Full text of Business Review (Federal Reserve Bank of Philadelphia) : First Quarter 2008

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

The Economic Logic of a Fresh Start
BY SATYAJIT CHATTERJEE

A

debtor’s right to have his or her debts
dismissed or discharged via a bankruptcy
proceeding is referred to as the law’s “fresh
start” provision. Fresh start has been — and
continues to be — a controversial feature of the U.S.
bankruptcy law. Lately, the law has come under scrutiny
because of the dramatic rise in personal bankruptcy
filings over the past 25 years. In this article, Satyajit
Chatterjee explains the economic logic underlying the
fresh start concept. He also argues that this logic can
explain why opposition to a discharge policy has waxed
and waned over time.

U.S. law gives debtors the right
to petition a bankruptcy court and
ask to be released from their financial
obligations to creditors. For reasons explained in this article, a debtor’s right
to have his or her debts dismissed or
discharged via a bankruptcy proceeding is referred to as the law’s “fresh
start” provision. Fresh start has been
— and continues to be — a controver-

Satyajit Chatterjee
is a senior
economic advisor
and economist
in the Research
Department of
the Philadelphia
Fed. This article
is available free of
charge at www.philadelphiafed.org/econ/br/.
www.philadelphiafed.org

sial feature of U.S. bankruptcy law. Of
late, the law has come under scrutiny
because of the dramatic rise in personal bankruptcy filings in the last 25
years. In 2005, roughly one out of every
75 U.S. households took advantage of
the fresh start provision; in 1980, only
one out of 375 households did.
The need to deal in some fashion
with people who cannot (or will not)
repay their debts was felt from the
earliest days of European settlement
in New England. By and large, the
colonists dealt harshly with defaulters
and were quite hostile to the idea of
the discharge of personal debts. But
this hostility appears to have waned by
the late 19th century, when Congress
enacted a federal bankruptcy law with
a fresh start provision. Unlike earlier
attempts at legalizing discharge, the
1898 law proved to be more perma-

nent, although later laws modified
many of its provisions. The latest
turn in this gradual evolution is the
Bankruptcy Abuse Prevention and
Consumer Protection Act of 2005, a
law that significantly curtails a debtor’s
right to a fresh start.
The objective of this article is
twofold. The first is to explain the
economic logic underlying the fresh
start concept. For there is an economic
logic – one that gets ignored when
advocates portray fresh start as a form
of protection against rapacious creditors or when opponents portray it as a
refuge for the morally bankrupt. The
economic logic puts debtors and creditors on an equal footing but argues
that, under certain circumstances,
society as a whole is better off when
discharge is permitted. The second
objective is to argue that this logic can
explain why opposition to a discharge
policy has waxed and waned over time.
Why did the colonists view discharge
with hostility? Why did this opposition wane by the turn of the previous
century? Why has opposition to a fresh
start reappeared now? The economic
logic of a fresh start suggests that these
shifts in attitude reflect an evolving
tradeoff between the economic costs
and benefits of a fresh start.
WHAT IS A FRESH START?
U.S. bankruptcy law permits an
individual debtor to be released from
his or her financial obligations to current creditors. The main requirement
for obtaining this release, or discharge,
of debt is that the debtor must surrender to creditors whatever property
he or she has at the time the discharge
is sought. By surrendering all existing
Business Review Q1 2008 1

property to creditors, a debtor who
is unable or unwilling to repay all of
his or her debt can obtain permanent
protection from collection efforts by
current creditors.
There are some exceptions to this
general provision of the law. On the liability side, not all financial obligations
are eligible for discharge. Examples of
nondischargeable obligations include
student loans and judgments incurred
in judicial court cases. On the asset
side, debtors are not required to surrender certain assets to creditors. For
instance, in Florida and Texas, home
equity is exempt from seizure by creditors. In addition, any property essential
to a person’s livelihood or dignity (such
as tools used by a carpenter to do his
job, ordinary clothes, and so forth) is
generally exempt from seizure by creditors as well.1
The term fresh start is used to
describe this provision of bankruptcy
law because it neatly encapsulates the
spirit of an oft-cited justification for
discharge given in a 1934 ruling by the
U.S. Supreme Court. According to the
court, discharge of debt “gives to the
honest but unfortunate debtor who
surrenders for distribution the property
which he owns at the time of bankruptcy a new opportunity in life and
a clear field for future effort, unhampered by the pressure and discouragement of preexisting debt.”2
To understand the economic logic
of the fresh start provision, we should
view the need for this provision from
two closely related, but distinct, per-

1

Other exceptions exist to prevent abuse of the
provision. For instance, shifting one’s wealth
into nonexempt assets shortly before filing for
bankruptcy is viewed as an abuse and will make
the debtor ineligible for a fresh start. Similarly,
the right to a discharge is not available to a
debtor who has used this provision in the previous six years — so “serial” discharge is viewed
as an abuse and is not permitted.

2

Local Loan Co. vs. Hunt U.S. 234, 244 (1934).

2 Q1 2008 Business Review

spectives. The first is the situation as
it relates to a debtor and creditor after
debt has been incurred (what economists call the ex-post perspective). The
second perspective is the situation as it
relates to potential debtors and creditors before any debt is incurred (what
economists call the ex-ante perspective). The desirability of a fresh start

In most modern
societies, contract
law gives creditors
the right to seize the
property of a debtor
who does not repay
his or her debts.
can be argued from either perspective,
but the nature of the argument is different in the two cases and therefore
best discussed separately. As we will
see, both perspectives are implicit in
the famous Supreme Court justification for a fresh start quoted earlier.3
FRESH START FROM THE
POST-DEBT PERSPECTIVE
To understand the post-debt logic
for having a fresh start provision, we

3

The language of the Supreme Court ruling
points to a set of core issues that any discussion of fresh start should cover. A complete and
thorough discussion of all aspects of fresh start
would be well beyond the scope of this article.
The last two chapters of Thomas Jackson’s book
and the article by Michelle White provide more
details about the costs and benefits of fresh
start. The article by Michel Robe, Eva-Maria
Steiger, and Pierre-Armand Michel provides
further background on the nature of fresh start.
Also, I do not discuss a second form of bankruptcy — called Chapter 13 — in this article. In
a Chapter 13 bankruptcy, the debtor is allowed
to keep his or her assets in return for agreeing to
a new repayment schedule that involves a partial discharge of debt. Historically, only a third
of bankruptcy filings in the U.S. have been
Chapter 13 filings; the rest are of the fresh start
variety. For a nice discussion of Chapter 13
bankruptcy, see the article by Wenli Li.

need to be clear about what transpires
in its absence. In most modern societies, contract law gives creditors the
right to seize the property of a debtor
who does not repay his or her debts. If
the debtor lacks sufficient property, the
law permits creditors to garnish the
debtor’s earnings in excess of what is
needed by the debtor to meet nondiscretionary expenses. Importantly,
these creditors’ rights continue to be
in force as long as there is some unmet
financial obligation. It is against these
creditors’ rights that the fresh start
provision extends protection.
Since this article is about the
economic logic of a fresh start, and the
logic can be somewhat subtle, it helps
to talk about the issues by using an
example. I will introduce the example
in this section and progressively extend
it in the following two sections. In this
section, I use the example to make
clear one reason why unrestricted
creditors’ rights can be bad for society.
Consider the case of a debtor,
whom we shall call D, who has borrowed from a creditor, C, and her
payment on the debt is now due.
Assume also that D has no assets and
her obligation to C amounts to $5000.
Further assume that D’s monthly takehome pay from her regular full-time
job totals $2000 and her monthly
nondiscretionary expenses are $1800.
Since D does not have the funds to
pay off her obligation, she is in default.
According to the law, C has the right
to seize $200 from D each month for
the next 25 months in order to recover
what is owed to him.4
However, matters may unfold differently. Imagine that D has the option
to reduce her hours at her job so as

4
Actually, the law would permit C to recover
more than $5000 because recovery takes time
and C loses interest on the part of the debt yet
to be repaid. Taking compensation for lost interest into account would require D to pay $200
each month for more than 25 months.

www.philadelphiafed.org

to make her monthly take-home pay
exactly $1800. If D chooses to do this,
C can no longer seize any income from
D because D does not have any discretionary income. D may prefer this “less
work” option to the option of working
full-time and having her additional
$200 a month “taxed” away by C for
the next two years. Of course, this less
work option will keep D under the
threat of garnishment indefinitely, but
the benefits of working full-time and
eventually becoming debt-free come
too far in the future for D to make the
extra effort.
This outcome is inefficient because D clearly values the extra $200 a
month more than the effort required to
earn it — which is why she was working full-time in the first place. But now
C’s right as a creditor stops D from
doing so. Society’s loss is the $200 D
could earn, net of her efforts to earn
it. One might wonder why the loss to
society does not include the $5000 loss
to C, but from a societal point of view,
C’s loss is exactly offset by D’s gain.5
In this example C’s rights as a
creditor force D to obtain her discretionary income in a form that C
cannot seize, that is, in the form of
leisure. One can also imagine D’s being induced to engage in activities that
allow her to hide her earnings from C,
for instance, doing informal work for
friends and relatives or engaging in illegal activities. If these alternatives are
inferior (from society’s point of view)
to D’s working full-time at her regular job, the inefficiency remains and
is perhaps compounded. In general,
whenever a debtor has the option to
substitute nonseizable forms of income

5

This is the sense in which economic logic
puts lenders and borrowers on an equal footing:
Any gain or loss to the lender that comes at
the expense of an equivalent loss or gain to the
borrower is viewed as being neutral with regard
to gains or losses to society as a whole.

www.philadelphiafed.org

for regular earnings, unbridled creditors’ rights can cause a costly distortion of work effort. In such situations,
an efficiency case can be made for
constraining a creditor’s rights. Indeed,
if D is given the right of discharge, she
will avail herself of it and continue
working full-time at her regular job
and the loss to society will be avoided.6 This is the economic justification
for discharge implicit in the Supreme
Court’s statement that discharge gives
a debtor “new opportunity in life and
a clear field for future effort, unhampered by the pressure and discouragement of preexisting debt.”
But this post-debt justification
ignores the fact that when discharge
is permitted, a creditor’s incentive to
lend is seriously blunted. Is not the
reduction in lending that is sure to
result an important loss from a societal
point of view? This is an important
objection, and it brings us to the issue
of whether discharge can be justified
from a pre-debt perspective. As we will
see, discharge can be justified from a
pre-debt perspective, but the argument
in favor of it must be amended in an
important way.
RETHINKING THE (POSTDEBT) LOGIC OF A FRESH
START FROM A PRE-DEBT
PERSPECTIVE
We will continue with our example of creditor C and debtor D, but

6
It could be argued that the inefficiency could
be avoided without permitting discharge if C
and D negotiated a better outcome. After all,
C must understand that if he insists on getting
all of his $5000 back, he will get nothing. Given
this, he might offer to partially forgive D’s debt,
and D might well accept such an offer and go
back to working full-time. However, it’s likely
that C is dealing not only with D but with many
other debtors, and C must be cognizant of the
fact that forgiving D’s debt might embolden
other debtors to demand a similar consideration. These external effects might prevent an
efficiency-restoring renegotiation between C
and D.

we will now focus on their situation
as they contemplate entering into a
lender-borrower relationship. To do so
requires extending the example. The
extended example explains that if the
lender understands the circumstance
under which his creditor will default,
and he can act to avoid that circumstance, then fresh start serves no useful
purpose. In fact, instituting a policy
of fresh start in such a situation could
make matters worse!
We will assume that both C and
D are forward-looking: Each person
fully understands how the other will
act in the future if a loan is made.
We will continue to assume that D’s
circumstances are exactly as before:
She has a full-time job earning $2000
a month with the option of working
fewer hours; she has no assets; and her
monthly nondiscretionary expenses
are $1800. To keep matters simple, we
will also assume that D does not plan
to save any portion of her resources for
the foreseeable future. This means that
she will spend any loan granted to her
and she will not have any property a
creditor can grab in the future. Finally,
we will assume that the best alternative use of C’s funds is a risk-free investment that will earn him 5 percent
per year.
Consider first the case where
discharge is not permitted. From our
previous discussion, we know that if D
borrowed a lot of money, she will not
pay it back. Being forward-looking,
C understands this fact and will not
lend a lot of money to D. Indeed, the
most C would be willing to lend is the
present value of the longest stream of
$200 monthly payments D can handle.
Let’s assume that the longest stream is
12 months; that is, if D is faced with
the prospect of making 13 or more
monthly payments of $200 each, she
will stop making payments and take
the less work option. But if she needs
to make fewer than 13 monthly pay-

Business Review Q1 2008 3

ments, she will continue making them.
In this case C can lend, at most, $2238
(rounded) to D without losing money
on the deal. This amount of $2238 is
simply the present discounted value
of 12 monthly receipts of $200 each
when the annual interest rate is 5 percent. This is the key difference when
matters are viewed from the pre-debt
perspective: If C is aware of the level
of debt beyond which D will default,
he will rationally lend less than that
amount and default will not occur.
If foresight can prevent default, the
post-debt rationale for discharge does
not apply.
In fact, having a discharge policy
in place can actually make matters
worse. Consider the above contract
that requires D to pay $200 a month
to C for 12 months. Would D have the
incentive to adhere to this contract if
she has the option to invoke discharge?
The answer depends on the pecuniary and psychological costs of invoking discharge. The pecuniary costs of
discharge include the out-of-pocket
expenses of going to court, the cost of
not being able to borrow again for an
extended period of time, and the cost
of being barred from certain types of
employment following default.7 The
psychological cost might stem from a
feeling of shame in having failed to
meet one’s obligations. If these costs
are high enough, D will adhere to the
contract.
But it is also possible that these
costs are too low to prevent D from
invoking discharge. If this is the
case — and C is aware that D’s costs
of discharge are low — the amount C
would be willing to lend to D will be

less than $2238. How much less? For
concreteness, suppose that the costs
of discharge are such that D would
adhere to the contract if she had six
or fewer monthly payments of $200 to
make. Then, the most that C would
be willing to lend would be $1183
(rounded). This amount is the present
discounted value of six monthly pay-

The risk of not being able to repay gives a
fresh start a benefit that may counter its costs.
ments of $200 each when the annual
interest rate is 5 percent.
Thus, a discharge policy may have
the effect of reducing D’s credit limit
from $2238 to $1183. However, it is
entirely possible that D would actually prefer to borrow more than $1183
and pay it back but the existence of a
discharge policy makes it impossible for
her to do so. The bottom line is that
a discharge policy can have adverse
effects on borrowers by making it too
easy for them to default and consequently make creditors less willing to
lend.
This objection to a discharge
policy is implicitly recognized and
countered in the Supreme Court ruling
quoted earlier. Recall that the ruling
made reference to the “honest but unfortunate debtor” in making the case
for discharge. Evidently, the court was
drawing attention to the fact that misfortune plays a role when people don’t
pay back their debts. And indeed, as
argued in the next section, the risk of
bad outcomes can provide a justification for a discharge policy.

7

Failure to meet debt obligations is recorded
in a person’s credit history. This history is
available to potential creditors and employers.
A tarnished credit history typically leads to
difficulties in obtaining new loans and could
lead to difficulty in obtaining certain types of
employment.

4 Q1 2008 Business Review

extend our example one more time.
The point of the extended example is
that the risk of not being able to repay
gives a fresh start a benefit that may
counter its costs. The benefit is that in
the event that the debtor is unable to
repay, she can invoke discharge and be
relieved of the burden of her debt. But
the debtor may choose to invoke dis-

THE ROLE OF RISK IN
RESTORING THE (PRE-DEBT)
LOGIC OF A FRESH START
To explain the role of risk, we will

charge even when she has the capacity
to repay. To avoid this outcome, the
creditor must reduce the amount lent,
which is the cost.
Imagine that in the month immediately after D takes out her loan there
is a small chance that her discretionary
income will fall permanently to $100.
We will assume that D is contemplating entering into a contract wherein
she promises to pay $200 each month
for some (to be determined) set of
months. The question we want to
answer is: How much would C be willing to lend to D, recognizing that D’s
discretionary income may fall to $100?
Let’s answer the question first
for the case where discharge is not
permitted. If D’s discretionary income
remains $200 in the first month of the
loan, she will be in a position to make
12 monthly payments of $200 (which
is the maximum number of months she
can promise, given that she can choose
the less work option and not pay anything). But if her discretionary income
falls to $100 in the first month, she
will be in default. At that point, C will
have the right to “tax” away all of D’s
discretionary income until all obligations are met. Faced with this “tax,” D
will choose the less work option (that
is, reduce her discretionary income to

www.philadelphiafed.org

zero by working fewer hours) and never
repay anything.8 Knowing this, C will
be willing to lend somewhat less than
$2238 against D’s promise to pay $200
each month for the next 12 months.
By offering to lend somewhat less than
$2238, C will get more than a 5 percent rate of return on his investment
in the event D actually pays back. This
return above the opportunity cost of
his funds (which by assumption is 5
percent) is C’s compensation for taking
on the risk that D will default on the
loan.
Now let’s answer the question
assuming that discharge is permitted.
In this case, D will adhere to her loan
contract as long as she has six or fewer
monthly payments of $200 to make
(which is the maximum number of
months she can promise to pay, given
that she has the option to invoke
discharge and walk away from her
debt) and her discretionary income is
$200. If her discretionary income falls
to $100 in the first month of the loan,
she will invoke her right to discharge
and walk away from her debt. Again,
anticipating this, C will be willing to
lend somewhat less than $1183 against
D’s promise to pay $200 each month
for the next six months because there
is the small chance that D will not
make any payments at all.
The bottom line is that without
the possibility of discharge, the “honest but unfortunate” debtor runs the
risk of being condemned indefinitely to
life under the threat of seizure, a situation that is both unpleasant and bad
for work effort. Permitting discharge

8
D’s options are to hand over all of her discretionary income for the next 24 months or
reduce her work effort. Since the less work
option is preferable to the prospect of handing
over all her discretionary income for the next
13 (or more) months, she will surely choose the
less work option when faced with the option of
handing over all of her discretionary income for
the next 24 months.

www.philadelphiafed.org

eliminates this possibility but reduces
the maximum amount a debtor can
borrow. The amount the debtor can
borrow is less because the lender must

then, the case for a discharge policy
rests ultimately on the risk of bad
outcomes and the fact that discharge
provides a form of insurance against

The economic logic of a fresh start provides
insights into the history of the evolution
of personal bankruptcy law in the U.S., in
particular, into Americans’ divergent attitudes,
over time and space, toward the efficacy of a
discharge policy.
make certain that the debtor has the
incentive to repay even when there
is no financial hardship. If debtors’
aversion to the risk of bad outcomes is
sufficiently strong, or if their need to
borrow is sufficiently weak, then from
the debtors’ perspective permitting
discharge will be preferable to prohibiting it.
This risk-based logic for discharge
is further strengthened if we recognize
that lenders need not bear any losses
from having a discharge policy in
place. This is so because in the (likely)
event that D does make all six of her
monthly payments of $200 each, C
earns more than a 5 percent rate of return (annualized).9 Therefore, by lending on similar terms to many people,
C can use the additional returns from
the above-5-percent interest rate paid
by nondefaulting debtors to offset the
losses inflicted by defaulting debtors
and still obtain an average return of 5
percent on his investments.
From the pre-debt perspective,

9

This is so because C lent an amount less than
$1183 but insisted that D pay back $200 for six
months. For instance, if C lent only $1500, the
implicit interest rate in the event D paid back
the loan would be about 15 percent.

this risk. The economic logic of a fresh
start then comes down to a comparison between the benefits of insurance
and the costs of a reduced borrowing capacity. If enough people in the
economy value the insurance benefit
of a discharge policy more than the
cost of a reduced borrowing capacity, a
policy of discharge, or fresh start, will
be socially desirable.
FRESH START AND THE
EVOLUTION OF U.S.
BANKRUPTCY LAW
The economic logic of a fresh start
provides insights into the history of
the evolution of personal bankruptcy
law in the U.S., in particular, into
Americans’ divergent attitudes, over
time and space, toward the efficacy
of a discharge policy. An obvious but
nevertheless important point about the
economic logic is that it states conditions under which having a discharge
policy is socially desirable. Thus, if we
observe an economy that looks upon
a discharge policy with disfavor — or
vice versa — the economic logic of a
fresh start suggests reasons this might
be so.
The colonial history of the United
States affords a unique opportunity to
observe the economic logic of a fresh

Business Review Q1 2008 5

start at work.10 Each of the 13 original
colonies started out with debtor and
creditor rights based on English laws.
These laws gave creditors the right to
seize the property of insolvent debtors
and, if there was any suspicion that
the debtor was hiding property, to
imprison him or her. There are ample
records of impoverished insolvent
debtors spending years in jail petitioning colonial legislatures for relief. A
discharge of debt was possible at the
behest of the creditor only.
But as time progressed, the colonies altered these laws to suit their own
needs.11 As one would expect from the
post-debt logic of a fresh start, the law
that came under pressure first was the
law permitting the imprisonment of
debtors. It was clear to everyone that
keeping insolvent debtors in jail for
years served no useful purpose. It was
unlikely that someone who had been
imprisoned for several years had any
hidden wealth, so creditors were not
being served by this imprisonment and
society lost the labor of the imprisoned
debtors. Bowing to public pressure,
Massachusetts enacted a debtor relief

10

The discussion in this section draws from
three sources. Most heavily, it draws from Peter
Coleman’s highly regarded history of insolvency,
imprisonment, and bankruptcy in colonial
America. It also draws from historian Bruce
Mann’s recent book on colonial America’s attitude toward debt and debtors, and it draws
from David Skeel’s fascinating account of the
century-long legislative struggle to establish a
federal bankruptcy law.

11

Thus, with regard to bankruptcy laws, the
colonies acted like small, largely independent
democratic countries with legislatures attuned
to the needs of their respective citizenry, a fact
that imperial authorities in Britain did not
always care for but put up with nevertheless.
Interestingly, this situation continued after
independence because even though the Constitution gave the federal government the right to
devise “uniform laws regarding bankruptcy,” no
long-lasting federal bankruptcy law was devised
until 1898. Thus, the states (and territories)
continued to enact local bankruptcy laws to
meet their own individual needs until the dawn
of the 20th century.

