Do Gender Disparities in Employment Increase Profitability? Evidence from the United States

Abstract

This paper investigates whether the contribution of the declining share of wages in national income to the upswing in profitability between 1982 and 1997 in the United States was aided by the growing incorporation of women into employment. The analysis finds that women helped moderate the decline in the aggregate wage share. The reduction in gender pay disparity overwhelmed the negative effect of women's growing share of market work on the wage share. However, in (one-digit) sectors where wage shares fell, women did not contribute to restraining the fall, indicating that the aggregate outcome was the net result of distinct sectoral trends in women's employment conditions. We argue that the perverse process of labor productivity falling faster than the real wage in the service sector may have played a key role in shaping the aggregate outcome. The post-1997 trends in the US are discussed in a postscript.

Keywords:

• Profitability
• feminization of the labor force
• gender wage gap

ACKNOWLEDGMENTS

We are grateful to the anonymous referees and the editors of the Special Issue of Feminist Economics for their comments and suggestions. We would also like to thank Korkut Ertürk and Ramaa Vasudevan for their comments on earlier drafts of this paper. We alone remain responsible for any errors and omissions.

Notes

The period since 1997 has not displayed any sustained movement in the US's profit rate till 2005 (the latest year of data that was available at the time of writing); it has also not seen a steady deterioration of the wage share and rising feminization (see the postscript for details).

While the empirical validity of the observable implication of a causal mechanism is not the sole criterion by which a theoretical hypothesis is judged in normal scientific practice, it is definitely an important consideration. Regression models are also open to similar caveats regarding causality.

The lower average wage could also trigger changes in the methods of production employed in the economy and via that channel influence the rate of profit and relative prices. This effect of the growth in female labor supply, though worthy of study, is probably much harder to quantify than the direct effect on the average wage.

The rate of capacity utilization is considered in this tradition as the main determinant of the cyclical fluctuations in the profit rate (Anwar Shaikh 1999).

The data on value added and employee compensation in private industries was taken from the “Industry Economic Accounts” section of the website of the US Department of Commerce, Bureau of Economic Analysis (BEA, n.d.a.). Gross value added is the “gross product originating” definition of the US Department of Commerce, BEA. The data on capital stock in private industries was downloaded from the “National Economic Accounts” section of the US Department of Commerce, BEA website (n.d.b.). Capital stock refers to “the current-cost net stock of private nonresidential fixed assets” definition of the US Department of Commerce, BEA.

 The US Department of Commerce, BEA does not publish a series for the price deflator of net capital stock. We derived an implicit deflator by converting the current dollar series into a chained-dollar series using the quantity index for net capital stock (available from the BEA, Table 4.2) and then calculating the ratio of the current dollar series to the chained-dollar series. Such a procedure is justified because the percent change in the chained-dollar values will be identical to the percent change in the quantity index. The chain-type price index of value added was taken from the US Department of Commerce, BEA (n.d.a.).

The periodization of the upswings in the profit rate was performed by examining the smoothed profit rate series. (The smoothing was done using the standard Hodrick-Prescott filter.) Visual inspection of the filtered series suggested that a sustained upward movement in the profit rate begins from the first year of the chosen period and ends in the last year of the period. That is, for example, 1982 is the year from which the filtered profit rate in every year is higher than the profit rate of the previous year, and 1997 is the last year for which this observation is true. Since the latter period (1982–97) is the focus of this paper, we also conducted a Chow-test to examine whether the two chosen years could be considered as breakpoints in the profit rate series stretching from 1947–2002. The results of the Chow test confirmed the results from the visual inspection. The F-values for 1982 and 1997 were, respectively, 34.38 (0.0001) and 6.38 (0.003), with the numbers in parentheses indicating the p-values. Our periodization also generally accords with that followed in similar studies cited in this paper (Wolff 2003; Duménil and Lévy 2002).

We discuss the post-1997 trends in the postscript.

Consider the equation x = yz. The standard decomposition resolves the change in x between two periods (say 0 and 1) into the weighted sum of changes in the right-hand side variables according to the following equation: , where the bar over a variable indicates that it is the average of the values corresponding to the two periods.

