Cyclical Patterns of Employment, Utilization and Profitability
University of Massachusetts - Amherst
ScholarWorks@UMass Amherst Economics Department Working Paper Series
Economics
2010
Cyclical Patterns of Employment, Utilization and Profitability Ben Zipperer University of Massachusetts - Amherst, [email protected]
Peter Skott University of Massachusetts - Amherst, [email protected]
Follow this and additional works at: http://scholarworks.umass.edu/econ_workingpaper Part of the Economics Commons Zipperer, Ben and Skott, Peter, "Cyclical Patterns of Employment, Utilization and Profitability" (2010). Economics Department Working Paper Series. Paper 96. http://scholarworks.umass.edu/econ_workingpaper/96
This Article is brought to you for free and open access by the Economics at ScholarWorks@UMass Amherst. It has been accepted for inclusion in Economics Department Working Paper Series by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected].
DEPARTMENT OF ECONOMICS
Working Paper
Cyclical patterns of employment, utilization and profitability by Ben Zipperer and Peter Skott Working Paper 2010-02
UNIVERSITY OF MASSACHUSETTS AMHERST
Cyclical patterns of employment, utilization and pro tability Ben Zipperer
∗
and Peter Skott
†
February 27, 2010
Abstract The interaction between income distribution, accumulation, employment and the utilization of capital is central to macroeconomic models in the `heterodox' tradition. This paper examines the stylized pattern of these variables using US data for the period after 1948. We look at the trends and cycles in individual time series and examine the bivariate cycical patterns among the variables.
JEL classi cation: E12, E32, O41 Key words: growth, business cycles, aggregate demand, instability, income distribution, utilization rate, investment function, pricing. ∗ Department of Economics, University of Massachusetts Amherst, MA 01003, USA; [email protected]. † Department of Economics, University of Massachusetts Amherst, MA 01003, USA; [email protected].
1
Introduction
The `heterodox' tradition in macroeconomics contains a wide range of models. Kaleckian models treat the utilization rate as an accommodating variable, both in the short and the long run. Goodwin's celebrated formalization of Marx, by contrast, take the utilization rate as xed and looks at the interaction between employment and distribution. Distribution is also central to Kaldorian and Robinsonian theories which, like Goodwin, endogenize the pro t share and take the utilization rate as structurally determined in the long run but, like the Kaleckians, view short-run variations in utilization as an intrinsic part of the cycle. The dierences in these and other areas are important, and this diversity of views on core issues is no cause for celebration. Real-world economies dier widely, of course, and appropriate models will re ect these dierences.
One would not want to use the same model to analyze advanced, labor-
constrained economies and dual economies with large (hidden) unemployment.
But an
attempt should be made to resolve fundamental disagreements on the dynamics of accumulation, income distribution and utilization rates in an advanced capitalist economy. Theoretical and empirical work is needed to evaluate the theories and help create a coherent and convincing alternative to the existing orthodoxy. This paper makes no attempt to address the theoretical issues.
Instead, we explore
US data on some of the variables that most, if not all heterodox macroeconomic theories consider important: utilization.
employment rates, pro t shares, accumulation rates, and capacity
1 The paper complements and extends recent studies of US cycles by Barbosa-
Filho and Taylor (2006) and Mohun and Veneziani (2008).
Both of these contributions
use the Goodwin model as their theoretical framework; the former focusing exclusively on oscillations in pro ts and utilization, and the latter on pro ts and employment. The paper is in 5 sections. Section 2 discusses some general data issues. Trends and cycles of the variables are presented in section 3, and section 4 examines bivariate cyclical patterns among the variables. A nal section summarizes the main ndings.
2
Data
Benchmark versions of the heterodox models are typically cast in terms of closed economies without a public sector, and a one-good assumption implies that sectoral dierences are taken to be insigni cant. In short, the models describe a pure capitalist economy. Abstract models of this kind can be very useful.
They serve to highlight particular
issues and can be used to examine the logic of mechanisms that may operate within the more complex reality of real-world economies too. But questions arise if one wants to take the simple models to the data. It is hard to evaluate the models since their predictions can be negated by other in uences explosive cycles for instance may be tamed by the
1
Theoretical issues are discussed in a companion paper, Skott and Zipperer (2010). See also Skott (2010).
1
in uence of public policy and it may not even be obvious which data should be used as the empirical counterparts to the theoretical variables. Our approach in this paper is to look at sectors that, according to the theories themselves, should conform to the behavioral assumptions. In practice, this criterion suggests a focus on pro t shares, accumulation and capacity utilization for the corporate business sector.
Employment, however, is dierent.
The state of the labor market the size of
the reserve army of labor is important because it aects the balance of power between workers and capital, and the relevant measure is the average economy-wide employment rate. Other questions concern the precise de nition of the variables. How for instance should one treat taxes in the calculation of pro t shares? The answers to many questions of this kind depend on the precise purpose of the analysis.
There may be no uniquely correct
answers and even if there is, data limitations restrict what can be done in practice. Given these problems, the best one can do may be to be clear about the de nition of the chosen indicators and to test, when possible, the sensitivity of the observed patterns to changes in the choice of indicators. We are interested in cyclical uctuations as well as long-term trends, and high-frequency data therefore are desirable. US employment and utilization data are available monthly, but corporate-sector pro t data are published on a quarterly basis and unless noted, all data below are quarterly averages. Data are seasonally adjusted by the reporting agency, and we assume that these adjustments adequately correct for seasonal eects. The sample begins in the rst quarter of 1948, the earliest available US employment rates, and ends in the nal quarter of 2008, the last full year of available data at the time of writing.
2
In addition to actual data observations, we smooth the data to construct a short-term and a long-run trend. The smoothing procedure employs the Hodrick-Prescott (1997) lter with smoothing parameters 6.25 and 129,600 for short and long-run trends respectively. While the parameter choice is largely arbitrary any `small' and `large' numbers would be suitable for our descriptive analysis these speci c short- and long-run parameters are recommended by Ravn and Uhlig (2002) for business cycle analysis of annual and monthly data, respectively. Our smaller-than-usual smoothing parameter for the short-run trend is intended to lter out extreme peaks only, leaving considerable quarter-to-quarter variation. Our higher-than-usual smoothing parameter for the long-run trend re ects our eort to describe a long-term variation over decades. Because the actual 1948q1 and 2008q4 observations may bias the constructed endpoints of the long-term trend, we restrict the long-term trend to 1953-2001. That is, after creating the ltered series, we drop the years 1948-1952 and 2002-2008 from the constructed long-term trend, beginning and ending the long-term trend close to the 1953q2 and 2001q1 NBER business cycle peaks.
2
All US data in this paper were collected in September 2009. The US national accounts data used include the ` nal' 2008q4 gures for pro ts and output, as well as the BEA's July 2009 comprehensive revision. OECD data were collected in December 2009.
2
3
Trends and cycles
3.1 Pro t shares To measure the pro t share
π,
we use the surplus and compensation subcategories of
quarterly value added, net of depreciation, in the Bureau of Economic Analysis (BEA) National Income and Product Accounts (NIPA).
3 The largest private sector in the NIPA
tables for which quarterly compensation and operating surplus delineations are available is domestic corporate business. As a share of the total business sector, corporate business net value added rose during 1948-2008, from 57 percent in 1948q1 to 64 percent in 2008q4;
4 Although
as a share of total GDP it remained roughly constant at about 50 percent.
to some extent dictated by data availability, the use of corporate sector data can also, as noted above, be justi ed by a cleaner application of the theories to corporate than to noncorporate business. Net value added (gross value added net of depreciation) is fully decomposed into taxes on production, labor compensation, and net operating surplus.
In the BEA national
accounts, net operating surplus includes taxes on corporate income (which are dierent than production taxes). How one treats taxes can therefore in principle produce dierent proxies for the pro t share. For example, three candidate series are
π= π∗ = π0 =
net surplus + production taxes net surplus + production taxes + compensation net surplus net surplus + compensation net surplus - corporate income taxes net surplus - corporate income taxes + compensation
Over 1948-2008, the means for
π, π∗,
and
π0
are, respectively, 0.29, 0.21, and 0.15. The
decision to omit or include production taxes yields series with virtually identical variation: the correlation coe cient between
π
and
π∗
is
0.98.
In contrast, the treatment of corpo-
rate income taxes yields series with dierent long-term variation (but similar short-term uctuations). Figure 1 shows the actual quarterly evolution of
π
and
π0,
as well as their
short- and long-run Hodrick-Prescott ltered trends. Focusing on the pro t share inclusive of all taxes, pro ts accounted for one-third of output at their peak in the early 1950s. A substantial redistribution towards labor income followed, and the pro t share temporarily fell below one-quarter around the onset of the 1980 recession before swinging upwards again in the 1980s and 1990s.
Using the pro t
0 share π exclusive of all taxes, the long redistribution toward labor after the 1950s fails 0 to occur. The long-term series for π remained around 14% until around the 1980s, after which the share of pro ts began to increase substantially.
3 4
See http://www.bea.gov/national/nipaweb/index.asp See BEA NIPA Tables 1.3.5 and 1.14.
3
These long-term movements
capture most of the dierence between the two series. Although the correlation coe cient between actual quarterly observations for
π
and
π0
is 0.42, deviations of each series from
its long-term trend correlate at 0.92. In simple models pro t maximizing markups over cost are unaected by a change in the tax rate on pro ts, and for many purposes it will be reasonable to include taxes with
5 In section 4 unless otherwise noted we therefore take the pro t share 6 sum of the surplus and taxes, divided by value added.
pro ts.
FIGURE 1 - US quarterly pro t shares
π
and
π0;
π
to be the
actual, and short-term and long-term
trends A potentially large problem with the above de nition of the pro t share concerns its treatment of executive pay, which NIPA tables include as compensation.
Executive
compensation has increased dramatically and arguably a large part of this increase should be included with pro ts. Our measure of the pro t share fails to do this and may give a misleading picture of the trend in pro tability, especially for the period after the 1980s. Krueger (1999) attempts to account for this issue by modifying the NIPA data using a compensation series derived from the Bureau of Labor Statistics Employment Cost Index (ECI) whose data exclude jobs in which employees have a signi cant role in setting their own wages which grew more slowly over 1988-1995 than labor's share of income in the NIPA.
7 . After using the ECI data to modify the NIPA-based pro t share, Krueger nds
that instead of growing only 0.6 percentage points over 1988-1995, the modi ed pro t share grew somewhere between 1.9 and 4.6 percentage points, depending on the nature of the adjustment. A similar adjustment to our data would steepen the most recent long-term rise in the pro t share, although it is unclear,
a priori, what impact this adjustment would
have on the cyclical component of the pro t share around its trend.
3.2 Utilization rates We use the Federal Reserve capacity utilization series for manufacturing.
The Federal
Reserve also publishes a capacity utilization series for the total industrial sector but this series only exists since 1967.
8
The manufacturing series may be more reliable and the
movements in the total industrial series closely match those for the manufacturing series. The average over 1967-2008 of the former is 81.3% compared to 80.1% for the latter, and the correlation coe cient between the two series is 0.99. The Federal Reserve manufacturing series is monthly and seasonally adjusted, and we take quarterly averages for the 19482008 period. Over the period, utilization uctuates strongly with extreme values above 90
5
The invariance of the optimal markup is subject to quali ations. Consumer loyalty or other intertemporal elements in demand can make it optimal for a rm to vary its markup in response to uctuations in the tax rate. 6 See lines 4, 7, and 8 of NIPA Table 1.14. 7 The ECI data begin in 1975: http://www.bls.gov/news.release/eci.tn.htm 8 See Shapiro (1989) for a review of how the Federal Reserve calculates capacity and utilization. 4
percent and below 70 percent (Figure 2). In January 1986 the Federal Reserve switched from SIC to NAICS industry classi cation, possibly introducing a discontinuity that may aect statistical analysis. We ignore this complication. FIGURE 2 - US quarterly manufacturing utilization rate
u; actual, and short-term and
long-term trends
3.3 Accumulation and capital capacity To approximate changes in the capital stock, there are two candidate data series for the
9
US: Bureau of Economic Analysis (BEA) net xed assets , and the Federal Reserve (Fed)
10 Both series are available for the manufacturing sector. industrial capacity index.
The capacity index measures the greatest level of output each plant ... can maintain within the framework of a realistic work schedule. monthly. To calculate quarterly capacity changes
ˆ K
11
The capacity series is published
from the monthly published data, we
calculate the percent dierence between index values three months apart: fourth quarter
ˆ K
is the percent dierence between December and September values; rst quarter
ˆ K
is
calculated from March's index value and the previous year's December value. The average quarterly growth rate of manufacturing capacity was about 0.9 percent between 1948-2008. The growth rate generally fell over the period, with large peaks in the mid-to-late 1960s and mid-to-late 1990s. The same 1986 SIC to NAICS industrial classi cation change mentioned above for the utilization series may also aect the capacity series, a complication we ignore. FIGURE 3 - Quarterly Fed-based
ˆ; K
actual, and short-term and long-term trends
While the Fed series is monthly, the BEA data are annual. This is a serious disadvantage for the analysis of cyclical patterns and there may also be other reasons to prefer the Fed data. Heterodox models usually assume a xed coe cient production function and a constant rate of depreciation. If these assumptions are satis ed, indicators of production capacity (the Fed data) and capital stock (BEA) will coincide, if correctly measured. If the assumptions are relaxed, however, the two indicators can deviate, and the economic argument behind the standard investment functions concerns the desired increase in capacity. For some purposes, at least, the Fed data therefore may be preferable on theoretical grounds (as well as because of their high frequency). The BEA series is based on the end-of-year capital stock, and we calculate its annual
n − 1.
