FRBSF Economic Letter
1997-34 | November 14, 1997
Since early 1996, the U.S. economy has produced a combination of very low unemployment and stable or declining inflation. Many analysts and pundits regard this performance as surprising, because similar sustained low unemployment rates have been associated in the past with rising inflation. Some observers–including Federal Reserve Chairman Alan Greenspan–pointed to worker apprehension regarding the security of their current jobs as a key factor that may have helped to moderate wage and therefore price inflation in 1996.
In recent months, however, some analysts and the media have noted that attitudes toward job security may be shifting, and that this perception of rising job security may lead to rising wage and price inflation. Given the potentially important role of job security in the policy arena and in illuminating how the labor market functions, development of reliable current indicators of the level of job security would be useful. In this Economic Letter, we discuss the data and methods available for this purpose, and we use them to analyze long-run and recent trends in the degree of job security. These data suggest that job security has declined over the past 20 years, but the observable implications of this decline for worker behavior are small and have been eliminated by sustained labor market tightness this year.
Declining job security began to receive attention in the media as early as 1991. Much of the initial evidence was anecdotal, based either on specific, well-publicized instances of downsizing or on survey information regarding workers’ perceptions of their individual job security. These surveys showed that workers’ perceived probability of job loss during 1993-94 was high given the level of labor market tightness, but since then it has come back down to normal levels (Schmidt 1997). Although attitudinal survey data are useful, the responses are subject to perceptual and reporting biases. For example, workers’ reported job fears may be heightened by media coverage of downsizing, even though workers’ actual behavior does not change; such stability in worker behavior would limit the practical labor market implications of the perceived decline in job security. Because factors such as media coverage have an unknown and perhaps changing impact on survey responses over time, attitudinal survey data are of limited use for comparative purposes.
Data on actual job-related behavior and outcomes provide means to avoid the pitfalls associated with attitudinal surveys. One variable used by researchers interested in job security is average job tenure–i.e., the average length of time that workers stay with their current employer. The U.S. Bureau of Labor Statistics (BLS) recently reported that average job tenure for men declined between 1983 and 1996, after controlling for the aging of the male work force by examining average tenure within age groups (U.S. BLS 1997).
The recent decline in male job tenure is consistent with the hypothesis of declining job security. However, job security is best represented by workers’ perceived probability of involuntary job loss. Because average tenure is determined by both voluntary and involuntary turnover, it provides ambiguous information about job security per se. Data that identify or directly reflect the reasons for job change are necessary to make proper inferences about job security.
Another key source of data often cited in discussions of job security is the Displaced Workers Survey (DWS), conducted every two years as a supplement to the BLS’s monthly household survey, the Current Population Survey (CPS). The DWS asks respondents about their job loss experiences during the preceding five years. It has the advantage of directly measuring involuntary permanent job loss, which is the primary outcome on which workers’ job security fears appear to be focused. However, the DWS’s infrequency and lengthy time frame substantially reduce its usefulness as a current indicator of job security.
Other data that distinguish between worker and employer decisions, and that are nearly current, are provided through the CPS itself. The CPS is a primary source of U.S. labor market data, and it provides the basis for official government unemployment statistics. Among the data provided are counts of the unemployed, broken down by the reason for unemployment. These reasons include voluntary separations (quits), temporary layoffs, other involuntary separations, new labor force entrants, and re-entrants. Within each category, the BLS provides counts by duration; this includes individuals unemployed fewer than 5 weeks, which measures the monthly flow into unemployment. We use the quit and involuntary separation flows to measure job turnover decisions by workers and firms.
To analyze trends in job turnover, we formed yearly averages for the share of the total number of jobs each month that end in a “permanent dismissal” or a quit and that lead to a spell of unemployment, for the period January 1977 through August 1997. We define a permanent dismissal as an involuntary job loss for reasons other than temporary layoff. This category includes exits from both permanent and temporary jobs (the latter were separately identified beginning in January 1994). New permanent dismissals and quits measure involuntary and voluntary job turnover, respectively, and as such their cyclical movements and long-run trends offer quantitative evidence on changing job security over time. Our analysis accounts for changes made to the CPS in 1994 which altered survey features relevant to our unemployment tabulations.
Figures 1 and 2 show actual, trend, and predicted values of the permanent dismissal and quit series. The predicted values are based on the statistical relationship each series has with the contemporaneous unemployment rate (a measure of the business cycle) and an underlying trend over time. The dashed trend line corresponds to the predicted series with the monthly cyclical component removed. The figures show that both turnover series are closely tied to the business cycle. In particular, rising unemployment raises permanent dismissals and reduces quits. Furthermore, the figures reveal a close correspondence between the actual series values and the predictions based on the unemployment rate and a time trend.