6 Q1 2008 Business Review

law in 1698, which permitted insolvent
debtors who owed less than £500 to
obtain their release from jail upon
swearing an oath of poverty. But the
law did not discharge debts: Debtors
were responsible for paying back everything they owed, and creditors retained
the right to attach future property accumulated by insolvent debtors toward
satisfaction of any unmet obligations.
As the discussion of the post-debt
logic of discharge would lead us to
expect, this law had an adverse effect
on the work effort and savings of insolvent debtors released from jail. Why
would an insolvent debtor exert effort
and accumulate wealth if his creditors
could “tax” it all away? The Massachusetts legislature acknowledged this
problem when, in 1725, it noted that
the law had been a “…great encouragement to idleness and ill-husbandry, and
too much a temptation to perjury…”
and repealed it. Nevertheless, similar
laws were passed periodically until
1787, when relief of debtors who owed
moderate amounts of money was made
a permanent part of Massachusetts
law. The experience of Massachusetts in this regard is similar to that
of the other colonies that enacted
debtor-relief laws. But even though the
post-debt logic of discharge was quite
apparent to the Massachusetts legislature, neither it nor most other colonial
legislatures permitted discharge of debt
until much later.
This reluctance to permit discharge can be explained from the
perspective of the economic logic of
a fresh start. Recall that permitting
discharge leads lenders to lend less.
Hence, whenever the capacity to borrow is really important to people, we
can expect society to be hostile to the
idea of discharge. There is ample historical evidence that in the early years
of colonization, ordinary people acutely
felt the need to borrow. To quote
historian Bruce Mann: “Debt was an

inescapable fact of life in early America. One measure of how thoroughly
this was so is in the pervasiveness
of debts owed and owing in probate
inventories….Debt cut across regional,
class, and occupational lines. Whether
one was an Atlantic merchant or a
rural shopkeeper, a tidewater planter or
a backwoods farmer, debt was an integral part of daily life.” This one single
fact probably goes a long way toward
accounting for the general reluctance
of colonists to enact a discharge policy.
When a person’s ability to earn a living depends on his or her capacity to
borrow, it is not in the interests of a
people to erect barriers to the flow of
credit by making it difficult for creditors to collect on their loans.12
Nevertheless, a few colonies did
enact discharge laws fairly early in
their histories: Rhode Island, New
York, Maryland, and South Carolina.13
These exceptions may also be consistent with the logic of a fresh start. As
historian Peter Coleman notes, these

12

The colonies tempered the adverse consequences of not having a discharge policy by
modifying creditors’ rights. Many colonies
passed laws that prevented creditors from immediately seizing the assets of borrowers in default.
These “stay laws” gave insolvent debtors breathing room to meet their obligations. That way, if
the reason for their inability to meet debt payments was temporary, they were not deprived
of their assets. Also, during times of general
colony-wide financial distress, many colonies
passed laws that discharged the debts of people
who were insolvent on a particular date. In this
fashion, many colonies provided relief to debtors
on an ad hoc basis without having an official
discharge policy in place. See the article by Ian
Domowitz and Elie Tamer for a description of
how business conditions influenced bankruptcy
legislation.

13

Pennsylvania did not enact a discharge policy,
but a discharge policy for the residents of the
county of Philadelphia was enacted and allowed
to stand for a brief period. New Jersey had a
discharge policy in place during 1771-1787 but
abolished it thereafter. Delaware did not permit
discharge until 1900. Massachusetts, the place
where America’s industrial revolution took
root, enacted a discharge policy in the mid-19th
century.

www.philadelphiafed.org

colonies stood out for being heavily
commercialized. In these colonies we
would expect a discharge policy to
reflect the needs of entrepreneurs, a
class of people for whom risk-sharing is
more critical.14
After independence, the U.S.
Constitution granted the federal government the right to enact a uniform
bankruptcy law. But attempts to do so
failed miserably for almost a century.
Historians have puzzled over why it
took so long for a bankruptcy bill to
pass and what led to its passage and
success in 1898. As with any other
piece of legislation, special interest
groups had a lot of influence in shaping the character of various bankruptcy bills. But because of the ubiquity of
debt in early America, the fairness and
practicality (or lack thereof) of any
bankruptcy law became quickly apparent to people, and if the law performed
poorly, it did not last.
Put differently, the economic logic
of a fresh start was a constant reality
check on the interest-group logic of
relief laws passed by lawmakers.15 It is
significant that the 1898 bankruptcy
bill – the one that eventually lasted
long enough to become permanent –
allowed U.S. states to have a say in local discharge policy. Thus, a discharge
policy suited to local needs became
possible, and the law itself found acceptance. It is perhaps also significant
that by the end of the 19th century,
America was no longer a nation of
“rural shopkeepers, tidewater planters

14

It is important for entrepreneurs to be able to
borrow in order to get a venture going. If the
initial venture is successful, further expansion
can be financed by borrowing against accumulated assets.

15

Interest-group politics explains regulatory
capture: how interest groups can use the law
to their own benefit. It explains why regulation might end up serving the interests of the
industry it regulates rather than the interests of
society as a whole.

www.philadelphiafed.org

or backwoods farmers.” It was a nation
where three-quarters of industrial output was generated in business corporations. The rise of corporations meant
that ordinary people were much more
likely to be wage earners and thus less
reliant on credit to earn a living.
Now, at the start of the 21st century, the situation has changed. Once
again, debt has assumed greater importance in the lives of ordinary people.
People and businesses expect to buy
and sell all manner of consumer goods
and services on credit, a development

Recall that although a discharge policy
is primarily meant to give the borrower
an escape route if his or her capacity
to repay is impaired, a lender must
contend with the fact that a borrower
might invoke discharge even when
he or she has the capacity to repay.
Indeed, the reason the creditor (in our
example) had to restrict his lending to
the debtor was to ensure that the debtor invoked discharge only when her
capacity to repay was impaired. The
latest bankruptcy bill is an attempt
to relax this limitation on lending by

The latest bankruptcy bill puts significant
restrictions on who can avail themselves of
the right to discharge.
that began with the proliferation of
consumer durables in the 1920s, most
notably automobiles. Thus, credit for
ordinary people (consumer credit) has
again become an integral part of our
economic system, and, predictably, this
development has led to dissatisfaction
with the policy of discharging debts.16
The latest bankruptcy bill puts significant restrictions on who can avail
themselves of the right to discharge.17

16

There are more subtle changes at work as
well. Because fresh start is a form of insurance,
the need for it is not as great if there are other
forms of insurance available. Since the early
1940s, unemployment insurance has become
widely available in the U.S., and this development makes people more willing to accept limits
on discharge policy in return for an increased
capacity to borrow. See the article by Kartik
Athreya for more discussion of the interaction
between unemployment insurance and fresh
start.

insisting that the debtor cannot invoke
discharge if she has the capacity to
repay. In this spirit, the law does not
allow households with above-median
income to invoke discharge. The result
of the law will be to make more credit
available at cheaper terms – something
the average consumer presumably
wants and will benefit from.18
CONCLUSION
U.S. law gives individual debtors
the right to petition a bankruptcy
court and ask to be released from
their financial obligations to creditors.
This right is referred to as the law’s
fresh start provision – after a famous
Supreme Court ruling that succinctly
captured the basic reasons for having
such a policy. As discussed at length
in this article, the reasons fall into

17

The Bankruptcy Abuse Prevention and Consumer Protection Act of 2005 did more than
just restrict access to discharge for above median-income households. See Loretta Mester’s
article for a comprehensive discussion of the
reform (when the act was still in the proposal
stage) and the empirical research on bankruptcy
and credit that bears on it.

18

My article with Dean Corbae, Makoto Nakajima, and Vìctor Rìos-Rull examines this issue
in a numerically specified model of the U.S.
economy and concludes that restricting abovemedian-income households’ access to fresh start
will benefit the average U.S. household.

Business Review Q1 2008 7

two categories. One justification for a
discharge policy is that it eliminates
the adverse effects that unrestricted
creditor rights can have on an insolvent debtor’s work incentives. A deeper
justification is that a discharge policy
provides a form of insurance benefit:
The policy permits a debtor to not

repay his or her debts when doing so
would be very costly. Of course, these
benefits come at the cost of a reduced
capacity to borrow. Perhaps the most
important lesson to be gleaned from an
understanding of the economic logic of
a fresh start is that there are both costs
and benefits of adopting a discharge

policy. Furthermore, there is nothing
in the economic logic to suggest that
the benefits necessarily exceed the
costs. Indeed, the evolving calculus of
costs and benefits may well account for
Americans’ changing attitudes toward
the efficacy of a fresh start. BR

Jackson, Thomas H. Logic and Limits of
Bankruptcy Law. Cambridge, MA: Harvard
University Press, 1986.

Robe, Michel, Eva-Maria Steiger, and
Pierre-Armand Michel. “Penalties and
Optimality in Financial Contracts: Taking
Stock,” SFB 649 Discussion Paper 2006013, Humboldt University, Berlin.

REFERENCES

Athreya, Kartik. “Unemployment
Insurance and Personal Bankruptcy,”
Federal Reserve Bank of Richmond
Economic Quarterly (Spring 2003).
Chatterjee, Satyajit, Dean Corbae, Makoto
Nakajima, and José-Víctor Ríos-Rull.
“A Quantitative Theory of Unsecured
Consumer Credit with Risk of Default,”
Econometrica, 75(6), 2007, pp. 1525-89.
Coleman, Peter J. Debtors and Creditors
in America: Insolvency, Imprisonment for
Debt and Bankruptcy, 1607-1900. Madison:
State Historical Society of Madison, 1974.
Domowitz, Ian, and Elie Tamer. “Two
Hundred Years of Bankruptcy: A Tale of
Legislation and Economic Fluctuations,”
Working Paper 97-25, Institute of Policy
Research, Northwestern University (1997).

8 Q1 2008 Business Review

Li, Wenli. “What Do We Know About
Chapter 13 Bankruptcy Filings?” Federal
Reserve Bank of Philadelphia Business
Review (Fourth Quarter 2007).
Mann, Bruce H. Republic of Debtors:
Bankruptcy in the Age of Independence.
Cambridge, MA: Harvard University Press,
2002.
Mester, Loretta. “Is the Personal
Bankruptcy System Bankrupt?” Federal
Reserve Bank of Philadelphia Business
Review (First Quarter 2002) .

Skeel, David A. Jr. Debt’s Dominion: A
History of Bankruptcy Law in America.
Princeton: Princeton University Press,
2001.
White, Michelle. “A General Model of
Personal Bankruptcy: Insurance, Work
Effort, and Opportunism,” American
Law and Economics Association Annual
Meetings, Paper No. 46, 2005.

www.philadelphiafed.org

The Industrial Revolution and the
Demographic Transition

I

BY AUBHIK KHAN

n the 19th century, the United Kingdom
began a period of economic transformation
known as the Industrial Revolution. It’s
commonly believed that this era opened
as new inventions improved the technologies used to
produce goods and provide services. However, we now
know that such improvements affected only a relatively
small part of the economy. Nonetheless, output rose
during the first stage of the Industrial Revolution because
of capital accumulation. One explanation for this increase
in capital may be that another revolution occurred
in Britain around the same time: the demographic
transition. In this article, Aubhik Khan outlines
some evidence on the Industrial Revolution and the
demographic transition, then presents two economic
theories that link the two phenomena.

In the 19th century, the United
Kingdom began a period of economic
transformation known as the Industrial Revolution. While the typical
reader may think of Dickensian mills
when hearing of the Industrial Revolu-

Aubhik Khan
is a senior
economic advisor
and economist
in the Research
Department of
the Philadelphia
Fed. This article
is available free of
charge at www.philadelphiafed.org/econ/br/.
www.philadelphiafed.org

tion and of the end of a pastoral society, for most economists, the Industrial
Revolution is associated with a change
in the long-run or average rate of
growth of per capita income. Also, in
the 19th century, a steady rise in living
standards began that has, in some
sense, never ceased. As a result, people
are now accustomed to economic
growth. They expect it alongside the
sometimes gradual, sometimes abrupt
changes to the organization of industry
and society associated with technological change.
Prior to the Industrial Revolution, the notion that there would be

an improvement in people’s standards
of living almost every year would be
unfamiliar not only to laypersons,
whether common people or the nobility, but also to economists working in
that period.
It is commonly believed that the
Industrial Revolution began as new
inventions improved the technologies
used to produce goods and provide
services. However, there is a difficulty
with this account: We now know that
such improvements affected only a few
sectors that represented a small part of
the economy. In the absence of widespread improvements in technology,
output rose during the first stage of the
Industrial Revolution because of capital accumulation — that is, because
there was an increase in the quantity
of machines and tools available to each
worker.
Why did society suddenly choose
to increase capital at an increasing
rate? One answer may be that another
revolution occurred in Britain around
the same time: the demographic
transition. This demographic transition saw the rate of population growth
in the United Kingdom first rise, and
then later fall. During this period,
adult mortality fell, then child and
infant mortality, then finally fertility.
After presenting some evidence
on the Industrial Revolution and the
demographic transition, I present two
economic theories that link the two
phenomena. The first explains the
slowdown in population growth as a
result of technological progress. It
represents the conventional view that
the Industrial Revolution drove the
demographic transition. An influential
summary of this theory is contained in
Business Review Q1 2008 9

the 2002 book written by Nobel laureate Robert E. Lucas. The second economic theory — which is part of my
ongoing research with Michele Boldrin
and Larry Jones — suggests that causality runs in the opposite direction.
These different theories have different
implications for how modern developing economies may improve their rate
of growth. For example, to the extent
that demographic transitions affect
economic development, policy that reduces mortality and fertility may raise
the level of economic development.
THE INDUSTRIAL
REVOLUTION
Real Wages and Population
Stagnated Until 1800. Between 1250
and 1800 there was little sustained
improvement in the British economy.
The economic history of Great Britain
over this period is reasonably well captured by a model originally developed
by Robert Malthus.
Malthus’s theory suggested an
inverse relation between the real wage
(the wage paid to laborers measured
in terms of the goods it can provide)
and population. This inverse relation
stems from the value of labor. For
example, when population was lower
than its average level, labor would be
relatively scarce. This would drive up
real wages as landowners bid for scarce
laborers. Increases in real wages would
allow laborers to purchase more goods
and services, including better food
and shelter. Their standard of living
would rise. This rise in living standards would also increase the number
of children born that would survive
into adulthood. This would move
population back to its average level
and reduce the scarcity of labor. As a
consequence, landowners, no longer
having difficulty operating their farms,
would reduce the real wage back to its
average level. The resultant decline in
the living standards of workers would

10 Q1 2008 Business Review

end the growth in population.
Malthus’s theory could explain
the persistent rise in the real wage in
England during the 15th and 16th centuries. Over this time the Black Death
sharply reduced the number of laborers.1 However, the theory also implied
that society would always remain poor
and that the “perfectibility” of society

conditions there was over the 500-year
period. As stated above, the level of
the real wage in 1390 is very close to
that observed in 1740. Equally striking
to someone living today is that there
is little discernible difference in the
population of England between 1350
and 1740. For comparison, the population of the United States was 248

Malthus’s theory suggested an inverse relation
between the real wage (the wage paid to
laborers measured in terms of the goods it can
provide) and population.
was infeasible. Whenever living conditions temporarily improved, population
growth would bring them back down.
This somewhat bleak outlook on life
was consistent with the observation
that the real wage was about the same
in 1740 as it had been 350 years before.
Figure 1 is taken from the influential paper by Gary Hansen and Edward
Prescott. It shows the population of
England and the average real wage
paid on farms from the end of the 13th
to the middle of the 19th century.
Over this period farm laborers had
little to no assets, and they worked as
many hours as their employers demanded, subject to their health. As
a result, their real wage can be taken
as a very good indicator of their real
income.
What is striking from the figure,
when viewed through the eyes of
someone who lives in the 21st century, is how little net change in living

million in 1990, having almost doubled
in the 50 years since 1940, when it was
132 million.2
The small overall changes in real
wages and population provide support
for Malthus’s theory of a natural longrun level of population associated with
a particular real wage. Furthermore,
the rise in real wages in the 15th and
16th centuries, which occurred at the
same time that periodic outbreaks of
plague led to an extraordinary rise in
mortality and reduction in population,
is also consistent with the Malthusian
view.
After 1800 Both Real Wages
and Population Grew. This inverse
relationship between real wages and
population began to change around
the beginning of the 19th century.
Between 1780 and 1989, the real wage
rose 22-fold. The English Industrial
Revolution had arrived, bringing with
it a sustained improvement in living
conditions.

1

Catastrophic outbreaks of plague afflicted
the English periodically between the mid-14th
century and the 17th century. One of these
outbreaks, known as the Great Plague (16651666), is estimated to have cost between 75,000
and 100,000 lives in London, about one-fifth of
the city’s population.

2

The population data, which include
immigration, are taken from the U.S. census
and are available at http://www.census.gov/
population/censusdata/table-2.pdf.

www.philadelphiafed.org

FIGURE 1
The English Economy: Population and
Real Farm Wage
Population (Millions)*

Real Farm Wage
5.0

10

4.5

9

4.0

8

3.5

7
Wage

3.0

6

2.5

5

2.0

4

1.5

3
Population

1.0

2

0.5
0.0

1275

1
1350

1425

1500

1575

1650

1725

1800

0

* Dashed line indicates missing data.
Source: Figure 1 in Hansen and Prescott (2002)

As we know, before the Industrial
Revolution, there was little change in
living standards. If we set the real GDP
per person in Great Britain to 100 in
1566, it had risen to only 130 by 1806.
This implies an annual rate of economic growth in income per person of
0.11 percent over a 240-year period. In
other words, there was no discernible
improvement, at least on average, in
the quality of life for most people.
However, beginning in the early
19th century, growth rates began to
rise. Between 1806 and 1906, income
per person grew at an average of 0.9
percent a year, that is, more than
eight times faster. From 1906 to 1990,
income per person in the United Kingdom has grown at an annual rate of 1.5
percent a year. This is more than 13
times faster than the average growth
rate between 1566 and 1806.
Problems with the Technological Explanation. In the traditional
view, new inventions brought about

www.philadelphiafed.org

this new era of persistent growth. Examples include James Watt’s improved
steam engine, John Kay’s fly shuttle,
and James Hargreaves’ spinning jenny.
However, as famously argued by N.F.R.
Crafts and C. Knick Harley, while
these and several other well-known
discoveries were applied to production in the 19th century, their impact
was limited to just a few sectors in the
economy in the early part of the Industrial Revolution.
Gregory Clark’s quantitative assessment of the role of technological
progress in the 18th century supports
Crafts and Harley’s view. To assess the
impact of technological progress on
the economy, we must break overall
production per person into components that are attributable to capital,
labor, and total factor productivity.
This is the famous growth decomposition first used by Nobel laureate
Robert Solow in his 1957 paper. Solow
assumes that the output of goods and

services requires two inputs. The first
is labor. The total quantity of labor
used by a business is measured as the
number of workers times the average
hours worked by each. A rise in the
quantity of labor, either because there
are more workers or because they
work longer hours, increases the total
quantity of goods or services produced
by the business. The second input is
capital, the quantity of machines and
buildings used to produce goods and
services. An increase in capital means
that more machines and buildings are
used for a given method of production.
The third component is a change in
the method of production — that is,
in the overall level of technology —
and is called a change in total factor
productivity. Inventions that allow
more output to be produced without
increasing the quantity of inputs lead
to a rise in total factor productivity.
Gregory Clark extends the Solow
method to include land as a factor of
production. Separating out changes in
output per worker between 1700 and
1861, he finds that total factor productivity growth shows little rise until
the middle of the 19th century. This
means that the role of discovery and
innovation — that is, technological
progress — in spurring the Industrial Revolution was relatively minor.
Instead, for some reason, society as a
whole began to invest more heavily
in capital, that is, in machines. Since
capital is accumulated by using current
production to increase machines and
buildings instead of consuming it, an
increase in capital implies a rise in the
savings rate.3 I will discuss a possible
reason for this change in the rate at
which society saved output below.

3

This is strictly true only for a country that
can’t borrow from abroad to finance investment.
While there was international borrowing and
lending in 18th century England, access to such
funds was limited.

Business Review Q1 2008 11

THE DEMOGRAPHIC
TRANSITION
Over the same two centuries
associated with the English Industrial Revolution, there were dramatic
changes in population growth and life
expectancy driven by changes in the
underlying factors that explain them:
fertility and mortality. Population
growth rose in England around 1700
and continued to rise until reaching a
peak of 1.36 percent a year during the
period 1791 to 1831. Looking across
centuries, we find that between 1680
and 1820 the population of England
increased 133 percent. Next, between
1820 and 1900 it rose another 166
percent. When compared with other
large European nations, this represents
a dramatic increase in population. For
example, the corresponding increases
in France were 29 percent and 26
percent (Figure 2).
Two economic historians, E.A.
Wrigley and Robert Schofield, describe
a famous finding in their 1981 book:
Most of the increase in population
was the result of a rise in fertility. We
see little change in life expectancy
between 1700 and 1870 largely because
infant and child mortality did not fall
until late in the 18th century. For example, the expected life span was 36.8
years between 1701 and 1711; 160 years
later, between 1861 and 1871, it had
risen to only 40.7 years. Notably, the
mortality rates of people between the
ages of five and 20 fell markedly over
this period. For the years between 1735
and 1970, Figure 3 plots the fraction
of children that survived to their fifth
and 20th year of life.
Aside from the fall in child
mortality, a dramatic rise in fertility
occurred during this period. Over the
250 years before 1800, the crude birth
rate (a measure of fertility) first fell,
then rose. However, in 1796, at 35.51
births per 1000 people, it was no different from its level in 1551. Thereafter,

there is a notable increase in fertility
until it peaks in 1821 at 40.22 births
per 1000. Fertility remained high until
the beginning of the 20th century
when it began to decline, as mortality
had done earlier.
These changes in fertility, mortality, and population growth are known
as a demographic transition (Figure
4). A demographic transition involves
four stages, broad patterns that social
scientists have observed across countries. In the first stage, both fertility
and mortality are high, and population
growth is low. In the second stage,
mortality begins to fall first, without a
change in fertility. Population growth
rises over this second stage. Over the
third stage, fertility falls. In the fourth
stage, both mortality and fertility settle
at low levels, and population growth is
once again low (although the level of
population has now risen). The transition in England is exceptional in that
the high initial level of fertility, rather
than simply falling sometime after

the second stage, first rose only to fall
much later on.
THE LINK BETWEEN FERTILITY
AND ECONOMIC GROWTH:
TWO ECONOMIC THEORIES
Economists and other social scientists have produced a huge literature
about the Industrial Revolution. There
is also a large body of work that studies
the demographic transition. Here I
discuss only economic theories that
link the two events, and even then I
discuss only one example of each of
the two theories.
The first theory is by far the most
commonly accepted, and I will call it
the technology-led theory. This theory
suggests that improvements in technology led to the Industrial Revolution
and that the associated rise in the
standard of living reduced mortality.
Fertility fell as people began to invest
in the quality of their children.
The second theory is relatively
new and undeveloped and, therefore, is

FIGURE 2
UK Population and GDP Per Capita, 1565 to 1990
In(Pop), In(GDP)
12

10
In(Pop)

8

6
In(GDP)

4

2

0
1550

1650

1750

1850

1950

Source: Boldrin, Jones, and Khan (2005)
12 Q1 2008 Business Review

www.philadelphiafed.org

far less widely accepted. It argues that
the demographic transition preceded
economic development and, moreover, was responsible for some of the
improvement in living standards. I will
call it the demography-led theory.
Economic Models of Fertility.
Both theories rely on an economic
model of household fertility choice,
a theory of how parents decide how
many children to have. When studying fertility choices of households,
economists assume that parents care
about their children’s happiness or
welfare, as well as their own. With this
assumption, economists have gained
powerful insights about fertility choices
by a household that wishes to maximize its welfare. The most famous proponents of this view are Robert Barro
and Gary Becker, and I will describe
a very simple version of the approach
taken in their 1989 paper.
Barro and Becker developed a
model in which parents care about
both the number of children they have
and the welfare of those children. At
the same time, parents also value their
own direct consumption of goods and
services. Given their income and their
time, they must trade off their own
welfare from consuming goods against
their welfare from having children, as
well as their children’s welfare.4
In applications of the Barro and
Becker model to economic development, parents are able to affect the
welfare of their children by investing in
their education. Specifically, parents

FIGURE 3
English Survival Rates
1.2

1.0

0.8

0.6

0.4

0.2

0
1735

1765

1795

1825

1850

1870

probability of surviving to age 5

1890

1910

1930

1950

1970

probability of surviving to age 20

Source: Boldrin, Jones, and Khan (2005)

FIGURE 4
A Stylized Demographic Transition

Population

births and deaths per thousand

120

45
40

100
35
80

30
25

Stage 1

Stage 2

Stage 3

Stage 4

60

20
4

An alternative view of population growth
is discussed by Stephen Parente and Edward
Prescott in their chapter in the Handbook of
Economic Growth. They argue that fertility
choices are not made at the household level
but at the societal level and that they are
implemented through a range of policies that
either promote or hinder families’ choices
about how many children to have. Parente
and Prescott suggest that these policies arose
because pre-industrial societies had to defend
land.

www.philadelphiafed.org

40

15
10

20
5
0

0

20

40

60

80

100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400

0

time

crude birth rate

crude death rate

population

Business Review Q1 2008 13

choose how much costly human capital
to give to each child.5 Higher levels of
human capital, by increasing children’s
skills, allow them to earn more real
income. This, in turn, enables them to
raise their own consumption and thus
their welfare. Thus, parents face two
choices involving their children: They
must decide how many children to
bear, and they must determine the human capital investment in each child.
Technology Leads Demography.
The technology-led theory finds that
improvements in technology increase
the return to investment in human
capital. Prominent examples of this
theory are contained in the works of
Gary Becker, Kevin Murphy, and Robert Tamura, and in the work of Lucas.
Before the technological improvements
that led to the Industrial Revolution were implemented, the return
to investing in the human capital of
each child was relatively low, at least
given the costs, because the difference in the earnings of skilled and
unskilled workers was small. However,
the introduction of new technologies
brought with it more complex methods
of production, and total factor productivity increased. In such environments
skilled workers became more valuable
than they had previously been, and the
wage premium paid to skilled workers
rose.
The rise in the skill premium led
those parents who could afford it to invest more heavily in the human capital
of their children. Over time, improvements in income led to more and more
parents being able to afford to educate
their children. Both the rise in total
factor productivity and the increase in
human capital led to increases in the

real earnings of workers. Living standards improved. Moreover, the move
to increased investment in human
capital increased the cost of having
children for parents. As a result, the
number of children per family fell over
time.