To
determine annual changes using the Fed series, we calculate the rate of change in year
n as
changes in year
n
as the percent dierence between values in years
n
and
9 See the manufacturing industry entry in BEA Table 4.2 of non-residential xed assets: http://www.bea.gov/national/FA2004/ 10 See http://www.federalreserve.gov/releases/G17/caputl.htm. 11 Corrado and Mattey (1997) describe how the Federal Reserve capacity series is constructed. See also Shapiro (1989).
5
the percent dierence between December values in years
n and n−1.
Figure 4 shows actual
annual rates of change for the Fed and BEA series, as well as their smoothed, long-term trends. For the long-term trends of these annual data, we use a smoothing parameter of 1000, which results in long-term annual rates of change similar to the long-term annual
12
rates of change from quarterly data.
Over 1948-2008, the annual Fed-based rate of capacity changes average about 3.5%, whereas the annual BEA accumulation rates average 2.7%. The two annual series generally move together, with a correlation coe cient of 0.74, indicating some support for using xedcoe cient production functions. Signi cant dierences between the series occur in 19491950 and 1958-1960, when the Fed series exceed the BEA series by about 2-3 percentage points, and in 1996-1999, when the gap is about 3-5 percentage points. In 1974, the BEA series is more than 2 percentage points larger than the Fed series. Since 1982, the Fedbased measure has exceeded that for the BEA, a trend consistent with increasing reliance on computer-based technology. For example, the the large relative run-up in the Fed-series in the 1990s may be viewed in terms of rms realizing productivity gains from computer equipment. The BEA series may show small increases in the capital stock because prices for computers, relative to other machinery, fell dramatically over that period. FIGURE 4 - Annual Fed-based and BEA-based
ˆ; K
actual and long-term trend
3.4 Employment rates We measure the employment rate
e
as one minus the seasonally adjusted unemployment
rate from the Bureau of Labor Statistics (BLS) Current Population Survey (CPS).
13 This
economy-wide de nition of the employment rate hides all regional, sectoral and skill-speci c dierences in labor market conditions.
The implied BLS measure of movements in the
total labor force is in line with common simplifying assumptions of a constant growth rate in the labor force, but the de nition avoids complications from historical shifts in the US employment-to-population ratio that may have been endogenous (for example due to female labor market entry). As shown in Figure 5, the employment rate peaked at more than 97 percent in the early 1950s and dropped to less than 90 percent during 1982-1983. Over 1948-2008, however, it rarely escaped the 92-96 percent range. In a comparison of US-Census-based and BLS Current Population Survey-based employment rates, Schmitt and Baker (2006) noted that the BLS may increasingly be understating the unemployment rate around the order of several tenths of a percentage point. We ignore this complication, although accounting for it would slightly lower the most recent end of the long-term employment trend.
12
Using the same long-term smoothing parameter we used for quarterly data on the annual data would result in an approximately linear long-term series. 13 See http://stats.bls.gov/data/.
6
FIGURE 5 - US quarterly employment rate
e;
actual, and short-term and long-term
trends
3.5 Trends The considerable short-term variation in the utilization series may overshadow relatively modest long-term movements. Figure 6 graphs the percentage-point dierence between the mean and long-run Hodrick-Prescott values for the pro t share, utilization rate, employment rate and the annual accumulation rate. The span between maximum and minimum long-run dierences from the mean of the employment rate is modest at 2.7 percentage points.
The span of deviations from the mean is similar for both pro t shares and uti-
lization rates, 4.4 percentage points for the pro t share and 5.7 percentage points for utilization. But the mean values are dierent and the relative long-term variation for the pro t share is much larger than for the utilization rate: the span between the maximum and minimum of the proportional deviation from the mean is 2.8 percentage points for the employment rate, 7.0 percentage points for the utilization rate, and 15.6 percentage points for the pro t share. Annual Fed-based accumulation rates average about 3.6 percent, and the span of absolute dierences from this mean is about 2.0 percentage points (corresponding to a span of proportional dierences of 53.1 percentage points). FIGURE 6 - Dierence between the US mean and long-run trend, for utilization, employment, pro t shares, and quarterly accumulation rate
4
Bivariate cyclical patterns
4.1 Employment - pro tability Figure 7 contains two time-connected scatterplots of US employment rates and pro t shares. Dots in the top panel are actual quarterly observations in the
(e, π)-plane.
The
bottom panel is a slightly smoothed version using the short-run HP trend described above. Each rst-quarter observation is dated with its year. Line segments between the quarterly observations merely help to illustrate the time orientation. FIGURE 7 - US actual and smoothed
(e, π)-plane
cycles
The clockwise loops follow National Bureau of Economic Research (NBER) business
14 The cycles are most easily distinguished in the bottom panel containing smoothed
cycles.
data, and this panel also shows the long-run variation discussed in the previous subsection. Until the early 1970s the center of the loops shifted vertically as the pro t share fell while
14
See http://www.nber.org/cycles.html.
7
employment remained above 93 percent. The leftward shift towards higher unemployment over the 1970s and 1980s occurred while the pro t share remained below 30 percent. The 1980s began a shift in
e
and
π
to the northeast, towards greater employment rates and
higher pro t shares. The salient clockwise cycles in the smoothed data are not an artifact of the ltering process. In the actual quarterly data, the pro t share moves procyclically: the correlation between it and the employment rate is 0.51, but the correlation between the lags of the pro t share and the current employment rate is stronger (for instance, a two year lag of has a 0.66 correlation with
e).
produce clockwise cycles in the
π
Oscillations in one of these variables therefore necessarily
(e, π)-plane.
Figure 8 - Deviation of US employment and pro t shares from their long term trends in the
(e, π)-plane.
To get a clearer picture of the cyclical element, we follow the approach of Mohun and Veneziani (2008, 2009) and examine the deviations of the actual quarterly observations of
(e, π)
from their long-run trend.
This approach may be appropriate in particular if
the long-term variation in employment, pro t shares, and utilization re ect structural or institutional shifts in the economy (e.g., changes in the bargaining power of workers through a rise or fall in unionization) that are unrelated to the short cycles. Figure 8 displays the deviations. The separate panels correspond to NBER-dated peakto-peak business cycles, but we include observations from one year after the second peak to account for
(e, π)
cycles that `complete' after the NBER cycle has ended for example,
the cycle in the 1960s. All of the deviation-based cycles are qualitatively well-structured and clockwise-oriented, except for the short cycle starting around the beginning of the 1980q1 recession.
4.2 Employment - utilization and utilization - pro tability cycles Similar clockwise cycles exist in the 10.
(e, u)- and (u, π)-planes,
as displayed in Figures 9 and
The amplitudes of the uctuations (like the period length) dier across cycles, and
there are signi cant dierences in the average amplitude of the variables. The employment rate and pro t shares typically vary by less than 6 percentage points over a cycle, whereas utilization varies by up to 15-20 percentage points. Figure 9 - Deviation of US employment and utilization from their long term trends in the
(e, u)-plane.
Figure 10 - Deviation of US utilization and pro t shares from their long term trends in the
(u, π)-plane.
8
4.3 Other patterns Yˆ .
Visibly regular cycles are also present when considering output growth
To calculate
output growth, we use the quarterly growth rate of real net (of depreciation) value added for the corporate business sector. Since BEA does not publish real output or a de ator for the overall corporate sector, to adjust for in ation, we use the de ator implied by the BEA series on real and nominal output of the non nancial business sector.
15 This measure
of quarterly real output growth averaged 0.9 percent over 1948-2008, varying widely until the 1960s, after which it mostly remained within the range of -2 to 4 percent. Clockwise rotations appear in
(e, Yˆ )-, (π, Yˆ )-, and (u, Yˆ )-planes for the slightly- smoothed, short-term
trends for these variables (see Figure 11). Cycles, nally, seem to exist with quarterly Fed accumulation rates
ˆ K
plotted against employment, pro t shares, and utilization, but their
orientations are not as consistent (Figure 12). Figure 11 - Smoothed cycles in the
(e, Yˆ )-, (π, Yˆ )-,
Figure 12 - Smoothed cycles in the
ˆ -, (π, K) ˆ -, (e, K)
and
and
(u, Yˆ )-planes. ˆ -planes. (u, K)
4.4 Robustness The cycles are robust to modi cations of the underlying data. For example, cycles exist regardless of choosing either the Fed's manufacturing or the total industry utilization series. Data variations like these result in only trivial dierences because of the high correlation between the alternative and original data series. Using either the BEA-based and or Fedbased annual accumulation series produces similarly- oriented cycles.
The existence of
cycles is not altered by de ning the numerator of the pro t share the surplus to be inclusive or net of taxes on production or corporate pro ts, as well as including or omitting these taxes in the denominator of the pro t share. As noted above, this is likely due to the high correlation between the pro t share series. We also examined the implications of changing the method for constructing utilization data by looking at output deviations from a smoothed output series. Taking the ratio of corporate value added to its Hodrick-Prescott long-run trend yields a utilization series that is qualitatively dierent than the Fed series other, the output-based
u0
u.
u0
While the two series still track each
contains dierent variation: the correlation coe cient for
u and
u0 is 0.63, as opposed to 0.99 between the Fed's two series. While using the output based0 series u may yield dierent results for a more detailed econometric study, similar cycles emerge in
(u, π)- and (u0 , π)-planes.
Figure 13 illustrates these cycles using the short-term,
slightly smoothed series. FIGURE 13 - Smoothed cycles in Fed-based
15
See NIPA Table 1.14, lines 19 and 42. 9
(u, π)-
and output-based
(u0 , π)-planes.
4.5 Other countries The existence of cycles in pro t, employment, and utilization spaces are not peculiar to the US economy. As Harvie (2000), Mohun and Veneziani (2009), and others have observed, European economies can exhibit pro t-share/employment rate cycles.
Using quarterly
OECD data, we con rm the existence of cycles using utilization rates as well. With one exception, we limited our sample to those countries in the OECD.Stat Database
16 with
at least two decades of consecutive quarterly surplus, employment, and utilization data: the US, Belgium, Canada, Great Britain, France, Italy, the Netherlands, and Spain. The exception is Japan which was included because of its size and importance, and for which
17 Data
the OECD has employment and utilization series but no quarterly surplus data.
availability causes the sample period to vary by country, but in general the sample covers the 1980s into the 2000s. FIGURE 14 - Quarterly
e, π , u
for OECD countries
Figure 14 shows actual quarterly employment, pro t share, and utilization series for the
18 Utilization rates largely uctuate around 80%, with Japan's utiliza-
sample countries.
tion being much higher in the late 1960s and early 1970s. The highest country average is France's at 84.4%, and the lowest average is Italy's at 75.9%. Pro t shares oscillate between one-quarter and one-third for most countries, except for Italy and Spain, where shares average about 42.7% and 44.9%, respectively. There is considerable variation among countries' employment rates. Japan's employment rate was the highest in most of the period but fell below the US and the Netherlands in the mid-to-late 1990s. Spain's employment rate is usually the lowest and the most volatile, at extremes uctuating more than 4 percentage points from its long-term trend. Cycles for these countries are generally present but not as cleanly observed as in the US data.
This is not surprising.
Smaller countries are much more aected by foreign
trade, and most of the countries have public sectors that are substantially larger than in the US. Hence, cyclical patterns that arise from the private-sector interaction between accumulation, output and pricing decisions will not have the same regularity in these countries. Figures 15-20 depict the patterns for two countries, France and Spain. contains deviations of
(u, π)
from its long-term trend in France.
