by Julia Lane"
Turnover is job change — workers changing firms and firms shedding and hiring workers. Turnover is sometimes seen as an indicator of the dynamism of the economy, since without turnover, labor cannot get reallocated from less-productive to more-productive uses. Indeed, voluntary job change usually results in gains to the worker. Involuntary job loss also imposes costs on workers, however, particularly on low-wage workers who are least likely to be able to bear the cost of being without work.
In the subsequent sections, we discuss the impact on turnover on low-wage workers in more detail. We begin by describing turnover in more detail, examining why it occurs, and then discuss its pervasiveness. We then describe the consequences of turnover, particularly in the low-wage labor market, and evaluate potential policy recommendations.
What Is Turnover?
Turnover is the result of both quits and layoffs. Thus, some turnover is a result of jobs in one firm being destroyed and jobs in another firm being created — and hence due to the reallocation of jobs across the economy in response to changes in product demand. A majority of job changes, however, are because workers reshuffle across the same set of jobs, and this worker reallocation occurs over and above job reallocation. There are enormous amounts of both job and worker reallocation in the U.S. — the annual job reallocation rate is about 20 percent; the quarterly worker reallocation rate exceeds 40 percent.(1) In other words, almost one job in five is destroyed or created every year, while out of every 10 jobs, four will be occupied by new people within a quarter.
Although most jobs are short, most people are in long-term jobs. The picture of enormous job change must be tempered by the realization that while most job spells end relatively quickly, most people are in long-lasting jobs. Thus, although almost one in four new jobs ends within a quarter,(2) and roughly 20 percent of workers have been on the job for less than one year,(3) average tenure for men in 1991 was almost 8 years.(4)
Low-wage workers have shorter job tenure and a greater number of job spells than other workers. The impact of this on well-being is clear — the difference between poor and non-poor low-wage workers is that poor workers are employed for roughly the same number of hours per weeks, but 20 percent fewer weeks per year.(5) Thus, low wages, combined with frequent spells without a job, suffice to push workers below the poverty threshold.
The causes and consequences of turnover are complex. Clearly, turnover can be seen as a joint employer/worker decision: If the change is initiated by the employer, it is called a layoff; if by the worker, it is called a quit. Thus, understanding the causes requires understanding the decision-making process of both the employer and the employee. Describing the consequences is equally tricky, since turnover initiated by the worker is often likely to result in her getting a better job while that initiated by the firm may have adverse consequences.
Less-educated workers are less likely to jump jobs and more likely to be pushed out. One study in particular, has pointed out that this group of workers are more likely to have new jobs than the most educated (22.7 percent of high school dropouts were in jobs with less than one year's tenure in 1996 compared with 16.6 percent of college) and are also less likely to be offered health insurance when they change jobs (15.4 percent are offered coverage in new jobs, compared with 45.2 percent of college graduates), and this likelihood has declined dramatically in the past 15 years.(6)
Why Does Turnover Occur?
Turnover is a result of both worker and firm decisions. Thus, the answer depends on separating out not only turnover driven by job reallocation and by "churning" (worker turnover over and above job reallocation) but also, within the churning component, understanding the contribution made by the worker and by the firm.
Job Reallocation and Turnover
The job reallocation part of turnover is a result of job creation and destruction across firms and industries. This is clearly part of the market reallocation process — indeed, job creation and destruction rates have been used by the Organization for Economic Cooperation and Development (OECD) as an index of the flexibility of the labor market.
While job reallocation may be necessary in an efficient economy, this does not mean that there are no underlying costs. In particular, older, long-tenured individuals who are displaced suffer quite large and permanent earnings losses. One study estimates that earnings on new jobs are 25 percent lower even five years after job loss, and that the present discounted value of these losses can be as high as $80,000.(7) Furthermore, workers who have been displaced often suffer multiple job losses in the years after being displaced, with all the concomitant negative effects.(8) The dual labor market literature suggests that shocks to product demand are borne by a buffer, secondary labor market, comprised primarily of less-educated, less-tenured minority workers, who are easier to shed and rehire than workers in the primary labor market. The employment ratio for nonwhite men is almost three times as responsive to the prime-age male unemployment rate as for white men, and teenagers and young women bear 50 percent of the cyclical variation in employment.(9)
Turnover as a Firm Strategy
Why do firms churn workers — why do some firms explicitly have high turnover policies? Firms make different management decisions in setting an employment contract because there are different costs associated with hiring and firing workers. The hiring costs include advertising, screening, and training; the firing costs include work disruption, loss of the worker's firm-specific knowledge, and severance benefits. These costs clearly vary greatly depending on the type of worker and the nature of the production process. Consequently, we would expect to see quite different levels of turnover across firms, industries, and types of workers.