5

6

Economists use the term human capital to
describe a worker’s skills and ability. Investment
in human capital is usually believed to be timeintensive and includes years spent in formal
education as well as on-the-job training.

14 Q1 2008 Business Review

began in the 18th century. Investing
in a child’s human capital will turn
out to be a waste if he or she does not
survive long enough to benefit from it.
Thus, investment in human capital is
very risky when childhood mortality
is high. However, if children of school

Before the technological improvements
that led to the Industrial Revolution were
implemented, the return to investing in the
human capital of each child was relatively low.
It is convenient shorthand to
describe children with a higher level
of human capital as children with a
higher skill quality.6 According to
economic theories of fertility, there is
a tradeoff between the quality and the
quantity of children a family has. The
technology-led theory argues that new
inventions moved families to increase
quality at the expense of quantity and
that this reduced fertility.
This conventional view can explain the fall in fertility that occurred
at the end of the Industrial Revolution.
However, a weakness is that it relies,
to some extent, on the thesis that the
Industrial Revolution was spurred by
technological improvements. As I discussed above, there is some evidence
to suggest that this was not initially
true. It also suffers from another problem: Economic growth rose long before
fertility fell.
Demography Leads Technology.
The demography-led theory centers
on the effects of the fall in mortality,
for children age five and above, that

Obviously, a person’s quality can’t be reduced
to his or her skill level. Using the terminology
of a quality/quantity tradeoff, however, places
the family’s problem in a familiar economic
framework that allows for clarity of exposition.

age are likely to live on to adulthood,
costly expenditures on their schooling
become less risky.
The demography-led theory suggests that reductions in mortality for
children age five and older increased
the return to human capital investments for children, since, once they are
old enough to receive formal education
and specialized training in skills, they
would also be more likely to live on to
earn the higher wages of skilled workers. As before, this drives an increase
in parents’ investments in children and
a reduction in fertility.
As more skilled workers are
able to make better use of machines,
increases in human capital raise the
returns to investing in physical capital.
At the same time, the higher earnings
by households with skilled workers
raise average household income. This
allows for a rise in savings, which, in
turn, funds physical capital investment
in the economy. Driven by the rise
in human capital and the resultant
increase in income, the stock of physical capital grows. This availability of
better equipment for skilled workers
compounds the effects of the initial
rise in human capital, and there is
further accumulation of both human
and physical capital.
www.philadelphiafed.org

The demography-led theory’s appeal is that it doesn’t rely on total factor productivity growth to explain the
fall in fertility. However, it does not explain the reductions in mortality that
occurred during the Industrial Revolution. These are explained implicitly by
the technology-driven theory as the
natural consequence of improvements
in medical technology. Explaining
them explicitly is more important for
the demography-driven theory, since it
relies heavily on changes in mortality.
Another difficulty with the demography-led theory is that it is, as of yet,
insufficiently developed to evaluate it
against data.
Both the technology-led theory
and the demography-led theory explain
changes in growth and fertility through
parents’ decisions on how many children to have and how much to invest
in their education, skills, and general
well-being. Both emphasize the quality/
quantity tradeoff. What distinguishes
the two theories is why this tradeoff

changes. In the technology-led theory,
improvements in technology raise the
return to investment in the human
capital of children. In the demographyled theory, this return rises because
older children, who are the recipients
of such investments, live longer. This
increases the benefit they may expect
from human capital investment.
CONCLUSION
In the 17th and 18th centuries,
Great Britain experienced an economic transformation, the Industrial
Revolution, which began a period of
economic growth and prosperity that
defines the modern era. Standards of
living that had fluctuated for hundreds
of years now began to improve steadily.
Roughly over the same period, a
demographic transition occurred. First,
adult mortality fell; sometime later
there was a decline in child and infant
mortality. Fertility initially rose and
then fell alongside mortality. These
changes led to a sharp rise in popula-

tion growth rates, which subsided only
after many decades.
Economic theory offers explanations that uncover the links between
the Industrial Revolution and the
demographic transformation. I have
discussed two theories. The first,
the technology-led theory, is widely
understood and supported. The
second, the demography-led theory, is
relatively new. It has been developed
partly in response to several difficulties
with the technology-led theory. Most
notably, the timing of events suggests
some difficulty, though perhaps not an
insurmountable one, in explaining the
proposition that an increase in income
led to a fall in fertility. The Industrial
Revolution began at the end of the
18th century, but fertility did not fall
until 100 years later. This timing is
consistent with the demography-led
theory, but a full evaluation of the
relative merits of the two theories
will require a more careful empirical
examination. BR

REFERENCES

Barro, Robert J., and Gary S. Becker.
“Fertility Choice in a Model of Economic
Growth,” Econometrica, 57, 2 (1989), pp.
481–501.

Clark, Gregory. A Farewell to Alms: A Brief
Economic History of the World. Princeton,
NJ: Princeton University Press, 2007.

Lucas, Robert E., Jr. “The Industrial
Revolution: Past and Future,” in Lectures
on Economic Growth. Cambridge, MA:
Harvard University Press (2002).

Becker, Gary S., Kevin M. Murphy, and
Robert Tamura. “Human Capital, Fertility,
and Economic Growth,” Journal of Political
Economy, 98, 5 (October 1990), pp. S12-37.

Crafts, N.F.R., and C.K. Harley. “Output
Growth and the British Industrial
Revolution: A Restatement of the CraftsHarley View,” Economic History Review, 45
(1992), pp. 703-30.

Solow, Robert M. “Technical Change
and the Aggregate Production Function,”
Review of Economics and Statistics 39
(August 1957), pp. 312-20.

Boldrin, Michele, Larry Jones, and Aubhik
Khan. “Three Equations Generating an
Industrial Revolution,” work in progress,
May 2005.

Hansen, Gary D., and Edward C. Prescott.
“Malthus to Solow,” American Economic
Review, 92, 4 (September 2002), pp. 120517.

Wrigley, E.A., and Robert Schofield. The
Population History of England 1541-1871.
Cambridge, MA: Harvard University Press
(1981).

www.philadelphiafed.org

Business Review Q1 2008 15

Human Capital and Higher Education:
How Does Our Region Fare?
BY TIMOTHY SCHILLER

T

he number of people with a college education
in a given state or region varies across the
nation. States in the Third Federal Reserve
District (Pennsylvania, New Jersey, and
Delaware) compare favorably with the nation on measures
of college education, and the three states as a whole are
close to the national average. Despite its average ranking
in educational attainment, the area is a premier location
for colleges and universities. In this article, Tim Schiller
evaluates the region’s standing with respect to college
education by reviewing data on individual and social
returns to education, looking at college education as
a stimulant to local economic growth, and comparing
the tri-state area with the nation as a source of and a
destination for college graduates.
Human capital refers to the technical skills and knowledge acquired by
workers. Education is an investment
in human capital, that is, in the skills
and knowledge that produce a return
to the individual in the form of higher
earnings. Education also has social
returns or spillovers. The presence of
educated workers in a region enhances
the earnings of those who, regardless

Tim Schiller
is a senior
economic analyst
in the Research
Department of
the Philadelphia
Fed. This article
is available free of
charge at www.philadelphiafed.org/econ/br/.
16 Q1 2008 Business Review

of their own educational level, work
with or near educated workers. This
is especially true for spillovers from
college-educated workers. Research
shows that having large numbers of
college graduates in a region increases
that region’s economic growth and that
spillovers (also called externalities)
are an important factor in generating more rapid growth. Aware of this
connection, educators, state and local
governments, and businesses around
the country are making efforts to
increase the educational attainment of
their local work forces, especially the
number of college graduates.
The number of people in a region
who have a college education varies
significantly across the nation. Parts of
the three-state region (Pennsylvania,

New Jersey, and Delaware) compare favorably with the nation on measures of
college education, and the three states
as a whole are close to the national
average. In spite of its average ranking
in the nation, the region is one of the
premier locations for college education.
The area’s colleges and universities are
important sources of college-educated
workers for the nation and the world.
In evaluating the region’s standing
with respect to college education, we
must consider its important role as a
producer of college graduates as well
as its role as a user of college-educated
workers.
To help with this evaluation, I will
review what we know about individual
and social returns to education, look
at college education as a stimulant to
local economic growth, and compare
our region to the nation as a source
of, as well as a destination for, college
graduates.
EDUCATION: AN INVESTMENT
IN HUMAN CAPITAL
Education represents an investment in the knowledge and skills that
increase people’s ability to earn. The
cost consists of the direct outlays for
education as well as the opportunity
cost of forgone income during the time
spent acquiring the education. The
return is the increase in earnings that
results. Economists have measured the
return to education over many years
and found that it increases steadily
for each level of education attained.1
Data from the U.S. Bureau of Labor

1

See the articles by Jacob Mincer; Gary Becker;
and James Heckman, Lance Lochner, and Petra
Todd.
www.philadelphiafed.org

TABLE 1
Unemployment and Earnings of Workers 25 Years and Older (2006)
Unemployment Rate
Percent

Education

Median Weekly Earnings
Dollars

Doctoral Degree

1.4

1,441

Professional Degree

1.1

1,474

Master's Degree

1.7

1,140

Bachelor's Degree

2.3

962

Associate's Degree

3.0

721

Some College, No Degree

3.9

674

High School Graduate

4.3

595

Less Than High School Diploma

6.8

419

Source: Bureau of Labor Statistics

Statistics show that earnings rise and
unemployment declines for each higher
level of education (Table 1).
The economic importance of
education has been growing. Even as
the number of college graduates in the
labor force has increased, the wage
gap between these workers and those
with less education has widened. The
increased wage reflects an increase in
demand that has been greater than the
increase in supply. Firms have been
investing in new technologies that require more workers with the education
and skills to use them, and more and
more of the nation’s economic growth
has been originating in sectors with
high demand for skilled workers.2 The
investment in new technology could
reflect firms’ desire to take advantage of the increase in the supply of
college-educated workers. Or it could
be a result of the development of new

2

See the article by Keith Sill.

www.philadelphiafed.org

general-purpose technologies, such as
advances in computers and telecommunications, that either require or are
most productively used by educated
workers. In either case, the increasing
premium for college-educated workers
in the face of rising supply indicates
that the growth in demand for collegeeducated workers has exceeded the
growth in supply.
EDUCATION SPILLOVERS
AND REGIONAL ECONOMIC
PERFORMANCE
In addition to providing a return
to the individual, investment in
education results in spillovers that
benefit others who work with or near
individuals who have made the investment. Spillovers provide the economic
justification for public subsidies for
education and motivate community
interest in improving the educational
attainment of the population.3 Since
3

See the article by Robert Topel.

spillovers appear more likely to stem
from college-educated workers than
from those with less education, much
of the economic research on spillovers
has focused on the extent of college
education among the population under
study.4
Social interaction is the primary
way in which spillovers occur, whether
by chance or by plan. This interaction
is most likely to lead to productive
spillovers if it occurs in a work context.
This context can be provided in a metropolitan area with a high concentration of firms in the same industry, and
it can also be provided in an area with
a diversity of industries.5 In the first
case, employees from different firms
in the same industry can exchange
ideas about new products and production methods more readily because of

4
See the article by Susana Iranzo and
Giovanni Peri.
5

See the article by Gerald Carlino.

Business Review Q1 2008 17

the dense concentration of employees
who work in the same industry. In the
second case, the diversity of industries
allows ideas developed in one industry
to be more widely disseminated to
other industries, where the new ideas,
perhaps with some modifications, can
also be productively applied. In both
cases, exchanges of information about
productivity-enhancing possibilities
are more likely in areas with greater
population size, density, and industrial
variety.
Innovation, spillovers, and improved productivity are more likely in
metropolitan areas with large concentrations of workers with higher
education. Empirical research supports
this insight, demonstrating that earnings, which are based on productivity,
are greater in metropolitan areas that
have greater concentrations of college
graduates. Research by Enrico Moretti
estimates that a one-percentage-point
increase in the supply of college graduates in a metropolitan area raises wages
for workers in that area: 1.9 percent for
high school dropouts, 1.6 percent for
high school graduates, and 0.4 percent
for college graduates.6 Furthermore,
research by Edward Glaeser and David
Maré finds that growth in earnings appears to be more rapid in urban areas;
an initial wage increase of about 7
percent when workers first move from
rural to urban areas rises to an urbanrural difference of about 10 percent in
three to five years.
By making workers more productive, education enables faster earnings
growth for the educated individual.
Additionally, various research studies

have revealed that areas with concentrations of educated residents are more
likely to have faster growth in population, employment, and productivity
than areas where college-educated residents are less concentrated.7 Of course,
college graduates are likely to relocate
to obtain employment early in their careers; therefore, rapidly growing areas

Pennsylvania ranks high among all states
in the U.S. in the number of colleges and
universities and in the number of degrees
awarded, both absolutely and when adjusted
for total state population.
are likely to attract them. Thus, there
is a certain counterbalance between
influences: Concentrations of college graduates influence growth, and
growth influences the concentration of
college graduates. I discuss this in more
detail later when I talk about local area
efforts to increase the college-educated
shares of their populations.
RAISING THE LEVEL OF
EDUCATIONAL ATTAINMENT
IN A REGION
As we have seen, college education is beneficial to the individual
who possesses it. It also has spillover
benefits for co-workers and residents
of a region where large numbers of
college graduates work and live. What
are some of the factors that affect the
educational attainment of an area’s
population? At first glance, it would
seem that an area that produces a large

6

The increase in wages for college graduates
is the net effect of two offsetting factors:
spillovers, which raise wages, and the increase
in supply of college graduates, which tends to
reduce wages. The small positive net result
indicates that the spillover effect slightly
overcomes the supply effect.

18 Q1 2008 Business Review

number of college graduates would
have a greater percentage of population with bachelor’s degrees or higher.
The production of college graduates is notably evident in the Pennsylvania-New Jersey-Delaware region. A
large number of colleges and universities produce large numbers of college
graduates, although there is variation

7

See the articles by Curtis Simon and Clark
Nardinelli; Edward Glaeser, Jose Scheinkman,
and Andrei Schleifer; James Rauch; and
Christopher Wheeler.

among the three states. Pennsylvania
ranks high among all states in the U.S.
in the number of colleges and universities and in the number of degrees
awarded, both absolutely and when
adjusted for total state population.
New Jersey ranks somewhat above
average on both measures absolutely,
but below average when adjusted for
total state population. Delaware ranks
below average on both measures absolutely; however, when the measures are
adjusted for total state population, the
percentage moves above average in the
number of degrees awarded but not in
the number of institutions. (See Tables
2 and 3 for state data and rankings.)
Pennsylvania and Delaware
“produce” more college graduates than
they “consume,” and New Jersey “produces” fewer graduates. That is, the
total number of freshmen enrolled in
Pennsylvania and Delaware is greater
than the number of college freshmen
among those states’ population. (Pennsylvania and Delaware bring in some
students from out of state.) The total
number of freshmen enrolled in New
Jersey is lower than the number of colwww.philadelphiafed.org

TABLE 2
Four-Year Colleges and Universities Relative to Total Population (2005)
State

Institutions per 1,000 Population State

Institutions per 1,000 Population

Vermont

0.0353

South Carolina

0.0092

District of Columbia

0.0272

Virginia

0.0091

South Dakota

0.0232

Alaska

0.0090

North Dakota

0.0188

Illinois

0.0088

Nebraska

0.0154

Total

0.0087

New Hampshire

0.0153

Wisconsin

0.0087

Iowa

0.0152

Arkansas

0.0086

Massachusetts

0.0152

Connecticut

0.0083

Maine

0.0151

Alabama

0.0081

Missouri

0.0145

North Carolina

0.0076

Minnesota

0.0131

Georgia

0.0074

West Virginia

0.0127

Delaware

0.0071

Pennsylvania

0.0120

Idaho

0.0070

Kansas

0.0117

Utah

0.0069

New York

0.0115

Mississippi

0.0068

Rhode Island

0.0112

Michigan

0.0068

Tennessee

0.0111

Maryland

0.0068

New Mexico

0.0109

Arizona

0.0067

Oregon

0.0107

Louisiana

0.0066

Oklahoma

0.0107

Washington

0.0065

Montana

0.0107

California

0.0064

Hawaii

0.0102

Florida

0.0062

Indiana

0.0099

Nevada

0.0054

Kentucky

0.0096

Texas

0.0048

Colorado

0.0094

New Jersey

0.0044

Ohio

0.0093

Wyoming

0.0039

Source: National Center for Education Statistics and Census Bureau

lege freshmen in the state’s population.
(On net, New Jersey residents go out of
state for their college education.) The
percentage breakdown is: Delaware
enrolls about 20 percent more freshmen in total; Pennsylvania enrolls 10
percent more; and New Jersey enrolls
www.philadelphiafed.org

about 30 percent less.8
The region’s production of college
graduates is concentrated in certain
metropolitan areas, such as Philadel8
See the reference to the U.S. National Center
for Education Statistics.

phia, State College, and Princeton,
which is in the Trenton metropolitan
area. These centers of education
export their output to the rest of the
world and, in this respect, are similar
to some other well-known educational
centers in the nation, such as Raleigh-

Business Review Q1 2008 19

TABLE 3
Bachelor’s and Higher Degrees Awarded Relative to Total Population (2005)
State

Degrees per 1,000 Population

State

Degrees per 1,000 Population

District of Columbia

38.69

Ohio

7.22

Massachusetts

12.56

Oklahoma

7.19

Rhode Island

11.39

West Virginia

7.14

Vermont

11.07

Virginia

7.10

Utah

10.45

Montana

6.99

North Dakota

10.42

Oregon

6.77

Nebraska

9.83

Louisiana

6.76

Missouri

9.57

Idaho

6.45

New York

9.53

Maine

6.36

Delaware

9.30

Kentucky

6.24

Iowa

9.26

Washington

6.23

Pennsylvania

9.02

North Carolina

6.23

Minnesota

8.64

Tennessee

6.18

New Hampshire

8.56

Hawaii

6.00

Arizona

8.54

California

5.99

Kansas

8.40

South Carolina

5.90

Illinois

8.40

New Mexico

5.77

Colorado

8.20

Georgia

5.72

South Dakota

8.10

Texas

5.67

Indiana

8.10

Mississippi

5.66

Michigan

7.77

Florida

5.43

Connecticut

7.74

New Jersey

5.43

Wisconsin

7.57

Arkansas

5.32

Alabama

7.30

Wyoming

4.57

Total

7.25

Nevada

3.29

Maryland

7.23

Alaska

3.17

Source: National Center for Education Statistics and Census Bureau

Durham and Boston. Like these other
areas, the centers of higher education
in the region do not obtain all of their
“raw material” locally, nor do they
“consume” all of their “finished products” locally. For example, Pennsylvania imports a significant portion of

20 Q1 2008 Business Review

the raw material for producing college
graduates — it has the second highest
number of enrolled college freshmen
from out of state — and the college
graduates, the “finished products,” are
re-exported. Delaware also re-exports
college graduates, but a much smaller

number than Pennsylvania.
As the data described earlier
demonstrate, Pennsylvania produces
a large number of college graduates.
Because Pennsylvania supplies workers
with undergraduate and advanced degrees for the nation and many foreign
www.philadelphiafed.org

countries, it is not surprising that it
does not retain a large share of them.
Indeed, only four of the top 10 states
ranked by degrees awarded (adjusted
for total population) also rank among
the top 10 states in the percentage of
population with a bachelor’s degree or
higher. This fact suggests that there
is not a strong relationship between
the number of degrees awarded in a
state and the proportion of the state’s
population holding degrees. Empirical research supports this impression.
One statistical estimate indicates that
the percentage increase in a state’s
college-educated population will be
only about one-third of the percentage
increase in its production of college
graduates in the long run.9 Obviously,
merely producing college graduates in a
state does not guarantee that they will
remain there.
ATTRACTING GRADUATES:
AMENITIES AND JOBS
If producing college graduates in
a state does not result in a commensurate increase in college graduates
among that state’s population, we need
to look beyond the supply side to find
ways to increase the number of college
graduates in a state or a metropolitan
area’s population and labor force.
If we look beyond the supply side,
what do we observe on the demand
side? The importance of the demand
side can be clearly seen within our
region in the contrast between New
Jersey and Pennsylvania. New Jersey is
the leading state in providing college
freshmen to other states, but the high
percentage of college graduates among
its population indicates that New
Jersey attracts college graduates even
if many of them have been educated
outside the state.

States and metropolitan areas
seeking to increase their collegeeducated populations need to consider
two major aspects of the demand side:
the amenity aspect, which relates to
which features of an area are attractive
to college graduates, and the economic
aspect, which relates to which areas
have high demand for college-educated
workers. The amenities most promi-

Large metropolitan areas are more likely
than small ones to possess the amenities
and economic prospects that attract college
graduates.
nently highlighted by survey research
and analyses of population movements
are those associated with cultural and
recreational opportunities and warm,
dry climates.10 The economic aspect
is related to job opportunities and
salaries.
Various studies around the country have identified specific examples of
these two aspects that are important to
college graduates. A survey of Philadelphia-area college graduates discovered
that the availability and affordability
of housing are features of the area that
are important to graduates who remain
here; geographic location, job opportunities, recreation, and climate are
features that are important to graduates who leave the area.11 These results
match those of surveys conducted in
other states and metropolitan areas.12
They are also consistent with research on college graduates’ interstate

10

See the article by Richard Florida.