16
Figure 15
For the sample time
http://stats.oecd.org For all countries, utilization rates are for the manufacturing sector, except in the cases of Canada and Japan, which report total industrial utilization. All rates are capacity utilization rates (measured 0% to 100%), except in the case of Japan, which reports a utilization index with a base year of 2005. To convert Japan's index to a rate comparable to other countries' rates, we arbitrarily assume that the 2005 capacity utilization rate value for Japan is 80.0%. 18 There are cross-country correlations between the business cycles and utilization rates sometime track each other across countries and sometimes not for example, Great Britain's series correlates well with the US but not with France. 17
10
period, France's clockwise cycles in
(u, π)
correspond well to the country's rst three
19 Arguably there is no well-de ned cycle present for the 1998q2-2000q4
business cycles.
peak-to-peak, and a relatively rough clockwise cycle appears for the 2000q4-2008q1 cycle. Similarly, gures 16 and 17 show for France that clockwise cycles usually, but not always, exist in
(e, π)-
and
(e, u)-planes.
FIGURE 15 - Deviation of utilization and pro t shares in France from their long term trends in the
(u, π)-plane
FIGURE 16 - Deviation of employment and pro t shares in France from their long term trends in the
(e, π)-plane
Figure 17 - Deviation of employment and utilization in France from their long term trends in the
(e, u)-plane
Spain exhibits clear clockwise
(u, π)
cycles in the 1983q4-1989q2 and 1991q4-1995q1
peak-to-peak business cycles. Less clear are whether there exist
(u, π)
cycles during the
1979q3-1983q4 and 1991q4-1995q1 business cycles. The 1995q1-1998q2 period exhibits a clear counter-clockwise cycle.
Within the 1998q2-2008q1 cycle, there appear to be two
cycles, the rst counter-clockwise and the second clockwise. Figures 19 and 20 show mixed results for clockwise cycles in the
(e, π)-
and
(e, u)-planes.
FIGURE 18 - Deviation of utilization and pro t shares in Spain from their long term trends in the
(u, π)-plane
FIGURE 19 - Deviation of employment and pro t shares in Spain from their long term trends in the
(e, π)-plane
Figure 20 - Deviation of employment and utilization in Spain from their long term trends in the
(e, u)-plane
Pro t shares are unavailable for Japan, but cycles can be observed in employment and utilization. Figure 21 displays deviations from long-term trends in the
(e, u)-plane.
Japan's
cycles are mostly well-de ned and clockwise oriented over most of the eight business cycles during the 1970-2008 sample period. The only exception is the last 2004q2-2008q2 peakto-peak business cycle, where no
(e, u)-cycle
seems apparent. Interestingly, while Japan
19 Business cycle turning points for OECD countries can be http://www.oecd.org/document/29/0,3343,en_2649_34349_35725597_1_1_1_1,00.html
11
found
at
is widely recognized to have suered a liquidity-trap or at least a slump since the 1990s, su cient volatility in the economy exists to generate
(e, u)
20
cycles.
Figure 21 - Deviation of employment and utilization in Japan from their long term trends in the
5
(e, u)-plane
Conclusion
The US economy since the second world war may provide the best arena for an evaluation of dierent approaches within heterodox macro.
The US is as close as one gets to a
closed economy, the size of the public sector is relatively modest, and unlike Japan and many European economies, the US did not have large amounts of hidden unemployment in backward sectors for a good part of the post-war period. With respect to data, moreover, quarterly series are available for some of the key variables in heterodox models. This paper describes stylized patterns in the data on employment, pro tability, capacity utilization and accumulation. The patterns are quite clear and consistent in the case of the US economy, and many of them exist for other OECD countries too, but are generally not as clean and consistent. For the US we nd that
•
the US employment rate, the pro t share and the utilization rate uctuate around a mean of about 0.94, 0.81 and 0.28, respectively
•
the long-term trends of the variables as measured by Hodrick-Prescott lter exhibit modest variation.
The percentage point dierence between maximum and
minimum values of the HP trend is 2.7, 4.4.
and 5.7 for employment, the pro t
share and the utilization rate, respectively. In proportional terms, the variation is largest for the pro t share (15.6 percent) followed by utilization (7.0 percent ) and employment (2.8 percent).
•
short-term uctuations are signi cant for all the variables but the amplitudes dier: typically less the 6 percentage points over a cycle for employment and the pro t share and up to 15-20 percentage points for utilization. In proportional terms, the amplitude is similar for utilization and the pro t share but much smaller for employment.
•
there is strong evidence of clockwise short-term cycles in three bivariate spaces:
(e, u), •
and
(e, π),
(u, π) (e, Yˆ ), (π, Yˆ ) and (u, Yˆ ) too, ˆ ˆ and (u, K) ˆ spaces are less (e, K), (π, K)
clockwise short-run cycles exist for
while the orienta-
tions of the cycles in the
consistent.
20
For dierent views on the reasons behind Japan's liquidity trap, see Krugman (1998) and Nakatani and Skott (2007). 12
•
the short-term cycles are synchronized with the standard NBER dating of business cycles
•
the cyclical patterns appear to be quite robust to changes in the precise de nition and measurement of the variables.
The empirical analysis needs to go much further than a simple analysis of stylized patterns. But the presence of strong and consistent patterns can provide an input into the development and evaluation of economic theories. It is beyond the scope of this paper to discuss these issues. In a companion paper, Skott and Zipperer (2010a), we take a step in this direction by looking at the consistency of three post Keynesian benchmark models with the empirical evidence.
References [1] Barbosa-Filho, N.H., Taylor,L. (2006): Distributive and Demand Cycles in the US Economy - A Structuralist Goodwin Model, in
Metroeconomica
[2] Corrado, C., Mattey, J. (1997): Capacity utilization, in
tives, 11(1), 151-167.
Journal of Economic Perspec-
[3] Goodwin, R.M. (1967): A growth cycle, in C.H. Feinstein (ed.)
and growth, Cambridge:
57(3), 389-411.
Socialism, capitalism
Cambridge University Press.
[4] Harvie, D. (2000): Testing Goodwin: growth cycles in ten OECD countries, in
bridge Journal of Economics, 24(3), 349-376.
Cam-
[5] Hodrick, R.J., Prescott, E.C. (1997): Postwar U.S. Business Cycles: An Empirical Investigation, in
Journal of Money, Credit and Banking, 29(1), 1-16.
[6] Krugman, P. (1998): It's Baaack: Japan's Slump and the Return of the Liquidity Trap, in
Brooking's Papers on Economic Activity, 29(2), 137-205.
[7] Krueger, A.B. (1999): Measuring labor's share, in
and Proceedings, 89(2), 45-51.
American Economic Review, Papers
[8] Mohun, S., Veneziani, R. (2008): Goodwin cycles and the US economy, 1948- 2004, in
Mathematical Economics and the Dynamics of Capitalism - Goodwin's Legacy Continued, Routledge.
P. Flaschel and M. Landesmann (eds.),
[9] Mohun, S., Veneziani, R. (2009): Social democracy and class compromise, mimeo. [10] Nakatani, T., Skott, P. (2007): Japanese growth and stagnation, in
and Economic Dynamics, 18(3), 306-332.
13
Structural Change
[11] Ravn, M. O. and Uhlig, H. (2002): On adjusting the Hodrick-Prescott lter for the frequency of observations, in
Review of Economics and Statistics, 84(2), 371-375.
[12] Schmitt, J., Baker, D. (2006): Missing inaction: evidence of undercounting of nonworkers in the Current Population Survey,
Brie ng Paper.
[13] Shapiro, M.D. (1989): Utilization, in
Center for Economic and Policy Research
Assessing the Federal Reserve's Measures of Capacity and
Brookings Papers on Economic Activity, 1:1989, 181-225.
[14] Skott, P. (1989):
Con ict and Eective Demand in Economic Growth.
Cambridge:
Cambridge University Press. [15] Skott (2010) "Theoretical and empirical shortcomings of the Kaleckian investment function" [16] Skott, P., Zipperer, B. (2010): Dynamic patterns of accumulation and income distribution, Political Economy Quarterly (Kikan Keizai Riron), 46 (4), pp.34-53 (in Japanese) [17] Skott, P., Zipperer, B. (2010a) "An empirical evaluation of three post Keynesian models". Working paper 2010-xx, University of Massachusetts Amherst.
14
Figures for Cyclical Patterns paper February 27, 2010
1
.1
.15
.2
.25
.3
.35
FIGURE 1 - US quarterly pro t shares π and π 0 ; actual, and short-term and long-term trends
1950q1
1960q1
1970q1
1980q1 pi st_pi lt_pi
2
1990q1 pi’ st_pi’ lt_pi’
2000q1
2010q1
.7
.75
.8
.85
.9
FIGURE 2 - US quarterly manufacturing utilization rate u; actual, and short-term and long-term trends
1950q1
1960q1
1970q1
1980q1 u lt_u
3
1990q1 st_u
2000q1
2010q1
0
.005
.01
.015
.02
ˆ ; actual, and short-term and long-term trends FIGURE 3 - Quarterly Fed-based K
1950q1
1960q1
1970q1
1980q1
fed_khat lt_fed_khat
4
1990q1 st_fed_khat
2000q1
2010q1
−.02
0
.02
.04
.06
.08
ˆ ; actual and long-term trend FIGURE 4 - Annual Fed-based and BEA-based K
1940
1960
1980
fed_khat lt_fed_khat_descrip
5
2000 bea_khat lt_bea_khat_descrip
2020
.9
.92
.94
.96
.98
FIGURE 5 - US quarterly employment rate e; actual, and short-term and long-term trends
1950q1
1960q1
1970q1
1980q1 e lt_e
6
1990q1 st_e
2000q1
2010q1
−.04
−.02
0
.02
.04
FIGURE 6 - Dierence between the US mean and long-run trend, for utilization, employment, pro t shares, and quarterly accumulation rate
1950q1
1960q1
1970q1
1980q1
u_deviation e_deviation
1990q1 pi_deviation khat_deviation
7
2000q1
2010q1
.35
FIGURE 7 - US actual and smoothed (e, π)-plane cycles
.33
1951 19651949 1955
1950
.31
1964
1966 1948
1952
.27
.29
2006
1983
1953
.23
.25
1982
1963 1962 1956 1967 1959 1960 2007 19571968 2005 1969 1997 1996 2008 1954 1984 1976 1985 1961 1972 1995 1973 1998 1971 2004 1999 1989 1988 1986 19941958 1991 1977 1990 1992 1978 2003 1970 2000 1987 1979 1993 1975 1974 1981 2002 2001 1980
.91
.93
.95
.97
.35
.89
.33
1951
.29
.31
1950
1984 1976
1985
.27
19771986 1992 1993 1975
1983
.25
1982
19491965 1966 1948 1952 1964 1955 1956 2006 1963 1967 2007 1953 1968 1962 1959 1960 1957 2005 1997 1969 2008 1961 1954 1996 1998 19581995 1972 1973 1999 1978 1971 1994 2004 1988 1989 1970 1990 1991 1987 2000 1979 2003 1974
1981
2002
2001
.23
1980
.89
.91
.93
8
.95
.97
.04 .02
.02
.04
Figure 8 - Deviation of US employment and pro t shares from their long term trends in the (e, π)-plane.
1951
1955
−.02
1952 1953
0
0
1949
1956 1957
−.02
1950
1954
1954
−.04
−.04
1958
0
.02
.04
−.04
−.02
0
.02
.04
1965
.02
.02
.04
−.02
.04
−.04
1964 1968 1967 1969
0
0
1963 1962
−.02
19591960
−.02
1966
1961
1961 1970
−.04
−.04
1958
0
.02
.04
−.04
−.02
0
.02
.04
.02
.02
.04
−.02
.04
−.04
1976
0
−.04
1974 1970
−.04
−.02
0
.02
1977
.04
9
1978 1979
1975
−.02
−.02
1972 1971
1974 1980
−.04
0
1973
−.04
−.02
0
.02
.04
1989 1988
1986 1991
.04 .02 0
.04 .02 0
1984 1985
1987
1983
1990
1981 1982
−.02
−.02
1982
−.04
−.04
1980
0
.02
.04
−.04
−.02
0
.02
.04
2006
.02
.02
.04
−.02
.04
−.04
1997 1996
2005 2007
1993
−.02
0
2000
2004 2008 2003
−.02
0
1995 1998 1994 1999 1991 1992
2002
−.04
2001
−.04
2001
−.04
−.02
0
.02
.04
−.04
10
−.02
0
.02
.04
.1
.1
Figure 9 - Deviation of US employment and utilization from their long term trends in the (e, u)-plane.
