Recently, many labor market observers have asserted that the link between workers and employers has become more tenuous and short term. The reasons given for this transformation usually relate to the need to remain competitive in an age of changing technology and growing international trade or to accommodate a growing desire for flexibility among workers. The evidence cited in support of such trends includes well-publicized corporate restructuring and the rise in employment in certain sectors of the economy, such as business services and, especially, the temporary help industry.

One of the more intriguing recent business practices has been the use of contingent workers (see the Contingent and Alternative Arrangements Special Report). Though time-series data are not available, it is believed that the use of such workers has increased significantly over time. It is also believed that the direction of most business organizations is toward a smaller core of permanent workers, with contingent workers making up the balance during peak times. Intuitively, the net result of these practices should be a measurable loss of job stability and worker tenure.

Job Tenure. Data from the Current Population Survey (CPS) show that the length of time workers have been with their current employers has changed very little in recent years. Among wage and salary workers, median tenure—the point at which half the workers had been with their employer longer and half shorter—was 4.1 years in January 1991. This was essentially the same as in January of 1987 (3.9 years) and 1983 (4.0 years). Although the overall median remained essentially the same, changes occurred for some demographic groups. In particular, median tenure declined among men in age groups above 40, and this perhaps has contributed to the public perception that job stability has declined in recent years. In contrast to men, however, median tenure generally rose for women above age 40 and changed little for women in most other age groups (see Figure 12).

Within educational groups, median tenure declined among men and women with less than a high school diploma, perhaps as a result of rising skill requirements for more stable jobs. Among men age 25 and older, the median tenure of school dropouts fell from 7.3 years in 1983 to 5.5 years in 1991. The median tenure for female dropouts fell from 5.4 to 4.1 years during the same span (see Figure 13) . Median tenure also fell for men with a high school education but no college training, although the median remained essentially unchanged for women in this educational category. Among women with at least some college experience, median tenure increased slightly.

Given the employment shifts since the 1960s away from goods-producing industries and toward service-producing industries, it is also useful to analyze tenure data from an industry perspective. Median tenure changed very little during the 1983-91 period in most of the major industries (see Figure 14) . Two exceptions were construction, which showed a moderate increase between 1987 and 1991, and mining, which experienced a sharp rise between 1983 and 1987.

The median for construction, which was 2.9 years in 1983 and 2.8 years in 1987, reached 3.4 years in 1991. The tenure distribution for the construction industry showed that, in both 1983 and 1987, 40 percent of workers had been with their current employer 1 year or less. In 1991, however, only a third of construction workers had 1 year or less tenure. The reduced building activity during the 1990-91 recession undoubtedly contributed to the decline in the proportion of short-tenured construction workers, and this decline resulted in an increase in the median.

The median for mining rose sharply, from 3.9 years in 1983 to 6.7 years in 1987, before receding slightly to 6.3 years in 1991. In 1983, 41 percent of workers in mining had 2 years of tenure or less. This large proportion of short-tenured workers stemmed from the substantial hiring in the oil and gas extraction industry, mining's largest component, during the oil boom of the late 1970s and early 1980s. By 1987, the oil boom had ended, mining employment had fallen considerably, and many less-senior workers had lost their jobs. In fact, the proportion of workers with 2 years or less tenure had fallen to 23 percent. In the meantime, many of the workers who were newly hired at the time of the 1983 survey had, by 1987, achieved 5 or 6 years of tenure. As a result of these developments, median tenure in mining rose sharply from 1983 to 1987.

Other than the increases in construction and mining, no meaningful changes in median tenure occurred in any of the other major industry groups during the 1983-91 period. This is, perhaps, a surprising result, given the job reductions that have taken place in such industries as manufacturing and the rapid employment growth in services and retail trade. Another way to identify changes in job stability is to examine the tenure distribution of workers. For example, examining the proportion of workers who had been with their current employer 10 years or longer is one way to assess whether the prevalence of long-term employment relationships has changed. Here again, the evidence is somewhat mixed. Overall, the proportion of workers age 25 or older with 10 years or more of tenure was 32 percent in both 1983 and 1991 (see Figure 15) . Among men in the 25 to 29 year and 30 to 34 year age groups, the proportion who had been with their current employer 10 years or longer rose from 1983 to 1991. In contrast, men in each of the older age groups exhibited declining proportions with such long tenure. Among adult women, the proportion with tenure of 10 years or longer rose during the 1983 to 91 period in nearly every age group. This reflects the increasing attachment of women to the labor force and their reduced likelihood of interrupting their careers for child rearing or other reasons.

It should be recognized that there are limitations in using tenure data as a gauge of employment stability. For example, median tenure paradoxically could rise during recessions, when many jobs are less secure, because less-senior workers are more likely to be dismissed by their employers than are those with longer tenure. Likewise, median tenure could fall during periods of economic growth, when job opportunities are more abundant and workers can easily transfer their skills from one organization to another. Of course, high median tenure also could result from stable economic conditions and good worker-employer matches.

