Full text of Economic Review (Federal Reserve Bank of Cleveland) : 1986 Quarter 2

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

E C O N O M I C
1 9 8 6

2

Metropolitan Wage Differentials:
Can Cleveland Still Compete? How

high are Cleveland’s wages, relative to those
in other cities, when differences in worker
skills are held constant? Skill-adjusted wage
differentials for 44 cities show that Cleve­
land’s wages are higher than the national
average, but lower than in some “ fastgrowth” cities. In addition, wage differen­
tials for Cleveland and other cities have nar­
rowed substantially in the last decade.

The Effects of Supplemental Income
and Labor Productivity on Metropolitan
Labor Cost Differentials. Author Thomas F.

9

Luce examines the effects that a combina­
tion o f supplemental income and labor
productivity have on the measurement o f
metropolitan labor-cost differentials in
manufacturing. Using data for the 20 largest
Standard Metropolitan Statistical Areas
(SMSAs), he finds that controlling for these
factors increases the measured labor-cost
differentials among these SMSAs. He also
finds that the data do not support the pro­
position that higher-than-average wage rates
are associated with greater-than-average
labor productivity.

Reducing Risk in Wire Transfer Sys­
tems. Wire transfer provides an effi­

cient electronic method for moving huge
sums o f money— sometimes as much as
$500 billion per day— in the nation’s pay­
ment system. Some users have com e to rely
on daylight credit generated by wire transfer
systems. This practice creates risk for those
extending credit as well as a systemic risk
o f disrupting worldwide financial markets.
Author E. J. Stevens discusses wire transfer
systems, a new Federal Reserve Board of
Governors’ risk control policy, and some
institutional changes that might be expect­
ed to reduce risk in this policy environment.

R E V I E W

Q U A R T E R

Economic Review

2

is published quar­

terly by the Research Department of
the Federal Reserve Bank of Cleve­
land. Copies of the issues listed
here are available through our Public
Information Department,
216/579-2047.

Editor: William G. Murmann.
Assistant editor: Meredith Holmes.
Design: Michael Galka.
Typesetting: Liz Hanna.

Opinions stated in

Economic Review

are those of the authors and not
necessarily those of the Federal
Reserve Bank of Cleveland or of the
Board of Governors of the Federal
Reserve System .

Material m ay be reprinted provided
that the source is credited. Please
send copies of reprinted materials to
the editor.

IS S N 0013-0281

Metropolitan Wage
Differentials: Can Cleveland
Still Compete?
by Randall W. Eberts
and Joe A. Stone

Randall W . Eberts is Associate Pro­

This paper was prepared for the

fessor, and Joe A . Stone is W . E.

Labor Conference, which was held

Miner Professor, Department of Eco­

M ay 2 ,1 9 8 5 , in Cleveland, and which

nomics, University of Oregon.

was sponsored jointly by the
Regional Economic Issues (REI) Pro­
gram and the Federal Reserve Bank
of Cleveland. The study was sup­
ported in part by a grant from the
Cleveland Foundation through the REI
Program. The authors would like to
acknowledge useful comments by
Michael Fogarty, director of the REI
Program, and by Joseph Antos. In
addition, the authors are especially
indebted to Ralph Day for technical
computer assistance.

Introduction
Labor costs are often cited as one of the primary7
reasons for the economic hardships plaguing
many older industrial cities, such as Cleveland. O f
course, other factors, such as local taxes, proxim­
ity to markets, product cycles, and energy costs
may also contribute to the area’s diminished abil­
ity to compete with other regions in attracting
and retaining businesses. Nonetheless, since
labor costs represent an important part of total
production costs, the initial presence of signifi­
cant wage differentials among metropolitan areas
may have been a major factor in the economic
expansion of Sunbelt cities and the relative
decline of Snowbelt cities. In turn, divergent pat­
terns of growth resulting partly from firms relocat­
ing in low-wage areas may have caused wage lev­
els to converge.
With respect to the effect of differ­
ential labor costs on firm location and on
regional employment growth, two aspects of
labor costs must be considered. First, there is
more to examining labor cost differentials across
regions than simply looking at regional differen­
ces in wage rates. Firms consider not only the
amount they pay workers, but also the productiv­
ity of their workers. Stated simply, an employer is
willing to pay a worker in Cleveland a higher
hourly wage than a worker in Atlanta, for exam­
ple, if the Cleveland worker is more productive
than the Atlanta worker. Therefore, a comparison
of regional wage differentials is much more
meaningful when these wages are adjusted for
differences in worker skills.

Second, the advantage to a firm in
searching for a low-wage area is directly propor­
tional to the degree of regional dispersion in
labor costs. A large regional variation in labor
costs would make it advantageous for firms to
search for low-wage areas, since the relative cost
savings would be sizable. O n the other hand, if
wage differentials, adjusted for worker skills, are
observed to converge over time, then the competi­
tive disadvantage of relatively high-wage areas,
such as Cleveland, would diminish over time.
The purpose of this paper is three­
fold: to provide estimates of variations across
metropolitan areas in the wage employers pay a
worker of given skills and training, to compare
these “skill-adjusted” wage differentials with
observed differentials, and to examine how these
differentials may have changed over the past
decade. The Cleveland metropolitan labor market
is used as a point of comparison to highlight how
labor costs in a major industrial city in the Fourth
Federal Reserve District fare with respect to other
U. S. cities.

I. Theoretical Framework
Metropolitan areas in the United States are charac­
terized by many firms that act as price-takers
when they sell to national markets and that con­
sider the rental prices of capital to be fixed by ex­
ternal conditions (see Borts and Stein [1964];
and Muth [1968 and 1983] )• This demand-side in­
terpretation of regional labor markets fixes local
nominal wages by the horizontal labor demand
curve of firms competing in national or interna-

tional product markets. Long-run equilibrium
levels of local wages are determined by the
demand for labor, under the technical condition
that the level of output changes in constant pro­
portion to changes in labor and capital. Shifts in
labor supply have no long-run effect on local
nominal wages in this model, but supply changes
do cause changes in total employment and even-

Estimates of Wage Equations for 1974 and 1983
( Current Population Survey data)
Variable

1974

1983

Intercept

1.26
(39.08)

1.58
(115.27)

Schooling

0.12
(9.15)

0.13
(30.70)

Schooling squared

0.007
(2.17)

0.004
(3.18)

0.024
(31.39)

0.026
(114.22)

Potential experience

Potential experience squared

Employment status (full­
time = 1)
Gender (female =1)

Race (nonwhite =1)

Occupation dummy variables
(omitted for brevity)
R-square
Number of observations

-0.0004

-0.0004
(-25.05)

(-86.03)

0.14
(14.25)

0.16
(58.34)

-0.31
(-35.15)

-0.23
(-96.52)

-0.05
(-3-90)

-0.02
(-6.82)

------

------

0.49
13,733

0.49
175,268

NOTES: Coefficients are follow ed by /-statistics in parentheses. The 1983 re­
gression also contains quarterly dum my variables to control for variations
during 1983. See text for definition o f variables and further explanation o f
data. All coefficients are statistically significant at the 0.01 percent level,
except for schooling squared in 1974, which is statistically significant at the
0.05 percent level.

TABLE

1

tually in total population. O f course, other influ­
ences on local wages are possible in short-run dis­
equilibrium and even in long-run equilibrium, if
local products are relatively unique or sold in geo­
graphically limited markets, if local natural re­
sources are a significant input into the production
of exportable goods, or if any of the other condi­
tions of the demand-side model above are violated.
Johnson (1983) provides an extensive theoretical
and empirical analysis of many of these factors,
including local costs of living, environmental
amenities important to workers, taxes, income

transfers, moving costs, unionization, transporta­
tion costs, discrimination, and various human
capital and skill variables. Most of the previous
studies of geographical wage differentials have
allowed a dominant role for labor supply in
determining local wages (see Coelho and Ghali
[1971 ]; Bellante [1979]; Sahling and Smith
[1983]; Scully [1969]; and Johnson [1983]).
Without necessarily denying a role
for nondemand factors, the purpose of our study
is to obtain estimates of metropolitan wage dif­
ferentials relevant for identifying demand-side
effects and to explore the possible significance of
such effects over the past decade. To do this, we
first estimate the demand-side differentials for
1974 and 1983, and then examine the trends in
the differentials between the two periods. Under
the demand-side model, the change in skilladjusted wage differentials during this period is
expected (all else the same) to be inversely
related to subsequent rates of economic growth
via firm locations, expansions, and contractions.
We have found in Eberts and Stone (1985), for
example, a significant inverse relationship
between metropolitan wage differentials in the
1970s and subsequent firm locations. Therefore,
one would expect wage differentials measured in
1974 to narrow by 1983.

II. Data and Empirical Results
The data used to estimate the metropolitan wage
differentials are obtained from 1974 and 1983
(CPS) compiled by
the Bureau of Labor Statistics. The 1974 data
come from the May survey, which contains sup­
plementary questions regarding employment.
The 1983 information is derived from questions
asked of one-quarter of the individuals in each of
the 12 monthly surveys. Because of this differ­
ence (and also because of other changes in the
CPS between 1974 and 1983), the total number of
workers with sufficiently complete records for
analysis is much smaller in 1974 than in 1983
(13,733 workers in 1974 versus 175,268 in 1983).
The sample allows us to identify 43 of the largest
metropolitan areas— Standard Metropolitan Statis­
tical Areas (SMSAs)— for both years of data.
Our first step in obtaining skilladjusted wage differentials is to specify estimable
wage equations that reflect appropriate demand
determinants of the wages of individual workers.
This approach follows the human capital specifi­
cation of individual wages set forth by Hanoch
(1967) and Mincer (1974). Thus, we specify indi­
vidual wages (expressed in logarithms) as a func­
tion of observed determinants of individual
productivity— education level (entered as a quad­
ratic), potential experience (age, minus years of
education, minus six, also entered as a quad-

Current Population Surveys

ratic), a binary7dummy variable indicating full­
time employment status, and 46 binary occupa­
tion dummy variables (with one of these omitted
as a constant). Binary dummy variables are also

1974 Metropolitan Wage Differentials
(percentage difference from national average)

Rank

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

SMSA

New York
Paterson
San Francisco
Detroit
Chicago
Nassau-Suffolk
Rochester
San Jose
Portland
Gary
San Diego
Anaheim
Seattle
Los Angeles
Albany
Akron
Cleveland
Atlanta
Denver
New Orleans
Baltimore
Sacramento
Indianapolis
Minneapolis-St. Paul
Milwaukee
Columbus
Boston
San Bernardino
Houston
Newark
Philadelphia
St. Louis
Pittsburgh
Cincinnati
Miami
Kansas City
Dallas
Ft. Worth
Birmingham
NonSMSAs and other SMSAs
Buffalo
Norfolk
Greensboro
Tampa

Skilladjusted

18.6
17.9
17.5
17.1
15.7
15.5
14.5
13.4
13-3
12.9
12.9
12.2
9.1
8.4
8.3
7.9
7.5
6.5
5.9
5.9
5.8
5.7
5.5
5.1
4.9
4.3
4.1
3.9
3.8
3.7
3.1
1.4
0.6
0.5
-0.6
-1.8
-2.9
-4.4
-4.7
-5.8
-6.9
-7.1
-8.1
-15.9