.05
1953
.05
1951
1956 1957
1952
0
0
1955
1954
−.05
−.05
1954 1949 1950
−.1
−.1
1958
0
.05
.1
−.1
−.05
0
.05
.1
.05
.1
.05
.1
.1
−.05
.1
−.1
.05
.05
1966 1965 1967 1969 1968 1964
0
0
1960 1959
−.05
−.05
1963 1970 1962
0
.05
.1
1961
−.1
−.05
0
.05
.05
.1
−.05
.1
−.1
−.1
−.1
1958 1961
1973 1974
1979 1974
0
0
1980 1978 1977
−.05
−.05
1972 1970
1971
1976
−.1
−.05
0
.05
.1
11
−.1
−.1
1975
−.1
−.05
0
.1
.1
.05
.05
1989 1988 1990
0
0
1980
1987 1984 1985 1986 1991
−.05
−.05
1981
1982
1982
−.1
−.1
1983
0
.05
.1
−.1
−.05
0
.05
.1
.05
.1
.05
.05
.1
−.05
.1
−.1
2006 2007 2008 2005
0
0
1995 1998 1997 2000 1996 19941999 1993 1992 2001 1991
−.05
2002
−.1
−.1
−.05
20042001 2003
−.1
−.05
0
.05
.1
−.1
12
−.05
0
.1 .05
.05
.1
Figure 10 - Deviation of US utilization and pro t shares from their long term trends in the (u, π)-plane.
1951 1952
1955
0
0
19501949
1953
1956 1957 1954
−.05
1958
−.1
−.1
−.05
1954
.1
−.1
.1
.05
.05
0
.1
−.05
.05
−.1
−.05
0
.05
1965 1966
1960
1961
1961
1968 1967 1969
1970
−.1
−.1
−.05
1958
−.05
1963 1962
0
0
1964 1959
.1
0
1971
1972
.05
.1
−.1
−.05
0
.05
.1
.05
1977 1978 1975
1979
1980 1974
−.1
−.05
0
.05
.1
13
−.1
−.1
−.05
1974
−.05
1970
1976
1973
0
0
.05
.1
−.05
.1
−.1
−.1
−.05
0
.05
.1
.1 .05 0
.1 .05 0
1983
1981
1982
−.1
−.1
−.05
1980
−.05
1982
1984 1985 1986 19881989 1991 1987 1990
0
.05
.1
−.1
−.05
0
.05
.1
.05
.05
.1
−.05
.1
−.1
0
0
1997 1996 1995 1998 1994 1999 1991 1992 1993 2000
2006 2005 2007 2008
2004 2003 2002
−.05
2001
−.1
−.1
−.05
2001
−.1
−.05
0
.05
.1
−.1
14
−.05
0
.05
.1
.04
Figure 11 - Smoothed cycles in the (e, Yˆ )-, (π, Yˆ )-, and (u, Yˆ )-planes.
1950 1955
1983
1977
1976 1993 1985 1992 1981 1986 1975
0
.02
1959 1984
1982
1965 1972 1962 1964 1978 1968 1997 1963 1987 199819991966 1994 1961 1951 1996 1973 2004 1971 1952 1988 1995 2005 2006 2003 1967 1969 1953 2000 1956 1954 1960 2002 1989 1979 1990 1991 1958 1980 2007 1949 1970 2008 1974
−.02
1957 2001
.92
.94
.96
.98
.04
.9
1950
1951
−.02
0
.02
1955 1959 1984 1965 1977 1972 1962 1983 1964 1978 1968 1963 1976 1997 1987 1998 1994 1966 1999 19612005 1996 2004 1971 1952 1988 1973 1985 1993 1995 1992 2003 19672006 2000 1956 1981 1953 1969 1986 1954 1960 1975 2002 1989 1958 19791991 1990 1980 2007 1949 1970 1982 2008 1957 2001 1974
.26
.28
.04
.24
.3
.32
.34
1950
.02
1959
1955
1984
1983
0
2003 2002
1965 1968 1951 1973 1956 1969
1966 1967 1953
1974
−.02
1982
1972 1977 1962 1964 1976 1997 1978 1963 1987 1998 199919941996 1961 2004 2005 1971 1952 1988 1993 1995 1985 1992 2006 2000 1981 1986 1954 1960 1975 1989 1958 1979 1991 1980 1990 2007 1949 1970 2008 1957 2001
.7
.75
.8
15
.85
.9
.005 .01 .015 .02
ˆ -, (π, K) ˆ -, and (u, K) ˆ -planes. Figure 12 - Smoothed cycles in the (e, K) 1998 1997 1996
1982
1976 1984
1981 1985 1977 1975 1992 1993 1986
1965
1967
1968 2000 1949 1956 1969 1995 1952 19571951 1953 1960 1950 1964 1955 1970 2001 1973 1961 1978 1954 1974 1963 1958 1959 1979 1971 1962 1972 1994 1980 1990 1987 1991 2007 1989 2008 1988 2006 2002 2005
0
1983
1966 1999
2004 2003
.005 .01 .015 .02
.9
.92
.94
.96
.98
1998 1997 1999
1996
0
1995 1970 2001 1973 1978 1974 1958 1979 1971 1972 1981 1985 1977 1994 1980 1990 1975 1982 1976 1987 1992 1993 1991 19861989 1984 1988 2002 1983
1949 1956 1969 1952 1957 1960 1953 1955 1964 1961 1954 19591962 1963
1950
1951
2007
2008
2006
2005
2004
2003
.005 .01 .015 .02
1965
1968
2000
.24
1966 1967
.26
.28
.3
.32
.34
1998 1997 1999
1995 1952 1957 1960 1970 1955 1964 1950 2001 1961 19541978 1963 1958 1979 1971 19851959 1981 1962 1972 1977 19801994 1990 19761975 1987 1992 1993 2007 1986 1991 1989 2008 1984 1988 2006 2005
1982
0
1983 2002
.7
1996
2000
1949
2003
1966 1967 1965 1968 1956 1969 1951 1973 1974
1953
2004
.75
.8
16
.85
.9
.34
FIGURE 13 - Smoothed cycles in Fed-based (u, π)- and output-based (u0 , π)-planes.
1951
.32
1950 1949
.3
2006 2007 1962 1959 1960
2005 2008 1961
.28
1958 1984 1976 2004
.26
1968 1997 1954 1996 1998 1995
1985
1957
1972 19991994 1989 1978 1977 19701988 1990 1991 1987 1992 1993 2000 1979
2003
1966
1956
1963
1967 1953
1969
1973
1971 1986
1975 1983
1965
1948 1952 1964 1955
1974
1982 1981
2002
2001
.24
1980
.75
.8
.85
.9
.34
.7
1951
.32
1950 1965 1949 1964
1966 1952
1955
1963 2007
.3
1962 1959
1960
1995
.28
1958
1976 1994
.26
19921975 1993 1983
2006 1956 1967 1968 1953
1957 2005 2008 1997 1954 1996
1961
1984 2004 1971 1977
1991
1948
1969
1998 1985 1972 1970 1986 1990 1987
2003
19991973 1989 1978 1988
1979
1974
2000
1982 1981 2002
2001
.24
1980
.9
.95
17
1
1.05
1.1
.8
.85
.9
.95
1
FIGURE 14 - Quarterly e, π , u for OECD countries
1970q1
1980q1
1990q1
2000q1
2010q1
1970q1
1980q1
1990q1
2000q1
2010q1
1970q1
1980q1
1990q1
2000q1
2010q1
.2
.3
.4
.5
1960q1
.6
.7
.8
.9
1
1960q1
1960q1
18
.03
.04
FIGURE 15 - Deviation of utilization and pro t shares in France from their long term trends in the (u, π)-plane
1989
1990
.02
.02
1990
.01
1988
1987
1992 1991
0
1995
0
1986
−.02
1984
−.01
1994 1985
.02
.04
−.06 −.04 −.02
.02
.04
1998
2001
.005
1995
2000
0
.005
0
.01
0
.01
−.02
1993
0
1999
−.005
−.01 −.005
1996
1997
−.02
−.01
0
.01
.02
.01
−.03
.005
2001 20072008
−.005
0
2006
2003 2002
−.01
2004 2005
−.04
−.02
0
.02
.04
19
1999
−.01
0
.01
.02
.03
.04
.03
.04
FIGURE 16 - Deviation of employment and pro t shares in France from their long term trends in the (e, π)-plane
1989
1990
.02
.02
1990
.01
1988
1987
1992 1991
0
1995
1984 1994
1993
−.01
−.02
0
1986
1985
.005
.01
.015
−.02
−.01
0
.01
.02
.01
0
.01
−.01 −.005
2001
1995
.005
.005
1998
0
2000
0
1999
1997
.01
−.014 −.012 −.01 −.008 −.006 −.004
2008
2007
2006
−.005
0
.005
2001
2003 2002
2004
−.01
2005
−.005
0
.005
.01
.01520
−.005
−.01 −.005
1996
1999
−.01 −.005
0
.005
.01
.015
Figure 17 - Deviation of employment and utilization in France from their long term trends in the (e, u)-plane
1986
−.02
1987 1984 1985
0
.005
.01
.04
.015
1993 1994
−.02
−.01
0
.01
.04
.02
−.01 −.005
1992
1995
−.06 −.04 −.02
0
1988
.02
2001
.02
.03
.01
1999
0
1995
2000 1999
0
.01
1996
−.01
−.03 −.02 −.01
1991
0
.02
1989
1998
1997
.04
−.014 −.012 −.01 −.008 −.006 −.004
.02
2001
2008 2007
2002
0
2003 2005
2004
2006
−.04
−.02
1990
.02
.04
1990
−.005
0
.005
.01
.01521
−.01 −.005
0
.005
.01
.015
.03
.02
FIGURE 18 - Deviation of utilization and pro t shares in Spain from their long term trends in the (u, π)-plane
1986
1984
0
1987
.01
.01
.02
1988 1989
1985 1984
0
1980
1990
−.02 −.01
−.01
1983
−.02
1982
1981
.01
.02
.03
−.02
0
.02
.04
.01
0
.02
−.01
−.01
0
0
.01
1995
1991
1992
1994
1992
−.02
−.03 −.02 −.01
1990
−.03
1993
−.02
0
.02
.04
−.06
−.04
−.02
0
.02
.005
.01
−.04
.005
19971996
2004 2002
1995
2003
0
1998
1999
−.02
−.01
0
.01
.02
.03 22
−.005
−.005
0
2001
2000 1999
−.02
0
.02
.04
.03
.02
FIGURE 19 - Deviation of employment and pro t shares in Spain from their long term trends in the (e, π)-plane
0
1984
1988 1987
.01
.01
.02
1986
1989
1985
1980
0
1984
−.02 −.01
−.01 −.02
1982
1981
.02
.04
−.04
−.02
0
.02
.04
.01
0
.02
−.02
−.01
−.03 −.02 −.01
1990
1991
1992
1994
1992
−.02
0
0
.01
1995
−.03
1993
.015
.02
.025
.03
.035
−.04
−.02
0
.02
.005
.01
.01
1996
.005
1990
1983
1997
2004 2002
1995 2003
0
1998
1999
−.03
−.02
−.01
0
.01 23
−.005
−.005
0
2001
1999
−.01
0
.01
2000
.02
.03
.04
.03
Figure 20 - Deviation of employment and utilization in Spain from their long term trends in the (e, u)-plane
1990
.01
.02
.02
1989
1987 1988
0
1982
1981
0
1983
1980
1985
1984
1984
−.02
−.01
1986
.02
.04
−.04
−.02
0
.02
.04
0
1990
1992
1992 1994
1993
−.04
−.06
−.02
1995
−.04
0
1991
−.02
.02
.02
0
.04
−.02
.02
.025
.03
.035
−.04
−.02
0
.02
.04
.015
.03
.01
.02
1999
.01
.02
1999
2001
0
0
1998
2003
−.02
1995
−.02 −.01
2000
1996 1997
−.03
−.02
−.01
0
.01 24
2004 2002
−.01
0
.01
.02
.03
.05
Figure 21 - Deviation of employment and utilization in Japan from their long term trends in the (e, u)-plane .06
1973
0
.02
−.05
.04
1974
1974
1976
−.1
1971
0
1977
−.15
−.02
1972
.002
.003
.004
.005
1980
−.004
−.002
0
.002
.05
.001
.05
0
1975
.004
1980
1985 1984
0
0
1979
−.05
1981 1978
1982
1977
−.1
−.05
1983
0
.001 .002
.1
−.003 −.002 −.001
1991 1990
.05
1989
1988 1985
0
1986
−.05
1987
−.005
0
.005
.01 25
−.004
−.002
0
.002
.05
.1 .05
1991
0
1997 1992
2000
0
1998 2001
−.05
1995
1998 1996
−.05
1993
1999
−.1
−.1
1994
.002
.004
.006
.008
−.01
−.005
0
.05
.05
0
.005
2008 2007 2006
2005 2005
0
0
2004
2001
−.05
−.05
2003
−.1
−.1
2002
−.015
−.01
−.005
0
−.005
26
0
.005
.01
ScholarWorks@UMass Amherst Economics Department Working Paper Series
Economics
2010
Cyclical Patterns of Employment, Utilization and Profitability Ben Zipperer University of Massachusetts - Amherst, [email protected]
Peter Skott University of Massachusetts - Amherst, [email protected]
Follow this and additional works at: http://scholarworks.umass.edu/econ_workingpaper Part of the Economics Commons Zipperer, Ben and Skott, Peter, "Cyclical Patterns of Employment, Utilization and Profitability" (2010). Economics Department Working Paper Series. Paper 96. http://scholarworks.umass.edu/econ_workingpaper/96
This Article is brought to you for free and open access by the Economics at ScholarWorks@UMass Amherst. It has been accepted for inclusion in Economics Department Working Paper Series by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected].