Adjustment costs are high if production processes are complex. Thus, turnover in retail trade is higher than in manufacturing — and turnover should increase economy wide as the economy moves from manufacturing and toward services. Similarly, if the task is complex and difficult to monitor, it may make economic sense to pay a higher efficiency wage to get the worker to work harder — which has the additional effect of reducing turnover. Firms may also offer implicit contracts to those workers who are averse to wage variability in order to guard against economic downturns. Wages may be set high and turnover lower for "insider" workers who possess valuable amounts of firm-specific human capital. These factors all have consequences for workers — turnover should be higher for those workers performing simple, easily monitored tasks, junior workers who have little firm-specific human capital, and younger workers who are less averse to wage variation.
Different firms have different production processes and different turnover rates. Even within the retail trade industry, a coffee franchise like Starbucks has explicitly chosen a relatively high-wage, high-benefit, low- turnover strategy to sell its coffee, while other shops will produce a different type of product with a different personnel strategy.(10)
Turnover as a Worker Choice
Why might workers have different turnover rates? Different types of workers may also have different levels of attachment to the job, and hence have different quit rates. Young workers are more likely to search for new jobs than old; unmarried workers are less attached than married; on average, women are less attached than men. Clearly, if job search is the motive, we would expect quits to be cyclical — higher in a boom period, where there are more jobs available, and lower in a recession.
Turnover as a Joint Worker-Firm Decision
Turnover can also be seen as a joint decision. The discussion above has essentially dichotomized the worker and firm decisions. If one thinks of the job match process as being a little like a marriage, undertaken by two consenting parties, then the separation process can be seen as a divorce initiated by one side as a result of the dissatisfaction with the returns from the match. If this is the view of the world, then many simple matches will dissolve quickly — turnover will be high — because it will be evident to both sides that it is not working. It is also possible that workers in low-wage jobs are less likely to be equal partners in the marriage, and are therefore subject to more decisions decided by the more powerful partner, the employer.(11) For very complex matches, such as baseball managers, lawyers, or accountants, it may take a long time for this to become evident.(12) Again, the end result is that workers who are less educated, or who engage in tasks that are less complex are more likely to have high turnover and have short-tenure jobs.
Extent of Turnover
What is the level of turnover in the economy? Just as there is both a worker and a firm side to the source of turnover, there are two ways of determining the extent of turnover. The first is to ask, through worker-based surveys, how many jobs a worker has had within a given time period. The second is to use employer-based surveys or administrative records to determine how many workers have left or been hired, again, within a given time period. Researchers using the former approach, with Survey of Income and Program Participation (SIPP) data, estimate that the average monthly turnover rate in the U.S. economy in 1991 was 7.1 percent.(13) Using the latter, other analysts estimate that quarterly turnover in the early 1980s was 23 percent.(14), (15) Turnover thus depends on the time period within which it is measured, as well as the unit of analysis. It is worth noting that the source of turnover also depends on the business cycle — downturns in the economy, while not affecting overall turnover, will increase the involuntary component of turnover by increasing layoffs and reducing quits.(16)
Turnover also varies by industry, skill, sex, and age —the average number reported in the previous paragraph hides a great deal of variation. For low-wage workers, the news is almost all bad. Low-skilled and young workers experience more turnover than older, more highly educated workers. Workers in complex jobs, such as in manufacturing, are likely to experience lower turnover and higher wages than workers in retail sales, who tend to be low-wage workers. New and small firms, which are also more likely to pay lower wages, should have higher turnover; new job matches, paying entry-level wages, should dissolve faster than long ones. The consequences of this are well illustrated by Burtless (elsewhere in this volume) who notes that although over 1.2 million cashier jobs will be available every year until 2006, fewer than 1 out of 6 of these represent new job openings — the rest is turnover. Alternatively, of 11 low-skill occupations with 6.5 million job openings per year, only 1 million are new jobs.
