Use of hourly earnings proposed to revive spendable earnings series Essay

In 1982, the Bureau of Labor Statistics announced the discontinuation of its statistical series on “real spendable weekly earnings of
workers with three dependents,” which had long been used as an
indicator of trends in the purchasing power of U.S. workers. This
monthly series covered all production and nonsupervisory workers in the
private nonfarm economy, and was based on data from the Bureau’s
establishment survey and information on Federal income tax and social
security contribution rates.



According to the series, workers’ real spendable earnings grew
rapidly from 1948 through the mid-1960’s, oscillated around a very
slightly increasing trend for the next decade, and finally dropped
sharply in the late 1970’s. By 1981, the last year for which data
were published, average real spendable earnings had fallen to levels
recorded during the late 1950’s. The implication that the average
worker was no better off in the early 1980’s than in the late
1950’s was profoundly troubling to many economists. Evidence based
on other statistical indicators (such as real per capita disposable
personal income, or the gross weekly earnings of male full-time workers
age 25 and older) suggested no stagnation, let alone decline, in
workers’ purchasing power. Economic statisticians were moved to
scrutinize more carefully the real spendable earnings series, which had
already begun to meet criticism during the early 1970’s, and they
identified a number of apparently serious shortcomings. Criticism of
the old series


The chief concerns of the critics were summarized by BLS economist
Paul Flaim in a January 1982 article in the Monthly Labor Review:



* Since the mid-1960’s, there has been a significant shift in
the composition of the U.S. labor force, with both women and young
workers accounting for an increasing share of the total. Both of these
groups hold part-time jobs with much greater frequency than older male
workers, and tend to have lower paying jobs as well. As a result, a
series based on average weekly earnings for all workers understates the
rate of growth of (a) average hourly earnings, because hours worked per
week have tended to decline; and (b) earnings of any given subgroup of
workers (in particular male family breadwinners), because these better
paid workers constitute a declining fraction of the labor force.



* Many of the assumptions made by the BLS in calculating the
Federal income taxes paid by the “typical” worker were no
longer appropriate. Most importantly, the typical worker is no longer
the head of a household with three nonearning dependents. Moreover, a
sizable minority of workers itemize deductions on their tax returns,
rather than taking the standard deduction as assumed in the calculation
of the BLS spendable earnings series.



* The BLS did not make any allowance for State and local income
taxes paid by workers, deducting from gross earnings only an estimate of
Federal income taxes and social security contributions.



* The BLS Consumer Price Index for Urban Wage and Clerical Workers
(CPI-W), used to deflate current-dollar earings, was a misleading
indicator of the impact of inflation on workers’ purchasing power,
especially (but not exclusively) because of its treatment of housing
costs.



* The whole concept of “spendable” earnings was
inadequate. In addition to take-home pay, one should include in a
measure of a worker’s economic well-being an estimate of the (not
immediately spendable) benefits accruing from (a) employer-provided
medical insurance coverage and private pension plans; (b) social
security benefits; and even (c) public services provided by Federal,
State, and local governments.



Some of the criticisms levied at the old spendable earnings series
are no doubt justified. But others are far from compelling. Following a
discussion of the possible relevance of each of the points noted above,
this article presents a new spendable earnings series that avoids the
genuine shortcomings of the discontinued BLS series.


It is an indisputable fact that adult male workers constitute a
decreasing fraction of the U.S. labor force. But the implication that
one should ignore declines in the average worker’s purchasing power
that result from such a compositional shift (as opposed to declines in
the average purchasing power of particular subgroups of workers) strikes
me as mistaken. While for certain purposes one may wish to inquire into
the changing economic status of particular subgroups of workers, it is
certainly a matter of general interest to know what has been happening
to the purchasing power of the average worker, however the
characteristics of that worker may be changing in other respects.



Nevertheless, there has been a gradual decline in average weekly
hours of work for production and nonsupervisory workers in the U.S.
economy, in part because of the changing composition of the labor force,
and trends in weekly earnings therefore do not accurately reflect trends
in hourly earnings. Because workers presumably derive greater benefits
from the same income if it is received for fewer hours of work, having
thereby more time available for other pursuits, it would appear to make
more sense to base a measure of workers’ purchasing power on hourly
rather than weekly earnings.



There are also problems in using tax formulas applicable to a
household with one earner and three dependents, when the structure of
the typical U.S. household has changed so much in recent decades. And it
would be desirable to avoid the rather arbitrary assumptions about the
Federal income tax return of the typical worker that BLS made in its
calculations. Thus, there is a clear need for an alternative approach
to measuring the fraction of workers’ earnings that is paid in
Federal income taxes. One would also want to take into account the
State and local income taxes paid by workers, given the increasing
importance of these taxes both in absolute terms and relative to Federal
income taxes.



