8

---

Women and Labor Market Flexibility

 

The Cases of Japan and the Former West Germany in the Postwar Years

 

David Kucera

 

 

There are two primary reasons for the increased interest in labor market flexibility since the early 1970s: first, the interest in production flexibility, often associated with the just-in-time production techniques used by larger Japanese manufacturers; and, second, the worsening as well as diverging unemployment patterns in the advanced capitalist countries after the early 1970s. While unemployment increased in Japan and Europe as well as the United States after the early 1970s, the increase was considerably greater in Europe, for which the experience of the former West Germany was typical. (All data in this study refer to the former West Germany or, in most recent years, to the regions of the former West Germany.) Even though the rate of unemployment in 1990 was lower in West Germany than in the United States, it was the manifold increase in unemployment in the former that drew particular attention (Bureau of Labor Statistics 1994).

These differences are widely held to result from differences in labor market flexibility, and thus increasing flexibility is argued to be vital to improving the European unemployment problem. The supposed lack of labor market flexibility in Europe is often attributed to the effects of strong government social policies and is commonly referred to as Eurosclerosis. One of the most prominent recent studies on unemployment is the OECD Jobs Study. The emphasis that the Study gives to the issue of labor flexibility in accounting for the different unemployment experiences of Europe and Japan is reflected in the following quotation, which considers the destabilizing effects in the 1970s and 1980s of the oil crises, the breakdown of Bretton Woods, financial market liberalization, the increased globalization of production, and the rapid pace of technical change: “[I]n the midst of this tumultuous period when so many forces were testing the flexibility of economies, policies to achieve social objectives were extended, with the unintended side-effect of making markets, including importantly labour markets, more rigid. This erosion of the ability to adapt to change was probably most pronounced in continental Europe and Oceania … Only the Japanese economy handled the problems of this period of adjustment relatively well” (Organization for Economic Cooperation and Development) [hereafter OECD] 1994a, 30).

Reflecting an influential viewpoint, an editorial in the May 30, 1986, Financial Times stated that “Japan’s low unemployment is primarily a reflection of very high internal mobility, America’s a reflection of high external mobility. Europe suffers high unemployment because it lacks either sort of mobility,” where internal refers to the mobility of workers within firms, the internal labor market, and external to the mobility of workers in the labor market at large (quoted in Metcalf 1987, 51–52).

A fundamental shortcoming with these points of view is that they neglect the very gendered nature of labor flexibility and labor market adjustment in Japan. Japan does in fact rely to a high degree on external labor flexibility, but this flexibility is disproportionately provided by women workers. These women serve very strongly as a buffer work force, moving in and out of employment and the labor force with much greater volatility than Japanese men. This study, by paying careful attention to the role of women in Japanese and West German labor markets, provides a substantially different view of both Japanese labor market flexibility and, just as important, Japanese macroeconomic performance since the early 1970s.

This study uses techniques drawn from Women and Recession, a 1988 volume edited by Jill Rubery, which provide a means of comparing men and women’s labor force and employment volatility over business cycles. By using standardized data and tests for Japan and West Germany, this study reveals that women served more strongly as a buffer work force in Japan than in West Germany. More than that, this study reveals strongly diverging patterns of women’s relative labor force and employment volatility over the post-1973 years of slower growth. There was a dramatic increase in these measures in Japan and no significant change in West Germany, indicating that Japanese women served more strongly as a buffer work force in the post-1973 years relative to Japanese men. Combined with data on the extremely high proportions of discouraged workers in Japan in recent years, 80 percent of whom are women, this study argues that the Japanese unemployment picture is a good deal less favorable than indicated by standard measures and probably not so different from West Germany, once one more fully accounts for the role of Japanese women as a buffer work force. As a result, one should be cautious of policy proposals suggesting that the West German labor system become more like the Japanese, in terms of general labor flexibility as well as government social policy.

