Long-term attachments and long-run firm rates of return.

• Author: Orazem, Peter F.
• Position: Labor force relationship with firms

Long-term attachments between workers and firms are common. Numerous studies have examined worker returns to tenure, but little is known of firm returns to firm-worker matches. Yet these attachments represent a human capital asset quasi-held by the firm, which is not captured by traditional accounting measures of firm assets. Firms with large quasi-holdings of human capital will have higher measured return on assets, other things equal. Analysis of data on 250 large manufacturing firms supports the view that firms profit from long-term attachments with their workers. Consequently, unmeasured human capital assets contribute to the explanation of persistence in measured long-run excess profits across firms.

JEL Classification: J31, J24, M4

1. Introduction

   One of the stylized facts of the U.S. labor market is that long-term attachments between firms and workers are common. Hall (1982) first noted the prevalence of long-term attachments in the labor force. Despite reports in the popular press of declining firm and worker loyalty to one another, longterm attachments have remained common. Farber (1999) reported that, in 1996, 35% of workers aged 35-64 had over 10 years of job tenure and 21% of those aged 45-64 had over 20 years of job tenure. Evidence presented by Diebold, Neumark, and Polsky (1997), Gottschalk and Moffitt (1999), and Neumark, Polsky, and Hansen (1999) suggests that job stability in the 1990s is similar to that in the 1970s and 1980s. Explanations for the prevalence of long-term attachments have concentrated on the role of firm-specific human capital (Becker 1964; Parsons 1972) and the quality of the match between the firm and the worker (Jovanovic 1979). In either case, the ongoing relationship between the firm and the worker generates increased productivity or lower production costs that would not persist if the worker were to switch to another firm. Because workers would be expected to receive some of the rents from the long-term attachment between firm and worker, the theories predict upward-sloping wage profiles with job tenure. (1)

   Tests of the theories have concentrated on estimating worker wage returns to tenure. Some studies (e.g., Abraham and Farber 1987; Altonji and Shakotko 1987) found negligible effects on wage growth, while others (Antel 1985; McLaughlin 1991; Topel 1991; Parent 1999; and Abowd and Kang 2002) find more substantial returns. Studies that have attempted to measure firm-specific training directly (Brown 1989; Parent 1999) found significant positive wage responses. Another line of research examined the impact of permanent layoffs on subsequent wages. These studies (Hamermesh 1987; Addison and Portugal 1989; Topel 1991; Jacobson, LaLonde, and Sullivan 1993) have found substantial wage losses associated with exogenous disruptions of the firm-worker match, although some of the loss may be industry specific and not firm specific (Neal 1995).

   Although the models explaining long-term attachments between the firm and the worker typically suggest that the firm and the worker share in the rents from the match, few studies have explored the link between long-term attachments and firm profitability. The most common approach has been to examine the link between firm-provided training and measures of firm performance. Barrel's (2000) review reports that, of the five cross-sectional studies that examined whether training can explain variation in measures of firm output, two found no effect and three found positive effects. A sixth study, by Ichniowski, Shaw, and Prennushi (1997), found mixed effects of individual training measures on productivity. However, these studies do not directly address the issue of firm returns to long-term attachments. First, firm returns are typically measured in terms of output rather than profit. This makes it difficult to distinguish firm returns from worker returns because training would raise labor productivity even if workers were getting all the returns to training. (2) Second, training may not capture variation in long-term worker attachment to firms. Training is typically measured by expenditures over one year or less. Consequently, the stock of human capital is not measured, but rather the increment to the stock. In addition, labor turnover may be influenced by many factors other than measured training including benefits packages, compensating differentials, and informal training. While training may be an indicator of long-term attachments, it is not necessarily the only indicator.