Turnover by Industry
Table 1 demonstrates that turnover, however measured, varies dramatically by industry. It is highest in the construction and retail trade industries; lowest in manufacturing, finance insurance and real estate, and in public administration. Two high-turnover sectors, retail trade and professional services, while accounting for 1 in 5 jobs, account for almost half of all worker-based turnover. It is also worth noting that turnover rates are more than twice as high for small and new employers as for large, older employers.(17)
These employer differences have important implications for the employment of low-wage workers. Clearly, high-turnover firms are more likely to have job openings than low-turnover firms, but, just as clearly, employment with these firms is less likely to last and less likely to be one in which wages grow. For example, the new and small firms with high turnover are also likely to have less capital and train their workers less, since this reduces hiring and firing costs and makes turnover less costly. It is also possible that small firms have less efficient personnel screening mechanisms — which is what leads to a greater number of bad hires, and, consequently, greater turnover — but this also may mean that there are fewer or poorer skills acquired on the job. The high-turnover industries are also those where the production tasks are less complex, so on average the skills acquired in retail trade, for example, are less valuable than those acquired in manufacturing or finance, insurance and real estate.
In sum, the high-turnover industries are also those most likely to hire low-wage workers. Some researchers point out that forced exiters from welfare are twice as likely to get jobs in the retail trade sector than are other workers;(18) similarly 28 percent of jobs offered to welfare recipients were in the highest-turnover industry (business services).(19)
|Turnover Actions||Proportion of Turnover
|Quarterly Turnover Rate from UI data*
|Agriculture, Forestry, and Fishing||1,698||245||3.26||14.43||47.96|
|Transportation, Communication, and Utilities||7,463||350||4.66||4.69||17.43|
|Finance, Insurance, Real Estate||6,621||387||5.16||5.85||14.88|
|Business and Related Services||5,589||712||9.49||12.74||21.83|
|Entertainment and Recreation||1,241||218||2.90||17.57|
|Professional and Related Services||25,441||1,431||19.07||5.62|
|Source: Ryscavage (1997), table E, p. 5 and author's calculations
* Anderson and Meyer (1994).
Turnover by Worker Characteristics
Turning to the worker side of the story, table 2 demonstrates that turnover is much higher for young than for old workers. It should be noted that this is not necessarily a negative for this group: many writers have documented that mobility among younger workers leads to higher earnings and greater earnings growth. In fact, the average length of time spent looking for work after a job change is only 1.7 months for young women; 2.1 months for young men.(20) However, adult men with a high school education change jobs almost 40 percent more often than do college-educated males. This is of particular concern, both because this group takes almost twice as long to find a new job as does youth and because Farber (1997a) shows that prior mobility is a good predictor of the probability of leaving a new job.(21)
Thus, those workers who have moved a lot continue to move a lot, and if there are intervening spells of joblessness, those workers have fewer weeks worked in the year. There are consequences other than lost earnings capacity. In 1996 almost one in four workers with less than high school education was in a job lasting less than a year, compared with only one in six college-educated workers.(22) This is important, because predicted job tenure is an important determinant of the training decision and training is very closely related to wage growth. Workers who are likely to leave the job for nonemployment are 17 percent less likely to receive training.(23)
|Both sexes ages 16-24||19,366||15.8|
|Men ages 25-54||39,892||4.9|
|Women ages 25-54||37,172||5.8|
|Source: Ryscavage, table F, p. 5|
Consequences of Turnover
What are the consequences of turnover for the firm, for the worker, and for the economy? They can be huge. Job loss has been valued at $76 billion at its peak in 1991 — by estimating the value of lost earnings for the worker and multiplying by the number of workers experiencing job loss — although, as noted, some of this loss may be efficient for the firm and the overall economy.(24)
Effect on Firms
The discussion above suggests that firms choose the optimal amount of job turnover, and consequently that there should be no impact on firm survival. Empirical work, however, suggests that employers with high turnover rates are less likely to survive — possibly because the entrepreneur was not omniscient.(25) Quite apart from the survival issue, however, the consequences of a firm's decision to have high turnover are that it is less likely to invest in human capital and training, there will be less worker-to-worker transfer of firm-specific knowledge, and it is less likely to offer fringe benefits, including health insurance.
