• Negative Consequences of Early Career Job Turnover
• Positive Aspects of Early Career Job Turnover
• The Role of Values and Assumptions
• Previous Empirical Research
• Plan of the Report

Although the vast majority of our young people leave high school to go directly to work, we typically offer them little or no assistance in this transition. . . . The result is that typical high school graduates mill about in the labor market, moving from one dead-end job to another until the age of 23 or 24.

--Commission on the Skills of the American Workforce, 1990, p. 46

In the debate about U.S. competitiveness, the U.S. educational system is frequently accused of preparing students poorly for the school-to-work transition. This poor preparation has two components: the content of the education the students receive, and (given their skills) the assistance the educational system provides to the students in finding jobs (Commission on the Skills of the American Workforce [CSAW], 1990; Rosenbaum et al., 1990; General Accounting Office, 1991; Prewo, 1993; Osterman and Iannozzi, 1993).

Non-college-bound youth in the United States are described as drifting from activity to activity until their mid-twenties, when (it is hoped) they settle into long-term commitments to full-time jobs. During the time between leaving school and finding primary-sector jobs, young people are perceived as spending a long period of unproductive time in school, in dead-end jobs, unemployed, or not even looking for work, with a "consequent loss of training and productivity" (Rosenbaum et al., 1990). For example, Osterman and Iannozzi (1993, p. 4) write that

[t]he early years in the labor market for many graduating students are characterized not by an absence of jobs but rather by a "churning" process. High turnover and frequent job change are evident during this period when youth sample different jobs or simply move from one low-skill job to another. The phenomenon of churning represents a characteristic of the youth labor market that has important implications for program design. . . . What happens when the period of churning has concluded? Evidence suggests that a substantial fraction of this cohort has been unable to "settle down" into quality jobs. In the past, most youth in their late twenties--even if they did not attend college--could expect eventually to obtain stable employment; this is no longer true . . . as many as 50 percent of high school youth had not found a steady job by the time they reached their late twenties.

This characterization implies that the transition period is spent unproductively. An alternative perspective characterizes this period as productive "job shopping" (Johnson, 1978; McCall, 1990): In the individual-choice-oriented U.S. society, young people try out various jobs until they find something amenable to their tastes and abilities (see also Meyer and Wise, 1982; Manski and Wise, 1983; Topel and Ward, 1992). Perhaps in other, less individual-oriented societies, the worker adjusts to the job rather than vice versa.

In this section, we draw on the human capital and job-matching literatures to provide theoretical perspectives for our empirical work on the early labor market career. We begin by considering the arguments that job turnover in the early career has negative consequences. The counter arguments that we summarize next view early career job turnover in a more positive light. We end this discussion by noting that differing values and underlying assumptions may play an important role in the divergence among the theoretical results of economists, assessments by policymakers, and the observed behavior of U.S. youth. A brief review of previous empirical research on the school-to-work transition concludes the section.

Negative Consequences of Early Career Job Turnover

The perspective that early career job turnover has negative consequences for youth is based on three concerns: those leaving jobs are at risk of being unemployed; those leaving jobs forfeit accumulated firm-specific human capital; and high turnover discourages firms from providing training. These concerns are discussed in turn.

The Risk of No Employment

Some of the concern about job turnover is prompted by the simple fact that, when jobs end, young people are at risk of being unemployed or out of the labor force. Such non-employment is intrinsically wasteful (disregarding the value of the leisure to the young person); potential output is sacrificed (Slichter, 1919, as quoted in Topel, 1991). Furthermore, inasmuch as workers build up skills on the job that can be used on other jobs (sometimes referred to as "general training"; Becker, 1964), no such skills are accumulated during subsequent periods spent looking for work or out of the labor force. Existing skills may even decay (i.e., be lost or forgotten) when youth are between jobs (Mincer and Polachek, 1974; Sandell and Shapiro, 1980). By itself, this argument implies that the crucial issue is not job continuity but being employed. This view is consistent with the general concerns in the literature about the fraction of youth at any given time who are unemployed or out of the labor force (Freeman and Wise, 1982; Rees, 1986).

