Item Response Theory in Stata
Rafal Raciborski
StataCorp
[email protected]
Stata 14 provides several commands for fitting item response theory (IRT) models. IRT has a long history in test development and psychometrics and is now being adopted more broadly in fields such as health services research. In this presentation, I will provide an overview of IRT, demonstrate how to fit models with binary items, and discuss postestimation tools such as plotting characteristic curves and information
functions. I will also demonstrate briefly how to fit Bayesian IRT models using Stata. This is a non-technical talk with an emphasis on concepts and no prior knowledge of IRT or Bayesian statistics is assumed.
Matching in the multivalued treatment environment
Jan Zwierzchowski
Warsaw School of Economics
[email protected]
The literature on statistical methods of program evaluation is mainly focused on estimating effects of binary treatments. Moreover, even papers focused on multivalued treatments effects implicitly assume that there is a certainty about the emergence of an alternative treatment. As an effect the uncertainty concerning the choice of an appropriate hypothetical counterfactual outcome has not been modelled.
A new, generalized concept of counterfactual causality has been proposed – average treatment effect on the treated is defined as a difference between observed income and a convex linear combination of all possible counterfactuals weighted using estimated propensity score values. Under this framework not only counterfactual incomes can be estimated but also the hypothetical emergence of a counterfactual treatment can be modelled, as it depends on similar characteristics as the potential outcomes.
The
mmatch Stata program has been written in order to implement the proposed framework. The proposed concept of causality has been illustrated using the data on unemployment rates and level of formal education using EUSILC data for Poland.
Covariate Balancing Propensity Score Estimator in a Context of Women Labor Supply
Filip Premik
Warsaw School of Economics, Narodowy Bank Polski
[email protected]
Since the seminal paper of Rosenbaum and Rubin, propensity score (PS) has played a significant role in causal inference framework. It aims to indicate similar units which are to be matched or to provide appropriate weights. PS has gained its great popularity by allowing for a dramatic reduction of dimensionality in estimation. Further development of related methods has turned the attention of researchers to the dual nature of PS as a covariate balancing score and conditional probability of treatment assignment. Imai and Ratkovic (2014) exploit the aforementioned duality by deriving a set of appropriate moment conditions and thereby introduce a PS estimator that optimizes the covariate balance – Covariate Balancing Propensity Score (CBPS). The paper introduces a new Stata user-written function CBPS which implements the CBPS method within a generalized method of moments framework. The short description of the estimator and the function is presented. Additionally, an empirical exercise which concerns a relationship between woman‘s fertility and her labor supply using the exogenous variation due to twin births (Rosenzweig and Wolpin, 1980; Angrist and Evans, 1998) is provided. The paper examines also performance of the CBPS method compared to classical PS estimators in unfavorable data environment of high degree of heterogeneity among women, low fraction of twin births and exogeneity of the treatment variable with respect to covariates. Moreover, for the best of author‘s knowledge it is the first paper that concerns the labor supply of Polish women using the first-birth twins identification strategy.
How (not) to use Propensity Score Matching? A guidance for the implementation of PSM in Stata
Gergő Tóth
Central European University, Hungary
[email protected]
Propensity Score Matching (PSM) has become by far the most commonly used matching method to estimate causal treatment effects. The goal of the use of matching is twofold, on the one hand, we use PSM to overcome that counterfactual situation when we want to compare the outcome of the treated observations with the results of the treated observations if they were not treated. Propensity Score helps us to find close matches to compare by using observations’ corresponding features. On the other hand, PSM can be used to reduce the imbalance within the dataset which is an obvious source of model dependence. However, after the researcher has taken a position to apply PSM she is faced with lots of questions related to its implementation - which alternative matching algorithm to choose or trimming to determine the common support - for the actual dataset. The aim of this presentation is to provide a brief summary about the implementation of PSM and show some trade-offs regarding bias and efficiency on a real life dataset.