DEPARTMENT OF ECONOMICS
Working Paper
Cyclical patterns of employment, utilization and profitability by Ben Zipperer and Peter Skott Working Paper 2010-02
UNIVERSITY OF MASSACHUSETTS AMHERST
Cyclical patterns of employment, utilization and pro tability Ben Zipperer
∗
and Peter Skott
†
February 27, 2010
Abstract The interaction between income distribution, accumulation, employment and the utilization of capital is central to macroeconomic models in the `heterodox' tradition. This paper examines the stylized pattern of these variables using US data for the period after 1948. We look at the trends and cycles in individual time series and examine the bivariate cycical patterns among the variables.
JEL classi cation: E12, E32, O41 Key words: growth, business cycles, aggregate demand, instability, income distribution, utilization rate, investment function, pricing. ∗ Department of Economics, University of Massachusetts Amherst, MA 01003, USA; [email protected]. † Department of Economics, University of Massachusetts Amherst, MA 01003, USA; [email protected].
1
Introduction
The `heterodox' tradition in macroeconomics contains a wide range of models. Kaleckian models treat the utilization rate as an accommodating variable, both in the short and the long run. Goodwin's celebrated formalization of Marx, by contrast, take the utilization rate as xed and looks at the interaction between employment and distribution. Distribution is also central to Kaldorian and Robinsonian theories which, like Goodwin, endogenize the pro t share and take the utilization rate as structurally determined in the long run but, like the Kaleckians, view short-run variations in utilization as an intrinsic part of the cycle. The dierences in these and other areas are important, and this diversity of views on core issues is no cause for celebration. Real-world economies dier widely, of course, and appropriate models will re ect these dierences.
One would not want to use the same model to analyze advanced, labor-
constrained economies and dual economies with large (hidden) unemployment.
But an
attempt should be made to resolve fundamental disagreements on the dynamics of accumulation, income distribution and utilization rates in an advanced capitalist economy. Theoretical and empirical work is needed to evaluate the theories and help create a coherent and convincing alternative to the existing orthodoxy. This paper makes no attempt to address the theoretical issues.
Instead, we explore
US data on some of the variables that most, if not all heterodox macroeconomic theories consider important: utilization.
employment rates, pro t shares, accumulation rates, and capacity
1 The paper complements and extends recent studies of US cycles by Barbosa-
Filho and Taylor (2006) and Mohun and Veneziani (2008).
Both of these contributions
use the Goodwin model as their theoretical framework; the former focusing exclusively on oscillations in pro ts and utilization, and the latter on pro ts and employment. The paper is in 5 sections. Section 2 discusses some general data issues. Trends and cycles of the variables are presented in section 3, and section 4 examines bivariate cyclical patterns among the variables. A nal section summarizes the main ndings.
2
Data
Benchmark versions of the heterodox models are typically cast in terms of closed economies without a public sector, and a one-good assumption implies that sectoral dierences are taken to be insigni cant. In short, the models describe a pure capitalist economy. Abstract models of this kind can be very useful.
They serve to highlight particular
issues and can be used to examine the logic of mechanisms that may operate within the more complex reality of real-world economies too. But questions arise if one wants to take the simple models to the data. It is hard to evaluate the models since their predictions can be negated by other in uences explosive cycles for instance may be tamed by the
1
Theoretical issues are discussed in a companion paper, Skott and Zipperer (2010). See also Skott (2010).
1
in uence of public policy and it may not even be obvious which data should be used as the empirical counterparts to the theoretical variables. Our approach in this paper is to look at sectors that, according to the theories themselves, should conform to the behavioral assumptions. In practice, this criterion suggests a focus on pro t shares, accumulation and capacity utilization for the corporate business sector.
Employment, however, is dierent.
The state of the labor market the size of
the reserve army of labor is important because it aects the balance of power between workers and capital, and the relevant measure is the average economy-wide employment rate. Other questions concern the precise de nition of the variables. How for instance should one treat taxes in the calculation of pro t shares? The answers to many questions of this kind depend on the precise purpose of the analysis.
There may be no uniquely correct
answers and even if there is, data limitations restrict what can be done in practice. Given these problems, the best one can do may be to be clear about the de nition of the chosen indicators and to test, when possible, the sensitivity of the observed patterns to changes in the choice of indicators. We are interested in cyclical uctuations as well as long-term trends, and high-frequency data therefore are desirable. US employment and utilization data are available monthly, but corporate-sector pro t data are published on a quarterly basis and unless noted, all data below are quarterly averages. Data are seasonally adjusted by the reporting agency, and we assume that these adjustments adequately correct for seasonal eects. The sample begins in the rst quarter of 1948, the earliest available US employment rates, and ends in the nal quarter of 2008, the last full year of available data at the time of writing.
2
In addition to actual data observations, we smooth the data to construct a short-term and a long-run trend. The smoothing procedure employs the Hodrick-Prescott (1997) lter with smoothing parameters 6.25 and 129,600 for short and long-run trends respectively. While the parameter choice is largely arbitrary any `small' and `large' numbers would be suitable for our descriptive analysis these speci c short- and long-run parameters are recommended by Ravn and Uhlig (2002) for business cycle analysis of annual and monthly data, respectively. Our smaller-than-usual smoothing parameter for the short-run trend is intended to lter out extreme peaks only, leaving considerable quarter-to-quarter variation. Our higher-than-usual smoothing parameter for the long-run trend re ects our eort to describe a long-term variation over decades. Because the actual 1948q1 and 2008q4 observations may bias the constructed endpoints of the long-term trend, we restrict the long-term trend to 1953-2001. That is, after creating the ltered series, we drop the years 1948-1952 and 2002-2008 from the constructed long-term trend, beginning and ending the long-term trend close to the 1953q2 and 2001q1 NBER business cycle peaks.
2
All US data in this paper were collected in September 2009. The US national accounts data used include the ` nal' 2008q4 gures for pro ts and output, as well as the BEA's July 2009 comprehensive revision. OECD data were collected in December 2009.
2
3
Trends and cycles
3.1 Pro t shares To measure the pro t share
π,
we use the surplus and compensation subcategories of
quarterly value added, net of depreciation, in the Bureau of Economic Analysis (BEA) National Income and Product Accounts (NIPA).
3 The largest private sector in the NIPA
tables for which quarterly compensation and operating surplus delineations are available is domestic corporate business. As a share of the total business sector, corporate business net value added rose during 1948-2008, from 57 percent in 1948q1 to 64 percent in 2008q4;
4 Although
as a share of total GDP it remained roughly constant at about 50 percent.
to some extent dictated by data availability, the use of corporate sector data can also, as noted above, be justi ed by a cleaner application of the theories to corporate than to noncorporate business. Net value added (gross value added net of depreciation) is fully decomposed into taxes on production, labor compensation, and net operating surplus.
In the BEA national
accounts, net operating surplus includes taxes on corporate income (which are dierent than production taxes). How one treats taxes can therefore in principle produce dierent proxies for the pro t share. For example, three candidate series are
π= π∗ = π0 =
net surplus + production taxes net surplus + production taxes + compensation net surplus net surplus + compensation net surplus - corporate income taxes net surplus - corporate income taxes + compensation
Over 1948-2008, the means for
π, π∗,
and
π0
are, respectively, 0.29, 0.21, and 0.15. The
decision to omit or include production taxes yields series with virtually identical variation: the correlation coe cient between
π
and
π∗
is
0.98.
In contrast, the treatment of corpo-
rate income taxes yields series with dierent long-term variation (but similar short-term uctuations). Figure 1 shows the actual quarterly evolution of
π
and
π0,
as well as their
short- and long-run Hodrick-Prescott ltered trends. Focusing on the pro t share inclusive of all taxes, pro ts accounted for one-third of output at their peak in the early 1950s. A substantial redistribution towards labor income followed, and the pro t share temporarily fell below one-quarter around the onset of the 1980 recession before swinging upwards again in the 1980s and 1990s.
Using the pro t
0 share π exclusive of all taxes, the long redistribution toward labor after the 1950s fails 0 to occur. The long-term series for π remained around 14% until around the 1980s, after which the share of pro ts began to increase substantially.
3 4
See http://www.bea.gov/national/nipaweb/index.asp See BEA NIPA Tables 1.3.5 and 1.14.
3
These long-term movements
capture most of the dierence between the two series. Although the correlation coe cient between actual quarterly observations for
π
and
π0
is 0.42, deviations of each series from
its long-term trend correlate at 0.92. In simple models pro t maximizing markups over cost are unaected by a change in the tax rate on pro ts, and for many purposes it will be reasonable to include taxes with
5 In section 4 unless otherwise noted we therefore take the pro t share 6 sum of the surplus and taxes, divided by value added.
pro ts.
FIGURE 1 - US quarterly pro t shares
π
and
π0;
π
to be the
actual, and short-term and long-term
trends A potentially large problem with the above de nition of the pro t share concerns its treatment of executive pay, which NIPA tables include as compensation.
Executive
compensation has increased dramatically and arguably a large part of this increase should be included with pro ts. Our measure of the pro t share fails to do this and may give a misleading picture of the trend in pro tability, especially for the period after the 1980s. Krueger (1999) attempts to account for this issue by modifying the NIPA data using a compensation series derived from the Bureau of Labor Statistics Employment Cost Index (ECI) whose data exclude jobs in which employees have a signi cant role in setting their own wages which grew more slowly over 1988-1995 than labor's share of income in the NIPA.
7 . After using the ECI data to modify the NIPA-based pro t share, Krueger nds
that instead of growing only 0.6 percentage points over 1988-1995, the modi ed pro t share grew somewhere between 1.9 and 4.6 percentage points, depending on the nature of the adjustment. A similar adjustment to our data would steepen the most recent long-term rise in the pro t share, although it is unclear,
a priori, what impact this adjustment would
have on the cyclical component of the pro t share around its trend.
3.2 Utilization rates We use the Federal Reserve capacity utilization series for manufacturing.
The Federal
Reserve also publishes a capacity utilization series for the total industrial sector but this series only exists since 1967.
8
The manufacturing series may be more reliable and the
movements in the total industrial series closely match those for the manufacturing series. The average over 1967-2008 of the former is 81.3% compared to 80.1% for the latter, and the correlation coe cient between the two series is 0.99. The Federal Reserve manufacturing series is monthly and seasonally adjusted, and we take quarterly averages for the 19482008 period. Over the period, utilization uctuates strongly with extreme values above 90
5
The invariance of the optimal markup is subject to quali ations. Consumer loyalty or other intertemporal elements in demand can make it optimal for a rm to vary its markup in response to uctuations in the tax rate. 6 See lines 4, 7, and 8 of NIPA Table 1.14. 7 The ECI data begin in 1975: http://www.bls.gov/news.release/eci.tn.htm 8 See Shapiro (1989) for a review of how the Federal Reserve calculates capacity and utilization. 4
percent and below 70 percent (Figure 2). In January 1986 the Federal Reserve switched from SIC to NAICS industry classi cation, possibly introducing a discontinuity that may aect statistical analysis. We ignore this complication. FIGURE 2 - US quarterly manufacturing utilization rate
u; actual, and short-term and
long-term trends
3.3 Accumulation and capital capacity To approximate changes in the capital stock, there are two candidate data series for the
9
US: Bureau of Economic Analysis (BEA) net xed assets , and the Federal Reserve (Fed)
10 Both series are available for the manufacturing sector. industrial capacity index.
The capacity index measures the greatest level of output each plant ... can maintain within the framework of a realistic work schedule. monthly. To calculate quarterly capacity changes
ˆ K
11
The capacity series is published
from the monthly published data, we
calculate the percent dierence between index values three months apart: fourth quarter
ˆ K
is the percent dierence between December and September values; rst quarter
ˆ K
is
calculated from March's index value and the previous year's December value. The average quarterly growth rate of manufacturing capacity was about 0.9 percent between 1948-2008. The growth rate generally fell over the period, with large peaks in the mid-to-late 1960s and mid-to-late 1990s. The same 1986 SIC to NAICS industrial classi cation change mentioned above for the utilization series may also aect the capacity series, a complication we ignore. FIGURE 3 - Quarterly Fed-based
ˆ; K
actual, and short-term and long-term trends
While the Fed series is monthly, the BEA data are annual. This is a serious disadvantage for the analysis of cyclical patterns and there may also be other reasons to prefer the Fed data. Heterodox models usually assume a xed coe cient production function and a constant rate of depreciation. If these assumptions are satis ed, indicators of production capacity (the Fed data) and capital stock (BEA) will coincide, if correctly measured. If the assumptions are relaxed, however, the two indicators can deviate, and the economic argument behind the standard investment functions concerns the desired increase in capacity. For some purposes, at least, the Fed data therefore may be preferable on theoretical grounds (as well as because of their high frequency). The BEA series is based on the end-of-year capital stock, and we calculate its annual
n − 1.
