Search

We show that within a life-cycle skill accumulation model, IV identification of the return to schooling parameter is either achieved at any point in the life-cycle where the level of skills accumulated beyond school completion for compliers is exactly equal to the post-schooling skill level of non-compliers (the Skill-Equality condition), or when the skill-ratio is equal to the relative population proportions of non-compliers over compliers (the Weighted-Skill-Ratio condition). As a consequence, it is generally impossible to tie IV identification to any specific phase of the life-cycle and there cannot exist a generally acceptable optimal age to sample earnings for IV estimation. The practical example developed in the paper shows precisely how an instrument may fulfill identification at a multiplicity of ages, and how different instruments may achieve identification with specific sampling designs and fail to do so with others. Within a life-cycle skill accumulation data generating process, identification of the return to schooling requires not only implicit assumptions about the underlying model, but also assumptions about the validity of the specific age sampling distribution implied by the data.

The literature estimating returns to education has often utilized spousal education and parental education as instrument variables (IV). However, due to usual survey designs, both IVs are available only for the individuals whose spouse or parents are present in the same household. The IV estimates based on these selective sub-samples may be inconsistent, even when the IVs satisfy the standard assumptions. In this paper, we examine the empirical relevance of this issue in the Chinese context. To our surprise, unlike the selection issue in other situations, this kind of selection does not appear particularly worrisome, suggesting that the previous IV results are robust. In particular, using China Household Income Project 1995 and 2002, we find that correcting for this potential issue has only a modest impact on the magnitude of the standard IV estimates using parental education as an IV, but a negligible impact on those using spousal education. Using the specification tests proposed, we find that these impacts are generally not statistically significant. These results are further confirmed by our analysis using U.S. data. We believe that these results are of use to both policymakers and practitioners.

In many countries the demand for health care services is of increasing importance. Especially in the industrialized world with a changing demographic structure social insurances and politics face real challenges. Reliable predictors of those demand functions will therefore become invaluable tools. This article proposes a prediction method for the distribution of the number of visits to the medical doctor for a determined population, given a sample that is not necessarily taken from that population. It uses the estimated conditional sample distribution, and it can be applied for forecast scenarios. The methods are illustrated along data from Sidney. The introduced methodology can be applied as well to any other prediction problem of discrete distributions in real, future or any fictitious population. It is therefore also an excellent tool for future predictions, scenarios and policy evaluation.

This paper addresses the estimation of a semiparametric sample selection index model where both the selection rule and the outcome variable are binary. Since the marginal effects are often of primary interest and are difficult to recover in a semiparametric setting, we develop estimators for both the marginal effects and the underlying model parameters. The marginal effect estimator only uses observations which are members of a high probability set in which the selection problem is not present. A key innovation is that this high probability set is data dependent. The model parameter estimator is a quasi-likelihood estimator based on regular kernels with bias corrections.We establish their large sample properties and provide simulation evidence confirminng that these estimators perform well in finite samples.

This paper considers a setting in which the acquisition of human capital entails a change of location in social space that causes individuals to revise their comparison groups. Skill levels are viewed as occupational groups, and moving up the skill ladder by acquiring additional human capital, which in itself is rewarding, leads to a shift in the individual’s inclination to compare himself with a different, and on average better-paid, comparison group, which in itself is penalizing. The paper sheds new light on the dynamics of human capital formation, and suggests novel policy interventions to encourage human capital formation in the aggregate and, at the same time, reduce inter-group income inequality.

This paper provides evidence from Internet-based, large-scale survey data of hypothetical choice experiment on the relative utility hypothesis. The methodology exploited here complements previous empirical results from happiness studies, incentivized choice experiment studies, and neuroscience studies in such a way that methodological problems among previous studies within these fields are resolved. We show that not only the intensity but also the distribution of relative utility are different across specific comparison benchmarks (internal reference group), and across types of reference groups people are facing in the experiments (external reference group). The relative utility effect among Japanese respondents, while shown to exist in the form of jealousy, is found to be not as strong as can validate the Easterlin paradox. Comparison benchmark with daily contacts is related to stronger jealousy. We also provide empirical evidence, which nuances that the reference group is chosen endogenously.