Actual

21.6
18.6
19.8
23.6
16.1
24.8
19.9
23.7
16.8
10.5
21.2
27.3
24.4
10.8
18.7
3-8
14.4
2.8
11.4
-0.8
5.4
9.0
8.9
9.8
8.0
3.9
9.4
5.0
10.4
3.6
6.3
1.7
-1.6
-0.3
0.2
3.6
-0.9
-0.5
0.1
-8.7
-4.9
-7.6
-8.6
-17.9

entered to control for race and gender differences
in wages. Under the assumptions of the demand
model, the separate wage regressions for 1974
and 1983 yield coefficients that reflect national
average marginal productivities in specific occu­
pations and for particular human capital compo­
nents. Industry dummy variables and union mem­
bership status are not included, because these
variables are not viewed as productive attributes.
Detailed information on other components of
labor compensation (pensions, health insurance,
and the like) is not available in the data.
The predicted wage level for each
worker in the sample is obtained by multiplying
the estimated coefficients by each worker’s char­
acteristics. The predicted wage can be interpreted
as the compensation a worker could expect to
receive, given his or her characteristics, regardless
of geographic location. Subtracting the predicted
wage from the actual wage, then, nets out the
portion of the actual wage that is related to the
worker’s skills. The skill-adjusted metropolitan
wage differentials are then obtained by averaging
the wage residuals (actual, minus predicted
wage) in each year for all workers in a particular
metropolitan area. Average wage differentials are
calculated for each of the 43 SMSAs for each year.
The national average wage differential is, of
course, equal to zero by the property of leastsquares regression. For purposes of comparison,
an additional average is calculated jointly for
nonSMSAs and other excluded SMSAs.

Wage regressions.

The estimated (log) wage
equations for both 1974 and 1983 are presented
in
except that the 45 estimated coeffi­
cients for the occupation dummy variables are
omitted for brevity. These equations are pres­
ented only to document the results of our
demand-oriented wage regressions. Except for
the absence of nondemand factors (for example,
controls for union membership), these are famil­
iar regressions (with minor variations) in the
labor literature.
The estimated coefficients in
are as expected in both years. Schooling (with
a value equal to 1 for eight to 11 years, a value of
2 for 12 to 15 years, a value of 3 for 16 to 17
years, and a value of 4 for more than 18 years)
enters with a significantly positive coefficient.1
Schooling squared also enters with a significantly
positive coefficient; years of potential experience

table 1,

table

1

NOTE: Wage differentials are derived from C urrent Population Survey’ files,
using the technique described in the text.

I

This specification of education permits greater nonlinearity in
the effects of different education levels than the use of individ­
ual years of education, although the difference is trivial for our estimated

wage differentials.

enters with a positive coefficient; experience
squared enters with a negative coefficient; a
dummy variable for full-time employment enters
with a positive coefficient; and dummy variables

1983 Metropolitan Wage Differentials
(percentage difference from national average)

Rank

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

SMSA

San Francisco
San Jose
Anaheim
Seattle
Minneapolis-St. Paul
Nassau-Suffolk
Houston
Los Angeles
Chicago
San Bernardino
Detroit
Gary
Dallas
Portland
Paterson
Sacramento
Denver
Newark
Milwaukee
New York
San Diego
Cleveland
Rochester
New Orleans
St. Louis
Ft. Worth
Pittsburgh
Atlanta
Boston
Kansas City
Baltimore
Philadelphia
Cincinnati
Akron
Greensboro
Columbus
Indianapolis
Buffalo
NonSMSAs and other SMSAs
Albany
Birmingham
Miami
Norfolk
Tampa

Skilladjusted

18.1
18.1
15.5
14.9
12.0
11.3
10.8
10.7
10.4
10.0
9.3
8.4
8.0
7.9
7.8
7.8
7.6
7.3
7.1
7.1
5.7
5.1
5.1
4.8
3.9
3-4
2.7
2.7
2.2
2.1
1.6
1.5
1.3
-1.3
-1.6
-2.2
-2.5
-2.6
-4.8
-6.0
-6.9
-6.9
-7.3
-10.7

Actual

25.4
28.4
23.0
22.1
12.6
16.0
14.5
13.0
14.5
7.1
9.1
3.6
12.2
9.7
13.5
7.0
12.7
12.7
7.9
11.4
4.5
7.0
11.0
8.8
4.0
3.0
5.4
6.2
5.7
4.5
4.8
4.7
1.4
1.4
-3.5
-2.1
-2.6
-4.5
-7.1
-5.4
-5.1
-11.4
-7.3
-11.7

NOTE: Wage differentials are derived from C urrent P opulation Survey files,
using the technique described in the text.

for race (with nonwhite equal to 1) and gender
(with female equal to 1) enter with negative coef­
ficients. All listed coefficients are significant at the
5 percent level.
With the exception of the decline in
absolute value for the race and gender coefficients
in 1983, the 1974 and 1983 regressions are basic­
ally the same. The similarity extends, by and
large, to the 45 occupation dummy variables as
well, although a few of these coefficients do
change. Intercepts in the two equations, of course,
differ significantly, due to both nominal and real
wage growth between 1974 and 1983 for the Unit­
ed States as a whole. Both regressions explain 49
percent of the variation in actual wages.

Metropolitan wage differentials.

Skill-adjusted
and actual metropolitan wage differentials
(expressed as the percentage deviation from the
national average) are presented in
for
1974 and in
for 1983- The SMSAs are
ranked according to the size of the skill-adjusted
differential. Because of the semilogarithmic speci­
fication of the wage equation, residuals are expo­
nentiated to obtain percentage differentials.
The rankings offer a perspective on
how Cleveland’s wages compare with regions
against which the area might compete for eco­
nomic development. In 1974, Cleveland’s skilladjusted wage was 7.5 percent above the national
average, which put Cleveland in seventeenth place
among the cities considered. A number of cities
usually associated with rapid growth, such as San
Jose, San Diego, and Anaheim, had wage differen­
tials that were higher than Cleveland’s. O n the
other hand, Cleveland’s skill-adjusted wages are
consistently higher than they are in southeastern
cities. About one-quarter of the cities with wage
rates below Cleveland’s level were in the South­
east, and no southeastern city had a skill-adjusted
wage differential higher than Cleveland’s. More­
over, small SMSAs and nonSMSA regions showed
much lower skill-adjusted wage differentials than
Cleveland’s— over 12 percent lower.
In 1983, Cleveland’s skill-adjusted
wage fell to only slightly above 5 percent of the
national average, which brought its ranking down
to twenty-second place. All the southern cities in
the sample still had wage differentials below
Cleveland’s. A few additional cities, such as San
Bernardino and Sacramento, were added to the
1974 list of west coast cities that surpassed Cleve­
land in the skill-adjusted wage differential.
Wage differences between metro­
politan areas can be broken down into two compo­
nents: differences in the skill-adjusted wages and
differences in the value of skills (measured in
dollars). Consider the difference in actual wages
between two SMSAs (
and
). Recall that:

table 2

table 3

w\

w0

5

(1)

log 0 0) =
log (

u\ )

=

bS0 + ?0,
bS, + e x,

b

where is the regression coefficient associated
with the skill-related variables (5), and denotes
the residual or skill-adjusted wages (actual wage,
minus predicted wage). We assume that the
appropriate aggregation has been done, so that
each equation represents wages in a specific met­
ropolitan area.
The difference in the actual (log)
wages between the two metropolitan areas is:

e

( u \ ) - log ( w 0) =
b(Sx - S 0) + (e v e0)

(2)

log

The first component on the right-hand side is the
difference in levels of skills normalized in wage
units between the two areas. The second is the
difference in skill-adjusted wages between the
two metropolitan areas. If, for example, the actual
wage differential is greater than the skill-adjusted
differential, then the skill level is necessarily
greater in area 1 than in area o. Consider the
wage differentials displayed for San Francisco in
1983. The actual wage in San Francisco is 17.2
percent higher than the actual wage in Cleveland,
but the skill-adjusted wage is only 12.4 percent
higher.2 The difference of 4.8 percentage points is
due to the higher skill levels of San Francisco
workers relative to Cleveland workers. Since
employers are willing to pay workers the value of
their contribution to the production of each unit
of output, the higher wages associated with
higher productivity do not affect the relative
competitiveness of the two areas. Rather, it is the
difference in wages over and above the differen­
tial associated with higher labor productivity that
affects competitiveness among regions. In the
case of San Francisco, a 12.4 percent wage differ­
ential exists, which is not accounted for by skill
differentials. On the other hand, Rochester’s 3-2
percent wage differential relative to Cleveland is
due entirely to higher skill levels in Rochester.

2

The percentage difference in wages between any tw o metropol­
itan areas can be easily calculated from

tables I and 2,

by

using the following formula:

WvW°

= I WyWus -

\ ^us

wus J

(1/(1+(w0-wus)/wus))

w here ( w , - w0) /w 0 is the percentage difference in wages
between area 1 and area o and (w j -w us)lw us is the percentage
deviation in w a g e s in area i from the n ation's (the differential dis­
played in tables 2 and 3).

Although the results for 1974 show
a rough correspondence between skill-adjusted
and observed (actual) wage differentials, substan­
tial differences are also clearly evident. Detroit,
Anaheim, Birmingham, San Diego, Cleveland,
Houston, and Boston, for example, all have
observed wage differentials that exceed the skilladjusted differential by at least 8 percentage
points, which is the approximate differential
required for statistical significance at the 5 per­
cent level. Only Akron exhibits the opposite
phenomenon— a skill-adjusted differential that is
at least 8 percentage points higher than the
observed differential. The five SMSAs with the
highest skill-adjusted wages are New York, Pater­
son, San Francisco, Detroit, and Chicago. The five
lowest SMSAs are Tampa, Ft. Worth, Greensboro,
Norfolk, and Buffalo.
The results for 1983 show a
stronger correspondence between skill-adjusted
and observed wage differentials. By this year, no
SMSA except San Jose has an observed wage dif­
ferential that differs from the skill-adjusted differ­
ential by at least 8 percentage points. Only one of
the five highest-wage SMSAs in 1974 (San Francis­
co) remains in the top five in 1983- The remain­
ing four in 1983 are San Jose, Anaheim, Seattle,
and Minneapolis-St. Paul. Two of the lowest-wage
SMSAs in 1974 (Tampa and Norfolk) remain among
the five lowest SMSAs in 1983. The remaining
three in 1983 are Albany, Birmingham, and Miami.
The changes in the differentials
between 1974 and 1983 are presented in
SMSAs with the largest increases are Dallas, Ft.
Worth, Houston, Minneapolis-St. Paul, and
Greensboro. SMSAs with the largest decreases are
Albany, New York, Paterson, Rochester, and
Akron. Most of the cities associated with rapid
growth during the last decade exhibit increases in
both skill-adjusted and observed wage differen­
tials. In some instances, the skill-adjusted and
actual changes in wage differentials differ sub­
stantially. SMSAs that show increases in the skilladjusted differential, but a decline in the actual
wage differentials, are Houston, Anaheim, and
Sacramento. For these SMSAs, the skill-adjusted
increase is presumably offset by a decline in
average skill level.
Cleveland’s skill-adjusted and ob­
served wage differentials fell between 1974 and
1983; the actual wage declined more rapidly than
the skill-adjusted wage. Since the relative decline
in the actual wage differential, with respect to the
skill-adjusted wage differential, has to be offset by
a decline in average skill level of the area’s work
force, this indicates that Cleveland suffered a de­
cline in the average skill of the area’s labor force.
New Orleans, Philadelphia (trivial­
ly), Atlanta, and Akron show decreases in the
skill-adjusted wage differential, but an increase in

table 4.