This study is noteworthy not only in that it is directly comparative but because Japan and West Germany particularly merit careful comparison. The postwar experience of these countries was broadly similar, from the rebuilding of devastated economies to rapid output and productivity growth and eventual world economic leadership. The similarities held not just for broad macroeconomic factors, but for patterns of female labor force participation. In the early 1950s, just over 40 percent of the labor force in Japan was female, compared with about 36 percent in West Germany. These measures converged by the mid-1970s at about 37 percent, after which they increased in nearly lockstep fashion to 40 percent in 1990.¹ In terms of overall parallel movement, the pattern is even more similar for all employees (all workers minus self-employed and family workers) and all nonagricultural employees, with female percentages in Japan consistently 2 to 3 percent below those in West Germany over the postwar years. Female percentages for all and nonagricultural employees showed overall increases throughout the postwar years, with more rapid increases since 1970 in West Germany and 1975 in Japan.

The next section of this chapter considers the data on unemployment and discouraged workers as well as different kinds of labor flexibility—internal, external, numerical, and functional—particularly as they apply to the Japanese labor system. The section after presents data and methods used, followed by a section presenting the results of the data analysis.

Unemployment and Labor Market Flexibility

Over the years 1983 to 1991, unemployment rates averaged 7.2 percent in West Germany and 2.5 percent in Japan. Counting discouraged workers as unemployed over these same years would increase the Japanese unemployment rate by 125 percent, to 5.6 percent (OECD 1993a, 35; 1993b, 32–33). As noted, fully 80 percent of these discouraged workers were women. Data on discouraged workers are not, unfortunately, collected in West Germany, making direct comparison impossible. However, it is worth noting that among the eighteen OECD countries for which data on discouraged workers are available, Japan and, to a lesser extent, Sweden, are truly anomalous. These are also the two OECD countries with the lowest official rates of unemployment in 1991, with unemployment rates of 2.1 percent in Japan and 2.7 percent in Sweden, compared with the OECD average of 7.1 percent (for twenty OECD countries). The proportion of discouraged to unemployed workers was 90.8 percent in Japan that year and 54.1 percent in Sweden, compared with the OECD average of 14.3 percent (including Japan and Sweden). In no other countries did the proportion of discouraged to unemployed workers exceed 25 percent (OECD 1993a, 10). Unless West Germany also had exceptionally high proportions of discouraged workers, then, counting discouraged workers as unemployed would bring substantial convergence between Japanese and West German rates of unemployment. In addition, a large majority of involuntary part-timers in Japan and West Germany were women (70 and 80 percent, respectively, in 1991), and there were nearly twice as many involuntary part-timers in relation to unemployed workers in Japan as in West Germany over the span of years from 1983 to 1991 (OECD 1993a, 10, 16–17).

Toshiaki Tachibanaki writes, “There is a common understanding outside of Japan that the Japanese labour market is more flexible than those of other industrialized nations, and that this flexibility has facilitated to lower the rate of unemployment and to provide a better performance of the macroeconomy in general” (1987, 647). The neoclassical theory of unemployment suggests that the primary cause of long-term, nonfrictional unemployment is wage rigidity and that this would account for the very low rates of unemployment in Japan compared with West Germany.

However, a study by Robert Gordon suggests the similarity between eleven Western European countries and Japan in terms of nominal wage flexibility in relation to output changes. Gordon writes that “the differences among the United States, Japan, and Europe are minimal. Responses of the nominal wage rate to the output ratio are of roughly the same order of magnitude in the three aggregate economies” (1987, 689). Ronald Schettkat is also skeptical of the role of labor market flexibility in accounting for the large increase in unemployment in West Germany. He writes: “Neither intersectoral wage rigidity nor intrasectoral wage rigidity increased over recent decades in West Germany and thus neither can explain the rise in unemployment…. International comparison shows that countries that developed their institutional framework in the direction suggested by the ‘eurosclerosis’ thesis came up with the worst unemployment results (e.g., Great Britain) whereas economies with a strong welfare state (e.g., Sweden) did very well” (1990, 207; cf. Abraham and Houseman 1994, and Blank 1994, for similar arguments regarding the relationship between government social policy and labor market flexibility and unemployment). Schettkat also notes that both the inflow and outflow ratios for employment (measured by gross flows divided by initial employment) increased in West Germany from 1970 to 1986 (in spite of the increase in long-term unemployment), at odds with the argument that the increase in West German unemployment rates were caused by labor market rigidity (1990, 210, 221).

