117
273
103

-3

-1

103
10
3235
40

131

127
268
99

42333
2
1980-2010
1980 -90
1990-2000
2000-10
1980-2008
200 8-10

B. An nu al av erag e in fla tio n

42333
1

101

127
130
245

Note: All price changes are based on August-to-August inflation rates.
Source: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Price Index Database.

1

4

6
3
3

5

2

2

3
5
2
6
8
3

4

246
94

3
0

101

97
1

105

5220

533

273
58
135
128
266
103
1

200
159
120

157
126
109
217
100

216

514
218
159
148
482
107

230

143
127
126

128
34
103
93
128
100
1

265
59
1

255
149
135
127
246
104
249

151

1980-2010
1980 -90
1990-2000
2000-10
1980-2008
200 8-10

A. Cum ul at ive price ch an ge s

To tal

M isce lla ne ou s
To ba cc o

Edu ca tio n

( ........................................................................................................................................................pe rcen t........................................................................................................................................................... )

Re ad ing
m at er ia ls

serv ic es

pro ducts
Food

A lc ohol

Hou sing

Ap pa re l

Tra ns po rtat io n

He alt h

Ent er ta in m en t

Pe rso na l
ca re

Pe rson al
ca re

Inflation, by expenditure category, 1980-2010

on average. In particular, total expenditures are
higher for those with a high school diploma than
those without one, for the nonelderly than the
elderly, for the nondisabled than the disabled, for
non-single-mothers than single mothers, and for
the nonpoor than the poor. In short, individuals in
our groups of interest spend less on average than
individuals in the remainder of the population.
We measure the inflation of a group as the
weighted average of the price changes of the items
purchased by households in that group, with the
weights depending on the market basket of the group
in question. For example, because the elderly spend
more on health care than the nonelderly, the price
changes in health items get a larger weight in the
calculation of the inflation of the elderly than the
nonelderly. Given the differences in consumption
patterns shown in table 6, panel A (p. 125), we would
expect to find differences in inflation experiences if
price changes across categories differ dramatically.
For example, in a period of rapidly increasing oil
prices, we would expect that the inflation experi­
enced by households that commute less, such as
the elderly and the disabled, would be lower than
that experienced by commuting households, all
else being equal.
Before calculating inflation experiences of dif­
ferent groups, we would like to develop some intu­
ition for the results by displaying price changes of
goods in different categories. In table 7, panels A
and B, we show how prices have changed in the
broad expenditure categories displayed in table 6,
panels A and B. We show price changes for six dif­
ferent periods: 1980-2010, 1980-90, 1990-2000,
2000-10, 1980-2008, and 2008-10. As noted in the
introduction and summary, we divide the period
1980-2010 into 1980-2008 and 2008-10 because
of the unusual patterns of price changes and COLAs
during the recent recession and subsequent recovery.
All price changes are based on August-to-August
inflation rates. Panel A of table 7 shows the total
price change over the periods, while panel B shows
average annual rates during the periods. For example,
the 249 percent for food inflation over the period
1980-2010 in panel A of table 7 means that nominal
food prices in August 2010 were 249 percent of their
August 1980 level. In addition, the average annual
rate of food inflation over the period 1980-2010
was 3 percent, as shown in panel B of table 7.
We see in both panels of table 7 that inflation
rates have differed across the expenditure categories.
For some categories, in particular health, education,
and tobacco, price growth has been above the total

4Q/2011, Economic Perspectives

price growth (as shown in the final column) during all
three decades of the 1980-2010 period. In contrast,
apparel inflation has been lower than the overall price
growth during all three decades. Transportation price
growth has been lower than or equal to total price growth
in all of the periods we consider. Because of these pat­
terns, we would expect groups that concentrate high
portions of consumption on health, education, and
tobacco to have experienced higher inflation than the
average consumer, while groups that concentrate high
portions of spending on apparel and transportation
would have experienced lower inflation.
Now, we combine the expenditure share data and
the price change data to calculate group inflation. We
calculate group inflation in two ways. Our first infla­
tion calculation is based on the annual market basket
consumed by a particular group. The inflation rate for
a group in a particular month is calculated as the yearover-year price change of the market basket consumed
by that group in the prior year. For example, inflation
for the elderly in August 2010 is equal to the price change
between August 2009 and August 2010 of the market
basket purchased by the elderly in 2009.11 Put differently,
inflation is the weighted average price change of the
goods and services purchased by the elderly, with the
weights being the elderly’s expenditure shares (as dis­
played in table 6, panel A, p. 125). We label such cal­
culations “annual-weighted inflation.” This differs from
the way in which the official CPI is calculated because
the official CPI uses weights that are fixed over a period
longer than a year and are derived from expenditures
across multiple years. For example, the CPI from
January 2006 through December 2007 is based on
the 2003 and 2004 market basket. Our second inflation
calculation follows the BLS’s methodology as closely
as we are able (U.S. Bureau of Labor Statistics, 2007).12
For this second measure, we only tabulate inflation from
1987 onward because earlier inflation data would require
the use of older Consumer Expenditure Survey data
(in particular that for 1972-73) than we have used.
We label such calculations “fixed-weighted inflation.”13
In table 8, panel A, we show annual-weighted
inflation calculations, and in panel B, we show fixedweighted inflation calculations. We show cumulative
inflation experiences based on inflation during the month
of August. We choose August because many of the
COLAs are based on year-over-year third-quarter in­
flation. In table 8, panel A (first row and first column
of data), we show that for the overall population, prices
were 255 percent of their August 1980 level in August
2010. This does not mean that a fixed set of goods that
cost $100 in 1980 costs $255 in 2010 because our cal­
culations of inflation are based on a market basket
that is redetermined every year.

Federal Reserve Bank of Chicago

Over the 1980-2010 period, the highest levels of
inflation have been experienced by the elderly, followed
by the disabled, the poor, and those with less than a
high school diploma, as shown in table 8, panel A
(as well as in panel B over the 1987-2010 period).
This pattern is due in part to the findings presented
in panel A of table 6 (p. 125) that the elderly and the
disabled spend more than on average in the health
category, which had quickly growing prices, while those
with less than a high diploma and the poor spend less
than on average in the transportation category, which
had slowly growing prices. This general pattern persists,
more or less, across the different periods displayed in
table 8, panel A. This finding is consistent with other
research that has focused on the elderly as a group
that has faced high inflation (Hobijn and Lagakos,
2005; and Amble and Stewart, 1994).
Based on the calculations using annual weights
in panel A of table 8, we note that over the 30-year
period from 1980 through 2010, inflation faced by the
elderly has been 15 percentage points higher than that
experienced by the overall population. Inflation faced
by the elderly has been higher in each of the three
decades displayed in panel A of table 8 as well. We
generally find smaller gaps between the inflation of
the poor, those with less than a high diploma, and the
disabled and that of the overall population. We also
find that single mothers have experienced slightly lower
inflation than the overall population. The results using
fixed weights, in panel B of table 8, are similar. Note
that the numbers in the first row of panel B of table 8
are smaller than the numbers in the first row of panel A
of table 8 because we are measuring cumulative infla­
tion over a shorter period in panel B.
In the final column of both panels A and B of
table 8, we show cumulative August-to-August infla­
tion according to the official CPI-U. Our measure of
inflation for “all” over the period 1987-2010 in panel
B of table 8 (190 percent in the first row and first
column of data) should be close to the official CPI-U
over the same period (191 percent in the first row and
final column) because for this data point we are using
the same BLS data and methodology. We would expect
our inflation measure for “all” over the period 1980-2010
in panel A of table 8 (255 percent in the first row and
first column) to be lower than the official CPI-U over
the same period (262 percent in the first row and final
column) because we are updating market baskets more
quickly than the CPI-U and taking into account the fact
that households may change their behavior in response
to rising prices by purchasing more of those goods
and services whose prices are increasing less quickly.

127

128

127
190
100
127
190
100

131

126
189
100

131

127
190
100
129
195
100

127
190
100

131
131

131

Notes: For descriptions of the demographic group variables, see box 1 on p. 119. CPI-U means Consumer Price Index for All Urban Consumers. The cumulative inflation experiences are based on inflation during the month
of August. Please see the text for further details on inflation based on annual and fixed weights. The different weighting methodologies do not apply to the calculations for the official CPI-U, which is created by the U.S. Bureau
of Labor Statistics using one weighting methodology.
Sources: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Price Index Database and C o n s u m e r E x p e n d it u re S u rv e y .

191

100

126

126
189
100

115

131

131

195
115
132
129
195
100

197

194
115

w ei gh ts
1987-2010
1987-90
1990-2000
2000-10
1987-2008
2008-10

190
114

190
114

198
115
133
129
198
100

188
114

115
133
129
197
100

190
114

187

247
100

114
130
126
187
100

190
114

129
261
100
125

131

189
114

191

126
263
100

131

262
158

254
156
129
126
254
100
262
158
129

256
156
130
126
256
100
247
155
127

255
156
129
126
255
100
262
158
129
128
260
100

253
156
129
126
253
100
270
160
132
128
269
100
255
156
129
126
255
100
261

157
130
128
262
100
B. Fix ed

All

255
156
129
126
255
100

1980-2010
1980-90
1990-2000
2000-10
1980-2008
2008-10

A. An nu al w eig hts

CPI-U

N on po or
Po or

Non -s in gl em ot he r
Si ng le
m ot he r
Non disa bled
Di sa ble d

Non elde rly
Elderly

High sc ho ol
gr ad ua te
or more
Le ss th an
hi gh sc ho ol
di pl om a

Cumulative inflation experiences, by demographic group

Social Security and SSI
We begin by looking at the Social
Security and SSI COLA and the inflation
experiences of the elderly and the disabled.
Social Security and SSI benefits (for both
the elderly and the disabled) have been in­
dexed to the (seasonally unadjusted) CPI-W
since 1975. Benefits for the Civil Service
Retirement System, military retirement, and
veterans service-connected compensation
are all indexed in the same way.
In table 9, panel A, we show the increase
in the CPI-W in the first column of data. The
number 256 in the first row and first column
of data means that, according to the CPI-W,
prices in August 2010 were 256 percent of their
nominal August 1980 value. In the next two
columns, the numbers displayed for the various
periods are the same as those measuring the
inflation faced by the elderly and the disabled
with annual weights and fixed weights in
table 8, panels A and B, respectively.
From table 9, panel A, we see that for
both annual-weighted and fixed-weighted in­
flation measures, the inflation experienced by
the elderly has been almost always higher
than the CPI-W, both over the entire period
and for each of the three decades covered by
the data. Over the entire 30-year period, based
on annual weights, elderly inflation has been
14 percentage points above the CPI-W. For each
of the three decades presented in the part of
panel A of table 9 using annual weights, the
gap has been between 2 percentage points
and 5 percentage points. Because individuals
tend to benefit from the program for a number
of years (the life expectancy of an American
65 years old in 1980 was 16.4 additional
years),14 these decade-long gaps lead to de­
clines in the purchasing power for the same
individual. For the disabled, the inflation ex­
perienced by the group has also tended to be
higher than aggregate inflation for both the

Offic ial

Our next goal is to compare the inflation
experiences of different groups to increases in
benefit payments. Benefit payment increases
arise either because a program has an explicit
cost-of-living adjustment or because legislators
enact increased benefit amounts. We focus on
four programs—Social Security, SSI, TANF,
and SNAP.

( ................................................................................................................................. ....................................................................... p e r c e n t ................................................................................................................................. ....................................................................... )

Group inflation and program COLAs

4Q/2011, Economic Perspectives

TABLE 9

Consumer Price Index, benefit adjustments, and inflation experiences of select demographic groups
A. Social Security and SSI and the elderly and the disabled
Official
CPI-W

Elderly

Social
Security/
SSI COLA

Disabled

(........................................................ percent........................................................... )
Annual weights
1980-2010
1980-90
1990-2000
2000-10
1980-2008
2008-10

256
155
130
126
257
100

270
160
132
128
269
100

262
158
129
128
260
100

265
152
133
131
250
106

Fixed weights
1987-2010
1987-90
1990-2000
2000-10
1987-2008
2008-10

189
115
130
126
190
100

198
115
133
129
198
100

197
115
133
129
197
100

198
113
133
131
187
106

AFDC/TANF
maximum in
Illinois

Official
CPI-W

B. AFDC/TANF and single mothers
AFDC/TANF
maximum in
Alabama

AFDC/TANF
maximum in
Connecticut

Single
mother

(..............................................................................percent............................................................................... )
Annual weights
1980-2010
1980-90
1990-2000
2000-10
1980-2008
2008-10

182
100
139
131
182
100

143
137
99
106
143
100

150
127
103
114
150
100

256
155
130
126
257
100

247
155
127
125
247
100

Fixed weights
1987-2010
1987-90
1990-2000
2000-10
1987-2008
2008-10

182
100
139
131
182
100

124
119
99
106
124
100

126
107
103
114
126
100

189
115
130
126
190
100

187
114
130
126
187
100

Poor

Less than
high school
diploma

C. SNAP and the disabled, single mothers, poor, and those with less than a high school diploma
Thrifty
Food
Plan

CPI-food

Official
CPI-W

Disabled

Single
mother

SNAP
(food stamp)
monthly
maximum

percent...............................................................................................)
Annual weights
1980-2010
1980-90
1990-2000
2000-10
1980-2008
2008-10

250
149
128
131
260
96

249
151
127
130
245
101

256
155
130
126
257
100

262
158
129
128
260
100

247
155
127
125
247
100

262
158
129
129
261
100

261
157
130
128
262
100

320
158
129
157
259
123

Fixed weights
1987-2010
1987-90
1990-2000
2000-10
1987-2008
2008-10

202
120
128
131
210
96

193
117
127
130
190
101

189
115
130
126
190
100

197
115
133
129
197
100

187
114
130
126
187
100

195
115
132
129
195
100

194
115
131
129
195
100

246
122
129
157
200
123

Notes: For descriptions of the demographic group variables, see box 1 on p. 119. CPI-W means Consumer Price Index for Urban Wage Earners and Clerical
Workers; CPI-food means the Consumer Price Index for all food. COLA means cost-of-living adjustment. SSI means Supplemental Security Income. AFDC means
Aid to Families with Dependent Children, and TANF means Temporary Assistance for Needy Families. SNAP means Supplemental Nutrition Assistance Program
(Food Stamp Program). The Thrifty Food Plan is the basis for food stamp allotments. Please see the text for further details on inflation based on annual and fixed
weights. The different weighting methodologies only apply to the inflation calculations for the demographic groups.
Sources: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Price Index Database and Consumer Expenditure Survey,
and data on benefit determination from the U.S. Social Security Administration (panel A), U.S. Department of Health and Human Services (panel B), and U.S.
Department of Agriculture (panel C).