Effect on Workers
On the worker side, for some workers — particularly young men — voluntary job mobility increases earnings and earnings growth. Job changing can account for one-third of the increase in real wages in the first 10 years in the labor market.(26) However, the effects of job change are fundamentally different for less-skilled workers — affecting work time, skills, wage levels, wage growth, and fringe benefits.
Unskilled workers suffer longer subsequent spells of unemployment after a job change and consequently lower annual earnings. The lost work time is compounded by lost skills, since long spells of unemployment, even for skilled workers, lead to a depreciation of skills pushing previously high-wage workers into the low-wage, unskilled category.(27) If the source of the turnover is job loss, workers have lower employment probabilities, higher probabilities of part-time work and lower earnings, and that these costs are higher for the least-educated workers.(28) Earnings losses from displacement are quite large and persistent;(29) estimates range from 10 to 25 percent several years after the displacement. The effect on the growth of wage, as well as wage levels can be important, too. Another effect of higher turnover probability is reduced training, which results in flatter earnings.(30) The effect of being laid off can also stigmatize a worker as a "lemon."(31) Finally, there is some evidence that new jobs have fewer benefits, such as health insurance and other fringe benefits.(32)
Three Roles for Policy Makers
High and pervasive turnover is a characteristic of the U.S. labor market. The discussions above have indicated that while turnover may be an efficient way of reallocating workers from one part of the economy to another, or from one job to another, less-educated workers are disproportionately affected, and they may not be well placed to bear the costs. Certain types of firms, particularly new and small firms, also have disproportionately high turnover, possibly because of inefficient management, and this may affect their survival probabilities. This suggests three roles for policy: one aimed at changing worker characteristics, another at changing job characteristics, and the third at job placement strategies.
The Role of Education
A major role for policy makers is to improve training, both prior to and during employment. The results of every study point directly to the importance of education — high school dropouts are much more likely to be churned through the labor market than college graduates. Since it is clearly not feasible to turn all dropouts into college graduates, it may be more useful to look directly at what employers want. This would reduce turnover by making the employees more valuable and more costly to fire. A study using matched surveys of employers and employees suggests that workers want employees to be able to read, do arithmetic, deal with customers, possess motivation, and be polite.(33)
Classroom-based training programs have had mixed success, but there is some evidence in both the U.S. and Germany that job-based training programs help reduce turnover and teach basic skills.(34) Although it is difficult to identify the direction of causality between turnover and training, low-turnover firms have a much higher incidence of training, particularly formal training, than do high-turnover firms.(35) The average annual expenditure on training could be as high as $1,433 per worker; clearly expenditures at this level would provide a strong disincentive to shed workers in economic downturns.
The Role of the Firm and Job Characteristics
There is also a role for policy makers in changing job characteristics. Some suggest that U.S. policy has a "clear pro-layoff bias" and recommend policy to encourage companies to reduce hours, rely on attrition, and look at other alternatives to layoff.(36) In particular, they recommend expanding the use of short-time compensation programs, encourage work sharing, provide incentives for workplace training, and adjust the unemployment insurance tax schemes in different states.
The use of short-time compensation and work-time sharing schemes are possibly the most interesting of these ideas. The essential concept is relatively clear. If a firm needs to lay off the equivalent of 100 full-time workers, it could either shed 100 workers, forcing them to bear the full cost of layoff, or it could reduce the work time of 200 workers by half. The advantage of the latter situation is both that the cost of job loss is not concentrated on just a few workers, and that workers have the time to look for another job if the downturn in demand is temporary.
How can policy makers encourage such programs? Abraham and Houseman suggest that there are three major impediments to encouraging such work sharing plans in the U.S. The first is the structure of the unemployment insurance system, which effectively subsidizes some of the cost of job loss. The second is the fact that employers are likely to continue to pay fringe benefits to workers on short-time compensation schemes, which is extremely costly. The third is that there is a great deal of red tape associated with instituting a short-term compensation scheme, which effectively discourages employer participation.