the actual wage differential. For these SMSAs, the
skill-adjusted decrease is presumably offset by an
increase in the average skill level. Large diver­
gences between the skill-adjusted and actual

Change in Wage Differentials from 1974 to 1983
(percentage point change)

Rank

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

SMSA

Dallas
Ft. Worth
Houston
Minneapolis-St. Paul
Greensboro
San Bernardino
Seattle
Tampa
San Jose
Buffalo
Kansas City
Newark
Anaheim
St. Louis
Los Angeles
Milwaukee
Pittsburgh
Sacramento
Denver
NonSMSAs and other SMSAs
Cincinnati
San Francisco
Norfolk
New Orleans
Philadelphia
Boston
Birmingham
Cleveland
Atlanta
Nassau-Suffolk
Baltimore
Gary
Chicago
Portland
Miami
Columbus
San Diego
Detroit
Indianapolis
Akron
Rochester
Paterson
New York
Albany

Skilladjusted

10.9
7.8
7.0
6.9
6.5
6.1
5.8
5.2
4.7
4.3
3.9
3-6
3.3
2.5
2.3
2.2
2.1
2.1
1.7
1.0
0.8
0.6
-0.2
-1.1
-1.6
-1.9
-2.2
-2.4
-3.8
-4.2
-4.2
-4.5
-5.3
-5.4
-6.3
-6.5
-7.2
-7.8
-8.0
-9.2
-9.4
-10.1
-11.5
-14.3

Actual

13-1
3.5
4.1
6.9
5.1
2.1
-2.3
6.2
6.2
0.4
0.9
-9.1
-4.3
2.3
2.2
-0.1
7.0
-2.0
0.7
1.6
1.7
5.6
0.3
9.6
-1.6
-3.7
-5.2
-7.4
3.4
-8.8
-0.6
-6.9
-1.6
-7.1
-11.6
-6.0
-16.7
-14.5
-11.5
-2.4
-8.9
-5.1
-10.2
-24.1

NOTE: Wage differentials are derived from Current Population Survey files,
using the technique described in the text.

changes (even if the changes are in the same
direction) have similar interpretation. Other
SMSAs with large differences between the two
measures are San Jose, Birmingham, Gary, San
Diego, Detroit, and Albany.
Based on the estimates above,
have skill-adjusted metropolitan wage differen­
tials converged since 1974? This question can be
answered by calculating the change in the coeffi­
cient of variation from 1974 to 1983. The coeffi­
cient of variation is the standard deviation (com­
puted from the sample of SMSA-level wage
differentials) divided by the mean; thus, it is an
index of the degree of dispersion in the sample.
This measure indicates substantial convergence
for both sets of differentials, declining by 22 per­
cent for the skill-adjusted differentials and by 46
percent for the actual wage differentials. Because
the observed wage differential is composed of
the skill-adjusted wage differential and a differen­
tial related to differences in actual skills, the fact
that observed wages converged more than twice
as much as skill-adjusted wages suggests that
variations across metropolitan areas in actual skill
levels also declined during the period.3
Why do we observe relatively
strong wage convergence during the 1974-1983
period? Following our demand-side approach,
one could attribute convergence to the expand­
ing scope of most product markets (both domes­
tically and internationally), increased competition
faced by geographically concentrated firms that
may have had some power to influence price, the
relative decline of industries that make products
using relatively large amounts of local natural
resources, and the emergence of manufacturing
industries that require smaller-scale plants.

III. Conclusion
The objective of this paper was to provide esti­
mates of variations across metropolitan areas in
the wage employers pay a worker of given skills
and training, and then to compare how these dif­
ferentials have changed over the past decade.
Based upon 1974 and 1983 data from the
we find substantial variations
in skill-adjusted wages in both 1974 and 1983, as
well as significant deviations between skilladjusted and observed wage levels. We also find
that the wage differentials and skill differentials
converged significantly during this same period.
Cleveland’s skill-adjusted and actual wage levels

Population Survey,

3

Current

The change in average skill level could be the result of changes
in actual skills or of changes in the market compensation of the

skills between 1974 and 1983. The general similarity of the 1974 and
1983 wage regressions, however, suggests that most of ihe change in
skill level reflects actual changes in skills.

also converged toward the national averages.
Over the last decade, Cleveland closed the gap by
2.4 percentage points for skill-adjusted wages and
by nearly 9 percentage points for actual wages.
The reduction in Cleveland’s wage
differentials and the general convergence in
wages and skills could influence Cleveland’s
economic future in at least two ways. First, the
incentive for firms to move out of Cleveland
might diminish, since convergence in wages
reduces the potential cost savings of a move.
Second, the wage differential might not be as crit­
ical a factor in economic growth as it once was.
In fact, labor supply-side factors, such as labor
climate and local amenities and public services,
might become more important influences on
economic development.

References
Bellante, Don. “The North-South Differential and
the Migration of Heterogeneous Labor,”
vol. 69, no. 1 (March
1979), pp. 166-75.

Amer­

ican Economic Review,

Borts, George H. “The Equalization of Returns
and Regional Economic Growth,
vol. 50, no. 3 (June I960),
pp. 319-47.

"American

Economic Review,

Economic Growth

------- , and Jerome L Stein.
New York: Columbia Univer­
sity Press, 1964.

in a Free Market.

Coelho, Philip R, and Moheb A Ghali. “The End
of the North-South Wage Differential,”
vol. 6, no. 5 (December
1971), pp. 932-37.

can Economic Review,

Ameri­

Eberts, Randall W., and Joe A Stone. “The Deter­
minants of Firm Location,”
Department of Economics, University of
Oregon, 1985.

Working Paper,

Hanoch, Giora. “An Economic Analysis of Earn­
ings and Schooling,”
vol. 2, no. 3 (Summer 1967), pp.
310-29.

Resources,

Journal of Human

Johnson, George E. “Intermetropolitan Wage Dif­
ferentials in the United States,” in Jack E. Tri­
plett, ed.,
Chicago: University of Chicago Press, 1983, pp.
309-30.

The Measurement oj Labor Cost.

Schooling Experience, and Learn­

Mincer, Jacob.
New York: National Bureau of Economic
Research, Distributed by Columbia University
Press, 1974.

ing.

Muth, Richard F. “Differential Growth Among
Large U.S. Cities,” in James P. Quirk and Arvid
M. Zarley, eds.,
Lawrence, KS: University of Kansas
Press, 1968.

nomics.

Papers in Quantitative Eco­

_________ “Intermetropolitan Wage Differentials
in the United States. Comment,” in Jack E. Tri­
plett, ed.,
Chicago: University of Chicago Press, 1983, pp.
330-32.

The Measurement oj Labor Cost.

Sahling, Leonard G., and Sharon P. Smith.
“Regional Wage Differentials: Has the South
Risen Again?”
vol. 65, no. 1 (February 1983), pp. 131-35.

tics,

Review oj Economics and Statis­

Scully, Gerald W. “Interstate Wage Differentials: A
Cross Sectional Analysis,”
vol. 59, no. 5 (December 1969), pp.
757-73.

Review,

American Economic

The Effects of Supplemental
Income and Labor
Productivity on
Metropolitan Labor
Cost Differentials
by Thomas F. Luce
Thomas F . Luce is associate direc­
tor, Wharton Philadelphia Economic
Monitoring Project, University of
Pennsylvania.
The author would like to thank
Anita A . Summers and Michael F.
Fogarty for their helpful comments
on preliminary drafts.

Introduction
Regional variations in labor costs attract a great
deal of attention because of their potential to
affect the regional distribution of economic activ­
ity. Because of the major role that labor costs play
in total production costs, regional differences in
labor costs may translate into dramatic regional
differences in profitability. Profitability, in turn, is
a major determinant of whether or not existing
firms will expand and of whether or not new
firms are likely to locate in a given region.
Most studies of regional variations
in labor costs are based solely on data for payroll
per employee or wages. However, the measure of
labor costs that is most relevant to the profitability
of a firm is the total cost of the labor needed to
produce its output. There are at least two steps
involved in getting from simple wage data to
estimates of labor costs per unit of output.
First, nonwage income to workers
( “supplemental income”) must be added to
wages to get total labor costs per hour. In 1977,
the value of measurable supplemental income
paid to manufacturing workers was, on average,
about 20 percent of the value of wages paid to
those workers. This percentage showed consider­
able variation, however, among regions and
states. (See Garofalo and Fogarty [1984].) By
1982, supplemental income had increased to
nearly 25 percent of wages.
Second, labor costs per hour must
be scaled by the amount of output per hour that
is attributable to labor inputs, as opposed to
other inputs, if the objective is to measure labor
costs per unit of output. The amount of output

generated by a unit of labor input varies for dif­
ferent workers and for different production pro­
cesses. Labor productivity will be different among
workers possessing different skills or other per­
sonal characteristics. Similarly, productivity will
vary for a single worker according to the amount
of other factors of production (machinery,
energy, or raw materials) used in a particular
production process.
The research described here has
been directed toward incorporating these two
corrections to the raw wage rates to obtain a
more accurate measure of labor costs per unit of
output. The supplemental income estimates are a
direct extension of work done at the state level
by Garofalo and Fogarty.
The strategy employed to control
for labor productivity differs significantly from
that used in most other studies. Researchers in­
terested in analyzing regional labor productivity
patterns face hard decisions regarding the tech­
niques and data available to them. One option is
to use indirect indicators of productivity that can
be measured reliably, but which may or may not
be reliable proxies for labor productivity. This op­
tion uses the personal characteristics of workers
to measure productivity. The strength of the tech­
nique is that it uses data that are relatively access­
ible and reliable. It has two major weaknesses: (1)
Strong assumptions are required about the rela­
tionship between the indirect measures (personal
characteristics) and labor productivity. (2) The
method does not control for differences across
industries or regions in capital intensity.