Schettkat’s mention of inflow and outflow ratios suggests the many aspects of labor flexibility. These ratios are related to external mobility or flexibility, noted above. Another widely used distinction is between functional and numerical flexibility, with functional flexibility referring to qualitative issues of labor flexibility within the firm and numerical flexibility referring to quantitative issues of labor flexibility both within and without the firm. Functional flexibility refers, more specifically, to issues of flexibility in job design and definition and the mobility of workers across tasks. In this sense, functional flexibility is closely associated with flexibility in production, such as the just-in-time system of production, for which functional flexibility is argued to be essential (Piore 1986, 162–63).

A recent OECD study describes what is seen as the increased emphasis of firms on functional rather than numerical flexibility, writing that functional flexibility “appears to have clearly outdistanced the Fordist overemphasis on numerical flexibility and the quantity of labor, controlling the number of hours worked and the organization of working time, regulation of hiring and dismissals, and the use of part-time and temporary workers” (quoted in Brodsky 1994, 59). The actual shift from numerical to functional flexibility is probably overstated, for the growing emphasis on functional flexibility is, at least in some circumstances, consistent with the growth of numerical flexibility. For example, some U.S. manufacturers have implemented alternative work schedules (breaking down the eight-hour day and five-day week) simultaneously with implementing so-called labor—management cooperation schemes such as quality circles, which generally emphasize aspects of production associated with functional flexibility. (Cf. Slaughter 1995, 3, regarding the Saturn Corporation’s plant in Tennessee, and Lund 1994, 7, regarding the NUMMI plant, a GM—Toyota joint venture in California.)

The organization of production in larger Japanese manufacturing firms has provided a widely emulated model for functional flexibility. At the same time, Japanese firms make extensive use of numerical flexibility. This is one of the primary themes of this study: in terms of hiring and firing, numerical flexibility in Japan is provided disproportionately by women workers, as indicated by their greater employment volatility relative to men. And many of the Japanese women who provide such a buffer work force are among the growing number of part-time and temporary workers, which the OECD study associates with numerical flexibility.

This study follows up on the body of literature that argues that the Japanese lifetime employment system, which applies almost exclusively to men, imposes a rigidity in the labor adjustment process, a rigidity accommodated by the highly flexible nature of Japanese women’s employment and labor force participation (Jones 1976–77, 592; Kawashima 1987, 611; Tachibanaki 1987, 669; Edwards 1988, 249; Carney and O’Kelly 1990, 127; Ono 1990, 87–88; Peterson and Sullivan 1990, 172–73; Hashimoto 1993, 141; Houseman and Abraham 1993, 45; Lam 1993, 198, 218). In a recent study, Susan Houseman and Katharine Abraham describe the relation of women to the lifetime employment system as follows: “The Japanese lifetime-employment system gives the average Japanese worker stronger job security than the average American worker. However, life-time-employment guarantees apply only to the core, “regular” work force. Peripheral workers, including temporary, day, and part-time workers, are more vulnerable to contract termination. These workers are predominantly female and are widely believed to serve as a buffer for the regular workforce” (1993, 45). In a similar manner, Linda Edwards writes: “It should be pointed out that the present system of employing men in full-time jobs and women in more casual, part-time jobs serves the Japanese economy well. With this system, Japanese firms are willing to guarantee many of their male workers job security over the business cycle because the firms have access to a pool of transitory (female) workers who readily enter and exit the labor force and are able to take on part-time employment…. True employment equality between men and women would upset the way the Japanese labor market currently adjusts to cyclical variations in aggregate demand” (1988, 249).