Respondent driven sampling (RDS) is a network sampling technique typically employed for hard-to-reach populations (e.g. drug users, men who have sex with men, people with HIV). Similar to snowball sampling, initial seed respondents recruit additional respondents from their network of friends. The recruiting process repeats iteratively, thereby forming long referral chains. Unlike in snowball sampling, it is crucial to obtain estimates of respondents' personal network size (i.e., number of acquaintances in the target population) and information about who recruited whom. Markov chain theory makes it possible to derive population estimates and sampling weights. We introduce a new Stata program for RDS and illustrate its use.

This note presents the R package bayesGARCH (Ardia, 2007) which provides functions for the Bayesian estimation of the parsimonious and effective GARCH(1,1) model with Student-t innovations. The estimation procedure is fully automatic and thus avoids the tedious task of tuning a MCMC sampling algorithm. The usage of the package is shown in an empirical application to exchange rate logreturns.

The probability of selection into treatment plays an important role in matching and selection models. However, this probability can often not be consistently estimated, because of choice-based sampling designs with unknown sampling weights. This note establishes that the selection and matching procedures can be implemented using propensity scores fit on choice-based samples with misspecified weights, because the odds ratio of the propensity score fit on the choice-based sample is monotonically related to the odds ratio of the true propensity scores.

During the ongoing financial crisis the analysis of similar historical crises has gained more and more attention among economic researchers and forecasters. Existing studies, however, do not tackle the immense heterogeneity that is present in cross-country samples in a formal and consistent way. In this paper, we propose a standardization approach to estimate the typical impact of economic crises. We show that our approach leads to estimates that are much less dependent on the sample used to estimate the typical shape and, hence, should give more reliable information about the typical macroeconomic impact of economic crises.

It is often believed that without instrument, endogenous sample selection models are identified only if a covariate with a large support is available (see Chamberlain, 1986, and Lewbel, 2007). We propose a new identification strategy mainly based on the condition that the selection variable becomes independent of the covariates when the outcome, not one of the covariates, tends to infinity. No large support on the covariates is required. Moreover, we prove that this condition is testable. We finally show that our strategy can also be applied to the identification of generalized Roy models.

This paper presents the R package AdMit which provides functions to approximate and sample from a certain target distribution given only a kernel of the target density function. The core algorithm consists in the function AdMit which fits an adaptive mixture of Student-t distributions to the density of interest via its kernel function. Then, importance sampling or the independence chain Metropolis- Hastings algorithm are used to obtain quantities of interest for the target density, using the fitted mixture as the importance or candidate density. The estimation procedure is fully automatic and thus avoids the time-consuming and difficult task of tuning a sampling algorithm. The relevance of the package is shown in two examples. The first aims at illustrating in detail the use of the functions provided by the package in a bivariate bimodal distribution. The second shows the relevance of the adaptive mixture procedure through the Bayesian estimation of a mixture of ARCH model fitted to foreign exchange log-returns data. The methodology is compared to standard cases of importance sampling and the Metropolis-Hastings algorithm using a naive candidate and with the Griddy-Gibbs approach.

Highly non-elliptical posterior distributions may occur in several econometric models, in particular, when the likelihood information is allowed to dominate and data information is weak. We explain the issue of highly non-elliptical posteriors in a model for the effect of education on income using data from the well-known Angrist and Krueger (1991) study and discuss how a so-called Information Matrix or Jeffreys' prior may be used as a `regularization prior' that in combination with the likelihood yields posteriors with desirable properties. We further consider an 8-dimensional bimodal posterior distribution in a 2-regime mixture model for the real US GNP growth. In order to perform a Bayesian posterior analysis using indirect sampling methods in these models, one has to find a good candidate density. In a recent paper - Hoogerheide, Kaashoek and Van Dijk (2007) - a class of neural network functions was introduced as candidate densities in case of non-elliptical posteriors. In the present paper, the connection between canonical model structures, non-elliptical credible sets, and more sophisticated neural network simulation techniques is explored. In all examples considered in this paper – a bimodal distribution of Gelman and Meng (1991) and posteriors in IV and mixture models - the mixture of Student's t distributions is clearly a much better candidate than a Student's t candidate, yielding far more precise estimates of posterior means after the same amount of computing time, whereas the Student's t candidate almost completely misses substantial parts of the parameter space.

