1. American Economic Association
Long Swings and the Nonreproductive Cycle
Author(s): David M. Gordon, Thomas W. Weisskopf, Samuel Bowles
Source: The American Economic Review, Vol. 73, No. 2, Papers and Proceedings of the Ninety-
Fifth Annual Meeting of the American Economic Association (May, 1983), pp. 152-157
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/1816831
Accessed: 26/10/2010 17:13
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=aea.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.
http://www.jstor.org
2. Long Swings and the Nonreproductive Cycle
By DAVID M. GORDON, THOMAS E. WEISSKOPF, AND SAMUEL BOWLES*
The U.S. and world capitalist economies
are currently in the midst of the third long
swing crisis of the past century. A notable
and, we shall argue, defining characteristicof
this and prior long swing crises is the failure
of the business cycle, through its normal
functioning, to restore conditions for rapid
accumulation. We develop a theoretical mod-
el of the relation of the business cycle to long
swing expansions and crises, and present
some evidence supporting our hypotheses
about the relationship between the long swing
crisis and what we term the nonreproductive
cycle. Our approach implies that the most
theoretically coherent basis for dating long
swings in capitalist economies is not to be
found in the movements of total output or
investment, but rather in an investigation of
what Gordon has called the social structure
of accumulation and its ability to restore
profitability during cyclical downturns.
This model of the relationship of cycles to
swings not only helps clarify the characteris-
tic patterns of long swings, but may also help
resolve a number of anomalies which have
occasionally puzzled economists. The consid-
erable rise in real wages during the sharp
cyclical downturns of the 1890's and 1930's
is no longer anomalous, for example, but is
rather an expected feature of business cycles
in a period of long swing crisis. (This does
not explain, of course, why real wages rose.)
Further, the somewhat inconclusive debate
in the Keynesian literature about the pro-
cyclical or anticyclical behavior of real wages
may potentially be clarified by our observa-
tion that real wages tend to move procycli-
cally during long swing expansions and anti-
cyclically during long swing crises-and that
for this reason the prevailing debate has been
clouded by an underspecification of the de-
terminants of cyclical behavior during long
swings.
These anomalies are not surprising. Most
studies of the relationship of the business
cycle to long swings in economic activity
have failed largely because of the absence of
any theory of the relationship between these
two forms of economic fluctuation; the same
models of investment and output determina-
tion are typically applied to both. Recent
Marxian analyses of the dynamics of long
swings offer the possibility of overcoming
this weakness. (See Gordon, Richard Ed-
wards, and Michael Reich, ch. 2, for a
review.) These analyses have stressed the cru-
cial importance of a periodically recon-
stituted set of institutions, called the social
structureof accumulation (SSA), which pro-
vides the economic stability and moderation
of political economic conflict essential for
favorable profit expectations and therefore
for rapid capital accumulation. These institu-
tions include, for example, systems of labor
management, the international monetary sys-
tem, and structures mediating raw materials
supply. Their erosion sets the stage for eco-
nomic crisis.
Despite the promise of this approach,
however, the "normal" operations of the
business cycle have not been integrated into
the broader analysis of the SSA and long
swings. We begin by distinguishing the re-
productive (or well-behaved) cycle from the
nonreproductive (or perverse) cycle. The re-
productive cycle is one in which a downturn
in economic activity is corrected by the func-
tioning of the cycle itself. We call this cycle
reproductive because it endogenously re-
stores conditions for rapid accumulation
without requiring fundamental changes in
the structure of the accumulation process.
The nonreproductive cycle, by contrast, is
one in which a downturn does not correct
itself endogenously, and which therefore re-
quires basic changes in the institutions that
*New School for Social Research, University of
Michigan, and University of Massachusetts-Amherst,
respectively. The order of our names has been selected
randomly. We thank Peter Alexander for research assis-
tance, and Carol Heim and Robert Zevin for helpful
comments and suggestions.
152
3. VOL. 73 NO. 2 LONG WAVES IN ECONOMIC ACTIVITY 153
regulate the accumulation process and estab-
lish the conditions for profitability.
This distinction allows us to define the
difference between long swing expansions
and crises in a particularly simple manner.