Data and Methods

The primary data sources for this study are drawn directly from Japanese and West German government publications. For Japan, labor force and employment data are from the Japanese Ministry of Labor’s Yearbook of Labour Statistics (based on household labor force surveys for aggregate data and on establishment surveys for manufacturing data, with results for the latter discussed in the section “Japanese Women as a Buffer Workforce in Recent Years”); for West Germany, labor force and employment data are from the Bundesministerium fuer Arbeit und Sozialordnung’s Arbeits- und Sozialstatistik: Hauptergebnisse (based on a combination of household and establishment surveys; the surveys are combined in the data source itself²). All data are annual.

In order to derive measures of women’s relative labor force and employment volatility, regressions are based on ordinary-least-squares estimates of the following equation:

 

 

where F is defined as the number of female employees or labor force participants; T, the number of male and female employees or labor force participants (with both F and T evaluated in terms of logarithmic growth rates); TIME, a linear time trend; and ε, an error term. The model is identical to that used in Women and Recession. The key issue in this model is whether estimates of β are significantly greater than 1, in which case it can be said that women function as a buffer work force.

A central element of this analysis is that the equation is also estimated with a dummy-modified T variable, providing a test of structural change for the post-1973 years of slower employment growth. In this case, the estimating equation is as follows:

 

 

where DUMMY is defined as 0 for years before and 1 for years after the slowdown. (The equation was also estimated by adding the dummy variable itself, allowing the intercept to vary, but coefficient estimates on the dummy variable did not prove significant.) In this model, estimates of β provide a measure of women’s relative labor force or employment volatility for the pre-1973 period, estimates of θ provide a measure of the change in such volatility between the pre- and post-1973 periods, and, by inference, the sum of estimates of β and θ provide a measure of women’s relative labor force or employment volatility for the post-1973 period. A key consideration in this second model is the sign and magnitude of estimates of θ. If, for example, an estimate of θ is positive, this indicates that women’s relative labor force or employment volatility increased in the post-1973 years, with the extent of increase given by the magnitude of the estimate.

For estimates of α’s, θ’s, and γ’s, the null hypothesis is that these equal zero and the non-null hypothesis that they do not; for estimates of β’s, the null hypothesis is that these equal one and the non-null hypothesis that they are greater than one, with two-tailed tests for estimates of α’s, θ’s, and γ’s, and one-tailed tests for estimates of β’s, as well as for confidence intervals comparing women’s relative labor force and employment volatility in Japan and West Germany. In the tables that follow, statistical significance is indicated by the number sign (#), one asterisk (*), and two asterisks (**) for 10, 5, and 1 percent significance, respectively (though a result deemed significant indicates at least 5 percent significance, unless indicated otherwise).

Last, standard deviations of the logarithmic growth rates of the time-series used in this model are derived, providing a different measure of women’s labor force and employment volatility in Japan and West Germany.

Women in Japan and West Germany as Buffer Work Forces

Japan

Table 8.1 shows results at the aggregate level for Japan. Columns numbered 1 through 3 show results based on equation (1) above; columns numbered 4 through 7 show results based on equation (2). Estimates of β are shown in columns 1 and 4, headed by the span of years for which β’s are estimated; estimates of γ are shown in columns 2 and 6, headed by “TIME”³ estimates of θ are shown in column 5, headed by “CHANGE.” Below each of these coefficient estimates are their associated standard errors, in parentheses. To facilitate presentation, results for a estimates are not shown. In columns 3 and 7 are adjusted R-squares, below which are indicators of any autocorrelation (ARMA) corrections, in parentheses, as well as Durbin-Watson statistics.

Headings for levels of aggregation, to the leftmost, are followed by “(D)” when an additional level of differencing is required to attain stationarity. Below these headings are average female percentages for the entire period under estimation. (This format is also followed in Table 8.2 for West Germany.) Results are shown for the labor force, all workers, all employees (minus self-employed and family workers), and nonagricultural employees.

