Marcel Das

A panel data model for subjective information on household income growth

Subjective expectations about future income changes are analyzed, using household panel data. The models used are extensions of existing binary choice panel data models to the case of ordered response. We consider static models with random and fixed individual effects. We also look at a dynamic random effects model which includes a measure for permanent and transitory income. We find that income change expectations strongly depend on realized income changes in the past: those whose income fell, are more pessimistic than others, while those whose income rose are more optimistic. Expected income changes are also significantly affected by employment status, family composition, permanent income, and past expectations. Expectations are then compared to the head of households ex post perception of the realized income change for the same period. The main finding is that rational expectations are rejected, and that in particular, households whose income has decreased in the past underestimate their future income growth.

A structural labour supply model with flexible preferences

We show how non-parametric flexibility can be attained in a structural labour supply model that can be used to analyse all sorts of (non-linear) tax and benefits reforms. The direct utility function is approximated with a series expansion. For given length of the expansion, the model is estimated by smooth simulated maximum likelihood, using Dutch data on labour supply of married females. Estimates of own and cross wage elasticities and tax reform effects suggest that a series expansion of order two is enough. Monte Carlo simulations show that the estimator performs very well, unless there is measurement error in the hours variable.

Expected and realized income changes: Evidence from the Dutch socio-economic panel

Income expectations play a central role in household decision making. In the life cycle model for example, consumption and savings decisions reflect expectations of future income. In empirical applications where direct information on expectations is not available, it is usually assumed that expectations are rational, and reflected by observed future realizations. In this paper, we analyze direct subjective information on expected changes of household income in one panel wave of Dutch families. First, we describe these data and investigate how the expectations can be explained by, among other variables, income changes in the past. Second, we combine these data with information on realized income changes in the next panel wave, and analyze the dierences between expected and realized changes. We find that, on average, households underestimate their future incomes signiantly. In particular, this holds for those families whose income has fallen in the past.

Nonparametric Tests of Panel Conditioning and Attrition Bias in Panel Surveys

Over the past decades there has been an increasing use of panel surveys at the household or individual level. Panel data have important advantages compared to independent cross sections, but also two potential drawbacks: attrition bias and panel conditioning effects. Attrition bias arises if dropping out of the panel is correlated with a variable of interest. Panel conditioning arises if responses are influenced by participation in the previous wave(s); the experience of the previous interview(s) may affect the answers to questions on the same topic, such that these answers differ systematically from those of respondents interviewed for the first time. In this study the authors discuss how to disentangle attrition and panel conditioning effects and develop tests for panel conditioning allowing for nonrandom attrition. First, the authors consider a nonparametric approach with assumptions on the sample design only, leading to interval identification of the measures for the attrition and panel conditioning effects. Second, the authors introduce additional assumptions concerning the attrition process, which lead to point estimates and standard errors for both the attrition bias and the panel conditioning effect. The authors illustrate their method on a variety of repeated questions in two household panels. The authors find significant panel conditioning effects in knowledge questions, but not in other types of questions. The examples show that the bounds can be informative if the attrition rate is not too high. In most but not all of the examples, point estimates of the panel conditioning effect are similar for different additional assumptions on the attrition process.

Methods for Probability-Based Online and Mixed-Mode Panels

This special issue is devoted to discussion of probability-based survey panels that collect data either solely or partly through online questionnaires. Panels of this kind have been around for a long time, though they have been few in number, but recent years have seen several new panels start up in Europe. This has led to renewed interest in the methodology of such panels and also to deeper questioning of the role of these panels. On one hand, the probability-based panels are to some extent competing against cheaper non-probability access panels. On the other hand, the probability-based panels are increasingly being seen as possible alternatives to more expensive probability-based survey methods. In both cases, clients and data users want to better understand the relative advantages and disadvantages of the probability-based panels.

A Comparison of Four Probability-Based Online and Mixed-Mode Panels in Europe

Inferential statistics teach us that we need a random probability sample to infer from a sample to the general population. In online survey research, however, volunteer access panels, in which respondents self-select themselves into the sample, dominate the landscape. Such panels are attractive due to their low costs. Nevertheless, recent years have seen increasing numbers of debates about the quality, in particular about errors in the representativeness and measurement, of such panels. In this article, we describe four probability-based online and mixed-mode panels for the general population, namely, the Longitudinal Internet Studies for the Social Sciences (LISS) Panel in the Netherlands, the German Internet Panel (GIP) and the GESIS Panel in Germany, and the Longitudinal Study by Internet for the Social Sciences (ELIPSS) Panel in France. We compare them in terms of sampling strategies, offline recruitment procedures, and panel characteristics. Our aim is to provide an overview to the scientific community of the availability of such data sources to demonstrate the potential strategies for recruiting and maintaining probability-based online panels to practitioners and to direct analysts of the comparative data collected across these panels to methodological differences that may affect comparative estimates.

