Many studies have examined the impact of immigration on native wages. Some of these studies have relied upon education-experience groups to define labor markets and identify the wage elasticity with respect to immigrant labor supply. However, evidence suggests that immigrants’ educational attainment is treated differently in the labor market and constructing labor markets based upon this characteristic leads to potentially biased conclusions. We utilize O∗NET occupational characteristics to form a different set of labor markets. Our analysis finds higher partial equilibrium effects on native wages than prior work using education-experience skill groups, as expected. These larger effects, however, are shown to be concentrated on the least skilled natives. Estimates of the total wage effect along the distribution of occupational skills confirm that the negative wage effect is concentrated on native workers in the bottom tail of the distribution. Natives in the upper tail of the distribution experience wage gains as a result of immigration. The distributional impact is likely due to the distribution of skills among recent immigrants.

In this paper we present a Bayesian approach to estimate the mean of a binary variable and changes in the mean over time, when the variable is subject to misclassification error. These parameters are partially identified and we derive identified sets under various assumptions about the misclassification rates. We apply our method to estimating the prevalence and trend of prescription opioid misuse, using data from the 2002-2014 National Survey on Drug Use and Health. Using a range of priors, the posterior distribution provides evidence that among middle-aged white men, the prevalence of opioid misuse increased multiple times between 2002 and 2012.

Earnings nonresponse in household surveys is widespread, yet there is limited evidence on whether and how nonresponse bias affects measured earnings. This paper examines the patterns and consequences of nonresponse on earnings gaps and inequality using Census internal Current Population Survey individual records linked to Social Security administrative data on earnings for calendar years 2005-2010. We find that the common assumption that earnings are missing at random is rejected, and that nonresponse across the earnings distribution is U-shaped—left-tail “strugglers” and right-tail “stars” are least likely to report earnings. Approximately one-third to one-half the difference in inequality measures between the CPS and administrative data is accounted for by nonresponse in the CPS. Earnings differentials by race, gender, and education in lower and upper quantiles are biased from nonresponse as much as 20 percent. We recommend users of public CPS data adopt flexible copula-based models to correct for nonrandom nonresponse for distributional research.

…Lake Wobegon, where all the women are strong, all the men are good looking, and all the children are above average.—Garrison Keillor, A Prairie Home Companion (Lee 1991)

Over 20 years ago, Maxwell and Lopus (1994) identified what has since been referred to as the “Lake Wobegon” effect in economic education research. Given that student self-reported data is a ubiquitous feature of empirically based economic education research, the potential for systematic misreporting raises significant concerns about biased estimates and misinterpretation of results when nonrandom reporting occurs. Using data from 509 students surveyed at two universities, we too find the Lake Wobegon effect. We consider factors that influence the degree and direction of the potential misreporting of current cumulative GPA. Consistent with past studies, we examine differences in means and use OLS estimation. Going beyond earlier work, as our survey was administered in the classroom, we incorporate information about situational factors relevant to the student while completing the survey. We also evaluate density, deciles, and use quantile analysis to more completely characterize the sources of the Lake Wobegon effect. This additional analysis shows that what might at first glance appear to be a difference in mean values might actually be driven by heteroscedasticity associated with the GPA, which differs across gender. Also, survey conditions have a systematic impact on self-reported values of academic performance. A better understanding of these influences might more aptly inform economic education research in both study techniques and conclusions drawn from data.

We present a Bayesian analysis of a regression model with a binary covariate that may have classification (measurement) error. Prior research demonstrates that the regression coefficient is only partially identified. We take a Bayesian approach which adds assumptions in the form of priors on the unknown misclassification probabilities. The approach is intermediate between the frequentist bounds of previous literature and strong assumptions which achieve point identification, and thus preferable in many settings. We present two simple algorithms to sample from the posterior distribution when the likelihood function is not fully parametric but only satisfies a set of moment restrictions. We focus on how varying amounts of information contained in a prior distribution on the misclassification probabilities change the posterior of the parameters of interest. While the priors add information to the model, they do not necessarily tighten the identified set. However, the information is sufficient to tighten Bayesian inferences. We also consider the case where the mismeasured binary regressor is endogenous. We illustrate the use of our Bayesian approach in a simulated data set and an empirical application investigating the association between narcotic pain reliever use and earnings.

We consider Bayesian inference about the mean of a binary variable that is subject to misclassification error. If the error probabilities are not known, or cannot be estimated, the parameter is only partially identified. For several reasonable and intuitive prior distributions of the misclassification probabilities, we derive new analytical expressions for the posterior distribution. Our results circumvent the need for Markov chain Monte Carlo simulation. The priors we use lead to regions in the identified set that are aposteriori more likely than others.