Long swing expansions are characterized by
reproductive cycles, sustaining the effective-
ness of the SSA in promoting profitability,
investment, and growth. Long swing crises
are characterized by nonreproductive cycles,
leading to prolonged periods of economic
stagnation or disaccumulation and eventu-
ally, if capitalism is to continue, to the con-
struction of a new SSA capable of rekindling
profitability, investment, and growth. The
theoretical distinction between long swing
expansions and crises thus resides in the
reproductive or nonreproductive nature of
the business cycle; slower economic growth
or reduced accumulation are therefore a
probable consequence of a long swing crisis
rather than its defining characteristic.
These definitions presuppose a specific set
of interconnections among the SSA, the ex-
pected profit rate, the cycle, and investment
activity. The expected profit rate depends on
the effectiveness of those institutions which
make up a given SSA. The level and pattern
of investment depends upon the SSA and the
expected profit rate. Cyclical downturns are
induced by a decline in the expected rate of
profit. Reproductive cyclical downturns re-
store the expected rate of profit, and thus
investment activity, while nonreproductive
cyclical downturns do not. (We provide some
econometric evidence for these propositions
in our 1983 book, and in Weisskopf's 1979
article. One effort at explaining the emer-
gence of the nonreproductive cycle in the
post-World War II period is found in Bowles
and Herbert Gintis.)
Why would a reproductive cycle become
nonreproductive?We may examine the func-
tioning of the reproductive cycle through an
analysis of the determinants of the profit
rate. Abstracting from taxes, we may repre-
sent the profit rate of the individual firm, r,
as the product of the share of profits in firm
value-added SrI the ratio of output to utilized
capital stock yu, and the ratio of utilized to
owned capital stock k*, or
(1) r-sryuk*,
where
(2) r- II/Ko; sr -I / Y; Yu Y/Ku;
k*-KJIKo;
and II is firm profits, Ko is the value of the
firm's owned capital stock, Y is firm value-
added, and KUis the portion of the owned
capital stock which is currently utilized.
The expected profit rate rewill by parallel
reasoning depend on the "full capacity profit
rate" srYu, and expected capacity utilization
k*, or
(3) re- SrYuk*e
The expression in (3) suggests that a cycli-
cal downturn may restore expected profits in
either of two principal ways, by raising the
profit share, Sr, or by raising the ratio of
output to utilized capital, yu. (Downturns do
not generally raise expected capacity utiliza-
tion.) The cyclical downturn may raise yuby
eliminating high-cost firms and by inducing
the nonuse of high-cost processes within
surviving firms. The effect of the cycle on yu
is closely associated with the effect of the
cycle on competitive pressures among busi-
ness and, by implication, with the business
failure rate.
The effect of a cyclical downturn on the
profit share involves a somewhat more com-
plicated set of connections. It may be in-
vestigated by representing the profit share as
unity minus what we call real unit labor
costs, or
(4) Sr- [(wlp)lqe] ,
where w is the nominal wage,p is the price of
output, q is output per unit of labor effort,
and e is labor effort input per hour.
A cyclical downturn may lower the prod-
uct wage (w/p), raise output per unit of
effort (q), or raise labor effort per hour (e).
The effects on w/p and e are derived from
the increased power of capital over labor
associated with an increase in the size of the
reserve army of the unemployed and the
consequent increase in the cost to the worker
of losing his or her job. (These effects are
estimated econometrically in our 1983 paper,
and in Juliet B. Schor and Bowles.) The
4. 154 AEA PAPERS AND PROCEEDINGS MAY 1983
effect of the downturn upon q may arise
because of the firm elimination and competi-
tive pressure effects outlined above, or be-
cause of increased power of capital in de-
termining work rules and technical changes,
or for other reasons.
The nonreproductive cycle is one in which
these restorative effects of the downturn on
the profit share and the ratio of output to
utilized capital fail to operate. The failure of
these effects signals an erosion of the ef-
fectiveness of the institutions of the SSA and
therefore the onset of a long swing crisis.
Under what conditions might they fail to
operate?
The cyclical downturn might not raise the
ratio of output to utilized capital stock if
high-cost firms and high-cost operations were
not eliminated, or if the downturn was char-
acterized by an especially severe contraction
in sectors with relatively high output/capital
ratios, or if the reserve army effects failed to
raise output per effort unit or effort per
hour, or for other reasons.