Design of Web Questionnaires : An Information Processing Perspective for the Effect of Response Categories

In this article, an information-processing perspective is used to explore the impact of response categories on the answers respondents provide in Web surveys. Response categories have a significant effect on response formulation in questions that are difficult to process, whereas in easier questions (where responses are based on direct recall) the response scales have a smaller effect. In general, people with less cognitive sophistication are more affected by contextual cues. The Need for Cognition and the Need to Evaluate indexes for motivation account for a significant part of the variance in survey responding. Interactions of ability to process information and motivation combine in regulating responses for questions that are more difficult to process. The results hint at a substantial role of satisficing in Web surveys.

Design of Web Questionnaires: A Test for Number of Items per Screen

This paper presents results from an experimental manipulation of one versus multiple-items per screen format in a Web survey.The purpose of the experiment was to find out if a questionnaire s format influences how respondents provide answers in online questionnaires and if this is depending on personal characteristics.Four different formats were used, varying the number of items on a screen (1, 4, 10, and 40 items).To test how robust the results were, and to find out whether or not a specific format shows more deviation in answer scores, the experiment was repeated.We found that mean scores, variances and correlations do not differ much in the different formats.In addition, formats show the same deviation of item scores between repeated experiments.In relation to non-response error, we found that the more items appear on a single screen, the higher the number of people with one or more missing values.Placing more items on a single screen a) shortens the duration of the interview, b) negatively influences the respondents evaluation of the duration of the interview, c) negatively influences the respondents evaluation of the layout, and d) increases the difficulty in completing the interview.We also found that scrolling negatively influences the evaluation of a questionnaires layout. Furthermore, the results show that differences between formats are influenced by personal characteristics.

A Structural Labour Supply Model with Nonparametric Preferences

Nonparametric techniques are usually seen as a statistic device for data description and exploration, and not as a tool for estimating models with a richer economic structure, which are often required for policy analysis. This paper presents an example where nonparametric flexibility can be attained in a fully structural model. A structural labour supply model with a nonparametric specification of preferences is introduced, which can be used for the analysis of all sorts of (non-linear) tax and benefits changes. Moreover, the model can deal with several other problems in estimation of structural labour supply models, such as non-convex tax rules, benefits, unobserved wages of non-workers, and model coherency. The utility maximization problem is solved by discretizing the budget set and choosing the optimal leisure and income combination from a finite set of alternatives. The direct utility function is approximated with a series expansion. For a given length of the expansion, the model is estimated by smooth simulated maximum likelihood. The wage equation is estimated jointly with the labour supply model, and measurement errors in wage rates are allowed for. The model is estimated with Dutch data on labour supply of married females, for various lengths of the series expansion. Estimates of labour supply elasticities and effects of a proposed tax reform suggest that the results do not change much once the order of the series expansion is extended beyond two, even though the second order model is statistically rejected against higher order models. Monte Carlo simulations are used to show that the estimation strategy has remarkably good finite sample properties for the size of our sample. On the other hand they lead to some concern about the potential bias to measurement error in the hours variable.

Can I use a Panel? Panel Conditioning and Attrition Bias in Panel Surveys

Over the past decades there has been an increasing use of panel surveys at the household or individual level, instead of using independent cross-sections. Panel data have important advantages, but there are also two potential drawbacks: attrition bias and panel conditioning effects. Attrition bias can arise if respondents drop out of the panel non-randomly, i.e., when attrition is correlated to a variable of interest. Panel conditioning arises if responses in one wave are in°uenced by participation in the previous wave(s). The experience of the previous interview(s) may affect the answers of respondents in a next interview on the same topic, such that their answers differ systematically from the answers of individuals who are interviewed for the first time. The literature has mainly focused on estimating attrition bias; less is known on panel conditioning effects. In this study we discuss how to disentangle the total bias in panel surveys due to attrition and panel conditioning into a panel conditioning and an attrition effect, and develop a test for panel conditioning allowing for non-random attrition. First, we consider a fully nonparametric approach without any assumptions other than those on the sample design, leading to interval identification of the measures for the attrition and panel conditioning effect. Second, we analyze the proposed measures under additional assumptions concerning the attrition process, making it possible to obtain point estimates and standard errors for both the attrition bias and the panel conditioning effect. We illustrate our method on a variety of questions from two-wave surveys conducted in a Dutch household panel. We found a significant bias due to panel conditioning in knowledge questions, but not in other types of questions. The examples show that the bounds can be informative if the attrition rate is not too high. Point estimates of the panel conditioning effect do not vary a lot with the different assumptions on the attrition process.