Research

High-dimensional Bayesian nonparanormal dynamic conditional model with multivariate volatility applications

2023

Draft

Abstract: This paper proposes a Bayesian approach for the estimation of large conditional precision matrices instead of inverting conditional covariance matrices estimated, using, for example, the dynamic conditional correlations (DCC) approach. By adopting a Wishart distribution and horseshoe priors within a DCC–GARCH(1,1) model, our method imposes sparsity and circumvents the inversion of conditional covariance matrices. We also employ a nonparanormal method with rank transformation to allow for conditional dependence without estimating transformation functions to achieve Gaussianity. Monte Carlo simulations show that our approach is effective at estimating the conditional precision matrix, particularly when the number of variables $(N)$ exceeds the number of observations $(T)$. We investigate the utility of our proposed approach with two real-world applications. First, to study conditional partial correlations among international stock price indices. Second, to test for $\boldsymbol{\alpha}$ in the context of CAPM and Fama-French 5 factor models with a conditional precision matrix-based Wald-type test. The results indicate stable conditional partial correlations through market disruptions. When there are market disruptions, blue chip stocks chosen from S&P 500 daily returns provide statistically significant evidence against the CAPM and Fama-French five models.

Female labor force participation and gender role attitudes, joint with Junghyuk Lee

2023

[Manuscript in preparation]

Bayesian dynamic factor augmented structure learning: cross-sectional dependence for residuals

2021

Draft

Abstract: We propose the Bayesian approach to estimate the dynamic factor-augmented VAR model. As a result, we can obtain the contemporaneous connectedness as a graphical model of the cross-sectional dependence. In this paper, we estimate unobserved factors as a principal component given the known number of factors. Then, we draw factors through the Gibbs sampler using the forward-filtering backward-sampling algorithm. For the transition matrix, we use a rescaled version of the spike and slab priors for our coefficients of lagged variables, which solves the matrices’ collinearity (or possible rank deficiency) when the number of variables is high-dimensional. We check the properties of the estimators derived from the rescaled spike and slab prior by converting the original Bayesian problem into the Frequentists’ ridge estimation problem. We show that the posterior mean asymptotically maximizes the posterior distribution by analyzing the sensitivity of the choice of priors of coefficients. Lastly, we use the fractional Bayes factor to implement the Bayesian graphical model selection based on the graphical VAR. MC simulation shows the performance of our estimation strategy, and we consider weak cross-sectional dependencies in U.S. house prices.

Individual heterogeneity in the returns to schooling: instrumental variable quantile regression

2018

Draft

Abstract: The main focus of this paper is to investigate whether people with varying levels of unobserved ability obtain different earnings based on their years of schooling. This paper's contribution to the literature is to use the instrumental quantile regression method to capture the heterogeneity of returns on the twins' sample while controlling for ability and measurement error biases. After controlling all covariates and biases, the range of estimates is between 9 percent and 15 percent. Although there is a weak identification problem, the results from both the levels and the proxy models are statistically significant. This paper shows the existence of heterogeneity across individuals through the general Wald-type location shift test. This indicates the complementary relationship between education and schooling in the generation of earnings. Furthermore, I check the positive ability bias, negative measurement error, linearity of schooling, and the heterogeneity of returns of other covariates, including age, race, gender, union membership, and tenure.