View original document

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

April 1,1995

Federal Reserve Bank of Cleveland

Are Wages Inflexible?
by Ben Craig

Low often does an average worker's
hourly wage change? Do employees
and employers avoid wage cuts? When
wage changes occur, do they tend to
come in gradual, small adjustments or
in the form of a single large increase?
Finally, do pay cuts follow the same
pattern as pay raises?
The answers to these questions are of
intense interest to macroeconomists,
because the nature of market wageadjustment processes lies at the heart of
many long-standing debates about the
effects of monetary policy. For example,
workers may resist nominal (actual dollar) wage cuts, but be willing to accept
unchanged nominal wages as the overall
price level rises (and thus see a decline
in their real, or inflation-adjusted,
wages). In this case, it may be in society's best interest to achieve positive,
moderate rates of inflation.
Beyond monetary policy questions are a
host of other important policy issues that
depend crucially on the nature of wage
adjustments in U.S. labor markets. For
instance, cross-regional differences in
returns to workers may persist because
wage compensation patterns respond
sluggishly to changing economic conditions. Also, labor market rigidities may
critically interfere with the private sector's ability to withstand the effects of
increased international competition.
This Economic Commentary documents
two important facts about how wages
change for hourly workers. First, contrary to the common presumption that
wages are "sticky" (that is, they respond

sluggishly to changing circumstances),
nominal wage changes are in fact prevalent. Furthermore, there is little evidence
to suggest the presence of downward
wage rigidity: Actual wage cuts are
common in the data examined here.
Second, we explore the possibility that
only some wages are sticky by examining the relationship between the frequency of wage changes and their magnitude. For example, some might expect
adjustment pressures in sticky-wage sectors to build over time, resulting in relatively large changes when they do occur.
However, our investigation suggests
that, if anything, infrequent wage adjustments tend to be smaller in magnitude
than frequent ones.
Taken together, these facts present a
challenge for proponents of the view that
labor markets are subject to substantial
wage rigidity. Indeed, the evidence presented here makes a compelling case for
the view that wage flexibility is pervasive in the U.S. economy.


The SIPP Data

The characteristics of wage adjustment
can be evaluated with real-world data
that carefully measure hourly earnings.
A particularly useful data set for this purpose is the Survey of Income and Program Participation (SIPP) from the U.S.
Census Bureau. The SIPP examined
15,000 households from October 1983
through May 1986 to observe how people use government transfer programs
such as welfare (or AFDC) to balance out
changes in their lives, such as divorce or
loss of a job. Every four months, people

Using data adapted from the U.S.
Census Bureau's Survey of Income
and Program Participation, the
author shows that wage adjustments
for hourly employees are much more
variable than is commonly supposed.
Workers whose wages change often
are likely to experience larger fluctuations, while those whose wages
change infrequently tend to see smaller movements. Wage flexibility would
appear to be the rule, rather than the
exception, in U.S. labor markets.

were asked if they were paid on an hourly basis and, if so, the amount of that rate
on their last paycheck. The interviews
provided eight separate wage measurements over the period.
The SIPP is especially valuable because
it covers many people at separate moments in a short time frame. Other data
sets survey workers' wages over longer
sample periods, but measure their hourly
pay only once a year.1 This may be less
accurate than the SIPP's measurements
for two reasons. First, people may have
only a hazy recollection of their pay rates
over the entire year. There is no substitute for being able to refer to a recent pay
stub. Second, surveys that depend on
annual recollections often assign a wage
rate from annual earnings (taken from
W-2 income tax forms) divided by the
person's estimate of hours worked over
the past year. Artificial variation may be

introduced into constant annual salaries
as people report different hours worked
from year to year. This estimate also
might change from year to year even if
the hours worked did not.
From the 15,000 original households
chosen for the SIPP survey, we developed a sample that isolated wage
changes for hourly employees. We
excluded salaried workers and those
paid on commission, and removed anyone not continuously employed with the
same firm for the entire period because
of layoff, job switch, or retirement.
Finally, we included only those workers
who responded to all eight interview
cycles in the survey, leaving us with a
final sample of 2,700 people who
reported an hourly wage for all eight
Because salaried workers are excluded,
this sample does not a represent a "snapshot" of the U.S. labor force. A person
on salary is more likely to be older, to
have less variability in earnings, and to
be better educated (and thus able to
negotiate a steadier income) than the
hourly wage earners represented in our
sample. Further, by excluding people
who were not continuously employed
during the two years of the SIPP interviews, we may have left out people who
had sticky wages, but were laid off during some portion of this period because
of their sticky wages. However, the
wage measurements for the SIPP sample
are much closer to the actual hourly
price of labor than are the usual measures of annual earnings divided by a
guess at annual hours.

