Monte Carlo evidence on the choice between sample selection and two-part models

doi:10.1016/0304-4076(87)90081-9

Journal of Econometrics

Volume 35, Issue 1, May 1987, Pages 59-82

https://doi.org/10.1016/0304-4076(87)90081-9 Get rights and content

Abstract

This Monte Carlo study examines the relative performance of sample selection and two-part models for data with a cluster at zero. The data are drawn from a bivariate normal distribution with a positive correlation. The alternative estimators are examined in terms of means squared error, mean bias and pointwise bias. The sample selection estimators include LIML and FIML. The two-part estimators include a naive (the true specification, omitting the correlation coefficient) and a data-analytic (testimator) variant.

In the absence of exclusion restrictions, the two-part models are no worse, and often appreciably better than selection models in terms of mean behavior, but can behave poorly for extreme values of the independent variable. LIML had the worst performance of all four models. Empirically, selection effects are difficult to distinguish from a non-linear (e.g., quadratic) response. With exclusion restrictions, simple selection models were significantly better behaved than a naive two-part model over subranges of the data, but were negligibly better than the data-analytic version.

References (29)

T. Amemiya
Tobit models: A survey
Journal of Econometrics
(1984)
M. Hurd
Estimation in truncated samples when there is heteroscedasticity
Journal of Econometrics
(1979)
W.G. Manning et al.
The demand for dental care: Evidence from a randomized trial in health insurance
Journal of the American Dental Association
(1985)
D.J. Poirier et al.
On the appropriateness of endogenous switching
Journal of Econometrics
(1981)
W.P. van de Ven et al.
The demand for deductibles in private health insurance: A probit model with sample selection
Journal of Econometrics
(1981)
T. Amemiya
Regression analysis when the dependent variable is truncated normal
Econometrica
(1971)
N. Duan
Smearing estimate: A nonparametric retransformation method
Journal of the American Statistical Association
(1983)
N. Duan et al.
A comparison of alternative models for the demand for medical care
Journal of Economic and Business Statistics
(1983)
N. Duan et al.
Choosing between the sample-selection model and the multi-part model
Journal of Business and Economic Statistics
(1984)
N. Duan et al.
Comments on selectivity bias

A.S. Goldberger

Abnormal selection bias

J.W. Hay et al.

Let them eat cake: A note on comparing alternative models of the demand for health care

Journal of Business and Economic Statistics

(1984)

J. Heckman

Shadow prices, market wages, and labor supply

Econometrica

(1974)

J. Heckman

The common structure of statistical models of truncation, sample selection, and limited dependent variables, and sample estimator for such models

Annals of Economic and Social Measurement

(1976)

Cited by (254)

