How important is where you work in determining how much you earn? Standard labor models take wages as given. Moreover, differences across firms have an impact on who is hired, but the models are silent about the level of pay for any particular worker. A major implication is that inequality trends in wages are driven by skill prices. Among economists the conception why wages vary largely among equally skilled workers remains up for debate.
Labor economists have traditionally argued that there is scope for employers in setting wages. Early empirical studies supported this view. However, the main difficulty is to disentangle industry wage premiums from the degree of sorting of different skill groups across firms. Due to the growing availability of matched employer-employee datasets, wage inequality questions can more readily be studied at the level of the firm. However, fundamental identification problems that plagued earlier studies remain present in these new datasets.
At the highest level, David Card, Ana Rute Cardoso, Joerg Heining, and Patrick Kline first summarize in their BSE Working Paper (No. 976) “Firms and Labor Market Inequality: Evidence and Some Theory” what has been learned so far about the role played by firms in setting wages from these new datasets. Second, the authors develop a model of wage setting.
The authors’ starting point is the observation that similar firms display large differences in productivity. The question then is whether some of these productivity differences carry over to wages. Evidence on the aggregate level supports this link. Nonetheless, the aggregate relationship is potentially driven by different groups of workers being assigned to different firms, i.e. the prevalence of sorting.
The authors portray two lines in the literature on factors at the firm level that drive wage inequality. Both literature lines attempt to circumvent the sorting issue by using matched employer-employee data.
The first line of literature studies the impact of firm productivity shocks on the wage of employees. A review of this line of literature by the authors suggests that estimated wage-productivity elasticities are quite robust and in the range between 0.05 and 0.15. The authors then present new evidence on the wage-productivity relationship using data from Portugal. They also address various specification issues such as techniques for dealing with unobserved worker differences and various approaches to measuring rents.
The second line of literature builds on the additive worker and firm effects wage model by Abowd, Kramarz, and Margolis (AKM). In a nutshell, data on wages by workers that move between firms is used to estimate firm-specific pay premiums. The literature typically finds that about 20% of the wage variance can be assigned to firm-specific effects. With respect to the second line of literature, the authors’ contribution is to provide evidence of the validity of the assumptions underlying the AKM model and to discuss some issues that arise when implementing AKM’s two-way fixed effects estimator.
In a next step, the authors proceed to provide a more direct link between the two related literatures. By using data from Portugal, they find that more productive firms not only pay higher wages on average but also have the tendency to hire more productive workers. In particular, 40% of the difference in wages between more and less productive firms is due to the sorting of higher-ability workers to more productive firms. This result highlights the importance of controlling for differences in workers.
In order to interpret the findings of the two lines of literature addressing wage inequality, the authors develop in a last step a model of wage setting. Workers have heterogeneous preferences over the work environments of firms. Examples are firm location and job characteristics, e.g. corporate culture or starting times for work. Consequently, firms are imperfect substitutes in the eyes of potential employees, which in turn gives firms more bargaining power over wages. Thus the model introduces imperfect competition in the labor market. Firms set the same wage for each skill group. Wages are inversely related to the elasticity of labor supply.
The model is able to predict many empirical observations. For example, heterogeneity in firm productivity affects the firm size distribution, the wage premium distributions, and the degree of sorting of different skill groups. When applied to the data, the model suggests a higher wage-productivity elasticity than search-models typically do. The authors also show that the firm-specific wage premium will be constant across skill groups if the said are perfect substitutes in production.
The authors conclude with some thoughts on unresolved empirical and theoretical issues in the literature. One major empirical challenge is to come up with research designs to credibly identify the causal link from firm-specific shocks to wages. Currently, studies rely on debatable identifying assumptions. An important question that needs to be answered by theory is how to model strategic labor market interactions between firms. The authors plead that future research on labor market inequality ought to pay attention to institutional details of particular labor markets, for example firms arguably have more saying in setting wages in some labor markets than in others, rather than having a “theory of everything”.