The decomposition of the determinants of the profit rate into wage share and output-capital ratio is also not a causal separation because these variables are not independent of each other. Any substantial change in the output-capital ratio due to technological change can be expected to have an effect on the wage share via its effect on labor productivity; while substantial changes in the real wage can influence the kind of methods of production that are employed (as well as the relative prices) and hence have an impact on the output-capital ratio.

We borrow the term “gendered macroeconomic variables” from Nilüfer Çağatay, Diane Elson, and Caren Grown (1995: 1830). In a sense, the approach taken in this paper can be viewed as an empirical implementation of the “gendered macroeconomic variable method” that they discuss.

The employment-population ratios referred to in this paragraph pertain to individuals who are 16 years and older. They are based on annual household surveys and can be accessed at US Department of Labor, Bureau of Labor Statistics (BLS n.d.b.).

Focusing specifically on the labor-supply behavior of married US women, Francine D. Blau and Lawrence M. Kahn (2005: 21–5) find that the elasticity with respect to husband's earnings was small and falling even prior to the 1980s and that the decline continued in the later years (see also Richard Blundell and Thomas MaCurdy [1999] for related analyses and references).

The relative wage reported here pertains to full-time, year-around workers. It is based on annual household surveys conducted by the US Census Bureau (n.d.).

An interesting issue is the effect of the rise in overall wage inequality on the gender wage gap over the period. One would expect that since women tend to have less marketable skills than men, the factors that aggravated wage inequality – technological change and international competition – would have a negative impact on the gender wage gap. Blau and Kahn (1997) found that over the 1980s the impact of rising inequality and greater returns to skill, while sizeable, was more than offset by improvements in women's measurable qualifications and either a reduction in gender discrimination or an improvement in women's unmeasured labor market skills.

Stephen Rose and Heidi I. Hartmann (2004) look at a number of alternative measures of earnings disparity in contrast to the conventional measure that considers only full-time, year-round workers in a given year. They utilized the longitudinal data available from the Panel Study of Income Dynamics to develop their measures and found that the conventional measure may considerably understate the size of the gender wage gap. For example, they found that the gender gap in average annual earnings over the 1983–98 period ranged widely from about 62.1 percent to 36 percent depending on the population for which the earnings were computed. The former figure measured the gender gap in annual earnings calculated over all prime-age workers with reported earnings for at least one year between 1983 and 1998, whereas the latter measured the gap computed over all full-time, full-year workers, with earnings reported each year over the period. Much of the discrepancy between these measures is due to the fact that women are traditionally less likely to hold full-time, full-year jobs and have continuous work histories. In addition, there is persistent segmentation of the labor market by sex in which women are more likely to earn less than their male counterparts both within occupations and across traditionally female and male job types.

In contrast, in standard marginalist theory, there is not much room for social and political factors to intervene in shaping how effective the workers are in capturing productivity gains because there are strict limits within which the wage share can vary. For example, if the aggregate production function is assumed to be Cobb-Douglas – that is, Y = AK ^1–α L ^α– then it is easy to show that under the normal assumptions of perfect competition and homogenous factors of production (K and L represent, respectively, capital and labor), the equilibrium wage share will be equal to the elasticity of output with respect to labor (α). With more general production functions, the constraint on the wage share is still technologically determined (see, for example, Samuel Bentolila and Gilles Saint-Paul [2003]).

The ADS (also known as the March CPS) is the source of official data on US household income. For more information about the ADS, please see US Department of Labour, Bureau of Labor Statistics (BLS n.d.a). The calculations described in the text were carried out using the public-use version of the ADS micro-data. Sample sizes for men range from 45,048 in 1982 to 36,583 in 1997 and for women from 36,942 in 1982 to 33,169 in 1997.

Employer contributions for social insurance consist mainly of mandatory payments by employers to the government for Social Security, unemployment insurance, and workers' compensation.

Other labor income consists of employer contributions to private group health and life insurance, private pension and profit-sharing plans, private supplemental unemployment insurance, private workers' compensation, and publicly administered government employee retirement plans.

