Labor economists like numbers. With datasets containing tens of thousands or hundreds of thousands of observations readily available on the NBER website (here)and from numerous other sources, and with increasingly powerful computers to crunch the numbers, there is no shortage of numbers to examine economic issues large or small.
When I began this review, I intended to write about the major topics and ideas that have emerged in the four years since I last examined how the labor program was doing ( NBER Reporter Winter 1997/8). But I quickly realized that my labor colleagues would squirm in their chairs at any exclusively qualitative review of ideas. The more aggressive labor economists would demand "Where are the numbers?" To assuage this potential critical audience, I examined the data in the NBER labor studies archives and listings of NBER papers in labor studies and did some statistical calculations.
Like the rest of the U.S. economy, the Labor Studies program expanded in the late 1990s to early 2000s. From 1999 to 2001, the program averaged over 103 papers per year and it is on pace to produce some 120-130 papers in 2002. By contrast, from 1981, when the program began to take form, through 1989, there were 60 papers per year, and from 1990 to 1998, the program produced 75 papers a year. In part, the growth reflects increased numbers of people in the program. But even these figures do not fully represent the expansion of the labor research. For example, the NBER has a new Program on Children, a Working Group on Higher Education, and a new project on the science work force, all of which deeply involve labor-oriented specialists.
In any case, having established that this review is empirical, I turn next to some of the topics and ideas that have characterized labor studies research in the past several years. Because of the plethora of papers, this review is necessarily limited.
A substantial number of NBER researchers have been examining issues related to education. Some have continued the long-standing human capital interest in estimating returns to education, largely by examining the effects of different econometric procedures on earnings ( 7444, 7457, 7769, 7820, 7950, . 7989). Going a step toward assessing the impact of education beyond earnings, Lance J. Lochner and Enrico Moretti have documented the strong link between education and crime, which can increase the social returns to schooling by 14-26 percent above the individual returns.
Much recent work has focused on the higher education market. Orley C. Ashenfelter and David Card provide evidence that the end of forced mandatory retirement of faculty will increase the number of older faculty ( 8378) . John Bound and co-authors find that U.S. states awarding more BA degrees in a cohort tend to have higher concentrations of college-educated workers ( 8555) . Ronald G. Ehrenberg finds that collective bargaining coverage raises staff salaries in colleges and universities ( 8861) ; he has examined the growing dispersion of faculty salaries across institutions in terms of dispersion in endowments and appropriations for public institutions ( 8965) ; and he shows that the widely reported US News and World Report rankings of colleges in fact affect how colleges do in the market for students ( 7227) . Card and Thomas Lemieux analyze the link between cohort wage differentials and supplies in higher education over time ( 7655, 7658) , while Daron Acemoglu and Jorn-Steffen Pischke find that changes in the distribution of family income have had substantial effects on enrollments ( 7986) . Focusing on one of the most striking developments in higher education -- the increased proportion of women among graduates -- Kerwin K. Charles and Ming Ching Luoh argue the expected future earnings dispersion helps explain the data ( 9028) .
An additional body of work has explored the determinants of educational outputs, finding for the most part that increased expenditures ( 8269) or other reforms, ranging from voucher systems in Colombia (8343) to compulsory attendance laws ( 8563) to smaller class size (7656, 8875) , all affect outcomes. But Joshua D. Angrist and Frank Levy ( 7424) report that classroom computers did little for educational performance in Israel; while Julian R. Betts and Jeffrey T. Grogger report that grading standards improved educational achievement but lowered attainment (7875) . Peter Temin ( 8898) argues that failure to raise teacher pay has produced low-quality teachers, with adverse effects for outcomes, while Eric A. Hanushek and co-authors report that teacher mobility is more related to the characteristics of students than to salary ( 8599) . Caroline M. Hoxby uses a new survey of charter school teachers and other data to show that greater school choice could improve the quality of the teaching work force ( 7866) and provides evidence that peer effects are important in classrooms ( 7867) .In an analysis of Dartmouth College roommates, Bruce Sacerdote finds that peer effects are very important in determining levels of academic effort and in decisions to join social groups such as fraternities, but not in choice of college major ( 7469) . Finally, in a series of papers (7217, 7126, 6439, 6537) Claudia Goldin and Lawrence F. Katz have studied the returns to education and inequality from 1914 through 1999, concentrating on the US's development of universal and publicly funded secondary school education. They argue that the social capital embodied in relatively homogeneous small mid-Western towns fueled this expansion (three cheers for Iowa!).
Labor Institutions Around the World
NBER labor researchers also have studied labor markets in countries outside the United States. They do this for the sake of learning about the strengths and weaknesses of the American labor market and the strengths and weaknesses of our economic models, and, as Daniel S. Hamermesh argues, to take advantage of the greater variation in institutions across countries and of larger or better datasets in some other countries than in the United States ( 8757) .
Among these international projects, Card and I, working with Richard Blundell, directed a major investigation of the British economy in conjunction with the Centre for Economic Performance of the London School of Economics and the Institute of Fiscal Studies. We found that the Thatcher and ensuing economic reforms increased market freedoms and flexibility, improving employment rates, and ending the U.K.'s downward slide in GDP per capita, but with the cost of higher inequality ( 8801, 8253, 8448, 8413) . But in the United Kingdom, real wages grew more rapidly than did inequality, so that the standard of living of even those at the bottom of the earnings distribution improved during this period. Several researchers have examined German institutions and markets in some detail [7564, 7697, 7611, 8051, 8797] while others have looked at Japanese employment practices (7965, 7939) . Others have examined labor market institutions in Latin America;and still others have studied how the Canadian labor market and social insurance system have performed (7371, 7370, 8658) . Going from empirical work to theory, Steven J. Davis examined the relationship between the compression of wages under centralized bargaining and the distribution of employment in Sweden; he then proposed a search equilibrium model in which centralized bargaining improves productivity and welfare ( 7502, 8343) . Marianne Bertrand and Francis Kramarz have shown that entry regulation to the French retail industry reduced employment in the sector by about 10 percent ( 8211) . Nina Pavcnik has found that increases in the price of rice in Vietnam were a major factor in reducing child labor ( 8760) . Overall, the main theme that emerges from this stream of work is that labor market institutions have substantial effects on inequality and employment, and thus must be part of any empirical analysis, along with the two blades of Marshall's scissors.