Federal Reserve Bank of Chicago

129

annual-weighted and fixed-weighted measures, although
to a lesser degree (for example, by 6 percentage points
over the entire period for the annual-weighted measure).
Like the elderly on Social Security Old-Age and Sur­
vivors Insurance, disabled individuals tend to benefit
from the Social Security Disability Insurance and SSI
programs for long durations—with the average stay on
disability lasting over ten years (Rupp and Scott, 1995).
The comparison of inflation experiences of the
elderly and the disabled with the actual COLAs im­
plemented by the Social Security and SSI programs
is more complicated. Over the period 1976-83, the
Social Security/SSI COLAs were based on increases
in the CPI-W from the first quarter of the prior year
to the first quarter of the current year and became ef­
fective with June benefits paid to recipients in July.
After 1983, the COLAs were based on increases in
the CPI-W from the third quarter of the prior year to
the third quarter of the current year and became effec­
tive with December benefits paid in January. Figure 1
shows August-over-August growth in the CPI-W,
Social Security/SSI COLA, and inflation faced by
the elderly. There are three notable features of the
Social Security/SSI COLA relative to the CPI-W.
First, CPI-W increases are reflected in the COLA with
a lag because the COLA has been implemented one
quarter after the price change is measured. Second,
there was no COLA in 1983; in other words, benefits
in August 1983 were the same as benefits in August
1982. This is due to the shift, beginning in 1983, from
implementing COLAs in July to implementing COLAs
in the following January (that is, the 1983 COLA was
implemented in January 1984). Third, recent Social
Security/SSI COLAs have been somewhat unusual.
The COLA for 2008—first paid in January 2009—was
5.8 percent. Prices in 2008:Q3 were 5.8 percent above
their 2007:Q3 level. The magnitude of this increase was
in part due to the timing of the COLA determination.
Energy prices spiked over the summer of2008. For all
of2008, CPI-W inflation was 4.1 percent, but the COLA
was based on the 2008:Q3 measurements. The COLA
for 2009—first paid in January 2010—was zero because
prices fell between 2008:Q3 and 2009:Q3 and the COLA
cannot be negative. This fall was due in part to the tem­
porary nature of the energy price spike. The COLA for
2010, payable in January 2011 (not shown in figure 1),
was also zero because, although prices increased
modestly (1.5 percent) between 2009:Q3 and 2010:Q3,
they were still about half a percentage point below their
2008:Q3 level. In effect, recipients were compensated
beginning in January 2009 for the inflation experienced
between 2009:Q3 and 2010:Q3. Because of all these

130

factors, the CPI-W and the Social Security/SSI COLA
have differed modestly over this period.
We divide our comparison of the Social Security/
SSI COLA with elderly inflation into the periods
1980-2008 and 2008-10 because the forces at work
in these two eras differ. In the 1980-2008 period,
elderly inflation was above the Social Security/SSI
COLA by 19 percentage points (see table 9, panel A,
fifth row, p. 129). This is in part due to the following
factors: the gap between elderly inflation and overall
inflation, the lack of a COLA in 1983, and the fact that
the price increases in 2008 had not yet been incorpo­
rated into the COLA. In 2008-10, the Social Security/
SSI COLA was higher than the inflation faced by the
elderly. This is due to the large COLA in January 2009
and the fact that the January 2010 COLA could not be
negative. The pattern for the disabled is similar, although
the gap in the 1980-2008 period is smaller.
Overall, the inflation experienced by the elderly
and the disabled has generally been higher than the
price index used to adjust their most substantial income
support benefits. However, the path for the actual
Social Security/SSI COLA has differed from that for
the index it tracks because of idiosyncrasies in the
determination of the Social Security/SSI COLA.
Temporary Assistance for Needy Families
Benefit payments for the Temporary Assistance
for Needy Families program (which replaced the Aid
to Families with Dependent Children in 1997) are set
by the states. States set maximum benefit payments
for families of different compositions, and subtract
some portion of family income to set the actual bene­
fit payment for a given family. In table 9, panel B
(p. 129), we compare nominal increases in maximum
monthly AFDC/TANF benefits for a family of three
in three states—Alabama, Connecticut, and Illinois—
with the inflation experiences of single mothers, based
on both annual and fixed weights. We choose these
three states because one was a relatively high-benefit
state in 1980 (Connecticut’s maximum benefit was $475),
one was a moderate-benefit state (Illinois’s maximum
was $288), and one was a low-benefit state (Alabama’s
maximum was $118). While states determine benefit
levels, TANF payments are partially funded by federal
block grants that have been fixed in nominal terms
since they were established in 1996.
If we compare the increases in AFDC/TANF
benefits with the inflation experiences of single
mothers based on annual weights, we find that while
single mothers were facing prices in 2010 that were
247 percent of their 1980 level, benefits in these three
states were between 143 percent and 182 percent of

4Q/2011, Economic Perspectives

FIGURE 1

Social Security/SSI COLA, elderly inflation, and CPI-W

Social Security/SSI COLA

A

Elderly inflation

B

CPI-W

Notes: SSI means Supplemental Security Income. COLA means cost-of-living adjustment. CPI-W means Consumer Price Index for Urban
Wage Earners and Clerical Workers. August-over-August growth is displayed for all data.
Sources: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Price Index Database and Consumer
Expenditure Survey; and data on benefit determination from the U.S. Social Security Administration.

their 1980 level (see table 9, panel B, first row, p. 129).
Growth in the price of the market basket of single
mothers was 65 percentage points above the growth in
benefits in the state with the largest percentage growth
in benefits among the three selected—Alabama. We also
see large gaps, particularly for Connecticut and Illinois,
when we investigate the 1987-2010 period and use fixed
weights. These gaps between benefit growth and price
growth are far larger than that between the Social
Security/SSI COLA and elderly inflation, and represent
substantial erosion in the purchasing power of program
beneficiaries. These three states are fairly representative
of the 50 states. In no state did the value of benefits
keep up with the annual-weighted price increases faced
by single mothers over the 1980-2010 period. This
erosion in the real value of welfare benefits has been
noted elsewhere (for example, Schott and Levinson,
2008). For 1990-2000 and 2000-10, growth in benefit
payments in Alabama (the low-benefit state) slightly ex­
ceeded the price growth faced by single mothers (see
table 9, panel B, third and fourth rows). However, the
maximum benefit in Alabama had been unchanged be­
tween 1980 and 1990 (see table 9, panel B, second row).
In figure 2, we show August-over-August increases
in AFDC/TANF benefits in the three states, overall

Federal Reserve Bank of Chicago

inflation (as measured by the CPI-W), and single-mother
inflation. In most years, benefits have been unchanged,
but there have been occasional changes. For the
AFDC/TANF population, the duration of benefit re­
ceipt differs before and after the implementation of
the TANF program in 1997, since federal funding for
TANF recipients is limited to 60 months. Prior to wel­
fare reform in 1996, over 50 percent of the caseload
was expected to stay on the program for over a decade
(Rupp and Scott, 1995). The real erosion in welfare
benefits translates into both lower real benefits for
individuals who enter AFDC/TANF at later dates and
a decline in the purchasing power of benefits for an
individual during her stay on AFDC/TANF.
Supplemental Nutrition Assistance Program
In panel C of table 9 (p. 129), we compare in­
creases in monthly maximum benefits from the
Supplemental Nutrition Assistance Program (formerly
called the Food Stamp Program) with price increases
faced by those with less than a high school diploma,
the disabled, single mothers, and the poor—all based
on both annual and fixed weights. Individuals in all
of these groups are heavily represented among SNAP
recipients (see the Food Stamp Program column in

131

FIGURE 2

AFDC/TANF benefit growth in select states, single-mother inflation, and CPI-W
percent

35 r

-10 __ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i__ i
1980

’82
♦

’84

’86

’88

’90

’92

’94

A

Alabama AFDC/TANF benefit growth

—■- Connecticut AFDC/TANF benefit growth

’96

’98

2000

’02

’04

Illinois AFDC/TANF benefit growth

’06
#

’08

’10

CPI-W

—H— Single-mother inflation

Notes: AFDC means Aid to Families with Dependent Children; TANF means Temporary Assistance for Needy Families, and it replaced the
AFDC in 1997. CPI-W means Consumer Price Index for Urban Wage Earners and Clerical Workers. August-over-August growth is displayed
for all data.
Sources: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Price Index Database and Consumer
Expenditure Survey; and data on benefit determination from the U.S. Department of Health and Human Services.

table 2, pp. 120-121). As noted in table 1 (pp. 115-117),
SNAP maximum benefits are currently indexed to in­
creases in the cost the U.S. Department of Agriculture’s
Thrifty Food Plan (TFP). In particular, benefits are
based on the cost of a low-cost nutritious diet for a
family of four with one child aged 6-8 and one child
aged 9-11. June-to-June increases in prices are reflect­
ed in October SNAP benefits.
In table 9, panel C (p. 129), we display increases in
the cost of the TFP in the first column of data. The fourth
through seventh columns of data display price increases
for our groups of interest. The increases in the cost of
the TFP are slightly above the annual-weighted price
increases faced by single mothers over the 1980-2010
period, while the increases in the cost of the TFP are
slightly below the inflation experienced by the other
groups in panel C. For the 1980-90 period, when
food price growth overall (as measured by the Consumer
Price Index for food, or CPI-food, and displayed in the
second column) was below total inflation (displayed
in the third column), all groups experienced inflation
that was higher than the growth in the cost of the TFP.
This is by design, in that growth in benefits is meant

132

to match increases in the price of food rather than
increases in the cost of all items.
We graph annual August-over-August increases
in SNAP benefits, the cost of the TFP, the cost of food
(as measured by the CPI-food), and the price of the
market basket consumed by the poor in figure 3. Two
notable patterns emerge from the figure. First, TFP
inflation is more volatile than overall food inflation
(as measured by the CPI-food). This is due to greater
weighting in the TFP toward vegetables, milk products,
and fruit, whose prices tend to be less stable than those
of other foods and food away from home (McGranahan,
2008; and Carlson et al., 2007). Second, SNAP benefit
growth and TFP cost growth have differed quite sub­
stantially at times. In fact, over the period 1980-2010,
these measures have been negatively correlated. This
differential is due to the four-month lag in implementing
benefit changes and to frequent policy changes in the
methods of determining maximum benefits. The cur­
rent method of indexing benefits was first put in place
in October 1996, based on June 1996 prices.15 Prior to
that, inflation adjustments had been a done in a variety
of different ways. Since the Food Stamp Act of 1964,

4Q/2011, Economic Perspectives

FIGURE 3

SNAP benefit growth, Thrifty Food Plan cost growth, inflation of the poor, and food inflation

♦

SNAP (food stamp) benefit growth

B

Thrifty Food Plan cost growth

A
X

Inflation of the poor
CPI-food

Notes: Supplemental Nutrition Assistance Program (SNAP) was previously known as the federal Food Stamp Program (and it is still commonly
referred to by this name). The Thrifty Food Plan is the basis for food stamp allotments. CPI-food means Consumer Price Index for all food.
August-over-August growth is displayed for all data.
Sources: Authors’ calculations based on data from the U.S. Bureau of Labor Statistics, Consumer Price Index Database and Consumer
Expenditure Survey; and data on benefit determination from the U.S. Department of Agriculture.

food stamps have been indexed annually, indexed semi­
annually, and frozen. Benefits have been set from
99 percent to 103 percent of the cost of the Thrifty Food
Plan. For example, from October 1992 until October
1996, food stamp benefits were set at 103 percent of
the cost of the TFP. As another example, food stamp
benefits were fixed between January 1981 and September
1982, and the benefit adjustment for October 1, 1982,
was based on 21 months of price changes.
By contrast, in 2010, the food stamp maximum
was 123 percent of its 2008 level, although the cost
of the TFP was 96 percent of its 2008 level in 2010
(table 9, panel C, sixth row, p. 129). This disparity
is due to a provision of the American Recovery and
Reinvestment Act of 2009 that set benefits beginning
in April 2009 at 113.6 percent of the cost of the TFP
as of June 2008. Because of the ARRA increase, SNAP
benefit growth exceeded the inflation faced by all pop­
ulation groups over the entire 1980-2010 period—in
particular, the 2008-10 period.
SNAP COLAs will also be unusual going forward.
Under current legislation, SNAP benefits are set to re­
main at 113.6 percent of the June 2008 TFP cost until

Federal Reserve Bank of Chicago

October 2013 unless food inflation is so high that
100 percent of the contemporaneous TFP cost exceeds
113.6 percent of the June 2008 TFP cost prior to that
date. In other words, from now until October 2013,
unless inflation is very high, there will be no benefit
increases. In October 2013, benefits will revert to
being set at 100 percent of the June 2012 TFP cost.
Assuming total TFP inflation between June 2008
and October 2013 is less than 13.6 percent (about
2.6 percent per year), benefits will fall in October 2013.
This schedule for future benefit adjustments has been
changed twice since the passage of the ARRA. Origi­
nally, the provision was going to end whenever the
TFP cost exceeded 113.6 percent of the June 2008 level.
It was then set to end in March 2014 and is now set to
end in October 2013. These changes in the timing of
the added benefits’ phaseout are akin to what has been
seen in other periods where food stamp benefit adjust­
ments were subject to frequent policy shifts. In general,
the relationship of the cost of the TFP and maximum
SNAP allotments has been influenced by policy deci­
sions. This relationship was altered through the ARRA
and through two additional pieces of legislation since

133

the ARRA’s passage, as well as numerous times prior
to 2009.

Reviewing the four transfer programs
The relationship between the experience of pro­
gram recipients and the computation of benefit levels
has been quite different for Social Security and SSI,
TANF, and SNAP With the exception of a change in
the timing of COLA determination between 1982 and
1983, the Social Security/SSI COLA has been calcu­
lated in a consistent manner. As a result, for the most
part, the Social Security/SSI COLA has been close to
the inflation measure, the CPI-W, it is intended to track.
However, the inflation experiences of both the elderly
and the disabled have been generally higher than the
inflation of the population covered by the CPI-W.
The gap between the Social Security/SSI COLA
and the inflation of the elderly and the disabled pales
in comparison with the gap between the growth of
TANF benefits and the inflation faced by single mothers.
TANF beneficiaries have seen substantial declines in
the purchasing power of their benefits. Although the
inflation faced by single mothers has been moderately
below inflation for the CPI-W population, the growth
in TANF benefits has been far below the inflation
faced by single mothers. This gap is so large because
neither the block grants from the federal government
to the states nor the state benefits themselves are in­
dexed. As a result, the growth in nominal TANF ben­
efits at the state level has been modest and uneven.
For SNAP benefit recipients, inflation over the
entire period has been close to growth in the cost of
the TFP in part because food inflation has been close
to overall inflation. However, SNAP benefit growth
and TFP cost growth have diverged because of policy
decisions. SNAP benefit levels and the relationship
between these benefit levels and the cost of the TFP
have been policy levers that are frequently used. Be­
cause of a major increase in benefits enacted as part
of the ARRA, benefit increases far exceed the infla­
tion of the groups that depend on SNAP. However,
the history of the ARRA benefits is also indicative of
the frequency with which SNAP benefits are altered—
the timing of the phaseout of the added ARRA bene­
fits has been changed twice since the ARRA passed.

Conclusion
We compare the inflation indexation used in gov­
ernment transfer programs with the inflation experi­
ences of households that are dependent on those programs
for income support.
We find that the inflation experienced by differ­
ent demographic groups differs from aggregate infla­
tion because of differences in consumption patterns

134

across the demographic groups and differences in price
changes across expenditure categories. Demographic
groups that concentrate a higher portion of their spending
in categories whose prices have grown rapidly over the
past three decades, such as health care, have experienced
higher inflation than demographic groups that concen­
trate a higher portion of their spending in categories
whose prices have grown more slowly, such as trans­
portation. Because of their high demand for health care
and low commuting costs, elderly households have
experienced the highest inflation of all the groups
we investigate.
We also find that the evolution of transfer program
benefits has differed substantially across the four pro­
grams we investigate. Social Security and SSI benefit
growth has been moderately lower than the inflation
experiences of the elderly and the disabled because of
the consumption patterns of the elderly and the disabled.
However, TANF benefit growth has been far below the
inflation experienced by single mothers because of
the absence a routine COLA for most state-level ben­
efits; and SNAP benefit growth has diverged from the
inflation experienced by its beneficiaries because of
frequent changes in the way the SNAP COLA has
been calculated.
Much of the policy debate concerning COLAs has
revolved around the Social Security program. This is
in part due to the high inflation experienced by the el­
derly and in part due to the fact that Social Security is
the single largest income support program, represent­
ing 31 percent of all federal expenditures on payments
to individuals in 2010.16Because the elderly have ex­
perienced higher inflation than the overall population,
their inflation experiences have exceeded inflation as
measured by the CPI-W, the index upon which increases
in Social Security benefits are based. Given the gap
between inflation experienced by the elderly and the
Social Security COLA, the elderly have experienced
a decline in their ability to purchase their preferred
market basket, even in the presence of a fully indexed
benefit. Using an alternative COLA that indexed Social
Security benefits to the inflation faced by the elderly
could eliminate this gap. However, such a policy change
would be extremely costly. Researchers at the Federal
Reserve Bank of New York (Hobijn and Lagakos, 2003)
estimated the potential costs of using a CPI based on
the consumption patterns of the elderly to adjust Social
Security benefits. According to this research, had an
elderly-specific index been adopted in 1984, benefits
would have been 3.84 percent higher than they actually
were in 2001. The New York Fed researchers anticipated
that changing to an elderly-specific index in 2003 would
likely have increased future benefit levels and have

4Q/2011, Economic Perspectives

rendered the Social Security trust fund insolvent five
years sooner than projected at the time.
Alternatively, changing the Social Security COLA
to one that led to more modest increases in benefits,
as proposed by the National Commission on Fiscal
Responsibility and Reform, would magnify the gap
between the inflation of the elderly and the Social
Security COLA.17 At the same time, such a change
would relieve some budget pressures.
Past attempts to change the Social Security COLA
to one that reflected the purchasing habits of the elderly
have not gotten much traction. Legislation to change
the Social Security COLA has been introduced in

every Congress since the 105th in 1997-98, but this
legislation has never made it to the floor of either
chamber. Legislation has also been introduced in the
current Congress. The National Commission on Fiscal
Responsibility and Reform’s proposal to use a chainweighted Consumer Price Index (in particular, the
chain-weighted Consumer Price Index for All Urban
Consumers, or C-CPI-U) has not yet been incorporat­
ed into any legislative proposal, and such a change
was not included in President Obama’s 2012 budget
proposals. However, this change has been incorporated
into some of the broad-based proposals to address the
federal deficit.

NOTES
Social Security is officially referred to as the federal Old-Age,
Survivors, and Disability Insurance Program, or OASDI.

9In McGranahan and Paulson (2006), we detail how the data are
merged to create these expenditure patterns.

2Chain-weighting is a method of measuring inflation that takes into
account the fact that people tend to buy less of things whose prices
have increased a lot and instead buy more of substitutes whose prices
have risen less. Different inflation measures are discussed in more
detail later in this article.

10For the official BLS definition of “consumer unit,” see www.bls.
gov/cex/csxfaqs.htm#q3.

3Social Security and SSI are administered by the U.S. Social Security
Administration; TANF is administered by the U.S. Department of
Health and Human Services; and SNAP is administered by the U.S.
Department of Agriculture, Food and Nutrition Service.

12We cannot perfectly mirror the BLS’s calculation because we
lack some of the information needed to do so. Area information is
missing in the public use data. In addition, some prices are not pub­
licly released.

4These are our calculations based on data from White House, Office
of Management and Budget (2011a, b).