The role of placement services
Policy makers can also use placement services to help reduce turnover — by using administrative data to identify low-turnover employers. Despite the fact that some note that state employment offices are viewed with some suspicion by employers, who regard referees as "lemons" and prefer to rely on word of mouth,(37) placement services can provide a valuable information function in either helping workers retain jobs or choose jobs that have a higher probability of retention. Clearly, high-turnover employers are the ones most likely to be hiring workers, everything else equal, and these are exactly the kind of employers who are likely to churn workers back out into nonemployment. If placement services can identify these firms, there are several options. Program counselors can warn workers about the firm's track record (if permitted by law); they can investigate the source of such high-turnover strategies; they can counsel workers about how to use these jobs as steps to better jobs; or they can avoid placing workers in these types of firms. Identification is feasible at the state level, by means of administrative unemployment insurance record data. Indeed, this was the focus of a Joint Center for Poverty Research conference, "Evaluating State Policy: The Effective Use of Administrative Data," which was held at Northwestern University in June 16-17, 1997.
This approach can be effective. Welfare recipients who are placed in low-turnover firms are much less likely to leave those firms and return to welfare.(38) They also develop in some detail ways to identify firms that offer welfare recipients jobs with "successful" outcomes — jobs where workers stay for at least four quarters and do not return to welfare. Placement of welfare recipients in some industries (notably, health services and professional services) increases the probability of the worker still being employed in the subsequent year, in contrast to those placed in retail trade.(39) The characteristics of the job, particularly the wage and industry, matter.
Abraham, Katharine, and Susan Houseman. 1992. Job Security in America: Lessons from Germany. Washington, D.C.: The Brookings Institution.
Anderson, Patricia, and Bruce Meyer. 1994. "The Extent and Consequences of Job Turnover." Brookings Papers on Economic Activity 177-248.
Arthur, Jeffrey B. 1994. "Effects of Human Resource Systems on Manufacturing Performance and Turnover." Academy of Management Journal 37 (3): 670-87.
Bartik, Timothy. 1997. "Short-Term Employment Persistence for Welfare Recipients: The 'Effects' of Wages, Industry, Occupation, and Firm Size." W.E. Upjohn Institute Working Paper 97-46.
Campbell, C. 1997. "The Determinants of Dismissals, Quits, and Layoffs: A Multinomial Logit Approach." Southern Economic Journal 63 (4): 1066-73.
Davis, S., and J. Haltiwanger. 1998. "Gross Job Flows." In Handbook of Labor Economics, Orley Ashenfelter and David Card, ed., forthcoming.
Fallick, Bruce. 1996. "A Review of the Recent Empirical Literature on Displaced Workers." Industrial and Labor Relations Review 50 (1): 5-16.
Farber, Henry. 1994. "The Analysis of Interfirm Work Mobility." Journal of Labor Economics 12 (4, August): 554-593.
Farber, Henry. 1997a. "The Changing Face of Job Loss in the United States, 1981-1995." Brookings Papers on Economic Activity, 55-142.
Farber, Henry. 1997b. "Job Creation in the United States: Good Jobs or Bad." Princeton University. Mimeo.
Farber, Henry. 1998. "Mobility and Stability: The Dynamics of Job Change in Labor Markets." In Handbook of Labor Economics. Forthcoming.
Frazis, Harley, Maury Gittleman, Michael Horrigan, and Mary Joyce. 1998. "Results from the 1995 Survey of Employer Provided Training." Monthly Labor Review 121 (6): 3-13.
Gibbons, Robert, and Lawrence Katz. 1991. "Layoffs and Lemons." Journal of Labor Economics 9 (4): 351-80.
Hall, Robert. 1995. "Lost Jobs." Brookings Papers on Economic Activity 1: 221-274.
Holzer, Harry. 1996. What Employers Want: Job Prospects for Less-Educated Workers. New York: Russell Sage Foundation.
Huff Stevens, Ann. 1997. "Persistent Effects of Job Displacement: The Importance of Multiple Job Losses." Journal of Labor Economics 15 (1): 165-88.
Jacobson, Louis, Robert LaLonde, and Daniel G. Sullivan. 1993. "Earnings Losses of Displaced Workers."
American Economic Review 83 (4): 685-709.
Jacobson, Louis, Robert LaLonde, and Daniel G. Sullivan. 1997. "The Returns to Classroom Training for
Dislocated Workers." Federal Reserve Bank of Chicago Working Paper WP-97-16.
Kremer, Michael. 1993. "The O-Ring Theory of Economic Development." Quarterly Journal of Economics 108 (3): 551-76.
Lane, Julia, Alan Isaac, and David Stevens. 1996. "Firm Heterogeneity and Worker Turnover." Review of
Industrial Organization 11 (3): 275-291.