11
See the reference to the Knowledge Industry
Partnership.

9

See the article by John Bound, Jeffrey Groen,
Gabor Kezdi, and Sarah Turner.

www.philadelphiafed.org

moves, which reveals that they tend
to leave states that have low employment growth, high unemployment, or
low pay and move to states that score
higher on one or more of these measures, with net migration to the South
Atlantic and Mountain states.13
Large metropolitan areas are more
likely than small ones to possess the
amenities and economic prospects

that attract college graduates. Data for
metropolitan areas indicate that the
percentage of the population with a
bachelor’s degree or higher is greater
in large metropolitan areas throughout
the nation than it is in small areas.
This is true for the three-state region.
Four of the 21 metropolitan areas that
are wholly or partially in the region
have above-average percentages of
population with a bachelor’s degree or
higher. Two of these are among the
largest: the New York metropolitan
area, which includes northern New
Jersey, and the Philadelphia metropolitan area (Table 4). These two
metropolitan areas are economically
diverse, and many firms that need
college-educated workers are located
there. The other two areas, which are
the highest ranked by this measure
among areas in the three-state region,
are State College, PA, and Trenton,
NJ. In both of these areas, colleges
and universities make up a large portion of the employment base, and the
large share of faculty and students
among the areas’ populations boosts

12

See the publication by Carnegie Mellon
University.

13

See the article by Yolanda Kodrzycki.

Business Review Q1 2008 21

TABLE 4
Percent of Population 25 Years and Older
with a Bachelor's Degree or Higher (2005)
Metropolitan Area

Percent

State College, PA

40.2

Trenton-Ewing, NJ

37.7

New York-Northern New Jersey-Long Island, NY-NJ-PA

34.8

Philadelphia-Camden-Wilmington, PA-NJ-DE-MD

31.7

Total U.S. Metropolitan Area Population

30.1

Harrisburg-Carlisle, PA

28.1

Pittsburgh, PA

27.1

Ocean City, NJ

26.3

Allentown-Bethlehem-Easton, PA-NJ

25.0

Erie, PA

24.2

Atlantic City, NJ

23.9

Lancaster, PA

23.0

Reading, PA

21.4

Scranton-Wilkes-Barre, PA

20.9

York-Hanover, PA

19.2

Lebanon, PA

18.9

Johnstown, PA

17.9

Dover, DE

17.6

Youngstown-Warren-Boardman, OH-PA

17.3

Williamsport, PA

16.9

Altoona, PA

15.8

Vineland-Millville-Bridgeton, NJ

15.7

Source: Census Bureau

their percentage of college-educated
residents. “College town” metropolitan
areas rank high among all areas in the
region by percentage of the population
with a bachelor’s degree, as such towns
do throughout the nation.
The high proportion of college
graduates in the New York-Northern
New Jersey metropolitan area clearly
drives up the statewide proportion,

22 Q1 2008 Business Review

demonstrating the influence of both
the amenity and economic aspects
of demand. (The availability of some
amenities, especially cultural ones, not
only results from the concentration of
college graduates but fosters such concentrations as well, because concentrations of college graduates constitute
a large market for cultural amenities,
which, in turn, attracts providers of

such amenities.) The area provides
cultural and recreational amenities,
and its concentrations of industries
with large and growing needs for
college-educated workers provide the
economic aspect, serving as sources of
demand for college graduates.
The Philadelphia metropolitan
area serves a similar role at the other
end of New Jersey and for southeastern
Pennsylvania. In fact, the Philadelphia
area has the high percentage of college
graduates that is typical of large metropolitan areas (Table 5). But it does not
figure as prominently in the statewide
picture in Pennsylvania as the New
Jersey portions of the New YorkNorthern New Jersey and the Philadelphia metropolitan areas do in New
Jersey. Consequently, the statewide
percentage in Pennsylvania is near the
national average, while the New Jersey
statewide percentage is above it.
If we examine the demand for
college graduates as indicated by
employment growth, the data indicate
that job growth for occupations that
typically require a college education
has been slower in Pennsylvania than
in the nation for several years, while it
has been faster in New Jersey.14 These
occupations are those in management,
business and finance operations, computers and mathematics, architecture
and engineering, sciences, community
and social service, legal, education,
arts and media, and health care. Obviously, many of these occupations are
more in demand in urban areas than
in rural areas. New Jersey, being more
densely urbanized than Pennsylvania,
will therefore have a greater base of
demand for these occupations, but the
difference in growth rates is striking.
From 1999 (when current occupational
definitions were established) until

14
See the 2007b reference to the U.S. Bureau of
Labor Statistics.

www.philadelphiafed.org

TABLE 5
Percent of Population 25 Years and Older
with a Bachelor's Degree or Higher (2005)
Ten Largest Metropolitan Areas
Metropolitan Area

Percent

Washington-Arlington-Alexandria, DC-VA-MD-WV

45.9

New York-Northern New Jersey-Long Island, NY-NJ-PA

34.8

Atlanta-Sandy Springs-Marietta, GA

34.3

Chicago-Napierville-Joliet, IL-IN-WI

32.1

Philadelphia-Camden-Wilmington, PA-NJ-DE-MD

31.7

Total U.S. Metropolitan Area Population

30.1

Dallas-Ft. Worth-Arlington, TX

30.0

Los Angeles-Long Beach-Santa Ana, CA

29.4

Houston-Sugar Land-Baytown, TX

27.8

Miami-Ft. Lauderdale-Miami Beach, FL

27.5

Detroit-Warren-Livonia, MI

26.4

Source: Census Bureau

FIGURE
Percent of Population 25 Years and Older with
Bachelor’s Degree or Higher
Percent

NJ
DE
US

PA

1990

1992

1994

Source: Census Bureau

www.philadelphiafed.org

1996

1998

2000

2002

2004

2006, employment in these occupations increased 13 percent in New
Jersey compared with 4 percent in
Pennsylvania (the national increase
was 6 percent).15 Although Pennsylvania produces large numbers of college
graduates, slow job growth for college
graduates in the state reflects, at least
in part, a relatively weaker demand for
them. New Jersey’s demand for college
graduates brings them into the state
(for the first time or as returning residents) even if they did not receive their
college education there.
REGIONAL EFFORTS TO BOOST
THE COLLEGE-EDUCATED
POPULATION
The gap between a college
graduate’s income and the income of
someone who hasn’t completed college
has been increasing, and the percentage of the population with a bachelor’s
degree or higher has been rising for
many years throughout the country. In
the past decade and a half, educational
attainment, by this measure, has increased somewhat more in Pennsylvania and New Jersey than in the nation,
but it has slipped slightly in Delaware
(see the Figure). Despite the general increase in the proportion of the
population with a college education,
there is still a great deal of variation in
this measure around the nation (Table
6). Consequently, regional variations
in educational attainment are increasingly influencing regional variations in
income.
The positive impact of an educated population on regional income
and economic growth is well known
to governments, businesses, and civic
groups around the country, and they
are making efforts to attract and retain
college students and graduates. This is
not an easy task, since recent college

2006
15
Data on occupational employment for
Delaware are not complete for these years.

Business Review Q1 2008 23

TABLE 6
Percent of Population 25 years and Older with a Bachelor’s Degree
or Higher (Ranked in 2006)
State

2006

2000

1990

State

2006

2000

1990

District of Columbia

49.1

38.3

33.3

New Mexico

26.7

23.6

20.4

Massachusetts

40.4

32.7

27.2

Pennsylvania

26.6

24.3

17.9

Colorado

36.4

34.6

27.0

Delaware

26.2

24.0

21.4

Connecticut

36.0

31.6

27.2

Michigan

26.1

23.0

17.3

Maryland

35.7

32.3

26.5

North Carolina

25.6

23.2

17.4

New Jersey

35.6

30.1

24.8

Texas

25.5

23.9

20.4

Vermont

34.0

28.8

24.3

South Dakota

25.3

25.7

17.2

Minnesota

33.5

31.2

21.9

Idaho

25.1

20.0

17.7

Hawaii

32.3

26.3

22.9

Montana

25.1

23.8

19.8

New York

32.2

28.7

23.1

Iowa

24.7

25.5

16.9

New Hampshire

32.1

30.1

24.3

Wisconsin

24.6

23.8

17.7

Virginia

32.1

31.9

24.5

Arizona

24.5

24.6

20.3

Kansas

31.6

27.3

21.1

Missouri

24.3

26.2

17.8

Washington

31.4

28.6

22.9

Ohio

23.3

24.6

17.0

Illinois

31.2

27.1

21.1

Oklahoma

22.9

22.5

17.8

Rhode Island

30.9

26.4

21.3

South Carolina

22.6

19.0

16.6

California

29.8

27.5

23.4

Tennessee

22.0

22.0

15.9

North Dakota

28.7

22.6

18.0

Indiana

21.9

17.1

15.6

Oregon

28.3

27.2

20.6

Louisiana

21.2

22.5

16.1

Georgia

28.1

23.1

19.3

Mississippi

21.1

18.7

14.8

U.S.

28.0

25.6

20.3

Alabama

20.8

20.4

15.6

Alaska

27.7

28.1

23.0

Nevada

20.8

19.3

15.3

Florida

27.2

22.8

18.3

Wyoming

20.8

20.6

18.8

Nebraska

27.2

24.6

19.0

Kentucky

20.2

20.5

13.6

Utah

27.0

26.4

22.2

Arkansas

19.0

18.4

13.4

Maine

26.9

24.1

18.8

West Virginia

15.9

15.3

12.3

Source: Census Bureau
24 Q1 2008 Business Review

www.philadelphiafed.org

graduates are among the most mobile
sectors of the population. States and
cities are using a variety of methods
to increase enrollment at colleges and
universities in their areas and to retain
graduates.16 These methods include
scholarships, marketing efforts, and
internships, among others. Programs
at the state and local levels around the
nation as well as in the region foster
internships, collaboration between colleges and industry, and new business
formation focused on college graduates.17 The Philadelphia Knowledge
Industry Partnership is spearheading
efforts in the Philadelphia region, and
there are programs in other large cities
in the District.
It is important for regional efforts
aimed at increasing the number of college-educated workers to concentrate
on the economic aspect of the demand
side by encouraging job growth focused
on industries and occupations that use
college graduates. This is clearly evident in our region: New Jersey ranks
high in college graduates, attracting
them from out-of-state colleges as its
highly educated labor force grows.
Demographic studies and surveys both

16

See the article by George Smith.

show that job opportunities are powerful determinants of college graduates’
location decisions, especially for those
more inclined to move from one area
to another; so it makes sense to focus
efforts to attract college graduates on
this factor.
Programs that succeed in attracting and retaining college graduates
can benefit the regions that undertake
them. But promoters of such programs
must keep in mind that some areas,
such as those in the regions mentioned
earlier, are — and are likely to remain
— exporters of college graduates, with
the associated relatively low retention
rate. Nevertheless, colleges and universities that send a relatively large share
of their graduates elsewhere still provide several important benefits to the
local economy.18 The college itself is a
source of employment. Both students
and faculty raise the educational attainment level of the local population
(demonstrated in our region by the
high percentage of college-educated
residents in the State College, PA, and
Trenton, NJ areas). The area can also
serve as a source of supply of collegeeducated workers for local employers,
even if most graduates go elsewhere.
But it is perhaps more appropriate
to view these areas as export centers,

17

See the references to the Pennsylvania
Economy League and the Knowledge Industry
Partnership.

www.philadelphiafed.org

18

See the reference to the Federal Reserve Bank
of Atlanta.

rather than local sources of supply
and, in turn, to view the region in
which they are located as an import
destination. If we view the situation
in this way, raising the percentage of
the college-educated population in the
region would best be accomplished by
raising demand for college graduates
— primarily by stimulating growth of
jobs requiring a college education (or
higher), not by raising supply through
efforts narrowly aimed at retaining or
attracting college graduates.
CONCLUSION
Education is an investment in
human capital that pays individual and
social dividends. The percentage of an
area’s population that has a bachelor’s
degree or higher is positively associated with the area’s total income and
growth. Recognizing this, civic leaders
in many areas of the country, including our region, are making efforts to
attract and retain college graduates.
Research shows that employment
opportunities are a key element for
successful attraction and retention
efforts. Thus, programs to boost the
college-educated population should not
be narrowly focused on the education
sector but should include broader efforts to boost employment growth, especially for occupations and industries
that require workers with bachelor’s
degrees and higher. BR

Business Review Q1 2008 25

REFERENCES

Becker, Gary S. Human Capital: A
Theoretical and Empirical Analysis, with
Special Reference to Education, Second
Edition. New York: Columbia University
Press, 1975.
Bound, John, Jeffrey Groen, Gabor Kezdi,
and Sarah Turner. “Trade in University
Training: Cross-State Variation in the
Production and Use of College-Educated
Labor,” Working Paper 8555, National
Bureau of Economic Research (2001).
Carlino, Gerald A. “Knowledge Spillovers:
Cities’ Role in the New Economy,” Federal
Reserve Bank of Philadelphia Business
Review (Fourth Quarter 2001), pp. 17-26.
Carnegie Mellon University, Center for
Economic Development. Plugging the Brain
Drain: A Review of Studies and Issues for
Attracting and Retaining Talent (2001).
Federal Reserve Bank of Atlanta. “Higher
Education Translates into Big Business,”
EconSouth (Second Quarter 2000), pp.
8-13.

Iranzo, Susana, and Giovanni Peri.
“Schooling Externalities, Technology
and Productivity: Theory and Evidence
from U.S. States,” Working Paper 12440,
National Bureau of Economic Research
(2006).
Knowledge Industry Partnership. www.
kiponline.org (accessed August 3, 2007).
Knowledge Industry Partnership. Should
I Stay or Should I Go? Survey of Recent
College Graduates (2004).
Kodrzycki, Yolanda K. “Migration of
Recent College Graduates: Evidence
from the National Longitudinal Survey of
Youth,” Federal Reserve Bank of Boston
New England Economic Review (January/
February 2001), pp. 13-34.
Mincer, Jacob. Schooling, Experience, and
Earnings. New York: National Bureau of
Economic Research, 1974.

Florida, Richard. The Rise of the Creative
Class. New York: Basic Books, 2002.

Moretti, Enrico. “Estimating the Social
Return to Higher Education: Evidence
from Longitudinal and Repeated CrossSectional Data,” Journal of Econometrics,
121 (2004), pp. 175-212.

Glaeser, Edward L., and David C. Maré.
“Cities and Skills,” Journal of Labor
Economics, 19 (2001), pp. 316-42.

Pennsylvania Economy League-Eastern
Division. Greater Philadelphia’s Knowledge
Industry (2000).

Glaeser, Edward L., Jose A. Scheinkman,
and Andrei Shleifer. “Economic Growth
in a Cross-section of Cities,” Journal of
Monetary Economics, 36 (1995), pp. 117-43.

Rauch, James E. “Productivity Gains from
Geographic Concentrations of Human
Capital: Evidence from the Cities,” Journal
of Urban Economics, 34 (1993), pp. 380400.

Heckman, James J., Lance J. Lochner,
and Petra E. Todd. “Fifty Years of Mincer
Earnings Regressions,” Working Paper
9732, National Bureau of Economic
Research (2003).

26 Q1 2008 Business Review

Sill, Keith. “Widening the Wage Gap: The
Skill Premium and Technology,” Federal
Reserve Bank of Philadelphia Business
Review (Fourth Quarter 2002), pp. 25-32.
Simon, Curtis J., and Clark Nardinelli.
“Human Capital and the Rise of American
Cities, 1900-1990,” Regional Science and
Urban Economics, 32 (2002), pp. 59-96.
Smith, George. “Collaborative Efforts in
Student Retention: A Snapshot of Best
Practices Across the Nation,” Knowledge
Industry Partnership, www.kiponline.
org/research_bestpractices.htm (accessed
August 3, 2007).
Topel, Robert. “The Private and Social
Values of Education,” in Proceedings of
Education and Economic Development:
A Conference. Federal Reserve Bank of
Cleveland, November 18-19, 2004, pp.
47-57.
United States Bureau of Labor Statistics.
“Education Pays,” www.bls.gov/emp/
emptab7.htm (accessed August 15, 2007).
United States Bureau of Labor Statistics.
Occupational Employment Statistics. www.
bls.gov/oes/home.htm (accessed August 20,
2007).
United States National Center for
Education Statistics. Digest of Education
Statistics, 2006.
Wheeler, Christopher H. “Human
Capital Growth in a Cross Section of U.S.
Metropolitan Areas,” Federal Reserve
Bank of St. Louis Review (March/April
2006), pp. 113-32.

www.philadelphiafed.org

Recent Developments in
Consumer Credit and Payments
BY MITCHELL BERLIN

O

n September 20-21, 2007, the Research
Department and the Payment Cards Center
of the Federal Reserve Bank of Philadelphia
held their fourth joint conference to present
and discuss the latest research on consumer credit
payments. Approximately 75 participants attended the
conference, which included six research papers on topics
such as liquidity constraints, the rise in bankruptcy, and
the financial mistakes made by credit card holders. In this
article, Mitchell Berlin summarizes the papers presented at
the conference.

In his opening remarks at the
conference, Charles Plosser, president of the Federal Reserve Bank of
Philadelphia, noted that innovation in
electronic payments has led to major
changes in the financial industry. The
process of innovation has allowed new
entrants into the industry, expanding the availability of consumer credit
and permitting more opportunities
for smoothing consumption over
time. Plosser reminded conference
participants that rapid growth in in-

Mitchell Berlin
is a vice president
and economist in
the Philadelphia
Fed’s Research
Department. He
is also head of the
Banking section.
This article is
available free of
charge at www.philadelphiafed.org/econ/br/.
www.philadelphiafed.org

novation often leads to excesses and
mistakes and that progress is necessarily uneven. He stressed that the Fed’s
mandate is to evaluate innovations in
the context of economic efficiency,
effective monetary policy, and an efficient payments system. This mandate
provides the rationale for this conference, which brought together researchers whose papers address fundamental
issues about consumer credit.
LIQUIDITY CONSTRAINTS
In the first paper, Jonathan
Levin of Stanford University reported
the results of a study (with William
Adams and Liran Einav) that
provided evidence for the economic
significance of liquidity constraints in
the market for subprime auto loans.
The authors also sought to uncover the
underlying sources of these constraints.
Broadly, liquidity constraints refer
to limits on an individual’s ability to

borrow because of various frictions in
credit markets, especially those due
to incentive problems that arise when
borrowers are better informed than
lenders about their risk of default.
When such borrowing limits are
significant, an individual’s ability to
make purchases depends heavily on his
or her cash on hand.
Levin and his co-authors examined a sample of applications for
loans at a large subprime auto lender
between June 2001 and December
2004. In addition, they examined the
details of the loan contracts for the applications that were accepted and the
repayment history on all loans through
April 2006.
First, they examined general
borrowing patterns for evidence of
liquidity constraints. They found that
44 percent of car buyers made the minimum down payment; that is, a large
share of buyers borrowed no more than
the absolute minimum, even though
a higher down payment would have
reduced their loan rate significantly.
Strikingly, the authors found that
both applications and sales revealed
a marked spike in February. Levin
explained that February is the time
of year when consumers receive tax
rebates and have more cash on hand to
make a purchase.1 When the authors
split their sample into customers who
were eligible for the earned income
tax credit and those who weren’t, the
February spike remained only for those
who were eligible. This provided fur-

1

He noted that this explanation for the timeof-year effect was actually provided by the
subprime lender.
Business Review Q1 2008 27

ther support for the tax rebate explanation for the spike in the data.
Levin and co-authors then
turned to formal econometric tests
for evidence of liquidity constraints.
They estimated the distinct effects of
higher loan payments (measured by a
higher car price) and of higher minimum down payments on customers’
probability of purchasing. Levin and
co-authors argued that a customer who
is not liquidity constrained would care
only about the present value of total
loan payments: A dollar spent today
to cover the down payment should
have the same effect on the borrower’s
purchasing decision as an appropriately
discounted dollar spent tomorrow
to repay the loan. On the contrary,
they found that a $100 increase in the
minimum down payment had the same
effect on the probability of purchase
as a $900 increase in the car price,
evidence that purchase decisions were
strongly affected by customers’ ability
to come up with the initial cash. Levin
and co-authors argued that the alternative explanation — that customers
discount future car payments at an
annual rate of 427 percent — was
implausible.
Next, the authors tried to uncover
the underlying sources of liquidity
constraints, in particular, the relative
effects of adverse selection and moral
hazard. The authors defined adverse
selection as the tendency for borrowers who have a higher risk of default
to take out larger loans, while they defined moral hazard as the tendency for
borrowers with larger loans to default
more often.2 In either case, contracting is made more difficult when the
borrower is better informed than the
lender about his or her risk of default.

To disentangle the effects of adverse
selection and moral hazard, the authors first estimated a Tobit model of
customers’ desired down payment and
found that observably riskier customers — for example, customers with low
credit scores or lower incomes — had a

Levin and co-authors suggested that
improvements in credit rating technologies
probably played an important role in the strong
growth of subprime markets in the 1990s.
lower desired down payment, a finding
consistent with adverse selection.
The authors then estimated the
effect of larger loan size on the probability of default. They argued that
this effect includes both moral hazard
— the higher probability of default
due to larger loan size — and adverse
selection — the tendency for riskier
borrowers to take out larger loans.
The authors proposed the following
procedure to disentangle these effects.
Since the explanatory variables
used to estimate customers’ desired
minimum down payment included
most of the observable factors that a
lender would use to estimate a borrower’s risk, the authors argued that
the residual from the Tobit regression
was a measure of the borrower’s private
information, including the borrower’s
private information about his or her
probability of default.3 This residual
could then be included along with the
loan size (and other control variables)
in a regression that explained the probability of default; the authors interpreted the coefficient on loan size as the

3

2
Note that, in this context, both adverse
selection and moral hazard operate through the
size of the borrower’s loan.

28 Q1 2008 Business Review

moral hazard effect and the coefficient
on the residual as the adverse selection effect. Regression results provided
evidence for both adverse selection
and moral hazard but showed that
moral hazard was twice as important
quantitatively.

In this context, the residual refers to the
portion of the customer’s down payment
decision that can’t be explained by factors that
the researcher can observe and include in the
regression, such as the customer’s FICO score.

Their regressions also showed
that the customer’s FICO score had a
very strong relationship to the probability of default; that is, observing
the borrower’s credit rating provided
lenders with a lot of information about
borrower risk. Levin and co-authors
suggested that improvements in credit
rating technologies probably played an
important role in the strong growth of
subprime markets in the 1990s.
THE RISE OF HOUSEHOLD
BANKRUPTCY
The next speaker, Borghan Narajabad of Rice University, discussed the
results of his work on the underlying
causes of the increase in consumer
bankruptcies in the mid-1990s. He
argued that prior research had failed
to adequately explain why the rise in
bankruptcies coincided with other
developments in credit markets. In
particular, the 1990s had also witnessed a significant rise in credit card
debt and usage and increased variation
in credit terms offered to customers.
His theoretical model was designed
to yield these empirical predictions in
addition to the rise in bankruptcies.
According to Narajabad’s explanation,
an improvement in lenders’ screening
technology permitted them to better
differentiate high-risk from low-risk

www.philadelphiafed.org

borrowers. In turn, lenders could
profitably offer more credit to all borrowers, but it was profitable to provide
the largest increases in credit limits to
lower risk borrowers. Also, the general
increase in the availability of credit
increased both credit card usage and
the number of bankruptcies.
The main elements of Narajabad’s
stylized theoretical model were: (i)
individuals have the need for credit
to cover their consumption needs for
an uncertain amount of time; (ii) they
are differentiated according to the risk
that their income would remain low
for a long time; (iii) individuals know
more about their underlying risk than
lenders; (iv) borrowers have an incentive to default on loans when the debt
burden is large compared with their
costs of defaulting; and (iv) at some
cost, lenders can screen borrowers and
become better informed about the borrower’s risk type.
Outlining the underlying logic of
his model, Narajabad first analyzed the
borrowing decision. He explained that
the amount borrowed in each period
prior to the (uncertain) time when the
borrower could repay was determined
by the following marginal condition.
The marginal utility of higher consumption financed by borrowing must
equal the marginal cost of borrowing.
This marginal cost has two elements:
first, the loss of future borrowing
capacity as the borrower moved nearer
to his or her credit limit and, second,
the higher debt payments once the
borrower has the capacity to repay.
From this marginal condition,
Narajabad showed that the model generated two of the patterns observed in
the data. The model predicted that an
increase in the credit limit would lead
to an increase in borrowing, mainly
because the constraint on future borrowing capacity is relaxed when the
credit limit is increased. The model
also predicted a rise in bankruptcies

www.philadelphiafed.org

with an increase in borrowing limits.
Since a borrower can reach the credit
limit before having enough income
to repay existing loans, he or she may
choose to default if the debt load is
sufficiently high.
Turning to the lender’s decision,
Narajabad explained that his model
predicted that an improvement in the
lender’s screening technology induced
lenders to increase borrowers’ credit
limits, with a disproportionate increase
in the credit limits for lower risk borrowers. Thus, the model explained
the increased variation in credit terms
observed in the data.
Narajabad then turned to a quantitative exercise to see how well the
model actually matched the data. The
estimation technique seeks to match
selected statistics describing consumer
use of credit cards in the 1990s, as
measured by the Survey of Consumer
Finances from 1992 and 1998. These
statistics included the ratio of credit
limits to income in both years, the
ratio of credit card debt to income in
both years, default rates for 1992, and
the variance of credit limits in the two
years. Narajabad explained that he explicitly chose not to match default rates
in 1998 when he estimated the model.
The model’s ability to match the actual
rise in defaults would be an important
test of its success.
Narajabad concluded that the
model was broadly successful in matching the data. He found that the model
could generate approximately one-third
of the increase in defaults from 1992
to 1998. Narajabad also explained that
his model rejected alternative explanations for the increase in bankruptcies.
A reduction in the stigma attached to
filing for bankruptcy predicts a counterfactual decline in credit to higher
risk borrowers, while a reduction in the
transaction costs of lending does not
predict the greater variation in credit
limits across different customers.

