Cyclical quality adjustment in the labor market.

• Author: Devereux, Paul J.

1. Introduction

   The U.S. unemployment rate fell to very low levels during the 1990s. Even more impressively, the unemployment rate for minorities and less educated individuals fell disproportionately. Convergence of relative unemployment rates in expansions and divergence in recessions has been a feature of the U.S. economy over the last 30 years. There are many possible reasons for this feature. For example, if the costs of hiring and firing less-skilled labor are low, one would expect these individuals to have very procyclical employment. In this paper, I explore an alternative model: when there is an excess supply of labor, high-skilled people take jobs that would normally be occupied by less-skilled people. In expansions, the process reverses, with less-skilled workers getting access to jobs that they would normally not attain. I refer to this process as cyclical quality adjustment.

   Reder (1955, p. 834) was one of the first to discuss how hiring standards adjust to business cycle conditions:

   Quality variations in labor markets arise through upgrading and downgrading of members of the labor force relative to the jobs they are to fill. When applicants become scarce, employers tend to lower the minimum standards upon which they insist as a condition for hiring a worker to fill a particular job--and vice versa when applicants become plentiful. Since then, many papers have developed this idea theoretically and many economic commentators appear to share this view of the labor market. For example, the Economic Report of the President states that "A strong labor market is particularly important to less advantaged groups in the labor market, such as workers with less education, younger workers, racial and ethnic minorities, and immigrants.... When employers find it hard to fill vacancies, they are more willing to hire and train workers whom they might pass over when they have fewer openings and an abundance of applicants." (Council of Economic Advisors 1999, p. 103). Likewise, the expansion during the 1990s spawned many newspaper articles about the improvement in the types of jobs available to less-skilled applicants. However, there is little systematic empirical work about changes in job assignment in the United States over the business cycle. The empirical work in this paper seeks to redress this situation.

   Cyclical assignment changes are important for many reasons. As mentioned earlier, job upgrading is a potential explanation for the greater cyclical variation of employment experienced by less-skilled workers. It is well known that the unemployment rates of less-skilled individuals and minorities are more countercyclical relative to more skilled individuals (Clark and Summers 1981; Kydland 1984; Keane and Prasad 1993; Hoynes 1999). The standard explanation for this effect is that low-ability workers are more likely to be laid off in downturns because training and hiring costs are lower for these workers. No commensurate attention has been paid to the issue of how the decision of who to hire changes over the business cycle. However, assignment models present a natural explanation for the greater procyclicality of employment of the less skilled. Both the adjustment cost hypothesis and the upgrading hypothesis imply that less-skilled workers will have more procyclical employment. Empirically, they are distinguishable because the upgrading explanation implies that jobs obtained by comparable workers should vary systematically over the cycle. The adjustment cost hypothesis relates to the composition of those whose jobs are terminated.

   Second, these mechanisms suggest a reason why wages are particularly procyclical for job changers. Recent studies using panel data have indicated that the wages of job starters are much more procyclical than the wages of workers who do not change jobs. (1) Researchers have suggested various explanations for this result. Beaudry and DiNardo (1991) and MacLeod and Malcomson (1993) present contracting models in which the wages of job starters depend on the state of the labor market when they join the firm. Subsequent to joining there is limited adjustment in wages, so the wages of stayers are less procyclical than the wages of changers. Barlevy (2001) presents a model whereby firms create temporary jobs in booms and pay a compensating differential to workers for the unemployment risk. Thus, the wages of job changers are procyclical for this reason. The quality adjustment hypothesis is a natural alternative to these explanations for the wage procyclicality of job changers.