13These weights are not fixed over the entire sample, but they are
fixed for longer than a year.

5These adjustments are automatic except for the case of veterans’
benefits. Congress enacts legislation every year that sets the COLA
for veterans’ benefits equal to that for Social Security benefits. This
legislation tends to pass unanimously.

14See www.cdc.gov/nchs/data/hus/hus2009tables/Table024.pdf.

6Market baskets refer to evolving selections of goods and services
purchased by individuals that are used to track inflation in an econ­
omy or specific market.
7For alternative discussions on the use of indexation in the federal
government, see Congressional Budget Office (1981, 2010).
8Our criteria do not lead us to look at the expenditure patterns of
veterans. However, even if we had wanted to do so, we would not
have been able to. The only question related to veteran status in the
Consumer Expenditure Survey is one that asks the amount of income
from workers’ compensation and veterans’ benefits combined. Workers’
compensation is a larger program paying out $55 billion in annual
benefits in 2007 (U.S. Census Bureau, 2010) versus $32 billion in
veterans service-connected compensation (see table 1 on pp. 115-117
of this article for more on the latter). We do not include workers’
compensation in the list of programs discussed in this article because
it is largely funded by private insurance carriers and employers’
self-insurance and not by the federal government.

Federal Reserve Bank of Chicago

1’Doing the calculation in this way ensures that the inflation measure
is not influenced by seasonal patterns in expenditures or prices.

15The current method of indexing SNAP (food stamp) benefits was
enacted in the Personal Responsibility and Work Opportunity Recon­
ciliation Act of 1996. See www.fns.usda.gov/snap/rules/Legislation/
timeline.pdf for details on the Food Stamp Program’s evolution.
16This is our calculation based on data from White House, Office
of Management and Budget (2011b).

17It would be worthwhile to compare the Social Security COLA with
increases in elderly inflation by using the type of chain-weighted
measures that the National Commission on Fiscal Responsibility
and Reform has proposed. While chain-weighted inflation experi­
enced by the elderly will most likely be lower than the measures of
elderly inflation we present, the rapid increases in health care costs
will still lead to chain-weighted inflation experienced by the elderly
being greater than chain-weighted aggregate inflation.

135

REFERENCES

Amble, Nathan, and Kenneth Stewart, 1994,
“Experimental price index for elderly consumers,”
Monthly Labor Ttewew, Vol. 117, No. 5,May,pp. 11-16.
Carlson, Andrea, Mark Lino, WenYen Juan,
Kenneth Hanson, and P. Peter Basiotis, 2007, Thrifty’
Food Plan, 2006, U.S. Department of Agriculture,
Center for Nutrition Policy and Promotion, report,
No. CNPP-19, April.

Congressional Budget Office, 2010, “Using a differ­
ent measure of inflation for indexing federal programs
and the tax code,” Economic and Budget Issue Brief,
Washington, DC, February 24.

__________ , 1981, Indexing with the Consumer Price
Index: Problems and Alternatives, report, Washington,
DC, June.
Hanson, Kenneth, and Margaret Andrews, 2008,
“Rising food prices take a bite out of food stamp ben­
efits,” Economic Information Bulletin, U.S. Department
of Agriculture, Economic Research Service, No. 41,
December.

Haveman, Robert, Karen Holden, Kathryn Wilson,
and Barbara Wolfe, 2003, “Social security, age of
retirement, and economic well-being: Intertemporal
and demographic patterns among retired-worker
beneficiaries,” Demography, Vol. 40, No. 2, May,
pp. 369-394.

Hobijn, Bart, and David Lagakos, 2005, “Inflation
inequality in the United States,” Review ofIncome
and Wealth, Vol. 51, No. 4, December, pp. 581-606.
__________ , 2003, “Social Security and the Consumer
Price Index for the elderly,” Current Issues in Economics
and Finance, Federal Reserve Bank of New York,
Vol. 9, No. 5, May.

Internal Revenue Service, Statistics of Income,
2010, Individual Income Tax Returns 2008, IRS
Publication, No. 1304 (Rev. 07-2010), Washington,
DC: U.S. Department of the Treasury, available at
www.irs.gov/pub/irs-soi/08inalcr.pdf.
McGranahan, Leslie, 2008, “Food inflation and the
consumption patterns of U.S. households,” Chicago
Fed Letter, Federal Reserve Bank of Chicago, No. 255,
October.
McGranahan, Leslie, and Anna Paulson, 2006,
“Constructing the Chicago Fed Income Based Economic

136

Index—Consumer Price Index: Inflation experiences
by demographic group: 1983-2005,” Federal Reserve
Bank of Chicago, working paper, No. WP 2005-20,
revised November 2006.

Meyer, Bruce D., and Dan T. Rosenbaum, 2001,
“Welfare, the earned income tax credit, and the labor
supply of single mothers,” Quarterly Journal of
Economics,Nol. 116, No. 3, August, pp. 1063-1114.
National Commission on Fiscal Responsibility
and Reform, 2010, “The moment of truth,”
report, Washington, DC, December, available at
www.fiscalcommission.gov/sites/fiscalcommission.
gov/files/documents/TheMomentofTmthl2_l_2010.pdf.
Rupp, Kalman, and Charles G. Scott, 1995, “Length
of stay on the Supplemental Security Income disability
program,” Social Security’ Bulletin, Vol. 58, No. 1,
Spring, available at www.ssa.gov/policy/docs/ssb/
v58nl/v58nlp29.pdf.

Schott, Liz, and Zachary Levinson, 2008, “TANF
benefits are low and have not kept pace with inflation,”
Center on Budget and Policy Priorities, report, November
24, available at www.cbpp.org/pdl711-24-08tanf.pdf.

U.S. Bureau of Labor Statistics, 2007, “Chapter 17.
The Consumer Price Index (Updated 06/2007),” in
BLS Handbook ofMethods, Washington, DC, available
at www.bls.gov/opub/hom/pdf/homch 17.pdf.
U.S. Census Bureau, 2010, Statistical Abstract of
the United States: 2011, Washington, DC, available
at www.census.gov/compendia/statab/2011/
2011edition.html.
U.S. Department of Housing and Urban
Development, 2011, FY2010 Annual Performance
Report, Washington, DC, March, available at
http://portal.hud.gov/hudportal/documents/
huddoc?id=fy2010apr.pdf.
White House, Office of Management and Budget,
2011a, Analytical Perspectives, Budget of the United
States Government, Fiscal Year 2012, Washington, DC:
U.S. Government Printing Office, available at www.
whitehouse.gov/omb/budget/Analytical_Perspectives.

___________ , 2011b, Historical Tables, Budget of the
United States Government, Fiscal Year 2012, Wash­
ington, DC: U.S. Government Printing Office, avail­
able at www.whitehouse.gov/omb/budget/Historicals.

4Q/2011, Economic Perspectives

Clearing over-the-counter derivatives
Ed Nosal

Introduction and summary
The recent financial crisis highlighted some of the
potential problems associated with over-the-counter
(OTC) derivatives markets. Prior to the financial crisis
of 2008, the over-the-counter market was not required
to “clear” transactions. This changed with the signing
of the new financial reform legislation, the Dodd-Frank
Act, on July 21,2010. Going forward, most OTC deriv­
atives will be cleared through a particular set of insti­
tutional arrangements: a regulated clearinghouse.
The financial crisis exposed some significant cracks
in the OTC derivatives markets,1 as exemplified by the
case of insurance company AIG. We now know that
AIG sold a gigantic amount of OTC credit default swaps
(insurance contracts against defaults), with notional
value over $440 billion. Since AIG did not take posi­
tions to offset their credit default swap exposure, one
might conjecture that this could cause problems for
AIG and financial markets.2 Namely, if the economy
were hit by an adverse aggregate shock that either
caused a very large number of swap payouts to move
against AIG or weakened its balance sheet (or both),
then AIG may not have the funds to perform on its OTC
obligations. And even if it did have the funds, AIG may
choose to default (referred to as a strategic default)
on its credit default swap obligations. In either case,
AIG’s counterparties would not receive the payoffs
they were expecting, and this could have a ripple effect
on their counterparties and the broader markets.
In fact, during the financial crisis, AIG had to raise
money in debt markets in order to make payments on
the real-estate-related credit default swaps that moved
against it. This new debt issue, along with the large
potential future losses that AIG could experience, re­
sulted in AIG’s debt being downgraded by the credit­
rating agencies. As a result of this downgrade, AIG was
required to post billions of dollars of collateral for its

Federal Reserve Bank of Chicago

existing swap contracts. If the collateral had not been
posted, then cross-default clauses in other contracts that
AIG had written would have been activated, requiring
AIG to settle these contracts immediately. AIG did not
have the liquidity on hand to make all these payments,
even though its core business activities were in fine
shape. Without some help, AIG would have defaulted
on its contractual obligations. Hence, the deterioration
of the real estate market put significant financial stress
on AIG, putting not only AIG at risk, but also AIG’s
counterparties and their counterparties, and so on. In
the end, collateral was posted for AIG’s positions thanks
to a massive government bailout.3
How could this scenario have been avoided? One
could argue that clearing AIG’s OTC derivatives con­
tracts would have prevented this negative outcome.
Intuitively, one can think of clearing as a set of insti­
tutional arrangements that are designed to enhance con­
tractual performance. This includes a wide range of
procedures that are implemented after a buyer and seller
agree to the contract terms and before final settlement
occurs. For example, if AIG had been required to clear
its credit default swaps, it would have had to set aside
collateral when it initially negotiated the contracts.
When the swap contracts turned against AIG, it could
have used the collateral to satisfy its positions, instead
of borrowing, which, in turn, would have prevented
the credit downgrade in its debt.

Ed Nosal is a vice president and senior research advisor in
the Financial Markets Group of the Economic Research
Department at the Federal Reserve Bank of Chicago. The
author would like to thank Robert Steigerwaldfor many
hours ofconversation on clearing andfor discussing earlier
drafts of this article. He would also like to thank Lisa Barrow,
JeffCampbell, Adam Cooper, Randall Costa, Godfried De Vidts,
Darrell Duffie, Marcus Katz, Helen Koshy, Adi Li, David
Marshall, Bob McDonald, and Ketan Patelfor discussions
and useful comments on this article.

137

Clearing would also have led to greater transparency.
Even absent posting collateral, if information regarding
AIG’s credit default swap portfolio was publicly avail­
able, it is likely that AIG’s credit default swap portfolio
would have been much smaller as counterparties would
have been reluctant to enter into arrangements with such
a highly leveraged entity. These sorts of remedies have,
in fact, been mandated in the recent Dodd-Frank Act,
which requires the vast majority of OTC derivatives
to be cleared through a regulated clearinghouse.
Since clearing is a not a costless activity, two
questions naturally arise: 1) Who should be allowed
to participate in a clearinghouse as a clearing member?;
and 2) Which derivative contracts should be cleared?
A recent public policy symposium held at the Federal
Reserve Bank of Chicago addressed these questions.4
In this article, I review the topic of clearing from an
economics perspective and provide the reader with a
framework to think about clearing issues. The analysis
can offer answers to basic but important questions,
such as: What is clearing? Why is it important? What
is the role of a clearing member? And why is risk
management important for clearing? Second, I use
this framework to review some of the ideas that were
expressed at the symposium regarding who should be
allowed to participate in the clearing process and what
should be cleared.
I conclude that: 1) although certain criteria must
be met for individuals or institutions to participate in the
clearing process, one criterion that should not be im­
posed is that participants must possess massive amounts
of capital; and 2) derivative contracts negotiated by
“end users,” corporations that hold the contracts until
they expire, should not be treated any differently from
contracts that are negotiated by other firms that do
not necessarily hold them until expiration.
In the next section, I provide a simple economic
environment to think about the clearing concept. The
economic environment presented in this section has the
benefit of being transparent, but at the cost of being
quite simple. In the two following sections, I compli­
cate the environment slightly, in order to discuss the
behavior of OTC participants and risk management—
important concepts for clearing. I describe how a
clearinghouse might be structured. Then, I present and
analyze the overarching themes of the symposium.
Finally, I analyze the debate about clearinghouse mem­
bership and end-user exemptions.

A simple model of clearing
A natural starting point for any economic investi­
gation is the Arrow-Debreu model. In the model, people
are fully aware of all possible future contingencies

138

and are able to write contracts, at the beginning of time,
for delivery and acceptance of all possible commodities,
where a commodity is distinguished by date, location,
and state of the world. Contracts are mediated by mar­
kets, and there is a market and price for each com­
modity, that is, markets are “complete.” The model
determines one of the most important concepts in
economics, which is that of relative prices.
Since people are fully aware of all future contin­
gencies and markets are complete, all decisions regard­
ing how much to buy and sell can be made at the
beginning of time. Hence, all trading of contracts occurs
at one point in time—at the beginning of time—and as
time moves on, people simply make or accept delivery
of commodities based on their contracts. Spot markets
are not needed at future dates since, as the economy
moves in time, people do not learn anything that they
did not already know at the beginning of time. If spot
markets opened up at a future date and trade occurred,
that would imply that people made unexplainable mis­
takes at the beginning of time, that is, they are irrational.
So, in the Arrow-Debreu world, all deals are struck at
the beginning of time, and people trust one other to
perform as specified in the contingent contracts.
The elegance and simplicity of the Arrow-Debreu
model owe a lot to the absence of frictions in the en­
vironment. For example, people do not have to find
one another or bargain over prices and quantities or
worry about contractual performance. All of these fric­
tions are assumed away. In fact, the lack of frictions in
the Arrow-Debreu model greatly simplifies the nature
of social interaction: There is none! People observe
equilibrium prices and trade only against their own
budget constraint, which means that the value of what
they buy cannot exceed the value of what they sell. The
model, however, cannot explain a number of impor­
tant things, such as how goods are exchanged or why
the institutions of money, banking, and clearing exist.
If we want to understand these important institu­
tions, we must introduce some explicit frictions into
the benchmark model. For example, if we want to
have spot markets open up over time, we can intro­
duce a search friction. If it takes time and effort to
find and purchase goods and services, then it will not
be possible to contract for all purchases and sales of
commodities at the beginning of time. If there are in­
formational frictions, then contracts can only depend
upon things that are verifiable, and complete statecontingent contracts are not feasible. If there is a com­
mitment friction, then things like money, banks, and
other institutions may arise to help alleviate the com­
mitment problem. And finally, we might also want to
take into account legal frictions. These frictions may

4Q/2011, Economic Perspectives

explain some risk-management practices that would be
a puzzle if a lack of commitment was the only friction.
Although I use all of the above-mentioned frictions in
my analysis, the commitment friction—people cannot
commit to undertake future actions—and the infor­
mation friction play prominent roles. Now, I illustrate
the importance of two prominent clearing processes—
novation and the posting of collateral.
A farmer plants seeds today that produce wheat
tomorrow, and a baker needs wheat tomorrow to bake
bread. The price per bushel of wheat tomorrow can
take one of two values, say, $5 or $15, that are equally
likely. The farmer and baker are risk averse, meaning
that they prefer to agree today to exchange one bushel
of wheat tomorrow for tomorrow’s expected price of
$10, as opposed to buying or selling at the spot price
tomorrow of $5 or $15.
The farmer and baker may be able to get their mu­
tually preferred outcomes if they enter into a forward
contract. A forward contract is a particular kind of deriv­
ative contract, where the farmer promises to deliver a
commodity, one bushel of wheat, tomorrow in exchange
for $10; and the baker promises to deliver $10 tomorrow
in exchange for the commodity. If the farmer and baker
can commit to these promises, then they can get their
preferred outcomes—wheat for $10—and that’s the
end of the story. But if the farmer and baker cannot
commit, then delivery and exchange of wheat for $10
won’t happen. To see this, suppose that the price of
wheat turns out to be $5 and the baker does not accept
delivery from the farmer and, instead, purchases wheat
on the spot market, that is, the baker strategically defaults
on the agreement because the spot price is lower. This
strategy gives the baker an extra $5, compared with
the strategy of performing his contractual obligation.
Similarly, the farmer can get a net benefit of $5 per
bushel by strategically defaulting when the price of
wheat tomorrow is $15. Although the farmer and baker
would like to exchange wheat for $10 tomorrow, their
lack of commitment prevents this from happening. If
the farmer and baker could somehow bind themselves
to a $10 per bushel agreement, then they would do so,
so long as the cost of binding isn’t too great. This is
where the notion of clearing comes in.
One way the baker and farmer may be able to
bind themselves to the contract is for each of them to
provide $5 of collateral upfront. The $5 of collateral
is used to cover any losses incurred by a counterparty
should the other counterparty fail to perform on the
contract. For example, if the price of wheat is $5 and
the baker reneges on the contract, then he loses his $5
of collateral, which is given to the farmer. In effect,
the baker pays $10 for the wheat. It would seem that

