Contributions published at Statistics and Empirical Economic Research (Rainer Winkelmann)

The book provides graduate students and researchers with an up-to-date survey of statistical and econometric techniques for the analysis of count data, with a focus on conditional distribution models. Proper count data probability models allow for rich inferences, both with respect to the stochastic count process that generated the data, and with respect to predicting the distribution of outcomes. The book starts with a presentation of the benchmark Poisson regression model. Alternative models address unobserved heterogeneity, state dependence, selectivity, endogeneity, underreporting, and clustered sampling. Testing and estimation is discussed from frequentist and Bayesian perspectives. Finally, applications are reviewed in fields such as economics, marketing, sociology, demography, and health sciences. The fifth edition contains several new topics, including copula functions, Poisson regression for non-counts, additional semi-parametric methods, and discrete factor models. Other sections have been reorganized, rewritten, and extended.

Most economic models are limited to analyzing the behavior of a representative agent and, consequently, make the implicit assumption that either individuals are homogeneous or that individual heterogeneity does not matter for the aggregate outcome. However, recent empirical evidence in experimental economics indicates that under strategic complementarity a minority of irrational agents may, indeed, drive the market's outcome. To avoid potential aggregation bias, researchers in empirical economics should take individual heterogeneity into account, which results in the following trade-off: on the one hand, a representative agent approach, which is parsimonious and easy to interpret, completely neglects heterogeneity, but on the other hand, estimating economic behavior at the individual level, which requires a lot of parameters and results in a plethora of estimates, may demand too much from the data. To escape this dilemma, empirical economists may apply finite mixture models, which offer a compromise between completely ignoring individual heterogeneity and running into difficulties when estimating individual by individual. This thesis comprises four independent applications of finite mixture regression models. The first three experimental studies are part of a comprehensive research project, funded by the Swiss National Science Foundation, and discuss the identification and stability of two different behavioral types of decision makers in the domain of risk. The last essay applies a finite mixture model to the German Socio-Economic Panel to segregate the share of altruists from the rest of the population which is assumed to be selfish. Deutsche Zusammenfassung: Die in der experimentellen Wirtschaftsforschung gängigen ökonometrischen Modelle schätzen häufig das Verhalten eines repräsentativen Individuums. Dem liegt die implizite Annahme zu Grunde, dass sich die Individuen entweder homogen verhalten, oder aber individuelle Heterogenität keinen Einfluss auf das aggregierte Verhalten ausübt. Neuere empirische und theoretische Evidenz zeigt jedoch, dass in imperfekten Märkten - insbesondere unter strategischer Komplementarität - eine Minderheit irrationaler Agenten das Marktgleichgewicht stark beeinflussen kann. Um Verzerrungen bei der Aggregation zu vermeiden, sollten in solchen Situationen individuelle Unterschiede zwingend berücksichtigt werden. Typischerweise stehen aber bei Experimentaldaten zu wenige Beobachtungen pro Individuum zur Verfügung, um komplexe Verhaltensmodelle auf individueller Ebene schätzen zu können. Finite Mixture Modelle bieten hier einen guten Kompromiss zwischen einem Repräsentativen Agenten Modell und einer Schätzung auf individueller Ebene: Sie erlauben es eine bestimmte Anzahl Verhaltenstypen zu identifizieren und jedes einzelne Individuum endogen einem dieser Verhaltenstypen zuzuordnen. Damit erfassen sie einerseits den entscheidenden Teil der individuellen Heterogenität, nämlich die Existenz verschiedener Verhaltenstypen, und benötigen andererseits deutliche weniger Parameter als eine Schätzung auf individueller Ebene. Diese Arbeit umfasst vier unabhängige Anwendungen von Finite Mixture Modellen. Die ersten drei experimentellen Studien sind Teil eines vom Schweizerischen Nationalfonds unterstützten Forschungsprojekts zur Charakterisierung verschiedener Verhaltenstypen bei Entscheidungen unter Risiko. In der vierten Studie wird mit einem Finite Mixture Modell der Anteil an Altruisten in einer Teilstichprobe des Deutschen Haushaltspanels identifiziert.

High parental income, while undeniably causing beneÞts for a child in terms of better access to education and more favorable labor market outcomes, may at the same time increase a childÕs income aspirations and thereby reduce Þnancial satisfaction, ceteris paribus. In this paper, we investigate the relationship between Þnancial satisfaction and parental income with data from the German Socio-Economic Panel. The results indicate that there is indeed a negative well-being externality of parental income, and that children appear to compare their actual income situation with the aspiration level acquired while growing up.

It is a widely held belief that apprenticeship training represents a net investment for training firms, the cost of which needs to be recouped after the training period. A new firm-level data set for Switzerland reveals large variation in net costs across firms and, remarkably, negative net costs for 60 per cent of all firms. We use these data to estimate the effect of net costs on the number of apprentices hired by a firm. The results show that the costs have a significant impact on the training decision but no significant influence on the number of apprentices, once the firm has decided to train. For policy purposes, these results indicate that subsidies for firms that already train apprentices would not boost the number of available training places.