WHO MAKES MISTAKES?
Barry Scholnick of the University
of Alberta discussed the results of
his study (with Nadia Massoud and
Anthony Saunders) of financial
mistakes made by credit card holders.
They examined the prevalence of
certain types of mistakes, as well as
the types of customers who made
these mistakes. The main question
motivating their study was whether
mistakes were made predominantly by
wealthy customers, who might make
mistakes because the impact on their
total wealth is trivial, or by poor and
less educated customers, who might
make mistakes because of a lack of
financial sophistication.
Scholnick and his co-authors
constructed a database extending from
December 2004 to June 2006 that
combined: (i) confidential data (from
a Canadian bank) about individual
cardholders that included customers’
credit card accounts, deposit accounts,
and credit scores; (ii) demographic
information about the individuals
in a customer’s postal code, which
the authors used as a proxy for the
individual customer’s demographic
traits;4 and (iii) information about
residential property transactions in the
postal code. Scholnick emphasized
the unique features of this data set.
The small number of households in the
Canadian postal zones (approximately
200) minimizes the measurement error
created by using an aggregate in place
of the individual’s actual wealth.5 Furthermore, monthly data on customer
balances provided a detailed picture
of the evolution of customers’ liquid
wealth holdings over time. The authors
viewed the comprehensiveness and
4

To protect customers’ privacy, the bank
identified customers’ postal zones but not their
addresses.

5

By comparison, U.S. ZIP codes have 10,000
households.

Business Review Q1 2008 29

detail of this data set as one of the
paper’s main contributions.
The authors considered four types
of mistakes: a cash advance, a delinquent payment, a transaction that
exceeds the credit limit, and a simultaneous delinquency and overlimit in
a single month. These mistakes range
from frictional to moderately consequential. The cash advance triggers
only a moderate fee, while the other
mistakes may affect a consumer’s credit
report in addition to triggering a fee.
In each case, the authors considered
the transaction a mistake only if the
customer had adequate bank balances
to avoid the penalty, for example, if the
customer could have avoided a delinquency by making a payment from a
deposit account.6
The authors showed that a significant fraction of total transactions were
mistakes. For example, while delinquencies occurred in 10.3 percent of
observations, mistakes accounted for 4
percent of the observations (adjusting
for precautionary balances). In addition, they found that consumers make
consequential mistakes more often
than frictional mistakes. The authors
argued that this provides evidence
that the mistakes were not caused by
rational inattention. If customers were
simply not paying attention because it
was not worth their time, the authors
expected frictional mistakes to be
made more often than more costly
mistakes.
The authors then turned to the
question: Who makes credit card
mistakes? The authors estimated
panel logit regressions for 75,000
customers, a separate one for each

6

Of course, a customer might be delinquent
or go over the limit on credit card payments
to maintain his or her bank balance above
some level, so-called precautionary balances.
Accordingly, the authors used various
definitions of mistakes, each corresponding to a
different measure of precautionary balances.

30 Q1 2008 Business Review

type of mistake and for each definition of precautionary cash balances.
In general, the authors found that less
wealthy cardholders were more likely
to make mistakes. More specifically,
renters were significantly more likely
to make mistakes than homeowners,
and individuals with more business
and investment income were, for the
most part, significantly less likely to
make mistakes. Those individuals with

Consumers make
consequential
mistakes more
often than frictional
mistakes.
a larger share of total income derived
from government payments — another
indicator of lower wealth — were more
likely to make mistakes. Scholnick
argued that these results were not
consistent with the view that mistakes
were mainly committed by wealthier
customers, rationally allocating their
attention.
Although individuals with higher
assessed risk were more likely to make
mistakes, the authors found no evidence that mistakes were associated
with subsequent defaults. Since mistakes typically trigger fees, the authors
argued that this result is inconsistent
with bankers’ claims that fees are
assessed to compensate the bank for
defaults.
THE AGE OF REASON
John Driscoll of the Federal
Reserve Board reported on recent
research (with Sumit Agarwal,
Xavier Gabaix, and David Laibson)
that explored the pattern of financial
decision-making over an individual’s
lifetime. He and his co-authors found
that across a wide range of financial

transactions, the quality of financial
decisions followed a U-shaped pattern; that is, financial decision-making
improved until an individual reached
his or her 50s and then declined as
he or she aged. Driscoll and his coauthors argued that this age of reason
effect could be explained by a model in
which an individual’s analytic capabilities decline roughly linearly from
age 20 onward, while experience with
financial matters increases thoughout
the individual’s life, but at a decreasing
rate over time. The net effect yields
improvements in financial performance until (roughly) age 53; beyond
this age, the decline in cognitive ability dominates.
Using proprietary data sets from
a national financial institution, the
authors considered financial decisionmaking in 10 separate contexts,
including a number of decisions involving home equity loans, auto loans, and
credit cards. In addition to providing
information about the terms of the
financial transaction, for example,
fees and rates, the data sets include
substantial demographic information
about the individuals.
Driscoll provided a detailed
discussion of the empirical results for
home equity loans and home equity
lines of credit. The authors examined
the average annual percentage rate
paid by borrowers in each of six age
buckets, controlling for various demographic characteristics and various
measures related to a borrower’s risk
of default, including FICO score. The
authors found that the rate paid by
borrowers followed a U-shaped pattern,
declining continuously until age 50
to 60 and rising subsequently. This is
precisely the same pattern the authors
discovered for many other products,
but for home equity loans and lines of
credit, the authors had additional evidence supporting their hypothesis that
this U-shaped pattern was related to

www.philadelphiafed.org

the quality of the borrower’s financial
decision-making.
Specifically, the loan rate depended on the borrower’s loan-to-value
ratio (LTV); the lender charged a
higher rate for higher LTVs, although
it increased in discrete jumps. As part
of the application, the lender required
borrowers to estimate the value of
their homes, and the lender subsequently performed its own appraisal.
If the lender’s estimated LTV was
significantly higher than the borrower’s
estimated LTV, the loan officer would
direct the borrower to a higher-priced
loan. But if the lender’s estimated
LTV was significantly lower than the
borrower’s estimated LTV, the loan
officer would not direct the borrower
to a lower-priced loan. The authors
defined a rate-changing mistake as one
in which the borrower’s estimate was
significantly different from the lender’s
estimate, and they found that such
mistakes led to an average increase
of 125 basis points for loans and 150
basis points for lines of credit (holding constant the borrower’s risk and
other demographic characteristics).
The authors found that the U-shaped
pattern existed only for customers who
made rate-changing mistakes. This
supported the authors’ claim that the
quality of the customer’s financial
decision-making — in this case, the
ability to accurately value one’s house
— underpins the higher loan rates
paid by the young and the old.
The authors also studied balance
transfer offers in which customers
received low teaser rates for balances
transferred to a new card. However,
this rate applied only to the balances
transferred; all new purchases were
charged a high rate, and all payments
on the new card were applied first
to the transferred balances. For the
customer, the optimal strategy during
the teaser rate period is to make all
purchases and payments on the old

www.philadelphiafed.org

card. The authors found that one-third
of the customers who transferred balances identified the optimal strategy
within the first month, one-third figured out the strategy before the sixth
month, and one-third never learned
during the teaser rate period. They
also found that the percent of borrowers who discovered the optimal policy
at some point was first increasing in
age and then decreasing in age — an
inverse U-shape — with the highest
percentage for borrowers between ages
35 and 44. On the other hand, the

an earlier period is inconsistent with
the argument that cohort effects were
driving results.
IS IT OPTIMAL TO FORGET
DEFAULTS?
Ronel Elul of the Federal Reserve
Bank of Philadelphia presented his
research (with Piero Gottardi) that
examined the rationale for laws requiring bankruptcies to be erased from an
individual’s credit files. Elul explained
that the Fair Credit Reporting Act
(FCRA) requires bankruptcies to be

For the customer, the optimal strategy during
the teaser rate period is to make all purchases
and payments on the old card.
fraction that never learned the optimal
strategy displayed a U-shape, again
with the lowest percentage for borrowers between ages 35 and 44. Driscoll
and co-authors’ interpretation of these
results was that middle-aged people
were both most likely to act optimally
and least likely to remain permanently
confused. They also argued that the
somewhat younger peak of financial
performance may reflect the greater
analytic skill required to determine the
optimal strategy.
Driscoll said that he and his coauthors had explored and rejected a
number of alternative explanations for
the U-shaped pattern. In particular,
the pattern could not be explained by
age-related variation in default risk or
by borrowing to meet medical expenses. Driscoll argued that they could
not rule out cohort effects but that for
credit card and auto loans the data
indicated the same U-shaped pattern
for 1992 data, 10 years earlier than the
sample considered in the paper. He argued that replicating the U-shape for

expunged after 10 years and that laws
restricting the use of old information
are common outside the U.S. Elul and
Gottardi’s research showed that forgetting is optimal under some conditions
and also that it must be imposed by
government mandate; that is, it would
never arise through private contractual
arrangements.
Elul and his co-author examined a
model with both adverse selection and
moral hazard.7 In particular, there were
two types of borrowers: a safe borrower
who never defaulted and a risky borrower who could lower his or her probability of default by expending costly
effort. The authors assumed that all
loan contracts were single-period contracts, and they focused their attention
on Markov perfect equilibria, those
in which lenders can observe only

7

In this context, adverse selection refers to
borrowers’ being better informed about their
intrinsic risk than lenders when loan contracts
are signed and moral hazard refers to borrowers’
knowing more about risk-taking behavior
subsequent to receiving the loan.

Business Review Q1 2008 31

whether a borrower has defaulted or
repaid a loan in the previous period.8
In their model, reputational concerns lead risky borrowers to exert high
effort. By assumption, a risky borrower
would always choose low effort without
reputational concerns, and no lender
would make a loan to a risky borrower
who chooses low effort. Although
lenders are unable to directly observe
a borrower’s type, they can observe
whether the borrower has defaulted
in the previous period. A default by a
borrower indicates to all lenders that
the borrower is a risky type, and once
a borrower has defaulted he would
automatically be excluded from the
loan market. After a number of initial
periods of low effort, a risky borrower
who has not yet defaulted may choose
to exert high effort to maintain his
or her reputation.9 As long as the
borrower doesn’t default, he or she is
indistinguishable from a safe borrower
and receives the same loan rate.10
Into this setting Elul and Gottardi
introduced a stylized representation
of the FCRA. Once a borrower has
defaulted, the default is stricken from
the record with some probability. The
authors asked: Under what conditions
would introducing a positive probability of forgetting increase consumer
welfare?
Elul explained that the possibility
of forgetting introduces a tradeoff. On

8
Elul and Gottardi argued that such equilibria
are more realistic than those that can arise
when contracts might be conditioned on past
behavior in complicated ways.
9

The risky borrower doesn’t begin to exert
high effort until enough risky borrowers have
defaulted. At this point, lenders assess a high
probability that someone who has not yet
defaulted is a safe borrower. Said differently, the
value of establishing a reputation rises as the
fraction of high-risk borrowers in the population
decreases.

10

That is, the model always yields a pooling
equilibrium.

32 Q1 2008 Business Review

the one hand, forgetting reduces the
risky borrower’s incentive to exert high
effort because it reduces the penalty
for default (exclusion from the loan
market forever). This negative effect
on incentives is manifested as a larger
initial number of periods in which the
risky borrower exerts low effort, before
reputation-building incentives kick
in. However, permitting borrowers to
re-enter (which requires forgetting)

Consumers are quite
sensitive to the price
differential between
fixed- and adjustablerate mortgages.
also has a beneficial effect, because
aggregate output is reduced when risky
borrowers who would have exerted
high effort are excluded from the loan
market.
Elul presented his and Gottardi’s
general result that forgetting will be
efficient under certain conditions.
In particular, they showed that some
amount of forgetting will be efficient
when agents don’t discount the future
too heavily; when gains from exerting
high effort are sufficiently high; when
low effort is not too inefficient; and
when the fraction of low-risk borrowers
is high enough. Under these conditions, the additional output when borrowers re-enter the market outweighs
the negative effect on incentives in
initial periods.
The authors also explained that
forgetting requires a government
mandate; that is, it could not be implemented through private contracts. The
reason is that any lender who unilaterally chose a policy of forgetting would
attract only risky borrowers and would
suffer losses.

SUBSTITUTION BETWEEN
FIXED- AND ADJUSTABLERATE MORTGAGES
In the final presentation of the
conference, James Vickery of the
Federal Reserve Bank of New York
presented the results of his research
into the elasticity of substitution
between fixed-rate mortgages (FRMs)
and adjustable-rate mortgages (ARMs).
He argued that consumers are quite
sensitive to the price differential
between fixed- and adjustable-rate
mortgages; specifically, he found that a
20-basis-point increase in the rate on
FRMs (relative to ARMs) would cause
a 17-percentage-point decline in the
market share for FRMs.
Vickery explained that the regulatory cutoff for conforming mortgages
— the maximum size for loans that
can be purchased and insured by the
government sponsored enterprises
(GSEs) — creates a discontinuity at
the conforming loan limit. He argued
that the supply of fixed-rate mortgages
falls discontinuously at the conforming loan limit because loans can’t be as
easily securitized without a guarantee
from the GSEs. The greater difficulty
of securitizing loans affects the supply of FRMs more than the supply
of ARMs because FRMs subject the
lender to interest rate risk if they are
kept on the lender’s balance sheet. As
long as the relative demand for FRMs
and ARMs is affected by their rates,
but not by the conforming loan limit
per se, the discontinuity permitted
Vickery to identify the demand curve
for FRMs.
Vickery estimated the coefficient
of substitution between fixed- and
adjustable-rate mortgages in two steps:
First, he estimated the change in the
market share of fixed-rate loans at
the conforming loan limit; second, he
estimated the size of the difference in
the rates on conforming and nonconforming loans. The coefficient of

www.philadelphiafed.org

substitution is simply the ratio of the
change in market share of FRMs to a
given difference in the rates.
To carry out the first step, he estimated an empirical model in which the
probability of a loan’s having a fixed
rate depends on the relative rates on
FRMs and ARMs — permitting the
relationship to differ before and after
the conforming loan limit — as well as
a number of control variables. Vickery
used two different techniques to deal
with the likelihood that households
with a greater preference for fixed-rate
loans might adjust their behavior to
keep their loan below the conforming
limit. The first approach was an instrumental variable approach in which
Vickery constructed a dummy variable
indicating whether 80 percent of the
house value exceeded the loan limit.
Vickery argued that the house price
was plausibly exogenous with respect
to consumers’ relative preference over
types of mortgage loans. His second
approach was to simply drop observations near the conforming limit.
Vickery presented his main
results using the Monthly Interest
Rate Survey (MIRS), a sample
collected monthly from depository
institutions (sample period 19922005), which includes important
contractual characteristics but which

has no information about borrower
characteristics. Using the MIRS data,
Vickery found that the FRM share
fell discontinuously by 14.3 percent
using the instrumental variable
specification and by 20.4 percent using
the specification dropping observations
near the conforming loan limit.
Vickery reported similar, but somewhat
smaller, effects using a different
data set, the Survey of Consumer
Finances, in which respondents
provide extensive information about
household characteristics. He argued
that data about mortgage rates and
sizes are likely to be more accurate
when reported by financial institutions
rather than households and that the
MIRS estimates were more likely to be
correct.
Vickery then estimated the difference between the rates on conforming
and nonconforming loan limits. His
preferred estimates used data from
Bankrate.11 This data set contains a
detailed description of the loan contract associated with a particular rate,
especially information about any preexisting customer relationship between
the lender and borrower and information about the borrower’s FICO score.
11
Bankrate, Inc. is a company that provides
information on financial rates.

Using this data set, Vickery estimated
that the difference in the rates on a
conforming and nonconforming loan
ranged from 27 basis points for a 30year FRM to nine basis points for an
ARM that reprices after the first year.
Using pricing information from the
MIRS data set, Vickery found similar,
but somewhat smaller, estimates. Vickery argued that the Bankrate estimates
were more likely to be correct because
of the greater contractual detail.
The coefficient of substitution is
the ratio of the change in the market
share of FRMs to the difference in the
rates between FRMs and ARMs at
the conforming loan limit. Using his
preferred estimates, Vickery calculated
that (holding constant macroeconomic factors such as the yield curve)
a 20-basis-point increase in the rate
on a fixed-rate loan would lead to a 17
percent decline in the market share of
fixed-rate mortgages.
Vickery then explained the results
of a thought experiment in which he
asked how much the share of fixedrate loans would decline in the U.S.
if mortgage rates were the same as
in England, where adjustable-rate
mortgages are much more common.
He estimated that the average share
of FRMs would decline from 76 to 37
percent using UK rates. BR

REFERENCES

Adams, William, Liran Einav, and
Jonathan Levin. “Liquidity Constraints
and Imperfect Information in Subprime
Lending,” Working Paper (April 2007).
Agarwarl, Sumit, John C. Driscoll, Xavier
Gabaix, and David Laibson. “The Age
of Reason: Financial Decisions over the
Lifecycle,” Working Paper (June 2007).

www.philadelphiafed.org

Elul, Ronel, and Piero Gottardi.
“Bankruptcy: Is it Enough to Forgive or
Must We Also Forget?” Working Paper 0710, Federal Reserve Bank of Philadelphia
(March 2007).
Massoud, Nadia, Anthony Saunders, and
Barry Scholnick. “Who Makes Credit Card
Mistakes?” Working Paper (August 2007).

Narajabad, Borghan. “Information
Technology and the Rise of Household
Bankruptcy,” Working Paper (February
2007).
Vickery, James. “Interest Rates and
Consumer Choice in the Residential
Mortgage Market,” Working Paper
(September 2007).

Business Review Q1 2008 33

Financing Community Development:
Learning from the Past, Looking to the Future

Summary of the 2007 Federal Reserve System Community Affairs Research Conference
BY LORETTA J. MESTER

T

he Federal Reserve System’s 2007 Community
Development Research Conference,
“Financing Community Development:
Learning from the Past, Looking to the
Future,” was held in Washington, D.C., on March 29-30,
2007. This conference was the fifth in a biennial series
that the Federal Reserve System established in 1999.
The responsibility for organizing the conference program
rotates among the Federal Reserve Banks. The staffs of
the Federal Reserve Bank of Philadelphia’s Community
Affairs Department and Research Department took the
lead in organizing the 2007 program. The intention of
the conference series is to encourage the application
of rigorous economic analysis to issues related to
community development because without such state-ofthe-art research, policymakers cannot hope to devise
effective economic development policies and programs.
In this article, Loretta Mester provides a summary of the
conference.
The conference was organized
around six key questions: (1) Is subprime loan pricing fair or predatory?