The other choice—the one used in
this analysis— is to attempt to estimate labor pro­
ductivity directly from data that measure output
and input levels. The strength of this approach is
that the effects of differences in the quality of
labor
of differences in the mix of inputs
(capital intensity) are both captured by the mea­
sure. This approach also has two major weakness­
es: (1) Reliable data are not easily available at the
regional level for some of the measures, especial­
ly capital inputs. Some variables must, therefore,
be estimated (with some error) from the data that
are available. (2) The option requires one to make
fairly strong assumptions about the nature of pro­
duction processes across industries and regions.
Neither approach is entirely satis­
factory7since each requires strong assumptions. It
could be argued that the first approach (based on
characteristics of the labor force) provides the
more relevant measure for new firms seeking a
location, because these kinds of firms are not tied
to an existing technology or physical plant. On
the other hand, the second approach, because it
controls for the effects on labor productivity of
existing capital intensities, may be the better
measure for capturing the potential of existing
firms to expand in their current location.
Indexes for wage rates, supplement­
al income, and labor productivity have been gen­
erated for each of the 20 commonly reported Man­
ufacturing subsectors (two-digit SIC industries) in
the 20 largest (based on 1980 employment) Stan­
dard Metropolitan Statistical Areas (SMSAs).

and

I. Wages
The wage data collected for this research support
the finding by other researchers that wage rates
vary significantly among metropolitan areas and
regions in the United States. The first column of
shows 1982 Manufacturing wage indexes
for the 20 largest SMSAs. The indexes represent
production worker hourly wages in the SMSA as a
percent of the national average. In order to con­
trol for the fact that different SMSAs have different
industrial structures, each SMSA’s index compares
the cost of that SMSA’s employment mix com­
puted from the SMSA’s wage structure to the cost
of the same mix at national average wages. This
means that an artificially high index number will
not be produced simply because an SMSA has
greater-than-average concentrations of employ­
ment in industries that have wages that are higher
than the average for Manufacturing as a whole.
The first thing that is clear from the
wage indexes is that, in 1982, there was a great
deal of variation in Manufacturing wages among
these large SMSAs. Wages in the lowest-wage
SMSA (Nassau) were only 75 percent of those in
the highest-wage SMSA (Pittsburgh). Eight of the

table 1

SMSAs showed wages less than the national aver­
age, but the weighted average wage for the 20
was 2 percent greater than the national average.
The most striking feature of the
regional averages is that all of the SMSAs in the
North Central region showed Manufacturing wages
in 1982 that exceeded the national average.
Wages in Cleveland approximately matched the
regional average at about 8 percent above the
national average and were greater than in all but
six of the 20 largest SMSAs. Wages in the North­
east, South, and West were close to the national
average, but the indexes are far from being uni­
formly distributed within these regions. The
Northeast, for instance, shows the lowest regional
average despite the fact that it contains the
highest-wage SMSA in the sample (Pittsburgh).
Similarly, wages in the South and West range
from a low of 91 percent of the average (Dallas)
to a maximum of 115 percent (San Francisco).
These regional patterns, particu­
larly the finding that the Northeastern SMSAs
showed lower wages on average than those in
the South and West, are somewhat surprising. In
light of the often-cited difference between wages
in the “Sunbelt” and the “Frostbelt,” one might
have expected greater regional differences than
those revealed by the data. One possible explana­
tion for the patterns is that wages have converged
over time as the result of equilibrating forces at
work in the national economy. In regions where
Manufacturing employment is in decline, one
would expect downward pressure on relative
wages. Examination of column (7) of
provides some support for this view. Wages in
the slower-growing Northeast and North Central
regions have indeed declined relative to those in
the South and West. However, the decline was
significantly greater in the Northeast than the
North Central and there are some clear excep­
tions at the SMSA level— for example, relative
wages in Pittsburgh increased between 1977 and
1982 despite significant declines in Manufactur­
ing employment during the period.

table 1

n. Supplemental Income
Estimates of supplemental income for the 20
largest SMSAs showed even more variation than
the wage indexes, ranging from about 75 percent
of the national average to more than 130 percent
of the average. The supplemental income data
include
available in the
both mandatory supplements to wages like social
security and worker’s compensation, and voluntary
supplements like health and life insurance. Other,
less easily measured fringe benefits, such as free
parking or subsidized cafeterias, are not included.
Unfortunately, the regional data are
reported at the state level only, with no detail

Census of Manufactures

across industries. Industry detail is available only
in the national data. Estimates of supplemental
income for each of the 20 Manufacturing sectors
in the SMSAs have been generated by combining
wage data from the SMSAs with the information
about regional variations in fringe benefits rates
contained in the state level data and with the
information about variations among industrial
sectors contained in the more detailed national
data. The procedure assumes supplementary in­
come in a given industry and SMSA to be the
product of (1) the level of wages in that industry'
and SMSA (2) the average supplemental income
rate (supplemental income divided by wages) for
the industry in the nation, and (3) the average

supplemental income rate in total Manufacturing
in the SMSA’s home state (controlling for the
Manufacturing employment mix in the state). The
supplemental income estimates for each industry
in an SMSA are then combined in the same way
as the wage estimates to get the mix-controlled
index for total Manufacturing in the SMSA
The supplemental income indexes
in
column (2), compare the supplemen­
tal income cost of the SMSA’s employment mix to
the cost of the same mix at the national average
supplemental income rates. The estimates for
1982 show the differences among the SMSAs’
fringe benefits rates to be much more substantial
than for wages. The supplemental income rate in

table 1,

Manufacturing Wage, Supplemental Income and Labor Productivity Indexes:
20 Largest SMSAs, 1982
United States =100
(1 )
Simple
wage
index

Twenty largest SMSAs3 102.2

(2 )

(3 )

(4 )

(5 )

Supplemental
income
index

Total
labor cost
index

Labor
productivity
index

Corrected
labor cost
index

(6 )

(7 )

Column ( 5 ) Column ( 1 )
minus
change
column ( 1 )
from 1977

(8 )
Column ( 5 )
change
from 1977

98.1

101.5

99.1

102.4

0.2

-0.8

-0.7

0.4
-6.0
1.6
3-4
-6.4
3.8
23.5

-2.6
-0.8
-2.2
-2.0
-3.6
-2.8
5.5

-3.3
-5.5
-1.8
-5.9
-4.9
20.6

Northeast3
Boston
Nassau
Newark
New York
Philadelphia
Pittsburgh

98.6
96.5
89.1
96.5
95.5
101.6
118.9

93.3
85.4
74.1
85.4
83.8
110.2
131.4

97.7
94.5
86.4
94.4
93-5
103.3
121.6

98.7
104.3
95.2
94.6
105.0
98.0
85.3

99.0
90.5
90.7
99.9
89.1
105.4
142.4

North Central3
Chicago
Cleveland
Detroit
Minneapolis
St. Louis

108.0
103.2
107.7
118.4
109.0
104.8

109.9
106.5
114.9
129.0
93.2
97.2

108.5
103.9
109.2
120.6
106.1
103.2

95.9
94.8
94.1
92.0
103.4
104.8

113-0
109.5
116.0
131.1
102.6
98.5

5.0
6.3
8.3
12.7
-6.4
-6.3

-0.9
-0.1
-1.7
-1.3
3.9
0.1

5.3
1.8
-0.4

South3
Atlanta
Baltimore
Dallas
Houston
Washington, DC*3

100.4
93.8
111.3
91.3
109.1
101.4

87.8
77.9
119.2
74.3
88.8
137.5

98.0
90.7
112.9
88.0
105.1
107.6

101.6
102.6
99.3
103.9
100.1
99.2

96.4
88.4
113.7
84.7
105.0
108.5

-4.0
-5.4
2.4
-6.6
-4.1
7.1

1.7
-2.4
4.5
1.0
4.9
-4.5

-0.1
0.2
0.7
-3.2
6.6
-4.0

West3
Anaheim
Denver
Los Angeles
San Francisco

101.1
99.1
105.7
98.3
114.8

96.7
94.8
92.0
95.0
111.0

100.3
98.3
103.1
97.7
114.1

102.5
98.4
101.0

97.9
99.9
102.1
94.0
112.2

-3.2
0.8
-3.6
-4.3
-2.6

2.5
4.0
4.8
2.4
1.5

0.2
4.2
2.2
-1.2
2.5

103.9
101.7

-3.2

3-4
7.4
2.2

SOURCE: Com puted from Census o f M anufactures, 1977 and 1982, Bureau o f the Census, U.S. Department o f Com m erce.
a. Aggregate indexes are weighted averages o f the SMSA estimates with weights based on manufacturing em ploym ent in the 20 largest
SMSAs.
b. Supplemental incom e index for Washington, DC based on 1977 supplemental incom e data.

11

Nassau (the SMSA showing the lowest index
number) was only 54 percent of that in Washing­
ton, DC (the SMSA with the highest rate).
The overall pattern of supplemen­
tal labor costs reflects the wage pattern. Higherthan-average wage rates tend to be associated
with higher-than-average supplemental income
rates. Eight of the 12 SMSAs with higher-thanaverage wages had higher-than-average supple­
mental income rates. Similarly, all eight of the
lower-than-average wage SMSAs also showed
lower-than-average supplemental income. Cleve­
land’s standing in supplemental income rates
mirrored its position in wages— only three of the
20 SMSAs showed supplemental income rates
higher than Cleveland’s.
Total labor costs in a specific
industry and SMSA are calculated as the sum of
wages and supplemental income. The industryspecific estimates are then combined in the same
way as the wage and supplemental income mea­
sures to get a mix-corrected estimate of the laborcost index for total Manufacturing. Because of the
rough correspondence between the wage in­
dexes and the supplemental income indexes, and
because supplemental income represents a much
lower percentage of total compensation than
wages, the total labor-cost indexes (colum n [3])
do not differ dramatically from the wage indexes.
In only four of the 20 SMSAs is the difference
greater than three percentage points (Atlanta, Dal­
las, Houston, and Washington, DC). In general,
adding supplemental income to the labor-cost
indexes increases the spread among the SMSAs,
but not by a substantial amount.

III. Labor Productivity
Factoring labor productivity into the labor-cost
estimates is also very important. If higher-thanaverage wages in an SMSA reflect higher-thanaverage labor productivity, then the index for the
SMSA from column (3) of
will overstate
any relative disadvantage that the SMSA might
have in competing with other regions for jobs.
The potential for labor productivity
to significantly alter the competitive position of
an SMSA is greater than it is for supplemental
income. Supplemental income represented only
about 24 percent of total compensation in 1982.
Any adjustment made for labor productivity,
however, affects 100 percent of total compensa­
tion. Consequently, equivalent percentage differ­
ences in the two factors in an SMSA will have dif­
ferent effects on the overall measure of labor
costs in the SMSA, with the productivity adjust­
ment being the greater of the two.
Differences among SMSAs in labor
productivity can arise from two different sources.
First, differences in labor quality due to skill lev­

table 1

els, education, or experience are likely to be
reflected in differences in productivity. Many ana­
lysts, therefore, use various labor-force character­
istics as proxy measures for productivity. How­
ever, there is another important source of
productivity differences. The amounts of other
factors of production that are used in combina­
tion with labor will affect labor productivity inde­
pendently of the quality of labor. The productivity
of otherwise identical workers will be different
depending on the amount of capital (such as
machinery and equipment) used in combination
with them in the production process.
Ideally, controlling for labor pro­
ductivity differences arising from both sources
would require that the researcher have industryand SMSA-specific data for output, capital inputs,
labor inputs, and all other inputs. In addition,
knowledge about the production process, itself—
how capital and labor are combined at different
levels of output— is required. With this kind of
data, it would be possible to separate the portion
of output directly attributable to labor from the
part attributable to the other inputs.
These kinds of data are not readily
available, particularly at the SMSA level. However,
the
reports value-added
data by industry and SMSA From these data, it is
possible to estimate the amount of output that is
directly attributable to a unit (one hour) of labor
input, after controlling for the amount of capital
used in the production process. The labor cost
per unit of output can then be estimated by div­
iding the cost per unit of labor input by output
per unit of labor input.
The procedure used in this
research to make this calculation involved two
steps. First, capital inputs were estimated by sub­
tracting total labor costs (including the costs for
nonproduction workers) from value added and
dividing this difference by an estimate of the rate
of return to capital. Second, the value of output
attributable to a unit of labor input was estimated
by assuming that capital and labor are combined
in a specific way in the production process.1The

