It is now well established in the literature that similar workers earn very dissimilar wages when working for different employers. Early research made a strong case for the existence of inter- and intra-industry wage differentials for similar workers. The work of post-war neo-institutional labor economists such as Richard Lester (1952), Lloyd Reynolds (1949), and Sumner Slichter (1950) theorized about and offered empirical evidence in support of the claim that firms possess different pay policies. Contemporary economic theorists (Shapiro and Stiglitz 1984) and empirical labor economists (Dickens and Katz, 1987; Groshen 1991; Groshen and Krueger 1990) have returned to these same themes. Now sophisticated empirical techniques and far better data allow us to say with virtual certainty that wage differences across workers are in no small part the result of the institutional practices of firms rather than entirely due to the human capital skills of workers (Gibbons and Katz 1992; Abowd, Kramarz and Margolis 1999).
What exactly accounts for these differences? Neoclassical economics models of perfectly competitive labor markets hypothesize that pay differences across workers are entirely accounted for by the human capital characteristics of workers, the quality of the working conditions under which they labor, and the non-wage components (e.g., fringe benefits) of the compensation package. This analysis dates back to the pioneering work of Adam Smith and the so-called theory of compensating differentials. Historically, it has been the writings of institutional labor economists that emphasized pay differences based on noncompetitive features of the labor market--such as unions and the voluntary rent-sharing of employers--or the human resource practices and strategies of firms, wherein pay may be positively related to firm size, the cost of labor turnover, the difficulty of monitoring labor effort, or the need to attract a more talented applicant pool. However, with the emergence of noncompetitive and asymmetric information models in neoclassical theory, many of these institutional claims are also now part of the mainstream economics literature. (1)
For example, modern neoclassical models of efficiency wages hypothesize that firms may offer high wages as a strategy to reduce costly labor turnover and to increase the work effort of employees. It is now commonly accepted, and copious empirical evidence verifies, that unionized workers earn more, ceteris paribus. Wages are also hypothesized to be higher the greater the establishment's ability to pay out of profits, although if the appropriative power of workers is held constant, the voluntary sharing of economic rent by firms would appear to be ruled out in mainstream models by the profit motive. Only more institutional theories, wherein the motive of establishments is more complex, might posit the existence of voluntary rent sharing on the part of firms.
Regardless of the precise theory, however, essentially two empirical methodologies have been employed for exploring the question of pay differences across similar workers: (1) ask employers why their wage policies take the form they do or whether those policies conform to the predictions of economic theory (Blinder 1990; Bewley 1999); or (2) utilize existing survey data that are not ideally suited for the task and clever empirical strategies to generate statistical tests of particular hypotheses (Krueger and Summers 1988; Arai 2003).
In this paper, we explore theories of inter-establishment wage differentials using an original establishment survey. It neither asks firms if they ascribe to a particular theory in practice nor relies on imprecise measures of establishment characteristics, such as the capital-labor ratio or the ratio of managerial to non-managerial workers to capture monitoring capability, to shed light on the various hypotheses. Instead, we adopt a middle ground position: we ask specific questions about the nature of production (whether it is difficult to monitor workers), labor supply (the difficulty of recruiting workers or the cost of worker replacement), and the profitability of the establishment which allow more precise statistical tests than are possible with the second approach above, but without subscribing to the view, characteristic of the first approach, that human resource managers know why they do what they do.
The survey is confined to a limited geographical area--the city of Los Angeles at a particular point in time--2002--and to a select set of detailed industries, which allows for a more careful analysis of intra-industry, inter-establishment wage differentials. We explore the rent-sharing hypothesis by exploring the relationship between average establishment wages and profits. Because efficiency wage theories rarely make sense for an entire establishment, let alone an industry, these particular hypotheses are tested using information on the starting wage, monitoring capabilities, and recruitment difficulties in the largest low-wage occupation of establishments.
The results suggest that differences in the skills of workers account for a portion of intra-industry average and occupational wage differences, but so too do unions, profits, monitoring difficulty, and recruitment difficulty. Variation in non-monetary benefits sometimes exacerbates intra-industry wage differentials, contrary to the theory of compensating differentials, and differences in establishment size either have no effect on wages or affect wages negatively. While far from definitive, in large part due to a limited sample size, the results are suggestive of the importance of rent sharing and efficiency wage theories in explaining establishment wage differences and also illustrative of the value of this approach to understanding why wages vary across employers.
Data and Empirical Method
The data come from an original survey of establishments in Los Angeles, which we have entitled the Survey of Diversity in Human Resource Practices (SDHRP). The survey was meant to document differences in wage and benefits packages; human resource outcomes such as turnover, training, and absenteeism; production conditions and labor supply constraints; and profits among establishments in narrowly-defined industrial groups. Because the survey was also meant to serve as a control group for studying the impact of the Los Angeles Living Wage Ordinance, it focused on employers of low-wage workers in sectors affected by the ordinance-namely, janitorial, landscaping, security, parking, retail, food, childcare, and social services. Establishments covered by the living wage ordinance have been excluded from this analysis in an effort to eliminate wage variation due to government regulation. (2)
The experience of workers and firms in Los Angeles is probably not representative of the national experience, and so the results presented here may not generalize to similar industries in other metropolitan areas. Los Angeles has a high proportion of Hispanics and immigrants of both legal and illegal status. Also, compared to other cities, a higher percentage of low-wage jobs are unionized in Los Angeles, and especially among janitorial and hotel workers.
We began by creating a sample frame of establishments from the targeted set of industries. With the sampling frame established, a two-stage stratified sampling approach was invoked in which establishments were first divided into industry sectors, and then within each sector further divided into large (more than 50 employees) and small establishments. The survey was conducted by mail during the spring and summer of 2002, but there was also significant follow up phone contact with establishments both to answer questions about the survey and to urge surveyed firms to send in responses. Of the universe, consisting of 32,128 establishments in these sectors, a sample of 860 establishments was surveyed. There were 207 completed surveys, of which 190 were useable for the empirical analysis. The response rate of 24 percent is low but perhaps respectable for establishments in low-wage industries such as these. Survey weights are utilized to render the results representative of the population of establishments in these industries in Los Angeles.
Our analytical approach is to run a series of wage equations, in which average establishment wage and starting occupational wage are the dependent variables. We account for industry in the average wage analysis and both industry and occupation in the starting...