We analyze the completed highest education degree of two birth cohorts (1934-1943 and 1964-1973) in Switzerland, using data from the 1999 Swiss Household Panel. The fraction of tertiary graduates has increased over time, for women more so than for men. Educational attainment depends strongly on the educational attainment of parents. For women, we find that a substantial fraction of the overall increase in participation in tertiary education can be explained by the narrowing gap in participation rates between women with lowly educated parents and women with highly educated parents. Logit models show that financial problems have become more important as an impediment for higher education.

null

We propose a test for the presence of money illusion based on subjective survey information on individual satisfaction with income. Using data from the German Socio-Economic Panel for the period 1993–2003, we cannot reject the null hypothesis of no money illusion.

We discuss regression models for ordered responses, such as ratings of bonds, schooling attainment, or measures of subjective well-being. Commonly used models in this context are the ordered logit and ordered probit regression models. They are based on an underlying latent model with single index function and constant thresholds. We argue that these approaches are overly restrictive and preclude a flexible estimation of the effect of regressors on the discrete outcome probabilities. For example, the signs of the marginal probability effects can only change once when moving from the smallest category to the largest one. We then discuss several alternative models that overcome these limitations. An application illustrates the benefit of these alternatives.

Increasing evidence from the empirical economic and psychological literature suggests that positive and negative well-being are more than opposite ends of the same phenomenon. Two separate measures of the dependent variable may be needed when analyzing the determinants of subjective well-being. We argue that this conclusion reflects in part the use of too restrictive econometric models. A flexible multiple-index ordered probit panel data model with varying thresholds can identify response asymmetries in single-item measures of subjective well-being. An application to data from the German Socio-Economic Panel for 1984-2004 shows that income has only a minor effect on positive subjective well-being but a large effect on negative well-being.

Sorting of people on the labor market not only assures the most productive use of valuable skills but also generates individual utility gains if people experience an optimal match between job characteristics and their preferences. Based on individual data on reported satisfaction with life it is possible to assess these latter gains from matching. We introduce a two-equation ordered probit model with endogenous switching and study self-selection into government and private sector jobs. We find considerable gains from matching amounting to an increase in the fraction of very satisfied workers from 53.8 to 58.8 percent relative to a hypothetical random allocationnof workers to the two sectors.

In this paper, we analyze the completed highest education degree of two birth cohorts (1934-1943 and 1964-1973) in Switzerland, using data from the 1999 wave of the Swiss Household Panel. As expected, the fraction of tertiary graduates has increased over time, for women more so than for men. Also, the educational attainment depends strongly on the educational attainment of parents. We then decompose the overall trend into a parental background effect, a general expansion effect and a distribution effect. For women in particular, we find that a substantial fraction of the overall increase in participation in tertiary education can be explained by the fact that the gap in participation rates between women with lowly educated parents and women with highly educated parents has narrowed. We then investigate the role of financial constraints in explaining these trends. Although the number of individuals suffering financial hardship during youth has declined over time, logit models show that financial problems have become more important as an impediment for higher education.

null

This paper employs recent data for Germany and a new outcome variable to assess the consequences of parental separation on the well-being of youths. In particular, it is considered how subjective well-being, elicited from an ordinal 11-point general life satisfaction question, differs between youths living in intact and non-intact families, holding many other potential determinants of well-being constant using ordered probit regressions. The main finding of this study is that living in a non-intact family has not the hypothesised large negative effect on child well-being.

We discuss regression models for ordered responses, such as ratings of bonds, schooling attainment, or measures of subjective well-being. Commonly used models in this context are the ordered logit and ordered probit regression models. They are based on an underlying latent model with single index function and constant thresholds. We argue that these approaches are overly restrictive and preclude a flexible estimation of the effect of regressors on the discrete outcome probabilities. For example, the signs of the marginal probability effects can only change once when moving from the smallest category to the largest one. We then discuss several alternative models that overcome these limitations. An application illustrates the benefit of these alternatives.

Much progress has been made in recent years in developing and applying a direct measure of utility using survey questions on satisfaction with income and with life in general. In this paper we apply this new type of measurement to the study of money illusion. Using data from the German Socio-Economic Panel for the years 1993 to 2003, we cannot reject the hypothesis of no money illusion.

In present day Germany, one in seven children is raised in a single parent household. We investigate the effect of single parenthood on children’s educational attainment, measured by the school track at the age 14, using ordered probit models. We study whether the effect of living in single parenthood during early or late childhood differs. Finally, we ask whether the family effect operates through resources – fewer income and parental time available for the child –, or through adverse effects on psychological well-being. The data used in this study are a nationally representative sample of 14 year old children drawn from the German Socio-Economic Panel.