The link between turnover and non-employment, however, may not be so strong. Based on a model of Burdett (1978), Parsons (1991) develops a model of on-the-job search behavior. In Parson's framework, it is often optimal to search for a new job without leaving the current job. Using the National Longitudinal Survey--Youth (NLS-Y) data for 1980 (which we also use below), Parsons (1991) found that half of all job quitters had already arranged their next job. Mattilla (1974) quoted similar statistics for the late 1960s.

The Loss of Firm-Specific Human Capital

A second concern about job turnover stems from the relationship between wages and time worked for a specific employer (often referred to as job tenure). If workers' wages rise because they stay longer on a particular job--perhaps because they accumulate human capital (e.g., they learn a firm's procedures or move up a job ladder) or perhaps because the firm can find a job that better uses the person's skills--then when the worker leaves the firm, he or she forfeits those higher wages and, assuming the higher wage was due to higher productivity, society loses output.

At face value, the empirical evidence is strongly supportive of the proposition that wages rise with tenure. Higher wages for more labor market experience, and an incremental premium for time with the current employer for workers with the same overall labor experience, are among the most robust findings in labor economics. However, there is controversy about whether the increased wages with greater tenure are evidence of a causal relationship (Abraham and Farber, 1987; Altonji and Shakoto, 1987; Topel, 1991). The alternative explanation is that the observed increase in earnings with tenure is merely a result of selection: If the people who left an employer had stayed, their wages would have been lower. Those who remain with an employer are increasingly a select group who are more productive with the given employer.

To consider the selection argument further, imagine a world with two types of workers: those who are highly productive and highly likely to stay with an employer (perhaps because they stayed in school longer and acquired more skills), and those who are less productive and less likely to stay on the job. If the more productive people earn more than the less productive people, and the more productive people stay two periods with an employer and the less productive people stay only one period, then the average earnings of all people in their first period of tenure with the firm is a weighted average of the wages of the high- and low-productivity people. Only the high-productivity people are left in the second period, so the wage in the second period of tenure is equal to the wage of the high-productivity people. Average wages of people with two periods of tenure are higher than those for people with one period of tenure. The rising wages with tenure are, however, not the result of tenure per se. If the job leavers had stayed, their wages would not have risen (see Topel, 1991, for more on this argument).

A job-matching story yields a similar type of selectivity. Suppose all workers are ex ante identical, but ex post some workers find that they can do a particular job better than others. Then at the end of the first period, people who find they cannot do the work well will leave to try a different job. Again, only the workers best suited for a particular job are left in the second period, and, by virtue of their higher productivity, they earn higher wages than those who leave the employer. Again, wages appear to rise with job tenure; and, again, if the job leavers had stayed, their wages would not have risen. Note that this second example does not require the assumption of the first--that more-productive workers are also more likely to stay longer (by their own choice).

Topel (1991) considers these selectivity arguments in detail and concludes that most of the returns to job tenure are causal. Despite these selection arguments, the longer a randomly chosen worker would stay on a job (on average), the higher would be his/her wages. This finding is consistent with declining turnover rates as workers gain job tenure. The more job tenure there is, the larger is the increment to firm-specific capital forfeited when workers leave a job.

Perhaps the most intuitively appealing evidence for this conclusion is the plant-closing literature. When a plant closes, workers lose their jobs through no fault of their own. If the selection model was correct, these displaced workers would be able to quickly find new jobs at approximately the same wage they earned on their previous job. In fact, even after controlling for the fact that plant closings tend to be serially and spatially correlated, it appears that displaced workers experience large wage losses, suggesting the importance of firm-specific capital or job matching (Topel, 1991; Carrington, 1993; Jacobson, LaLonde, and Sullivan, 1993).