To
determine annual changes using the Fed series, we calculate the rate of change in year
n as
changes in year
n
as the percent dierence between values in years
n
and
9 See the manufacturing industry entry in BEA Table 4.2 of non-residential xed assets: http://www.bea.gov/national/FA2004/ 10 See http://www.federalreserve.gov/releases/G17/caputl.htm. 11 Corrado and Mattey (1997) describe how the Federal Reserve capacity series is constructed. See also Shapiro (1989).
5
the percent dierence between December values in years
n and n−1.
Figure 4 shows actual
annual rates of change for the Fed and BEA series, as well as their smoothed, long-term trends. For the long-term trends of these annual data, we use a smoothing parameter of 1000, which results in long-term annual rates of change similar to the long-term annual
12
rates of change from quarterly data.
Over 1948-2008, the annual Fed-based rate of capacity changes average about 3.5%, whereas the annual BEA accumulation rates average 2.7%. The two annual series generally move together, with a correlation coe cient of 0.74, indicating some support for using xedcoe cient production functions. Signi cant dierences between the series occur in 19491950 and 1958-1960, when the Fed series exceed the BEA series by about 2-3 percentage points, and in 1996-1999, when the gap is about 3-5 percentage points. In 1974, the BEA series is more than 2 percentage points larger than the Fed series. Since 1982, the Fedbased measure has exceeded that for the BEA, a trend consistent with increasing reliance on computer-based technology. For example, the the large relative run-up in the Fed-series in the 1990s may be viewed in terms of rms realizing productivity gains from computer equipment. The BEA series may show small increases in the capital stock because prices for computers, relative to other machinery, fell dramatically over that period. FIGURE 4 - Annual Fed-based and BEA-based
ˆ; K
actual and long-term trend
3.4 Employment rates We measure the employment rate
e
as one minus the seasonally adjusted unemployment
rate from the Bureau of Labor Statistics (BLS) Current Population Survey (CPS).
13 This
economy-wide de nition of the employment rate hides all regional, sectoral and skill-speci c dierences in labor market conditions.
The implied BLS measure of movements in the
total labor force is in line with common simplifying assumptions of a constant growth rate in the labor force, but the de nition avoids complications from historical shifts in the US employment-to-population ratio that may have been endogenous (for example due to female labor market entry). As shown in Figure 5, the employment rate peaked at more than 97 percent in the early 1950s and dropped to less than 90 percent during 1982-1983. Over 1948-2008, however, it rarely escaped the 92-96 percent range. In a comparison of US-Census-based and BLS Current Population Survey-based employment rates, Schmitt and Baker (2006) noted that the BLS may increasingly be understating the unemployment rate around the order of several tenths of a percentage point. We ignore this complication, although accounting for it would slightly lower the most recent end of the long-term employment trend.
12
Using the same long-term smoothing parameter we used for quarterly data on the annual data would result in an approximately linear long-term series. 13 See http://stats.bls.gov/data/.
6
FIGURE 5 - US quarterly employment rate
e;
actual, and short-term and long-term
trends
3.5 Trends The considerable short-term variation in the utilization series may overshadow relatively modest long-term movements. Figure 6 graphs the percentage-point dierence between the mean and long-run Hodrick-Prescott values for the pro t share, utilization rate, employment rate and the annual accumulation rate. The span between maximum and minimum long-run dierences from the mean of the employment rate is modest at 2.7 percentage points.
The span of deviations from the mean is similar for both pro t shares and uti-
lization rates, 4.4 percentage points for the pro t share and 5.7 percentage points for utilization. But the mean values are dierent and the relative long-term variation for the pro t share is much larger than for the utilization rate: the span between the maximum and minimum of the proportional deviation from the mean is 2.8 percentage points for the employment rate, 7.0 percentage points for the utilization rate, and 15.6 percentage points for the pro t share. Annual Fed-based accumulation rates average about 3.6 percent, and the span of absolute dierences from this mean is about 2.0 percentage points (corresponding to a span of proportional dierences of 53.1 percentage points). FIGURE 6 - Dierence between the US mean and long-run trend, for utilization, employment, pro t shares, and quarterly accumulation rate
4
Bivariate cyclical patterns
4.1 Employment - pro tability Figure 7 contains two time-connected scatterplots of US employment rates and pro t shares. Dots in the top panel are actual quarterly observations in the
(e, π)-plane.
The
bottom panel is a slightly smoothed version using the short-run HP trend described above. Each rst-quarter observation is dated with its year. Line segments between the quarterly observations merely help to illustrate the time orientation. FIGURE 7 - US actual and smoothed
(e, π)-plane
cycles
The clockwise loops follow National Bureau of Economic Research (NBER) business
14 The cycles are most easily distinguished in the bottom panel containing smoothed
cycles.
data, and this panel also shows the long-run variation discussed in the previous subsection. Until the early 1970s the center of the loops shifted vertically as the pro t share fell while
14
See http://www.nber.org/cycles.html.
7
employment remained above 93 percent. The leftward shift towards higher unemployment over the 1970s and 1980s occurred while the pro t share remained below 30 percent. The 1980s began a shift in
e
and
π
to the northeast, towards greater employment rates and
higher pro t shares. The salient clockwise cycles in the smoothed data are not an artifact of the ltering process. In the actual quarterly data, the pro t share moves procyclically: the correlation between it and the employment rate is 0.51, but the correlation between the lags of the pro t share and the current employment rate is stronger (for instance, a two year lag of has a 0.66 correlation with
e).
produce clockwise cycles in the
π
Oscillations in one of these variables therefore necessarily
(e, π)-plane.
Figure 8 - Deviation of US employment and pro t shares from their long term trends in the
(e, π)-plane.
To get a clearer picture of the cyclical element, we follow the approach of Mohun and Veneziani (2008, 2009) and examine the deviations of the actual quarterly observations of
(e, π)
from their long-run trend.
This approach may be appropriate in particular if
the long-term variation in employment, pro t shares, and utilization re ect structural or institutional shifts in the economy (e.g., changes in the bargaining power of workers through a rise or fall in unionization) that are unrelated to the short cycles. Figure 8 displays the deviations. The separate panels correspond to NBER-dated peakto-peak business cycles, but we include observations from one year after the second peak to account for
(e, π)
cycles that `complete' after the NBER cycle has ended for example,
the cycle in the 1960s. All of the deviation-based cycles are qualitatively well-structured and clockwise-oriented, except for the short cycle starting around the beginning of the 1980q1 recession.
4.2 Employment - utilization and utilization - pro tability cycles Similar clockwise cycles exist in the 10.
(e, u)- and (u, π)-planes,
as displayed in Figures 9 and
The amplitudes of the uctuations (like the period length) dier across cycles, and
there are signi cant dierences in the average amplitude of the variables. The employment rate and pro t shares typically vary by less than 6 percentage points over a cycle, whereas utilization varies by up to 15-20 percentage points. Figure 9 - Deviation of US employment and utilization from their long term trends in the
(e, u)-plane.
Figure 10 - Deviation of US utilization and pro t shares from their long term trends in the
(u, π)-plane.
8
4.3 Other patterns Yˆ .
Visibly regular cycles are also present when considering output growth
To calculate
output growth, we use the quarterly growth rate of real net (of depreciation) value added for the corporate business sector. Since BEA does not publish real output or a de ator for the overall corporate sector, to adjust for in ation, we use the de ator implied by the BEA series on real and nominal output of the non nancial business sector.
15 This measure
of quarterly real output growth averaged 0.9 percent over 1948-2008, varying widely until the 1960s, after which it mostly remained within the range of -2 to 4 percent. Clockwise rotations appear in
(e, Yˆ )-, (π, Yˆ )-, and (u, Yˆ )-planes for the slightly- smoothed, short-term
trends for these variables (see Figure 11). Cycles, nally, seem to exist with quarterly Fed accumulation rates
ˆ K
plotted against employment, pro t shares, and utilization, but their
orientations are not as consistent (Figure 12). Figure 11 - Smoothed cycles in the
(e, Yˆ )-, (π, Yˆ )-,
Figure 12 - Smoothed cycles in the
ˆ -, (π, K) ˆ -, (e, K)
and
and
(u, Yˆ )-planes. ˆ -planes. (u, K)
4.4 Robustness The cycles are robust to modi cations of the underlying data. For example, cycles exist regardless of choosing either the Fed's manufacturing or the total industry utilization series. Data variations like these result in only trivial dierences because of the high correlation between the alternative and original data series. Using either the BEA-based and or Fedbased annual accumulation series produces similarly- oriented cycles.
The existence of
cycles is not altered by de ning the numerator of the pro t share the surplus to be inclusive or net of taxes on production or corporate pro ts, as well as including or omitting these taxes in the denominator of the pro t share. As noted above, this is likely due to the high correlation between the pro t share series. We also examined the implications of changing the method for constructing utilization data by looking at output deviations from a smoothed output series. Taking the ratio of corporate value added to its Hodrick-Prescott long-run trend yields a utilization series that is qualitatively dierent than the Fed series other, the output-based
u0
u.
u0
While the two series still track each
contains dierent variation: the correlation coe cient for
u and
u0 is 0.63, as opposed to 0.99 between the Fed's two series. While using the output based0 series u may yield dierent results for a more detailed econometric study, similar cycles emerge in
(u, π)- and (u0 , π)-planes.
Figure 13 illustrates these cycles using the short-term,
slightly smoothed series. FIGURE 13 - Smoothed cycles in Fed-based
15
See NIPA Table 1.14, lines 19 and 42. 9
(u, π)-
and output-based
(u0 , π)-planes.
4.5 Other countries The existence of cycles in pro t, employment, and utilization spaces are not peculiar to the US economy. As Harvie (2000), Mohun and Veneziani (2009), and others have observed, European economies can exhibit pro t-share/employment rate cycles.
Using quarterly
OECD data, we con rm the existence of cycles using utilization rates as well. With one exception, we limited our sample to those countries in the OECD.Stat Database
16 with
at least two decades of consecutive quarterly surplus, employment, and utilization data: the US, Belgium, Canada, Great Britain, France, Italy, the Netherlands, and Spain. The exception is Japan which was included because of its size and importance, and for which
17 Data
the OECD has employment and utilization series but no quarterly surplus data.
availability causes the sample period to vary by country, but in general the sample covers the 1980s into the 2000s. FIGURE 14 - Quarterly
e, π , u
for OECD countries
Figure 14 shows actual quarterly employment, pro t share, and utilization series for the
18 Utilization rates largely uctuate around 80%, with Japan's utiliza-
sample countries.
tion being much higher in the late 1960s and early 1970s. The highest country average is France's at 84.4%, and the lowest average is Italy's at 75.9%. Pro t shares oscillate between one-quarter and one-third for most countries, except for Italy and Spain, where shares average about 42.7% and 44.9%, respectively. There is considerable variation among countries' employment rates. Japan's employment rate was the highest in most of the period but fell below the US and the Netherlands in the mid-to-late 1990s. Spain's employment rate is usually the lowest and the most volatile, at extremes uctuating more than 4 percentage points from its long-term trend. Cycles for these countries are generally present but not as cleanly observed as in the US data.
This is not surprising.
Smaller countries are much more aected by foreign
trade, and most of the countries have public sectors that are substantially larger than in the US. Hence, cyclical patterns that arise from the private-sector interaction between accumulation, output and pricing decisions will not have the same regularity in these countries. Figures 15-20 depict the patterns for two countries, France and Spain. contains deviations of
(u, π)
from its long-term trend in France.
