In theory, non wage job characteristics (e.g. type of work, working conditions, job security) are potential determinants of wage dispersion and labor market turnover (see Rosen, 1986). However, very different estimates of workers’ Marginal Willingness to Pay (MWP hereafter) for these amenities have been obtained using either cross-sectional data on wages and amenities, or job duration data. Hwang, Mortensen and Reed (1998) build a structural on-the-job search model that provides an explanation for these conflicting results. In this paper, we take partial equilibrium version of their model to data on European countries.
In a perfectly competitive labor market, there must exist positive wage differentials for disamenities (Smith, 1976). The literature on hedonic models, initiated by Rosen (1974, 1986), provides a relevant theoretical framework for the analysis of these compensating differentials, suggesting to estimate workers’ MWP with cross-sectional hedonic wage regressions. However, this method has not yielded strong empirical evidence of compensating differentials. Typical estimates in this literature, starting with Thaler and Rosen (1975), are of small order of magnitude, often less than five percent of the wage, if not insignificantly different from zero or wrong-signed.
The presence of labor market frictions can explain why preferences are not reflected in cross-section. If searching for job offers is costly and subject to incomplete information, hedonic prices and workers’ MWP need not coincide. Therefore, low wage/amenity correlations must not be interpreted as reflecting weak preferences for job attributes. Gronberg and Reed (1994) use job duration data in order to estimate workers’ MWP. They find large and significant MWP for two non wage attributes– measuring several aspects of working conditions out of four. Since then, several authors have used their methodology and found MWP estimates of similar orders of magnitude.
To reconcile these two approaches from an empirical perspective, we estimate a partial equilibrium version of the model of Hwang et al. (1998). Time is discrete, and workers belong to a stationary environment. At each period a job offer arrives with some probability, drawn from an exogenous distribution of wages and amenities. Workers accept every offer yielding more utility, where utility is a linear combination of the (log) wage and amenities. The weight of each attribute in the utility function is the worker’s MWP for that amenity. Transitions to employment and non employment also occur with some probability at each period. A difference with Hwang et al. (1998) is that we also allow for shocks that exogenously reallocate workers between jobs. Hence, there are two types of job-to-job mobility in the model: voluntary and constrained.
In this type of on-the-job search models, compensating wage differentials may arise because workers select themselves between jobs and may then trade lower (resp. higher) wages for better (resp. worse) amenities. At the steady-state equilibrium, this produces a cross sectional distribution of wages and amenities among employed workers. Our analysis will focus on the relation that maps individual preferences (MWP for amenities) onto this distribution. The mapping depends on the distribution of job offers, in particular the wage/amenity correlation posted by firms. It also depends on search frictions, measured using an index that counts the average number of outside offers received by employed workers between two adverse shocks (constrained job reallocation or job-to-non employment transition). The lower the search frictions index, the less workers select themselves between jobs, and the greater the difference between wage/amenity correlation in cross-section and workers’ MWP for amenities. The main contribution of this paper is to take the relation between workers’ preferences for amenities and compensating wage differentials in cross-section to the data, with a special emphasis on the role of search frictions.
Download
PDF Ebook The Pervasive Absence of Compensating Differentials