Estimates of Japanese women’s relative labor force and employment volatility in column 1 of Table 8.1 indicate that Japanese women served as a buffer work force throughout the postwar years. Results are particularly striking for the entire labor force and all workers, with the estimates of women’s relative volatility 86.7 and 71.9 percent greater, respectively, than if men’s and women’s volatility were identical (that is, if estimates of β were equal to one).

For each of the four aggregate levels, equation (2) with the test of structural change provides a better specification, as indicated by higher adjusted R-squares than for equation (1). Using 1973 as the turning point, tests of structural change are significant at the 5 percent level or better for the labor force, all workers, and nonagricultural employees, though at only the 10 percent level for all employees. Alternative and multiple turning points are also tested for years in the 1970s and early 1980s, based on recursive residuals tests as well as the more stable growth pattern following the second oil shock of the late 1970s. For the labor force and all workers, the 1973 turning point yields the largest coefficients of structural change. For all employees and nonagricultural employees, 1974 and 1975 turning points yield slightly larger coefficients of structural change, but not significantly so (at even the 10 percent level, based on confidence interval tests) and do not change the significance of regression results. The 1973 turning point was combined with turning points from the late 1970s and early 1980s, but only the 1973 turning point proved significant.

 

---

 

 

For ease of visual representation, coefficient estimates of women’s relative labor force and employment volatility are shown in Figure 8.1. The dark bars represent coefficient estimates for the 1952 to 1973 years, and the light bars for the 1973 to 1990 years.

The result most worthy of notice is that for the labor force. As argued in the opening of this study, the success of labor adjustment processes is usefully evaluated in terms of differences by gender. Results for Japan indicate that women function more strongly as a buffer work force by a large and significant measure. Were Japanese women’s labor force participation not so tenuous, Japan’s unemployment rates would be a good deal higher. Just as important, Japanese women’s relative labor force participation became a good deal more volatile after the 1973 slowdown in economic growth. For the 1952 to 1973 period, Japanese women’s relative labor force volatility was 67.7 percent higher than if men’s and women’s volatility were identical; for the 1973 to 1990 period, the figure increased to 150 percent (based on the sum of estimates for the 1952–73 period and for the test of structural change, or β plus θ from equation [2]). That is, the volatility of Japanese women’s labor force participation in the 1973 to 1990 period was an estimated two-and-one-half times greater than the volatility of total (men’s and women’s) labor force participation. This suggests that one should view Japan’s seemingly favorable unemployment patterns with skepticism.⁴

West Germany

Results at the aggregate level for West Germany are shown in Table 8.2. For the labor force and all employees, there is significant evidence that women served as a buffer work force, though at only the 10 percent level for all employees over the entire 1952 to 1990 period. There is no significant evidence of structural change for the post-1973 period, and adjusted R-squares are identical across equations (1) and (2). For all workers, all employees, and nonagricultural employees, estimates of structural change are negative, such that estimates of women’s relative employment volatility for the 1973 to 1990 period are all less than one.

Coefficient estimates of women’s relative labor force and employment volatility are shown in Figure 8.2. As with the graph for Japan, the dark bars represent coefficient estimates for the 1952 to 1973 years, and the light bars for the 1973 to 1990 years. A comparison of Figures 8.1 and 8.2 makes apparent the strongly diverging patterns of women’s relative labor force and employment volatility in Japan and West Germany, with substantial increases in these measures in the former and, generally, declines in the latter over the post-1973 years of slower employment growth.

Confidence Intervals Testing Differences Between Japan and West Germany

Coefficient estimates for women’s relative labor force volatility are consistently and significantly larger in Japan than in West Germany. This is revealed most clearly in Table 8.3, which shows the results of confidence interval tests. For the entire span of years considered and for both the pre- and post-1973 subperiods, estimates for Japan are larger than for West Germany and larger at the 1 percent level of significance. The same pattern is observed for all workers. For all employees, coefficient estimates of women’s relative employment volatility are consistently larger, but significantly larger for only the post-1973 subperiod. For nonagricultural employees, coefficient estimates are significantly larger for both the pre- and post-1973 subperiods, though at only the 10 percent level for the entire 1952 to 1990 period. These confidence interval tests provide strong evidence that Japanese women served more as a buffer work force than did German women.