In both instances, we applied population weights to obtain aggregate estimates.

Employer contributions for health insurance amounted to about 55 percent of all other labor income in 1997. Pension and profit-sharing plans accounted for another 40 percent of other labor income in the same year; however, roughly 60 percent of these were publicly administered retirement plans for government employees. The remaining 10 percent was distributed among workers' compensation, life insurance, and supplemental unemployment insurance.

A full investigation of the relative importance of these processes in shaping the substitution observed at the aggregate level as well as their relation to the processes of globalization and structural and technological changes is a task that is well beyond the scope of this paper.

The increase in US women's share of market work also reflects the growth of, and women's employment in, sectors producing market substitutes for the products of household production (for example, childcare, adult care, and prepared foods). The effect of the latter on the aggregate wage share and profitability would depend on the institutional arrangements under which these substitutes are provided (for example, public provisioning of childcare in the Scandinavian countries versus the predominantly private provisioning in the US) and their conditions of production (for example, the labor intensity of the activities and the presence of collective bargaining).

This is because the partial derivatives of the wage share with respect to the female share of employment and wage disparity are, respectively,

For example, if the degree of gender segmentation of the labor market had remained unchanged, the growing share of women in employment would occur mostly in the bottom rungs of the wage distribution. Other things remaining the same, such a process can indeed result in a rising, rather than falling, gender pay gap.

The data for the value added and employee compensation for the individual sectors were also drawn from the US Department of Commerce, BEA (n.d.a., n.d.b.) as discussed in note 4. A minor point to note regarding the sectoral disaggregation is that the sum of value added of the individual sectors will be different from the aggregate value added because of the very small amount of “statistical discrepancy,” the amount of value added that could not be allocated across sectors. As a result, the aggregate wage share reported in this section differs slightly from that reported earlier.

As we pointed out in the context of the discussion of the decomposition of the US's aggregate wage share, we are not suggesting that the relationships examined by the decompositions are necessarily causal. We are taking the changes in the sectoral labor ratios and compensation rates as given and then estimating the impacts of these factors on the US's wage share. However, following the same logic as in the instance of aggregate decomposition, this does not mean that the results from the decomposition are purely tautological.

The sectoral hourly compensation rates and labor ratios by gender were estimated by utilizing the ADS in conjunction with the US's national accounts. Details about the calculations are available from the authors upon request.

The characterization of services sector as a low-wage sector holds only on the average. Several sub-sectors in it include some of the highest-paid occupations, such as legal and health services.

The decline in the US service sector's labor productivity appears to fit its frequent characterization as the culprit in the familiar “cost-disease” model originally proposed by William Baumol (William J. Baumol 1967; William J. Baumol, Sue Allen Bateman, and Edward N. Wolff 1987). The poor wage trends in the service sector, however, need not be the effect of low productivity in a causal sense, since the availability of a low-wage labor pool can make the less capital-intensive, low productivity methods of production more profitable to firms (Lester Thurow 2000).

Capital-saving technological innovation over the US's “leaden age” is likely to have strengthened the positive contribution to the rise in the profit rate made by real output-capital ratio and the price of output relative to capital stock.

Compensation growth for both men and women was notably stronger in the US between 1997 and 2005 than between 1982 and 1997. Between 1982 and 1997, average compensation for men rose at a paltry 0.7 percent per annum and at 1.7 percent for women; in contrast, the annual growth rate was much higher at 2.7 percent for both women and men between 1997 and 2005. However, wage gains were not distributed evenly across the income distribution. Thomas Piketty and Emmanuel Saez (2003), for example, find that the share of wage and salary income going to the top 1 percent of tax filers in the US increased between 1980 and 2005 from 6.4 percent to 11.6 percent. Though a full analysis of the gender dimension of the distribution of wage gains is beyond the scope of this study, findings by Frank Levy and Peter Temin (2007) suggest that women may have contributed considerably to the gains made at the top of the US's wage distribution. They find that between 1980 and 2005, college-educated women in the US were the only major labor market group whose weekly wages kept pace with productivity gains in the US economy.