The cyclical downturn might not raise the
profit share if producer price cutting out-
weighed wage cutting, giving rise to an in-
crease in the product wage, or if the contrac-
tion failed to have positive effects on output
per unit of effort or effort per labor hour.
We do not have space to present a model
of price determination over the cycle which
incorporates both mark-up and limit-pricing
behavior: our model suggests, in brief, that
the negative effects of economic downturns
on product prices are likely to be limited.
The effect of the contraction on the other
components of real unit labor costs-the
nominal wage rate, output per effort unit,
and effort per labor hour-may be derived
from a microeconomic analysis of the capi-
tal-labor conflict over wage setting and work
intensity as modeled in Bowles. For our pur-
poses here, it is sufficient to observe that the
effects of the cyclical downturn on the above
determinants of real unit labor costs will be
more favorable to capital, the more the con-
traction increases the expected duration of
unemployment and the greater is the dif-
ference between the worker'scurrentafter-tax
wage and the worker's expected level of un-
employment insurance and other income-
replacing payments from the government.
In an accounting sense, the effect of a
cyclical downturn on the expected profit rate
will obviously depend on the sum of the
above contradictory effects. More substan-
tively, the overall effect will depend on the
manner in which the social structureof accu-
mulation regulates product markets, labor
markets, the labor process, international ex-
changes, state expenditures, and so forth.
Elaboration of an adequate model of these
relations cannot be pursued in such a short
essay. (Gordon, Edwards, and Reich provide
this kind of analysis for the labor process
and labor market.)
We confine ourselves, much more simply,
to a brief empirical demonstration that our
hypothesized relationships between the busi-
ness cycle and long swings are at least partly
confirmed by the historical data. Because
data on the capital stock and its utilization
are inadequate for cyclical analysis until rela-
tively recent years, we cannot explore the
effects of cyclical downturns on the level of
output per utilized unit of capital stock y,
But available data do permit investigation of
the movements of real unit labor costs in the
U.S. economy for the period from 1890 to
1982. (Our pre-1948 data refer to the manu-
facturing sector alone; data for 1948 and
later years are for the private business sector.
Sources and methods are fully described in a
data appendix, available from the authors.)
We use these data to classify cycles as
reproductive or nonreproductive. A nonre-
productive cycle is defined as one in which
the ratio of the product wage to output per
hour, or real unit labor costs [(w/p)/qel
rises rather than falls between the business
cycle peak and the year following the trough.
(We measure changes from peak to one year
after the trough to allow time for the full
restorative impact of the downturns.)
Our results in column (3) of Table 1 indi-
cate alternating periods of nonreproductive
and reproductive cycles, with the nonre-
productive or crisis periods spanning the
years 1890-1903, 1926-37, and 1969 to the
present- the cycles numbered 1-4, 11-12,
and 19-21, respectively. There is also a
noticeable long swing pattern in the data in
column (3), as one can see by graphing its
cycle values against time. (We omit the graph
for reasons of space.) This impression is sus-
5. VOL. 73 NO. 2 LONG WAVES IN ECONOMIC ACTIVITY 155
TABLE 1-THE RESTORATION OF PROFITABILITY IN CYCLICAL DOWNTURNS, 1890-1981
Percentage Average Annual Change
Change in
Product Output Real Unit Unemployment
Cyclical Downturn Wages per hour Labor Costs Rate
Cycle Peak Trough (1) (2) (3) (4)
1 1890 1891 0.78 0.00 0.78 1.4
2 1892 1894 3.84 -0.70 4.54 15.4
3 1895 1897 1.63 -0.67 2.30 0.8
4 1899 1901 2.03 1.45 0.58 - 2.5
5 1903 1905 -0.93 2.55 - 3.48 0.4
6 1907 1909 - 3.24 2.67 - 5.91 2.3
7 1910 1912 3.05 4.66 - 1.61 - 1.3
8 1913 1915 - 1.69 6.28 -7.97 4.2
9 1919 1922 5.87 10.83 -4.96 5.3
10 1923 1925 2.28 6.30 -4.02 0.8
11 1926 1928 4.29 3.32 0.97 2.4
12 1929 1933 1.50 1.35 0.15 21.7
13 1937 1939 2.40 3.72 - 1.32 2.9
14 1944 1947 -2.94 - 1.34 - 1.60 2.7
15 1948 1950 3.92 4.46 - 0.54 1.5
16 1953 1955 0.96 2.73 - 1.77 1.5
17 1957 1959 2.57 3.13 -0.56 1.2
18 1960 1962 3.14 3.48 -0.34 0.0
19 1969 1971 2.33 2.17 0.16 2.4
20 1973 1976 1.17 1.07 0.10 2.8
21 1979 1981 0.85 0.45 0.39 1.8
Notes: The NBER peak-trough-peakcycle of 1918-1919-1920 was ignored as a cycle, since the respective unemploy-
ment rates for these years were 1.4, 1.4, and 5.2 percent; we designate 1919 as the peak instead. Cycle peaks are those
identified by the NBER; troughs are one year after the NBER trough, except that if this is another peak year, the
NBER trough itself is used.
