Proposed spendable earnings series retains basic faults of earlier one Essay

On the surface, the new spendable earnings series proposed by
Professor Weisskopf appears to be a considerable improvement over the
series published by the Bureau of Labor Statistics until 1981. Upon
close scrutiny, however, the proposed series is found to share some of
the basic deficiencies that led to the discontinuation of the old one.

Because the proposed series uses gross hourly earnings as its
principal ingredient, it is certainly free of much of the downward
pressure on earnings levels that the secular decline in the length of
the workweek had applied to gross weekly earnings averages, the backbone
of the old spendable earnings series. The fact that Professor Weisskopf
attemtps earnings series. The fact that Professor Weisskopf attempts to
account for average deductions for State and local income taxes–in
addition to those for Federal income taxes and social security
contributions–marks another departure from the old series.

Because of these changes–and, I suspect, primarily because of the
first one–Professor Weisskopf’s series does show a somewhat
steeper upward trend in spendable earnings over the 1950’s and
1960’s than did the discontinued BLS series. To this extent, the
new series would appear to yield a more accurate picture of the actual
trend in earnings for the average full-time worker than was given by the
old series, which was being held down by the expansion of the part-time
work force.

Of more interest, however, is what the two series tell us about the
changes in spendable earnings after both turned downward from their 1972
peaks. Specifically, while the old BLS series showed a decline, of 16.6
percent in real spendable earnings during the 1972-81 period, Professor
Weisskopf’s new series shows a somewhat comparable decline of 13.5
percent over the same period. (See chart 1, p. 41.) The fairly
parallel movement of the two series over this period can lead to only
one conclusion. If the old series was biased downward in portraying the
trend in spendable earnings for the average worker during the
1970’s–and there was ample evidence indicating a large bias–then
the new one, although constructed differently, must also be seriously
biased downward for the period in question.

It must be remembered that the 1970’s were a period during
which the age-sex composition of the work force was changing
significantly, with the proportions accounted for by women and youth
growing very rapidly. The fact that many of these newcomers to the job
market took only part-time jobs had an obvious dampening effect on the
weekly earnings average for all workers. But the hourly earnings
average was also affected–in similar direction, if not in similar
magnitude–by the changing mix of workers and by the growing proportion
receiving lower, entry-level wages.

The extent to which the changing mix of workers affected the
overall earnings average is difficult to quantify. However, some notion
of its impact can be obtained merely by comparing the earnings trends
for all workers with the separate trends for men and women. The
tabulation below shows the percent changes–in constant dollar
terms–over the 1972-81 period both for the payroll-derived series on
gross weekly and hourly earnings (which do not provide any information
by sex) and for the household survey-derived series on weekly earnings,
which are available with some age-sex detail:

Percent change, 1972-81 Payroll series: Mean gross weekly earning
-14.3 Mean gross hourly earnings -9.9 Household series: Median usual
weekly earnings of full-time workers: Total -8.6 Men, age 2k and over
-2.8 Women, age 25 and over -1.4 Men, age 16 to 24 -11.6 Women, age 16
to 24 -12.6

While all of these earnings trends point downward for the period in
question, the gross weekly earnings series, which was the cornerstone of
the BLS spendable earnings series, shows a drop that far exceeded the
decline in weekly earnings among most full-time workers as measured in
the household survey. And the decline in gross hourly earnings,
although somewhat smaller, also appears to overestimate by a
considerable amount the true decrease in real earnings among most

While the household series on median weekly earnings for all
full-time workers did show a decline almost as large as that found in
the payroll series on gross hourly earnings, such was not the case for
the medians for workers age 25 and over. For these workers–who still
make up the bulk of the U.S. work force, and who are still visualized as
the “typical” or “average” workers–real median
weekly earnings showed only minimal declines over the 1972-81 period.
Only for persons 16 to 24 years of age, who are but a small portion of
the full-time work force, was the drop in weekly earnings of the same
magnitude as the changes shown by the two payroll series.

The above comparisons raise serious questions as to whether an
earnings average for all worker groups combined is a good indicator of
the long-term trend in the earnings of most workers, particularly over
periods when the composition of the labor force is changing rapidly.
The problem is that the changes in the earnings averages for a given
group of workers are not always representative of the changes in the
earnings of the “average worker” in the group.

to illustrate, take the following example of a group of workers,
consisting initially of five persons and expanding subsequently to six,
with their individual earnings behaving as follows:

In this case, he earnings average for this group of workers has not
changed at all between the two periods. But could we say the same with
regard to the earnings of the average worker in this group? Would we
not have to conclude that the average worker enjoyed a 10-percent
increase in earnings regardless of what is shown by the average for the
group? (Incidentally, an analogous situation could well develop in
those industries where, on the basis of recently concluded contracts,
newly hired workers are brought on at wages much lower than those
received by workers already on board. In other words, the institution
of a two-tier wage system may bring down the earnings average for the
industry without a decline in the earnings of any of the individual

SUMMING UP, in examining earnings trends it is important to go
beyond the overall averages and to disaggregate the data as far as
possible. While we cannot actually track the earnings of individual
workers (except in isolated experiments), disaggregation of the data by
sex, age, or other characteristics becomes vital when we are dealing
with long-term trends spanning decades. (Where such disaggregations are
not possible, we should be careful not to automatically equate the
changes in earnings averages with the changes in the earnings of the
average worker.)

The use of aggregate numbers is the basic problem with Professor
Weisskopf’s analysis, but it is not the only issue complicating the
analysis of earnings trends and the computation of a “spendable
earnings” series. The fact that more and more of a worker’s
remuneration–or an employer’s labor cost–is in the form of fringe
benefits which are not captured in most earnings data renders the
meaning of any “spendable earnings” series ever more difficult
to conceptualize and explain. And the anchoring of such series to the
earnings information from the establishment survey–which is the case
for the proposed series as it was for the old one–handicaps them with
yet other limitations. For example, the computation of the tax burden
is seriously hindered by the lack of any information on family
composition and total family income. And coverage would be limited to
production and nonsupervisory workers in the private sector–a still
large but gradually declining proportion of the work force.

A better alternative to such series is now available in the form of
the studies of “after-tax money income” initiated recently by
the Bureau of the Census. These studies, based on microdata from the
Current Population Survey, provide very detailed estimates of the
year-to-year changes in the purchasing power of U.S. workers and of the
differences in purchasing power among the principal population groups.
While these studies do not yet provide us the historical perspective on
spendable earnings that Professor Weisskopf’s series attempts to
give us, they are built on much more solid foundations.