Several researchers have used cross-country data to compare labor market outcomes around the world. Remco H. Oostendorp and I developed the Occupational Wages around the World data file ( 8058, and at "NBER Data") to provide better information on levels and structures of pay. Francine D. Blau and her co-authors have examined gender pay differences across countries ( 8200) and differences in employment patterns among OECD countries (8526, 9043) ; Paul Beaudry and co-authors have also worked on this topic ( 8149, 8754) . Using the International Adult Literacy Survey, Blau and Lawrence Kahn ( 8210) and Daniel Devroye and I ( 8140) have tested and rejected the claim that differences in inequality in cognitive skills explain cross-country differences in income inequality. Alan B. Krueger has explored the belief that investing in education is critical for macroeconomic growth, pointing out the importance of measurement error in schooling across countries, in affecting results ( 7190) . Acemoglu argued that one can explain differential patterns of change in inequality across countries in terms of the effect of labor institution-induced wage compression on the basis of technological change (8287, 8832) .
Building on the interest in markets in other advanced countries, Edward P. Lazear has organized a new working group that seeks to exploit datasets in other countries that match employers and employees or have other advantages over U.S. datasets.
Social Policies and Supply Responses
Economists in the labor program have examined the economic impact of US institutions and policies as well. Katz, Jeffrey R. Kling, and Jeffrey Liebman ( 7973) studied the impacts of the moving-to-opportunity program through which a random lottery gave some inner city residents eligibility for housing vouchers. They found improvements in children's social behavior but no effects on the labor market performance of parents. Bound and Sarah E. Turner ( 9044) found that the availability of G. I. benefits had a substantial and positive impact on the educational attainment of both white and black men born outside the South, but did not help Southern black GIs go to college; the result was exacerbation of the educational differences between black and white men from southern states. Jonathan Guryan ( 8345) found that the 1954 Supreme Court decision desegregating schools benefited the black students for whom the plans were designed, reducing their high school dropout rates, while having no effect on the dropout rates of whites. He suggests that peer effects explain at least some of the decline in the dropout rates of blacks attributable to desegregation plans.
Much work on economic programs and policies focuses on intended or unintended responses of the labor supply. In general, we would like to identify a single labor supply elasticity that we could apply to all problems. However, the specifics of programs and the contexts in which incentives change appear to rule out such an efficient solution to actual programs and decisions. Reviewing estimates of labor supply elasticities, Krueger and Bruce D. Meyer ( 9014) show that elasticities of lost work time that incorporate both the incidence and duration of claims for unemployment insurance and workers' compensation are substantially larger than the labor supply elasticities typically found for men in studies of the effects of wages or taxes on hours of work. Jeffrey Smith and co-authors ( 8825) show that the Worker Profiling and Reemployment Services, which "profiles" UI claimants to determine their probability of benefit exhaustion (or expected spell duration), affect behavior through its administration. The estimated treatment effect differs dramatically across the population. In a review of the relationship between health insurance, labor supply, and job mobility, John Grubert and Brigitte Madrian ( 8817) find that health insurance is a central determinant of retirement decisions; also, the labor supply of secondary earners has some impact on job mobility but is not a major determinant of the labor supply and welfare exit decisions of low income mothers.
Tools and More
Simon Kuznets did not need much in the way of econometric tools to get the numbers to speak, but most economists rely on formal statistics to help interpret data. In the past several years, some labor analysts have begun to use regression discontinuity models (8993, 8269, 8441) to deal with imperfect data. Bertrand, Esther Duflo, and Sendhil Mullainathan ( 8841) have examined the effect of serial correlation in outcomes variables on difference-in-difference estimates and find that in realistic CPS-based data the standard errors are severely biased. The basic theme of much of the statistical analysis has been toward applying less a priori structure and finding tools that better mimic experiments. On the notion that economists can do more than just mimic experiments, Alvin E. Roth and co-authors have been analyzing the matching market for physicians, auction sites, and the market for gastroenterologists, using data from the actual events or natural experiments and tests of hypotheses from laboratory experiments (6963, 7729, 8616) .
Finally, since my last Labor Studies report, one of the early members of the program, James J. Heckman, received a Nobel Prize. Although the prize was for his contribution to econometrics, we know that it is his grounding in labor issues and data that has spurred his creativity and prolificness. Jim's first paper in the Labor Program series was in 1978, while his first NBER paper was in 1974, before the NBER categorized papers by programs. More striking, perhaps, Jim's latest paper in the series as of this writing is dated July 2002. Who said scientific productivity falls with age? Sadly, over the same period, the labor program and economics more broadly lost an outstanding economist: Sherwin Rosen, one of the deepest thinkers and creative theorists in our midst, died in 2001. Sherwin's first paper in the labor series was in 1979, his first NBER paper was in 1977, and his last NBER paper in 2000 was on labor markets in professional sports ( 7573) , where he pondered alternative institutional ways to organize property rights and incentives.
*. Freeman is the Director of the NBER's Program on Labor Studies and is the Herbert Ascherman Professor of Economics at Harvard University.