Using SEM in Stata in analysis of relations between trust and public institutions’ performance in Poland and Germany based on ESS
Magdalena Burdach
Warsaw School of Economics
Jolanta Perek-Białas
Jagiellonian University
[email protected]
The aim of analysis was to check if in the same way the public institutions’ performance is evaluated by two different societies and how it is related with their satisfaction and trust towards the institutions based on data from the European Social Survey (ESS) in Poland (PL) and Germany (GE). The hypothesis of equal coefficients/means within and between countries for adequate variables representing the constructs mentioned above were checked via testing configural, metric and scalar invariance. Our analysis based on three Rounds of ESS (2010, 2012, 2014) separately as well for three Rounds jointly, for single country and between countries (MGCFA with the ADF method of estimation) performed in Stata. The analysis were evaluated for the stability of results obtained in all rounds, and so far the various models without and with restrictions were evaluated at least to obtain the partial invariance and with control the requested quality of models (ie. RMSEA for PL and GE for 2012 was equal to 0.043). Additionally, not only comparability with the latest Round will be presented but as well as how and in what way using various weights available for correct analysis in ESS will change final SEM results. At the end, there will be given conclusions if and how it is possible based on ESS to make cross-country comparisons with SEM analysis in the analyzed topics, in Stata and what we learn from this analysis.
Public-private sector wage gap: an evidence for Poland
Kamila Sławińska
Narodowy Bank Polski and Warsaw School of Economics
[email protected]
The aim of this study is to provide a broad empirical assessment of the public-private wage sector wage gap in Poland. This is done by using the data from the European Union Statistics on Income and Living Conditions (EU-SILC) for years 2005-2012. To measure the differences in earnings between public and private sector employees I apply four different estimation techniques: a linear regression, a Oaxaca-Blinder decomposition, a quantile regression and a quantile decomposition. The preliminary results support the hypothesis of a positive wage premium for public sector workers, which, depending on the used method, varies from 8% to 5% for the OLS and from 15% to 7 % for Oaxaca-Blinder decomposition. Regarding wage differences in the entire earnings distribution, the findings are in line with the results obtained for most Western European countries and suggest that the premium is the highest for the low-paid workers, whereas those at the upper tail of income distribution tend to be penalized for being employed in the public sector. The estimates also indicate an existence of a ’sticky floor’ for the poorly remunerated workers. Focusing on the gap evolution in time, its analysis with respect to relative GDP growth in Poland in years 2005-2012 suggest that the wage gap acts counter-cyclical.
Travel mode choices of the citizens of Łódź – microeconometric analysis
Szymon Wójcik
Department of Econometrics, University of Łódź
[email protected]
Travel mode choices made by citizens of the city are important factor determining congestion, level of pollution and noise, especially in rapidly developing agglomerations. Analysis of these decisions and factors by with they are determined should be considered as meaningful step in the projecting city’s urban and infrastructural policy.
Nowadays, city of Łódź experiences the process of deep infrastructural changes, progressive aging of society and demographic structure shift. Therefore we claim that Łódź can be interesting case of studying travel behaviour of its inhabitants.
The purpose of this study is to identify the determinants of decisions made by the citizens of Łódź in their daily travel activity. The database used in empirical part of this paper was established in the study Quality of life of the citizens of Łódź and its spatial diversification. Such choice of the dataset allows us to include more explanatory variables than in standard travel mode choice studies. To find the determinants of travel behaviour we estimate ordered logistic regression models and where needed their generalized versions. Presenting our results we try to compare the outcomes for different districts of the city of Łódź in order to investigate spatial differences.
The results show that there are significant differences between the determinants of different modes of transport also in spatial dimension. As expected, we observe high impact of socio-demographic variables on mode choice decision. Also the attitudes and opinions concerning the state of city’s infrastructure and effectiveness of the functioning of public transportation system have the effect on the frequency of the usage of particular travel mode.