The Changing Face of Job Loss in the United States, 1981-1995 IN THE FIRST WEEK of January 1996, AT&T announced it was restructuring its operations and reducing its managerial work force by 40,000. This was only the latest in a string of widely publicized large labor force reductions announced by major American corporations. The public perceives that corporations are responding to increased competitive pressure by restructuring and downsizing their work forces, particularly their white-collar

more »

... rk forces, to an unprecedented degree and that the workers so displaced are suffering substantial economic hardship. I In this study I examine evidence from Displaced Workers Surveys (DWSs) from 1984 to 1996 to provide a comprehensive picture of the incidence and consequences of job loss between 1981 and 1995 to determine the extent to which labor force data support these perceptions. Data limitations make it difficult to know what groups of workers lost jobs before the 1980s. The DWSs, which have been regular supplements to the Current Population Survey (CPS) at two-year intervals since 1984, have useful information on job loss, however.2 Specifically, these surveys ask workers if "in the past five years" (past three years in the 1994 and 1996 DWSs) they have "lost or left a job because of a 55 56 Brookings Papers: Microeconomics 1997 plant closing, an employer going out of business, a layoff from which [the worker] was not recalled, or other similar reason." These data have much to say about job loss, and they form the core of my empirical analysis. My earlier paper in this journal used the five DWSs from 1984 through 1992 to examine job loss, and the current paper is a natural extension of that earlier work in two ways.3 First, I bring the earlier analysis up to date by using the two most recent surveys to examine job loss through 1995. Second, I focus on the distinctions regarding the stated reason for (or cause of) the job loss. The three substantively important classifications considered are job loss due to plant closing, slack work, and elimination of a position or shift. Several other, less common, options coded in the survey, including seasonal job ended, self-employment ended, and other, are combined into a fourth category, "other." 4 A specific model of job loss by reason is not provided here but, more generally, the evolution of the incidence and costs of job loss due to these various stated reasons is investigated during the 1981-95 period.5 One factor that makes investigation of job loss by stated reason interesting is that the term "position abolished" resonates with the well-publicized round of corporate downsizing and restructuring of the past several years. A worker's self-report of the reason for job loss is bound to be arbitrary to some degree, however. Workers who lose their jobs because of corporate downsizing or restructuring could conceivably report any of the three reasons noted here. If the employer ceased operations at the site where the worker was employed, then the worker would likely report a plant closing. If the employer reduced employment across all or many workplace functions because of a decline in demand without ceasing operations, then the worker could report either slack work or an abolished position. If the employer "streamlined" operations by reducing employment in certain functions-management, for example-then the worker is likely to report that the position was abolished. 3. Farber (1993). 4. I discuss the coding of the reason for job loss and how the survey design deals with these reasons in more detail later. 5. As I discuss in the next section, an earlier paper by Gibbons and Katz (1991) uses the distinction between plant closing and slack work to test an adverse-selection model. 6. These mobility supplementsinformation on how long workers have held their current jobs. Only the data since 1973 are available in machine-readable form. 58 Brookings Papers. Microeconomics 1997 such a lifetime job.7 Ureta used the January 1978, 1981, and 1983 mobility supplements to recompute retention rates using artificial cohorts rather than contemporaneous retention rates.8 Like Hall, she found that lifetime jobs are an important feature of the U.S. labor market, but she finds smaller differences by sex. Several more recent papers have used CPS data on job tenure to examine changes in employment stability.9 Swinnerton and Wial, using data from 1979 through 1991, analyzed job retention rates computed from artificial cohorts and found a secular decline in job stability in the 1980s.'0 In contrast, Diebold, Neumark, and Polsky, using CPS data on tenure from 1973 through 1991 to compute retention rates for artificial cohorts, found that aggregate retention rates were fairly stable over the 1980s but declined for high school dropouts and for high school graduates relative to college graduates. A direct exchange between Diebold, Neumark, and Polsky and Swinnerton and Wial appears to support the view that job stability did not generally decrease during the 1979-91 period.'2 Farber, using CPS data on job tenure from 1973 through 1993, found that the prevalence of long-term employment had not declined over time, but the distribution of long jobs had shifted. Less educated men were less likely to hold long jobs than they had been previously, a finding that is offset by a substantial increase in the rate at which women hold long jobs. 13 Rose used data from the Panel Study of Income Dynamics (PSID) to measure job stability by examining the fraction of male workers who do not report any job changes in a given time period, typically ten years. The fraction of workers who reported no job changes in a given length of time was higher in the 1970s than in the 1980s. He argued this is evidence of increasing instability of employment."' Jaeger and Stevens used data from the PSID and the CPS mobility and benefit 7. Hall (1982). 8. Ureta (1992). 9. In addition to the January mobility supplements, information on job tenure was collected in pension and benefit supplements to the 59 supplements on (roughly) annual rates of job change to try to reconcile evidence from the CPS and PSID on job stability. They found little evidence in either survey of a trend in job stability, although the estimates from the PSID are rather imprecise. '5 Unfortunately, because of the design of the PSID, neither of these studies examined the mobility experience of women. In a paper in an earlier issue of this journal, I used the five DWSs from 1984 to 1992 to examine changes in the incidence and costs of job loss during the 1982-91 period. 