 

---

 

 

Japanese Women as a Buffer Work Force in Recent Years

In their 1993 article titled “Female Workers as a Buffer in the Japanese Economy,” Houseman and Abraham test women’s relative employment volatility in Japan and the United States for the manufacturing sector as a whole. Comparing the 1980s with the 1970s, their empirical work for Japan indicates that analogous measures of women’s employment volatility declined relative to men’s in the more recent period, though Japanese women continued to serve significantly more as a buffer work force than Japanese men even in the later period. Using monthly data, the authors construct a finite distributed lag model, regressing the logarithmic differences of, separately, women’s and men’s employment on the logarithmic differences of current and lagged manufacturing output.⁵

 

---

Table 8.3

Confidence Interval Results, 1952–1990

 

	(1)	(2)	(3)
		Subperiods
	1952–1990	1952–1973	1973–1990
Labor force	**	**	**
All workers	**	**	**
All employees	—	—	*
Nonagricultural employees	#	*	*

 

Sources: Japan Ministry of Labor, Yearbook of Labour Statistics, Bundesministerium fuer Arbeit und Sozialordnung, Arbeits- und Sozialstatistik.

Notes #, *, and ** indicate significance at 10, 5, and 1 percent levels, respectively; “—” indicates no significant difference between coefficient estimates.

---

For the 1970 to 1979 period, the twelve-month elasticity (the sum of coefficient estimates on current output and output lagged through twelve months) is 0.456 and 0.160 for women’s and men’s employment, respectively; for the 1980 to 1989 period, the twelve-month elasticity is 0.316 and 0.195 for women’s and men’s employment, respectively. For both periods, women’s employment elasticity is significantly greater than men’s at the 5 percent level. Yet female-to-male ratios based on these elasticities declined substantially, from 2.85 in the 1970s to 1.62 in the 1980s (Houseman and Abraham 1993, 47).

In an effort to address this issue, equation (2) from above is estimated using data for the manufacturing sector with both 1973 and either 1979 or 1980 turning points. Consistent with Houseman and Abraham’s findings, coefficient estimates for the 1979 and 1980 turning points are negative. They are not, however, significantly so, with t-statistics of 1.11 and 1.00 for the 1979 and 1980 turning points, respectively. Coefficient estimates for the 1973 turning point remain positive and significant at the 1 percent level, consistent with the view that the early 1970s marked a more decisive turning point over the span of the postwar years.⁶

That the 1980s represented only a temporarily altered situation for women workers in Japan is also suggested by comparing growth rates of male and female labor force participation and employment across the upswing of 1987 to 1991 and the downswing of 1991 to 1993. Table 8.4 shows average annual growth rates of labor force participation, total civilian employment, and industrial employment for men and women over the 1987 to 1991 and 1991 to 1993 periods. The table shows both that women’s labor force participation and employment grew more rapidly than men’s in the 1987–91 upswing, and that women’s labor force participation and employment grew less rapidly than men’s in the 1991–93 downswing. This is also shown by the gap between employment growth in the up- and downswing years in row C, showing the difference between values in row B and row A. By all categories—labor force participation, total employment, and industrial employment—the gap is much larger for women, indicating the greater difference across periods. Last, ratios are constructed by dividing the female gap by the male gap, shown in row D of the table. These ratios reveal that the difference in average annual growth rates of labor force participation across the up- and downswing was 5.7 times greater for women than for men. For total employment, the ratio was 2.9 and for industrial employment 4.9. These measures reveal that women continue to serve strongly as a buffer work force in Japan into the present.