• Are Wage Changes Infrequent?
Table 1 indicates the percentage of
workers in the sample who experienced
nominal wage changes (out of a possible
seven changes) during the eight interview cycles. For example, the third column (labeled column 2) identifies people whose wages changed twice over the
sample period. The bottom row of this
column tells us that 8.1 percent of the
sample experienced a total of two
changes. The other entries in the column
break down this percentage according to
the number of negative wage changes:
6 percent of the sample experienced two

Total Wage Changes
Wage Changes




























SOURCE: Author's calculations from the SIPP. Numbers may not add up exactly because of
rounding error.

positive wage changes, 2 percent saw
one negative and one positive change,
and only 0.2 percent had two changes
that were both negative.
The far right column summarizes the
share of the sample that experienced
negative wage changes. For example,
because the first row of the table pertains to individuals who had no negative changes, the first entry in the last
column tells us that 22 percent of the
sample experienced no negative adjustments in their nominal pay over the
sample period.
More than three-quarters of the sample
had their wages change four or more
times over the two-and-a-half-year span.
For those who think wages are sticky,
this may seem like an astounding amount
of variation. Equally surprising is that
half of the sample shows two or more
declines in wages during this period.
This is especially interesting because the
sample covers October 1983 to May
1986—the middle of the longest postwar
economic expansion in U.S. history.
A tiny fraction (less than 1 percent of
those with negative wage changes) saw
only negative changes. In other words, a
negative wage change for a person
almost always occurs in the context of
other positive wage changes. Even the
small number of respondents who reported only negative wage changes experienced two or fewer changes during this

period. This is consistent with the view
that wage changes are somewhat independent from period to period. Incurring
a pay cut is a bit like throwing a fair die
and getting a six. If the die is thrown
seven times, many people will get at least
one six. Fewer than one in 100,000 will
roll seven consecutive sixes.3

• Are Wage Changes
Like Earthquakes?
Perhaps variations in the nature and
degree of wage stickiness in various
occupations might explain differences in
the number of wage changes observed
in this sample. If we accept this image of
a sticky wage, then a wage behaves
somewhat like a geological fault. Pressure builds up along the fault until the
earth can no longer hold it back, and the
fault snaps to a new equilibrium level.
The "stickier" the fault, the more violent
the eventual earthquake.
In the case of a sticky wage, wages with
fewer changes should experience more
violent motion when they are finally
forced to change. Wages that change
often should vary only a little each time.
In this case, they are like "slippery"
faults with many mild earthquakes. The
result of this image is that fewer wage
changes imply a large wage change if
people remain in the same job. 4


Larger than 6% cut
5% to 6% cut
4% to 5% cut
3% to 4% cut
2% to 3% cut

One or two wage changes
Six wage changes

1% to 2% cut
0% to 1 % cut
0% to 1% raise
1% to 2% raise


Concluding Remarks

The wage changes in the sample of
workers drawn from the SIPP data show
much greater variability than is often
supposed. Further, because we directly
examine intrayear wage changes, our
analysis points up much variability that
would be disguised in an annual observation. For example, over the two-and-ahalf-year period, less than 10 percent of
the observations posted a net wage
decrease.7 Only 1.4 percent of our sample experienced no wage changes, but
half saw their wages decline two or more
times during part of the nation's longest
postwar economic expansion.
Fewer wage changes are associated
with smaller wage changes, and frequent
wage changes are usually accompanied
by a greater net change overall. The
flexible-wage viewpoint gets strong
support from data in which short-run
hourly wages are measured precisely.8

2% to 3% raise
3% to 4% raise
4% to 5% raise
5% to 6% raise
6% to 8% raise
8% to 10% raise
10% to 15% raise
15% to 30% raise
30% to 60% raise
More than 60% raise

12 14
Percent of observations
SOURCE: Author's calculations from the SIPP, October 1983-May 1986.