tained by evidence of autocorrelation when
the data in column (3) are tested against the
null hypothesis of constancy over time. The
Durbin-Watson statistic for such a test is
0.90, significant at 5 percent and substan-
tially lower than for comparable tests on the
between-cycle movements of data on pro-
ducers' prices, aggregate output, or aggregate
investment.' Our model of the effects of
cycles on expected profits, in short, reveals
clearer evidence of long swing behavior than
other macroeconomic indicators on which
economists have previously concentrated.
Additional analysis supports our hypothe-
ses about the differences between repro-
ductive and nonreproductive cycles. In the
reproductive cycle, we would expect a sys-
tematic inverse relationship between the
peak-to-trough changes in unemployment
and real unit labor costs: the greater the
reserve armyjolt, the larger the reduction in
real unit labor costs during the downturn. In
the nonreproductive cycle, given our hy-
pothesized erosion of the SSA, we would
expect a breakdown of this effect; we should
therefore find no evidence of any systematic
statistical relationship between the move-
ments in unemployment and real unit labor
costs. We have combined the reproductive
cycles (5-10 and 13-18) in one group, and
the nonreproductive cycles (1-4, 11-12, and
19-21) in another. Regressing the data in
column (3) on the data in column (4), we
find a negative and statistically significant
coefficient for the group of reproductive
cycles and accept the hypothesis of statistical
independence for the nonreproductive group.
'We performed these tests on data for the wholesale
price index, gross private domestic product, and gross
private domestic investment. The Durbin-Watson statis-
tics for these tests were 1.28, 1.37, and 1.12, respectively.
6. 156 AEA PAPERS AND PROCEEDINGS MAY 1983
TABLE2-GROWTH OF OUTPUT AND INVESTMENTIN THEU.S.ECONOMY,1860's- 1979a
Boom Crisis Boom Crisis Boom Crisis
1860's 1892 1899 1929 1937 1973
to to to to to to Averages
1892 1899 1929 1937 1973 1979 Boom Crisis
Gross Domestic
Nonfarm Product 6.6 2.9 3.7 -0.8 4.2 2.9 4.8 1.8
IndustrialProduction 5.1 3.4 4.5 0.4 4.9 2.7 4.8 2.2
Gross Domestic Private
Fixed Nonres. Investment 8.0 1.3 2.4 -4.7 4.7 2.8 5.0 -0.2
aAll entries are percentage average annual changes in constant dollars.
TABLE3-AVERAGE EFFECTSOF CYCLICALDOWNTURNSa
Real Unit Unemployment
Labor Costs Rate
Cycles (1) (2)
Crises:b 1-3 2.54 5.9
11-12 0.56 12.0
19-21 0.22 2.3
Average 1.18 6.1
Expansions: 4-10 - 3.91 1.3
13-18 - 1.02 1.6
Average - 2.58 1.4
aColumns (1) (shown in percent) and (2) represent the averages for the respective
cycle groups of the data presented in columns (3) and (4) in Table 1.
bWe include for this comparison the cycle identified as the onset of the crisis, or the
first nonreproductive crisis, whereas we excluded this cycle in our presentation of the
data in Table 2.
