Of Pencils and Computers
(Although a number of people, including most of the key players in the debate, have given me very positive feedback about this paper, it was written primarily for students in my graduate labor classes and has no obvious outlet. This website seems like a good place for it.)
This paper discusses the well-known papers by Alan Krueger and by John DiNardo and Steve Pischke on the relation between computers and the rise in wage inequality. It addresses what can be learned from the exercise in the context of simple Roy and hedonic models. I think there has been a great deal of misinterpretation of the Krueger/DiNardo/Pischke exchange. This paper helps to clarify the issues but does not break new theoretical or empirical ground.
The Pricing of Job Characteristics When Markets Do Not Clear: Theory and Policy Implications (with Sumon Majumdar) extended version of paper in International Economic Review
This paper examines a model of nonsequential search when jobs can vary with respect to nonpecuniary characteristics. We find that in the presence of frictions in the labor market, the equilibrium distribution need not show evidence of compensating wage differentials. The model also generates several pervasive features of labor markets: (a) unemployment and vacancies, (b) apparent discrimination, and (c) market segmentation. When workers are homogeneous, there is no positive role for policy -- restrictions on the range of job offers must decrease welfare and cannot reduce unemployment. However, when workers have heterogeneous preferences, such restrictions may lower unemployment and even lead to a Pareto-improvement in welfare. In particular, we consider the impact of policies banning discrimination and regulating working-conditions.
Effort and Wages: Evidence from the Payroll Tax (Canadian Journal of Economics, accepted subject to minor revisions)
Under the standard competitive model, if a tax change affects a group of workers with highly inelastic labor supply, their earnings will fall by essentially the entire nominal employer share of the tax increase. Allowing the wage to play a motivational role but maintaining the market-clearing assumption broadens the range of possible outcomes. With a 50/50 split in the nominal share, given a reasonable estimate of the elasticity of demand, earnings could fall from anywhere between 0 and more than 100% of the employer's nominal share but would not rise. In contrast, because there is excess labor (involuntary unemployment) in equilibrium, efficiency wage models function very much like models in which the supply of labor is perfectly elastic, and thus earnings rise by more than the worker's nominal share. I argue that the 1968, 1974 and 1979 increases in the taxable earnings base for FICA provide good opportunities to test the models. This tax increase affected only those workers earning significantly more than the median earnings for male full-time/year-round workers. Such workers' labor force participation is likely to have been highly inelastic. In addition, low earnings workers did not experience this tax increase. The results are supportive of models in which the motivational effects of wages are important but cannot clearly distinguish between the efficiency wage and market-clearing versions of those models.
If firms hire heterogeneous workers but must offer all workers insurance benefits under similar terms, then in equilibrium, some firms offer free health insurance, some require an employee premium payment and some do not offer insurance. Making the employee contribution pre-tax lowers the cost to workers of a given employee premium and encourages more firms to charge. This increases the offer rate, lowers the take-up rate, increases (decreases) coverage among high (low) demand groups, with an indeterminate overall effect. This pattern is consistent with trends in the U.S. economy following the creation of section 125 plans.
This is a paper that I wrote a long time ago and for which I still get frequent requests. Although I got a “revise and resubmit” from a prestigious journal, I never rewrote the paper because I thought there were errors in the approach. There are definitely mistakes in the paper, but in retrospect I regret not revising it. I believe the paper was very influential in making people think about the proper interpretation of instrumental variables when coefficients are not constant and, in particular, laid some of the groundwork for the work of Josh Angrist and Guido Imbens on LATE and the work of Jim Heckman and his coauthors on the interpretation of instrumental variables. I post it here for the historical record.
The Effect of High School Exit Exams on Graduation, Employment and Incarceration (with Olesya Baker)
We evaluate the effects of high school exit exams on high school graduation, incarceration, employment and wages. We construct a state/graduation-cohort dataset using the Current Population Survey, Census and information on exit exams. We find relatively modest effects of high school exit exams except on incarceration. Exams assessing academic skills below the high school level have little effect. However, more challenging standards-based exams reduce graduation and increase incarceration rates. About half the reduction in graduation rates is offset by increased GED receipt. We find no consistent effects of exit exams on employment or the distribution of wages.