This paper proposes a method to implement maximum likelihood estimation of the dynamic panel data type 2 and 3 tobit models. The likelihood function involves a two-dimensional indefinite integral evaluated using “two-step” Gauss-Hermite quadrature. A Monte Carlo study shows that the quadrature works well in finite sample for a number of evaluation points as small as two. Incorrectly ignoring the individual effects, or the dependence between the initial conditions and the individual effects results in an overestimation of the coefficients of the lagged dependent variables. An application to incremental and radical product innovations by Dutch business firms illustrates the method.

Using data from the European Social Survey (ESS), we examine the link between income and subjective well-being. We find that, for the whole sample of nineteen European countries, although income is positively correlated with both happiness and life satisfaction, reference income exerts a negative effect on individual well-being, a result consistent with the relative utility hypothesis. Performing separate analyses for some Eastern European countries, we also find some evidence of a tunnel effect', in that reference income has a positive impact on subjective well-being. Our findings support the view that in environments with stable income and employment, reference income serves as a basis for social comparisons, whereas in relatively volatile environments, it is used as a source of information for forming expectations about future status.

Selectivity bias caused by protest responses in Contingent Valuation studies can be detected and corrected by means of sample selection models. This paper compares two methods: the Heckman 2-steps method and the full ML, applied to data on forest recreation - where WTP is elicited as a continuous variable. Either method has its own drawback: computational complexity for the ML method, susceptibility to collinearity problems for the 2-steps method. The latter problem is observed in our best fitting specification, with the ML estimator outperforming the 2-steps. In this application, overlooking the effect of protest responses would cause an upwards bias of the final estimates of WTP.

When modelling data generated from a discrete choice contingent valuation question, the treatment of zero bids affects the welfare estimates. Zero bids may come from respondents who are not interested in the provision of the public good; alternatively, some zero-bidders may be protesting about the valuation exercise, but hold positive values for the good. In this paper we investigate the effect of different levels of information on zero-bidders on welfare estimates for the population. We find that different strategies of identification may have non-trivial effects. We recommend use of full debriefing questions for zero-bidders, and use of sample selection models to correct for bias caused by protest behaviour.

I develop a new approach for sample selection problems that allows parametric forms of any type to be chosen for both for the selection and the observed variables. The Generalized Parametric Selection (GPS) model can incorporate both duration and count data models, unlike previous parametric models. MLE does not require numerical integration or simulation techniques, unlike previous models for count data. I discuss application to common duration models (exponential, Weibull, log-logistic) and count models (Poisson, negative binomial). I demonstrate the usefulness of the model with an application to the effects of insurance status and managed care on hospitalization duration data. The example indicates that the GPS model may be preferred even in cases for which other parametric approaches are available.

In recent years a number of panel estimators have been suggested for sample selection models, where both the selection equation and the equation of interest contain individual effects which are correlated with the explanatory variables. We review and compare some of these estimators, and apply them to estimating the return to actual labour market experience for females, using a panel of twelve years. All these estimators rely on the assumption of strict exogeneity of regressors in the equation of interest, conditional on individual specific effects and the selection mechanism. This assumption is likely to be violated in many applications. Also, life history variables are often measured with error in survey data sets, because they contain a retrospective component. We show how non-strict exogeneity and measurement error can be taken into account within the estimation methods discussed.

In this paper the motivation and various concepts of statistical systems for assisting case workers in assigning unemployed persons to active labour market programmes (ALMP) are examined and the particular implementation of such a statistical system in Switzerland, which was introduced in the form of a randomized pilot study, is discussed.