16 I found slightly elevated rates of job loss for older and more educated workers in the slack labor market in the latter part of the period compared with the slack labor market of the earlier part of the period. But job-loss rates for younger and less educated workers were substantially higher than those for older and more educated workers throughout the period. These findings are consistent with the long-standing view that younger and less educated workers bear the brunt of recessions. Gardner carried out the first analysis that incorporated the 1994 DWS. She examined the incidence of job loss from 1981 to 1992. Although she found roughly comparable overall rates of job loss in the 1981-82 and 1991-92 periods, the industrial and occupational mix of job loss changed over this period. Job loss decreased among blue-collar workers and workers in manufacturing industries and increased among white-collar workers and workers in nonmanufacturing industries.'7 A substantial literature uses the DWS to study the postdisplacement employment and earnings experience of displaced workers. 18 This work demonstrates that displaced workers suffer substantial periods of unemployment and that earnings on jobs held after displacement are substantially lower than predisplacement earnings. In my earlier paper in this journal, I found no difference on average in the consequences of job loss between the 1982-83 recession and the 1990-91 recession.'9 The earnings loss suffered by displaced workers is positively related to tenure on the predisplacement job. Kletzer finds further that the 15. Jaeger and Stevens 60 Brookings Papers: Microeconomics 1997 postdisplacement earnings level is positively related to predisplacement tenure, suggesting that workers displaced from long jobs are more able on average than those displaced from shorter jobs.20 In recent work Neal, using the DWS, and Parent, using the National Longitudinal Survey of Youth, found that workers who find new employment in the same industry from which they were displaced earn more than industry switchers do.2' This new work suggests that Kletzer's finding, that postdisplacement earnings are positively related to predisplacement tenure, results from the transferability of industry-specific capital. Workers who are reemployed in the same industry "earn a return" on their previous tenure; those reemployed in a different industry do not. Gibbons and Katz take a different approach, analyzing the consequences of job loss in the context of an adverse selection model. Specifically, they argue that workers displaced because of "slack work" are subject to selection on the part of their employer. Within the limits of human resource management policies that give preference in retention to high tenure workers, employers are likely to lay off less productive workers when demand declines. In contrast, workers displaced because of a "plant closing" are not subject to such selection. Employers must lay off all workers in such situations. On this basis, Gibbons and Katz argue that workers displaced because of slack work will fare worse after displacement than workers displaced because of a plant closing. They present evidence from the 1984 and 1986 DWSs consistent with this adverse selection model.22 Each DWS from 1984 to 1992 asks workers if they were displaced from a job at any time in the preceding five-year period. The 1994 and 1996 DWSs ask workers if they were displaced from a job at any time in the preceding threeyear period. Displacement is defined in the interviewer instructions as 20. Kletzer (1989). 21. Neal (1995), Parent (1995). 22. Gibbons and Katz (1991). 23. U.S. Department of Commerce, 1988, Section II, p. 4. 24. This is consistent with work by Jacobsen, LaLonde, and Sullivan (1993), who find that displaced workers suffer wage declines even before they are displaced. 62 Brookings Papers: Microeconomics 1997 not rare) for workers to have lost more than one job in a five-year (or three-year) period, the DWS cannot be used to measure the total quantity of job loss. At best, it measures the number of workers who have lost at least one job in the relevant time period.25 Defining the Rate of Job Loss Even if all agree that the focus of the analysis should be on those workers who have lost at least one job, the problem remains of how to compute the job-loss rate. Consider some category of workers (defined by such characteristics as age, sex, and education). The DWS directly measures the number of workers in that category who have lost at least one job, which is a reasonable numerator for the category-specific jobloss rate. But the pool of workers who were at risk to lose a job during the relevant time period is not easily measurable. I take the straightforward approach, as I did in my earlier study,26 of using the number of workers in the given category employed at the survey date as the measure of the relevant pool, and this number serves as the denominator in the calculation of the job-loss rate. This is likely to be a good approximation unless employment in the group is changing rapidly over the relevant time period (three years). All job-loss rates presented in the next section are computed on this employment basis. Later I carry out multivariate analyses of the probability of job loss using disaggregated data. The sample consists of all workers employed at the survey date and all workers who reported a job loss in the relevant period (whether employed or not). This results in estimates analogous to the employment-based job-loss rates I present for specific groups. Changes in Survey Design and Comparability: The Adjusted Job-Loss Rate A second issue is the relevant time period over which to compute the rate of job loss. I use three-year rates of job loss, which are computed as the number of workers who report having lost a job in the three 25. There also is the commonly noted problem of recall bias because workers fail to report job loss that occurred long before the interview date. See Topel (1990) for evidence suggesting that recall bias is an important problem in the DWS. Farber (1993) also presents some evidence on this issue. 26. Farber (1993). Figure 3. Rate of Job Loss by Age and Reason A.