Wage determination and the gender pay gap: A feminist political economy analysis and decomposition

Abstract

This paper develops a heterodox analytical framework of wage determination and a new method of decomposition of the gender pay gap drawing on Marxian and feminist theories. The proposed framework utilizes two wage equations for the analysis of the gender gap: the first equation refers to average occupational wages and the second to individual wages as deviations from occupational wages. Using a data set for wages from industries in Greece, this paper demonstrates and explains differences in results between this proposed decomposition of the gender pay gap and that of Oaxaca-Blinder, and discusses the merits of this new technique compared to the Brown-Moon-Zoloth method. The authors argue that the main advantage of this proposed method of decomposition over the other two methods is that the proposed method allows for separate estimates of the impact of social and individual gender wage discrimination on the gender pay gap.

Keywords:

• Wage determination
• gender pay gap
• decomposition method
• JEL Codes: J16, J31, E11

Acknowledgments

We are grateful to the five anonymous reviewers and the editors of this journal for their very detailed comments and constructive suggestions for the improvement of this paper. We also thank Professors Jill Rubery, Robert Plasman, and Maria Luisa Moltó for their comments at previous stages of our research.

Notes

¹ John R. Hicks (1932, 1955) and the neo-institutionalists in the 1940s and 1950s also recognized the social determination of wages, as a result of the growing impact of unions and collective bargaining. However, this did not lead them to reject the methodological individualism of the neoclassical theory regarding wage determination. They claimed, instead, that a combination of social factors and market forces produces wage indeterminacy (Clark Kerr 1994).

² For a presentation of the pre-Becker literature on discrimination, see David Sapsford and Zafiris Tzannatos (1993).

³ Becker (1957) has defined three such types of taste discrimination stemming from employer, employee, and consumer preferences.

⁴ Edmund S. Phelps (1972) first put forward the statistical explanation of discrimination. Statistical discrimination may arise from differential expectations of employers regarding the productivity of men and women on particular jobs, based on knowledge of actual differences of average male and female productivity.

⁵ The argument for monopsonistic discrimination was first advanced by Florence in 1931 and then put in context by Robinson in 1933. They theorize that this form of discrimination stems from the fact that women are more prone than men to face monopsonistic conditions in the labor market because they are more constrained than men by family obligations.

⁶ Following Bergmann (1974), employment discrimination can be defined as unequal access of equally qualified individuals or groups of workers to prestigious jobs and occupations.

⁷ In literature, the econometric estimations of the relative importance of gender-specific factors, on one hand, and the wage structure, on the other, in explaining international differences and overtime changes of the gender pay gap are based on a decomposition method first developed by Chinhui Juhn, Kevin M. Murphy, and Brooks Pierce (1991) and then adapted by Blau and Kahn (1992).

⁸ In order to shorten our analysis we omit the concept of “regulating capitals” that would make our theoretical framework more comprehensive (Botwinick 1993).

⁹ Duménil and Lévy (1993, 1999, 2002) and Javier Herrera (1990) have shown, using US data, that industry profit rates do actually gravitate around a common value.

¹⁰ Following a well established theoretical tradition that understands the concept of tendency as the result of opposed forces of equalization and differentiation working simultaneously, our theoretical framework is not based on the assumption that profit rates actually equalize across and within industries. We rather assume that there are forces always at work drawing profit rates towards equalization and others pushing them towards differentiation.

¹¹ For the notion of productivity of capital, see Marx (1894/1993), vol. 3, “Economy in the Employment of Constant Capital.”

¹² Depending on the forms of competition, capacity utilization rates, and price elasticity of demand, a rise in an industry's price of production is possible in the short or the medium term. If the rise in price causes a severe fall in demand, it is more probable that firms will suffer a rather prolonged period of lower than average profitability, and profitability will be restored in the long term through capital movement between industries, lower capital accumulation rates in industries facing lower than average profitability, and eventually through devaluation of capitals for which profit margins are completely wiped out by high labor cost.

¹³ In Marx's terminology, the social need for a product is the volume of that product for which society is able and willing to pay. It corresponds to Adam Smith's concept of “effectual demand.”

¹⁴ Competition in the product market may be restricted, depending on the presence of monopolies/oligopolies in the industry and the openness of the market to imports and foreign investment.

¹⁵ An industry is considered to be expanding when the volume of product for which society is able and willing to pay exceeds supply for long periods of time.

¹⁶ Industry labor and capital productivity rises gradually as more firms in the industry adopt the new norms of production and each firm climbs its learning curve.

¹⁷ Numerous researchers have analyzed inter-industry wage differentials in OECD countries and found that the industrial wage structures have shown remarkable stability over time for all the countries examined (Maury Gittleman and Edward N. Wolff 1993).

¹⁸ For simplicity we omit the effect of lower p*_cap vs. p*_c, although this is an important element of Marxian theory. We, therefore, consider that p*_cap is the same across industries.

¹⁹ Minimum industry wages, extension of collective agreements to non-union firms, internal labor markets, and unemployment benefits are some of these institutional factors.

²⁰ Persistently excess demand is a situation where the volume of a product for which society is able and willing to pay exceeds supply for long periods of time.

²¹ Oaxaca and Blinder use the equation ln W _t  = Z _i β + u _i where ln W _t is the log of the hourly wage rate of the i-th worker, Z _i is a vector of individual characteristics, β is a vector of coefficients, and u _i is a disturbance term. That leads to a measure for discrimination referring to observed characteristics.

²² Joseph E. Zveglich, Jr. and Yana Rodgers (2004) have recently developed a decomposition technique that extends the Brown-Moon-Zoloth approach to explain trends in the gender pay gap. They attribute changes of the latter to changes in the relative employment distribution of men and women, the wage profile across occupations, the female employment share within each occupation, and relative pay between women and men within each occupation. However useful this technique may be for trend analysis of the gender pay gap, it does not improve the Brown-Moon-Zoloth method with regard to the analysis of the gender pay gap at a given point in time.

²³ From the total number of individuals in our two data sets, we have excluded only those with missing values for wages and occupation.

²⁴ Numerous studies have shown that productivity is correlated with plant size, at least in the manufacturing sector. Marx insisted on economies of scale and on the ability of worker collectives to accumulate knowledge and skills. Recent research produced more evidence on higher productivity in large establishments. For example, research and development investment is closely related to the firm size, and the increasing returns associated with research and development are an advantage of large firms. Furthermore, large firms have easier access to resources and information, which enables them to raise productivity. We therefore take establishment size as a proxy for labor and capital productivity. Using the average establishment size as a proxy for degree of competition in an industry is obviously a much less satisfactory solution, but we think it is a fair choice when there is no other proxy available and since large plants create higher barriers to entry. If established industrial capital already operates large plants, then potential new competitors must invest larger amounts of capital and wait longer to reach the same level of market productivity. Economies of scale are another barrier to entry that works to the advantage of large plants. Thus, the variable “establishment size” in our equations is designed to capture productivity of labor, productivity of capital, and product market power of leading firms. In a study of the determinants of industry wages the substitution of all the above-mentioned missing variables by “establishment size” would be unacceptable. However, in our paper, the substitution causes minor problems. This is because these particular industry characteristics, together with many other variables, serve only to estimate the coefficient in the equation for average occupational wages in industries, which allows for the decomposition of the segregation effect of the gender pay gap in two separate effects. Industry characteristics do not affect individual wage equations at all, since average reference wages used in individual wage equations are calculated directly from the sample.

²⁵ The first legalization procedure of undocumented immigrants in Greece in the 1990s took place in 1998.

²⁶ We have based our decomposition on the assumption that (γ) and are independent.