Figure 1 shows a small (but statistically significant) upward trend in the rate of permanent dismissals. This trend indicates that for any given unemployment rate, the number of jobs that end in permanent dismissal and generate a spell of unemployment has been rising during the past 20 years. The upward trend is about half as large as that reported in Valletta (1996), which used data on the share of permanent dismissals in the total number unemployed. The less pronounced time trend identified here suggests that the earlier results are partially attributable to rising durations of unemployment for permanently dismissed individuals (relative to those unemployed for other reasons).
In contrast, the trend line in Figure 2 shows a pronounced downward trend in the rate of quits into unemployment, conditional on the unemployment rate; this trend is about twice as large as that for permanent dismissals. One explanation for this pattern is consistent with the declining job security hypothesis. As noted above, quits decline as unemployment–a measure of adverse labor market conditions–rises. If workers feel that the trend toward rising permanent dismissals reflects long-run deterioration in labor market conditions, they may be increasingly reluctant to quit jobs and enter unemployment. Alternatively, because the CPS data do not account for job-to-job transitions, the downward trend in quits to unemployment may reflect an improvement in workers’ ability to search for new jobs while employed.
Although the long-run trends in dismissals and quits are relevant, discussion of declining job security largely has focused on workers’ experiences during recent years. In Figures 1 and 2, comparison of the predicted and actual series values during 1994-96 reveals only limited departures from their long-run relationships. In particular, the rates of permanent dismissals and quits in recent years have not diverged substantially from their expected levels based on the contemporaneous unemployment rate and 20-year time trend. We reach a similar conclusion if we predict the quit and dismissal rates for the 1990s based on data through 1989. These findings conflict with the common belief that permanent dismissals in the past few years have been above their expected levels due to corporate downsizing, mergers, and other factors leading to involuntary job loss.
The only noticeable departure from the long-run relationships is a slight downturn in the quit rate during 1996 (Figure 2), which was unexpected given the declining unemployment rate that year. Because layoffs and downsizing received tremendous media attention in 1996, workers’ perceptions of increased dismissal risk, whether justified or not, may have led them to feel more insecure about their jobs. However, sustained labor market tightness during 1996-97 appears to have reversed this downturn in quit rates and general sense of job insecurity. Monthly data for 1997 reveal that since March the quit rate has been steadily increasing and that this rate of increase has been more rapid than the corresponding rate of decline in permanent dismissals. These recent trends suggest that the perceived and actual levels of job security are rising.
The CPS data on the flow into unemployment due to permanent dismissals and quits suggest that job security has been undergoing a long-run decline. However, the impact on observable behavior by workers and firms has been relatively small, and their behavior over the past few years does not appear to represent a substantial departure from or acceleration of the long-run trend.
These results contrast with evidence from other data sources and with the broader perception held by many observers that the decline in job security has been both large and recent. Several factors may help to explain the discrepancy. First, the decline in job stability reflected in job tenure data holds only for men; because the CPS data used here mix information on men and women, we were unable to verify whether the upward trend in permanent dismissals also is more pronounced for men. Second, although the DWS data used in other work reveal large increases in permanent job loss in the mid-1990s, these increases were most pronounced for skilled white-collar workers (Valletta 1997a), who are more likely than other workers to move directly to a new job without an intervening spell of unemployment. This pattern may explain much of the discrepancy between findings based on the DWS and the CPS, since the DWS data include job-to-job transitions but the CPS data exclude them.
Despite the small recent effects of declining job security, the long-run rise in permanent dismissals into unemployment and fall in quits into unemployment suggest ongoing changes in labor market behavior by both firms and workers; this finding has been verified and analyzed with more rigor and depth in Valletta (1997b). Such developments, and their broader implications for firms’ hiring and firing decisions and workers’ career goals, are likely to remain a topic of conversation and analysis in the future.
Schmidt, Stefanie. 1997. “…Despite the Teamsters-UPS Settlement.” Wall Street Journal, p. A12 (August 27).
U.S Bureau of Labor Statistics. 1997. “Employee Tenure in the Mid-1990s.” Department of Labor News Release 97-25 (January 30).
Valletta, Robert G. 1996. “Has Job Security in the U.S. Declined?” FRBSF Economic Letter 96-07 (February 16).
_______. 1997a. “Job Loss during the 1990s.” FRBSF Economic Letter 97-05 (February 21).
_______. 1997b. “Declining Job Security.” Unpublished manuscript, Federal Reserve Bank of San Francisco (May).
Opinions expressed in FRBSF Economic Letter do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System. This publication is edited by Sam Zuckerman and Anita Todd. Permission to reprint must be obtained in writing.
Please send editorial comments and requests for reprint permission to
Attn: Research publications, MS 1140
Federal Reserve Bank of San Francisco
P.O. Box 7702
San Francisco, CA 94120