Federal Reserve Bank of Chicago

with the introduction of collateral, the parties should not
have an incentive to renege on their contracts. Maybe.
One tricky issue is who or what is to hold the
collateral? Notice that the lack of commitment cannot
be overcome by simply having each party hold the
other’s collateral. To see this, suppose the spot price
turns out to be $15. In this situation, the farmer does
best for himself by selling his wheat on the spot market
and keeping the baker’s collateral (and the baker keeps
the farmer’s collateral). Here, one holding of collateral
simply offsets the other one and does not guarantee
performance. It appears that a third party is needed to
hold the collateral of the baker and farmer.
With the introduction of a third party, things could
work as follows. If the baker and farmer perform on
their contract, then the third party returns the collateral
to each of them. If, however, one party defaults and,
as a result, harms the other party, then the third party
can use the collateral of the nonperforming party to
compensate the other party for his losses. For example,
if the farmer defaults when the price of wheat is $15,
then the third party gives the baker the farmer’s collateral,
as well as his own. From the farmer’s point of view,
he will pay $10 for wheat if he chooses to default, so
he now has no incentive to default. It appears that the
introduction of a third party that holds the collateral
of the farmer and baker implies that they will each
perform their contractual obligations. Maybe.
In order for the three-party arrangement to work,
it is necessary for the third party to be able to verily
which party reneges in the event of contractual non­
performance. Suppose that there is an informational
friction. In the last example, where the spot price is
$15, the farmer can renege and claim to the third party
that he attempted to deliver the wheat to the baker
but, for some reason, the baker refused to take deliv­
ery. The farmer, then, could argue because of the bak­
er’s nonperformance, he had no choice but to sell his
wheat on the market and he should not forfeit his col­
lateral. Hence, if the third party cannot perfectly ob­
serve or verily the actions of the farmer and baker,
then it will be unable to determine which party in fact
reneged. This implies that a simple third-party mech­
anism—one that simply holds collateral—cannot
guarantee performance.
One way around this verifiability problem is to
have all the transactions related to the contract go
through the third party. That is, instead of having the
farmer deliver wheat to the baker and the baker deliver
$ 10 to the farmer, all deliveries are made to the third
party. The third party then “redelivers” the wheat and
money according to the original contract. Under this
scheme, the initial contractual obligation between the

139

baker and farmer is discharged and replaced with two
new contracts: one between the third party and the baker
and another between the third party and the farmer.
This process, called novation, makes the third party a
central counterparty, or CCP, to the original parties
of the contract. Notice that novation circumvents the
verifiability problem: Since the farmer and baker make
deliveries to the CCP, the CCP is able to determine
whether each party has performed its obligations. The
CCP is now able to transfer the collateral it holds to
the appropriate party in the event of nonperformance.
The above example illustrates how novation and
collateral can substitute for commitment. Collateral and
novation are only two possible ingredients or processes
that fall under the rubric of clearing. Things like trade
matching and confirmation (that is, checking to make
sure party A entered into a contract and that party B is,
in fact, on the other side of the contract), information
warehousing (that is, compiling in an accessible manner
the set of all the trades that have taken place), and
risk management are also part of the clearing process.
In the example, the commitment friction is needed
for a CCP to emerge; without it counterparties would
not strategically default and there would be no need
for a CCP. In practice, one might think that even if
counterparties cannot commit, concern about their
reputations might provide sufficient inducement for
performance in the absence of collateral and CCPs.
For example, a major financial player will not strate­
gically default on a contract (which is not backed by
collateral) for only a small gain today, since that would
damage its reputation as a reliable institution in the
future. There is a cost associated with being viewed as
unreliable: Potential future counterparties will choose
not to enter into mutually beneficial contracts with an
“unreliable” institution. (These counterparties fear the
institution will not perform on its contract.) If a counter­
party’s concern for its reputation is sufficiently strong,
then it will perform on its contracts even though it lacks
commitment. If this is always the case, then my claim
that a commitment friction is responsible for the emer­
gence of CCPs seems inappropriate. That is, reputation
will enforce performance and a CCP is not needed. I
now examine this issue.

Commitment and reputation
An implication of the commitment friction, as
illustrated in the farmer-baker example, is that if
counterparties do not post collateral with the CCP,
then one of the counterparties will always strategically
default. Yet, in practice, we routinely observe people
and organizations honoring contracts that are not backed
by collateral. We also observe contractual performance

140

when collateral is posted on a bilateral basis, that is, a
third-party CCP does not hold the collateral. Do these
observations imply that CCPs and other clearing pro­
cesses arise for reasons other than a commitment fric­
tion? One could argue that counterparties perform on
their contracts because they care about their reputations,
and concern for one’s reputation outweighs the commit­
ment friction. Below, I argue that a commitment friction
is still relevant even when parties care about their
reputations. Hence, we should take the commitment
friction seriously when thinking about clearing.
Here, I modify the farmer-baker example so that
I can discuss the notion of reputation. Instead of having
a “one-shot” contracting relationship between the
farmer and baker, suppose that the farmer and baker
repeatedly interact with one another. The repeated nature
of the farmer-baker relationship may provide an incen­
tive for them to perform their contractual obligations
even when there is a commitment friction and no
collateral is posted. Why? Because performance today
implies that parties will choose to contract with each
other in the future, and this interaction is beneficial to
both parties. Or, put another way, a party will only
enter into a new contract with a counterparty that has
a reputation for performing.
To see this, suppose the farmer and baker agree
to exchange wheat for $10. Suppose further that if the
farmer fails to deliver wheat when the spot price is $15,
then both parties decide that they will never enter into
another contractual relationship with each other. The
baker believes that if they do, then the farmer will always
fail to deliver when the price is $15. With this belief,
the baker has no incentive to enter into a contract with
the farmer. So, in contrast to the original example,
where the interaction between the farmer and baker
was a one-shot affair, there is now a cost associated
with not performing, which is the inability for the
farmer and baker to stabilize the price of wheat at $ 10
in the future. In deciding whether or not to perform
when the price of wheat is $15, the farmer compares
the immediate benefit associated with nonperformance,
which is $5, with the future benefit associated with per­
formance, which is the ability to sell wheat at $10 in
future periods. If the latter exceeds the former, the farmer
will perform; otherwise he will strategically default.
One might expect in “normal” times that counter­
parties will perform their contractual obligations even
when no collateral is posted. This is because the benefit
associated with future contracting outweighs the short­
term gain associated with default. However, in times
of “stress,” which can be characterized by either large
price movements and/or weakened balance sheets, the
benefit associated with reneging can look quite attractive.

4Q/2011, Economic Perspectives

For example, if there is a huge price movement, a
counterparty may find it in its best interest to avoid
taking a large loss today, even though it will be unable
to enter into a contract in the future. (And this may be
the case even when the reneging counterparty has a
strong balance sheet and the resources to perform.)
Alternatively, if a counterparty has a weak balance
sheet, fulfilling a contractual obligation may make the
firm insolvent or vulnerable, whereas reneging will
allow it to “live” for another day or become stronger.
Faced with these choices, the counterparty may choose
to strategically default. It is important to emphasize that
during stressful times, parties that anticipate receiving
positive contractual payments are really counting on
their counterparties to perform. Yet, it is precisely at
such times that a counterparty is most likely to (strate­
gically) default.5 If, from a social perspective, it is im­
portant to mitigate contractual defaults during times
of stress, then we know from the earlier example that
performance can be enhanced if collateral is posted
with the CCP. Hence, even though parties will perform
on their contractual obligations in the absence of posting
collateral in most circumstances, a CCP will still emerge
and require counterparties to post collateral. The reason
is that performance is particularly important to counter­
parties (and society) during times of stress, but this is
a time when their counterparties may not care about
their reputations and may strategically default.6
The above discussion implies that when there is
a commitment friction, a CCP may emerge to enhance
performance even though counterparties care about
their reputations.

Risk management
In practice, CCPs devote considerable resources
to risk management. Unfortunately, the example I in­
troduced earlier is too stylized to illustrate the notion
of risk management because the CCP does not face any
risk. In the example, the size of the CCP’s collateral
holdings eliminates the risk that it will fail to perform
in the event that one of the counterparties fails to per­
form. The CCP holds $10: $5 from the farmer and $5
from the baker. In the event that either the farmer or
baker defaults, the CCP needs $5 to guarantee perfor­
mance, which it has. In the real world, however, prices
of goods or assets do not move in nice, finite, discrete
amounts over a specific time interval. Although short­
term price movements are typically not that large in
relative terms, one can’t rule out enormous price move­
ments over a period of a few hours. One can imagine
modifying the example to allow for a continuous dis­
tribution of spot prices characterized by low probabilities
of large price movements. In that case, the CCP can

Federal Reserve Bank of Chicago

only guarantee performance if it holds enormous amounts
of collateral. But this particular solution to the perfor­
mance problem may not be an attractive one. Posting
collateral is costly for counterparties, as they have more
productive uses for their resources. If a CCP demands
huge amounts of collateral—so as to guarantee perfor­
mance for any conceivable price movement—then the
farmer and baker may simply stop using forward con­
tracts. In this situation, forward contracts would become
extremely costly to use. The farmer and baker may
therefore prefer to transact on spot markets and face the
(lower cost) price risk. This is where risk management
comes in.
The CCP can economize on collateral by “guaran­
teeing” performance with a high probability (but less
than one). So, although there is a possibility that the
CCP will default on its obligations, it minimizes this
possibility by managing the risk of failure that it faces.
A CCP can do this by, for example, requiring adjust­
ments to collateral when it perceives that risk has
changed and making provisions for additional resources
should its collateral holdings prove to be insufficient
to cover a default by one of its counterparties. I will
now develop these ideas a bit further.
The amount of collateral that counterparties post
can depend on a number of things. It can depend on
the volatility of the price of the commodity or assets
that underlie the derivative contract.7 There is a posi­
tive relationship between the volatility of the price of
the underlying asset and the volatility of the value of
the derivative contract. If the underlying asset has low
price volatility, then it will require less collateral than
an underlying asset with high price volatility. The idea
here is that if a counterparty fails, then, on average, a
low-price-volatility underlying asset will require a smaller
amount of the CCP’s resources to ensure contractual
performance than a high-price-volatility underlying
asset. Also, if the price volatility of the underlying asset
changes over time, then so should the amount of collateral
that counterparties post with the CCP. For example, if
the volatility decreases, the CCP will transfer some of the
collateral back to the counterparty; if it increases, then
it will require additional collateral from the counterparty.
For longer-lived derivative securities, the value of
a particular position will change over time. For example,
if the price of the underlying asset, say, wheat, in a
forward contract increases over time, then the value of
the forward contract for the counterparty who delivers
it, the farmer, falls, since he is delivering an asset
whose value is much greater than the delivery price.
If this counterparty fails to deliver, then the CCP faces
a very high performance cost. The CCP can eliminate
these “accumulated liabilities” by requiring counterparties

141

to settle these liabilities on a periodic basis. That is, the
CCP marks-to-market its contracts—directly passing
resources from counterparties whose contracts have
lost value to those whose contracts have gained value.
The amount of collateral required by the CCP
can also depend upon the “liquidity” of the contracts
it novates. In practice, if a counterparty defaults, the
CCP sells the defaulting counterparty’s positions. If a
position is “liquid,” it can be sold quickly and at a low
cost. For example, contracts that are exchange traded
and cleared, such as futures contracts, are very liquid.8
The market provides a fairly accurate estimate of the
value of a position, and even a very large position can
be sold to the market over a relatively short period.
On the other hand, specialized OTC derivative contracts
are much less liquid because their estimated value is
subject to great variation and they are traded on a bi­
lateral basis. Hence, it may be difficult to sell a large
position of OTC contracts on short notice at a price that
is at or near their estimated value. Since less liquid
contracts are more costly to trade than liquid ones, a
CCP will require higher collateral for OTC contracts.
The CCP may have access to resources other than
collateral to help it facilitate contractual performance
in the event of a counterparty failure. The CCP may
require all of its members, that is, members that have
their contracts novated by the CCP, to contribute to a
guarantee fund. Members make this contribution when
they join the CCP, before the CCP novates any of their
contracts. This fund could be accessed in the event a
member defaults and its collateral is insufficient to
guarantee performance. (This happens when the value
of the defaulting member’s portfolio—which includes
current payments—plus the value of the member’s collat­
eral is negative.) Such an arrangement is sometimes
called loss mutualization, because losses are shared or
mutualized among the nondefaulting or surviving mem­
bers. If the defaulting member’s collateral and the entire
guarantee fund are insufficient for the CCP to perform,
then the CCP members’ agreement may require them
to provide additional resources to ensure performance.
Except for the requirement that members provide
additional resources in the event of a member default,
the risk-management strategies described above are
consistent with the commitment and informational fric­
tions. That is, contributions to the guarantee fund, the
posting of collateral, and marking-to-market represent
various payments that members make and receive that
are designed to enhance commitment. Importantly, these
payments are made before any default occurs. The re­
quirement to provide additional resources in the event
of a member’s default is subject to a commitment fric­
tion. Since the additional payment to the CCP occurs

142

after a default, members may choose to honor their
promises or not. In particular, in times of stress, members
may choose not to honor their promises because the
benefit associated with being a CCP member in the
future is less than the resources they have to sacrifice
today. The analysis regarding whether members will
contribute extra resources to a CCP after a member
defaults is the same as in the farmer-baker example in
the previous section, “Commitment and reputation.”
To summarize, risk management is an important
element of the clearing process because the CCP’s perfor­
mance guarantee is only as good as its risk-management
strategy. Risk-management strategies, such as markingto-market, making contributions to a guarantee fund,
and adjusting collateral holdings when perceived risk
changes, are consistent with the commitment friction.

The structure of central counterparties
In practice, a CCP has a set of members. Only
members can clear contracts with the CCP. That is, a
CCP novates only those contacts that are presented by
its members. Counterparties that are not members of a
CCP—let’s call them customers—have their contracts
novated by a CCP member. In this arrangement, the CCP
guarantees contractual performance for its members,
and a CCP member guarantees contractual performance
for its customers. Hence, there is a tiered, but separated,
relationship between customers, CCP members, and
the CCP. There is no direct contractual relationship
between the CCP and customers.
If a clearing member defaults on its contractual
obligations with the CCP, the CCP guarantees perfor­
mance of the defaulting member’s contracts. That is,
the CCP will perform its contractual obligations for all
of its nondefaulting or surviving CCP members. As a
result, all surviving CCP members will be able to per­
form their contractual obligations with their customers
and the CCP. It must be pointed out, however, that the
CCP guarantee does not (typically) extend to customers
of a defaulting CCP member. Since customers’ contracts
are guaranteed by CCP members, they effectively lose
any performance guarantee when the entity that guaran­
teed performance for them—the defaulting member—
no longer exists.9
Just as in the farmer-baker example, CCP mem­
bers post collateral with the CCP, and customers post
collateral with CCP members. So, the collateral that
CCP members post with the CCP can come from their
customers or from the members directly for contracts
that they entered into on their own account.
A clearing member is responsible for the perfor­
mance of the contracts that it brings to the CCP. So, if
a customer defaults on its contractual obligation to a

4Q/2011, Economic Perspectives

CCP member, the member must step in and ensure
performance or be in default with the CCP. Etfectively,
the defaulting customer’s contracts become the con­
tracts of the member. The CCP member, however, does
not have to hold onto the contracts associated with
the defaulting customer as part of its portfolio; the
member can always sell them. In either case, the CCP
member will receive the collateral that was posted for
the defaulting customer’s position. If the collateral
requirements were appropriately calculated, then they
should cover both the payments made by the member
and any losses associated with either holding or selling
the defaulting customer’s position. If, for some reason,
the collateral is insufficient to cover the losses, the
CCP member must absorb the losses or be in default
to the CCP. The member will typically be willing to
absorb these losses for the same reason as the farmer
is willing to deliver wheat to the baker at $10 when
the spot price is $15 in the absence of collateral. The
member values the future benefit associated with being
a CCP member more than the short-term benefit of
walking away from the losses. In times of stress,
however, a CCP member may choose to default on its
performance obligations.