   A closely related hypothesis is that job switchers have very procyclical wages because high-wage industries tend to have the most procyclical employment and workers achieve wage gains from moving to these industries in booms (Vroman 1977; Okun 1981; McLaughlin and Bils 2001). Industry is defined by the output of the firm, and the skill requirements of jobs in the same industry may differ enormously. Also, the evidence suggests that the wage gains result from moving to industries that pay rents rather than from moving to jobs that have greater skill requirements. Thus, it is not clear that employers are choosing different types of workers for particular jobs, and so this evidence does not speak to the fundamental question that I address. For these reasons, I define jobs by both industry and occupation in this paper.

   Third, there are important policy implications. If employers respond to excess supply of labor by increasing hiring standards, then low-skill workers will disproportionally experience unemployment in recessions. If government wants to help these individuals acquire jobs, it does not need to stimulate the bottom end of the labor market. Instead increasing labor demand at the middle or at the skilled end of the market will cause fewer skilled workers to take unskilled jobs and hence increase the employment of unskilled workers. On the other hand, if hiring standards remain unchanged, increasing the demand for highly skilled workers will increase their wages but have no effect on the employment rates or wages of the unskilled.

   Previous Empirical Work

   There is a substantial European empirical literature on this topic. Using Dutch data, Teulings (1993) and van Ours and Ridder (1995) have found evidence that workers get less attractive jobs in recessions. Teulings (1993) shows that the transition rates from unemployment to employment are more cyclical for low-skill workers. (2) However, given the greater flexibility of the U.S. economy compared to European ones, it is not clear how much these results generalize to the U.S. economy. (3) Thus, a separate investigation for the United States is required.

   There is little systematic U.S. empirical work on cyclical assignment changes. Bowlus (1995) shows that matches formed in expansions last longer than matches formed in recessions. Also, there are some papers about position changes over the business cycle within matches. Researchers have examined the hypothesis that employers move workers between positions over the cycle in order to adjust their wages procyclically. Solon, Whatley, and Stevens (1997) found that promotions increased wage cyclicality during the great depression era. Analysis using recent data by Wilson (1996) and Devereux (2000) finds no such support for this hypothesis. However, Devereux (2000) does show that, within matches, workers are engaged in tasks that require more skill in expansions. I add to this literature by examining whether there are changes in assignment across matches as well as within matches.

   Interviews conducted by Bewley (1999) suggest that employers faced overqualified job applicants during the deep recession in the Northeast in the early 1990s. Most employers in the primary sector expressed a reluctance to hire overqualified workers since they would be likely to leave once the economy improved. However, employers in the secondary sector, where turnover is high in any case, were more open to hiring overqualified workers. My research complements this qualitative research.

   In section 2, I present a model describing the search behavior of the unemployed. This provides a framework for analyzing and interpreting the empirical results. Following this, in section 3, I describe the data set, and in section 4 I describe the empirical analysis designed to test the implications of the model of cyclical quality adjustment. Section 5 concludes.

2. Modeling the Process of Quality Adjustment

   The theoretical literature on cyclical quality adjustment is dominated by theories of wage rigidity. A lone exception is Mortensen (1970), who uses a more neoclassical framework. Reynolds (1951), Reder (1955), Hall (1974), Thurow (1975), Okun (1981), Ohashi (1987), and Teulings (1995) have all presented fixed wage analyses of quality adjustment. Because reducing starting wages is a costly option for employers, they instead increase hiring and promotion standards in times of high unemployment. Hence, employers assign better workers than usual to particular jobs during recessions and searching workers attain lower quality jobs than they would normally attain. Recent evidence has challenged the idea that wages are generally sticky. (4) Therefore, I develop a flexible wage search model that formally generates quality adjustment in new hires over the business cycle. (5) In this model wages equal marginal product and so employers have no incentive to choose the best available workers. I use this model to guide the empirical work.

   Consider an economy that contains two distinct submarkets and a group of unemployed workers of varying skill levels (s). Each unemployed worker can search in only one submarket and may choose not to search at all. There is a fixed per period cost of search in that searchers lose the value of leisure or home production. The submarkets are ordered by increasing level of skill in that submarket 2 has a minimum skill requirement of...