Covariance analysis further confirms the hy-
pothesis of differences in both intercepts and
slopes between the two groups.
We can next explore the relationship be-
tween the data in Table 1 and conventional
hypotheses about long swings in economic
activity. The first nonreproductive cycle in
each sequence in column (3) (cycles 1, 11,
and 19) appears to precede by one cycle
those periods which are generally char-
acterized as crises or depressions.2 We pre-
sent in Table 2 the results of dating eco-
nomic crises proper, as opposed to their
onsets, as commencing after one nonrepro-
ductive cycle. (We have dated the end of the
1890's crisis at the NBER peak of 1899,
discounting the nonreproductive nature of
the 1899-1903 downturn, since the un-
employment rate actually fell during this
"contraction.") Data for aggregate output,
industrial production, and aggregate invest-
ment all demarcate distinct epochs of relative
growth and stagnation.
Table 3 presents the average effects of
cyclical downturns, averaging the data from
columns (3) and (4) in Table 1, for crisis and
expansion periods defined by this dating sys-
tem. There are sharp and consistent dif-
ferences between the crisis and expansion
periods by these measures as well.3
2Since we do not have comparable data for the years
before 1890, we cannot be certain, of course, that the
cycle beginning with the peak in 1890 was the first
nonreproductive cycle in this particular sequence.
3Some readers may question our use of the NBER
dating of peaks and troughs rather than a cycle dating
by peaks and troughs in the unemployment rate. We
prefer our choice since we are concerned with the effects
of downturns in business activity in general, not simply
with movements in unemployment; there are two
7. VOL. 73 NO. 2 LONG WAVES IN ECONOMIC ACTIVITY 157
We note, finally, that our dating identifies
crisis periods which correspond to eras in
which historians have identified intense in-
stitutional conflict and innovation. What
have been termed "key" presidential elec-
tions, those of 1892 and 1896, and of 1932
and 1936, fall within our first two crisis
periods. All three crisis periods, defined by
our method, have witnessed intensified class
conflict and sharp debates over major eco-
nomic policy issues.
We conclude that this empirical evidence,
however provisional, establishes strong initial
support for our model of the relationship
between the business cycle and long swings.
"perverse" business cycle downturns, indeed, which did
not result in increases in the unemployment rate. Our
results are robust if we use an analogous unemployment
dating scheme, in any case: the numbers corresponding
to the averages in column (I) of Table 3 are + 2.66,
+ 0.94, + 1.23, + 1.55, - 1.93, -0.98, and - 1.49 per-
cent, respectively.
REFERENCES
Bowles, Samuel, "Competitive Wage De-
termination and Involuntary Unemploy-
ment," mimeo., University of Massachu-
setts-Amherst, 1981.
and Gintis, Herbert, "The Crisis of
Liberal Democratic Capitalism: The Case
of the United States," Politics and Society,
Winter 1982, 11, 51-92.
, Gordon, David M., and Weisskopf,
Thomas E., Beyond the Waste Land: A
Democratic Alternative to Economic De-
cline, New York: Doubleday Books, 1983.
Gordon,David M., "Stages of Accumulation
and Long Economic Cycles," in T.
Hopkins and I. Wallerstein, eds., Processes
of the World-System, Beverly Hills: Sage
Publications, 1980.
, Edwards,Richard,and Reich, Michael,
Segmented Work, Divided Workers: The
Historical Transformationof Labor in the
UnitedStates, New York: Cambridge Uni-
versity Press, 1982.
Schor,JulietB., andBowles,Samuel,"The So-
cial Wage and the Labor Process: Measur-
ing Some Influences on Worker Resis-
tance," mimeo., Williams College, 1982.
Weisskopf,ThomasE.,"Marxian Crisis Theory
and the Rate of Profit in the Postwar U.S.
Economy," CambridgeJournal of Econom-
ics, December 1979, 3, 341-78.
, Bowles,Samuel,andGordon,DavidM.,
"Hearts and Minds: A Social Model of
Aggregate Productivity Growth in the
United States, 1948-1979," mimeo., Eco-
nomics Institute of the Center for Dem-
ocratic Alternatives, 1983.