Debate about central counterparty
membership
At the heart of the debate over CCP membership
criteria is the liquidity of OTC contracts. The debate
can be loosely characterized as follows. One side be­
lieves that because OTC contracts are not very liquid,
a CCP member must be able to assume the portfolio
of a defaulting member. This necessarily implies that
members must have significant capital (typically at
least $5 billion) available for clearing purposes. The
other side believes that the illiquidity of OTC deriva­
tives is overstated; for example, OTC interest rate
derivatives and credit default swaps are quite liquid.
If a CCP member defaults, then there are methods avail­
able to dispose of the portfolio, other than requiring
another member to purchase it. And finally, if a require­
ment for membership is significant financial resources,
then membership will be limited to a very small set of
financial institutions, which in turn could give these
institutions undue market power. I examine these views,
starting at the heart of the debate: liquidity.
Contracts are said to be liquid if the value, or “fair
price,” of the contract can be accurately determined on
an ongoing basis and large amounts of contracts can be
bought or sold at or near the fair price in a short period.
A CCP adopts risk-management strategies to en­
hance its performance guarantee. These strategies—
such as collateralizing positions, marking-to-market,

Federal Reserve Bank of Chicago

and disposing of a defaulting member’s portfolio—are
easier to implement when contracts are liquid. The
amount of collateral posted depends, in part, on the
volatility of the price of the underlying asset and of the
value of the OTC contract. Higher volatility implies
that more collateral should be posted, and changes in
volatility imply that the amount of collateral posted
should also change. When contracts are liquid, good
estimates for levels of volatility can be obtained. The
CCP can be reasonably assured that the process of
periodically marking-to-market members’ positions
will not leave the CCP with additional liabilities in the
event of a member’s default when contracts are liquid.
Since large positions can be sold quickly at or near
the fair price when contracts are liquid, the CCP will
be able to efficiently dispose of a defaulting member’s
portfolio. A final risk-management strategy, contribu­
tions to a guarantee fund, provides an additional buffer
for the CCP against losses from a member’s default.
The size of a member’s contribution determines the
maximum notional value of contracts (or risk) that the
member can bring to the CCP.
If all contracts were liquid, there would be no debate
associated with CCP membership. Any counterparty
that could post the required collateral for its positions
and contribute to a guarantee fund would be able to
become a member.
Because some contracts are not very liquid, however,
any sensible risk-management strategy, independent of
the structure ofmembership, will require higher levels
of collateral from CCP members.
An important aspect of guaranteeing performance
is the CCP’s ability to sell a defaulting member’s port­
folio quickly, at or near the fair price.10 When contracts
are highly liquid, the CCP does this by simply selling
the portfolio to the market. What can the CCP do when
the contracts are not very liquid and there is no market
to sell to? One side of the debate over CCP member­
ship says the solution is to require each CCP member
to assume, that is, purchase, part of the defaulting mem­
ber’s portfolio. Since the notional value of a defaulting
member’s portfolio may be quite large, a CCP member
will require substantial capital to assume its share.
Hence, this solution would restrict CCP membership
to those who possess significant capital that can be
used for clearing purposes.
There may be problems associated with this solu­
tion. Because of the significant capital requirement for
membership, the number of individuals or institutions
that can qualify for membership will be small. This
implies that the amount of liquidity that can be brought
into the clearing process is limited (by the wealth of the
small number of clearing members). Perhaps more

143

importantly, if there is only a small number of clearing
members, then they could use the resulting market power
to adversely influence the pricing of clearing services
and the pricing of the OTC contracts themselves.
If a clearing member defaults, one side of the debate
advocates that the CCP auction the defaulting mem­
ber’s portfolio among a small number of institutions,
the CCP members. Auctions share many of the desir­
able properties associated with exchanges or markets,
such as price discovery and a place that brings buyers
together with a seller. In other words, an auction can
provide (some) liquidity for the objects that are being
sold. But it is neither clear nor obvious why the auc­
tion would operate more efficiently from society’s
point of view if it is restricted to only “wealthy” bidders.
Since a CCP also serves as an information warehouse—
collecting and disseminating the prices of cleared
contracts—this information could be used by anyone,
that is, a CCP member or nonmember, who would like
to bid on part or all of the defaulting member’s port­
folio. Opening up the auction to nonmembers would
make more liquidity available to the clearing process.
(Of course, if a nonmember purchases part of the port­
folio, then those contracts would have to be cleared
through a CCP member.) There does not appear to be
a rationale to limit the sale of a defaulting member’s
portfolio to only (wealthy) clearing members.
Another problem created by the wealth restriction
for membership lies in the pricing of products. There
is always a public policy concern regarding the pricing
and supply of services when the number of service
providers—in our case, CCP members—is small. That
is, prices will be too high and quantities will be too
low. Perhaps a bigger potential problem is that since
CCP members are free to choose their customers, they
may choose to clear only those OTC contracts for
which they are direct counterparties, thereby limiting
competition in the OTC derivative markets.11 Hence,
the wealth restrictions for CCP membership, which are
motivated by clearing considerations, can have adverse
affects on the pricing of the OTC derivative contracts.
Ironically, the wealth restriction could ultimately pre­
vent these contracts from become more liquid.
The wealth restriction for membership seems arti­
ficial. As long as an institution can cover the risk that
it brings into the CCP, by providing appropriate levels
of collateral and making contributions to the guarantee
fund, there does not appear to be any reason to exclude
it from membership. In an unrestricted membership
environment, CCP members would compete for cus­
tomers by appropriately pricing their services. With this
structure, there would not be any obstacles to clearing
all “clearable” contracts (not just those for which a

144

member is a counterparty) or moving current OTC
contracts onto exchanges. Exchange trading would
improve both the liquidity of the contracts and the
CCP’s performance.
In summary, restricting CCP membership limits the
amount of liquidity in the clearing process. In addition,
membership restrictions can have adverse affects on
the provision and pricing of both clearing services
and OTC contracts.

End-user exemptions
“Since we weren’t part of the problem, we shouldn’t
have to pay.” This statement nicely summarizes the
sentiment of many nonfinancial corporate end-users
of OTC derivative products regarding that part of the
Dodd-Frank legislation that mandates clearing for
most OTC derivative contracts.
In most cases, nonfinancial corporations purchase
OTC derivative contracts to hedge their business risks.
These contracts are attractive because, unlike exchangetraded derivative contracts, they can be tailored to the
firm’s business needs (for example, in terms of timing
of payments). Because these firms are using the con­
tracts for hedging rather than speculative purposes,
they usually hold onto them until they expire.
Currently, if a nonfinancial corporation wants to
purchase an OTC derivative, such as a swap, it nego­
tiates the terms directly with a dealer. The dealer typically
does not require the nonfinancial firm to post collateral.
However, the firm does pay a premium over similar
products that require the buyer to post collateral. Non­
financial firms claim that a requirement that all OTC
contracts be centrally cleared will raise their cost of
hedging, because they will now be required to post
collateral. These firms argue that since they use the
contracts to hedge their business risks and, by and
large, they did not default on their derivative contracts
during the financial crisis, they should not have to
bear this cost.12
I conclude that, even in the best case scenario, the
cost of hedging for nonfinancial corporate end-users
will increase when their contracts are cleared. In the bestcase scenario, the dealer posts collateral and the enduser does not. This is the best-case scenario from an
end-user’s perspective, because it does not have to post
collateral. I now provide the details for this scenario.
Suppose that the dealer that negotiates the corporate
end-user’s swap has a large diversified portfolio of
end-users. As well, assume that corporate end-users will
never (strategically) default on their swap obligations.
That is, whenever end-users make the calculation, the
benefit associated with continued access to the swap
market exceeds the benefit of defaulting on a swap

4Q/2011, Economic Perspectives

payment. Hence, the only time an end-user defaults on
a swap payment is when it is insolvent (or bankrupt).
If one assumes that the probability of a firm becoming
insolvent is independently distributed across firms,13
then the dealer understands that a certain proportion
of its swap contracts will end up in default. The dealer
can charge a premium that reflects the proportion of swap
contracts that will fail. From the end-user’s perspective,
this can be interpreted as an insurance premium. A
default by an end-user will not create any problems for
the broker-dealer since defaults are anticipated and
priced. In this ideal world—where end-users never have
an incentive to strategically default and (nonstrategic)
defaults are uncorrelated—end-users do not have to
post any collateral. Note that this result is consistent
with my “simple” model of clearing, where collateral
was required to guarantee performance because either
the farmer or baker would always have an incentive
to strategically default.
Prior to the recent financial crisis, the dealer typically
would not have been required to post collateral for its
positions with the end-user. Most of the time, dealers
would contractually perform because it was in their
best interests to do so. However, in times of stress, the
dealer might default on its swap obligations. As a result,
the end-user would not receive any payment—which
may be critical during times of stress—and would be
one of many creditors seeking remedy from the dealer.
Clearly, contractual performance would be enhanced if
the OTC contracts were cleared by having dealers post
collateral for their positions with corporate end-users.
But posting collateral is not cheap. When dealers
post collateral for their swap contracts with corporate
end-users (as mandated by the Dodd-Frank legislation),
they will pass some proportion—possibly all—of the
associated cost to the end-users. Therefore, even if cor­
porate end-users do not post collateral for their positions,
their cost of hedging will increase.
Up to this point, I have assumed that corporate endusers don’t strategically default and that defaults by
nonfinancial firms are uncorrelated. These assumptions
are unrealistic. In times of stress, when there are large

Federal Reserve Bank of Chicago

price movements and/or weakened balance sheets, endusers may find it in their best interests to strategically
default on a swap payment, even though they have
the resources to pay. If many end-users default, then
the insurance premium that dealers charged them will
be insufficient to cover the dealers’ losses. If, from a
social perspective, it is important that dealers do not
experience large-scale loses on their swap contracts
during times of stress, then end-users’ swap positions
should be properly cleared through a CCP, which will
require them to post collateral.
Although nonfinancial firms, by and large, purchase
OTC contracts for the purpose of hedging, it is not at
all obvious that these entities do not pose a threat to the
stability of the financial system. Nonfinancial corporate
end-users represent a relatively large share of the OTC
market, 10 percent to 15 percent. If these firms receive
a correlated shock that weakens their ability to perform,
they may transmit this adverse shock to the balance
sheets of the dealers. The potential effects of this shock
can be greatly mitigated by requiring them to post
collateral for their positions.

Conclusion
In the first part of this article, I sketched out a frame­
work for thinking about clearing. I used the insights
from this framework to examine two prominent themes
from a recent symposium on clearing. In terms of CCP
membership, there is an alternative (unrestricted) CCP
structure that is at least as effective as one that requires
members to have substantial capital. The alternative
structure has the added benefit (which could be huge)
of promoting both competition and the provision of
liquidity in clearing and in the OTC derivatives markets.
In terms of end-user exemptions, the cost of hedging
will increase for end-users even if they are not required
to post collateral. Since end-users’ positions are non­
trivial, in the sense that (correlated) defaults by endusers can weaken the ability of their dealers to perform,
they should post collateral in order to strengthen dealers’
ability to perform.

145

NOTES
‘Derivatives are financial contracts whose value is linked to the price
of an underlying commodity, asset, rate, index, or the occurrence or
magnitude of an event. These contracts are traded both on traditional
exchanges and over the counter.

2In 2008, the notional value of all AIG’s derivative contracts, including
credit default swaps, was as high as $2.7 trillion.
3See www.freakonomicsmedia.com/2008/09/18/diamond-and-kashyapon-the-recent-financial-upheavals/ for a helpful Q&A about the
financial market disruption in late 2008.

4The symposium agenda is available at www.chicagofed.org/webpages/
events/2010/public_policy_symposium_on_OTC_derivatives_
clearing.cfm.
5Some people argue that large companies do not strategically default
because the reputational cost is too great. So, if these companies do
default, it is because they are either bankrupt or insolvent. But, if a
company senses it may become insolvent, it may attempt to counteract
this by taking defensive actions such as rescheduling debt and other
payments in an attempt to save itself. One can interpret this as strate­
gically defaulting—the company has the resources to pay current
bills, but chooses to withhold these payments.

6We observe this in the context of sovereign debt. For example, a
country has the resources to pay for debt, but chooses not to because
the current costs of doing so are too high, say, because the country
would have to increase taxes on its citizens when the economy is
weak. This is a costly decision—when a country repudiates debt
payments, its reputation will take a hit, in the sense that it will be
shunned in international debt markets in the foreseeable future.

146

7In the earlier example, the derivative contract is a forward contract
and the commodity that underlies the derivative contract is wheat.

8 A measure of liquidity for exchange-traded contracts is the bid-ask
spread. A low bid-ask spread indicates that the contract is liquid; a
high bid-ask spread indicates that it is not very liquid.
9If the customers’ collateral payments are segregated from the member’s
own collateral and the member defaults on its own positions, then
the customers’ collateral is protected. In practice, customers of the
defaulting member would transfer their accounts to one or more
nondefaulting members, who then can “novate” the customer con­
tracts of the defaulting member.
10The CCP may be able to avoid receiving a low selling price if addi­
tional time is taken to sell the position. There are two possible prob­
lems, however. First, there is a chance that the price of the contracts
can move against the CCP during a protracted selling period. This
implies that the CCP would require additional collateral from its
members. Second, the CCP may not have the luxury of time on its
side to dispose of the portfolio.

“If a clearing member buys or sells a contract from customer A, then
I say that customer A is a direct counterparty of the clearing member.
If customer A buys or sells a contract from customer B, a clearing
member can limit competition by choosing not to novate the contracts.
“However, there have been defaults by nonfinancial firms that dis­
rupted the broader markets—for example, Enron.
“This assumption is almost certainly false, that is, insolvencies will
be correlated. I make this assumption because it provides the bestcase scenario for the corporate end-user. Later, I discuss the impli­
cations of relaxing this assumption.

4Q/2011, Economic Perspectives

Worker flows and matching efficiency
Marcelo Veracierto

Introduction and summary
One of the best known facts about labor market dynamics
in the U.S. economy is that unemployment and vacancies
are strongly negatively correlated, an empirical relation­
ship called the “Beveridge curve.”1 In recent times, how­
ever, large deviations from the Beveridge curve have
been observed. In particular, vacancies have increased
quite significantly since mid-2009, but this phenomenon
has not been accompanied by a substantial decrease in
the unemployment rate (see figure 1). This failure of
the Beveridge curve has surprised many economists and
has been interpreted as evidence of mismatch, that is, of
increased frictions in the process through which workers
meet job opportunities (for example, Kocherlakota,
2010). The purpose of this article is to provide a mea­
sure of mismatch in U.S. labor markets and to assess
its importance in determining the behavior of the un­
employment rate and other labor market outcomes since
the start of the latest recession. The framework that
I use is a simplified version of the Mortensen and
Pissarides (1994) model. Since my purpose is to use
it as a tool for organizing and interpreting data, I will
abstract from any explicit decision making and focus
on the essential structure of the model.
The basic structure of the Mortensen and Pissarides
model has three main components. First, it has an aggre­
gate matching function that summarizes the process
through which unemployed workers and employers
with open vacancies search for each other and meet.
It functions very much like a standard production func­
tion, with unemployed workers and vacancies entering
as inputs of production and the number of matches
formed appearing as output. The second element is a
free-entry condition for the creation of vacancies. In
particular, it is assumed that there is a fixed cost to post
a vacancy and that employers create vacancies up to
the point at which the expected discounted value of a
filled job equals this fixed cost. The expected value of
a job is given by the probability of filling the job, which

Federal Reserve Bank of Chicago

is determined by the aggregate matching function, and
by the value of a job. In a full-blown version of the
model, the value of a job is endogenously determined
by the expected revenues that the job will generate and
by the bargaining power of the worker. However, in
the simplified version considered in this article, I am
silent about the explicit process through which the value
of a job is determined. The third main component is a
simple accounting relationship that states that the total
flows in and out of each labor market state must be equal.
A standard approach in the literature is to allow only
for two labor market states (employment and unemploy­
ment) and to assume that the model is always at its
steady state (that is, its long-run equilibrium). However,
I consider more flexible specifications in this article.
I use this simple version of the Mortensen and
Pissarides model to measure mismatch and evaluate
its consequences during the post-2007 recession period.
This is not the first article to do this. Two closely re­
lated papers are Barlevy (2011) and Bamichon and
Figura (2010). Barlevy follows the standard approach
by postulating two labor market states, assuming a con­
stant separation rate (that is, the rate at which workers
transit from employment into unemployment), and by
assuming that the model is always at its long-run steady
state. On the contrary, Bamichon and Figura incorpo­
rate a third labor market state (nonparticipation) and
allow the transition rates between the three labor market
states to vary over time. However, similar to Barlevy,
Bamichon and Figura assume that the model is always
at its steady state and that only unemployed workers
enter the matching function.2
Marcelo Veracierto is a senior economist in the Economic
Research Department ofthe Federal Reserve Bank ofChicago.
The author is grateful for comments from Gadi Barlevy,
Andy Neumeyer, and Juan Pablo Nicolini, as well as seminar
participants at the Federal Reserve Bank of Chicago and
the Universidad Torcuato Di Telia. He also thanks Hao Zou
for research assistance.

147

Given the different assumptions made in the liter­
ature, I use my model to evaluate how sensitive the re­
sults are to the different specifications. I consider the
following dimensions. First, I assess the importance of
allowing the separation rate to vary over time instead
of assuming it to be constant. Second, I evaluate the con­
sequences of specifying three labor market states instead
of two. Third, I assess the consequences of assuming
that the model is always at its steady state instead of
allowing for transitionary dynamics. Fourth, I evaluate
the consequences of allowing nonparticipants to enter the
matching function instead of assuming that the matching
function solely applies to unemployed workers.
I find that the results are extremely sensitive to
the alternative specifications. However, in the preferred
scenario (which has three labor market states, variable
transition rates, transitionary dynamics, and nonpartici­
pants entering the matching function), I obtain the fol­
lowing findings.3 First, the matching efficiency has
been quite volatile throughout the whole sample period
(2001:l[January]-2011:2[February]). Second, the
matching efficiency has been drifting down since the
start of the last recession. Third, the value of filled jobs
plummeted between 2007:12 (the start of the latest
recession) and 2009:6, but it has recovered quite signif­
icantly since then. Fourth, conditional on the observed
paths for the value of a job and all transition rates, the
drop in matching efficiency since the start of the re­
cession has had only a moderate impact on the unem­
ployment rate: The current unemployment rate would
be 1 percentage point lower if the matching efficiency
had stayed unchanged. Fifth, the bulk of the increase in
the unemployment rate since the start of the recession
is accounted for by changes in the transition rates across
labor market states. Sixth, the matching efficiency, the
value of a job, the transition rates, and the search in­
tensify of nonparticipants all have significant effects
on the dynamics of nonparticipation. Since they deemphasize the importance of matching inefficiencies
in explaining the large increase in the unemployment
rate since the start of the last recession, the results in
this paper are consistent with a greater role for policy
in achieving improvements in labor market conditions.
In the next section, I consider the case of two la­
bor market states. In the following section, I consider
the case of three labor market states and a matching
function with only unemployed workers. Then, I con­
sider the case of three labor market states but allow
for nonparticipants to enter the matching function.
Readers solely interested in learning about the rela­
tive contributions to unemployment dynamics should
jump to the last section of the paper, which uses the

148

preferred scenario. The first two sections report results
using alternative but less satisfactory methodologies.