Examining tenure distributions may not always prove to be informative either. Take the case of the manufacturing industry. As noted earlier, median tenure in manufacturing changed little during the 1983-91 period, even though manufacturing had undergone considerable job reductions in that span. These employment declines might be consistent with stable median tenure, however. For example, if the cutbacks had been achieved by permanently dismissing less-senior workers and offering early-retirement incentives to long-tenured workers, median tenure may not have changed at all, although the proportion of short- and long-tenured workers would have declined. In fact, that did not happen. As the following percentage distributions of manufacturing tenure show, little change occurred over the 1983-91 span (see Figure 16) .

Clearly, this indicates that changes in employment stability, whether increases or decreases, are not always easily discernible from CPS tenure data. In a 1995 paper, Henry S. Farber recognized the difficulty of interpreting changes in tenure over the business cycle and introduced a method to control for those changes. Examining CPS tenure data from 1973-93, Farber concluded that overall job stability was unchanged in the two decades studied. However, Farber did find that, for less-educated men, the probability of being in a long job (defined alternatively as a job lasting at least 10 years or at least 20 years) had deteriorated. In contrast, women with at least a high school education experienced a greater likelihood of holding a long job.

Retention Rates. The BLS tenure statistics only examine current tenure, that is, the length of time a worker already has been with his or her employer. The data say nothing about how long a worker eventually will be with his or her employer. What concerns many observers is that workers today may be less likely to achieve such long eventual tenure than their counterparts in decades past. Unfortunately, there is no direct measure of eventual tenure. Instead, estimates of eventual tenure must be inferred from the data available on current tenure.

Several researchers have used CPS data to estimate retention rates and the distribution of eventual tenure. Robert E. Hall used tenure data from the January 1978 CPS to estimate contemporaneous retention rates. Hall found that long-term employment is an important characteristic of the U.S. labor market. He noted that, while any particular new job is unlikely to last a long time, one that has lasted 5 years has a high probability of lasting 20 years. He found that more than a quarter of all workers are in jobs that eventually will last 20 years or longer, and among workers age 30 or older, 40 percent are in jobs that will last 20 years or longer.

Manuelita Ureta used historical retention rates to estimate eventual tenure. In Hall's analysis, women were considerably less likely than men to have long tenure, and blacks were at least as likely as whites to have long tenure. Using historical retention rates, Ureta's calculation showed that women indeed were less likely than men to achieve long tenure, though this gap was not as large as Hall's contemporaneous retention rates indicated. Ureta's analysis also indicated that blacks were less likely than whites to achieve long tenure.

Other researchers have analyzed historical retention rates to examine whether employment relationships have become more or less stable over time. A 1995 study by Kenneth A. Swinnerton and Howard Wial examined retention rates for three 4-year periods: 1979-83, 1983-87, and 1987-91. Swinnerton and Wial found that retention rates were higher in the 1983-87 period than in the earlier span. In the 1987-91 period, retention rates had declined from 1983-87 and, in fact, were lower than in 1979-83.

Turnover. Turnover is another measure that can be used to examine employment stability. A study by the National Commission for Employment Policy (NCEP) examined changes in job turnover by using data from the University of Michigan Survey Research Center's Panel Study of Income Dynamics (PSID). The NCEP study compared the employer stability of prime age individuals in the 1970s and 1980s. A person of "prime age" was defined to be 24 to 48 years old at the beginning of each decade (that is, 1970 and 1980) and thus ages 34 to 58 at the end of each decade (1979 and 1989). Workers were assigned to one of the following three categories of employer stability:

• Strong stability—individuals with 0 or 1 year of employer change during a decade
• Medium stability—individuals with 2 or 3 years of employer change
• Weak stability—individuals with 4 years or more of employer change

	1970s	1980s
Strong stability	67	52
Medium stablity	21	24
Weak stability	12	24

Early career changes and job market churning. Researchers have estimated that, on average, workers will have 8 to 10 jobs in their working lives. However, the pacing and timing of an individuals' job changes are far from even; the rate of job change declines with age and labor market experience. Data from the National Longitudinal Survey of Youth (NLSY) indicate that the average individual will have had 7.5 jobs between his or her 18th and 30th birthdays.

Estimates from the Longitudinal Employee-Employer Data (LEED) file indicate that the labor market attachment for young white men was extremely fragile. Two-thirds of young men's new jobs lasted less than a year, and only 44 percent of the young men spent less than one year (cumulative) without a job during the first 10 years after they entered the labor market full time.

Frequent job changes and labor market transitions by the young increase the turnover rates and contribute substantially to the low median tenure observed for the population as a whole. In January 1991, workers younger than 35 years old constituted 68.4 percent of those with a year or less tenure on the current job, 66.3 percent of those with 2 years or less tenure, and 61.6 percent of those with 5 years or less tenure.

Trends in part-year work. Yet another way to look at employment stability is to examine trends in the proportion of workers who worked only part of the year. A part-year worker is defined here as someone who worked at least 1 week but no more than 26 weeks during a given year. This measure can be viewed as the percentage of workers with a weaker attachment to the labor market.