Rajeev Dehejia

The impact of trade liberalization on manufacturing growth has been widely studied in the literature. What has gone unappreciated is that accelerated manufacturing growth has also been accompanied by accelerated services growth. Using firm-level data from India, we find a positive spillover from manufacturing growth to gross value added, wages, employment, and worker productivity in services, especially large urban firms and in service sectors whose output is used as a manufacturing input.

This paper investigates the impact of income and non-income shocks on child labor using a model in which the household maximizes utility from consumption as well as human capital development of the child. We also investigate if access to credit and household assets act as buffers against transitory shocks. Our results indicate significant effects of agricultural shocks on the child’s overall work hours and agricultural work hours, with higher effects for boys. Crop shocks also have significant adverse effects on school attendance, with girls experiencing a more-than 70% increase in the probability of quitting schooling. The results also indicate that access to a bank account has a buffering effect on the impact of shocks on child hunger. Having a bank account reduces both male child labor and household work hours of a girl child. While assets reduce working hours of girls, we do not find it having a significant effect on boys. We also do not see assets to act as a buffer against shocks

Experimental evidence on a range of interventions in developing countries is accumulating rapidly. Is it possible to extrapolate from an experimental evidence base to other locations of policy interest (from “reference” to “target” sites)? And which factors determine the accuracy of such an extrapolation? We investigate applying the Angrist and Evans (1998) natural experiment (the effect of boy-boy or girl-girl as the first two children on incremental fertility and mothers’ labor force participation) to data from International IPUMS on 166 country-year censuses. We define the external validity function with extrapolation error depending on covariate differences between reference and target locations, and find that smaller differences in geography, education, calendar year, and mothers’ labor force participation lead to lower extrapolation error. As experimental evidence accumulates, out-of-sample extrapolation error does not systematically approach zero if the available evidence base is naïvely extrapolated, but does if the external validity function is used to select the most appropriate reference context for a given target (although absolute error remains meaningful relative to the magnitude of the treatment effect). We also investigate where to locate experiments and the decision problem associated with extrapolating from existing evidence rather than running a new experiment at a target site.

In an experiment in non-formal schools in Indian slums, a reward scheme for attending a target number of school days increased average attendance when the scheme was in place, but had heterogeneous effects after it was removed. Among students with high baseline attendance, the incentive had no effect on attendance after it was discontinued, and test scores were unaffected. Among students with low baseline attendance, the incentive low- ered post-incentive attendance, and test scores decreased. For these students, the incen- tive was also associated with lower interest in school material and lower optimism and confidence about their ability. This suggests incentives might have unintended long-term consequences for the very students they are designed to help the most. 

This paper surveys six widely-used non-experimental methods for estimating treatment effects (instrumental variables, regression discontinuity, direct matching, propensity score matching, linear regression and non-parametric methods, and difference-in-differences), and assesses their internal and external validity relative both to each other and to randomized controlled trials. While randomized controlled trials can achieve the highest degree of internal validity when cleanly implemented in the field, the availability of large, nationally representative data sets offers the opportunity for a high degree of external validity using non-experimental methods. We argue that each method has merits in some context and they are complements rather than substitutes.

This paper investigates how fertility responds to financial incentives. We construct a large, individual-level panel data set of over 300,000 Israeli women during the period 1999–2005 with comprehensive information on their fertility histories, education, religious affiliation, ethnicity, and income. We exploit variation in Israel’s child subsidy program to identify the impact of changes in the price of a marginal child on fertility. We find a positive, statistically significant, and economically meaningful price effect on fertility. This positive effect is strongest for households in the lower range of the income distribution, weakens with income, and is present in all religious and ethnic subgroups. There is also a significant price effect on fertility among women who are close to the end of their lifetime fertility, suggesting that at least part of the effect that we estimate is due to a reduction in total fertility. Finally, we investigate how changes in household income affect fertility choices. Consistent with Becker (1960) and Becker and Tomes (1976), we find that the income effect is small in magnitude, and is negative at low income levels and positive at high income levels.

Public disclosure of labor conditions has been suggested as one way to encourage compliance with labor law and improvements in working conditions. Analyzing labor law compliance data in the apparel industry from Better Factories Cambodia, this paper finds that after the elimination of public disclosure of factory- level noncompliance the rate of increase in compliance slowed,but did not return to the baseline, even in the absence of a reputation sensitive buyer. 

“Best practice” in microfinance holds that interest rates should be set at profit-making levels, based on the belief that even poor customers favor access to finance over low fees.  Despite this core belief, little direct evidence exists on the price elasticity of credit demand in poor communities.  We examine increases in the interest rate on microfinance loans in the slums of Dhaka, Bangladesh.  Using unanticipated between-branch variation in prices, we estimate interest elasticities from -0.73 to -1.04, with our preferred estimate being at the upper end of this range. Interest income earned from most borrowers fell, but interest income earned from the largest customers increased, generating overall profitability at the branch level. 

 In this chapter, we take a first stab at analyzing the growth of services in India using firm-level data.