The Current Population Survey Annual Social and Economic Supplement (CPS ASEC) serves as the data source for official income, poverty, and inequality statistics in the United States. There is a concern that the rise in nonresponse to earnings questions could deteriorate data quality and distort estimates of these important metrics. We use a dataset of internal ASEC records matched to Social Security Detailed Earnings Records (DER) to study the impact of earnings nonresponse on estimates of poverty from 1997–2008. Our analysis does not treat the administrative data as the “truth”; instead, we rely on information from both administrative and survey data. We compare a “full response” poverty rate that assumes all ASEC respondents provided earnings data to the official poverty rate to gauge the nonresponse bias. On average, we find the nonresponse bias is about 1.0 percentage point.

This paper considers the use of multiple proxy measures for an unobserved variable and contrasts the approach taken in the measurement error literature to that of the model specification literature. We find that including all available proxy variables in the regression minimizes the bias on coefficients of correctly measured variables in the regression. We derive a set of bounds for all parameters in the model, and compare these results to extreme bounds analysis. Monte Carlo simulations demonstrate the performance of our bounds relative to extreme bounds. We conclude with an empirical example from the cross-country growth literature in which human capital is measured through three proxy variables: literacy rates, and enrollment in primary and secondary school, and show that our approach yields results that contrast sharply with extreme bounds analysis.

Earnings nonresponse in the Current Population Survey is roughly 30% in the monthly surveys and 20% in the March survey. If nonresponse is ignorable, unbiased estimates can be achieved by omitting nonrespondents. Little is known about whether CPS nonresponse is ignorable. Using sample frame measures to identify selection, we find clear-cut evidence among men but limited evidence among women for negative selection into response. Wage equation slope coefficients are affected little by selection, but because of intercept shifts, wages for men and, to a lesser extent, women are understated, as are gender gaps. Selection is least severe among household heads.

We study a regression model with a binary explanatory variable that is subject to misclassification errors. The regression coefficient is then only partially identified. We derive several results that relate different assumptions about the misclassification probabilities and the conditional variances to the size of the identified set.

We offer new evidence on earnings volatility of men and women in the United States over the past four decades by using matched data from the March Current Population Survey. We construct a measure of total volatility that encompasses both permanent and transitory instability, and that admits employment transitions and losses from self-employment. We also present a detailed decomposition of earnings volatility to account for changing shares in employment probabilities, conditional variances of continuous workers, and conditional mean variances from employment entry and exit. Our results show that earnings volatility among men increased by 15% from the early 1970s to mid-1980s, while women’s volatility fell, and each stabilized thereafter. However, this pooled series masks important heterogeneity in volatility levels and trends across education groups and marital status. We find that men’s earnings volatility is increasingly accounted for by employment transitions, especially exits, while the share of women’s volatility accounted for by continuous workers rose, each of which highlights the importance of allowing for periods of non-work in volatility studies.

The Appalachian region has experienced persistently higher poverty and lower earnings than the rest of the United States. We examine whether skill differentials or differences in the returns to those skills lie at the root of the Appalachian wage gap. Using census data, we decompose the Appalachian wage gap using both mean and full distribution methods. Our findings suggest that significant upgrading of skills within the region has prevented the gap from widening over the last 20 years. Additionally, we find that urban areas within Appalachia have not experienced the rise in returns to skills as in non-Appalachian urban areas.

In this paper, we consider GMM estimation of the regression and MRSAR models with SAR disturbances. We derive the best GMM estimator within the class of GMM estimators based on linear and quadratic moment conditions. The best GMM estimator has the merit of computational simplicity and asymptotic efficiency. It is asymptotically as efficient as the ML estimator under normality and asymptotically more efficient than the Gaussian QML estimator otherwise. Monte Carlo studies show that, with moderate-sized samples, the best GMM estimator has its biggest advantage when the disturbances are asymmetrically distributed. When the diagonal elements of the spatial weights matrix have enough variation, incorporating kurtosis of the disturbances in the moment functions will also be helpful.

The Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) dramatically changed the availability of food stamps to immigrants, but not to refugees. However, refugee usage of food stamps in the post-reform era has declined more than usage of either other immigrants or native citizens. We investigate food stamp program participation of immigrants and refugees to ascertain if the resulting decline was an unintended effect of the reform or is due to some other phenomena. We use data from the Immigration and Naturalization Service to construct an instrument for refugees. We find that refugees are more likely than other immigrants to use food stamps. While PRWORA clearly reduced non-refugee immigrant food stamp participation, the post-reform era decline in refugee usage is fully attributable to a strong refugee response to changing labor market conditions. We also find that refugees’ use of food stamps declines rapidly with time in the United States, unlike that of other immigrants.

It is common practice to use sensible rules of thumb for cleaning data. Measurement error is often the justification for removing (trimming) or recoding (winsorizing) observations where the dependent variable has values that lie outside a specified range. We consider a general measurement error process that nests many plausible models. Analytic results demonstrate that winsorizing and trimming are solutions for a narrow class of error processes. Indeed such procedures can induce or exacerbate bias. Monte Carlo simulations and empirical results demonstrate the fragility of cleaning. Even on root mean square error criteria, we cannot find generalizable justifications for
these procedures.