Loretta Mester
is senior vice
president and
director of
research in the
Philadelphia
Fed’s Research
Department. This
article is available
free of charge at
www.philadelphiafed.org/econ/br/.
34 Q1 2008 Business Review

(2) Are legislative remedies to limit
predatory lending really remedies?
(3) What determines who defaults or
goes bankrupt, and how do they fare?
(4) What should and can be done to
enhance borrowers’ knowledge of their
credit risk? (5) Does the financing of
small businesses differ for minorityowned businesses and for businesses
in low-income areas? and (6) Can
alternative financial services products
help the underbanked? Although the
research did not provide definitive
answers to these questions, the presen-

tations and discussions did advance
our knowledge and provided several interesting avenues for further research.*
Jeffrey Lacker, president of the
Federal Reserve Bank of Richmond
and chair of the Conference of Presidents’ Committee on Research, Public
Information, and Community Affairs,
opened the conference. He pointed
out the value of careful, objective
research on consumer financial markets, which have experienced much
innovation in recent years. Financial
innovation creates opportunities but
also entails risk. Lacker would like
researchers to study borrowing and
other household financial decisions
from an ex ante viewpoint, that is, to
look at the full distribution of possible
outcomes and their relative probabilities. Otherwise, it is difficult to
know whether any particular credit
market product is beneficial on net
or whether the benefits of any proposed method for curtailing adverse
outcomes outweigh the costs from
restricting credit that the method may
entail. He also pointed out one of
the limitations of the data collected
under the Home Mortgage Disclosure
Act. Even with recent enhancements,
these data include information from
lenders only and do not contain much
information about borrowers, so Lacker

*

Revisions of some of the papers presented at
this conference have been published in a special
issue of the Journal of Economics and Business, 60, Nos. 1-2, 2008. Part of this summary
is taken from my introduction to this special
issue. The conference papers are available
on the Federal Reserve System’s website at
www.federalreserve.gov/communityaffairs/
national/2007researchconf/default.htm.
www.philadelphiafed.org

is pessimistic about their usefulness
for understanding the effectiveness of
credit markets. Lacker suggested that
researchers try to partner with credit
rating bureaus so that lender-supplied
data can be combined with data on
households to better illuminate borrowers’ credit decisions and outcomes.
In his view, further research will help
us better understand the costs and
benefits of market practices and government interventions.
Indeed, turmoil in the subprime
mortgage market took center stage
in mid-2007, underscoring the importance of further research on this
market segment. Six papers at the
conference studied various aspects of
the subprime mortgage market, including pricing, possible predatory practices
and policy responses, foreclosures, and
delinquencies.
SESSION 1: IS SUBPRIME LOAN
PRICING FAIR OR PREDATORY?
“Predatory Lending Practices
and Subprime Foreclosures: Distinguishing Impacts by Loan Category,” by Morgan Rose, examines
the foreclosure behavior of subprime
mortgages. While the rise in subprime
mortgage lending has increased access
to credit for some borrowers, it has
also raised concerns about possible
predatory pricing practices within this
market segment. The recent increase
in subprime mortgage foreclosures has
prompted calls for more regulation to
curb predatory lending, and some municipalities and states have passed such
legislation. But distinguishing predatory lending from legitimate lending is
a difficult task. Rose’s analysis indicates that the impact of prepayment
penalty periods, balloon payments, and
reduced documentation — characteristics often cited as consistent with
predatory lending — on the foreclosure behavior of subprime refinance
and home purchase mortgages is not at

www.philadelphiafed.org

all straightforward. To the extent that
these factors are not associated with
foreclosures resulting in loss of wealth
and tax base, the empirical basis for
some of the new regulations enacted
at the municipal and state level is questionable. These laws might restrict
legitimate access to credit for low-income borrowers without offering much
benefit. The results also suggest that
our understanding of these loans must
advance before effective federal legislation to limit predatory lending can be

code on median household income,
race, education, and adult population.
Over 31,000 loans were used in the
empirical analysis, with over 200,000
loan-quarters of observations.
Rose estimates multinomial
logit models that explain for each of
four loan types (fixed-rate purchase,
fixed-rate refinance, adjustable-rate
purchase, adjustable-rate refinance)
the probability of a loan’s entering
foreclosure, prepayment, or remaining
active in the quarter. Explanatory vari-

The recent increase in subprime mortgage
foreclosures has prompted calls for more
regulation to curb predatory lending, and some
municipalities and states have passed such
legislation.
designed, and that the recent regulatory guidelines emphasizing prudent
loan terms and underwriting standards
may be a better approach than placing
restrictions on loan characteristics.
Rose uses quarterly data collected
by LoanPerformance, Inc. on subprime
refinance and home purchase mortgages originated in 1999Q1 through
2003Q2 on properties located in the
Chicago metropolitan area and which
have been securitized into private-label
mortgage-backed securities. Chicago
provides a good laboratory for study,
having experienced a significant
increase in foreclosures in recent years.
Focusing on a single geographic region
can help control for regional differences in housing markets. However, the
limited time period means the loans
studied are not seasoned and many
of the new types of mortgage instruments, like “piggyback” mortgages,
cannot be included. Rose combines
these data with 2000 Census Bureau
data, which include information by ZIP

ables include macroeconomic, demographic, and vintage control variables,
and features of the loans, including
whether the loan requires a balloon
payment, whether it has a prepayment
penalty period longer than 36 months
from origination, whether it is a low- or
no-documentation loan, the loan-tovalue ratio, interest rate at origination,
the borrower’s FICO score at origination, and, for refinance loans, whether
the borrower withdrew cash. The first
three of these loan characteristics are
often cited as features of predatory
loans. Standard errors were adjusted
to allow for clustering by loans, since
loans can remain in the data set for
multiple quarters.
The empirical findings indicate
that the relationship between outcome
(foreclosure, prepayment, active), loan
characteristics, and demographic variables differs among the four loan types,
making it difficult to reach a general
conclusion about whether particular
loan characteristics or combinations

Business Review Q1 2008 35

of characteristics are associated with
higher probability of foreclosure. For
example, having a prepayment penalty
period longer than 36 months is associated with a statistically significant
higher probability of foreclosure for
purchase fixed-rate mortgages and refinance adjustable-rate mortgages, but
not for refinance fixed-rate mortgages
or purchase adjustable-rate mortgages.
Low- or no-documentation is associated with a statistically significant
higher probability of foreclosure for
refinance loans of either type and a
statistically significant lower probability of foreclosure for purchase fixedrate mortgages, and is not significantly
associated with the probability of
foreclosure for purchase adjustable-rate
mortgages. Rose also examines the
impact of combinations of the three
loan characteristics often considered
characteristics of predatory loans. In
most, but not all cases, the results
indicate that the effect of the combination on the predicted probability of
foreclosure is greater than the sum of
the individual impacts.
Based on the analysis, Rose concludes that the relationships between
foreclosures and loan characteristics
often cited as predatory are much more
complex than previous analysis suggests, and that prohibitions on these
loan characteristics may not have the
desired effects intended by legislators.
This suggests the need for a model
of borrower and lender behavior to
better understand the consequences of
restricting various loan characteristics
on the supply and demand for these
types of credit.
The association between subprime
lending and minorities is the focus
of “Race, Ethnicity, and Subprime
Home Loan Pricing,” by Debbie
Gruenstein Bocian, Keith Ernst, and
Wei Li. The paper examines whether
African-American and Latino borrowers receive a disproportionately

36 Q1 2008 Business Review

larger share of higher-rate home loans,
controlling for borrower riskiness. This
paper uses the 2004 data collected
under the Home Mortgage Disclosure
Act (HMDA), which for the first time
included information on the costs
of subprime home loans. For firstlien loans, lenders were required to
report the spread between the annual
percentage rate (APR) of the loan and
the yield on a U.S. Treasury security of
comparable maturity if the spread was
three percentage points or higher. By
matching these data to a proprietary
database on subprime lending, the
authors are able to address a significant weakness of earlier studies of race
and loan pricing, namely, the inability
to control for the risk characteristics
of the borrowers and loans at the
time of origination. In particular, the
proprietary data allow them to control
for a borrower’s FICO score, loan-tovalue ratio, and whether the loan was
covered by private mortgage insurance.
The resulting data set contains over
177,000 subprime loans originated in
2004.
The analysis covers subprime
loans that have been securitized where
the loans are secured by first liens
on owner-occupied properties, and
excluding loans secured by manufactured housing units, backed by private
mortgage insurance, those with
nonstandard amortization schedules,
and those with origination amounts
above the Fannie Mae and Freddie
Mac limit (which was $333,700 in
2004). Separate analyses are performed
on six different subgroups of loans,
defined by whether the loan is fixed or
variable rate, included a prepayment
penalty or not, and was for purchase or
for refinancing. Following a method of
Ambrose et al. (2004), the authors use
three-stage least squares to estimate
a logistic model relating the probability of receiving a loan designated
as a higher-rate loan in the HMDA

data to borrower, loan, economic, and
geographic characteristics, allowing
for endogeneity between the loan-tovalue, loan amount, and loan interest
rate. (Unlike the Elliehausen et al.
paper discussed below, this paper does
not account for potential simultaneity
between the presence of a prepayment
penalty and other loan terms.)
Overall, the results of the analysis
suggest that for many types of loans,
African Americans and Latinos are
more likely to receive a higher-priced
loan compared to non-Latino white
borrowers with similar characteristics.
For example, the authors estimate that
African Americans are 1.84 times
and Latinos are 1.7 times more likely
to receive a higher-rate fixed-rate
purchase loan with prepayment penalties, all else equal, than a non-Latino
white borrower. These estimates are
statistically different from one at the 1
percent and 5 percent levels, respectively.
It is beyond the scope of the paper
to identify the causes for such a disparity in pricing. It could be that even
the better measures of borrower risk
that are used in the analysis still do
not completely control for differences
in risk. However, the results suggest
that other explanations must also be
considered, for example, are minority borrowers steered to higher-priced
loans? The authors suggest some
enhancements to the HMDA reports
that would aid in further research, for
example, including information on
loan-to-value and credit scores, and
also on the type of originator.
Alan White of Community Legal
Services, Philadelphia, discussed the
Rose and the Bocian et al. papers. In
his view, both papers provide further
evidence on the harm to consumer
welfare caused by deregulation of
mortgage markets. He thinks there
has been little empirical work documenting the welfare benefits of the

www.philadelphiafed.org

expansion of the subprime lending
market. Although their existence appears to be the received wisdom, he is
skeptical that on balance such benefits
outweigh the costs. Indeed, he proposes two alternative hypotheses: that
subprime loans have displaced other
credit products, like FHA loans, and
that subprime lending has expanded
credit not by bringing in more borrowers, but by increasing the amount of
funding available to individuals who
had access to credit before the rise of
the subprime market. Regarding discriminatory pricing, White suggested
that researchers evaluate whether the
loan-pricing matrices used by lenders
to match risk factors with price are
correctly calibrated. Do minority borrowers pay higher prices because their
cost to the lender is higher? White
also underscored one of the lessons
from Rose: the subprime market is
very heterogeneous — subprime loans
that were made in 2000 are different
from subprime loans that were made in
2006, and loans made for purchase and
loans made for refinance are different,
with the latter often better thought of
as a consumer credit product rather
than as a mortgage.
SESSION 2: ARE LEGISLATIVE
REMEDIES TO LIMIT
PREDATORY LENDING
REALLY REMEDIES?
“The Effect of Prepayment
Penalties on the Pricing of
Subprime Mortgages,” by Gregory
Elliehausen, Michael Staten, and
Jevgenijs Steinbuks, also investigates
prepayment penalties on subprime
loans. Similar to Rose’s research,
the results of this paper suggest that
restricting certain loan characteristics,
in particular prepayment penalties,
may have unintended consequences.
Previous research has indicated that
loans with prepayment penalties
have higher value to lenders, and the

www.philadelphiafed.org

prepayment penalty mitigates some
of the prepayment risk faced by the
lender. However, studies have yielded
conflicting results about whether the
rates that borrowers pay are lower
for loans that include prepayment
penalties. Elliehausen et al. advance
the existing literature by examining
the relationship between prepayment
penalties and loan rates using

variable-rate, and hybrid mortgages
with a 30-year term to maturity. A
three-equation simultaneous equation
system is estimated, with loan rate premium (the difference between the loan
rate and the rate on a Treasury security of comparable maturity), loan-tovalue ratio, and presence of a prepayment penalty as dependent variables.
Loan-to-value and prepayment penalty

Previous research has indicated that loans
with prepayment penalties have higher value
to lenders, and the prepayment penalty
mitigates some of the prepayment risk faced
by the lender.
simultaneous equation estimation
techniques, which recognize that
prepayment penalty, loan rate,
and loan-to-value ratios are set
simultaneously by the lender. Previous
studies have failed to recognize this
endogeneity and so have potentially
produced biased estimates of the effect
of a prepayment penalty on the loan
rate.
This study uses the subprime
mortgage database of the Financial
Services Research Program, which
contains data on all originations of
the subprime subsidiaries of eight large
financial institutions from 1995Q3
to 2004Q4. This database covers
nearly one-quarter of loans reported
as higher-priced mortgages made for
purchase or refinancing of owner-occupied homes in the 2004 HMDA data.
The analysis includes close-ended first
mortgages with loan-to-value ratios of
90 percent or less. The average loan
amount for these loans in 2004 was
$130,000.
Because pricing schedules differ by
loan type, the authors estimate separate loan pricing models for fixed-rate,

are included as explanatory variables
in the loan rate premium equation;
loan rate premium is included as an
explanatory variable in the loan-tovalue and in the prepayment penalty
equation. Loan characteristics included in the model as controls are loan
amount, home value, loan-to-value,
and whether the loan was a low-documentation loan. Borrower characteristics included are borrower income,
FICO risk score, and whether the
home is owner-occupied. The analysis
also controls for whether the mortgage
was originated by a mortgage broker
and whether the loan was used for
refinancing. Instruments are used to
identify the system. The prepayment
penalty equation is a probit equation
used to predict the probability that the
loan includes a prepayment penalty.
This predicted value is included in the
loan rate premium equation and then
the interest equation and loan-to-value
equations are estimated by two-stage
least squares.
The empirical results show that
controlling for potential endogeneity is important: The single equation

Business Review Q1 2008 37

ordinary least squares results and the
three-equation system results differ.
Results for the three-equation system
indicate that the presence of a prepayment penalty is associated with lower
loan rates: 38 basis points lower for
fixed-rate loans, 13 basis points lower
for variable-rate loans, and 19 basis
points lower for hybrid loans. The
authors report that these interest rate
reductions are similar to those found
in lenders’ wholesale loan pricing rate
sheets. This result raises the possibility
that a restriction on the use of prepayment penalties may have the unintended consequence of raising loan rates.
“State and Local Anti-Predatory
Lending Laws: The Effect of Legal Enforcement Mechanisms,” by
Raphael Bostic, Kathleen Engel,
Patricia McCoy, Anthony Pennington-Cross, and Susan Wachter, takes
another look at anti-predatory lending laws and their effect on subprime
mortgage lending. On the one hand,
such laws could restrict the availability
of this credit and raise its price. On
the other hand, they could allay consumer concerns about predatory lending by raising the cost to lenders that
engage in abusive practices, thereby
increasing the demand for this credit.
The authors’ analysis shows that in
order to understand the effect of these
laws, it is important to look at the individual provisions, including the types
of mortgages covered, restrictions on
pricing, and enforcement mechanisms.
The study finds that these components
have independent effects on the supply
of and demand for subprime mortgages. In particular, broader coverage,
which was a provision in the newer
anti-predation laws, and enhanced
enforcement are associated with a
greater likelihood of subprime origination, while restrictions on pricing are
associated with a lower likelihood of
subprime origination.
The Home Ownership and Equity

38 Q1 2008 Business Review

Protection Act (HOEPA), passed in
1994, is a federal law that regulates
loans considered to be “high-cost
loans.” The act defines these as first
mortgages with an annual percentage
rate at origination 8 percentage points
or more above the yield on Treasury
securities of comparable maturity;
subordinate liens with a spread of 10
percentage points or more; or loans
with total points and fees that exceed

The Home Ownership
and Equity Protection
Act (HOEPA), passed
in 1994, is a federal
law that regulates
loans considered to
be “high-cost loans.”
the greater of 8 percent of the loan
amount or $400 (subject to annual
indexing). While HOEPA imposes
significant restrictions on the credit
terms of these loans, it is estimated to
cover only a small portion of subprime
mortgages. Several states have passed
their own laws; many of these lower
the HOEPA pricing triggers, thereby
expanding coverage. The laws differ
in enforcement mechanisms: Some
allow only government enforcement,
and others allow borrowers to sue
particular parties, with some restricting private lawsuits to compensatory
damages only.
Bostic et al. examine the impact
of anti-predatory lending laws on the
three different outcomes: the probability of applying for a subprime loan
relative to a prime loan, the probability
of originating a subprime loan relative
to a prime loan, and the probability of
a subprime loan’s being rejected. The
analysis includes all types of anti-

predatory lending laws, both pre- and
post-HOEPA, and finds an additional
16 state laws that previous studies
in the literature have not identified.
Building on previous research (Ho and
Pennington-Cross, 2006), the authors
create two variants of a legal index
that measures the breadth of coverage,
type and severity of restrictions on
loan terms, and enforcement mechanisms. Higher values of the index
correspond to laws with broader coverage, more stringent restrictions, and
stronger enforcement mechanisms.
The authors use 2004 and 2005
HMDA data. They identify subprime
loans in two different ways. For 2004
and 2005, they designate loans as subprime if they are reported on HMDA
as having an annual percentage rate
in excess of the rate on a Treasury
security of comparable maturity of
3 percentage points or more. This
information is available only on loan
originations and not on applications
for loans that were not originated. For
2004, they also had a list of subprime
lenders that was generated by the U.S.
Department of Housing and Urban
Development (HUD) through industry trade publications, HMDA data
analysis, and phone calls to determine
the extent of the institutions’ subprime
lending. Thus, for 2004 they were able
to repeat their analysis for this definition of subprime, which also allowed
them to investigate applications for
subprime loans, as well as originations.
To focus on the effect of antipredatory lending laws on the market
and to help control for other factors
that might affect loan markets, the
analysis includes only loans that were
made in counties along a state border,
where at least one of the states has
an anti-predatory lending law. The
authors then estimate three separate
logit regressions to predict the three
outcomes described above (the probabilities of applying for, originating,

www.philadelphiafed.org

or being rejected for a subprime loan
relative to a prime loan), as a function
of the legal index, a fixed effect designating the state border pair in which
the loan is located, controls for borrower characteristics, such as borrower
income (but not borrower FICO score,
which is not available in the HMDA
data), and location characteristics such
as county unemployment rate. They
also include a control for whether the
institution is regulated by the Office
of the Comptroller of the Currency
(OCC), since the OCC has interpreted
the National Banking Act as exempting national banks from state and local
anti-predatory lending laws.
The empirical results indicate
that the existence of a state antipredatory lending law has little effect
on credit flows in the subprime
mortgage market: It has no effect on
the odds of applying for or entering
into a subprime loan, but it reduces the
odds of being rejected for a subprime
loan by 7 percent. However, the results
also show that individual components
of the laws can have significant and
sometimes offsetting effects. Although
the effects differ somewhat across year
(2004 vs. 2005) and subprime loan
definition (HUD list vs. HMDA price
criteria), in general, the results suggest
that tighter loan-term restrictions do
not have a significant effect on the
probability of a subprime loan application’s being made but do increase the
odds of a subprime loan application’s
being rejected, and they reduce the
odds of subprime loans’ being originated. These effects are somewhat
offset by provisions resulting in broader
coverage of the laws. Broader coverage is associated with lower odds of
subprime loan applications but also
with lower odds of rejection and higher
odds of origination. This is consistent
with the hypothesis that anti-predatory
laws help reassure potential borrowers, thereby attracting them to this

www.philadelphiafed.org

market. There is weak evidence that
stronger enforcement is associated with
higher probability of subprime origination and lower probability of rejection
of a subprime application. Similar to
the Rose and Elliehausen et al. papers
discussed above, one conclusion to
be drawn from the paper is that the
impact of laws intended to improve the
functioning of the subprime mortgage
market can be complex, resulting in
unanticipated outcomes.
Michael Calhoun of the Center
for Responsible Lending discussed
the Elliehausen et al. and Bostic et al.
papers. In Calhoun’s view, the mortgage delivery system is an important
component of the subprime mortgage
market, and he focused several of his
comments on the research results concerning mortgage brokers. One of the
many findings in Elliehausen et al. is
that loans from brokers are significantly more likely to carry a prepayment
penalty, all else equal, than loans from
retail lenders. Calhoun pointed out
that this is consistent with a hypothesis discussed in Ernst (2005), namely,
that brokers may be more likely to
place borrowers in subprime loans
with prepayment penalties in order to
maximize their own compensation.
Calhoun discussed three sources of
compensation for brokers: They can be
(but rarely are) paid in cash from the
borrower, their fees can be financed
into the loan amount, and they can
receive a payment from the lender for
placing a borrower with a higher interest rate than the lender requires to
compensate it for the given borrower’s
risk profile. The lender will be more
likely to make such a payment if the
loan includes a prepayment penalty,
which helps to ensure that the borrower remains in the loan long enough
for the lender to recoup this payment.
Calhoun calculates based on typical
prepayment penalties that the interest
rate reductions found for loans with

prepayment penalties in Elliehausen
et al. are not large enough to offset
the cost of the prepayment penalty
for many subprime borrowers with
hybrid adjustable rate mortgages. He
also suggests that loans that are more
profitable for the broker to deliver are
not necessarily the best deal for the
borrower. Calhoun also suggested that
it is important to consider the mortgage delivery system when assessing
anti-predatory lending laws as in Bostic
et al. In Calhoun’s view, the HOEPA
triggers for high-cost loans may be too
narrow, as they do not include prepayment penalties or payments to brokers
for delivering loans with rates above
the lender’s minimal acceptable rate.
Several states now include a broader
definition of high-cost loans in their
anti-predatory lending regulations.
The luncheon speaker on the
first day of the conference was Mary
Lee Widener, president and CEO of
Neighborhood Housing Services of
America, Inc. (NHSA). In her presentation, Widener said she expected the
fallout from the current problems in
the subprime market to be widespread
but noted that credit markets have
faced and handled large challenges in
the past. There are likely lessons to be
learned from the current experience
to help borrowers, lenders, community development organizations, and
policymakers handle future challenges.
In Widener’s view the most important
factors for advancing community development financing are collaboration,
affordability, and borrower support.
Collaboration between community
development organizations, regulators,
policymakers, and lenders was essential
for eliminating redlining, a common practice in the 1960s and 1970s.
Development of fair lending practices
followed, taking more collaboration.
By the mid-1980s, the Community
Reinvestment Act had resulted in
hundreds of local partnerships between

Business Review Q1 2008 39

lenders and nonprofits and local
governments that delivered capital into
many local communities. Collaboration with private-sector lenders was
important for achieving affordability,
and affordability included responsible
underwriting so that borrowers could
meet the long-term obligations of their
mortgages. Borrower support was also
needed — both pre-purchase and
post-purchase counseling. In Widener’s
view lenders’ commitment to forbear
and not foreclose when temporary
life events interrupted the borrower’s
ability to repay loans was also an
important element in helping families
in low-income communities remain
homeowners. Further advancements in
the low-income mortgage market were
made by NHSA through its collaboration with the mortgage insurance industry, the secondary market through
Freddie Mac and Fannie Mae, and the
rating agencies. This allowed loans to
low-income borrowers to be financed
through the capital markets.
Widener explained that several
challenges remain. One is trying to
overcome the reluctance of many
communities to allow development of
affordable housing. Another challenge
is making the borrower support and
development systems sustainable. One
aid to doing this is showing that loans
to low-income borrowers with proper
support systems perform better than
is commonly thought, which is what
NHSA has experienced. The subprime
lending market poses another challenge. When the terms under which
subprime lending is available become
predatory, such lending has a negative impact on communities. Better
consumer education and development
of alternative loan products better
suited to lower-income borrowers can
help. Widener discussed several such
products that have been developed via
collaborations among NHSA, other
nonprofits, and the private sector.

40 Q1 2008 Business Review

SESSION 3: WHAT DETERMINES WHO DEFAULTS OR
GOES BANKRUPT AND HOW
DO THEY FARE?
“The Delinquency of Subprime
Mortgages,” by Michelle Danis and
Anthony Pennington-Cross, analyzes
the dynamics of the payment behavior
of subprime mortgage borrowers using
more sophisticated econometric techniques than have heretofore been used
to study this issue. Payment dynamics
are an important determinant of loan
pricing. For example, delinquencies
will increase the price of these loans
to borrowers by increasing the cost of
servicing these loans and of guaranteeing timely payments. The paper’s goal
is to identify the key factors that drive
delinquency.
At any point in time a mortgage
can be current, delinquent, or terminated. Within each of these branches
of possibilities, there are further alternatives (called nests). If delinquent, the
mortgage can be 30, 60, 90, or more
days late. Termination can be due to
either prepayment or default (that is,
foreclosure). Notice that the status of
the mortgage is the result of actions
of both the borrower and the lender.
To capture the multiplicity of possible
outcomes, the authors estimate (via
full-information maximum likelihood)
a nested logit model of loan outcomes
as a function of explanatory variables,
including loan characteristics (age of
loan, loan-to-value, whether the loan
is a low-documentation loan, whether
the loan is a no-documentation
loan, and whether the loan includes
a prepayment penalty), borrower’s
FICO score at time of origination,
and variables controlling for economic
conditions in the state in which the
property is located (change in house
prices, volatility in house prices,
unemployment rate, and mortgage rate
change (which does not vary by state)).
The nested logit model has an advan-

tage over multinomial logit, which is
often used to investigate such multichoice situations. The multinomial
logit model requires that the ratio of
the probabilities of any two alternative
choices (that is, the odds ratio between
the two alternatives) be independent
of any other alternative. This makes
estimation easier but is often not a
good description of behavior. For
example, the multinomial logit model
would imply that if prepayment were
taken away as an option, we’d see
proportionate changes in the probabilities of all other alternatives. But
the nested logit model would imply
that any change in the probabilities of
delinquency is evenly distributed over
30, 60, or 90+ days, but there would
not need to be proportionate increases
in the probabilities of the remaining
alternatives in the other nests (that is,
default and current). Thus, the nested
logit model is less restrictive, and the
authors present tests indicating that
the more restrictive multinomial logit
model is rejected for their data.
The authors’ loan data are from
LoanPerformance, which provides data
on pools of nonagency, publicly placed
securitized loans. They use monthly
data on the payment status of singlefamily 30-year fixed-rate subprime
mortgages on owner-occupied property
originated between January 1996 and
May 2003. Over 97,000 loans are
included in the analysis. State-level
data on house price level, house price
volatility, and the unemployment rate,
and national prime mortgage rates are
matched to the loan data. However,
the time period is too early to cover
the recent period of sharp increases
in subprime mortgage delinquencies.
The authors present estimates of the
change in the probability of the outcome associated with a one-standarddeviation increase and one-standarddeviation decrease in an explanatory
variable, holding the other variables

www.philadelphiafed.org

constant at their means. (The changes
are not symmetric for increases and decreases in explanatory variables.) They
are unable to report standard errors
for these elasticity estimates, which
are highly nonlinear functions of the
explanatory variables and coefficients.
However, most of the coefficient estimates are significantly different from
zero at the 5 percent or better level.
The empirical results show that
some of the relationships between the
explanatory variables and the probability of delinquency, default, and prepayment are as expected but others are
not. A borrower’s credit score appears
to be a robust predictor of default and
delinquency, with higher credit scores
associated with lower likelihood of
delinquency or default. The estimated
probability of 90-day or more delinquency is 0.75 percent for a borrower
with a FICO score at the mean 649;
it is 1.89 percent for a borrower with
a FICO score one standard deviation
lower, at 579.
The empirical results also show
that for borrowers with credit scores
below 630, higher credit scores are
associated with higher likelihood of
prepayment. This might reflect the
borrowers’ ability to migrate to prime
loans as their credit scores improve.
However, for scores above 630, an increase in credit score is associated with
a lower probability of prepayment. This
seems counterintuitive. The authors
suggest this might reflect something
unique about these borrowers that is
not controlled for in the estimation –
these borrowers seem to have credit
scores that would qualify them for
prime mortgages, yet they have taken
out subprime mortgages.
Prepayments on mortgages are
known to be difficult to predict, and
the paper’s results do not contradict
this. As expected, the probability
of prepayment is very responsive to
changes in interest rates, with the

www.philadelphiafed.org

probability of prepayment increasing
as mortgage rates decline. But the
probability of prepayment is fairly unresponsive to changes in house prices,
which is an unexpected result.
An interesting finding is that
factors that imply increased probability of delinquency do not necessarily
imply increased probability of default.
For example, higher loan-to-value at
origination implies a higher probability of delinquency but not of default.