 

---

 

The continued role of women as a buffer work force in Japan is also suggested by examining the numbers of discouraged workers in relation to labor force participants over the 1987 to 1991 upswing and 1991 to 1993 downswing. As noted earlier, Japan had by far the highest ratios of discouraged to unemployed workers among the eighteen OECD countries for which data on discouraged workers are available, and from 1983 to 1991 the proportion of discouraged workers that were women averaged just over 80 percent (OECD 1993a, 10, 35; 1993b, 32–33).

Looking at discouraged workers as a percentage of the labor force, the proportions of discouraged workers declined by an average annual rate of 7.1 percent from 1987 to 1991. From 1991 to 1993, this pattern was reversed, with the proportions of discouraged workers increasing at an average annual rate of 4.5 percent (Sorrentino 1995, 34). These different patterns across the up- and downswing provide additional evidence indicating that women continue to play a vital role as a buffer work force in Japan, and that the difference in the employment situation of Japanese women in the 1980s was only temporary, resulting from the relatively favorable employment growth of the decade in the face of a labor shortage.

This continued role of Japanese women as a highly marginal buffer work force is also supported by more anecdotal evidence. In the December 1, 1992, New York Times, David Sanger wrote as follows:

When Japan’s economy boomed through the late 1980’s, doors opened and women edged into choice jobs that freed them from serving tea and answering telephones, the traditional duties of the clerical workers known as “office ladies.” But Japanese women today say that many of the changes were superficial, and that career opportunities are evaporating as the economy dries up …. Many of those [women entering the workforce in the 1980s] were middleaged women who re-entered the workforce after their children reached school age. And many were part of the huge corps of what Japan loosely terms its part-time workers, many of whom actually work more than 40 hours a week. When Japan’s economy was humming, such workers were evident everywhere. … With the boom’s abrupt end, though, part-time workers were the first to go.

Comparative Labor Force and Employment Volatility

Table 8.5 shows standard deviations of the logarithmic growth rates of labor force participation and employment for all workers, all employees, and nonagricultural employees, separating the 1952 to 1973 and 1973 to 1990 years as was done in the aggregate level regressions. These standard deviations are derived from the independent variables used in the time-series regressions. Examining these measures of volatility provides a somewhat different sense of women’s relative labor force and employment volatility. More specifically, even though women’s labor force and employment volatility relative to men’s is significantly higher in Japan than in West Germany, this is often relative to substantially lower labor force or employment volatility. That is, even though women unambiguously function more as a buffer work force than men in Japan than in West Germany, Japanese women are less unambiguously worse off than German women in certain cases. This results when total (men plus women) labor force or employment volatility is substantially lower in Japan.

 

---

Table 8.5

Standard Deviations of Logarithmic Growth Rates of Total (Men plus Women) Labor Force Participation and Employment, 1952–1990

 

	1952–1973		1973–1990
	Germany	Japan	Germany	Japan
Labor force	0.013	0.010	0.007	0.006
All workers	0.018	0.010	0.013	0.007
All employees	0.021	0.021	0.014	0.009
Nonagricultural
employees	0.022	0.021	0.014	0.009

 

Sources: Japan Ministry of Labour, Yearbook of Labour Statistics; Bundesministerium fuer Arbeit und Sozialordnung, Arbeits- und Sozialstatistik.

---

For the 1952 to 1973 period, the volatility of employment for all employees and nonagricultural employees was, by these measures, nearly identical in West Germany and Japan. For labor force participants and all workers, though, these measures are lower in Japan by 23.6 and 40.2 percent, respectively. In 1973 to 1990, years of slower employment growth, these measures of volatility are about the same for labor force participants but only half as high in Japan for all workers and 38.9 and 39.4 percent lower in Japan for all employees and nonagricultural employees, respectively. In those cases in which the measures of volatility are higher in West Germany, then, Japanese women’s high labor force and employment volatility is moderated relative to German women though not, of course, relative to Japanese men.

This issue is examined by comparing standard deviations of the growth rates of women’s labor force participation and women’s employment, shown in Table 8.6. With the exception of all workers in both periods, standard deviations are consistently larger for Japan. This biggest gap is for the labor force in the 1973 to 1990 period, for which the standard deviation for Japan is 65.1 percent larger than in West Germany. For all workers, all employees, and nonagricultural employees, though, the standard deviations are not much larger in Japan, suggesting that women in Japan were not, in this sense, much more strongly affected by employment volatility than women in West Germany.