While substantial literature now exists on poverty and inequality among social groups, until now, almost nothing has been known about how the socially dis-advantaged groups fare in entrepreneurship in terms of shares in the GVA, workers employed, and number of enterprises owned. Our chapter provides a first comprehensive look at these measures of entrepreneurship. We analyze the presence of the socially disadvantaged groups in proprietary and partnership enterprises in the economy as a whole, according to enterprises size, in rural and urban areas, according to sectors, and in different states.  

Despite the extensive literature on the determinants of child labor, the evidence on the consequences of child labor on outcomes such as education, labor, and health is limited. We evaluate the causal effect of child labor participation among children in school on these outcomes using panel data from Vietnam and an instrumental variables strategy. Five years subsequent to the child labor experience we find significant negative impacts on education, and also find a higher probability of wage work for those young adults who worked as children while attending school. We find few significant effects on health.

This paper examines whether participation in religious or other social organizations can help offset the negative effects of growing up in a disadvantaged environment. Using the National Survey of Families and Households, we collect measures of disadvantage as well as parental involvement with religious and other social organizations when the youth were ages 3 to 19 and we observe their outcomes 13 to 15 years later. We consider a range of definitions of disadvantage in childhood (family income and poverty measures, family characteristics including parental education, and child characteristics including parental assessments of the child) and a range of outcome measures in adulthood (including education, income, and measures of health and psychological wellbeing). Overall, we find strong evidence that youth with religiously active parents are less affected later in life by childhood disadvantage than youth whose parents did not frequently attend religious services. These buffering effects of religious organizations are most pronounced when outcomes are measured by high school graduation or non-smoking and when disadvantage is measured by family resources or maternal education, but we also find buffering effects for a number of other outcome-disadvantage pairs. We generally find much weaker buffering effects for other social organizations.

This article demonstrates that minimum wage laws need not induce unemployment even under the classic labor supply-and-demand paradigm. As a result, minimum wage laws can be welfare-enhancing under the basic labor supply and demand model, suggesting the presence of an optimal minimum wage. We discuss conditions under which the optimal minimum wage level is the subsistence wage level. As a consequence, minimum wages should vary across states or countries with the local subsistence levels.

This paper explores the relationship between the theory and practice of program evaluation as it relates to training programs. In practice programs are evaluated by mean-variance comparisons of the empirical distributions of the outcome of interest for the treatment and control programs. Typically, earnings are compared through the average treatment effect (ATE) and its standard error. In theory, programs should be evaluated as decision problems using social welfare functions and posterior predictive distributions for outcomes of interest. This paper considers three issues. First, under what conditions do the two approaches coincide? I.e., when should a program be evaluated based purely on the average treatment effect and its standard error? Second, under more restrictive parametric and functional form assumptions, the paper develops intuitive mean-variance tests for program evaluation that are consistent with the underlying decision problem. Third, these concepts are applied to the GAIN and JTPA data sets.

This paper examines whether involvement with religious organizations can help insure consumption and happiness. Using data from the Consumer Expenditure Survey (CEX), we find that households who contribute to a religious organization are better able to insure their consumption against income shocks. Using the National Survey of Families and Households (NSFH), we find that individuals who attend religious services are better able to insure their happiness against income shocks. Overall, our results suggest that religious organizations provide insurance though the form of this insurance may differ by race.

This paper studies the effect of state-level banking regulation on financial development and on components of state-level growth in the United States from 1900 to 1940. We use these banking laws to assess the findings of a large recent literature that has argued that financial development contributes to economic growth. We contend that the institutional mechanism leading to financial development is important in determining its consequences and that some types of financial development can even retard economic growth. For the United States from 1900 to 1940, we argue that the financial expansion induced by expanded bank branching accelerated the mechanization of agriculture and spurred growth in manufacturing. In contrast, financial expansions induced by state deposit insurance had negative consequences for both the agricultural and manufacturing sectors.

This paper examines the relationship between household income shocks and child labor. In particular, we investigate the extent to which transitory income shocks lead to increases in child labor and whether household access to credit mitigates the effects of these shocks. Using panel data from a survey in Tanzania, we find that both relationships are significant. Our results suggest that credit constraints play a role in explaining child labor and consequently that child labor is inefficient, but we also discuss alternative interpretations.

Even though access to credit is central to child labor theoretically, little work has been done to assess its importance empirically. Dehejia and Gatti examine the link between access to credit and child labor at a cross-country level. The authors measure child labor as a country aggregate, and proxy credit constraints by the level of financial market development. These two variables display a strong negative (unconditional) relationship. The authors show that even after they control for a wide range of variables-including GDP per capita, urbanization, initial child labor, schooling, fertility, legal institutions, inequality, and openness-this relationship remains strong and statistically significant. Moreover, they find that, in the absence of developed financial markets, households resort to child labor to cope with income variability.

This evidence suggests that policies aimed at increasing households'access to credit could be effective in reducing child labor.

