Gender Differentials in Industry Wage Premia, Affirmative Action, and Employment Growth on the Industry Level.(Statistical Data Included)

While most previous work has examined the gender wage gap by controlling for productivity-related factors, we focus on gender differences in industry wage premia (efficiency wages) on the detailed industry level and examine industry characteristics that might explain such differences. Using data from the 1988 CPS, we find that the extent to which firms in an industry were likely to be targeted for Affirmation Action compliance review, industry employment growth, and industry profitability have helped narrow the gender gap in industry wage premia. Other characteristics, such as capital intensity, mean education, and sales and payroll growth were much less important.

1. Introduction

Almost all studies of the gender wage gap have examined its evolution over time and space by controlling for productivity-related differentials between the two sexes (mainly, differences in education and experience). Yet, we know from recent work that one major factor explaining the variation of wages in the labor force is the existence of interindustry efficiency wage differentials. As a result, the inferences drawn from studies of the gender wage gap (such as the role of Affirmation Action programs) may be flawed by the failure to control for this source of wage variation.

The main contribution of this article is to explore the industry characteristics that are associated with significant differences between the genders in industry wage premia we found in Fields and Wolff, 1995. This is the first study to analyze gender disparities in efficiency wages on such a detailed level and the first, with two exceptions noted later, explicitly to consider efficiency wage differences by gender. Our analysis is based on data from the 1988 Current Population Survey (CPA) and data on the three digit industry level (a total of 158 industries).

To explain the difference between female and male industry wage premia, we focus on three industry characteristics in particular--the extent to which firms in each industry were likely to be targeted for Affirmation Action compliance review or investigation, industry employment growth, and industry profitability. We find strong evidence that all three factors helped narrow the gender gap in industry wage premia. Other characteristics that we have considered, including average plant size, the degree of industry unionization, the capital intensity of the production process, both the average level and variance in worker education, and changes in overall sales and wage levels, were statistically much less important.

The next section of the article (Section 2) provides background from the existing literature as well as a review of the various theories on why industry wage premia exist and why we might expect gender differences in them. Section 3 introduces our model and estimation techniques, and Section 4 discusses the regression results. Concluding remarks are made in the final part of the article, Section 5.

2. Literature Background and Theory

The fact that some industries pay higher (or lower) wages to the same grade of labor has been amply demonstrated in the literature, from Dunlop (1957), who first pointed out the existence of wage contours, Slichter (1950), and Weiss (1966) to the more recent studies reviewed in Dickens and Katz (1987). In this latter group, average industry wages (combining both male and female workers) are regressed on various industry characteristics, including the mean education and age of workers. Krueger and Summers (1988), for example, used industry wage premia--average industry wage levels adjusted for industry characteristics--as the dependent variable, by combining the wages of male and female workers. These studies also generally include the percent of females employed in the industry as an independent variable (typically yielding a negative regression coefficient).

One exception is Hodson and England (1986) who ran separate regressions on male and female earnings using average wages for each gender in each industry as the dependent variable. In contrast, our efficiency wage differentials are estimated from individual earnings data. A second is Blau and Kahn (1992) who examined male and female industry wage premia in different countries (including the U.S.), though they used only nine industry categories and only a small number of controls for occupation. However, where appropriate, we will compare our findings with theirs later.

The theoretical interpretation of why industry wage premia exist at all most widely accepted today is that of efficiency wage theory. Though there are some differences among versions of the theory, the common feature of efficiency wage theory is that paying higher wages helps to increase a firm's profitability. There are four factors cited to explain the benefits to the firm from paying higher wages: (1) the shirking model: eliciting increased effort (especially if the cost of monitoring workers is high); (2) the turnover model: reducing turnover costs; (3) the selection model: attracting a higher quality workforce; and (4) the sociological model: improving worker morale.

There is also the rent-sharing model (a variant on (4)). It holds that high profit firms in protected markets share some part of the profits with workers either because of strong union pressure or the threat of unionization or simply because they have the ability to pay higher wages.

Since the benefits of each of these factors will differ across industries, these theories all imply that different industries will, in general, pay different wages for the same work.

Krueger and Summers (1987), emphasize the importance of rent sharing as the explanation: More profitable industries, those with monopoly power, and those where labor's share is smaller, pay higher wages than less profitable ones. Managers maximize a utility function which includes both profits and the well-being of their workers. With an inelastic product demand curve, the cost of raising wages would thereby be reduced. They see this as essentially rent sharing which may coincidentally elicit greater effort from workers. They further point out that when individual workers change industries, their wage change is strongly correlated with the difference in average wage differentials between the industry they left and the one they entered (while it is unlikely that their ability has changed). Though industry wage premia may actually represent unmeasured differences in worker quality, Krueger and Summers (1987), point out: the pattern of research shows a remarkable similarity in the rank order of industries in te rms of wage premia across countries and also over time within the same country (also see Gittleman and Wolff, 1993, for similar evidence). So, if workers in high premia industries are more productive, there then must be some consistent, industry-specific characteristics that make them more productive.

A few words might be said about why there are differences in the industry wage premia for men and women. There are four potential reasons: (1) there are very likely to be geographic differences in the level and elasticity of female labor supply; male and female employees have different geographic distributions within the same industry. Although our earnings function regressions do control to some extent for place of residence (central city versus other; size of population of SMSA of residence; and four regions of the country), they do not control for differences in male and female labor supply by locality; (2) there are different elasticities of substitution of female for male labor across industries; given this and (1), above, in a locality with a limited supply of male labor, industries with a high elasticity of substitution of female for male labor would hire more females (with lower wage premia since female labor supply is more elastic) and industries with a low elasticity of substitution would have to p ay higher wage premia to attract male workers. Some supporting evidence is provided by Leonard (1996), who finds that changes in gender wage gaps over time varied widely across cities in the U.S. (3) Some industries may discriminate against female workers more or less than others; in the literature, Leonard (1984a) found more discrimination by gender than by race when he compared workers' wages with their human capital; he suggested that female workers may enter high wage industries in greater numbers at the cost of receiving pay below what is paid to men with similar human capital, and (4) the occupational composition of industries varies and male and female workers differ greatly in their occupational distribution. (We do partially control here for occupational differences in male and female employment but only by using 13 one-digit occupational dummy variables.) [1] Some evidence on this score is provided by Sicherman (1996), who finds a very high level of occupational gender segregation within the firms t hat he studied. His results show that women had higher quit rates than men in the firm, and he attributed this to the fact that women occupied lower-level jobs. If efficiency wages are paid to workers from whom long tenure is expected to be more likely, a disparity in gender wage premia within the same firm might result.

Finally, where earnings opportunities are …