Lane, Julia, David Stevens, and Simon Burgess. 1996. "Worker and Job Flows." Economics Letters 51: 109-13.
Lane, Julia, and Michael Parkin. 1998. "Turnover in an Accounting Firm." Journal of Labor Economics 16 (4): 702-716.
Lane, Julia, and David Stevens. 1995. "Family, Work and Welfare History: Work and Welfare Outcomes."
American Economic Review 85 (2): 266-70.
Lane, Julia, and David Stevens. 1997. "Welfare-to-Work Policy: Employer Hiring and Retention of Former Welfare Recipients." Northwestern/University of Chicago Joint Center for Poverty Research Working Paper. October #3.
Lawson, Leslie, and Christopher King. 1997. "The Reality of Welfare-to-Work: Employment Opportunities for Women Affected by Welfare Time Limits in Texas." University of Texas at Austin, Center for the Study of Human Resources. Mimeo.
Lindbeck, Assar, and Dean Snower. 1988. The Insider-Outsider Theory of Employment and Unemployment. Cambridge, Mass.: MIT Press.
Long, Sharon, and Alberto Martini. 1991. "Wages and Employment among the Working Poor: New Evidence from SIPP." No. 119, U.S. Department of Commerce, Bureau of the Census. Washington, D.C.
Lynch, Lisa, 1991. "The Role of Off-the-Job vs. On-the-Job Training for the Mobility of Women Workers." American Economic Review 81 (2): 151-57.
Lynch, Lisa, and Sandra Black. 1998. "Beyond the Incidence of Employer-Provided Training" Industrial and Labor Relations Review 52 (1): 64-81.
Nardone, Thomas, Jonathan Veum, and Julie Yates. 1997. "Measuring Job Security." Monthly Labor Review 120 (6): 26-33.
Royalty, Anne. 1996. "The Effects of Job Turnover on the Training of Men and Women." Industrial and Labor Relations Review 49 (3): 506-21.
Ryscavage, Paul. 1997. "Dynamics of Economic Well-Being: Labor Force and Income, 1991-1993." Current Population Reports, Household Economic Studies, P70. Bureau of the Census.
Summers, Lawrence., and Kim Clark. 1981. "Demographic Differences in Cyclical Employment Variation." Journal of Human Resources 16: 61-79.
Topel, Robert. 1993. "What Have We Learned from Empirical Studies of Unemployment and Turnover." American Economic Review 83 (2): 110-15.
Topel, Robert, and Michael Ward. 1992. "Job Mobility and the Careers of Young Men." Quarterly Journal of Economics 107: 441-79.
Valletta, Robert. 1997. "Declining Job Security." Federal Reserve Bank of San Francisco. Mimeo.
1. For a survey of job flows, see Davis and Haltiwanger (1998); the worker reallocation calculation is an average derived from different state level data (e.g., Anderson and Meyer 1994, Lane et al. 1996).
2. Anderson and Meyer (1994).
3. Farber (1998).
4. Valletta (1997).
5. Long and Martini (1991).
6. Farber (1998).
7. Jacobson et al. (1992).
8. Huff Stevens (1997).
9. Summers and Clark (1981) .
10. Kremer (1993).
11. Lindbeck and Snower (1988).
12. Lane and Parkin (1998).
13. Ryscavage (1997).
14. Anderson and Meyer (1994).
15. Lane, Stevens, and Burgess (1996) find a similar level for Maryland in the 1990s.
16. Campbell (1997).
17. Lane, Stevens, and Burgess (1996).
18. Lawson and King (1997).
19. Lane and Stevens (1997).
20. Ryscavage (1997).
21. Farber (1997a).
22. Farber (1997b).
23. Royalty (1996).
24. Hall (1995).
25. Lane, Isaac, and Stevens (1996).
26. Topel and Ward (1992).
27. Topel (1993).
28. Farber (1997).
29. Fallick (1996).
30. Royalty (1996).
31. Gibbons and Katz (1991).
32. Farber (1997b), p. 119.
33. Holzer (1996).
34. Lynch (1991); Lynch and Black (1998).
35. Frazis et al. (1998).
36. Abraham and Houseman (1992), p. 132.
37. Holzer (1996).
38. Lane and Stevens (1995); Lane and Stevens (1997).
39. Bartik (1997).