An interesting finding
is that factors that
imply increased
probability of
delinquency do not
necessarily imply
increased probability
of default.
This is a reminder that movement
from delinquency to default is partly
determined by actions of the lender.
Another surprise is that higher state
unemployment rates do not seem to
trigger higher probability of delinquency or default in the authors’ data. The
interaction between local economic
conditions and loan performance
presents an interesting avenue for
future research and is one of the issues
addressed in the Grover et al. paper
discussed below.
“The Anatomy of U.S. Personal
Bankruptcy Under Chapter 13,”
by Hülya Eraslan, Wenli Li, and
Pierre-Daniel Sarte, analyzes the
performance of consumers who file for
personal bankruptcy under Chapter
13, one of two chapters of the U.S.
bankruptcy code under which households can file for bankruptcy. Under

Chapter 7, filers turn over all of their
assets above an exemption level that
varies by state in exchange for having
their debts discharged. Under Chapter
13, filers need not turn over their assets but must complete a plan that indicates how they will repay their debts
out of future income. The repayment
plan under Chapter 13 must propose to
pay at least as much as the value of the
assets creditors would have received
under Chapter 7.
The Bankruptcy Abuse Prevention and Consumer Protection Act,
enacted in 2005, introduced a means
tests on filers, whereby filers deemed
to have sufficient income would be
required to file under Chapter 13. The
act presumes that higher-income filers
will end up paying off more of their
debt under Chapter 13, while at the
same time receiving a fresh start. But
there is little, if any, empirical evidence
about how debtors and their creditors
actually fare under Chapter 13. This
paper provides such evidence using a
data set that the authors painstakingly
constructed from public court docket
records of all Chapter 13 bankruptcy
filings between 2001 and 2002 in
Delaware. The analysis, based on data
from over 900 filings, casts doubt on
the success of Chapter 13 filings.
The authors approach their
investigation by constructing a
theoretical model of the bankruptcy
decision. Debtors, when considering
bankruptcy, decide first whether to file
under Chapter 7 or Chapter 13. The
authors do not model this decision and
focus only on the decisions filers make
after they have chosen Chapter 13.
These Chapter 13 filers must decide
on the length of the repayment plan to
propose (typically three years or five
years). Once the plan is proposed, the
court-appointed trustee must decide
whether to recommend that the court
confirm the plan or dismiss it. If the
plan is dismissed, the creditors can re-

Business Review Q1 2008 41

sume debt collection measured against
the filer. If the plan is confirmed, the
filer begins making payments according to the plan. Over time, the debtor
may experience unexpected changes in
income and the plan can be modified.
If the debtor completes the (perhaps
modified) plan, any remaining debts
are discharged. If for some reason the
debtor cannot or will not complete
the payments according to plan, the
case is dismissed. The debtor might
try to convert the case to Chapter 7
or go back to face his or her creditors
without the protection of the bankruptcy provisions. The authors use
maximum likelihood techniques to
estimate their structural model relating
several outcomes — the choice of plan
length, whether the plan is confirmed
or dismissed, the creditor recovery rate
under the plan, and whether the plan
is brought to conclusion — to exogenous debtor characteristics.
The Chapter 13 filers in the
sample have significantly more debt
but fewer assets than nonfilers — filers’
median total debt is about $121,000,
about 6.6 times the national median,
while the median value of their total
assets is about $103,000, less than half
the national median. The filers are
somewhat less likely to be unemployed
than the average Delaware resident,
but their average monthly income is
about 30 percent less than Delaware’s
average adjusted gross income and
they experienced a significant decline
in income prior to filing. The median
credit recovery rate under Chapter
13 is quite low, about 12 percent of
total debt; the mean recovery rate is
about 28 percent; and a relatively small
fraction of Chapter 13 filers are actually successful in getting their cases
discharged. Moreover, 20 percent of
the debtors who want to file under
Chapter 13 are never successful in getting their repayment plan approved by
the bankruptcy court – and this was at

42 Q1 2008 Business Review

a time when these filers were voluntarily choosing to file under Chapter
13 instead of Chapter 7.
The authors’ estimation results
indicate that the amount that creditors
ultimately recover from borrowers that
file for Chapter 13 is significantly related to whether debtors are experiencing bankruptcy for the first time, the
amount of their past-due secured debt
at the time of filing, and the amount
of income they have in excess of what

Who should bear
the responsibility for
medical problems
or job problems
that might trigger
bankruptcy?
is required for basic maintenance.
Also, changes in the debtors’ financial
conditions while in bankruptcy affect
their outcomes under Chapter 13. The
authors perform some policy experiments using their estimated model.
One of the provisions of the new law
prohibits debtors with income above
the state median to file a plan with less
than five years’ duration. Their model
suggests that this provision will likely
result in only a minimal increase in recovery rates for creditors but may lower
the likelihood that filers emerge from
the bankruptcy process with a fresh
start and their cases discharged.
Katherine Porter of the University of Iowa College of Law discussed the
Danis and Pennington-Cross and the
Eraslan et al. papers. Two key questions important to these papers are:
How do we define success in lending
markets, and what enables this success? As Porter pointed out, the definition of success will likely differ for
creditors and for debtors. From a policy

perspective, one must decide what a
tolerable level of failure is and then
determine how one might respond to
failure, be it via bankruptcy relief, government or private aid, or restrictions
on the availability of credit.
Porter suggested that it is not
altogether obvious how policymakers
should treat certain trigger events.
For example, who should bear the
responsibility for medical problems or
job problems that might trigger bankruptcy? In most cases, family income
plays a primary role in determining the
success of any type of remedy. But both
the level and the stability of income
have been shown to be important to
successful outcomes under Chapter
7 in previous research and under
Chapter 13 in the Eraslan, et al. paper.
Porter suggested that further investigation into the effect of income stability
on outcomes might prove to be fruitful
in furthering our understanding of the
bankruptcy process.
SESSION 4: WHAT SHOULD
AND CAN BE DONE TO
ENHANCE BORROWERS’
KNOWLEDGE OF THEIR
CREDIT RISK?
“Targeting Foreclosure Interventions: An Analysis of Neighborhood Characteristics Associated
with High Foreclosure Rates in Two
Minnesota Counties,” by Michael
Grover, Laura Smith, and Richard
Todd, examines the predictability of
outcome – in this case, the probability
that a mortgage moves into foreclosure
– based on neighborhood characteristics. If one can predict which neighborhoods are likely to have a high rate
of foreclosure, programs designed to
help sustain homeownership could be
targeted to neighborhoods with the
greatest need.
The paper uses public data on
foreclosures in two counties in Minnesota, Hennepin and Ramsey, in 2002.

www.philadelphiafed.org

(Minneapolis is located in Hennepin
and St. Paul is located in Ramsey.)
Data on 1,178 foreclosed properties
were used in the analysis. Street addresses of the properties involved were
matched to their census tract, so that
Census Bureau data from 1990 and
2000 could be matched to the foreclosure data. Additional data on lender,
interest rates, and mortgage riders and
conditions were obtained from the
property-records departments of the
two counties. Census-tract level credit
score data were obtained from PCI
Corporation and CRA Whiz; HMDA
data were also used. The authors found
that it was very difficult to determine
from the mortgage documents whether
the loan was for home purchase or
for refinancing, and it was sometimes
difficult to determine the lender. The
painstaking nature of the data collection limited the analysis to one year
and two counties. In the authors’ data
set, foreclosed mortgages are disproportionately of recent origin, with a
median duration from origination to
foreclosure sale of 2.6 years. Compared to other mortgages originated
in the same neighborhood during the
same period, the foreclosed mortgages
tended to have higher interest rates
and smaller loan amounts and were
more likely to have been originated
by a nonbank or subprime lender and
to have had another mortgage on the
property. Reflecting strong house
price appreciation in the time period
studied, the data also show that the
sheriff’s sale typically brought in more
than the outstanding mortgage balance. Thus, had borrowers chosen to
sell their homes before defaulting, they
could have paid off their mortgages
and gotten some equity. It remains an
interesting research question as to why
borrowers did not do this.
The authors’ analysis indicates
that of seven variables available in
advance of foreclosure, neighborhood

www.philadelphiafed.org

credit score is singly the most accurate
in identifying census tracts with the
highest foreclosure rates, which is consistent with the Danis and Pennington-Cross findings, discussed above.
In particular, the 1999 neighborhood
credit score correctly ranks 36 of the
50 tracts with the highest foreclosure
rates and its correlation with the
foreclosure rate is 0.64. The authors
also perform a multivariate analysis of
the association of foreclosure rate with
variables available in advance of or
concurrently with foreclosure. They
estimate a logit model that predicts
the probability of foreclosure with
census-level variables measuring credit
risk, minority homeownership transition, and other demographic factors.
Because foreclosure is a relatively rare
event, to accurately predict the probability of foreclosure, one needs a large
number of mortgaged units. Since
the number of mortgaged units varies
considerably over the census tracts in
the sample, the variance of prediction
error might vary systematically with
the number of mortgaged units in the
census tract. To allow for this potential heteroscedasticity in the error
term, the authors estimate the logit
regression using the minimum chisquared estimator.
This multivariate analysis
indicates that the percentage of
neighborhood adults with very low
credit scores and the change in
the share of minority homeowners
between 1990 and 2000 (a measure
of neighborhood transition) are the
strongest predictors of foreclosure
rate; both are positively associated
with foreclosure rate. Based on their
findings, the authors suggest that there
may be social benefits from making
mortgage and foreclosure records and
credit scores by neighborhood more
readily available to the public and
foreclosure mitigation practitioners,
but a cost-benefit analysis of this

suggestion is beyond the scope of the
paper.
Several papers in this volume
have found that a borrower’s credit risk
score at origination is associated with
mortgage outcome, with lower scores
associated with higher rates of delinquency and default. An interesting
question is whether borrowers have an
accurate assessment of their own credit
score and whether the accuracy of
their assessment varies with the level
of their score. If higher risk borrowers
have less accurate perceptions of their
own credit risk, they may be more
likely to enter into loan contracts for
which they are not well suited (if such
contracts are offered to them), and this
could partly explain the higher rates of
foreclosure and delinquencies seen for
these borrowers.
“Consumer Credit Literacy:
What Price Perception?” by Marsha
Courchane, Adam Gailey, and Peter
Zorn, tackles this interesting question.
The authors use data provided to them
by prime and subprime lenders on 1.2
million mortgage loans originated in
2004 and from a consumer survey conducted in 2000 by Freddie Mac. The
loan data include variables collected
under HMDA and loan-level variables
used in underwriting and pricing the
loans, such as FICO score, loan-tovalue ratio, and debt-to-income ratio.
The survey includes information about
consumers’ financial knowledge and
credit outcomes such as whether they
have been denied credit, been evicted,
had utilities turned off, or property
repossessed. The survey also asked
respondents how they would rate their
current credit record.
The empirical results suggest
that inaccurate self-assessment is not
always associated with bad financial
outcomes (which might include higher
likelihood of being denied credit, being
evicted, or declaring bankruptcy) and
that the direction of the inaccuracy

Business Review Q1 2008 43

matters. The authors use locally
weighted polynomial regressions to
examine the relationship between the
percent of respondents experiencing a
bad financial outcome and credit-risk
score as measured by FICO score, with
separate analyses for respondents that
correctly assessed their credit score
and for those who did not. They also
use probit regressions to investigate
this relationship when controlling for
other factors, including income and
net worth. Both analyses indicate that
consumers who assess their credit score
to be lower than it actually is (that
is, are pessimistic about their credit
record) are more likely to experience a
bad financial outcome than those who
accurately assess their credit score, but
consumers who assess their credit score
to be higher than it actually is (that is,
are optimistic) are less likely to have
bad financial outcomes than those
who correctly assess their score.
One possible explanation is that
there is reverse causality in the survey
data. That is, a bad financial outcome
might have caused the accuracy of the
self-assessment of credit score rather
than the other way around. However,
in a separate analysis that helps to
address this potential reverse causality, the authors still find that optimism
is associated with better financial
outcomes. The authors next explore
an alternative explanation — that
consumers are actually more accurate
in their assessments of their credit risk
than their FICO scores reflect. Using
their loan and survey data, the authors
construct an alternative credit score
and find some support for this alternative hypothesis: a regression of this
alternative credit score on FICO score
and accuracy of self-assessment (that
is, optimism and pessimism) indicates
that holding FICO score constant, optimism is associated with higher values
of the alternative credit score (that is,
lower risk) and pessimism is associated

44 Q1 2008 Business Review

with lower values of the alternative
credit score (that is, higher risk).
The authors interpret the results
of their research as supporting the
value of financial literacy programs to
the extent that these programs help
educate consumers about not only
their credit scores but also a broader
set of factors that are important for
assessing their credit risk. An alternative interpretation, which differs from
the authors’, is that consumers do not
need (or no longer need) these programs, as they appear to be accurate in
assessing their credit risk.
In his discussion, Glenn Canner
of the Federal Reserve Board staff
noted that concerns about foreclosures have increased over time as the
credit-quality of the borrower pool has
widened, new types of mortgages have
emerged, short-term interest rates have
risen, and house prices have flattened
or begun to fall. He agreed that it was
important to try to identify leading
indicators of neighborhood foreclosure
sales, given the adverse effects foreclosures can have on individuals and their
neighborhoods.
Canner discussed two theories
of default. The trigger-events theory
suggests that borrowers may default
when certain life-events – for example,
medical problems, divorce, job loss –
disrupt their ability or willingness to
pay. The options theory suggests that
when a borrower takes out a mortgage
it is like having a put option on the
value of the home – the borrower will
choose to default when the mortgage
balance exceeds the value of his or her
home. These two theories can suggest
alternative factors that Grover et al.
may want to incorporate into their
study of predicting foreclosures. The
options theory suggests that areas with
falling home prices or where borrowers have little or negative equity might
show higher rates of foreclosure. The
trigger-event theory suggests that fac-

tors that disrupt income flows or lead
to unexpected expenses might lead
to foreclosure. A trigger event might
also be a factor that could affect the
accuracy of a borrower’s assessment
of his or her own credit risk. Canner
discussed other factors that could affect self-assessment accuracy, including
expectations about one’s job prospects
and future income, financial literacy,
experience in obtaining credit, the reason a payment was missed (a one-time
event or a more habitual problem), and
changes in one’s credit score over time.
Charles Plosser, president of the
Federal Reserve Bank of Philadelphia,
opened the second day of the conference by discussing the theme that
brought together the diverse group of
individuals, including government policymakers, academic researchers, community leaders, consumer advocates,
and financial service providers. The
theme he discussed was that to ensure
opportunity for the economically
distressed and to promote economic
development, we must be guided by
accurate information, careful research,
and sound policy analysis.
In Plosser’s view, “public policy
driven by headlines rarely turns out
to be good policy” and research can
now make a greater contribution to
economic development efforts than
it could in the past because development efforts have been more diverse
and more local in nature. The efficacy
of these various programs cannot be
discerned without the proper research.
Plosser discussed the importance of
development strategies that work with
the marketplace as it tries to be more
responsive to the needs of lower-income households and cautions against
the law of unintended consequences
that might arise if policymakers try to
manipulate economic outcomes. Policies are likely to always have some surprising effects, but careful analysis of
proposed policies and careful monitor-

www.philadelphiafed.org

ing of implemented policies can help
keep such surprises to a minimum.
SESSION 5: DOES THE FINANCING OF SMALL BUSINESSES
DIFFER FOR MINORITYOWNED BUSINESSES AND FOR
BUSINESSES IN LOW-INCOME
AREAS?
The last two sessions of the
conference turned from mortgages to
other aspects of community lending
and development. “Tracing Access
to Financial Capital Among AfricanAmericans from the Entrepreneurial
Venture to the Established Business,” by Alicia Robb and Robert
Fairlie, empirically investigates the
relationship between wealth, access
to financial capital, and the outcomes
from African American-owned businesses from the start-up stage through
maturity. Business ownership rates for
African Americans are considerably
lower than those for whites. According to the 2000 census data, nearly 11
percent of white workers are selfemployed business owners, while less
than 5 percent of African-American
workers are. In addition, African
American-owned businesses appear
to be less successful on average than
those owned by whites or Asians, with
lower profits and higher closure rates.
Understanding the sources of such
disparities is an important step toward
determining whether entrepreneurship
is an effective way out of poverty for
minorities. The research can also help
in determining whether government
programs offering loans to minorityowned businesses can be made more
effective or whether a new approach is
needed. While previous studies have
found that the poorer performance of
African American-owned businesses
relative to white-owned businesses
stems from low levels of start-up capital, education, and business experience, these studies did not trace out

www.philadelphiafed.org

the relationship between wealth and
access to financial capital over the life
of the business.
The authors use data from several
sources, including the Census Bureau’s
Characteristics of Business Owners
Survey, the 1998 Survey of Small Business Finances, the Survey of MinorityOwned Businesses, the Survey of
Business Owners, and the Current
Population Survey, with sample-size

the owners prior to self-employment
– the authors find that lower levels
of assets among African Americans
account for 15.5 percent of the difference in the probability of becoming
self-employed between whites and
African Americans. However, a related
question is whether African Americans are less able to raise external
funds to start their businesses than
are whites and are thereby hampered

To ensure opportunity for the economically
distressed and to promote economic
development, we must be guided by accurate
information, careful research, and sound
policy analysis.
varying over the surveys and years.
For example, the 1998 Survey of Small
Business Finances includes about
3500 businesses that were not equally
owned by a minority and nonminority;
the 1997 Survey of Minority-Owned
Business Enterprises includes over 15
million white-owned firms and over
750,000 African American-owned
firms. All the data sets confirm that
African American-owned businesses
underperform white-owned businesses
and tend to be smaller in terms of sales
and employment.
Research on entrepreneurship
indicates that personal wealth is an
important determinant of self-employment. The differences in net worth
between whites and African Americans are large: The median net worth
of whites, at $67,000, is more than 10
times the median net worth of African
Americans, at under $6,200. Results
using the Current Population Survey
data from 1998 to 2003 indicate that
the largest single factor explaining
racial disparities in business creation
rates are differences in asset levels of

by undercapitalized businesses to
start with. The authors provide some
evidence consistent with this: The
Characteristics of Business Owner data
indicate that African American-owned
businesses have lower levels of start-up
capital compared to white-owned businesses. Less than 2 percent of African
American-owned businesses start with
$100,000 or more in capital, compared
with nearly 5 percent of white-owned
businesses, and 6.5 percent of African
American-owned businesses start
with $25,000 to $100,000 in capital,
compared with about 11 percent of
white-owned businesses. The empirical results also show that lower start-up
capital accounts for 14.5 percent of the
difference in profitability of whiteowned and African American-owned
businesses. However, as the authors
discuss, the amount of start-up capital
available for investment in new businesses may be related to the predicted
performance of the business. That is,
it could be that African Americanowned businesses have lower start-up
capital because investors perceive that

Business Review Q1 2008 45

their probability of success is lower. Or
they could have less access to capital
because they have less personal wealth
to borrow against.
The authors show that some
differences in racial borrowing patterns persist even as the businesses
mature. Data from the Survey of
Small Business Finances indicate that
African American-owned businesses
are less likely to have an outstanding loan or credit line and more likely
to have borrowed on a credit card
than white-owned businesses, but the
African American-owned firms also
have worse credit histories than whiteowned businesses, including higher
rates of delinquency and bankruptcy.
The authors estimated a multivariate
logistic equation and found that once
credit history is controlled for, the difference in the probability of having an
outstanding loan is not statistically significant. However, African American
owners are more likely to have been
denied credit and to have borrowed on
their credit card than white owners,
even controlling for credit history.
The causes of the differences in credit
experiences of white and African
American business owners, the effects
these differences might have on business outcomes, and the direction of
causality (does limited access to credit
cause poor performance or does poor
performance lead to limited access to
credit?) are potentially fruitful avenues
to pursue in future research.
Indeed, “Commercial Lending
Distance and Historically Underserved Areas,” by Robert DeYoung,
Scott Frame, Dennis Glennon,
Daniel McMillen, and Peter Nigro,
addresses the topic of access to credit
by small businesses located in minority and low- and moderate-income
neighborhoods, which have typically
been underserved by financial services. There is generally little publicly
available information about small