If additional tenure generates positive wage gains, then when young people move from job to job, they forfeit their accumulated tenure (and its accumulated wage premium). This argument, however, is unconvincing in one respect: The young person leaving a job will bear the cost of the higher wages he/she forgoes from the loss of job tenure with the job change. If the worker is able to gauge the magnitude of those forgone earnings, he/she will leave a job only when the benefits of leaving (presumably an even higher wage on the next job) outweigh the costs (including forfeiting accumulated firm-specific capital). One possible explanation for early career mobility (which seems to have received little empirical study) is that job leavers are in jobs that do not provide significant opportunities for earnings growth with tenure. This is essentially the argument of dual labor market theorists (Doeringer and Piore, 1971).

Disincentives for Firms to Provide Training

The third concern about early job mobility is that the incentives for firms to invest in young workers may be weakened. In contrast to the preceding argument, which implies that skill accumulation (e.g., moving up the career ladder) is costless, other models suggest that workers and firms must jointly make costly investments in order to develop the skills that yield higher wages. For example, workers may be sent for formal schooling or paired with more experienced employees for informal training. In this context, job tenure is crucial. Like any investment, such training has a cost that the "investor" expects to recover in the future. Becker's (1964) theory of human capital divides such investments into two types: (1) general human capital, which will be productive (and thus yield higher wages) for other employers as well as for this employer, and (2) specific human capital, which will be productive only for this employer.

The work site may be the most efficient place to provide job training. That it is is a basic tenet of apprenticeships, co-ops, tech-prep, and other programs in vocational education (Berrryman and Bailey, 1992): By receiving training on the job, students see the relevance of their classroom studies for the world of work, increasing their interest and motivation. The use of academic skills in the work setting reinforces the skills learned.

To the extent that such programs provide only general training, however, firms have little incentive to provide it. The firm bears the costs of such training programs up front, whereas the payback to the firm comes later in the form of higher productivity. If the employer could reasonably expect the trained worker to stay with the firm (e.g., under a norm of lifetime employment), the firm could pay the worker less than the value of his/her output for a period of time after the training until the up-front costs of training are recovered (through the difference between the value of output produced by the worker and the reduced wage). But if there is high turnover, as soon as the worker receives the general training, he/she can command higher pay elsewhere and will leave for a higher-paying job with another employer or will demand higher wages on the current job. Thus, the firm would not recover the costs of the general training (Lynch, 1993).

Long apprenticeship programs provide one way around this problem. Young workers receive a low training wage even after they have received considerable training. The difference between the value of their output in the later part of the apprenticeship and their low training wage is used to pay back the firm for the cost of the training. However, in the absence of formal binding apprenticeships (or of the threat of withholding a certificate until completion of the full apprenticeship period), firms will not recover their investment and thus will not provide the training. Without such binding apprenticeships, everyone is worse off: The skill level of the workforce, the earnings of workers, and the economy's productivity suffer (Lynch, 1993).

The argument is more subtle in the case of firm-specific training. Again, employers have little incentive to provide specific human capital unless workers will stay with the firm long enough to recoup their investment (over a period in which the worker's wage will be lower than his/her marginal product). In the case of specific training, the value of the worker to other firms is no higher following the training investment, so the training itself does not give the worker an incentive to leave the firm. Nevertheless, early turnover will result in the loss of the firm's investment if the worker leaves.

With firm-specific training, only workers who are expected to stay on the job for a long enough time will be trained by employers. As long as young people are perceived (correctly or incorrectly) to have high turnover, they will not be provided with training. To some extent, firms can counteract the turnover problem by offering higher wages to workers with specific training. Since another employer will not benefit from the specific training and will only pay the untrained wage (assuming that the training is completely firm-specific and has no value to another firm), workers then have an incentive to stay with their current employer (Lynch, 1993).

Alternatively, a firm could adopt contracting solutions, requiring workers to pay, at the time of the training, for the firm-specific training they receive. The firm would then pay higher wages once they are trained. The initial payment for training could take the form of a payment from the worker to the employer at the start of the training; or in the early years of the job, the firm could pay lower wages than the worker could demand in a firm not offering training. However, the low productivity and small asset holdings of young workers make such contracting solutions difficult to implement. Low wages for younger workers mean that firms cannot lower wages much further because of the minimum wage. Small asset holdings mean that young workers have difficulty posting large bonds or funding the training themselves.