Census of Manufactures

I

ln theory, value added represents only the contribution to the
total value of output that is made at the stage of production in
question. A n y contributions to value that are made at prior stages in the

total production process (such as by the refining of raw materials, or pre­
assembly of components) are not included in the “value added" at the
stage of production under analysis. In reality, the value-added data
reported by the Bureau of the Census includes some value that cannot
be directly attributed to the labor or capital brought to bear at the stage
of production in question. The capital estimates used here thus overstate
capital stock. H ow this overstatement affects the relative measures
used in this analysis depends on the extent to which the magnitude of
noise in the data varies from S M S A to S M S A , a piece of information
which is not available. The production process assumed for the analysis
is a constant retums-to-scale Cobb-Douglas production function with a
capital exponent of 0.282 (taken from Hulten and Schwab [1984]).

result is a measure of how effectively the SMSA’s
labor force is combined with the existing capital
plant. By estimating productivity directly from
output data (albeit with some strong assump­
tions), it is not necessary to make any assump­
tions about how labor force characteristics, such
as education or age, affect productivity. If an
SMSA’s labor force possesses productivityenhancing characteristics, the impact should be
captured in the estimate of output that is directly
attributable to labor inputs.
Labor productivity estimates
derived by using this procedure show much less
variation across the 20 largest SMSAs than either
the wage or supplemental income indexes.
reports the labor produc­
Column (4) of
tivity indexes for the 20 SMSAs. The index repre­
sents labor productivity in Manufacturing in the
SMSA as a percentage of national average labor
productivity in Manufacturing. Productivity in the
lowest-productivity SMSA (Pittsburgh) is about 85
percent of the national average and about 81 per­
cent of the value for the highest-productivity
SMSA in the group (New York).2
A primary reason for investigating
labor productivity is to test whether higher-thanaverage labor costs in an SMSA reflect higherthan-average labor productivity. Comparisons of
the third and fourth columns of
suggest
that this is not the case in the 20 largest SMSAs.
Indeed, the simple correlation coefficient— a mea­
sure of how closely two variables move together—
between the labor productivity indexes and the
wage indexes is negative, indicating that, in these
SMSAs, higher-than-average wage indexes are asso­
ciated with lower-than-average labor productivity.3
The result of this negative relation­
ship is that, when labor productivity is factored
into the labor-cost indexes, the spread among the
SMSAs increases. Column (5) shows the labor
cost per unit of output indexes. The lowest-cost
SMSA (Dallas) showed labor costs in 1982 that
were just under 60 percent of those in the
highest-cost SMSA (Pittsburgh).

table 1

table 1

2

The very low index for Pittsburgh is largely due to the index for
the S M S A 's dominant sector — Primary Metals. Reported

value added in this sector for 1982 w as less than total labor costs for
the sector, a relationship which is conceptually troublesome and which is
inconsistent with the labor productivity calculation. The difference
between reported value of shipments and cost of materials was there­
fore substituted for reported value added in the productivity estimation
procedure. Consequently, the productivity measure for Pittsburgh should
be viewed with caution, since it is likely that the problems resulting from
the use of available value-added data (see fn. 1) are particularly acute in
Pittsburgh's case.

V j

The correlation coefficient is 0.52.

IV. Combined Effects of Supplemental Income
and Labor Productivity
The supplemental income and labor productivity
adjustments to the simple wage index tend to op­
erate in the same direction. This was true for 17
of the 20 largest SMSAs. In each of the seven
SMSAs where the supplemental income adjustment
increased the labor-cost index, the productivity
adjustment also increased it. Similarly, in 10 of
the 13 SMSAs where the supplemental income cor­
rection decreased the labor-cost index, the pro­
ductivity correction also resulted in a decrease.
The net change in the labor-cost
measure resulting from the two adjustments is
shown in column (6) of
In 11 of the 20
SMSAs, the net effect of the two adjustments was
to decrease the labor-cost index. In these SMSAs,
the simple wage index overstates relative labor
costs. In the other nine SMSAs (including Cleve­
land), the simple index understates costs relative
to the national average. The magnitude of the
under- or overstatement varied substantially from
SMSA to SMSA, with the understatement being the
greatest for Pittsburgh, and the overstatement
being the greatest for Dallas.
Overall, these results suggest that
simple wage measurements will tend to distort
regional labor-cost differentials. O n average, the
wage indexes understate relative labor costs in
the higher-cost, North Central SMSAs, and over­
state them in the lower-cost SMSAs in the South
and West.
In addition, the productivity correc­
tion has a very significant effect on the measured
change in labor costs between 1977 and 1982.
The increases in costs in the South and West re­
flected in the simple wage indexes are largely off­
set by improving relative labor productivity during
the period (colum n [8],
O n the other
hand, the decline in relative wages in the North
Central region is overwhelmed by the decline in
the relative productivity measure. Only in the
Northeast does the productivity correction have
little effect on the measured change in labor
costs. The net effect is that the competitive posi­
tion (as measured by the productivity-corrected
labor-cost indexes) of the Northeastern SMSAs im ­
proved on average between 1977 and 1982, while
the North Central’s position deteriorated, and
those of the South and West remained unchanged.
What are the implications of laborcost differentials of the magnitude found in
7? Statistical analysis, relating employment growth
between 1977 and 1982 to relative labor costs in
1977 in the 20 largest SMSAs, suggests that they
have been significant in the past. (See Summers
and Luce [1985].)
The finding was that, after control­
ling for the effects of national employment
trends, unionization rates, right-to-work legisla­

table 1.

table 1).

table

tion, energy7costs, vulnerability to international
competition, state and local taxes, cost of living,
and local amenities, a labor-cost differential of 50
percent in 1977, like the one that existed
between Dallas and Detroit, was associated with a
subsequent employment growth differential of
almost 3 percent per year. The actual differential
for these two SMSAs for the period from 1977 to
1982 was about 10 percent per year, implying that
the labor-cost differential explained almost 30
percent of the total difference in growth rates.

case. Manufacturing employment declined much
more quickly in these six SMSAs between 1977
and 1982 than in the other 14, or in the nation as
a whole. In the six, total Manufacturing employ­
ment declined by more than 5 percent per year
over this time period, compared to a decline of
less than 1 percent per year in the other 14.

V. Relative Labor Costs in Cleveland
In Manufacturing as a whole, Cleveland fell into

Manufacturing Wage, Supplemental Income and Labor Productivity Indexes: Cleveland SMSA, 1982
United States =100
(1 )
Simple
wage
index

1 4

(2)

(3)

(4)

(5 )

Supplemental
income
index

Total
labor cost
index

Labor
productivity
index

Corrected
labor cost
index

(6 )

(7 )

Column ( 5 ) Column ( 5 )
minus
change
column ( 1 )
from 1977

Total manufacturing3

107.7

114.9

109.2

94.1

116.0

8.3

2.2

Durables3
Lumber products
Furniture and fixtures
Stone, clay and glass
Primary metals
Fabricated metals
Non-elec. machinery
Elec. machinery
Trans, equipm ent
Instruments
Other durables

109.0
108.2
110.8
100.2

116.5
115.5
118.3
106.9
118.7
119.2

94.6
102.3
107.8

116.9
107.0
103.9
104.3
121.7
118.9
123.8

7.9
-1.2

113.3
124.9
112.7
100.8
106.0

110.6
109.4
112.0
101.5
113-2
113.2
107.5
118.5
107.2
95.5
100.5

2.4
-2.7
-20.1
-0.5
23.5
2.8
11.9
-2.1

Nondurables3
Food and kindred
Textiles
Apparel
Paper and allied
Printing and publishing
Chemicals
Petroleum products
Rubber and plastics

102.7
102.0
85.1
161.6
89.1
110.3
95.7
74.2
91.6

107.5
108.9
90.8
172.4
95.1
117.7
102.1
79.2
97.7

103.5
103-3
86.0
163.1
90.2
111.4
97.0
75.2
92.8

92.7
93.7
112.0
125.8
81.1
87.1
92.5
94.0
91.9

111.3
111.7
106.2
117.0
105.6
94.5
99.4

97.3
93.0
95.2
86.9
114.9
98.4
90.5
87.0

103.1
109.0
105.6
115.5
111.7
110.3
76.8
129.7
111.3
128.0
104.9
80.0
101.0

-6.9
4.1
10.4
7.2
17.6
-13.9
3.4
11.1
16.1
9.0
8.3
-8.3
-31.9
22.2
17.7
9.2
5.8
9.4

-29.3
-2.8
-4.1
1.8
-22.1
-6.0
13.5
14.9
11.9
2.1
-18.7
-3-8

SOURCE: Com puted from Census o f M anufactures, 1977 and 1982, Bureau o f the Census, U.S. Department o f Com m erce,
a. Aggregate indexes control for industrial structure.

TABLE

2

For the SMSAs showing higherthan-average labor costs and lower-than-average
productivity in 1982 (Philadelphia, Pittsburgh, Chi­
cago, Cleveland, Detroit, and Baltimore) the im ­
plications of this finding are particularly sobering.
The statistical analysis implies that those SMSAs
would have had to possess very significant cost
advantages from other sources, such as greaterthan-average access to input or output markets, to
have been competitive with other areas in the Unit­
ed States. This does not appear to have been the

the group of SMSAs in 1982 (composed primarily
of the older SMSAs in the North and East) with
higher-than-average wages, higher-than-average
supplemental income, and lower-than-average
labor productivity. It is of interest to examine
whether this pattern carries over into the specific
industrial sectors that are of greatest importance
to the region.
shows the 1982 labor-cost
measures, described above, broken out by the 18
sectors for which data are available for Cleveland.

Table 2

Some caution should be exercised
in evaluating the results presented in
The
primary reason for this is the level of industrial
disaggregation used in the analysis. In the same
way that total Manufacturing measures that do not
control for different industrial structures across
SMSAs may over- or understate labor-cost differ­
ences, the two-digit SIC breakdowns in
may reflect differences between Cleveland and
the nation in industrial structure at a finer level of
disaggregation. This problem, in fact, appears to
be a factor in at least two of the sectors shown in
It is likely that the very low wage index
for Petroleum Products and the very high index
for Apparel are largely the result of this issue.
However, since these two sectors, together,
accounted for less than 5 percent of Manufactur­
ing employment in the region, they have very lit­
tle impact on the overall indexes.
Nearly 70 percent of 1982 produc­
tion worker employment in Manufacturing in the
Cleveland SMSA was contained in the five sectors
beginning with Primary Metals in
Each of
these sectors showed higher wages and supple­
mental income in Cleveland than in the nation as
a whole. In addition, only one of the five ( Elec­
tric Machinery) showed labor productivity signifi­
cantly above the national average. Two others
(Fabricated Metals and Transportation Equip­
ment) showed labor productivity within five per­
cent of the average. However, productivity advan­
tages in none of these sectors were large enough
to offset the significantly higher-than-average
wage and supplemental income rates.
Overall, productivity-conected labor
costs exceeded the national average in all but two
of the reported 18 sectors. In addition, the re­
gion’s competitive position deteriorated between
1977 and 1982 in eight of the 18 sectors and in
three of the region’s five largest sectors (Primary
and Fabricated Metals, and Nonelectric Machin­
ery). Labor costs clearly cannot be viewed as a
factor enhancing the region’s desirability to firms
competing in national and international markets.
What impact are differences of the
magnitude found in Cleveland likely to have on
future employment growth or decline in the
region? The research cited in previous sections
suggests that the impact was very significant
between 1977 and 1982. The findings implied
that a labor-cost differential like the one found
for Cleveland in 1977 (14 percent) was associated
with subsequent employment growth in Manufac­
turing, which was about 0.8 percent per year less
than it would have been if labor costs had been
equal to the national average. This represents
more than one-fifth of the total difference
between the growth rate in the Cleveland SMSA
and that in the nation between 1977 and 1982
(when the average difference was about 3 6 per­

table 2.

table 2

table 2.

table 2.

cent per year). Although other factors working to
Cleveland’s disadvantage explain the majority of
the region’s slower-than-average employment
growth in the period, the effect of higher-thanaverage labor costs cannot be ignored. A 0.8 per­
cent per year shortfall in growth represents about
7,000 Manufacturing jobs in the SMSA over the
five-year period from 1977 to 1982.