We examine whether this implication of
the sticky-wage scenario is evident in
our sample in figure 1, which shows the
frequency of the percentage wage
change for those who experienced one or
two changes (mauve bars) and for those
who experienced six changes (blue
bars).5 If the sticky-wage hypothesis is
true, the blue bars should be clustered
around zero, indicating small percentage
wage changes for people with frequent
pay adjustments. In other words, people
with many wage changes would reach
their final wage after a series of small
movements. The mauve bars, associated
with workers who experience just a few

adjustments in our sample, should not be
clustered around zero, because those
with large changes would tend to move
in a few large jumps.
As is quickly evident from figure 1, the
data do not support this particular
image of the sticky wage. People whose
wages change often are also more likely
to experience larger changes, whether
in the form of raises or cuts. On the
other hand, people whose wages change
infrequently are affected less often, and
when their wages move, they do so in
smaller amounts.6

Many of the stories told about wage
stickiness apply precisely to those workers included in the SIPP sample. Workers covered by collective bargaining
agreements, for example, are more likely
to be hourly wage earners who are
included in the sample. Yet, the empirical evidence strongly suggests that
wages are flexible. More than 87 percent
of the sample experienced three or more
wage changes for eight measurements
made from October 1983 to May 1986.
Wages decreased for more than 87 percent of the workers during the same
time. On balance, evidence seems to
indicate that substantial flexibility is the
rule, rather than the exception.

• Footnotes
1. For two excellent surveys of recent research on wage stickiness based on annual
data (such as the Michigan Panel Study of Income Dynamics), see D. Mitchell, "Keynesian, Old Keynesian, and New Keynesian
Wage Nominalism," Industrial Relations, vol.
32 (1993), pp. 1-29; and Gary Solon, Robert
Barsky, and Jonathan A. Parker, "Measuring
the Cyclicality of Real Wages: How Important Is Composition Bias?" Quarterly Journal
of Economics, vol. 109, no. 1 (February
1994), pp. 1-25.

2. Each person in our sample answered "yes"
to the question, "Were you paid by the hour
on this job?" The wage was the answer to the
question, "What was your regular hourly pay
rate at the end of last month?"
3. Independence of wage changes would be
consistent with wages adjusting quickly to an
independent external shock. The distribution
of wage changes is not quite independent.
There is a "memory" of past changes that
makes similar present changes slightly more
probable. A probability of one-sixth (or the
chance of rolling a six with a fair die) for wage
decreases is quite consistent with the sample.
4. An alternate view might be that an "earthquake" occurs only if the person changes his
job. In this case, the sticky-wage view offers
no prediction of the magnitude of the wage
5. We discarded frequencies for which no
wage change occurred, so that the percentages graphed in the figure represent an estimate of the probability density of the size of
the wage change, given that a change occurs.
Each person with six wage changes contributes six observations to the percentage
wage changes in the blue bars.

Federal Reserve Bank of Cleveland
Research Department
P.O. Box 6387
Cleveland, OH 44101
Address Correction Requested:
Please send corrected mailing label to
the above address.
Material may be reprinted provided that
the source is credited. Please send copies
of reprinted materials to the editor.

6. The same conclusions are evident when
the density of wage changes is graphed in
alternate ways. For example, when the total
wage change for the entire sample period is
graphed for the groups that experienced two
or fewer changes, the total change is much
more likely to be smaller than for the group
experiencing six wage changes.
7. Other studies using an annual sample frequency have exposed fewer wage decreases.
David Lebow, David Stockton, and William
Wascher, in "Inflation, Nominal Wage Rigidity, and the Efficiency of Labor Markets"
(Board of Governors of the Federal Reserve
System, unpublished manuscript, 1994), find
that annual differences in earnings are clustered around zero, which supports a mild
wage stickiness. The differences between
their findings and ours can be explained by
the variation in samples (they use the entire
Panel Study of Income Dynamics, including
salaried workers, as well as annual frequencies). High-frequency data are also important
in measuring relationships between wages
and behavior. See Jean Kimmel and Thomas
J. Kniesner, "The Intertemporal Substitution
Hypothesis Is Alive and Well (But Hiding in
the Data)," Indiana University, Working
Paper 93-014, 1993.

8. However, some care should be taken in
interpreting this observation. Perhaps some
wages are "sticky" exactly because the return
to the employer/firm wage adjustment is
small at any given time. Furthermore, small
deviations from optimal price-setting can
result in significant macroeconomic effects
in some "new Keynesian" frameworks. See,
for instance, Gregory N. Mankiw, "Small
Menu Costs and Large Business Cycles: A
Macroeconomic Model of Monopoly," in G.
Mankiw and D. Romer, eds., Imperfect Competition and Sticky Prices, Cambridge,
Mass.: MIT Press, 1985, pp. 2 9 ^ 2 .

Ben Craig is an economist at the Federal
Reserve Bank of Cleveland. He thanks
David Altig and Thomas J. Kniesner for
helpful comments and Kristin Roberts for
excellent research assistance.
The views stated herein are those of the
author and not necessarily those of the Federal Reserve Bank of Cleveland or of the
Board of Governors of the Federal Reserve

U.S. Postage Paid
Cleveland, OH
Permit No. 385