46 Q1 2008 Business Review

businesses with which to assess their
creditworthiness. The inability to
distinguish low-credit-risk small firms
from high-credit-risk small firms can
result in the rationing of credit to all
small firms. Banks, in particular local
banks, can help eliminate some of
these information problems through
repeated interactions with the firm. To
the extent that minority and low- to
moderate-income neighborhoods have
less access to local financial services,
they are potentially put at an even
greater disadvantage at overcoming
the imperfect information problems
and gaining access to credit. However,
the advent of new technologies, such
as credit scoring models for small businesses, can help alleviate the problem
of lack of proximate financial services
by giving lenders not necessarily physically located in the local neighborhood
the ability to distinguish more creditworthy firms from less creditworthy
firms. These new technologies can
substitute for the local bank-borrower
relationship in alleviating imperfect
information impediments to lending. Indeed, several previous studies
have found an increase in the distance
between U.S. small business borrowers
and their bank lenders in recent years.
The authors extend the previous literature by examining changes
in borrower-bank lender distance
for low- and moderate-income areas
and predominately (that is, over 50
percent) minority areas. Their data
are a random sample of over 27,000
small business loans originated by
U.S. commercial banks under the U.S.
Small Business Administration loan
program from January 1984 to April
2001 with term-to-maturity of three,
seven, and 15 years. The data include
locations for both borrower and lender,
so the authors computed as-the-crowflies distances for each pair. They then
used mapping software to determine
whether the borrower was located in

a low- and moderate- vs. middle- and
high-income census tract or a predominantly minority vs. nonminority
census tract.
The univariate analysis looks
at borrower-lender distance by type
of census tract over time. Their
multivariate ordinary least squares
regression analysis (which includes
loans originated in the period January
1992-April 2001) relates distance to
indicators of whether the borrower is
located in a low- and moderate-income
area, whether the borrower is located
in a minority area, a linear time trend,
interactions between type of census
tract and the time trend, and a set
of variables to control for borrower,
lender, and loan characteristics at the
time of loan origination.
The analyses indicate that during the 1980s and most of the 1990s,
borrower-lender distances tended to
be stable and shorter, on average, for
small businesses in low and moderateincome areas and in predominately
minority areas than for those in
middle- and upper-income areas and
nonminority areas, respectively. By
the late 1990s, however, all borrowerlender distances had increased, but
those for small businesses in low- and
moderate-income areas and in predominately minority areas had increased
more, so that the borrower-lender distances are now longer for firms located
in these areas compared to firms in
middle- and upper-income areas and
nonminority areas, respectively. The
timing is consistent with the introduction of automated small-business credit
scoring models, and smaller loans
in the sample (to which these credit
scoring models are most often applied)
seem to be driving the results. While
these results are suggestive, the authors
cannot directly test the hypothesis
that the introduction of small-business
credit scoring models has allowed for
increased distance between borrower

www.philadelphiafed.org

and lender. A definitive test is an
interesting topic for future research.
In discussing the DeYoung et al.
paper, Leora Klapper of the World
Bank said that two types of credit
scoring models are currently being
used for small-business lending. The
most common produces the personal
credit score of the business owner,
which measures the probability that
the owner will default and is based
on data on the owner, including the
owner’s credit history and indebtedness. The other model, which is
growing in usage, produces a business
survival score, which measures the
probability of business failure and is
based on data on the business or business’s industry, including information
on management quality and industry
risk. A business survival score can be
derived when owners don’t have much
personal credit history, and such models are becoming increasingly used in
emerging markets like India that don’t
have credit bureaus collecting data on
personal credit histories.
Klapper suggested that more research needs to be done to determine
whether the credit scoring models are
actually increasing access to credit in
low-income neighborhoods. Can these
models substitute for bank branches in
delivering credit to the smallest businesses? As the financial system moves
to more quantitative underwriting
models, are owners with limited credit
histories able to obtain as much credit
as they did under more qualitative
relationship lending by a loan officer?
As Klapper pointed out, access
to credit by African American business owners was a main theme in the
paper by Robb and Fairlie. There is a
growing international literature that
links access to financial services and
entrepreneurship. Aggregate level data
show a relationship between economic
growth and access to capital. Klapper showed that based on data on 90

www.philadelphiafed.org

countries, there is a strong significant
positive relationship between the ratio
of aggregate private credit to GDP (a
measure of financial development) and
entry rates of new businesses. However,
empirically it is difficult to separate
out the effects of personal wealth and
credit history from access to capital to
determine their independent effects.
Klapper cited some previous literature
that looked at the effect of windfall
gains on entrepreneurship as a way of
isolating the effect of access to capital

session, which focused on two particular products: payday lending and prepaid cards. As discussed in “Strategic
Pricing of Payday Loans: Evidence
from Colorado, 2000-2005,” by Robert DeYoung and Ronnie Phillips,
payday lending has arguably extended
credit availability to more households,
but at what price? In a typical payday
lending transaction, a customer
receives a specified amount of cash in
return for a personal check written to
the lender for that amount plus a fee;

As the financial system moves to more
quantitative underwriting models, are owners
with limited credit histories able to obtain as
much credit as they did under more qualitative
relationship lending by a loan officer?
on the self-employment decision. For
example, Lindh and Ohlsson (1996)
found that winners of the Swedish lottery are more likely to enter self-employment and remain successfully selfemployed, controlling for other factors
like demographics and inheritances.
This evidence is consistent with access
to credit being an important determinant of entrepreneurship.
Klapper suggested that as credit
scoring becomes more important in
the delivery of financial services and
credit to small businesses, helping
those in low-income and minority
neighborhoods to understand their
credit scores and learn ways to improve
them will likely become more important in expanding their economic
opportunities.
SESSION 6: CAN ALTERNATIVE
FINANCIAL SERVICES
PRODUCTS HELP THE
UNDERBANKED?
The theme of access to financial
services was also taken up in the last

the lender holds this check for a specified short period, often two weeks or
less. At the end of the holding period,
the transaction can be terminated by
the lender’s depositing the check or
the customer can pay another fee to
roll over the loan. Critics of payday
lending say it is credit offered at exorbitant prices — triple-digit APRs are
not uncommon — and marketed to
unsophisticated borrowers. Others say
such lending fills a need for immediate,
short-term credit. Why borrowers use
payday loans rather than alternative
forms of credit is not fully understood.
Surveys show, for example, that the
typical payday loan customer has a job
and a bank account, and half have a
credit card.
The paper investigates the pricing
patterns of payday lenders in Colorado and concludes that these lenders
behave strategically when setting their
terms and fees. The authors’ analysis
is based on information on nearly
25,000 payday loans made in Colorado
between June 2000 and August 2005.

Business Review Q1 2008 47

These loans were made after legislation
was passed that limited loan principal
to $500 for a term of 40 days or less,
limited the finance charge to a maximum of 20 percent of loan principal
up to $300 and to 7.5 percent above
$300, and permitted only one renewal
of the loan. The average APR on the
loans is nearly 460 percent, and nearly
90 percent of the loans carried the
maximum charge allowed by Colorado
law. Because payday loan prices are
constrained by the law, the authors
use Tobit regressions to investigate the
relationship between pricing, competition in the market, and demographic
characteristics of the geographic market (ZIP code area) in which the loans
are made. Since payday lenders appear
in less than a quarter of the ZIP code
areas in Colorado and this locational
choice of the lenders might be related
to the factors included in the Tobit
regression (for example, the income in
the market), there is a potential sample
selection bias; that is, the sample may
not be randomly selected. The authors
correct for this using the standard twostage Heckman procedure.
The analysis indicates that over
time, payday loan prices in Colorado
have drifted to the state-legislated
price ceiling, and that this occurred
more quickly in markets with more
payday lenders where explicit collusion
was more difficult. Thus, the legislated
price ceiling seems to have behaved
as a focal point and may have had an
unintended effect of facilitating implicit collusion. The authors’ empirical
results also suggest that lenders take
advantage of borrower switching costs
by offering lower prices on initial loans
than on refinanced loans (although
the difference is small). Lenders that
face fewer competitors appear better
able to exploit relationships in this
way; that is, they charged an even
lower initial price than did lenders
facing more competition. This inter-

48 Q1 2008 Business Review

temporal pricing strategy might be less
profitable for lenders in more competitive markets, since they face a higher
probability of losing their customers to
competitors before being able to make
up for the low initial price. Perhaps
more surprisingly, the authors also find
that payday loan prices are higher in
markets with more commercial bank
branches. This suggests that commercial bank products are not a substitute for payday loans. Indeed, to the
extent that borrowers need a checking
account to take out a payday loan,
commercial banking services serve as a
complement to payday lending.
While payday loans offer an alternative to other forms of credit, prepaid
cards offer an alternative to other
forms of payment. “Cardholder Use
of General Spending Prepaid Cards:
A Closer Look at the Market,” by
Sherry Rhine, Katy Jacob, Yazmin
Osaki, and Jennifer Tescher, studies
the current and potential use of this
rapidly growing payment instrument.
Traditional gift cards are typically
used to make small-dollar transactions
with specific retailers. In contrast,
general spending prepaid cards can
hold considerable value and can be
used to make payments at a variety of
establishments. For example, a firm
may offer payroll cards to its employees through which the firm will pay
employees their wages in lieu of direct
deposit into checking accounts, which
some employees may not have. Prepaid
cards have also been used to distribute
payments after natural disasters. As
the authors explain, network-branded
general spending reloadable cards offer
functions similar to traditional credit
and debit cards. Their transactions
are processed using the same systems
as these network brands (MasterCard,
Visa, American Express, or Discover)
and the cards can be used to withdraw
funds from ATMs, to make retail purchases, or to pay bills in person, online,

or by phone wherever the network
brand is accepted.
The study uses transactions and
cardholder demographic data from four
general spending prepaid card providers – a random sample of 500 cardholders was drawn from each of the
four firms, resulting in a sample of over
1900 active cardholders. Transactions
for each cardholder were tracked over
a 12-month period during 2005-2006.
These data were augmented with information obtained during discussions
with other industry providers. The
analysis suggests that many providers
are marketing their cards to underbanked customers, a potentially sizable
market. Most cardholders spend
nearly all of the funds loaded onto
their cards each month – they are not
using the cards as a store of value but
as a transactions method. They use
the cards mainly for point-of-sale purchases and not to withdraw cash from
an ATM, suggesting that the cards
may be acting as a substitute for cash.
The analysis indicates that the average
cardholder loads funds onto his card
once a month and the average amount
loaded is $217. The average cardholder
makes 3.5 point-of-sale transactions
per month, each averaging a little less
than $40. And he withdraws funds
from an ATM less than once a month,
with the average amount of withdrawal
a little more than $40. The authors’
study is one of the first to document
the usage of these types of cards. They
suggest that one avenue for future
research is to augment their data with
information from consumers about
their motivations for using such cards.
Victor Stango of Dartmouth
College discussed the two papers on
alternative financial services. As he
pointed out, there are clearly new
alternatives available to the underbanked, but given the high cost of
these alternatives, the question is
whether they are beneficial to their

www.philadelphiafed.org

users. The DeYoung and Phillips
paper discusses the high cost of payday
lending. Stango indicated that the
cost of prepaid cards is also very high.
He estimated, based on the data in the
Rhine, et al. paper, that the average
cardholder has a monthly balance of
between $100 and $200 and pays about
$20 in fees per month. Stango posed
some questions for future research:
Why do people use these alternative
financial services given their high cost?
Do consumers have sufficient information to make informed usage decisions?
Are the markets for these alternatives
operating as one might expect a competitive market to operate?
The conference concluded with
Federal Reserve Chairman Ben
Bernanke speaking on the Community Reinvestment Act (CRA). As
the Chairman explained, the CRA
affirmed the obligation of federally
insured depository institutions, which
benefit from access to the financial
safety net, including federal deposit
insurance and the Federal Reserve’s
discount window, to help meet the
credit needs of their communities, in
a safe and sound manner. But over
the 30 years since it was enacted, the
CRA has evolved with the financial
services industry. When the CRA
was passed in 1977, many felt that
poor conditions in American cities,
and in particular in lower-income and
minority neighborhoods, were partly
caused by limited credit availability.
As Chairman Bernanke explained, the
CRA and other legislation passed in
the 1970s, including the Equal Credit
Opportunity Act, the Fair Housing
Act, and the Home Mortgage Disclosure Act, were intended to reduce
credit-related discrimination, expand

www.philadelphiafed.org

access to credit, and increase the
information available to assess lending
patterns. The banking industry has
undergone significant changes since
then, with interstate banking and
branching, industry consolidation, the
rise of the secondary mortgage market,
and securitization. Banks have gained
experience in underwriting loans in
lower-income neighborhoods. Chairman Bernanke cited a Federal Reserve

rural areas and in disaster areas.
Chairman Bernanke said that the
CRA will have to continue to evolve
to reflect changes in financial markets
and in the economy. He concluded his
talk by pointing out some of the challenges that lay ahead. First, defining
“local community” is becoming more
difficult as institutions become more
national in scope and with the advent
of nontraditional delivery mechanisms

The CRA affirmed the obligation of federally
insured depository institutions, which benefit
from access to the financial safety net, to help
meet the credit needs of their communities, in
a safe and sound manner.
study that indicated that in general,
CRA-related mortgage lending was at
least somewhat profitable and usually did not involve disproportionately
higher default rates than non-CRA
mortgage lending (Avery, Bostic, and
Canner, 2000).
In 1995, the CRA regulations
were amended to emphasize
performance over process and to
lessen the compliance burden. Large
institutions’ compliance with CRA
would be judged based on their
performance with respect to lending,
investments, and services, and small
banks would be allowed to meet
their requirements via a streamlined
examination that focuses on lending
activities. In 2005, further refinements
were made, including expanding the
definition of community development
to cover activities that benefit middleincome communities in distressed

like the Internet. Second, nonbank
institutions are becoming more important providers of financial services to
lower-income communities. But these
institutions are not subject to CRA.
Third, access to credit in lower-income
communities has increased, but more
lending does not necessarily imply better outcomes. Distinguishing beneficial
from harmful lending poses a challenge for regulators as they seek to ensure that the CRA continues to assist
community economic development.
The presentations and discussion
at the 2007 Federal Reserve System
Community Affairs Research Conference help illuminate several aspects of
community reinvestment and development finance. They also suggest
that much remains to be learned. It is
hoped that this conference will inspire
further rigorous research in this area. BR

Business Review Q1 2008 49

REFERENCES

Ambrose, B. W., M. LaCour-Little, and
A.B. Sanders. “The Effect of Conforming
Loan Status on Mortgage Yield Spreads:
A Loan Level Analysis,” Real Estate
Economics, 32:4 (Winter 2004), pp. 541-69.
Avery, R. B., R.W. Bostic, and G.B.
Canner. “CRA Special Lending Programs,”
Federal Reserve Bulletin, 86 (November
2000), pp. 711-31.
Bernanke, Ben S. “The Community
Reinvestment Act: Its Evolution and New
Challenges,” speech at the Federal Reserve
System Community Affairs Research
Conference, Washington, D.C., March 30,
2007; www.federalreserve.gov/newsevents/
speech/bernanke20070330a.htm.
Bocian, D. G., K. Ernst, and W. Li. “Race,
Ethnicity, and Subprime Home Loan
Pricing,” Journal of Economics and Business,
60: 1-2, (2008), pp. 110-24.
Bostic, R., K. Engel, P. McCoy, A.
Pennington-Cross, and S. Wachter.
“State and Local Anti-Predatory Lending
Laws: The Effect of Legal Enforcement
Mechanisms,” Journal of Economics and
Business, 60:1-2 (2008), pp. 47-66.

DeYoung, R., and R. Phillips. “Strategic
Pricing of Payday Loans: Evidence
from Colorado, 2000-2005,” July 2006;
www.chicagofed.org/cedric/2007_res_
con_papers/ car_48_deyoungphillips_
fedcommaffairsconf_715_preliminary.pdf.
Elliehausen, G., M. Staten, and J.
Steinbuks. “The Effect of Prepayment
Penalties on the Pricing of Subprime
Mortgages,” Journal of Economics and
Business, 60:1-2 (2008), pp. 33-46.
Eraslan, H., W. Li, and P-D. Sarte. “The
Anatomy of U.S. Personal Bankruptcy
Under Chapter 13 (revision of the
conference paper formerly titled: “The
Realities of U.S. Personal Bankruptcy
Under Chapter 13), Federal Reserve Bank
of Philadelphia Working Paper 07-31
(September 2007).
Ernst, K. S. “Borrowers Gain No Interest
Rate Benefits from Prepayment Penalties
on Subprime Mortgages,” Research Report,
Center for Responsible Lending, Durham,
North Carolina (January 2005).

Courchane, M., A. Gailey, and P.
Zorn. “Consumer Literacy: What Price
Perception?” Journal of Economics and
Business, 60:1-2 (2008), pp. 125-38.

Grover, M., L. Smith, and R. Todd.
“Targeting Foreclosure Interventions: An
Analysis of Neighborhood Characteristics
Associated with High Foreclosure Rates
in Two Minnesota Counties,” Journal of
Economics and Business, 60:1-2 (2008), pp.
91-109.

Danis, M., and A. PenningtonCross. “The Delinquency of Subprime
Mortgages,” Journal of Economics and
Business, 60:1-2 (2008), pp. 67-90.

Ho, G., and A. Pennington-Cross. “The
Impact of Local Predatory Lending Laws
on the Flow of Subprime Credit,” Journal of
Urban Economics, 60:2 (2006), pp. 210-28.

DeYoung, R., S. Frame, D. Glennon, D.
McMillen, and P. Nigro. “Commercial
Lending Distance and Historically
Underserved Areas,” Journal of Economics
and Business, 60:1-2 (2008), pp. 149-64.

Lacker, J. M. “Opening Remarks,”
Community Affairs Research Conference,
Washington, D.C., March 29, 2007; www.
richmondfed.org/news_and_speeches/
presidents_speeches/index.cfm/id=96/
pdf=true.

50 Q1 2008 Business Review

Lindh, T. and H. Ohlsson. “SelfEmployment and Windfall Gains: Evidence
from the Swedish Lottery,” Economic
Journal, 106:429 (November 1996), pp.
1515-26.
Mester, L. J. “Introduction to the Special
Issue of the Journal of Economics and
Business on Community Reinvestment and
Development,” Journal of Economics and
Business, 60:1-2 (2008), pp. 1-12.
Plosser, C. I. “Opening Remarks,” Federal
Reserve Community Affairs Research
Conference, Washington, D.C., March
30, 2007; www.philadelphiafed.org/
publicaffairs/speeches/plosser/ 2007/03-3007_financing-comm-dev.pdf.
Rhine, S. L. W., K. Jacob, Y. Osaki, and
J. Tescher. “Cardholder Use of General
Spending Prepaid Cards: A Closer Look
at the Market,” January 2007; www.
chicagofed.org/cedric/2007_res_con_
papers/ car_22_rhine_prepaid_industry_
study_rhine_et_al_final_jan312007.pdf.
Robb, A.M., and R.W. Fairline.
“Tracing Access to Financial Capital
Among African-Americans from the
Entrepreneurial Venture to the Established
Business,” June 2006; www.chicagofed.org/
cedric/2007_res_con_papers/car_21_robb_
fairlie_draft.pdf.
Rose, M. “Predatory Lending Practices and
Subprime Foreclosures – Distinguishing
Impacts by Loan Categories,” Journal of
Economics and Business, 60:1-2 (2008), pp.
13-32.
Widener, M. L. “Luncheon Remarks,”
Federal Reserve System Community
Affairs Research Conference, Washington,
D.C., March 29, 2007; www. .chicagofed.
org/cedric/2007_res_con_papers/ widener_
federal_reserve_conf_final_remarks_29_
mar_07.pdf.

www.philadelphiafed.org

RESEARCH RAP

Abstracts of
research papers
produced by the
economists at
the Philadelphia
Fed

You can find more Research Rap abstracts on our website at: www.philadelphiafed.org/econ/resrap/. Or
view our Working Papers at: www.philadelphiafed.org/econ/wps/.

NET WORTH AND HOUSING
EQUITY
This paper documents the trends in the
life-cycle profiles of net worth and housing equity between 1983 and 2004. The
net worth of older households significantly
increased during the housing boom of
recent years. However, net worth grew by
more than housing equity, in part because
other assets also appreciated at the same
time. Moreover, the younger elderly offset rising house prices by increasing their
housing debt and used some of the proceeds
to invest in other assets. The authors also
consider how much of their housing equity
older households can actually tap, using
reverse mortgages. This fraction is lower at
younger ages, such that young retirees can
consume less than half of their housing equity. These results imply that “consumable”
net worth is smaller than standard calculations of net worth.
Working Paper 07-33, “Net Worth and
Housing Equity in Retirement,” Todd Sinai,
The Wharton School, University of Pennsylvania, and NBER, and Nicholas Souleles, The
Wharton School, University of Pennsylvania,
and Visiting Scholar, Federal Reserve Bank of
Philadelphia
INCOME TAX REBATES AND
CREDIT CARD ACCOUNTS
The authors use a new panel data set of
credit card accounts to analyze how consumers responded to the 2001 federal inwww.philadelphiafed.org

come tax rebates. They estimate the monthly
response of credit card payments, spending,
and debt, exploiting the unique, randomized
timing of the rebate disbursement. They find
that, on average, consumers initially saved
some of the rebate by increasing their credit
card payments and thereby paying down debt.
But soon afterward their spending increased,
counter to the canonical permanent-income
model. Spending rose most for consumers
who were initially most likely to be liquidity
constrained, whereas debt declined most (so
saving rose most) for unconstrained consumers. More generally, the results suggest that
there can be important dynamics in consumers’ response to "lumpy" increases in income
such as tax rebates, working in part through
balance-sheet (liquidity) mechanisms.
Working Paper 07-34, “The Reaction of
Consumer Spending and Debt to Tax Rebates:
Evidence from Consumer Credit Data,” Sumit
Agarwal, Federal Reserve Bank of Chicago;
Chunlin Liu, University of Nevada, Reno;
and Nicholas Souleles, The Wharton School,
University of Pennsylvania, and Visiting Scholar,
Federal Reserve Bank of Philadelphia
MEASURING BANK PERFORMANCE:
TWO EMPIRICAL APPROACHES
Great strides have been made in the
theory of bank technology in terms of
explaining banks’ comparative advantage in
producing informationally intensive assets
and financial services and in diversifying
or offsetting a variety of risks. Great strides
Business Review Q1 2008 51

have also been made in explaining sub-par managerial
performance in terms of agency theory and in applying
these theories to analyze the particular environment
of banking. In recent years, the empirical modeling
of bank technology and the measurement of bank
performance have begun to incorporate these
theoretical developments and yield interesting insights
that reflect the unique nature and role of banking in
modern economies. This paper gives an overview of
two general empirical approaches to measuring bank
performance and discusses some of the applications of
these approaches found in the literature.
Working Paper 08-1, “Efficiency in Banking: Theory,
Practice, and Evidence,” Joseph P. Hughes, Rutgers University, and Loretta J. Mester, Federal Reserve Bank of
Philadelphia, and The Wharton School, University of
Pennsylvania
BANK EFFICIENCY AND STRUCTURE:
RECENT RESEARCH
This paper discusses the research agenda on optimal bank productive efficiency and industrial structure.
One goal of this agenda is to answer some fundamental
questions in financial industry restructuring, such as
what motivates bank managers to engage in mergers
and acquisitions, and to evaluate the costs and benefits
of consolidation, which is essentially an empirical question. The paper reviews the recent literature, including techniques for modeling bank production and the
empirical results on scale economies, scope economies,
and efficiency in banking.

52 Q1 2008 Business Review

Working Paper 08-2, “Optimal Industrial Structure
in Banking,” Loretta J. Mester, Federal Reserve Bank of
Philadelphia, and The Wharton School, University of
Pennsylvania
INTERNATIONAL TRADE COSTS,
INVENTORIES, AND DEVALUATIONS
Fixed transaction costs and delivery lags are important costs of international trade. These costs lead firms
to import infrequently and hold substantially larger
inventories of imported goods than domestic goods.
Using multiple sources of data, the authors document
these facts. They then show that a parsimoniously
parameterized model economy with importers facing an
(S, s)-type inventory management problem successfully
accounts for these features of the data. Moreover, the
model can account for import and import price dynamics in the aftermath of large devaluations. In particular,
desired inventory adjustment in response to a sudden,
large increase in the relative price of imported goods
creates a short-term trade implosion, an immediate,
temporary drop in the value and number of distinct
varieties imported, as well as a slow increase in the
retail price of imported goods. The authors’ study of six
current account reversals following large devaluation
episodes in the last decade provides strong support for
the model’s predictions.
Working Paper 08-3, “Inventories, Lumpy Trade, and
Large Devaluations,” George Alessandria, Federal Reserve
Bank of Philadelphia; Joseph Kaboski, Ohio State University; and Virgiliu Midrigan, New York University

www.philadelphiafed.org