Compared with the two other negative consequences of turnover, this turnover-and-job-training relationship has consequences beyond the individual worker's own well-being. Although job leaving may be optimal from the individual worker's perspective, such turnover may create the perception that all youth have weak job attachment--a perception that will make firms less likely to provide training for young workers in the future. This potential externality (a cost imposed by this worker's actions on other workers) suggests a role for policies that provide new incentives or institutional mechanisms that overcome the barriers to investing in the skills of young workers and encourage young workers to stay longer with a given job (Lynch, 1993).

One attempt to encourage increased skills investment is proposals to formally certify job skills in particular areas. Job skill certification programs are motivated by the need for young workers to assure potential employers that they have acquired a necessary set of skills. Once a worker is certified, the worker will have a costless way to convince a new employer that he/she has the skills: Show the certificate. Thus, the training received becomes more "general" (rather than "specific"). Not only is the training inherently valuable to other employers, but now those employers can easily learn that a particular employee has the skills. By raising productivity both with the firm that provided the certification training and with other employers, the newly trained worker will be more likely to leave the current employer. This possibility will make employers less likely to offer the training, since they will not necessarily recover their training costs. Such certification might be good for other reasons--for example, it might encourage workers to get more training or make it easier for them to find appropriate jobs, thereby raising overall economic output--but it is likely to have the effect of reducing the amount of firm-provided training (Lynch, 1993).

This concern is consistent with the current institutional arrangement in the United States, whereby most training leading to a publicly recognized credential is paid for by the individual (e.g., at a technical school or other post-secondary institution). Firms are more likely to pay for training (on the job or otherwise) when that training does not lead to a credential. Formal apprenticeships, whereby workers pay for their training in the form of a training wage below the market wage (before they are certified), offer one way out of this dilemma. But such programs are rare in the United States.

Positive Aspects of Early Career Job Turnover

The labor economics literature contains several arguments that early career job turnover might be good. These arguments stem from a view of the early labor market career as an inherently dynamic process leading to an optimal match between the worker and a job.

A simple job-shopping model yields high levels of turnover at young ages, leading to a period of settling down as the worker ages (Burdett, 1978). In this model, workers begin their career qualified for a variety of jobs. Initially, they take the first job that becomes available but continue searching for a better job match. As new jobs open up, the worker compares subsequent job offers with the present job, taking the job offer when it is better (i.e., pays more) than the current job. As more job offers are received and some are accepted, the worker changes jobs at a decreasing rate, increasing his/her wage with each job change.

This simple job-matching model can be generalized to the case in which people as well as jobs vary (Jovanovic, 1979; Mincer and Jovanovic, 1981; Flinn, 1986). According to this formulation, different individuals have different productivities on the same job. While the worker may have some information about his/her productivity in a particular job, complete knowledge about the quality of the match between the worker and the job requires taking the job and gaining experience. This type of job shopping, then, involves taking a sequence of jobs; subsequent moves follow periods of learning about one's abilities. Job turnover leads to better and better job matches.

In addition to representing productive job shopping, employer changes may be an optimal career path. In the extreme case, if each firm had only a single job type, then, as workers matured and gained skills, they would have to change jobs to exploit their new skills. Obviously, firms offer more than one type of job, but it still follows that some career paths may require switching employers to attain the next job in the sequence (Sicherman and Galor, 1990).

The Role of Values and Assumptions

In our discussion of the potential negative and positive consequences of early job turnover, we have relied on two assumptions, both implicit in the current discussions of youth labor market experience. The first assumption, which is implicit in most discussions at the policy level, is that young people should either be working or in school; little value is placed on their leisure time. One interpretation of the observed high rates of non-employment among youth is that, in fact, young people place a high value on their leisure--it is worth more than the wages they are offered.

