Jaeho Kim, PhD - 김 재호

I am an assistant professor of Economics at the University of Oklahoma, Norman. My primary fields of research are Bayesian econometrics, and time series econometrics. My recent research interests also include machine learning algorithms and big data analysis.

Publications in Refereed Journal

[1] Bayesian Inference of Regime-Switching ARMA Models with Absorbing States: Dynamics of Ex-Ante Real Interest Rate under Structural Breaks, with C-J Kim at Journal of Business and Economic Statistics, Volume 33, Issue 4, 2015

[2] "Why are Bayesian Trend-cycle Decompositions of U.S.Real GDP so Different?" with Sora Chon (KDI) Forthcoming at Empirical Economics

[3] "An Efficient Sequential Learning Algorithm in Regime-switching Environments” with Sunhyung Lee, Forthcoming at Studies in Nonlinear Dynamics and Econometrics

Working Papers

[4] "Hidden Group Patterns in Democracy Developments: Bayesian Inference for Grouped Heterogeneity” (Original Version 2017, Recent Version 2018) with Le Wang, R & R at Journal of Applied Econometrics


We propose a nonparametric Bayesian approach to estimate time-varying grouped patterns of heterogeneity in linear panel data models. Unlike the classical approach in Bonhomme and Manresa (2015), our approach can accommodate selection of the optimal number of groups and model estimation jointly, and also be readily extended to quantify uncertainties in the estimated group structure. Monte Carlo simulations demonstrate the superior finite-sample performances of our proposed approach. Using our approach, we successfully replicate the estimated relationship between income and democracy in Bonhomme and Manresa (2015) and the group characteristics when we use the same number of groups. Furthermore, we find that the optimal number of groups could depend on model specifications on within-group heterogeneity and discuss ways to choose models in practice.

[5] “Heterogeneous Endogeneity”(Original Version 2015, Recent Version 2016) with Kevin Grier and Pallab Ghosh, R & R at Statistical Papers


We define heterogeneous endogeneity as a case where a potentially endogenous regressor is endogenous for some sub-groups of the data but exogenous for other subgroups. We derive an estimator and test procedure based on a control function approach to deal with the phenomenon. We show that accounting for heterogenous endogeneity can greatly increase the power of endogeneity tests and increase the precision of our estimator over traditional IV. The gains get larger as the instrument gets weaker and as the relative size of the non-endogenous subgroup gets larger. We illustrate our approach with an example using data from Abadie et al. (2002).

[6] "Non-Markovian Regime Switching Models” (Original Version 2017, Recent Version 2018) with CJ Kim, Under Review


This paper introduces a new class of regime switching model with a non-Markovian feature, namely a non-Markovian regime switching model. The latent variable that governs the underlying regime switching dynamics of the observed time series is modeled as a stochastic function of the entire past history of regime changes. We show that a conventionally used Markov switching model cannot properly approximate a non-Markovian switching process and results in biased parameter estimates. Therefore, we develop a Gibbs sampling algorithm to make an exact Bayesian inference on the proposed non-Markovian regime switching model and demonstrate its satisfactory performance via simulation and empirical illustrations.

[7] “Trend-cycle Decompositions of Real GDP Revisited: Classical and Bayesian Perspectives on an Unsolved Puzzle,” (Original Version 2014, Recent Version 2018) with CJ Kim, Under Review


While Perron and Wada (2009) show that postwar U.S. real GDP follows a trend stationary process (TSP) based on the maximum likelihood approach, our analysis based on the Bayesian approach shows that it is a difference stationary process (DSP) with the stochastic trend component explaining most of the variations in real GDP. One goal of this paper is to provide a comprehensive analysis on the sources of such different results and to suggest that more credibility should be conferred to the Bayesian inference. Another goal naturally is to re-investigate the trend-cycle decompositions within the Bayesian framework. This is done by employing a model that incorporates time variation in both the mean and variance. The Bayesian approach to model selection prefers a DSP model to a TSP model, even though the evidence is less than decisive. Empirical results also show that there exists convincing evidence that the cycle from the DSP model, which is small in magnitude and noisy, has out-of-sample predictive power for future output growth at short horizons. The highly persistent TSP cycle without any predictive power may be related to spurious periodicity discussed in Nelson and Kang (1981).

[8] "Permanent and Transitory Shocks to the U.S. Economy: Has Their Importance Been Constant over Time?" (2016) with Sora Chon (KDI)


The main purpose of this paper is to scrutinize the time-varying relative importance of the permanent and the transitory shocks to the U.S. real GDP. After accounting for the cointegration relation between real GDP, consumption, and investment, we find that i) the real permanent shock played a more prominent role than the nominal transitory shock from the 1950s to the 1960s, but the two shocks have almost equally contributed to the stochastic movements of real GDP since the 1970s; (ii) the long-run growth of the U.S. economy has substantially slowed since the recession in 2001. The annualized growth rate of real GDP has declined to approximately 1.6 %, falling from a peak of nearly 3.3 % in the last decade. For the empirical analysis, we employ a multivariate unobserved component model that accommodates stochastic volatility. The multivativariate model is estimated by an efficient particle Gibbs sampler with particle rejuvenation that simultaneously draws latent state variables all at once.

[9] “Efficient Bayesian Inference in Non-linear Switching State Space Models Using Particle Gibbs Sampling Approaches”(Original Version 2016, Recent Version 2018)


This paper develops a new Bayesian algorithm to efficiently estimate non-linear/non-Gaussian state space models with abruptly changing parameters. Within the Particle Gibbs framework developed by Andrieu et al. (2010), the proposed algorithm effectively combines two ideas: ancestor sampling and a partially deterministic sequential Monte Carlo method. In the proposed approach, the discrete latent state variable that governs the switching behavior of a complex dynamic system is deterministically generated to fully diversify particles, and ancestor sampling enables complete exploitation of the various generated particles. Without a large number of particles and sophisticated tailored importance distributions, the newly developed PG sampler is shown to be both easy to implement and computationally efficient, and it substantially outperforms a standard PG technique.

Working in Progress

[1] "Portfolio Sorts" with Drew Creal (U of Notre Dame)

[2] "Bayesian Inference for CART Models`` with Drew Creal (U of Notre Dame)

[3] "Time-varying Group Heterogeneity in Panel Data" with Le Wang (U of Oklahoma)

[4] "Price Discovery for Commodity Markets " with Scott Linn (U of Oklahoma)

[5] "A Unified Framework for Regime Switching Models" with C-J Kim (U of Washington)

[6] "Regime Switching Models in a Data-rich Environment"