There is continuing debate about the relative merits of the CPI-W
and alternative deflators, such as the U.S. National Income and Product
Accounts, as a measure of trends in the purchasing power of a dollar of
wages. The CPI-W has been criticized for its treatment of housing
costs; but it does have an advantage over the PCE series as a deflator for production and nonsupervisory workers’ earnings in that its
“market basket” of goods and services is designed to represent
the purchases of the typical worker of this kind rather than the typical
consumer. This issue might best be addressed by presenting and
comparing estimates of workers’ real purchasing power calculated
with alternative deflators.



Finally, criticism of the whole concept of spendable earnings as an
inadequate measure of a worker’s economic well-being has undeniable
merit. It should be noted, however, that once one opens up this welfare
economist’s Pandora’s Box, there are a host of other
considerations that begin to suggest themselves. Deferred income or
benefits in kind do not exhaust the factors that contribute to the
overall economic well-being of a worker; it would be impossible to
enumerate all the relevant factors, let alone measure their significance
with any accuracy. Under the cricumstances, it would appear most
desirable to track certain measurable indicators–such as spendable
earnings–while keeping quite clearly in mind their meaning and their
limitations. This I propose to do here; estimating the average
worker’s nonspendable earnings or benefits of any kind is beyond
the scope of this article. A new spendable earnings series



To chart trends in the purchasing power of U.S. workers, I have
developed a new annual time series measuring the average real spendable
hourly earnings of production and nonsupervisory workers in the
nonagricultural private business sector. The new series is not prone to
the bias inherent in a weekly earnings series because it focuses on
hourly earnings; it avoids the problems encountered by the BLS
statisticians in working with Federal income tax formulas for typical
families by making use of direct estimates of the actual effective rate
of income taxation on earners of the relevant income size class; and it
includes a (rough) allowance for State and local income taxes. The
basic series is deflated using the CPI-W but, for purposes of
comparison, an alternative series obtained using the fixed-weight PCE
deflator also is presented.



The basic annual series is calculated by deflating the BLS series
on average gross hourly earnings of production and nonsupervisory
workers in all private nonagricultural establishments by the CPI-W to
obtain the corresponding average gross real hourly earnings series. The
real earnings series is then multiplied by (1 — TRSS — TRFI — TRSI),
where TRSS is the estimated effective social security tax rate on the
average worker’s annual earnings; TRFI is the estimated effective
Federal income tax rate on the average worker’s annual earnings;
and TRSI is the estimated affective State and local income tax rate on
the average worker’s annual earnings.



The above tax rates are estaimted as follows. First, the average
worker’s annual earnings are estimated by multiplying the BLS
series on workers’ average gross hourly earnings by 52 times the
corresponding BLS series on average weekly hours. Then:



* TRSS is first set equal to the social security personal
contribution rate for each year (expressed as a fraction of unity). The
average worker’s annual earnings are then compared with the maximum
taxable wage for social security contributions; in years for which the
former exceeds the latter, TRSS is set equal to the social security
personal contribution rate multiplied by the ratio of the latter to the
former.



* TRFI is set equal to the effective Federal income tax rate on a
taxpayer with an adjusted gross income equal to the average
worker’s annual earnings. This tax rate is determined using
published Internal Revenue Service (IRS) data on sources of income,
deductions, and tax items by size of adjusted gross income (for taxable
returns only). “Total income tax” (after credits) is
expressed as a fraction of “adjusted gross income” (less
deficit) for each income size class, and the effective tax rate for the
average worker’s annual earnings level is determined by
interpolation between the tax rates for each income size class
(attributed to the midpoints of the respective classes).



* TRSI is roughly approximated by multiplying TRFI by the ratio of
total annual State and local government income tax receipts to total
annual Federal government income tax receipts.



The resulting annual real spendable hourly earnings series from
1948 to 1981 is presented alongside the original BLS annual real
spendable weekly earnings series in table 1. To facilitate comparison,
an index (1948 ” 100) is also shown for each series, and the two
indexes are plotted against time in chart 1. According to the chart,
the two series are not all that dissimilar. In both cases, spendable
earnings rise rapidly from 1948 to 1965, oscillate around a much more
modestly rising trend until 1977 (peaking in 1972), and then drop
sharply from 1977 to 1981. By 1981 (the last year for which data are
available in both series), the new series has fallen lower than at any
time since 1963, and the old series is at its lowest level since 1958.
The main difference is that the new series rises slightly more rapidly
over the postwar period as a whole. About half of this difference is
attributable to the fact that workers’ average weekly hours
declined fairly steadily from 40.0 in 1948 to 35.2 in 1981.