   Previous studies have used the persistence of sub- or supernormal firm profit over time as an indicator of firm monopoly power or union power to extract rents. However, atypical long-term worker attachments can also help to explain the persistence of measured excess supernormal profit across firms. Traditional accounting measures of firm assets do not include firm holdings of human capital because the firm cannot force workers to stay with the firm. However, the fact that specific human capital investment or productive firm-worker matches lead to long-term voluntary attachments between the firm and the worker, even in the face of business cycle fluctuations, implies that firms may have a quasi-hold on the specific human capital embodied in its workers. For such firms, traditional accounting measures of profit rates, such as firm revenue per unit of physical capital, will overstate the firm's rate of return because the denominator will only include measures of physical capital and not human capital.

   This study shows that variation in firm quasi-holdings of human capital does explain some of the variation in long-run measured firm profit rates. We exploit the theoretical link between firm's share of human capital investments and quit propensity to derive an observable estimate of a firm's quasi-holdings of specific human capital. (3) The model is tested using information on average profits for 250 manufacturing firms over the 1973-1983 period. The sample period was selected to coincide with a data series on industry turnover, which was discontinued in 1981, and to utilize information from the 1982 Census of Manufacturers, Results support the view that firms get returns from long-term attachments to their workers. (4)

   The article first examines a simple model of firm turnover as it relates to firm returns to long-term attachments. Next, an empirical strategy for testing the theory is proposed. The results of the analysis are reported in the final section.

2. Theory

   We need to establish an observable measure of firm quasi-holdings of specific human capital. To do this, we employ a model in the spirit of Becker (1964), Parsons (1972), Hashimoto (1981), and Antel (1985), which describes how workers and the firm share in the returns to the employment relationship. Let H be the expected value of production from a worker's general human capital and let h be the expected value of production from a worker's human capital specific to the firm. (5) The worker's opportunity wage (equal to the maximum of the value of the general training used in other firms or in nonmarket activities) is H, and the worker's gross value to this firm, V, is H + h. (6)

   It is generally accepted that the worker and the firm will share in the investment in specific human capital. By so doing, each party suffers a loss if the employee separates from the firm, whether by quit or layoff. However, the optimum amount of specific training, h, is not sensitive to the firm or worker shares. Both the firm and the worker will have an incentive to invest in specific training until the marginal return on additional h is equal to the return on other physical or human capital investments. However, as shown below, the level of quits or layoffs will be sensitive to the share of costs and returns.

   Let [beta] be the firm's share of training costs and returns to specific training. The worker's share of costs and returns is (1 - [beta]). The worker's wage in the firm will be

   (1) W = H + (1 - [beta])h.

   The firm's net return on worker human capital will be the worker's gross value to the firm less the wage, or

   (2) V - W = H + h (H+(1 - [beta])h)= [beta]h.

   Therefore, firms only make excess returns on the firm-specific human capital, h, and not H.

   With no uncertainty about the value of the worker inside or outside the firm, the employment relation can go on indefinitely. The existence of separations that were not planned from the beginning implies that the ex post realized net value of the job to the worker or the firm must differ from the ex ante expected value. Quits occur when the worker's productivity outside the firm rises sufficiently so that wages elsewhere turn out to be higher than the wage in the firm. Layoffs occur when the worker's value in the firm falls to a level such that the worker's wage exceeds the worker's value.

   To make these statements precise, define the value of the specific human capital as (h + [phi]), where [phi] has mean zero and density f([phi]). Thus, h may be viewed as the expected value of specific human capital in the firm and h + [phi] is the ex post realized value. Productivity shocks can occur in other firms as well. Let the opportunity wage in other firms be H + [theta], where [theta] has a zero mean and density g([theta]). As above, H may be viewed as the anticipated wage outside the firm, and H + [theta] is the ex post realized opportunity wage.

   Hashimoto (1981) argued that the wage may remain fixed as defined in Equation 1, even when the shocks occur. Renegotiation may be very costly. In addition, asymmetries in information in which firms know [theta] and workers know [theta] may make it difficult for the two parties to change the wage conditional on [theta] and [phi].

   Quit Condition

   Assuming wages remain fixed as in Equation 1, a worker will quit the firm when the opportunity wage rises above the worker's contracted wage in the firm. This occurs when

   (3) H + [theta] > W = H + (1...