Using state administrative records matched to the 1984 Survey of Income and Program Participation, we examine intertemporal relationships in response errors. False negative errors in reporting food stamps are highly correlated across interviews for the same household. Hypotheses that the error process can be explained by learning behaviour are not supported. Bivariate probit of response error in two periods reveals that responses to covariates are stable over time and the latent error terms are positively correlated. These findings support the hypothesis that respondents have a latent tendency to cooperate (or not cooperate) with surveys.

Making use of performance data for baseball players, this article provides empirical evidence in support of Lazear’s (1998) theoretical predictions that (1) risky workers will earn a premium for their upside potential, (2) this risk premium will be higher the longer a worker’s work life, and (3) firms must enjoy some comparative advantage in the labor market to be willing to pay a premium to risky workers. The validity of Lazear’s predictions carries implications for wage differentials between young and old workers and between men and women.

An annual panel of employment at the census tract level for the Atlanta region is used to estimate the change in a tract’s share of regional employment as a function of a variety of tax incentive programs, different transportation infrastructure investments, and crime. The results show that neighborhood-based property tax abatements, job tax credits, and highway improvements increase a tract’s employment share. Higher crime is found to reduce employment share.

Proxy variables arc frequently used in economics to control for unavailable variables in a linear regression setting. For example, AFQT scores have been used to control for human capital accumulation in measuring black-white wage differentials. This practice may bias the coefficient estimates for the correctly measured variables as well. This paper models proxy variables as a measurement error process and derives bounds for the coefficients on the correctly measured variables under a variety of assumptions. The results show that the coefficient on race in a linear regression is an overstatement of the actual black-white wage gap. Sensitivity analysis suggests that if human capital could be correctly measured it would be unlikely that the coefficient on black would be negative.

Broad variation in estimates of the union wage gap has perplexed labor economists. One specification error that is consistent with the observed variation is measurement error in reported union status. This article applies the results of Bollinger (1996) to estimate a range for the union wage gap. Both a cross-sectional model and a fixed-effects model are estimated. In order for the true coefficient in the fixed-effects estimates to be bounded below the true coefficient in the cross-sectional estimates, measurement error would have to be less than 0.8%. The difference between the fixed-effects estimates and the cross-sectional estimates is primarily due to measurement error rather than to unobserved heterogeneity. An examination of differences in returns to union membership by industry, occupation, and educational level shows that these differences are largely robust to measurement error. Many of these differences would be found even if error rates were as high as 10% or more.

Error in survey data originates from failure to contact the sample and from false answers to verifiable questions. These errors may be systematic and associated with uncooperative or unreliable respondents. Zabel modeled attrition in the Survey of Income and Program Participation and found systematic demographic and design effects. Bollinger and David modeled response error and identified correlations to income per capita. In this analysis, we link missing interviews in a panel and response error through a trivariate probit analysis. Robustness of the correlation between attrition and response error is examined by comparing variants of the model. The joint model of response error and attrition becomes the first stage of a pseudolikelihood estimate of a model of food-stamp participation. The model is significantly different from naive probit on the survey data.

Hedonic office rent models are estimated using data for Atlanta that span the years 1990- 1996. Controlling for typical building characteristics and lease terms, we find that variables measuring locational differences in wage rates, transport rates, and proximity to concentrations of support services and office workers play an important role in explaining spatial variation in office rents. No evidence is found in support of the hypothesis that technological advances in telecommunications have diminished the role played by face-to-face agglomeration economies in determining the intra-metropolitan location of office firms.

This article utilizes an exact match file between the 1978 March Current Population Survey and administrative records from the Social Security Administration to analyze errors in the reporting of annual income using nonparametric methodology. The article extends work of Bound and Krueger, and the results confirm many of the findings in Bound and Krueger. Three new findings are of interest: there is a higher measurement error in cross-sectional samples than in panels. The negative relationship between measurement error and earnings is driven largely by overreporting among low earners. Median response errors are not related to earnings.

This paper uses a simultaneous model of census tract population and employment to study the economic impacts of Atlanta’s MARTA rail transit system. The results indicated that MARTA has had no discernible impact on total population or employment in station areas, but it has altered the composition of employment in these areas in favor of the public sector.

Validation of Food Stamp program participation by Marquis and Moore revealed net bias of 13% in mean estimates of participation in the 1984 Survey of Income and Program Participation. We extend the analysis, incorporating demographic and economic covariates in models for underreporting and overreporting. We use the resulting models of response error to refine estimates of participation models for underreporting and overreporting. We use the resulting models of response error to refine estimates of participation models using pseudo-maximum likelihood estimation methods. We find that the probability of underreporting rises with increased family income. Estimates of the participation model indicate the presence of substantial bias in income- and asset- related parameters compared to values obtained without incorporating response error.

In this paper I examine identification and estimation of mean regression models when a binary regressor is mismeasured. I prove that bounds for the model parameters are identified and provide simple estimators which are consistent and asymptotically normal. When stronger prior information about the probability of misclassification is available, the bounds can be made tighter. Again, a simple estimator for these cases is provided. All results apply to parametric and nonparametric models. The paper concludes with a short empirical example.