From this perspective, for example, the early work patterns of non-college-bound youth can be rationalized as an effort to equalize leisure with that of their counterparts in college (Nolfi et al., 1986; Mare and Winship, 1986). The college-bound peers of non-college-bound youth spend four years in an environment with a long summer vacation, several other vacations during the year, and a relatively flexible weekly schedule. The intermittent employment pattern of non-college-bound youth allows them to reproduce the leisure pattern of their peers at college. In general, if young adults value their leisure more than do policymakers, then two of the negative consequences of job turnover cited above (more non-employment and lower wages) carry less weight in terms of young adults' own assessment of their well-being.

In a similar vein, the second assumption, which is inherent in the labor economics literature (following standard economic assumptions), is that the young worker's objective is to maximize income (with some, albeit low, discounting of future earnings). If young people value leisure highly (or apply a higher discount rate on future earnings), then the assumptions implicit in standard labor economics models are incorrect. The valuation of leisure is particularly salient because several recent model-based analyses of youth labor markets were estimated only on the subsample of continuously employed young workers (e.g., Topel, 1991). By eliminating youth who make a transition out of the labor market, analysts may be missing much of the interesting and important variation in behavior. There is the potential for additional insight by reformulating the economic models of young workers' careers to explicitly account for the fact that a sizable subset of individuals may value leisure highly (at least relative to work). If so, the income-maximization assumption of such labor economic models is wrong and the interpretation based on them--that the transition is smooth--may need to be reexamined.

Previous Empirical Research

Despite the stylized facts cited in the passage at the beginning of this section and the various explanations of those stylized facts, the published literature on the transition from school to work provides contradictory characterizations of actual behavior. Rosenbaum et al. (1990, p. 264) represent the failure perspective. They begin their paper as follows:

The transition from high school to work is a serious problem. Many high school graduates spend their first years after school unemployed or job hopping, with consequent loss of training and productivity.

Other analysts characterize the transition as proceeding smoothly (Meyer and Wise, 1982; Topel and Ward, 1992). For example, Manski and Wise (1983, p. 44), who analyze the employment experiences of male high school graduates, conclude that

[t]hese graduates by and large seem to have made a rather smooth transition to the labor force and to subsequent schooling without substantial periods out of school and without work.

A more complex picture is provided by Osterman (1980, p. 16), who contrasts an initial moratorium, "a period in which adventure seeking, sex, and peer group activities are all more important than work," with a subsequent period of "settling down." His analysis of the NLS-Young Men and personal interviews suggests that the period of settling down begins for most youths within a few years of leaving school; some youths, however, fail to make the transition. More recently, Osterman and Iannozzi (1993, p. 6) explicitly link the empirical facts of churning or milling about to program design:

For the bulk of youth not bound for college, the problem that public policy must address is not the simple absence of jobs but rather the difficulties these youth face in settling down into quality jobs in the adult labor market--a problem that has been exacerbated by rising skill requirements. If we accept a period of churning as part of the process, many of the ideas regarding improved information systems between schools and employers seem less compelling.

Before specifying policy approaches to address negative aspects of the early labor market career, we need to empirically examine the extent of milling about for the median youth and whether such churning is experienced by all youth. The empirical facts and their interpretation are relevant for policymakers and educators who seek to design programs to improve the transition from school to work. In the next subsection, we discuss our approach to addressing these empirical issues for a recent cohort of U.S. youth entering the labor market.

Plan of the Report

Our analysis of the early labor market transition for U.S. youth begins with a description of our main data source, the National Longitudinal Survey--Youth, and our methods. Section 3 presents the empirical results of the school-to-work transition in a static framework--labor force status at a given age; Section 4 presents results based on a dynamic perspective--number of jobs held and job duration. Section 5 uses time series of cross-sectional data from two supplements to the Current Population Survey (CPS) to put the narrow cohort data of the NLS-Y into historical context. The report concludes with a summary of the results, in Section 6.