Is energy aid allocated fairly? A global energy vulnerability perspective
2024, World Development
The escalating climate change crisis is causing many developing countries to become more vulnerable to energy-related challenges, highlighting the issue of fairness in allocating energy aid. Given this context, it is important to ensure countries that face higher energy vulnerabilities receive a greater share of energy aid funding. This study aims to assess the impact of energy vulnerability on energy aid allocation using a double-hurdle model and a panel of 124 countries from 2002 to 2019. We rely on an index of energy vulnerability to understand the differences in the amount of aid available to different energy-vulnerable recipients, and study the transmission mechanisms underlying this relationship. Our results suggest that the allocation of energy aid has been largely fair. Specifically, energy-vulnerable countries are more likely to be selected as recipients of energy aid during the selection stage, and subsequently receive a greater share of this aid during the allocation stage. However, this effect is heterogeneous and asymmetric. Donors tend to allocate energy aid to countries characterized by relatively low quality of government, low income levels, and limited access to energy assistance. Furthermore, disparities exist in the impact of energy vulnerability on different types of energy support. Specifically, countries with higher energy vulnerability receive a greater proportion of energy policy and energy distribution aid, but are allocated a relatively small share of aid to their non-renewable energy-generation sectors. Additionally, the relationship between energy vulnerability and renewable energy generation aid is not statistically significant. Our results highlight the importance of considering energy vulnerability when allocating energy assistance, and have both scholarly and practical significance. Our findings also have important policy implications for donors by providing guidance on how to promote a fairer allocation of energy aid.
Comparisons of observed and unobserved parameter heterogeneity in modeling vehicle-miles driven
2023, Transportation Research Part A: Policy and Practice
This study examines potential observed and unobserved parameter heterogeneity in modeling vehicle-miles driven (VMD). This application is both important in its own right, and a vehicle for exploring conceptual, mathematical, and empirical differences among three finite-segmentation ways of addressing such heterogeneity: deterministic segmentation and (endogenous) switching for observed heterogeneity, and latent class models for unobserved heterogeneity. Based on empirical data from about 3,000 Georgia residents, we model weekly VMD, and identify key explanatory variables as well as the different sensitivities to those variables exhibited by various population segments. The study posits that people have different sensitivities to explanatory variables by residence type (urban versus “less urban”), and supporting evidence was found in the deterministic segmentation and endogenous switching models. The switching regression model characterized those who are more likely to live in urban areas, e.g. workers, pro-environmental, and those who are less favorable to driving. In addition, the estimated error correlations in the endogenous switching model corroborated that there are unobserved factors common to the joint decisions of residential choice and VMD generation. The latent class model identified lower and higher VMD-inclined segments and had a better goodness of fit than the other two models. The results confirm that (1) there is notable heterogeneity in the population with respect to the weight given to relevant factors that influence VMD, and (2) accounting for both observed and unobserved heterogeneities is useful in modeling VMD.
Bivariate symmetric Heckman models and their characterization
2023, Journal of Multivariate Analysis
A sample selection bias problem arises when a variable of interest or response is correlated with a latent variable. This problem is presented when the response variable has part of its observations censored. The Heckman sample selection model is based on the bivariate normality assumption and fits both response and latent variables. Recently, this assumption has been relaxed to more flexible models based on the Student- $t$ distribution, which has appealing statistical properties. In this article, we introduce an extention of the Heckman sample selection model to the wide class of symmetric distributions. In the new class of sample selection models, covariates are used to describe its dispersion and correlation parameters explaining heteroscedasticity and sample selection bias, respectively. We derive mathematical and statistical properties of the introduced model, and estimate its parameters with the maximum likelihood method. The case of the bivariate Heckman-Student- $t$ model, as a special member of the family of symmetric Heckman models, is analyzed. Monte Carlo simulations are performed to assess the statistical behavior of the estimation method. Two real data sets are analyzed to illustrate our results.
Female leadership and borrowing constraints: Evidence from an emerging economy
2022, International Review of Financial Analysis
This study investigates the differences in credit access between male-managed and female-managed firms using two Enterprise Censuses in Vietnam. Our findings reveal that women-managed firms are less likely to borrow from commercial banks than their male counterparts, even when controlling for other determinants such as CEO education and experience, firm size, and ownership. No difference in credit access is documented for firms borrowing from non-commercial banks. Once we control for firm characteristics and CEO demographic factors, approved loan size is higher for firms managed by female CEOs regardless of the borrowing source. Using decomposition analysis, we find firm size contributes most in explaining the difference in credit access between female and male-managed companies.
Entrepreneurs’ legal infractions and hidden information: Evidence from small business bankruptcies
2022, International Review of Law and Economics
We study how entrepreneurs’ attitudes toward risk, as measured by their personal legal infractions, are related to information asymmetry and agency problems prior to bankruptcy, using a sample of 260 small Swedish firms. We find that auditor resignations are more likely in firms where entrepreneurs have legal infractions. Furthermore, we find that legal infractions are negatively related to the likelihood of a firm disclosing its annual report and the quality of the bookkeeping, suggesting that information asymmetry problems are more severe when the entrepreneurs have legal infractions. We also find that creditors’ recovery rates in bankruptcy are lower if the entrepreneurs have legal infractions related to serious traffic offences, such as driving under the influence of alcohol or without a license.
Can money buy friends? Evidence from a natural experiment
2021, European Economic Review
This study examines the causal effect of positive income shocks on friendships in Vietnam. We measure income shock by respondents’ lottery prizes, controlling for spending on the lottery. We find a positive effect from lottery winnings on the number of friends. People tend to increase the number of their non-colleague and long-term friends, rather than the number of friends among colleagues or new friends. The elasticity of the number of friends with respect to lottery winnings is estimated at 0.05%. Our estimate implies that the income shock needed to purchase one additional friend is nearly 5,000 USD.

View all citing articles on Scopus

^∗: We would like to thank Scott Cardell, Emmett Keeler, Lee Lillard, Carl Morris, Joe Newhouse, John Rolph, and Wynard van de Ven for their comments and support. Careful reviews by two anonymous referees have helped to markedly improve this paper. Bob Young and Karl Schutz provided programming support.

View full text

Monte Carlo evidence on the choice between sample selection and two-part models

Abstract

Journal of Econometrics

Journal of Econometrics

Journal of the American Dental Association

Journal of Econometrics

Journal of Econometrics

Regression analysis when the dependent variable is truncated normal

Econometrica

Smearing estimate: A nonparametric retransformation method

Journal of the American Statistical Association

A comparison of alternative models for the demand for medical care

Journal of Economic and Business Statistics

Choosing between the sample-selection model and the multi-part model

Journal of Business and Economic Statistics

Comments on selectivity bias

Abnormal selection bias

Let them eat cake: A note on comparing alternative models of the demand for health care

Journal of Business and Economic Statistics

Shadow prices, market wages, and labor supply

Econometrica

The common structure of statistical models of truncation, sample selection, and limited dependent variables, and sample estimator for such models

Annals of Economic and Social Measurement