VI. Conclusions
Manufacturing labor costs varied significantly
among large SMSAs in 1982. Most of the variation
was attributable to differences in wage rates.
When supplemental income was added to wages
to get total labor costs per hour, the spread
among SMSAs increased, but not by a substantial
amount. Correcting for differences among SMSAs
in labor productivity tended to increase the dif­
ferentials by more than the supplemental income
adjustment but by a magnitude that was less than
the original wage differentials. The data for the 20
largest SMSAs do not support the proposition that
higher-than-average wage rates are associated
with greater-than-average labor productivity.
Labor costs in 1982 for the Cleveland SMSA were significantly greater than the na­
tional average. O f the overall 16 percentage point
differential, about 50 percent (or eight percentage
points) was due to greater-than-average wage rates.
Another 40 percent of the total difference was
attributable to lower-than-average labor productiv­
ity, with the remaining 10 percent being due to
greater-than-average supplemental income rates.
The higher-than-average labor
costs in Cleveland are likely to have had a dam­
pening effect on employment growth in Manufac­
turing in the region. In the group of the 20 largest
SMSAs, labor-cost differentials of the magnitude
evident in Cleveland in 1977 were associated
with employment growth about 0.8 percent per
year less than if labor costs had equaled the
national average. This represents about one-fifth
of the total difference in Manufacturing employ­
ment growth rates between Cleveland and the
nation between 1977 and 1982.
The overall implication of this
research for the Cleveland area is that, in order to
compete effectively with other areas of the coun­
try for Manufacturing jobs, other characteristics of
the region must be sufficiently advantageous to
overcome the region’s relatively high labor costs.
Many of the same market forces that operated in
the past to create the higher-than-average wages
in the region are likely to lead in the future to
some moderation, but this is a slow and painful
process. Wages in Cleveland as a percent of
national average wages declined by only 2 per­
cent between 1977 and 1982 — a period when
Manufacturing employment in the region

15

decreased by 25 percent. In addition, the margi­
nal improvement in the region’s competitive
position embodied in the relative wage decline
was more than offset by a decrease in relative
labor productivity in the region.
Perhaps the most important mes­
sage from the analysis is that there is room for
improvement in the SMSA in one component of
labor costs— labor productivity—that can be en­
hanced over a shorter time horizon by actors
within the region. Any improvements in this direc­
tion will require both a commitment by labor to
productivity-enhancing changes in work rules and
incentive structures, and by management to
invest in the region to maintain and improve the
physical plant. Neither group, working alone, can
significantly improve the region’s ability to com­
pete in national and international markets.

References
Bureau of the Census,

Census of Manufactures,
Geographic Area Statistics U.S. Department of

Commerce, 1977 and 1982.
Garofalo, Gasper A., and Michael S. Fogarty,
“Census Facts,”
The Regional
Economic Issues Program, Cleveland, May,
1984, pp. 26-30.

REI Review,

Hulten, Charles R., and Robert M. Schwab,
“Regional Productivity Growth in U.S. Manu­
facturing: 1951-1978,
1984, pp. 152-162.
Summers, Anita A., and Thomas F. Luce,

nomic Review”Watch.

“The American Eco­

Eco­
nomic Report on the Philadelphia Metropoli­
tan Area, 1985, University of Pennsylvania

Press, Philadelphia, 1985.

Reducing Risk in Wire
Transfer Systems
by E. J. Stevens
E .J . Stevens is an assistant vice
president and economist at the
Federal Reserve Bank of Cleveland.
The author would like to acknowl­
edge useful conversations with
David B. Humphrey.

Introduction
Hundreds of billions of dollars in payments are
made each day in the United States. The system
that enables this enormous sum to change hands
includes several different mechanisms. Probably
the largest number of payments, but with the
smallest total dollar value, are made by using
coins and paper money. Another very large num ­
ber of payments, with a daily total value in the
neighborhood of $75 billion, are made by using
checks, credit cards, and direct transfers through
automated clearinghouses. The smallest number
of payments, but representing by far the largest
total dollar value— frequently $500 billion a day—
are made using so-called wire transfers of funds.
Wire transfers move balances elec­
tronically at Federal Reserve Banks from one
bank’s deposit account to another’s on the same
day. Transfers can be carried out over any of sev­
eral wire networks (large-dollar transfer systems)
connecting banks to one another and to the Fed­
eral Reserve Banks.1 In this way, banks make pay­
ments that handle their own short-term financing
transactions as well as payments on behalf of
themselves and their customers. These payments,
in turn, reflect much of the dollar-denominated
securities and foreign exchange market trading of
the world.
March 27, 1986 was the effective

date of a Federal Reserve Board of Governors’
policy to control risks in large-dollar transfer sys­
tems. Adjustment to that policy has been smooth,
as expected, for two reasons.2 First, consultation
and public comment on the need for and nature
of the program have been ongoing for a number
of years. The actual policy was announced in May
of last year. Since then, both the Federal Reserve
Banks and private consultants have been conduct­
ing informational meetings for banks across the
nation. Second, the risk-control mechanism that
became effective on March 27 embodies only a
modest initial effort at risk reduction. With the
mechanism in place, however, future steps to re­
duce risk become more feasible. How smoothly fu­
ture risk reduction can be assimilated will de­
pend on the ease with which financing practices
of banks and institutional arrangements for mak­
ing certain kinds of payments can adapt to the ris­
ing cost of risk im plied by the risk-control policy.
This article briefly describes sources
of risk in large-dollar transfer systems and dis­
cusses major features of the new mechanism for
risk control.3 Then, examples of potential changes

2

This expectation was supported by a survey done just before
March. See “ Findings: Survey on Implementation Status of

Reduction of Paym ent System Risk," Bank Administration Institute, Ja n ­
uary 23, 1986.

I

3

A full description of the policy m ay be found in “ Policy Statement
Regarding Risks on Large-Dollar Wire Transfer System s” (Docket

The word "bank” will be used here in a generic sense, and in­

No. R-0515), Board of Governors of the Federal Reserve System . Dis­

cludes commercial banks, thrift institutions, Edge A c t and Agree­

cussion of the risk problem is in: E . J . Stevens, "Risk in Large Dollar

ment Corporations, U . S . Branches and Agencies of Foreign Banks, and

Transfer S yste m s,” F R B of Cleveland

New York Article XII Investment Companies.

2-16.

Economic Review,

Fall 1984, pp.

17

in financing and payments practices that might
facilitate future risk reduction are examined.

I. Risk Exposure
The risk being controlled is the threat that pay­
ments made over one of the large-dollar transfer
systems can’t be settled. None of these systems
operates on a real time, cash-in-advance basis that
would continuously settle by deducting each pay­
ment, minute by minute, as it occurs, from the bal­
ance in an account. Instead, they are “batched” set­
tlement systems that update accounts only at the
end of the day by the net of payments and receipts
during the day. It is possible, therefore, for a de­
pository institution to transfer large sums during
the day before it has received all the funds needed
to settle its account at a Federal Reserve Bank. If
the needed funds can’t be acquired, a settlement
failure occurs.
A settlement failure is a rare event
in the United States. Many banks have failed to
open in the morning, but few in modem history
have failed to settle their accounts at a Federal
Reserve Bank the previous evening. Ultimately,
who stands to lose in the event of a settlement
failure depends in part on whether the largedollar transfer system involved is a net settlement
system, or Fedwire. The leading example of a net
settlement system is CHIPS (Clearing House
Interbank Payments System), a private telecom­
munications clearinghouse payments network
operated by the New York Clearing House. Partic­
ipants exchange provisional payments messages
during the day, but payments become final only
at the end of each day when the net position
(receipts, minus payments) of each participant is
settled through accounts at Federal Reserve
Banks.4 Inability of a participant to settle in this
type of system suggests that one or more other
participants or their customers are at risk because
the Federal Reserve will not effect a net settle­
ment order at day’s end if one or more partici­
pants have insufficient balances. O n the other
hand, Fedwire, a wire transfer system operated by
the Federal Reserve, makes payments by transfer­
ring funds directly from one depository’s account
at a Federal Reserve Bank to that of another.5
Inability of a Fedwire user to cover its payments
at the end of a day means that a Reserve Bank

takes the loss, because funds received by a bank
over Fedwire during the day are irrevocable once
notification of a payment is received.
In both cases, risk arises because a
bank can send more funds before the end of a
day than are covered by its initial balance, plus its
receipts, to that point during the day. Such a prac­
tice results in a “daylight overdraft.” For example,
consider a bank continuously borrowing overnight
in the federal funds market: each morning it returns
the previous day’s borrowing over Fedwire, but
can’t actually cover that return of funds until later
in the day when new borrowing has been arranged
and received. The risk is that a bank might be
unable to arrange sufficient new borrowing and
therefore fail to repay its daylight overdraft.
Daylight overdrafts reflect daylight
credit provided to the overdrafting bank either by
the Federal Reserve on Fedwire, or by other
banks on a net settlement system. The practice of
relying on daylight credit creates credit risk for
banks vis-a-vis their customers, for Federal
Reserve Banks vis-a-vis Fedwire users, and for par­
ticipants in net settlement systems vis-a-vis one
another. Systemic risk is also created in the last
case because the unexpected failure of one bank
to settle might have a ripple effect as that failure
makes it impossible for other banks to settle. In
such a case, there is the potential for causing a
classic banking crisis that could disrupt financial
markets worldwide.6 Rapid growth of large-dollar
transfers relative to reserve deposit balances sug­
gests that banks commonly resort to daylight
credit to finance payments during the day.7
The Federal Reserve does not con­
done daylight overdrafts and, until relatively re­
cently, they were probably rare. It was not until
1979 that the first measurement of daylight over­
drafts was taken. Therefore, aggregate values of
transfers relative to banks’ deposit balances at
Federal Reserve Banks is only suggestive of the
likely growth of daylight overdrafts. Transfers
were only about 20 percent of balances in 1950,
150 percent in 1970, but approaching 3000 percent
in the past few years. Now, with use of powerful
computerized accounting systems, it is possible
for a bank to maintain an on-line monitor of its
own and customers’ daylight overdrafts. The Fed­
eral Reserve is able to monitor the daylight over­
drafts of a bank across all large-dollar networks, at
least after the fact. In the future, large-dollar trans-

This brief description simplifies a more complex settlement pro­
cess. Only 22 banks' accounts actually receive a debit or credit at
Federal Reserve Banks. Ten of these banks settle for the remaining 112
participants. A failure might reflect the inability of one of the 22 to settle

6

A thorough analysis of systemic risk is in David B. Humphrey,
"Paym ents Finality and Risk of Settlement Failure: Implications

for Financial M arkets." Paper prepared for Conference on Technology

its own position, or of one of the associate banks to meet its settlement

and the Regulation of Financial M arkets, N ew York University, M ay

obligation with a settling participant.