I argue for thinking of program evaluation as a decision problem. There are two steps. First, a counselor determines which program (treatment or control) each individualjoins,based for example on maximizing the probability of employment or expected earnings. Second, the policymaker decides whether: to assign all individuals to treatment or to control, or to allow the counselor to choose.This framework has two advantages. Individualized assignment rules (known as profiling) can raise the average impact, improving cost effectiveness by exploiting treatment-impact heterogeneity. Second, it accounts systematically for inequality and uncertainty, and the policymaker's attitude toward these, in the evaluation.

This paper discusses propensity score matching in the context of Smith and Todd's (Does matching overcome Lalonde's critique of nonexperimental estimators, J. Econom., in (press) reanalysis of Dehejia and Wahba (J. Am. Statist. Assoc. 97 (1999) 1053; National Bereau of Economics Research working Paper No. 6829, Rev. Econom. Statist., 2002, forthcoming). Propensity score methods require that a separate propensity score specification be estimated for each treatment group-comparison group combination. Furthermore, a researcher should always examine the sensitivity of the estimated treatment effect to small changes in the propensity score specification; this is a useful diagnostic on the quality of the comparison group. When these are borne in mind, propensity score methods are useful in analyzing all of the subsamples of the NSW data considered in Smith and Todd (Does matching overcome Lalonde's critique of nonexperimental estimators, J. Econom., in press).

This paper investigates the incentive effects of automobile insurance, compulsory insurance laws, and no-fault liability laws on driver behavior and traffic fatalities. We analyze a panel of 50 U.S. states and the District of Columbia for 1970–98, a period in which many states adopted compulsory insurance regulations and/or no-fault laws. Using an instrumental variables approach, we find evidence that automobile insurance has moral hazard costs, leading to an increase in traffic fatalities. We also find that reductions in accident liability produced by no-fault liability laws have led to an increase in traffic fatalities (estimated to be on the order of 6 percent). Overall, our results indicate that, whatever other benefits they might produce, increases in the incidence of automobile insurance and moves to no-fault liability systems have significant negative effects on traffic fatalities.

In this paper we study the relationship between the unemployment rate at the time of a baby's conception and health outcomes at birth, and we explore whether this relationship is due to the effect of the unemployment rate on fertility decisions or on the health-related behavior of pregnant women. Economic models of fertility suggest that women who choose to have children in recessions may differ from women who choose to postpone fertility. To the extent that these parental characteristics are related to children's health, differential fertility may result in differences in the health of children over the business cycle. At the same time, evidence suggests that individuals' health may improve during recessions, because the overall effect of recessions is to increase health-related activities (and to decrease risky behaviors). Therefore, changes in parental behavior over the business cycle could also affect the health of infants, even in the absence of compositional change.

This paper discusses the problem of evaluating and predicting the treatment impact of a program that is implemented at multiple sites. Two issues arise: is information from other sites relevant in estimating the impact at a given site? and how can we account for predictive uncertainty in the site effects? Using data from the Greater Avenues for Independence evaluation, I develop a hierarchical model for earnings which allows both for site effects and for smoothing the estimated impact across sites. I show that the degree to which the estimated impact is smoothed across sites does not affect the estimate; i.e. most of the differences across sites are due to differences in the composition of participants. Second I show that predictive uncertainty regarding site effects is important; for example, when the predictive uncertainty regarding site effects is ignored, the treatment impact at the Riverside sites is significant, but when we consider predictive uncertainty, the impact for the Riverside sites is insignificant. Third, I show that the hierarchical model is able to extrapolate site effects with reasonable accuracy when the site for which the prediction is being made does not differ substantially from the sites already observed. For example, the San Diego treatment effects could have been predicted based on observable site characteristics, but the Riverside effects are consistently underestimated.

This paper considers causal inference and sample selection bias in nonexperimental settings in which (i) few units in the nonexperimental comparison group are comparable to the treatment units, and (ii) selecting a subset of comparison units similar to the treatment units is dif? cult because units must be compared across a high-dimensional set of pretreatment characteristics. We discuss the use of propensity score-matching methods, and implement them using data from the National Supported Work experiment. Following LaLonde (1986), we pair the experimental treated units with nonexperimental comparison units from the CPS and PSID, and compare the estimates of the treatment effect obtained using our methods to the benchmark results from the experiment. For both comparison groups, we show that the methods succeed in focusing attention on the small subset of the comparison units comparable to the treated units and, hence, in alleviating the bias due to systematic differences between the treated and comparison units.

The need to use randomized experiments in the context of manpower training programs, and in analyzing causal effects more generally, has been a subject of much debate. Lalonde (1986)considers experimental data from the National Supported Work (NSW) Demonstration and nonexperimental comparison groups drawn from the CPS and PSID, and argues that econometricmethods fail to replicate the benchmark experimental treatment effect. This paper applies propensity score methods, which have been developed in the statistics literature, to Lalonde'sdataset. In contrast with Lalonde's findings, using propensity score methods, we are able closely to replicate the experimental training effect. The methods succeed because they are able flexibly to control for the wide range of observable differences between the (experimental) treatment group and the (non-experimental) comparison group.