The case of two labor market states
There are two types of agents: firms and workers.
Each firm has one job available, which can either be
filled or vacant. The expected discounted value of
profits generated by a filled job is equal to J units of
the numeraire. Posting a vacant job requires k units
of the numeraire. There is an infinite number of poten­
tial firms. Workers can be in either of two states: em­
ployed or unemployed. Employed workers get separated
from their current jobs with probability X. Unemployed
workers and posted vacancies determine the total number
of new matches that are formed according to the fol­
lowing matching function:

1)

M,

where Mt is the total number of new matches, £7 is the
total number of unemployed workers, F is the total num­
ber of posted vacancies, At is the productivity of the
matching function, and 0 < a < 1.
Normalizing to one the total number of workers
in the economy, the evolution of unemployment over
time can be described by the following equation:

4Q/2011, Economic Perspectives

2)

U+l = U-Mt + (l-Ui-)\.

That is, the total number of workers that will be unem­
ployed tomorrow f/+1 is equal to the total number of
currently unemployed workers that do not find a match
Ut-Mt, plus the total number of currently employed
workers that get separated from their jobs (1 - Ut)'kf
Since firms are profit maximizers, the following
free-entry condition must be satisfied:

That is, the cost of posting a vacancy k must be equal to
the probability of filling a vacancy MJVt times the ex­
pected discounted value of profits generated by a filled
job Jt. If this condition was not satisfied, the total number
of vacancies created would be either zero or infinity,
depending on the direction of the resulting inequality.
Observe that the productivity of the matching
function At, the separation rate X(, and the expected
discounted profits generated by a filled job Jt are ex­
ogenous to the model. Given the total number of
workers unemployed at date zero Uo, the model gen­
erates an endogenous path for \Mt, V„Ul+l}t f)-

Steady state
Assuming a constant matching productivity A,
a constant separation rate X, and constant expected
discounted profits generated by a filled job J, a steady
state of the model economy can be defined as an initial
unemployment level Uo= U, such that the endogenous
path for {Mt, V, UM p that the model generates is
constant over time. That is, that Mf = M,V= V, and
Ut=Ufor every t > 0. A steady state (M, V, IT) can
be interpreted as the total matches, vacancies, and
unemployment that the economy will converge to in
the long run.
From equations 1-3, we have that the conditions
a steady state must satisfy are the following:
4)

5)

6)

Substituting equation 5 in equations 4 and 6 gives
the following simplified steady-state conditions:

7)

U=

X
x+<r’

8) k=A^J.

Equation 7 defines a negative relationship between
unemployment and vacancies and, for this reason, is
called the “Beveridge” curve. Equation 8 defines a
positive linear relationship between unemployment
and vacancies and, since it is defined by a free-entry
condition to the posting of vacancies, it is called the
“job creation” curve. The Beveridge and job creation
curves are depicted in figure 2. The intersection of
these curves determines the steady state (U\ F‘).
ft is particularly important to detennine what causes
shifts in each of these two curves, ft is possible to show
that an increase in the separation rate X shifts the
Beveridge curve up, an increase in the expected dis­
counted profits from a filled job J does not affect the
Beveridge curve, and an increase in the matching
efficiency parameter A shifts the Beveridge curve down.
In turn, the separation rate X has no effect on the job
creation curve, but an increase in either J or A rotates
the job creation curve clockwise. Given these shifts in
the Beveridge and job creation curves, we can now
detennine how changes in X, J, and A affect the steadystate pair (f/*, F‘). In particular, we can conclude that
that an increase in X increases both vacancies F and
unemployment U, that an increase in J increases F and
reduces U, and that an increase in A reduces F. The ef­
fects of an increase in A on U are unclear from the figure,
but substituting equation 8 in equation 7 gives that

M=Af/“F1-“,

M=(l —C/)X,

J.

Federal Reserve Bank of Chicago

Thus, we can safely conclude that an increase in J
reduces U.
To the extent that the transitionary dynamics in
response to a change in either X, A, or J are fast, busi­
ness cycle fluctuations in unemployment and vacancies
can be studied by performing the steady-state analysis
described in the previous paragraph. Assuming that
this is the case, we can make the following tentative

149

hypothesis. First, since there is a strong negative em­
pirical relationship between vacancies and unemploy­
ment between 2001:1 and 2007:12, fluctuations in the
value of a filled job J, together with a relatively constant
separation rate X and a relatively constant matching
productivity ,t. are the most likely scenario for explaining
this period. Second, significant changes in A and/or X
are necessary for explaining the substantial deviations
from the Beveridge curve observed after 2007:12,
especially after 2009:6 (see figure 1). Akey issue will be
to determine the behavior of the matching efficiency
parameter A during this later period. Another key issue
will be to evaluate the contribution of changes in A,
X, and J to the unemployment and vacancy dynamics
observed during this later period. Addressing these
issues will be the focus of the next two subsections.
Before proceeding, it will be convenient to rewrite
equations 7-8 as follows:

10>‘ = <K

This makes explicit the assumption that the economy
at any month t can be safely described by the steadystate equations 7-8, an assumption that will be main­
tained throughout the rest of this section.

Constant separation rate
Shimer (2005) has argued that the separation rate
X does not play an important role in generating unem­
ployment fluctuations. For this reason, I follow Barlevy
(2011) and consider in this section that the separation
rate X is constant over time. Under this assumption,
I use the model described by equations 9-10 to measure
the time paths for the efficiency parameter At and the
value of a job Jt.
In what follows, I set the separation rate X to 0.042,
which is equal to the average employment-to-unemployment transition rate plus the average employment-tononparticipation transition rate between 2001:1 and
2007:12. From equation 9, we have, for any two months
z and y, that
x-x

In

150

i

Using the above value of a and averaging the
values of At between 2001:1 and 2007:12 obtained
from equation 12 gives an estimate of A = 1.06.
Using this constant value for A, we can then use
equation 9 to construct the vacancies predicted by the
model economy (conditional on the observed unem­
ployment rate) as follows:

13) V, =

A-xi
U,
A

U,.

X-x]
-In

u,

7

In

Within the period 2001:1-2007:12 (which is a
period with relatively constant matching productivity
A), we can thus select the month z with the largest U/V
ratio and the month j with the smallest U/V ratio and
use them to get an estimate for a from equation 11,4
These months happen to be z = 2003:6 and j = 2001:1.
The estimated value of a turns out to be 0.4915.5
Equation 9 can also be used to measure the
matching efficiency at month t as follows:

-In

iw
J
74

JJ

The predicted vacancies are shown by the red
line in figure 3. We see that under a constant matching
efficiency parameter A, the model does a good job at
reproducing the behavior of vacancies between 2001:1
and 2007:12. However, beginning in 2009, the model

4Q/2011, Economic Perspectives

FIGURE 3

FIGURE 4

Matching efficiency
(2 states, constant separation rate)

unemployment
Note: Vacancies and unemployment are normalized by
total labor force.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

fails to keep track of the data using a constant A. This
suggests that the matching efficiency parameter A may
have experienced substantial changes in this later period.
To show that this could be the case, figure 4 reports
the values for ,1, (in logs) measured from equation 12
for the whole sample period (the vertical line corresponds
to 2007:12, that is, to the start of the past recession).
We see that the matching efficiency was relatively stable
before 2008:1. However, starting in 2008:1, the matching
efficiency has fluctuated quite substantially. In partic­
ular, we see that after an initial increase, the matching
efficiency has been decreasing continuously, reaching
a cumulative drop of 17.5 percent by 2010:11.
Normalizing the cost of posting a vacancy k to
one and using the path for At already found, equation
10 can be used to construct a time series for the value
of a job 7. In particular, we have that

Figure 5 reports the evolution of the value of a
job between 2001:1 and 2011:2 (in logs). We see that
7( dropped quite substantially during the recession:
Between 2007:12 and 2009:8, the value of a job de­
clined by 68 percent.6
I now turn to evaluate the contributions of changes
in 7; and 7( to the dynamics of unemployment and vacan­
cies since the beginning of the recession. In order to

Federal Reserve Bank of Chicago

Note: The log of At is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

do this, I compute adjusted unemployment (7* and
adjusted vacancies V* using equations 9-10 under
the assumption thatrt.= 720()7|2, for every month /.
That is, I let the value of a job 7 evolve as in figure 5
(that is, as in the data) but fix the matching productivity
to the value that it had at the start of the recession. For
this reason, U* and V* measure the unemployment
rates and vacancies that would have been obtained
had the matching productivity remained constant at
its December 2007 level but the path for the value of
a job J had remained the same. Observe that in a full­
blown model (in which J is endogenously determined),
a change in the path for At would generally affect the
path for./. As a consequence, comparing (//,*, fj‘) with
(£/, F) cannot be strictly interpreted as describing the
total effects of variations in7(; it should be interpreted
as describing the conditional effects of ,t; (that is, con­
ditional on the observed path for 7). In a full-blown
model, the variations in 7 would have to be accompanied
by variations in other variables (for example, in the
bargaining power of workers) in order to obtain an
unchanged path for 7(7
Figure 6 shows the path for U* (labeled “constant
A”) and for f/ (labeled “variable A”). We see that through
2009:1, the productivity of the matching function did
not play an important role in the unemployment dynamics
observed (both paths are quite similar). However, starting
in mid-2009, we see that the decline in matching pro­
ductivity reported in figure 4 played a significant role
in generating a significantly larger unemployment

151

FIGURE 5

FIGURE 6

Value of a job
(2 states, constant separation rate)

Effects on unemployment rate
(2 states, constant separation rate)

Note: The log of Jt is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

rate. In particular, we see that by February 2011 the un­
employment rate would have been 7.5 percent instead
of 8.9 percent had the matching productivity remained
constant at its beginning-of-recession level.
Figure 7 reports the paths for V* (labeled “con­
stant A”) and for Vt (labeled “variable A”). We also
see that through 2009:1, changes in the productivity
of the matching function had negligible effects on
vacancies. However, by mid-2009 both paths start to
diverge, and we see that by February 2011 vacancies
would have been 1.6 percent instead of 1.9 percent
had the matching productivity remained constant at
its 2007:12 level.

Variable separation rate
In this section, I allow the separation rate to vary
over time. Figure 8 reports the separation rate between
2001:1 and 2011:2 that is obtained from the Bureau
of Labor Statistics’ Current Population Survey (CPS)
data (once again, the vertical line depicts the beginning
of the last recession). We see that early on in the re­
cession the separation rate increased quite significantly,
reaching 4.9 percent by 2009:1, but that it subsequently
trended down toward its pre-recession level.
Given the data on
and Vt and the separation
rate 'kt reported in figure 8,1 compute the matching
efficiency At from the following equation:

152

which is analogous to equation 12, except that X is
allowed to vary over time. The resulting path for the
matching productivity At is reported in figure 9. We see
that contrary to figure 4, we now observe large fluctu­
ations in At previous to the start of the recession. Another
difference is that there is a sharp increase in matching
productivity early on in the recession that compensates
for the 2009:1 spike in the separation rate. Also, we
see that starting in 2009:2, the matching productivity
trends down much more sharply than in figure 4.
The value of filled jobs is computed from
equation 14 using the values obtained from equa­
tion 15. The resulting path is reported in figure 10.
We see that this path is not very different from that
in figure 5.
Figures 11 and 12 explore the conditional contri­
bution to unemployment and vacancies dynamics of
the matching productivity ,L the separation rate X(, and
the value of a job 7. In particular, I compute adjusted
unemployment U* and adjusted vacancies V* using
equations 9-10 under the assumption that At = .-'Lo07l2
and X = X2007 12, for every month t. That is, I let the value
of a job 7( evolve as in figure 10 (that is, as in the data),
but I fix the matching productivity to the value that it
had at the beginning of the recession A,OO712 and fix the
separation rate to the value that it had at the beginning
of the recession X2007 12. In other words, (7* and V*
measure the unemployment rates and vacancies that
would have been obtained had the matching productivity
and the separation rate remained constant at their
December 2007 levels.

4Q/2011, Economic Perspectives

FIGURE 7

FIGURE 8

Effects on vacancies
(2 states, constant separation rate)

Separation rate

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey and
Job Openings and Labor Turnover Survey.

I also compute adjusted unemployment U** and
adjusted vacancies V** using equations 9-10 under
the assumption that At = .-'h()()7l2. for every month t
(but I let the separation rate X( vary as in the data).
That is, I let the value of a job J evolve as in figure
10 and the separation rate 'kt evolve as in figure 8 but
fix the matching productivity to the value that it had

Federal Reserve Bank of Chicago

Note: The separation rate is equal to the sum of the
employment-to-unemployment and employment-tononparticipation transition rates.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey.

at the beginning of the recession X,007 12. In other
words, U** and V** measure the unemployment rates
and vacancies that would have been obtained had
the matching productivity remained constant at its
December 2007 level.
In figure 11, U* is labeled “constant A, constant
X,” U“ is labeled “constant A, variable X,” and Ut

153

FIGURE 11

FIGURE 12

Effects on unemployment rate
(2 states, variable separation rate)

Effects on vacancies
(2 states, variable separation rate)

------ Variable A, variable Zs
------ Constant A, constant Zs
Constant A, variable Zs
Sources: Author’s calculations based on data from the

Sources: Author’s calculations based on data from the

Bureau of Labor Statistics’ Current Population Survey

Bureau of Labor Statistics’ Current Population Survey

and Job Openings and Labor Turnover Survey.

and Job Openings and Labor Turnover Survey.

is labeled “variable A, variable X.” We see that U*
increases in the early part of the period and decreases
during the second part, mirroring the evolution of the
value of a job J described in figure 10. In turn, U"
increases much more than U* early on in the recession
because of the early increase in the separation rate X(
depicted in figure 8, but as the separation rate reverts
toward its beginning-of-recession level, U" starts to
behave very much like U*. Finally, since the difference
between U** and Ut is solely due to changes in the
matching productivity, we see that the large increase
in matching productivity early on in the recession
(reported in figure 9) played an important role in keep­
ing unemployment relatively low. However, the large
drop in matching productivity since early 2009 signif­
icantly contributed to maintaining an unemployment
rate of more than 9 percent.
In him, figure 12 shows that the matching produc­
tivity doesn’t play a crucial role in vacancy dynamics.
However, the large increase in the separation rate
in the early part of the recession played a noticeable
role in keeping vacancies relatively high early on
in the recession.

The case of three labor market states
The model used in the previous section had two
labor market states: employment and unemployment.
In this section, I allow workers to be in a third labor
market state: nonparticipation (that is, out of the labor

154

------ Variable A, variable Zs
------ Constant A, constant Zs
Constant A, variable Zs

force). A main reason for doing this is that in the CPS
data between 2001:1 and 2007:12, the total number of
people transitioning from nonparticipation to employ­
ment is almost twice as large as the total number of
people transitioning from unemployment to employment
(see figure 13), although the differences have become
much smaller since the start of the past recession. By
considering only two market states, the analysis in the
previous section completely missed these transitions.
Another reason for introducing three labor market states
into the model is that with two labor market states, it
is not clear what separation rates to consider: separations
into unemployment or separations into both unemploy­
ment and nonparticipation? Explicitly introducing three
labor markets states avoids this type of issue. More gen­
erally, introducing three labor market states allows me
to address worker flows data in a more satisfactory way.
In this section, I follow Bamichon and Figura
(2010) and assume that the matching function solely
describes transitions from unemployment into employ­
ment. In particular, I assume that the matching function
is given by
16) M, =A,U“V'~\

where Mt are the total flows from unemployment into
employment, Ut is unemployment, H are vacancies,
and 0 < a < 1.

4Q/2011, Economic Perspectives

(either from employment or unemployment), minus
all transitions out of nonparticipation (either to employ­
ment or unemployment).
In what follows, total population will be normal­
ized to one, that is,

FIGURE 13

Flows into employment
thousands of people

6,000
5,000

20) £,+ G, + IV,= 1,
4,000

for every period /.
Similar to the previous section, the following
free-entry condition must be satisfied:

3,000
2,000

1,000

2001

’03

'05

'07

'09

'11

------ Unemployment-employment
------ Nonparticipation-employment
Source: Bureau of Labor Statistics’ Current Population
Survey.

