Justin L. Tobias

Simple Estimators for Treatment Parameters in a Latent-Variable Framework

This note derives simply computed closed-form expressions for the average treatment effect, the effect of treatment on the treated, the local average treatment effect, and the marginal treatment effect in a latent-variable framework for both normal and nonnormal models. Asymptotic standard errors for versions of these parameters that average over observed characteristics are also obtained. The performances of the derived estimators are also evaluated in Monte Carlo experiments under correct specification and misspecification.

Controlling for Observed and Unobserved Site Characteristics in RUM Models of Recreation Demand

Random Utility Maximization (RUM) models of recreation demand are typically plagued by limited information on environmental and other attributes characterizing the available sites in the choice set. To the extent that these unobserved site attributes are correlated with the observed characteristics and/or the key travel cost variable, the resulting parameter estimates and subsequent welfare calculations are likely to be biased. In this paper we develop a Bayesian approach to estimating a RUM model that incorporates a full set of alternative specific constants, insulating the key travel cost parameter from the influence of the unobserved site attributes. In contrast to estimation procedures recently outlined in Murdock (2006), the posterior simulator we propose (combining data augmentation and Gibbs sampling techniques) can be used in the more general mixed logit framework in which some parameters of the conditional utility function are random. Following a series of generated data experiments to illustrate the performance of the simulator, we apply the estimation procedures to data from the Iowa Lakes Project. In contrast to an earlier study using the same data (Egan \textit{et al.} \cite{eganetal}), we find that, with the addition of a full set of alternative specific constants, water quality attributes no longer appear to influence the choice of where to recreate.

Model Uncertainty and Race and Gender Heterogeneity in the College Entry Decision.

This paper uses a flexible modeling strategy to examine the roles of measured ability, family characteristics and proxies for secondary schooling quality as determinants of the decision to enter college. While previous work on this topic has been careful to determine which explanatory variables to include when modeling college entry decisions, few studies have been concerned about appropriate distributional assumptions, (i.e. choice of link function). In this paper, I extend my binary choice analysis to the class of Student-t link functions, which enables me to approximately regard the often-used probit and logit models as special cases. Unconditional estimates which average over competing models and integrate out model uncertainty are also obtained. Using NLSY data, I apply these methods and find that the link functions and estimated impacts of ability and family characteristics on the probabilities of enrolling in college are not constant across race and gender groups.

Assessing Assessments of School Performance: The Case of California

The 1999 Public Schools Accountability Act set forth the Academic Performance Index (API) to assess the performance of Californias public schools. Under the Governors Performance Awards program, California schools showing adequate improvement in API scores become eligible for state funding. This article investigates the rule used by California which defines awards eligibility and determines if this rule is consistent with the act of rewarding schools that are performing better than expected. In addition, we introduce and estimate a multilevel hierarchical model to determine those school districts and counties that demonstrated the best performance over the 2000–2001 period.

The effects of cognitive ability and high school quality on college entry decisions: nonparametric estimation of parameters of interest

The decisions to attend college are analysed and nonparametric predictions compared to those obtained from the widely used logit model. The impacts of measured cognitive ability and proxies for high school quality on the decisions to attend college are examined for a sample of white and black males and females from the USA. Two different parameters of interest which isolate the effects of ability and high school quality on college entry decisions are described and estimated by ‘integrating out’ the effect of other covariates. It is found that measured cognitive ability is an extremely important determinant of college entry for all race and gender groups. At the same point in the ability distribution, blacks are more likely to select into college than whites, and females more likely than males of the same racial group. Proxies for high school quality such as teacher education, student teacher ratios, school enrolment and library size are shown to have little or no effects on the likelihood of college entry for all race and gender groups. Further, predictions obtained from the flexible nonparametric analysis are found to be quite similar to those obtained from the logit model, suggesting that simpler fully parametric binary choice models perform quite well as modelling college entry decisions.

A semiparametric investigation of the school quality-gs relationship

This article estimates a partially linear model that permits non-linearities of unspecified form in the school quality-earnings relationship. It examines the joint effect of teacher education and pupil-teacher ratios on 1990 earnings using NLSY data. It finds some evidence of non-linearities in this relationship, and that teacher education has a positive effect on log wages at some points in the pupil-teacher ratio support.

Explaining Trends in Body Mass Index Using Demographic Counterfactuals

The United States is experiencing a major public health problem relating to increasing levels of excess body fat. This paper is about the relationship in the United States between trends in the distribution of body mass index (BMI), including trends in overweight and obesity, and demographic change. We provide estimates of the counterfactual distribution of BMI that would have been observed in 2003–2008 had demographics remained fixed at 1980 values, roughly the beginning of the period of increasing overweight and obesity. We find that changes in demographics are partly responsible for the changes in the population distribution of BMI and are capable of explaining about 8.6% of the increase in the combined rate of overweight and obesity among women and about 7.2% of the increase among men. We also use demographic projections to predict a BMI distribution and corresponding rates of overweight and obesity for 2050.