Microeconometric Models : Applications for Topics in Labor Economics

Abstract

The main objective of this thesis is to examine Thailand’s informal labor and financial markets using intensive individual-level data. Different data structures require different econometric models to optimally extract the information and adjust for possible biases. In this thesis, we propose three microeconometric models to study households’ decisions and their market outcomes. In particular, in the first paper, we developed the evidence-theoretic k-NN rule (ET-kNN) model for rank-ordered data and applied it to predict an individual’s sources of loan. In the second paper, we developed the classifier chains generalized maximum entropy (CC-GME) model for multi-label choice problems and applied it to predict occupational hazards that each labor faces. In the third paper, we developed the multi-level sample selection quantile regression (MS-QR) model and applied it to study wage determination and compensating wage differentials in the informal sector. The contributions of this thesis are twofold. The first is the contribution of the applications on the informal labor and financial markets. The second is the contribution of the proposed methodologies, which can be adapted and applied to study other topics in applied microeconomics. To examine the informal financial market, we study factors determining households’ choices of loan sources and predict whether an individual would borrow from formal or informal sources. Since each individual can sequentially choose to borrow from several sources, the empirical model must extract information from all choices. For this problem, we adapted the nonparametric evidence-theoretic k-Nearest Neighbor rule, which was originally designed for multinomial choice data to rank-ordered choice data. The results show that the characteristics with highest contribution to the prediction of loan sources include total savings, college degree, total income and location, whether urban or rural. The prediction from the rank-ordered choice model outperforms that of the traditional multinomial choice model with only one observed choice. For the informal labor market, we examine two main parts. The first part is to identify factors determining the risk of facing occupational hazards. For this problem, we applied the multi-label classification technique to empirically study discrete choice problems, in which each individual faces more than one hazard. We developed the CC-GME model and the results show that the model is robust to distributional assumption of the errors and provide better predictions compared to other commonly used multi-label classification techniques. The second part is to study the compensating wage differential effects of those occupational hazards. Since heterogeneity is a key characteristic of the informal sector, we employed the quantile regression model. Because the sample selection of the labor force and of the informal sector biases the estimates, we propose the MS-QR model to estimate the wage equation. The results show different wage determinants across quantiles and compensating wage differential effects can only be observed in the middle and lower quantiles.