 

---

Table 8.6

Standard Deviations of Logarithmic Growth Rates of Women’s Labor Force Participation and Employment, 1952–1990

 

	1952–1973		1973–1990
	Germany	Japan	Germany	Japan
Labor force	0.017	0.018	0.008	0.013
All workers	0.020	0.018	0.015	0.014
All employees	0.025	0.031	0.015	0.016
Nonagricultural
employees	0.026	0.032	0.015	0.016

 

Sources: Japan Ministry of Labour, Yearbook of Labour Statistics; Bundesministerium fuer Arbeit und Sozialordnung, Arbeits- und Sozialstatistik.

---

Conclusion

Takafusa Nakamura writes, “Is it perhaps true that without a cushion against business slowdowns in the form of some kind of labor force that can be readily sacrificed, Japan’s employment system is untenable?” (1995, 162). This study provides evidence in the affirmative to Nakamura’s question, suggesting that the impediments to numerical flexibility imposed by the predominately male lifetime employment system are accommodated by the role of Japanese women as a buffer work force. Japanese women served all the more strongly as a buffer work force after the early 1970s, during which the Japanese labor market was lauded for its low unemployment and its flexibility in adapting to dramatic change in the world economy.

This study does not argue against the importance of internal and functional flexibility to the Japanese labor system. Nor does it argue that the superiority of Japanese over West German macroeconomic performance was mere illusion. There is no question, for example, that the Japanese economy did much better at creating jobs in the post-1973 period of slower growth (OECD 1994b, 4). All the same, the gender-blind approach typical of comparative studies of labor market flexibility and unemployment neglects the importance of external numerical flexibility in the Japanese economy, provided by women entering and exiting employment and the labor force at much higher rates that Japanese men. Such flexibility, reflected in the exceptionally high rates of discouraged women workers in Japan in recent years, indicates than Japanese unemployment rates are greatly underestimated. Were discouraged workers counted as unemployed in Japan, the average annual rate of unemployment in Japan over the 1983 to 1991 period would be more than double the official rate, giving a significantly different view of the success of labor adjustment in Japan in the face of slower economic growth.⁷

Notes

1. The decline in Japanese women’s labor force participation prior to the mid-1970s is accounted for by the declining importance of self-employed and family workers, particularly in agriculture.

2. In the 1991 volume of Arbeits- und Sozialstatistik [Employment and social statistics], the data are described as follows: “aufgrund der Ergebnisse der Volks- und Arbeitsstaettenzaehlung” [“on the basis of population and employment census”](26); in the 1975 volume, the data are described similarly as follows: “auf Grund der Ergebnisse der Volks- u. Berufszaehlung” [“on the basis of population and occupational census”] (26).

3. The linear time trend was included only when significant at the 5 percent level, following the analyses in Rubery’s Women and Recession.

4. The importance of the 1973 turning point is suggested by Fujita, who wrote that “in 1973’s oil crisis, companies tried all kinds of measures not to lay workers off, including shuffling, shorter work hours, etc. However, this was not the case for women workers (particularly the part-time workers) who were the first to be laid off in a recession. The state did not intervene in this practice of firing the female labor force, nor did the company union protect women workers who were to them only temporary workers” (1987, 593–94).

5. Houseman and Abraham’s model also includes a constant and a linear time trend.

6. Using 1973 and 1979 as turning points, the coefficient estimate for the 1973 turning point is 1.388, with an associated standard error of 0.260; the coefficient estimate for the 1979 turning point is −0.433, with an associated standard error of 0.389. Using 1973 and 1980 as turning points, the results are essentially the same.

7. See also Hamada and Kurosaka (1984, 72–76), Taira (1985), and Eatwell (1995) regarding other arguments that Japanese unemployment rates may be deceptively low.