16
Figure 15
For the sample time
http://stats.oecd.org For all countries, utilization rates are for the manufacturing sector, except in the cases of Canada and Japan, which report total industrial utilization. All rates are capacity utilization rates (measured 0% to 100%), except in the case of Japan, which reports a utilization index with a base year of 2005. To convert Japan's index to a rate comparable to other countries' rates, we arbitrarily assume that the 2005 capacity utilization rate value for Japan is 80.0%. 18 There are cross-country correlations between the business cycles and utilization rates sometime track each other across countries and sometimes not for example, Great Britain's series correlates well with the US but not with France. 17
10
period, France's clockwise cycles in
(u, π)
correspond well to the country's rst three
19 Arguably there is no well-de ned cycle present for the 1998q2-2000q4
business cycles.
peak-to-peak, and a relatively rough clockwise cycle appears for the 2000q4-2008q1 cycle. Similarly, gures 16 and 17 show for France that clockwise cycles usually, but not always, exist in
(e, π)-
and
(e, u)-planes.
FIGURE 15 - Deviation of utilization and pro t shares in France from their long term trends in the
(u, π)-plane
FIGURE 16 - Deviation of employment and pro t shares in France from their long term trends in the
(e, π)-plane
Figure 17 - Deviation of employment and utilization in France from their long term trends in the
(e, u)-plane
Spain exhibits clear clockwise
(u, π)
cycles in the 1983q4-1989q2 and 1991q4-1995q1
peak-to-peak business cycles. Less clear are whether there exist
(u, π)
cycles during the
1979q3-1983q4 and 1991q4-1995q1 business cycles. The 1995q1-1998q2 period exhibits a clear counter-clockwise cycle.
Within the 1998q2-2008q1 cycle, there appear to be two
cycles, the rst counter-clockwise and the second clockwise. Figures 19 and 20 show mixed results for clockwise cycles in the
(e, π)-
and
(e, u)-planes.
FIGURE 18 - Deviation of utilization and pro t shares in Spain from their long term trends in the
(u, π)-plane
FIGURE 19 - Deviation of employment and pro t shares in Spain from their long term trends in the
(e, π)-plane
Figure 20 - Deviation of employment and utilization in Spain from their long term trends in the
(e, u)-plane
Pro t shares are unavailable for Japan, but cycles can be observed in employment and utilization. Figure 21 displays deviations from long-term trends in the
(e, u)-plane.
Japan's
cycles are mostly well-de ned and clockwise oriented over most of the eight business cycles during the 1970-2008 sample period. The only exception is the last 2004q2-2008q2 peakto-peak business cycle, where no
(e, u)-cycle
seems apparent. Interestingly, while Japan
19 Business cycle turning points for OECD countries can be http://www.oecd.org/document/29/0,3343,en_2649_34349_35725597_1_1_1_1,00.html
11
found
at
is widely recognized to have suered a liquidity-trap or at least a slump since the 1990s, su cient volatility in the economy exists to generate
(e, u)
20
cycles.
Figure 21 - Deviation of employment and utilization in Japan from their long term trends in the
5
(e, u)-plane
Conclusion
The US economy since the second world war may provide the best arena for an evaluation of dierent approaches within heterodox macro.
The US is as close as one gets to a
closed economy, the size of the public sector is relatively modest, and unlike Japan and many European economies, the US did not have large amounts of hidden unemployment in backward sectors for a good part of the post-war period. With respect to data, moreover, quarterly series are available for some of the key variables in heterodox models. This paper describes stylized patterns in the data on employment, pro tability, capacity utilization and accumulation. The patterns are quite clear and consistent in the case of the US economy, and many of them exist for other OECD countries too, but are generally not as clean and consistent. For the US we nd that
•
the US employment rate, the pro t share and the utilization rate uctuate around a mean of about 0.94, 0.81 and 0.28, respectively
•
the long-term trends of the variables as measured by Hodrick-Prescott lter exhibit modest variation.
The percentage point dierence between maximum and
minimum values of the HP trend is 2.7, 4.4.
and 5.7 for employment, the pro t
share and the utilization rate, respectively. In proportional terms, the variation is largest for the pro t share (15.6 percent) followed by utilization (7.0 percent ) and employment (2.8 percent).
•
short-term uctuations are signi cant for all the variables but the amplitudes dier: typically less the 6 percentage points over a cycle for employment and the pro t share and up to 15-20 percentage points for utilization. In proportional terms, the amplitude is similar for utilization and the pro t share but much smaller for employment.
•
there is strong evidence of clockwise short-term cycles in three bivariate spaces:
(e, u), •
and
(e, π),
(u, π) (e, Yˆ ), (π, Yˆ ) and (u, Yˆ ) too, ˆ ˆ and (u, K) ˆ spaces are less (e, K), (π, K)
clockwise short-run cycles exist for
while the orienta-
tions of the cycles in the
consistent.
20
For dierent views on the reasons behind Japan's liquidity trap, see Krugman (1998) and Nakatani and Skott (2007). 12
•
the short-term cycles are synchronized with the standard NBER dating of business cycles
•
the cyclical patterns appear to be quite robust to changes in the precise de nition and measurement of the variables.
The empirical analysis needs to go much further than a simple analysis of stylized patterns. But the presence of strong and consistent patterns can provide an input into the development and evaluation of economic theories. It is beyond the scope of this paper to discuss these issues. In a companion paper, Skott and Zipperer (2010a), we take a step in this direction by looking at the consistency of three post Keynesian benchmark models with the empirical evidence.
References [1] Barbosa-Filho, N.H., Taylor,L. (2006): Distributive and Demand Cycles in the US Economy - A Structuralist Goodwin Model, in
Metroeconomica
[2] Corrado, C., Mattey, J. (1997): Capacity utilization, in
tives, 11(1), 151-167.
Journal of Economic Perspec-
[3] Goodwin, R.M. (1967): A growth cycle, in C.H. Feinstein (ed.)
and growth, Cambridge:
57(3), 389-411.
Socialism, capitalism
Cambridge University Press.
[4] Harvie, D. (2000): Testing Goodwin: growth cycles in ten OECD countries, in
bridge Journal of Economics, 24(3), 349-376.
Cam-
[5] Hodrick, R.J., Prescott, E.C. (1997): Postwar U.S. Business Cycles: An Empirical Investigation, in
Journal of Money, Credit and Banking, 29(1), 1-16.
[6] Krugman, P. (1998): It's Baaack: Japan's Slump and the Return of the Liquidity Trap, in
Brooking's Papers on Economic Activity, 29(2), 137-205.
[7] Krueger, A.B. (1999): Measuring labor's share, in
and Proceedings, 89(2), 45-51.
American Economic Review, Papers
[8] Mohun, S., Veneziani, R. (2008): Goodwin cycles and the US economy, 1948- 2004, in
Mathematical Economics and the Dynamics of Capitalism - Goodwin's Legacy Continued, Routledge.
P. Flaschel and M. Landesmann (eds.),
[9] Mohun, S., Veneziani, R. (2009): Social democracy and class compromise, mimeo. [10] Nakatani, T., Skott, P. (2007): Japanese growth and stagnation, in
and Economic Dynamics, 18(3), 306-332.
13
Structural Change
[11] Ravn, M. O. and Uhlig, H. (2002): On adjusting the Hodrick-Prescott lter for the frequency of observations, in
Review of Economics and Statistics, 84(2), 371-375.
[12] Schmitt, J., Baker, D. (2006): Missing inaction: evidence of undercounting of nonworkers in the Current Population Survey,
Brie ng Paper.
[13] Shapiro, M.D. (1989): Utilization, in
Center for Economic and Policy Research
Assessing the Federal Reserve's Measures of Capacity and
Brookings Papers on Economic Activity, 1:1989, 181-225.
[14] Skott, P. (1989):
Con ict and Eective Demand in Economic Growth.
Cambridge:
Cambridge University Press. [15] Skott (2010) "Theoretical and empirical shortcomings of the Kaleckian investment function" [16] Skott, P., Zipperer, B. (2010): Dynamic patterns of accumulation and income distribution, Political Economy Quarterly (Kikan Keizai Riron), 46 (4), pp.34-53 (in Japanese) [17] Skott, P., Zipperer, B. (2010a) "An empirical evaluation of three post Keynesian models". Working paper 2010-xx, University of Massachusetts Amherst.
14
Figures for Cyclical Patterns paper February 27, 2010
1
.1
.15
.2
.25
.3
.35
FIGURE 1 - US quarterly pro t shares π and π 0 ; actual, and short-term and long-term trends
1950q1
1960q1
1970q1
1980q1 pi st_pi lt_pi
2
1990q1 pi’ st_pi’ lt_pi’
2000q1
2010q1
.7
.75
.8
.85
.9
FIGURE 2 - US quarterly manufacturing utilization rate u; actual, and short-term and long-term trends
1950q1
1960q1
1970q1
1980q1 u lt_u
3
1990q1 st_u
2000q1
2010q1
0
.005
.01
.015
.02
ˆ ; actual, and short-term and long-term trends FIGURE 3 - Quarterly Fed-based K
1950q1
1960q1
1970q1
1980q1
fed_khat lt_fed_khat
4
1990q1 st_fed_khat
2000q1
2010q1
−.02
0
.02
.04
.06
.08
ˆ ; actual and long-term trend FIGURE 4 - Annual Fed-based and BEA-based K
1940
1960
1980
fed_khat lt_fed_khat_descrip
5
2000 bea_khat lt_bea_khat_descrip
2020
.9
.92
.94
.96
.98
FIGURE 5 - US quarterly employment rate e; actual, and short-term and long-term trends
1950q1
1960q1
1970q1
1980q1 e lt_e
6
1990q1 st_e
2000q1
2010q1
−.04
−.02
0
.02
.04
FIGURE 6 - Dierence between the US mean and long-run trend, for utilization, employment, pro t shares, and quarterly accumulation rate
1950q1
1960q1
1970q1
1980q1
u_deviation e_deviation
1990q1 pi_deviation khat_deviation
7
2000q1
2010q1
.35
FIGURE 7 - US actual and smoothed (e, π)-plane cycles
.33
1951 19651949 1955
1950
.31
1964
1966 1948
1952
.27
.29
2006
1983
1953
.23
.25
1982
1963 1962 1956 1967 1959 1960 2007 19571968 2005 1969 1997 1996 2008 1954 1984 1976 1985 1961 1972 1995 1973 1998 1971 2004 1999 1989 1988 1986 19941958 1991 1977 1990 1992 1978 2003 1970 2000 1987 1979 1993 1975 1974 1981 2002 2001 1980
.91
.93
.95
.97
.35
.89
.33
1951
.29
.31
1950
1984 1976
1985
.27
19771986 1992 1993 1975
1983
.25
1982
19491965 1966 1948 1952 1964 1955 1956 2006 1963 1967 2007 1953 1968 1962 1959 1960 1957 2005 1997 1969 2008 1961 1954 1996 1998 19581995 1972 1973 1999 1978 1971 1994 2004 1988 1989 1970 1990 1991 1987 2000 1979 2003 1974
1981
2002
2001
.23
1980
.89
.91
.93
8
.95
.97
.04 .02
.02
.04
Figure 8 - Deviation of US employment and pro t shares from their long term trends in the (e, π)-plane.
1951
1955
−.02
1952 1953
0
0
1949
1956 1957
−.02
1950
1954
1954
−.04
−.04
1958
0
.02
.04
−.04
−.02
0
.02
.04
1965
.02
.02
.04
−.02
.04
−.04
1964 1968 1967 1969
0
0
1963 1962
−.02
19591960
−.02
1966
1961
1961 1970
−.04
−.04
1958
0
.02
.04
−.04
−.02
0
.02
.04
.02
.02
.04
−.02
.04
−.04
1976
0
−.04
1974 1970
−.04
−.02
0
.02
1977
.04
9
1978 1979
1975
−.02
−.02
1972 1971
1974 1980
−.04
0
1973
−.04
−.02
0
.02
.04
1989 1988
1986 1991
.04 .02 0
.04 .02 0
1984 1985
1987
1983
1990
1981 1982
−.02
−.02
1982
−.04
−.04
1980
0
.02
.04
−.04
−.02
0
.02
.04
2006
.02
.02
.04
−.02
.04
−.04
1997 1996
2005 2007
1993
−.02
0
2000
2004 2008 2003
−.02
0
1995 1998 1994 1999 1991 1992
2002
−.04
2001
−.04
2001
−.04
−.02
0
.02
.04
−.04
10
−.02
0
.02
.04
.1
.1
Figure 9 - Deviation of US employment and utilization from their long term trends in the (e, u)-plane.