Discrimination and Worker Evaluation (with Costas Cavounidis)
We develop a model of self-sustaining discrimination in wages, coupled with higher unemployment and shorter employment duration among blacks. While white workers are hired and retained indefinitely without monitoring, black workers are monitored and fired if a negative signal is received. The fired workers, who return to the pool of job-seekers, lower the average productivity of black job-seekers, perpetuating the cycle of lower wages and discriminatory monitoring. Under suitable parameter values the model has two steady states, one corresponding to each population group. Discrimination can persist even if the productivity of blacks exceeds that of whites.
We use the exchange between Kearney/Levine and Jaeger/Joyce/Kaestner on “16 and Pregnant” to reexamine the use of DiD as a response to the failure of nature to properly design an experiment for us. We argue that 1) any DiD paper should address why the original levels of the experimental and control groups differed, and why this would not impact trends, 2) the parallel trends argument requires a justification of the chosen functional form and that the use of the interaction coefficients in probit and logit may be justified in some cases, and 3) parallel trends in the period prior to treatment is suggestive of counterfactual parallel trends, but parallel pre-trends is neither necessary nor sufficient for the parallel counterfactual trends condition to hold. Importantly, the purely statistical approach uses pretesting and thus generates the wrong standard errors. Moreover, we underline the dangers of implicitly or explicitly accepting the null hypothesis when failing to reject the absence of a differential pre-trend.
We develop a fairly general and tractable model of investment when workers can invest in multiple skills and different jobs put different weights on those skills. In addition to expected findings such as that younger workers are more likely than older workers to respond to a demand shock by investing
in skills whose value unexpectedly increases, we derive some less obvious results. Credit constraints may affect investment even when they do not bind it equilibrium. If there are mobility costs, firms will generally have an incentive to invest in some of their workers' skills even when there are a large number of similar competitors, and, in equilibrium, there can be overinvestment in all skills. Worker skill accumulation resembles learning by doing even in its absence. We demonstrate how the model can be simulated to show the effect of a shock to the price of individual skills.
Too Family Friendly? The Consequences of Parents' Right to Request Part-Time Work (with Daniel Fernandez-Kranz and Nuria Rodriguez-Planas)
We use differences-in-differences to evaluate a 1999 Spanish law granting employment protection to workers with children younger than 6 and who had asked for a shorter workweek due to family responsibilities. Only mothers took advantage of this arrangement. Consequently, relative to men of similar age, employers were subsequently 49% less likely to hire childbearing-age women, 40% more likely to separate from them, and 37% less likely to promote them to permanent contracts, lowering female employment by 3-8 percent. The results are similar using older women unaffected by the reform as controls. The reform affected both eligible and ineligible women, albeit more the former. Effects are largest where expected accommodation costs are greatest: small firms, low-skill jobs, and industries with few part-time workers. Employers passed on part of the costs through lower wages, with large adjustments in fixed-term jobs. Findings are robust to sensitivity analyses, and placebo tests.
The Determinants of Teachers' Occupational Choice (with Maria Dolores Palacios)
Among college graduates, teachers have both low average AFQT and high average risk aversion, perhaps because the compression of earnings within teaching attracts relatively risk-averse individuals. Using a dynamic optimization model with unobserved heterogeneity, we show that were it possible to make teacher compensation mimic the return to skills and riskiness of the nonteaching sector, overall compensation in teaching would increase. Moreover, this would make many current teachers substantially worse off, making reform challenging. Importantly, our conclusions are sensitive to the degree of heterogeneity for which we allow. Since even a model with no unobserved heterogeneity fits well within sample, one could easily conclude that allowing for two or three types fits the data adequately. Formal methods reject this conclusion. The BIC favors seven types. Ranking models using cross-validation, nine types is better although the improvements of going from six to seven, from seven to eight and from eight to nine types are noticeably smaller than those from adding an additional type to a lower base.
The Sad Truth about Happiness Scales (with Timothy N. Bond)
Happiness is reported in ordered intervals (e.g. very, pretty, not too happy). We review and apply standard statistical results to determine when such data permit identification of two groups’ relative average happiness. The necessary conditions for nonparametric identification are strong and unlikely to be ever satisfied. Standard parametric approaches cannot identify this ranking unless the variances are exactly equal. If not, ordered probit findings can be reversed by lognormal transformations. For nine prominent happiness research areas, conditions for nonparametric identification are rejected and standard parametric results are reversed using plausible transformations. Tests for a common reporting function consistently reject.