The transition rate from employment to unemploy­
ment Xf°, the transition rate from employment to non­
participation X™, the transition rate from unemployment
to nonparticipation X™, the transition rate from nonpar­
ticipation to unemployment X,"‘'\ and the transition rate
from nonparticipation to employment X,v/' are assumed
to be exogenous to the model.
The evolution of workers across labor market
states is then given by the following equations:
17) El+l = E^ApX^ + ^N,- (X^+Xfw)2?„

18) U,+l = U, + ^UE, + X,WTV, - 7^NU* ~A,UX~^

19)

TV,+1 = TV, + Xfw£, + X™F, - (X,W£ + X7)TV,.

Equation 17 states that next-period employment
is equal to current employment, plus all transitions
into employment (either from unemployment or non­
participation), minus total separations (either to un­
employment or nonparticipation). Equation 18 states
that next-period unemployment is equal to current
unemployment, plus all transitions into unemployment
(either from employment or nonparticipation), minus
all transitions out of unemployment (either to employ­
ment or nonparticipation). Equation 19 states that
next-period nonparticipation is equal to current non­
participation, plus all transitions into nonparticipation

Federal Reserve Bank of Chicago

Observe that from the point of view of a firm,
the probability of filling a vacancy is equal to
(AT, + X,W£TV,)/ Vt, because matches can be formed
with workers either coming from unemployment
or from nonparticipation.
Given the total number of workers unemployed
at date zero Uo and the total number of workers that
are nonparticipants at date zero No, the model generates
an endogenous path for {M,, F,, G,+1, JV,+1}" Q.

Steady state
Assuming a constant matching productivity A,
a constant value of a job 7, and constant transition
rates X, a steady state of the model economy can be
defined as an initial unemployment level Ua = U and
an initial nonparticipation level No, such that the en­
dogenous path for {AT,, Vt, Ul+l, TV J ” that the model
generates is constant over time.
From equations 17-19, we have that the condi­
tions that a steady state (F. TV, F) must satisfy are the
following:

22)

(X££/+

X£Af){l - F - TV) = 7F“F1_“ + X^TV,

23)

(xW£

Xw + X™)

+

TV =

XEN(l-U) + XUNU,

24) t=(.WT-- + vw)J

Similar to the previous section, it will be conve­
nient to rewrite these equations as:

25)

(Xf^ +

Xfw)(l- F,- TV,)

=A,U^~a+ 7,™N,,

155

26) N, =

FIGURE 14

^N(l-U,}+ Xu,NU,
WE+ C+x“)

Transition rates (/.s)

27) ^AtfC+^N,.
•A
This makes explicit the assumption that the econ­
omy at any month t can be safely described by the
steady-state equations 22-24, an assumption that will
be maintained throughout the following two subsections.

Constant transition rates
Figure 14 shows the transition rates X™,
X™
X,v/', and
. in logs, normalized by tbcir average value
for the period 2001:1-2007:12. We see that these
transition rates were relatively stable prior to 2007:12.
However, we see that with the onset of the recession,
there was a significant drop in the transition rate from
nonparticipation to employment X(VA, a drop in the
transition rate from unemployment to nonparticipation
a large increase in the transition rate from nonpar­
ticipation to unemployment X,"‘'v and a large increase
in the transition rate from employment to unemploy­
ment X{'v In turn, the transition rate from employment
to nonparticipation was not significantly affected.
Based on figure 14, and similar to the previous
section, here I assume that the transition rates
^EU yEN
^NU
constant oygr jjjg

------ UN
------ EU
EN

-----------

NE
NU

Note: All transition rates are reported in logs and
normalized to zero at the start of the last recession.
Source: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey.

33) A, =

1

u?v,'-a
(

xBV(i-c/,)+xMX~P|

(1EU + 1EN \ \-u, -

period 2001:1-2007:12. Taking simple averages
over this period gives the following values:

V

(XV£+Xw'+Xral
L i
) Jy

-)C

[kNE+XNU+XEN]

28) /.'v= 0.2258.

29) XK'= 0.0132,

Assuming that A is constant over the period
2001:1-2007:12, we can then use two months i and
j within this period to get an estimate of a as follows:

30) XHV = 0.0281,

31) X'v/; = 0.0505,

32) rkNU= 0.0253.
Substituting equation 26 in equation 25 under
the assumption of constant X values, we get that

x£w(i-tx)+xwtx
(XW£+XW+X£W)

x {In

-X

156

NE

XEN [i-Ui} + 'kUNUi

[kNE+1NU +1EN\

4Q/2011, Economic Perspectives

FIGURE 15

FIGURE 16

Predicted Beveridge curve

Matching efficiency
(3 states, steady state, constant transition rates)

unemployment
Note: Vacancies and unemployment are normalized
by total civilian noninstitutional population (16 years
and older).
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

yEN

i-u.-

Note: The log of Af is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

1-V + a.wu/

XNE +XNU +XEN

-In

(VC'+VA'|
\EN
-r

1-V + XunUj

[vA'(i-u,) + VA'u 1

XNE +XNU +XEN

(r£+xw+vw)
+ ln(t/. )-ln(U,)}.

-XA'£
Picking i = 2003:6 and j = 2001:1, which are
the months with the largest and smallest VIUratio,
respectively, gives an estimate of a = 0.16.8
In turn, equation 33 can be used to measure the
matching efficiency at month /. Using the above
value of a and averaging the values ofAf between
2001:1 and 2007:12 obtained from equation 33 gives
an estimate of A = 0.4533.
Using this constant value for A, I can then use
equation 33 to construct the vacancies predicted by
the model economy (conditional on observed unem­
ployment) as follows:

Federal Reserve Bank of Chicago

VA'(1-U,) + VA'U,
|XA'£+XW+VA'|

Figure 15 reports unemployment as a fraction of
total population and vacancies as a fraction of total
population between 2001:1 and 2011:2 (black dots),
as well as the vacancies predicted by equation 35. We
see that the steady state of the model with three labor
market states, constant transition rates, and a constant
A provides a good fit to the data through 2007:12. How­
ever, since the start of the latest recession there have
been large deviations from the stable Beveridge curve
predicted by the model. This indicates that the matching
efficiency parameter At must have experienced signif­
icant changes since then. Figure 16 shows that this has
been the case. It reports the matching efficiency levels
obtained by equation 33 between 2001:1 and 2011:2.
We see that before 2007:12, the matching efficiency
had been fairly stable, but it plummeted with the onset

157

FIGURE 17

FIGURE 18

Value of a job
(3 states, steady state, constant transition rates)

No np articip ation
(3 states, steady state, constant transition rates)

Note: The log of Jt is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

of the recession. Observe that the magnitude of the
fall is much larger than in figure 4 (p. 151).
Substituting equation 26 in equation 27 we get that

.<4° + V
36) Jt = kVt 1EN [l-U^ + ^U,

[kNE+1NU +1EN\
Figure 17 reports the path for J thus measured. We
see that it is very similar to that in figure 5 (p. 152), indi­
cating that having three labor market states does not sig­
nificantly affect the measurement of the value of a job.
Before decomposing the effects of the matching
efficiency parameter ,1, and the value of a job J, I would
like to point out that the results that follow should be
taken with a grain of salt. While I selected the paths for
and Jt to reproduce the observed path for Ut and F
(given the restrictions imposed by equations 33 and 36),
I made no attempt to reproduce the path for nonpartici­
pation TV, which according to the model is given by

37)

N, =

Figure 18 reports the path for nonparticipation in
U.S. data and the path for TV given by equation 37.

158

------ Model

------ Data

Note: Nonparticipation is normalized by total civilian
noninstitutional population (16 years and older).
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

We see that the model does a reasonable job at repro­
ducing the path for TV before 2007:12, but that it
largely overpredicts nonparticipation after that. This
suggests that either the assumption of constant transi­
tion rates or the assumption that the economy is al­
ways at the steady state of the model fails. I return to
this issue in the next section.
Similar to the previous section, I compute adjusted
unemployment U*, adjusted nonparticipation TV*, and
adjusted vacancies V* from equations 25-27 under the
assumption that At = ^2007 12, for every month /. That is,
I let the value of a job Jt evolve as in figure 17 (that is,
as in the data) but fix the matching productivity to the
value that it had at the start of the recession. In other
words, U*„ N*, and V* measure the unemployment, non­
participation, and vacancies that would have been ob­
tained had the matching productivity remained constant
at its December 2007 level but the value of a job had
evolved as observed.
The version of the model with constant transition
rates delivers the following results.9 Similar to figure 6
(p. 152), figure 19 indicates that starting in mid-2009,
the decline in matching productivity reported in fig­
ure 16 played an important role in generating a large
unemployment rate. This version of the model also
indicates that by February 2011, the unemployment
rate would have been 6.4 percent instead of 8.9 per­
cent had matching productivity remained constant at
its beginning-of-recession level. Figure 20 shows that

4Q/2011, Economic Perspectives

FIGURE 19

FIGURE 20

Effects on unemployment rate
(3 states, steady state, constant transition rates)

Effects on nonparticipation
(3 states, steady state, constant transition rates)

------ Variable A

------- Constant A

------ Variable A

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

the effects of matching productivity on nonparticipation
are very similar to those on unemployment. Finally, simi­
lar to figure 7 (p. 153), figure 21 shows that the effects
of matching efficiency on vacancies are negligible.

Variable transition rates
In this section, instead of assuming that transition
rates are constant, I allow them to fluctuate as in
figure 14. Given data on 1/ and all transition rates,
I compute matching efficiency as follows:

38) A=------ —
'
iitay'a
r t

1-G-

z7(i-c/,)+Cc/,

an expression obtained from substituting equation 26
in equation 25. Figure 22 reports the path for ,1, thus
obtained. We observe huge differences from figure 16.
Instead of relatively stable behavior before 2007:12
followed by a large drop, we observe significant vola­
tility throughout the sample period and a large increase

Federal Reserve Bank of Chicago

------- Constant A

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

after 2007:12. These differences indicate that the
predictions of the model rely critically on whether
transition rates are assumed to be constant or not.
The value of a job J is measured as

39) 7,=^ 4GX1-

zf(i-c/,)+z7c/,

and reported in figure 23. We see that the qualitative
behavior is similar to figure 17; however, the drop in
Jt after the start of the past recession is now some­
what larger.
Before turning to the decomposition of the different
effects, I revisit the issue of how well the model is able
to reproduce the path for nonparticipation, a path that
has not been targeted in the calibration. Figure 24 re­
ports the path for nonparticipation in U.S. data and the
path for Nt from equation 26. We see that contrary to
figure 18, the model now does a reasonable job at re­
producing the path for TV throughout the sample period.
In principle, this should be a reason for having more
confidence in the results obtained in this section.
In order to decompose the different effects, I com­
pute adjusted unemployment U*, adjusted nonparticipa­
tion N*, and adjusted vacancies V* from equations 2527 under the assumption that A=A200722, 7-4 = A, 2„07;12,
yEI
yEN
y Mi
y NE
,
A.^ : yEU
A 2007:12’ \

A.w =

/l2007:12’ h't

~ ^2007:12’

^2M7:i2 ’ f°r every month t. That is, I let the

159

FIGURE 21

FIGURE 22

Effects on vacancies
(3 states, steady state, constant transition rates)

Matching efficiency
(3 states, steady state, variable transition rates)

------ Variable A

-------- Constant A

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

Note: The log of At is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

FIGURE 24

No np articip ation
(steady state, variable transition rates)

------ Model

------ Data

Note: Nonparticipation is normalized by total civilian
noninstitutional population (16 years and older).
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

value of a job Jt evolve as in figure 23 (that is, as in the
data) but fix the matching productivity and the transi­
tion rates to the values that they had at the beginning
of the recession. In other words, U*, N, *, and V*
measure the unemployment, nonparticipation, and

160

vacancies that would have been obtained had the
matching productivity and transition rates remained
constant at their December 2007 levels but Jt had
evolved as it did.

4Q/2011, Economic Perspectives

FIGURE 25

FIGURE 26

Effects on unemployment rate
(3 states, steady state, variable transition rates)

Effects on vacancies
(3 states, steady state, variable transition rates)

------ Variable A, variable Zs
------ Constant A, constant Zs
Constant A, variable Zs
Sources: Author’s calculations based on data from the

Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

Also, I compute adjusted unemployment U**,
adjusted nonparticipation N**, and adjusted vacancies
V** from equations 25-27 under the assumption that
^=X,007.12, for every month t (but letting all X(s take
their actual values). That is, I let the value of a job J
evolve as in figure 23 and the transition rates evolve as
in figure 14 (p. 156), but I fix the matching productivity
to the value that it had at the beginning of the recession
A,007 12. In other words, U*'\ N“, and V** measure the
unemployment, nonparticipation, and vacancies that
would have been obtained had the matching productiv­
ity remained constant at its December 2007 level but
all transition rates and J had evolved as they did.
Figure 25 shows U* / (£’ + [/’) (“constant A,
constant X”), U" / (£** + I7,“) (“constant A, variable
X”), and Ut I (Et + Ut) (“variable A, variable X”). We
see that despite the large drop in the value of a job J
described in figure 23, U* / (£’ + [/’) increased only
moderately. In him, U" (E** + U**) increases by a
huge amount, indicating that the large increases in the
Xf'7 and X/'' observed after 2007:12 in figure 14 had
a large negative impact in the labor market. Actually,
the unemployment rate filmed to increase only as de­
scribed by Ut I (Et + 17) because of the large increase
in matching efficiency reported by figure 12 (p. 154).
Figure 26 shows that the increase in matching
productivity ,1, and the changes in transition rates played
a noticeable role in keeping vacancies relatively high.

Federal Reserve Bank of Chicago

------ Variable A, variable Zs
------ Constant A, constant Zs
Constant A, variable Zs
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

In turn, figure 27 shows that the increase in matching
productivity . played a crucial role in keeping nonpar­
ticipation (TV) relatively low, since the changes in tran­
sition rates would have increased it quite substantially.
Transitionary dynamics
The previous section showed that introducing vari­
able transition rates affects the results quite significantly
and that it allows one to keep track of the behavior of
nonparticipation much more closely. However, the analy­
sis of the previous section suffered two drawbacks. First,
while the calibration of the matching elasticity parameter
a assumed constant matching efficiency and constant
separation rates prior to 2007:12, we see from figures 14
(p. 156) and 22 that this is not quite the case. Second, the
analysis assumed that the steady state of the model could
be used to describe monthly data, while the large fluctua­
tions in transition rates observed in figure 14 (p. 156)
suggest that this may not be a good approximation. For
these reasons, in this section I take a more direct approach
to the calibration of the matching elasticity parameter
a and perform the analysis without imposing that the
model is always at its steady state. This allows me to
evaluate to what extent this affects the results.
Observe from equation 16 that
40) In (yj = In (H,) + a In

.

161

FIGURE 27

FIGURE 28

Effects on nonparticipation
(3 states, steady state, variable transition rates)

Matching function

------ Variable A, variable as
------ Constant A, constant Xs
Constant A, variable as
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

In what follows, I identify Mt with the total num­
ber of workers that transition from unemployment to
employment between months t and t + 1, as reported
by the CPS. Figure 28 plots ln(-y-) against ln(y-) for
the whole sample period. We see a strong linear relation,
suggesting that equation 40 provides a good description
of the data with a relatively constant At. Fitting equa­
tion 40 using OLS (ordinary least squares) over the
period 2001:1-2007:12 gives an estimate of a = 0.69.
Given this estimated value of a, the path for the
matching efficiency parameter ,1, implied by equation 40
is reported in figure 29. We see that this path is com­
pletely different from that in figure 22. The matching
productivity is much less variable and contrary to
figure 22, displays a large drop after the start of the last
recession, reaching by the end of the sample period a
value 12 percent lower than in 2007:12. There is no doubt
that measuring At directly from the matching function
in equation 16 gives a very different picture from mea­
suring it from the steady states of the model economy.
Figure 30 reports the value of a job obtained
from equation 27 using the matching efficiencies
obtained from equation 40 and reported in figure 29.
The figure is very similar to figure 23, again indicating
that the path for the value of a job is robust to the dif­
ferent ways of measuring it.
I now him to evaluating the relative contributions
of the value of filled jobs J, the matching efficiency

162

u/v
Note: M/V is total matches per vacancy; UAZ is the
unemployment-vacancy ratio.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

parameter
and the transition rates Xfy. X™, Xfy,
X”; and Xfy to unemployment dynamics since the be­
ginning of the recession. For this purpose, I proceed
as before and find a sequence E*„ U*, N*, and fy* that
satisfies equations 17, 18, 19, and 21 under the assumpUN .
yUN
y EU __ y EU
tionthat^ = 4007:12.X
2007:12’
^2007:12’
yEN _ ^E
yNE _ yNE
'\NU_ yNU
’ \
^2007:12 ’
^21
— A2
v2007:12 ’
for every month t. That is, I let the value of a job Jt
evolve as in figure 30 but fix the matching productivity
and all transition rates to the values that they had in
2007:12 (that is, at the beginning of the recession).
Similarly, as before, E*. U*. N*, and V*, describe the
employment, unemployment, nonparticipation, and
vacancies levels that would have obtained if the value
of a job had been the only variable changing over time.
Also, I compute the £,**,
A,**, and fy" that
satisfy equations 17, 18, 19, and 21 under the assump­
tion that At = ^2OO7 p, for every month /. That is, E" U**,