The new spendable earnings series thus paints just as troubling a
picture of recent trends in purchasing power as the discontinued BLS
series. The fact that the average U.S. worker has suffered a
significant decline in real spendable earnings cannot be dismissed as a
statistical illusion attributable to deficiencies in the BLS
methodology; rather, it reflects a genuine deterioration in an important
element of the average worker’s economic well-being. Some
additional data



Developments over time in the statistical series underlying the new
spendable earnings series also are of interest. First, chart 2 plots
real gross hourly earnings against real spendable hourly earnings (gross
earnings less estimated taxes). Note that the gross earnings series
displays a pattern similar to that of the spendable earnings series,
except that the slowdown after the mid-1960’s and the decline after
1973 are not as marked. This is clearly due to the fact that the ratio
of spendable to gross earnings fell significantly from the
mid-1960’s on.



Chart 3 shows trends in the three effective tax rates TRSS, TRFI,
and TRSI, as well as the total of the three, between 1948 and 1981. The
steady rise of the effective social security contribution rate is
clearly evident. The effective Federal income tax rate oscillates
around a more-or-less constant rate after rising during the Korean War,
but the corresponding State and local income tax rate shows a distinct
long-run upward trend (especially from the mid-1960’s on).



Finally, chart 4 compares the time pattern of the basic new
spendable ernings series with that of an alternative spendable earnings
series deflated by the fixed-weight PCE deflator rather than the CPI-W.
The overall shape–and the turning points–of the two series plotted in
the chart are very similar. However, the PCE-deflated series does not
turn down quite as sharply after 1972 and after 1977. As a result, is
peaks in 1977 rather than in 1972, and its 1981 value is the lowest
since 1969, rather than since 1963. Because the fixed-weight PCE
deflator did not rise nearly so rapidly over the past decade as the
CPI-W, its use in calculating a real earnings series yields a smaller
decline in purchasing power since 1972. But the alternative series
still conveys a very discouraging impression of the trend in
workers’ purchasing power in recent years. Conclusion



The new annual time series for the average real spendable hourly
earnings of production and nonsupervisory workers in the nonagricultural
private business sector of the U.S. economy avoids some of the
shortcomings for which the discontinued BLS series has been criticized.
And, over the postwar period, it displays a slightly more rapid rate of
growth in workers’ purchasing power. However, like the old BLS
series, the new one indicates that purchasing power declined sharply
through the late 1970’s to reach a 1981 level roughly comparable
with the recorded some two decades earlier.



There are a number of respects in which the new series could be
improved. First, it would clearly be desirable to have the values
available on a monthly as well as an annual basis, as in the case of the
old BLS series. To calculate monthly values for the new series, one
would only have to deflate BLS monthly estimates of workers’
average gross hourly earnings by the CPI-W. The resulting monthly
observations could then be multiplied by the ratio of spendable to gross
earnings (1 — TRSS — TRFI — TRSI) applicable to the year in question.



Second, the new procedure suffers from its dependence on published
IRS Federal income tax data for the estimation of TRFI and TRSI. Because
these data, even in preliminary form, are usually available only after a
lag of 1 to 2 years, it is not possible to provide monthly observations
on the same current basis as the old BLS series. To minimize this
problem, it would be necessary to develop a more approximative procedure
for estimating the current effective Federal income tax rate on the
average worker’s annual earnings. This could be done by
extrapolating from the most recently available annual observation using
data on legislated rates of Federal income taxation, thus borrowing from
the old BLS methodology for the purpose of providing timely preliminary
figures.



Third, there are some problems in using the effective Federal
income tax rate on the average workerhs annual earnings to calculate
TRFI. For example, if the typical worker has some nonwage income in
addition to his or her wages, the effective tax rate on that
worker’s total income will be understated because of the
progressivity of the tax structure. Also, if there are among the tax
returns in the relevant income size bracket some that have been filed
jointly by two-earner couples, the effective tax rate on that income
class will understate the tax rate that would be applicable to workers
who are sole wage-earners in their taxpaying unit. (The latter rate is
the relevant one for the purpose at hand.) Thus, the procedure I have
used to estimate TRFI is subject to a slight downward bias, and
spendable earnings are correspondingly overestimated. However, given
the very modest progressivity of the Federal income tax structure and
the relatively small fraction of workers for whom the above
considerations are likely to apply, the bias is surely very minor.



Fourth, the method I have used to estimate the impact of State and
local income taxation is very rough. A detailed examination of State
income tax data might yield improvement upon my simplifying assumption
of proportionality between Federal and State and local income taxation
across all income classes. However, the evidence in chart 3 indicates
that TRSI is substantially less significant than either TRSS or TRFI;
thus, any bias due to the rough methodology is unlikely to have much of
an impact on the spendable earnings series.



Finally, as one can tell by comparing the two series shown in chart
4, the choice of an appropriate earnings deflator is an important one
for a real purchasing power series–especially for assessing trends
during periods of rapid inflation such as the 1970’s. Because both
the CPI-W and the PCE deflator have their weaknesses, further efforts to
develop a better deflator for evaluating workers’ real spendable
earnings are clearly warranted.