.05
1953
.05
1951
1956 1957
1952
0
0
1955
1954
−.05
−.05
1954 1949 1950
−.1
−.1
1958
0
.05
.1
−.1
−.05
0
.05
.1
.05
.1
.05
.1
.1
−.05
.1
−.1
.05
.05
1966 1965 1967 1969 1968 1964
0
0
1960 1959
−.05
−.05
1963 1970 1962
0
.05
.1
1961
−.1
−.05
0
.05
.05
.1
−.05
.1
−.1
−.1
−.1
1958 1961
1973 1974
1979 1974
0
0
1980 1978 1977
−.05
−.05
1972 1970
1971
1976
−.1
−.05
0
.05
.1
11
−.1
−.1
1975
−.1
−.05
0
.1
.1
.05
.05
1989 1988 1990
0
0
1980
1987 1984 1985 1986 1991
−.05
−.05
1981
1982
1982
−.1
−.1
1983
0
.05
.1
−.1
−.05
0
.05
.1
.05
.1
.05
.05
.1
−.05
.1
−.1
2006 2007 2008 2005
0
0
1995 1998 1997 2000 1996 19941999 1993 1992 2001 1991
−.05
2002
−.1
−.1
−.05
20042001 2003
−.1
−.05
0
.05
.1
−.1
12
−.05
0
.1 .05
.05
.1
Figure 10 - Deviation of US utilization and pro t shares from their long term trends in the (u, π)-plane.
1951 1952
1955
0
0
19501949
1953
1956 1957 1954
−.05
1958
−.1
−.1
−.05
1954
.1
−.1
.1
.05
.05
0
.1
−.05
.05
−.1
−.05
0
.05
1965 1966
1960
1961
1961
1968 1967 1969
1970
−.1
−.1
−.05
1958
−.05
1963 1962
0
0
1964 1959
.1
0
1971
1972
.05
.1
−.1
−.05
0
.05
.1
.05
1977 1978 1975
1979
1980 1974
−.1
−.05
0
.05
.1
13
−.1
−.1
−.05
1974
−.05
1970
1976
1973
0
0
.05
.1
−.05
.1
−.1
−.1
−.05
0
.05
.1
.1 .05 0
.1 .05 0
1983
1981
1982
−.1
−.1
−.05
1980
−.05
1982
1984 1985 1986 19881989 1991 1987 1990
0
.05
.1
−.1
−.05
0
.05
.1
.05
.05
.1
−.05
.1
−.1
0
0
1997 1996 1995 1998 1994 1999 1991 1992 1993 2000
2006 2005 2007 2008
2004 2003 2002
−.05
2001
−.1
−.1
−.05
2001
−.1
−.05
0
.05
.1
−.1
14
−.05
0
.05
.1
.04
Figure 11 - Smoothed cycles in the (e, Yˆ )-, (π, Yˆ )-, and (u, Yˆ )-planes.
1950 1955
1983
1977
1976 1993 1985 1992 1981 1986 1975
0
.02
1959 1984
1982
1965 1972 1962 1964 1978 1968 1997 1963 1987 199819991966 1994 1961 1951 1996 1973 2004 1971 1952 1988 1995 2005 2006 2003 1967 1969 1953 2000 1956 1954 1960 2002 1989 1979 1990 1991 1958 1980 2007 1949 1970 2008 1974
−.02
1957 2001
.92
.94
.96
.98
.04
.9
1950
1951
−.02
0
.02
1955 1959 1984 1965 1977 1972 1962 1983 1964 1978 1968 1963 1976 1997 1987 1998 1994 1966 1999 19612005 1996 2004 1971 1952 1988 1973 1985 1993 1995 1992 2003 19672006 2000 1956 1981 1953 1969 1986 1954 1960 1975 2002 1989 1958 19791991 1990 1980 2007 1949 1970 1982 2008 1957 2001 1974
.26
.28
.04
.24
.3
.32
.34
1950
.02
1959
1955
1984
1983
0
2003 2002
1965 1968 1951 1973 1956 1969
1966 1967 1953
1974
−.02
1982
1972 1977 1962 1964 1976 1997 1978 1963 1987 1998 199919941996 1961 2004 2005 1971 1952 1988 1993 1995 1985 1992 2006 2000 1981 1986 1954 1960 1975 1989 1958 1979 1991 1980 1990 2007 1949 1970 2008 1957 2001
.7
.75
.8
15
.85
.9
.005 .01 .015 .02
ˆ -, (π, K) ˆ -, and (u, K) ˆ -planes. Figure 12 - Smoothed cycles in the (e, K) 1998 1997 1996
1982
1976 1984
1981 1985 1977 1975 1992 1993 1986
1965
1967
1968 2000 1949 1956 1969 1995 1952 19571951 1953 1960 1950 1964 1955 1970 2001 1973 1961 1978 1954 1974 1963 1958 1959 1979 1971 1962 1972 1994 1980 1990 1987 1991 2007 1989 2008 1988 2006 2002 2005
0
1983
1966 1999
2004 2003
.005 .01 .015 .02
.9
.92
.94
.96
.98
1998 1997 1999
1996
0
1995 1970 2001 1973 1978 1974 1958 1979 1971 1972 1981 1985 1977 1994 1980 1990 1975 1982 1976 1987 1992 1993 1991 19861989 1984 1988 2002 1983
1949 1956 1969 1952 1957 1960 1953 1955 1964 1961 1954 19591962 1963
1950
1951
2007
2008
2006
2005
2004
2003
.005 .01 .015 .02
1965
1968
2000
.24
1966 1967
.26
.28
.3
.32
.34
1998 1997 1999
1995 1952 1957 1960 1970 1955 1964 1950 2001 1961 19541978 1963 1958 1979 1971 19851959 1981 1962 1972 1977 19801994 1990 19761975 1987 1992 1993 2007 1986 1991 1989 2008 1984 1988 2006 2005
1982
0
1983 2002
.7
1996
2000
1949
2003
1966 1967 1965 1968 1956 1969 1951 1973 1974
1953
2004
.75
.8
16
.85
.9
.34
FIGURE 13 - Smoothed cycles in Fed-based (u, π)- and output-based (u0 , π)-planes.
1951
.32
1950 1949
.3
2006 2007 1962 1959 1960
2005 2008 1961
.28
1958 1984 1976 2004
.26
1968 1997 1954 1996 1998 1995
1985
1957
1972 19991994 1989 1978 1977 19701988 1990 1991 1987 1992 1993 2000 1979
2003
1966
1956
1963
1967 1953
1969
1973
1971 1986
1975 1983
1965
1948 1952 1964 1955
1974
1982 1981
2002
2001
.24
1980
.75
.8
.85
.9
.34
.7
1951
.32
1950 1965 1949 1964
1966 1952
1955
1963 2007
.3
1962 1959
1960
1995
.28
1958
1976 1994
.26
19921975 1993 1983
2006 1956 1967 1968 1953
1957 2005 2008 1997 1954 1996
1961
1984 2004 1971 1977
1991
1948
1969
1998 1985 1972 1970 1986 1990 1987
2003
19991973 1989 1978 1988
1979
1974
2000
1982 1981 2002
2001
.24
1980
.9
.95
17
1
1.05
1.1
.8
.85
.9
.95
1
FIGURE 14 - Quarterly e, π , u for OECD countries
1970q1
1980q1
1990q1
2000q1
2010q1
1970q1
1980q1
1990q1
2000q1
2010q1
1970q1
1980q1
1990q1
2000q1
2010q1
.2
.3
.4
.5
1960q1
.6
.7
.8
.9
1
1960q1
1960q1
18
.03
.04
FIGURE 15 - Deviation of utilization and pro t shares in France from their long term trends in the (u, π)-plane
1989
1990
.02
.02
1990
.01
1988
1987
1992 1991
0
1995
0
1986
−.02
1984
−.01
1994 1985
.02
.04
−.06 −.04 −.02
.02
.04
1998
2001
.005
1995
2000
0
.005
0
.01
0
.01
−.02
1993
0
1999
−.005
−.01 −.005
1996
1997
−.02
−.01
0
.01
.02
.01
−.03
.005
2001 20072008
−.005
0
2006
2003 2002
−.01
2004 2005
−.04
−.02
0
.02
.04
19
1999
−.01
0
.01
.02
.03
.04
.03
.04
FIGURE 16 - Deviation of employment and pro t shares in France from their long term trends in the (e, π)-plane
1989
1990
.02
.02
1990
.01
1988
1987
1992 1991
0
1995
1984 1994
1993
−.01
−.02
0
1986
1985
.005
.01
.015
−.02
−.01
0
.01
.02
.01
0
.01
−.01 −.005
2001
1995
.005
.005
1998
0
2000
0
1999
1997
.01
−.014 −.012 −.01 −.008 −.006 −.004
2008
2007
2006
−.005
0
.005
2001
2003 2002
2004
−.01
2005
−.005
0
.005
.01
.01520
−.005
−.01 −.005
1996
1999
−.01 −.005
0
.005
.01
.015
Figure 17 - Deviation of employment and utilization in France from their long term trends in the (e, u)-plane
1986
−.02
1987 1984 1985
0
.005
.01
.04
.015
1993 1994
−.02
−.01
0
.01
.04
.02
−.01 −.005
1992
1995
−.06 −.04 −.02
0
1988
.02
2001
.02
.03
.01
1999
0
1995
2000 1999
0
.01
1996
−.01
−.03 −.02 −.01
1991
0
.02
1989
1998
1997
.04
−.014 −.012 −.01 −.008 −.006 −.004
.02
2001
2008 2007
2002
0
2003 2005
2004
2006
−.04
−.02
1990
.02
.04
1990
−.005
0
.005
.01
.01521
−.01 −.005
0
.005
.01
.015
.03
.02
FIGURE 18 - Deviation of utilization and pro t shares in Spain from their long term trends in the (u, π)-plane
1986
1984
0
1987
.01
.01
.02
1988 1989
1985 1984
0
1980
1990
−.02 −.01
−.01
1983
−.02
1982
1981
.01
.02
.03
−.02
0
.02
.04
.01
0
.02
−.01
−.01
0
0
.01
1995
1991
1992
1994
1992
−.02
−.03 −.02 −.01
1990
−.03
1993
−.02
0
.02
.04
−.06
−.04
−.02
0
.02
.005
.01
−.04
.005
19971996
2004 2002
1995
2003
0
1998
1999
−.02
−.01
0
.01
.02
.03 22
−.005
−.005
0
2001
2000 1999
−.02
0
.02
.04
.03
.02
FIGURE 19 - Deviation of employment and pro t shares in Spain from their long term trends in the (e, π)-plane
0
1984
1988 1987
.01
.01
.02
1986
1989
1985
1980
0
1984
−.02 −.01
−.01 −.02
1982
1981
.02
.04
−.04
−.02
0
.02
.04
.01
0
.02
−.02
−.01
−.03 −.02 −.01
1990
1991
1992
1994
1992
−.02
0
0
.01
1995
−.03
1993
.015
.02
.025
.03
.035
−.04
−.02
0
.02
.005
.01
.01
1996
.005
1990
1983
1997
2004 2002
1995 2003
0
1998
1999
−.03
−.02
−.01
0
.01 23
−.005
−.005
0
2001
1999
−.01
0
.01
2000
.02
.03
.04
.03
Figure 20 - Deviation of employment and utilization in Spain from their long term trends in the (e, u)-plane
1990
.01
.02
.02
1989
1987 1988
0
1982
1981
0
1983
1980
1985
1984
1984
−.02
−.01
1986
.02
.04
−.04
−.02
0
.02
.04
0
1990
1992
1992 1994
1993
−.04
−.06
−.02
1995
−.04
0
1991
−.02
.02
.02
0
.04
−.02
.02
.025
.03
.035
−.04
−.02
0
.02
.04
.015
.03
.01
.02
1999
.01
.02
1999
2001
0
0
1998
2003
−.02
1995
−.02 −.01
2000
1996 1997
−.03
−.02
−.01
0
.01 24
2004 2002
−.01
0
.01
.02
.03
.05
Figure 21 - Deviation of employment and utilization in Japan from their long term trends in the (e, u)-plane .06
1973
0
.02
−.05
.04
1974
1974
1976
−.1
1971
0
1977
−.15
−.02
1972
.002
.003
.004
.005
1980
−.004
−.002
0
.002
.05
.001
.05
0
1975
.004
1980
1985 1984
0
0
1979
−.05
1981 1978
1982
1977
−.1
−.05
1983
0
.001 .002
.1
−.003 −.002 −.001
1991 1990
.05
1989
1988 1985
0
1986
−.05
1987
−.005
0
.005
.01 25
−.004
−.002
0
.002
.05
.1 .05
1991
0
1997 1992
2000
0
1998 2001
−.05
1995
1998 1996
−.05
1993
1999
−.1
−.1
1994
.002
.004
.006
.008
−.01
−.005
0
.05
.05
0
.005
2008 2007 2006
2005 2005
0
0
2004
2001
−.05
−.05
2003
−.1
−.1
2002
−.015
−.01
−.005
0
−.005
26
0
.005
.01