1985.

5

It is immaterial that the depositories m ay hold accounts at differ­
ent Reserve Banks because the Federal Reserve Banks "settle-

up” among themselves at the end of each day.

7

Marcia L . Stigum cites the example of a large m oney center bank
with daily payments 2V& times its

Market,

total assets. The Money

Hom ew ood, Illinois, D ow Jo nes, Irwin, 1983. p. 585-6.

fer systems conceivably could operate on-line
real time monitors that would prevent the use of
daylight credit completely, thus requiring that
cash be available in advance of each payment.
Daylight credit exposure is not a
unique indicator of risk. Risk depends on the
probability that institutions will not cover their
daylight overdrafts by the end of a day, as well as,
in the event of an actual failure to settle, the
probability that claimants won’t recover some or
all of their loss in the liquidation of a failed insti
tution. Payment system risk then depends jointly
on the amount of daylight credit, on the sound­
ness of institutions in daylight overdraft positions,
and on the ability of depository institutions to
control the amount of payments-related credit
extended to other depository institutions during a
day. Systemic risk—the risk that otherwise sound
institutions w ill be swept up in a cascade of set­
tlement failures— depends as well on the interre­
latedness of institutions in the payments system.
This is influenced heavily by the ability of the
central bank, in its role as lender-of-last-resort, to
prevent or isolate a settlement failure by provid­
ing overnight credit at the end of a day.
Reliance on daylight credit is not
troubling in itself. Rather, it is the uncontrolled
and unrationed provision of daylight credit that is
troubling. As long as daylight credit is unrationed,
risk creation is subsidized and daylight credit
becomes overused. Fedwire has no explicit price
for providing daylight credit and, because there is
no well-developed private market for daylight
credit, has little basis for setting such a price.
Until the current risk-control policy began to be
developed, Fedwire also did not have an effective
limit on daylight overdrafts for any but visible
problem banks.
It can be argued that there is im ­
plicit pricing of daylight credit in net settlement
systems.8 Receivers of funds transfers (suppliers of
daylight credit) face a cost in the form of some pro­
bability of loss. They therefore have an incentive
to lim it the amount of daylight credit they extend
to each other participant. However, this argument
is weak, unless the computerized net settlement
system provides a feature that both allows partic­
ipants to set such limits, and enforces them by
preventing transfers that would breach a limit.
Moreover, the whole argument breaks down
when, as appears to have been the case, there is a
widespread presumption among banks that the
Federal Reserve, as the lender-of-last-resort,
would lend to a participant that is otherwise
unable to settle rather than let a settlement failure
take place and risk a systemic wave of failures.

8

For discussion of risk and pricing, see David L . Mengle, “ Daylight
Overdrafts and Payment System Risks,"

Economic Review,

eral Reserve Bank of Richmond, M ay/June 1985 pp. 14-27.

Fed­

II. The Mechanism for Risk Control
The risk-control policy establishes three require­
ments for every net settlement system: 1) each
participant should be able to set a bilateral limit
on the net amount of daylight overdraft credit it is
willing to extend to each other participant; 2)
each participant should be subject to a lim it on
the amount of daylight overdraft credit it uses; 3)
the net settlement system should include an on­
line monitor to reject or hold payments that
would breach either limit.9
In the case of Fedwire, banks will
be subject to a daylight overdraft lim it in the form
of a dual “cap.” One part of the cap limits a
bank’s average daylight overdraft position during
a two-week required reserve maintenance period.
The other part limits a bank’s overdraft during
any single day of that two-week period.
A potential problem with inde­
pendent daylight overdraft caps for each largedollar system is that they would not distinguish
institutions using only one system from those
using two or more systems. Consequently, each
net settlement system must provide data to the
Federal Reserve so that it can monitor the risk
exposure each bank creates simultaneously over
all systems relative to that bank’s daylight over­
draft cap on Fedwire. If a bank’s overdrafts across
all systems exceed this limit, the Federal Reserve
Bank could counsel the bank and/or advise the
appropriate examiner about the situation, or the
Federal Reserve could reject a bank’s Fedwire
transfers that exceed its overdraft limit.
A bank seeking permission to run
daylight overdrafts must undertake a selfevaluation of its creditworthiness, credit policies,
and operational controls and procedures. This selfevaluation must include a review by its own board
of directors, and the bank must maintain records
as a basis for examiner inspection and comment
to the directors. The bank thereby will establish
its own overdraft limitations, but these must lie
within Federal Reserve guidelines. The guidelines
are expressed in terms of a multiple of the insti­
tution’s capital. (See box.) Should this volunatry
process not be taken seriously, “...the Board (of
Governors) will reconsider its options, including
the adoption of regulations designed to impose
explicit limits on daylight credit exposure.”10
In summary, each depository insti­
tution, including each Federal Reserve Bank, can
now manage the net amount of daylight credit it
extends to each other institution; each institution
must undergo self-evaluation necessary to obtain

9

This was a feature of the Board's interim risk-reduction policy
adopted in 1982.
Policy Statement Regarding Risks on Large-Dollar Wire
Transfer System s, p. 10.

The Cross-System Net Debit ‘Cap’
At the heart of the new risk-control policy is a cross-system
sender net debit cap. The sender is a bank, making pay­
ments over Fedwire. A net debit cap is a dollar lim it on the
amount of daylight credit a bank may draw by sending pay­
ments in excess of the sum of its opening balance and pay­
ments received up to any point during the day on Fedwire.
The lim it is “cross-system” in that, for banks that participate
in net settlement systems such as CHIPS, the amount of day­
light credit allowed under the lim it set on Fedwire will be
reduced by the net amount of daylight credit the bank has
drawn on those net settlement systems.3
Clearly, a bank’s cross-system daylight credit
use, or net debit position, must vary over a day, beginning
and ending at zero, but rising above zero whenever the
opening balance, plus payments received, fall short of pay­
ments made. The cross-system net debit cap has two forms.
One is a limit on the two-week average of a bank’s maximum
daily net debit position, with the average taken over each
two-week required reserve maintenance period. Averaging
provides flexibility for banks to operate within the unpredic­
table ebb and flow of payments traffic, while abiding by the
intent of the risk-control policy. The other form of the cap is
a lim it on a bank’s maximum net debit during each day of
the two-week period. This cap is higher than the two-week
average cap, but effectively puts a limit on the flexibility built
into the averaging process. If a bank is at the one-day limit
for one or more days of the period, then it must be below
the two-week average for one or more days in order to stay
within the average.

Dual Cap
Multiple of Adjusted Primary Capital
Cap Class

High
Above Average
Average
No Cap

Two-W eek Average

2.0
1.5
1.0
0

Plus

Single Pay

3.0
2.5
1.5

0

The Board’s Policy Statement includes a discussion of the
cap-setting procedure banks should employ and how self­
judgements of creditworthiness, credit policies, and opera­
tional controls and procedures might be combined into the
single summary self-classification required to obtain a cap
higher than zero.b

a. The Fedwire limit will not be raised if a bank has been a net supplier o f
credit on a net settlement system.
b. Other details o f the procedure also are included in the Statement, includ­
ing a definition o f adjusted primary capital; treatment o f Edge Act and
Agreement Corporations, U. S. Branches and Agencies o f Foreign Banks, and
New York Article XII Investment Companies; and implications for Book-

a nonzero lim it on the aggregate net amount of
daylight credit it draws from all systems during an
interval; the Federal Reserve Banks will monitor
the daylight overdraft positions of institutions on
Fedwire relative to their self-imposed caps, nor­
mally after the fact, but net of any daylight credit
obtained on other funds transfer networks.

III. Institutional Adjustments for Risk Reduction
Incredulity was a common reaction to early dis­
cussions of reducing risk on large-dollar transfer
systems.11How could half-a-trillion dollars or more
of daily payments possibly be resequenced so
that, with only $20-30 billion of cash deposits,
those payments could still be made, but with less
reliance on daylight overdrafts? Each sender might
wait until enough payments were received before
payments were sent, but every delayed send would,
of course, mean a delayed receipt for someone
else. Given the small cash base and limited time
during which transfer networks are open (the
working day sometimes extended into the even­
ing), the result seemed more likely to be “grid­
lock” than smoothly functioning transfers of funds.
The emphasis on creating a risk-control mechanism
first, with high overdraft limits based on selfevaluation, seems to have submerged this kind of
reaction. But when future steps are taken to use
the mechanism to reduce risk, how will smoothly
operating payments be maintained consistent with
reduced daylight overdrafts and reduced risk?
Two kinds of changes, induced by
market incentives, should take place that could
achieve the desired result. One kind would pur­
chase reduced risk directly, as individual banks
reallocate their operating and portfolio resources
to live within overdraft limits. The other kind
would result from innovations in standard arrange­
ments for interbank payments and financing.
Banks may
reduce the amount of daylight credit they extend
as well as reduce their own use of daylight credit
simply because nationwide attention has focused
on the problem. Heightened awareness and bet­
ter information may bring more prudent behav­
ior. W hile many banks have monitored and man­
aged their own and their customers’ daylight
overdraft positions for many years, others appar­
ently have not. As a result of the educational pro­
gram and preparation accompanying im plem en­
tation of the Board of Governors’ risk policy,
banks now may be less generous in accommodat­
ing other banks’ and customers’ use of daylight
credit, thereby reducing their own need for day­
light credit. Setting more prudent limits, or col-

Direct Risk Reduction:

entry Securities Transfers, Automated Clearinghouses, Net Settlement Ser
vices, and additional matters.