N*", and V** describe the employment, unemployment,
nonparticipation, and vacancies levels that would have
been obtained if the matching productivity parameter
had remained constant at its December 2007 level,
while all other variables (that is, J and all the X values)
had changed the way they did.
Figure 31 reports the paths for U* / (£’ + U*)
(“constant A, constant X”), U** / (£,“ + A ") (“constant
A, variable X”), and Ut / (Et + Ut ) (“variable A.
variable X”). From the U* / (£* + £/,*) path, we see

4Q/2011, Economic Perspectives

FIGURE 29

FIGURE 30

Matching efficiency
(3 states, transitionary dynamics)

Value of a job
(3 states, transitionary dynamics)

Note: The log of Af is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

that changes in the value of a job J played a very minor
role in unemployment dynamics: The red line is roughly
flat. From comparing the path for U** / (£” + U**)
with the path for U* / (£* + U*), I conclude that
changes in the transition rates X played a crucial role
in unemployment dynamics: The gray line is widely
different from the red line. In fact, we see that changes
in the transition rates X accounted for most of the un­
employment dynamics observed since the recession:
The black line is very close to the gray line, indicating
that the matching productivity played a minor role in
the observed unemployment rate dynamics.
In turn, figure 32 reports the path for N* (“con­
stant^, constant X”), TV" (“constant A, variable X”)
and Nt (“variable A, variable X”). From the N* path,
we see that far from accounting for the observed in­
crease in nonparticipation, changes in the value of a
job J would have accounted for a decrease in nonpar­
ticipation. The bulk of the increase in nonparticipation
is accounted for by changes in transition rates, since
changes in the matching productivity played a relatively
minor role: The gray line is very close to the black line).
Finally, figure 33 shows that none of the changes
in the matching efficiency parameter Af or in the tran­
sition probabilities were important determinants of
vacancies dynamics: The path for vacancies was mainly
determined by Jr

Federal Reserve Bank of Chicago

Note: The log of Jt is normalized to zero at the start of the
past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

Nonparticipants compete for vacancies
This section describes and uses the most satisfac­
tory specification of the model. Thus, it provides the
main results of the paper. Observe that the model used
in the previous section had three labor market states but
only unemployed workers were inputs to the matching
function: Nonparticipants made transitions to employ­
ment but without going through the matching function.
I view this feature as a weakness of the previous speci­
fication of the model. The workers transitioning from
nonparticipation to employment must be competing
for the same vacancies as the workers transitioning from
unemployment to employment and should therefore en­
ter the matching function in a similar way. This sec­
tion addresses this issue by modifying the matching
function of the previous section accordingly. Introduc­
ing a more satisfactory specification for the matching
function allows me to obtain better measurements of
the matching efficiency.
The matching function is now described as follows:

41)
where Mt is the total number of matches, t/ is unem­
ployment, TV is nonparticipation, Vt is vacancies, At
is the matching efficiency, 0 < y < 1, and 0 < a < 1.
Observe that y, can be interpreted as the fraction of
the total number of workers who report they are non­
participants but search for jobs anyway. Alternatively,

163

FIGURE 31

FIGURE 32

Effects on unemployment
(3 states, transitionary dynamics)

Effects on nonparticipation
(3 states, transitionary dynamics)

------ Variable A, variable Zs
------ Constant A, constant Zs
Constant A, variable Zs

------ Variable A, variable Zs
------ Constant A, constant Zs
Constant A, variable Zs

Sources: Author’s calculations based on data from the

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

V, can be interpreted as the search intensity of non­
participant workers.
The transition rate from unemployment into em­
ployment
is given by:
U,

\_Al(Ul+WlNXVll-a

ui+^n,

’

since a fraction (o-,+v,w,) of the total matches Mt is

formed with unemployed workers. Similarly, the
transition rate from nonparticipation into employment
k, is given by:

UJ

1

\

4([/,+vx)r-“
= V,-------------------------- ,
V+vX
since a fraction I V'N, inf the total matches M is
tr/.+vW, I
'
formed with nonparticipant workers.
The transition rate from employment to unemploy­
ment
the transition rate from employment to non­
participation X™, the transition rate from unemployment

164

to nonparticipation X™, and the transition rate from
nonparticipation to unemployment A-4 are assumed
to be exogenous to the model.
The evolution of workers across labor market
states is then given by the following equations:

44) E,+l=E, + M,-(^u + Xen)E,,

45) U,+l =U, + XeuE, + Xn,uN, - (X® +

46) 7V,+1 = N, + XenE, + T^U, -

,

+ 7^U}N,.

Equation 44 states that next-period employment
is equal to current employment, plus all new matches,
minus total separations (either to unemployment or non­
participation). Equation 45 states that next-period unem­
ployment is equal to current unemployment, plus all
transitions into unemployment (either from employment
or nonparticipation), minus all transitions out of un­
employment (either to employment or nonparticipation).
Equation 46 states that next-period nonparticipation is
equal to current nonparticipation, plus all transitions
into nonparticipation (either from employment or
unemployment), minus all transitions out of nonpar­
ticipation (either to employment or unemployment).

4Q/2011, Economic Perspectives

FIGURE 33

FIGURE 34

Effects on vacancies
(3 states, transitionary dynamics)

Search intensity of nonparticipants

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

------ Variable A, variable as
------ Constant A, constant Xs
Constant A, variable as
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

51)
The free-entry condition is given by

47) k =

k

Given the total number of workers unemployed
at date zero Uo and the total number of workers that
are nonparticipants at date zero No, the model gener­
ates an endogenous path for {Af,, , UM, NM j"Q.

J,

since from the point of view of a firm, the probability
of filling a vacancy is now equal to MJ Vf.
Observe that using equations 41, 42, and 43, we
can rewrite equations 44-47 as follows:

Results
From equations 42 and 43, we have that the search
intensity of nonparticipants can be measured as

48)7?,+1=7?, + 4(G,+vA)r’“-(^'

49) Ul+l = U,+

Figure 34 shows that the fraction of nonparticipants
that search has increased quite substantially since the
start of the latest recession.
From equation 41, we have that

lJuE, + Xj'N,
u,

53) In

j = In (4) + a In

•

50)Nl+l= N, + ^ E,+ KjU.
Z4(G,+vx)Xa
N,

Federal Reserve Bank of Chicago

In what follows, 1 identify Mt with the total num­
ber of workers that transition from unemployment
into employment between months t and t + 1, plus the
total number of workers that transition from nonpar­
ticipation into employment between those same months,

165

FIGURE 35

FIGURE 36

Matching function

Matching efficiency (transitionary dynamics,
N in matching function)

(u+v*N)V
Note: The log of At is normalized to zero at the start
of the past recession.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

Note: MA/ is total matches per vacancy; (u+y*N)/V is total
searchers per vacancy.
Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

as reported by the CPS. Figure 35 plots ln(-y-) against

In^.+yA j por qie w|10|e sample period. We see a
strong linear relation, suggesting that equation 53 pro­
vides a good description of the data with a relatively
constant Ar Fitting equation 53 using OLS between
2001:1 and 2007:12 gives an estimate of a = 0.62.
Given this estimated value of a, the path for the
matching efficiency parameter At implied by equation
53 is reported in figure 36. We see that the path is not
very different from figure 29 (p. 163). In him, the
value of a job, which is measured as

is reported in figure 37. We also see that its path is
not very different from figure 30 (p. 163).
In order to decompose the effects of the different
variables on labor market dynamics, I find the E*, U*,
N*, and V* that satisfy equations 48-51 under the as­
sumption that At = A2mTl2, <|/( = y2007:12’
= ^2007:12 ’
yEU _
~

yEU
yEN _ yEN
, y NU _ yNU
A2007:12’
~ A2007:12’ dnC*
~ ^2007:12’

p

every month /. That is, E*, U*, N*, and V* describe
the employment, unemployment, nonparticipation,
and vacancies levels that would have been obtained if
./had changed the way it did but all other variables
had remained constant at their December 2007 levels.

166

Also, I compute the E**, U*\ N**, and V** that
satisfy equations 48-51 under the assumption that
A= AoO7:12 alld V,= V2007:12’ for
111011111 A Tllat 1S’
E*\ U*\ N*\ and V** describe the employment, un­
employment, nonparticipation, and vacancies levels
that would have been obtained if./ and all X values had
changed the way they did but At and xg, had remained
constant at their December 2007 levels.
Similarly, I compute the E***, U**\ N**\ and V***
that satisfy equations 48-51 under the assumption
that x|/(= x|/2007 12’101 everY month t. That is,
(/"’,

N**\ and V*** describe the employment, unemploy­
ment, nonparticipation, and vacancies levels that
would have been obtained if At, Jt, and all X values
had changed the way they did but xg, had remained
constant at its December 2007 level.
Figure 38 reports the paths for U* I [E‘ + U*)
(“constant A, X, x|/”), £/,” / (£,” + (/") (“constant A,
xg”), (/"’ / (E***+ U*") (“constant x|/”), and (/ / (£ +
(/) (“everything variable”). From the U* I (E* + Uj
path, we see that changes in the value of a job ./played
a very minor role in unemployment dynamics: From
an initial unemployment rate of 4.9 percent, changes
in are only able to generate a peak unemployment
rate of 6.4 percent. From comparing the path for
U" I (£," + [/") with the path for [/,’ / (£/ +[/,’),
we see that changes in the transition rates X played
a crucial role in generating the large and persistent

4Q/2011, Economic Perspectives

FIGURE 38

FIGURE 37

Effects on unemployment
(transitionary dynamics, N in matching function)

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

increase in the unemployment rate. In fact, the path
for £/,’ ’/(£," + (7,") is very similar to Ut / (Et + Ut).
Comparing [/"’ /
+ [/’’’) with U** I [E** + V**),
we see that the large drop in matching efficiency shown
in figure 36 had a nontrivial role in the increase in the
unemployment rate: The largest difference between
[/"• / (£,"•+ [/,"•) and U" I (E" + U”} is 1.5 per­
cent. Comparing u*** / [E'"+U'"} with UJ (Ef + C7),
we see that the increase in the search intensity of non­
participants shown in figure 34 roughly offsets the
effects of the fall in matching efficiency.
Figure 39 reports the paths for N* (“constant A,
X, \|/”), N** (“constant v”), TV"’ (“constant y”) and
Nt (“everything variable”). We see that the value of a
job had a significant effect on nonparticipation: The
gray line first declines and then increases quite rapidly.
The changes in transition rates first increased nonpar­
ticipation but then lowered it: The red line is initially
above the gray line, but it crosses it in mid-2009. The
drop in matching efficiency of figure 36 has the effect
of increasing nonparticipation: The light red line is
significantly higher than the red line. However, the large
increase in the search intensity of nonparticipants
shown in figure 34 had a large effect on reducing
nonparticipation: The black line is much lower than
the light red line.

Federal Reserve Bank of Chicago

Lastly, figure 40 shows that the only important
determinant of vacancy dynamics was the value of
a job J (all other lines are quite similar to the gray).

Conclusion
This article has explored different approaches to
measuring matching efficiency and assessing its impli­
cations for labor market dynamics since the start of
the past recession. In particular, I evaluated the impor­
tance of allowing for a third labor market state, allow­
ing for variable transition rates, considering explicit
transitionary dynamics, and allowing nonparticipants
to enter the matching function. I find that the results
are quite sensitive to the different specifications.
In the preferred scenario (that is, the case with
three labor market states, variable transition rates,
nonparticipants entering the matching function, and
explicit transitionary dynamics), I obtained the following
findings. First, the matching efficiency parameter is
quite volatile throughout the sample period. Second,
the matching efficiency has been drifting down since
the start of the recession. Third, the value of filled
jobs plummeted between 2007:12 and 2009:6 but has
recovered quite significantly since then. Fourth, condi­
tional on the observed paths for the value of a job and
all transition rates, the drop in matching efficiency since
the start of the recession has had only a moderate im-

167

FIGURE 39

FIGURE 40

Effects on nonparticipation
(transitionary dynamics, N in matching function)

Effects on vacancies
(transitionary dynamics, N in matching function)

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

pact on the unemployment rate. Fifth, the large effects
on unemployment rate dynamics arise from changes
in the transition rates. Sixth, the matching efficiency,
the value of a job, the transition rates, and the search
intensity of nonparticipants all have significant effects
on the dynamics of nonparticipation.
The analysis performed in this article decomposed
the observed growth in unemployment and nonpartic­
ipation into contributions from changes in matching
efficiency, the value of a job, the search intensity of
nonparticipants, and the transition rates across different
labor market states. This decomposition was done very
much in the spirit of standard growth accounting ex­
ercises, in which GDP growth is decomposed into growth
contributions from total factor productivity, capital,
and labor. Interpreted as a growth accounting exercise,
the results in this paper should be considered as extremely
informative. However, care should be exercised in
providing a counterfactual interpretation to the results.

168

Sources: Author’s calculations based on data from the
Bureau of Labor Statistics’ Current Population Survey
and Job Openings and Labor Turnover Survey.

The reason is that if the matching efficiency had stayed
constant at its 2007:12 level (instead of dropping as it
actually did), this would have affected the value of a
job and the transition rates across labor market states,
but these secondary effects have not been considered
in the analysis. That is, the contributions to labor market
dynamics of the value of a job, the productivity of the
matching function, the different transition rates, and
the search intensity of nonparticipants have been cal­
culated as not affecting the other variables.10 To eval­
uate counterfactuals such as this, the explicit economic
decisions and wage determination process that this paper
has abstracted from would have to be incorporated and
the equilibrium of such full-blown models would have
to be analyzed. Of course, the counterfactual results
would depend on how those modeling choices are made.
Another caveat to the analysis is that it has not
incorporated job-to-job transitions. Since the workers
making these transitions are competing for the same
pool of vacancies as unemployed and nonparticipant
workers, their behavior may affect the measurement
of matching efficiency.

4Q/2011, Economic Perspectives

NOTES
’The negative relation between unemployment and vacancies is not
exclusive to the U.S.: It is present in a number of countries. See
Petrongolo and Pissarides (2001) for an empirical survey of the
Beveridge curve and the matching function.
2Strictly speaking, what these papers implicitly assume is that the
transitionary dynamics of the model economy are extremely fast.
Under this assumption, they use the steady state of the model to
analyze data, even when the matching function and transition rates
change over short periods.

3The preferred specification is the one with the least restrictive and/
or most appealing assumptions.
4We could choose alternative approaches. For instance, we could
take the average of the ratios in equation 11 for every pair of months
The problem with this approach is that measurement errors
would be severely amplified for pairs of months with similar unemployment/vacancy ratios. Another approach would be to take the average
of the ratios in equation 11 for pairs of months with sufficiently large
differences in unemployment/vacancy ratios. This approach would
lead to a similar estimate for a as the approach chosen here.

6This is a large number. However, Jt must be interpreted as the
value of creating an additional job and not as the average value of
all existing jobs. Obtaining such a large drop in the value of a mar­
ginal job should not be surprising, given the severity of the recession
experienced by the U.S. at the time.
7See the conclusions for a further discussion of how to interpret
the results of this article.
8See note 2 for a discussion of this estimation approach.

9These results should not be taken seriously. In what follows, I show
that the large drop in matching efficiency obtained in figure 16 is
an artifact of the constant transition rates.
10Observe that similar caution must be used in providing a counterfactual interpretation to standard GDP growth accounting exercises.
If total factor productivity had stayed constant during the last 50 years
(instead of growing at its actual rate), this would have affected capital
accumulation and labor supply. However, these secondary effects
are not taken into account when reporting the contribution to growth
of total factor productivity.

5For Uf, I use the unemployment rate at month t from the Bureau
of Labor Statistics’ (BLS) Current Population Survey (CPS). For
V, I use the average vacancies reported by the BLS’s Job
Openings and Labor Turnover Survey (JOLTS) in months t and
t—i, divided by the size of the labor force reported by the CPS in
month 1.1 average the vacancies reported by JOLTS because they
correspond to the number of vacancies at the end of the month,
while the CPS data roughly correspond to observations in the mid­
dle of the month.

REFERENCES

Barlevy, Gadi, 2011, “Evaluating the role of labor
market mismatch in rising unemployment,” Economic
Perspectives, Federal Reserve Bank of Chicago, Vol. 35,
Third Quarter, pp. 82-96.

Mortensen, Dale T., and Christopher A. Pissarides,
1994, “Job creation and job destruction in the theory
of unemployment,” Review ofEconomic Studies, Vol. 61,
No. 3, July, pp. 397-415.

Bamichon, Regis, and Andrew Figura, 2010, “What
drives movements in the unemployment rate? A decom­
position of the Beveridge curve,” Finance and Economics
Discussion Series, Board of Governors of the Federal
Reserve System, working paper, No. 2010-48.

Petrongolo, Barbara, and Christopher A. Pissarides,
2001, “Looking into the black box: A survey of the
matching function,” Journal ofEconomic Literature,
Vol. 39, No. 2, June, pp. 390-431.

Kocherlakota, Narayana, 2010, “Back inside the
FOMC,” speech in Missoula, MT, September 8.

Federal Reserve Bank of Chicago

Shimer, Robert, 2005, “The cyclicality of hires, sep­
arations, and job-to-job transitions,” Review, Federal
Reserve Bank of St. Louis, Vol. 87, No. 4, July/August,
pp. 493-507.

169