BOX

1

11

See, for example, Stigum.

lecting fees for scheduled extensions of daylight
credit to customers, would have this effect. Sim­
ilarly, with the ability to specify binding bilateral
net credit limits in net settlement networks,
banks may be less generous in accommodating
other banks’ use of daylight credit. Risk reduction
will then result both from reduced daylight over­
drafts and from improved credit quality resulting
from continuous, explicit risk management.
Banks also might delay making
some payments until later in the day in order to
reduce their reliance on daylight credit. O f
course, the resulting delayed receipts might in
crease reliance on daylight credit at other banks.
However, many depositories and customers never
use daylight credit and, in fact, maintain positive
balances throughout the day. Thus, some overall re­
duction in daylight credit is possible through more
careful management of the timing of payments.
Banks could elect to hold larger
overnight balances at Federal Reserve Banks from
which to make payments during the day. This
might seem to be an expensive adjustment cost­
ing a bank the foregone earnings on those extra re­
serves. However, a bank can elect to hold addition­
al sums as a clearing balance on which earnings
credits can be used to pay for priced services. In
either case, banks might make this a part of a leastcostly method of reducing daylight overdrafts.
Risk declines as bank capital
grows, providing more room for institutions to
operate within caps set on a “times capital” basis.
Maintaining a higher capital position might also
seem to be an expensive adjustment, but may be
worth the price. Moreover, many banks are
already adding, or planning to add, to capital as
they adjust to potential loan quality problems and
comply with regulatory guidelines for safety and
soundness. Even without any change in daylight
overdraft practices, more highly capitalized insti­
tutions might present lower risk.
Another fertile field for reducing
daylight overdrafts lies in the liability manage­
ment of depository institutions. About two-thirds
of Fedwire transfers reflect federal funds transac­
tions, as borrowing banks repay the previous
day’s borrowing and then, typically, replace that
with fresh borrowing for the current day. Extend­
ing the maturity of bank financing could yield
substantial dividends in reduced Fedwire traffic
and reduced daylight overdrafts of Federal
Reserve Bank accounts. Risk exposure of the
Federal Reserve Banks certainly would decline,
but risk exposure of others might grow. Longerterm financing would add to lenders’ risk of illiq­
uidity (that is, of using costly methods to meet
unexpected needs for cash) and, all else
unchanged, add to lenders’ and borrowers’ inter­
est rate risk (that is, of unexpected changes in
maturity rate spreads). Uninsured lenders, replac­

ing overnight with longer maturity loans, would
also face a slightly different credit risk. No longer
could they rely on Federal Reserve Banks to
assume credit risk each morning, as they had
when overnight loans had been returned. The
“musical chairs” of repayment thus would be
spaced further apart.
Moving the bearer of risk from
Federal Reserve Banks to private market lenders
does not represent evasion of risk-reduction pol­
icy. W idening the scope of market scrutiny and
the opportunity for risk pricing should be
expected to encourage more conservative behav­
ior by borrowing banks.
Substantial reductions
in daylight overdrafts at individual banks could
emerge from innovations in some long standing
market practices. Some of these innovations
might only evade the risk-control mechanism by
shifting risks outside the monitor, and will not be
acceptable.12 Others would, in fact, reduce risk
and are to be encouraged. Distinguishing
between the innovations will require careful
investigation. The three examples of suggested
changes discussed here might be acceptable if
carefully structured and are offered to indicate
the range of ideas being developed in the market
in response to the risk-control policy.
An alternative to replacing over­
night financing with longer-term borrowing
would be to develop a “rollover” practice in
overnight credit markets. Borrower and lender
might agree that, unless either wished to termi­
nate the entire credit, all or part would be rolled
over at the relevant daily rate each day. A single
daily transfer could cover interest, plus any agreed
change in the outstanding amount of the loan.
This would eliminate the need to transfer the full
amount of borrowing both back and forth each
day. Credit risk from overnight lending would
remain, but would not become a daily daylight
payment risk either for the Federal Reserve or for
participants in net settlement systems.
Access to a rollover loan, as well as
its price, presumably w ould depend on the credit­
worthiness of the borrower as viewed with more
intense lender scrutiny than for a typical over­
night loan today. In this way, the transfer of risk
from Federal Reserve Banks and participants in
net settlement systems should generate incentives
for more conservative behavior by borrowers.
Another substantial portion of the
traffic on large-dollar transfer systems flows
among banks that, for themselves or for dealer
customers, are settling securities or foreign

Innovations:

-1

The Policy Statement (pp. 30-31) specifically, “reaffirms its

_L L d

(earlier) policy that institutions m ay not use Fedwire or other

payments networks as a method of avoiding risk-reduction measures."

21

22

exchange transactions. Current practice typically
involves gross next-day settlement of securities
transactions, meaning that banks send one
another payments for each transaction. Each day,
two banks active in handling security market
operations typically will send each other m ulti­
m illion dollar payments that are more or less
offsetting. These payments are initiated and
received in automated systems on the basis of
trades known in advance because they were done
on the previous day.
The alternative would be for two
banks to offset the payments due to one another,
replacing those two payments with a single trans­
fer of the net difference due to one or the other
institution. Daylight credit risk would be reduced
if the banks adopted new legal agreements defin­
ing obligations to be for this net position rather
than for gross positions.13 Heretofore, the incen­
tive for this kind of economizing on payments
traffic was primarily the cost of a funds transfer—
at most a few dollars per transfer. The additional
incentive of avoiding more costly means of day­
light overdraft reduction might provide the impe­
tus for devising offset arrangements. As in the
case of federal funds rollover, offset payments
would not eliminate all risk. Banks would be ex­
posed to risk of a failure to settle the net amount
due, but the amount at risk would be much
smaller than the gross amounts now exposed.
Development of a day-loan market
is another institutional change frequently cited as
promising daylight overdraft relief. The Federal
Funds market is the source of one-day maturity
loans of cash in the form of deposit balances at a
Federal Reserve Bank. Similarly, a day-loan
market would be the source of loans of cash, but
with same-day maturity. Just as banks may charge
a fee to customers who daylight overdraft their
accounts, so too, for a fee, banks might be able to
borrow and lend cash for repayment later in the
day. Such a procedure seems technologically feas­
ible, especially if it were encouraged by provision
for priority-funds transfer messages that would
bypass a queue of payment orders on large-dollar
transfer systems. Some banks will always have
positive balances that might be loaned to others
who want to make payments but who are at their
daylight overdraft limits.
A day-loan market is not an institu­
tional development that would directly reduce
risk. Rather, it would transfer risk from the Fed­
eral Reserve Banks and the whole set of partici­
pants in net settlement networks to the institutions

-1

O

J.

Such an arrangement, said to be the first of its kind, is
expected to start operating soon in London, involving settle­

ment of foreign exchange transactions among major international banks.
See “ International Financing Review ," Issue 622, M ay 1 7 ,1 9 8 6 , pp.
1496-7.

making day loans. However, it may indirectly
reduce risk by making exposures more visible so
that market discipline would ration credit to risky
institutions with increased certainty.
These three examples of institu­
tional changes— rollover, offset, and day loans—
have not happened yet, but they, and others like
them, suggest promising ways in which market
practices might be expected to adjust to future
efforts to use the new risk-control mechanism to
reduce risk in large-dollar transfer systems.

IV. Concluding Remarks
An important result of the risk-control policy now
in place is that each depository institution’s cross­
system use of daylight credit can be monitored
relative to caps that are themselves related to the
institution’s self-evaluated creditworthiness. Initial
caps are not expected to result in any significant
disruption in large-dollar funds transfer service.
Nonetheless, some depository institutions are
having to adjust their operations to meet the pol­
icy limitations. This, plus the adjustments of other
institutions recently sensitized to the risks, should
at least dampen the growth of daylight overdraft
risk exposures. However, conclusions must await
experience under the new limitations because
payments patterns may change in response to
these initial adjustments, perhaps creating day­
light overdraft problems for institutions that had
not previously experienced them.
Once the situation settles down,
the Federal Reserve Board of Governors fully
expects to move further toward reducing risk,
perhaps, for example, by ratcheting down “timescapital” cross-system daylight overdraft limits. In
the meantime, banks can develop operational and
institutional changes that will reduce and redirect
risk without disrupting the payments system. In
return, Federal Reserve Banks’ risk exposure on
Fedwire should diminish and market discipline
should play a larger role in controlling risk.

Economic Commentary

The Difficulty in Explaining Postwar Stability
KJ. Kowalewski and Eric Kades
December 1, 1985

How Desirable Is Dollar Depreciation?
Gerald H. Anderson
December 15, 1985

A Correct Value for the Dollar?
Owen F. Humpage and
Nicholas V. Karamouzis
January 1, 1986

Bank Holding Company Voluntary
Nonbanking Asset Divestitures
Gary Whalen
January 15, 1986

Junk Bonds and Public Policy
Jerome S. Fons
February71, 1986

The Dynamics of Federal Debt

Can We Count on Private Pensions?

John B. Carlson and E. J. Stevens
July 1, 1985

James F. Siekmeier
February 15, 1986

Is Manufacturing Disappearing?

American Automobile Manufacturing: It’s Turn­
ing Japanese

Michael F. Bryan
July 15, 1985

Solutions to the International Debt Problem
Gerald H. Anderson
August 1, 1985

Medicaid: Federalism and the Reagan
Budget Proposals

Michael F. Bryan and Michael W. Dvorak
March 1, 1986

Should We Be Concerned About the Speed of
the Depreciation?
Owen F. Humpage
March 15, 1986

Paul Gary Wyckoff
August 15, 1985

The Government Securities Market and Pro­
posed Regulation

The Dollar in the Eighties

James J. Balazsyjr.
April 1, 1986

Owen F. Humpage and
Nicholas V. Karamouzis
September 1, 1985

Interstate Banking: Its Impact on Ohio Banks
Thomas M. Buynak
September 15, 1985

The M1 Target and Disinflation Policy
W illiam T. Gavin
October 1, 1985

International Trade and the Fourth
District’s Recovery
Robert H. Schnorbus
October 15, 1985

Reserve Borrowings and the Money Market
Richard L. Mugel
November 1, 1985

Bank Earnings: Comparing the Extremes
Paul R. Watro
November 15, 1985

A Revised Picture: Has Our View of the Econ­
omy Changed?
Theodore G. Bernard
April 15, 1986

Monetary Policy Debates Reflect Theoretical
Issues
James G. Hoehn
May 1, 1986

How Good Are Corporate Earnings?
Paul R. Watro
May 15, 1986

The Thrift Industry: Reconstruction in Progress
Thomas M. Buynak
June 1, 1986

The Emerging Service Economy
Patricia E. Beeson and Michael F. Bryan
June 15, 1986

Economic Review

Quarter I 1985
Beauty and the Bulls: The Investment Character­
istics of Paintings
by Michael F. Bryan

The Reserve Market and the Information Con­
tent of Ml Announcements
by W illiam T. Gavin and Nicholas V. Karamouzis

Quarter IV 1985
Stochastic Interest Rates in the Aggregate LifeCycle/Permanent Income Cum Rational Expec­
tations Model
by Kim J. Kowalewski

New Classical and New Keynesian Models of
Business Cycles
by Eric Kades

Quarter II 1985
Vector Autoregressive Forecasts of Recession
and Recovery: Is Less More?
by Gordon Schlegel

Revenue Sharing and Local Public Expenditure:
Old Questions, New Answers
by Paul Gary Wyckoff

Quarter I 1986
The Impact of Regional Difference in Unionism
on Employment
by Edward Montgomery

The Changing Nature of Regional Wage Differ­
entials from 1975 to 1983
by Lorie Jackson

Quarter III 1985
The Impact of Bank Holding Company Consoli­
dation: Evidence from Shareholder Returns
by Gary Whalen

The National Debt: A Secular Perspective
by John B. Carlson and E. J. Stevens

The Ohio Economy: A Time Series Analysis
by James G. Hoehn and James J. Balazsy, Jr.

Labor Market Conditions in Ohio Versus the
Rest of the United States: 1973-1984
by James L. Medoff