2023 Kansas Econometrics Workshop
Program for 2023 Kansas Econometrics Workshop
May 5, 2023
Department of Economics, University of Kansas
Venue: Gathering Room 2, The Oread Hotel
Hotel: Oread Hotel, located at 1200 Oread Avenue, Lawrence, KS 66044 (877) 263-6347 or (785) 843-1200 (https://theoread.com/)
Workshop web site: https://econometrics.ku.edu/
7:30-8:30 Breakfast (Gathering Room 2, Oread Hotel)
Session I: Chair, Zongwu Cai, University of Kansas
[1] 8:30-9:05 Cheng Hsiao, University of Southern California
“Statistical Inference for the Low Dimensional Parameters of Linear Regression Models in the Presence of High-Dimensional Data: An Orthogonal Projection Approach”
[2] 9:05-9:40 Hashem Pesaran, University of Southern California
“The Role of Pricing Errors in Linear Asset Pricing Models with Strong, Semi-strong,
and Latent Factors”
[3] 9:40-10:15 Arthur Lewbel, Boston College
“Estimating Social Network Models With Missing Links”
10:15-10:45 Coffee Break
Session II: Chair, John Keating, University of Kansas
[4] 10:45-11:20 Torben Andersen, Northwestern University
“Real-Time Detection of Local No-Arbitrage Violations”
[5] 11:20-11:55 Qiwei Yao, London School of Economics
“Autoregressive Networks”
[6] 11:55-12:30 John Keating, University of Kansas
“Asymmetric Vector Autoregressive Moving Average Models”
12:30-14:00 Lunch (Oread Hotel)
Session III: Chair, Yoon-Jin Lee, Kansas State University
[7] 14:00-14:35 Dacheng Xiu, Chicago University
“Prediction When Factors are Weak”
[8] 14:35-15:10 Qiankun Zhou, Louisianan State University
“Confidence Intervals of Treatment Effects in Panel Data Models with Interactive Fixed Effects”
[9] 15:10-15:45 Ben Sherwood, University of Kansas
“Penalized Quantile Regression”
15:45-16:15 Coffee Break
Session IV: Chair, Shahnaz Parseian, University of Kansas
[10] 16:15-16:50 Karen Yan, Georgia Institute of Technology
“A New Semiparametric Approach for Conditional Quantile Treatment Effects Under Ignorability”
[11] 16:50-17:25 Brantly Callaway, University of Georgia
“Difference in Differences With Time-Varying Covariates”
[12] 17:25-18:00 Shahnaz Parseian, University of Kansas
“Time-Varying Panel Data Models with Latent Group Structures”
18:15-21:00 Reception and Dinner at the Hotel Restaurant
SUMMARY:
The 2023 Kansas Econometrics Workshop, organized by the Department of Economics of KU, was successfully held on May 5 at The Oread Hotel on the Lawrence Campus. This is one of the series of the Kansas econometrics workshops focusing on recent developments of econometric theories and methodologies as well as their applications in both economics and finance. The main purpose of this series of workshops is to promote the advanced econometric research for faculty and students at KU. For this workshop, there were 10 invited speakers from outside of KU campus, including the famous econometricians from the world, Professors Cheng Hsiao (University of Southern California), Hashem Pesaran (University of Southern California), Arthur Lebwel (Boston College), Yiwei Yao (London School of Economics), Torben Andersen (Northwestern University), Dacheng Xiu (University of Chicago), Qiankun Zhou (Louisianan State University), Yoon-Jin Lee (Kansas State University), Karen Yang (Georgia Institute of Technology), and Brantly Callaway (University of Georgia), as well as Professor Ben Sherwood (School of Business, KU) and Professors John Keating and Shahnaz Parseian (Department of Economics, KU). It is so happy to note that there were more 30 participants from KU to the workshop, including about 20 Ph.D. students in economics and finance, and our students are very active to ask questions and to have discussions with the speakers. Therefore, this workshop is very helpful to our Ph.D. students. This workshop is excellent and the quality of presentations by the speakers is highly regarded from their presented topics, including econometric theories and methodologies, financial econometrics, micro-econometrics, and treatment effects & economic policy evaluation.
Cheng Hsiao is a professor at the Department of Economics of University of Southern California. He is a leading figure in econometrics, particularly in time-series analysis and panel data analysis. The title of his talk is “Statistical Inference for the Low Dimensional Parameters of Linear Regression Models in the Presence of High-Dimensional Data: An Orthogonal Projection Approach”, which is co-authored with Qiankun Zhou. He proposes a computationally simple orthogonal projection approach for estimating and inferring low dimensional parameters in a regression model with high-dimensional covariates. The model is written as a partial linear regression, with the error decomposed into two elements - one orthogonal to the regressors and the other expressed as a function of high-dimensional covariates. This approach is easy to implement and avoids the bias issues associated with two-stage regression. The proposed method is compared to other popular estimators through Monte Carlo simulations and empirical applications.
Hashem Pesaran holds the John Elliott Distinguished Chair in economics at the University of Southern California and is known for his research in quantitative analysis of financial markets, national and global macro-econometric modeling, energy demand, and the Middle East economy. He gave a talk on “The Role of Pricing Errors in Linear Asset Pricing Models with Strong, Semi-Strong, and Latent Factors”, a paper that is jointly written with Ron P. Smith. A two-step estimator is proposed to estimate parameters of interest, which are crucial for estimating risk premia and testing alpha. However, this estimator is biased and requires correction. The authors develop a bias-corrected estimator and investigate its properties under various assumptions. They also discuss the implications of factor strengths on the precision of estimating risk premia and show how to estimate the factor mean.
Arthur Lewbel is the inaugural holder of the Barbara A. and Patrick E. Roche Chair in economics at Boston College. His research interest spans the areas of nonparametric and semiparametric estimation, instrumental variables, and measurement error models. He gave a talk on the paper “Estimating Social Network Models with Missing Links”, which is co-authored with Xi Qu and Xun Tang. He proposes a method for estimating social network models with missing links. The traditional two-stage least squares approach requires perfect knowledge of the network structure, which is often unavailable in practice. The proposed method introduces a noisy measure H, which reports some existing links and randomly misses others. A penalized likelihood approach is used to estimate the model parameters by minimizing a penalized log-likelihood function that penalizes deviations of the adjacency matrix H from the true matrix G. The method is illustrated using simulations and an empirical application to peer effects in college roommate assignments.
Torben G. Andersen is the Nathan S. and Mary P. Sharp Professor of Finance, Northwestern University. Professor Anderson is a Faculty Research Associate of the National Bureau of Economic Research (NBER) and an International Fellow of the Center for Research in Econometric Analysis of Economic Time Series (CREATES) in Aarhus, Denmark. In addition, Professor Andersen was elected Fellow of the Econometric Society in 2008, and Fellow of the Society for Financial Econometrics, SoFiE, in 2013, Fellow of the Society for Economic Measurement (SEM) in 2018, Fellow of the International Association for Applied Econometrics (IAAE) in 2020, and Fellow of the Journal of Econometrics in 2021.And he is currently a co-editor of the Journal of Econometrics. Professor Andersen gave a talk titled “Real-time detection of local no-arbitrage violations”. This talk focuses on the task of detecting local episodes of violation of the standard Ito semi martingale assumption for financial asset prices in real-time. The proposed detectors are applied sequentially to continually incoming high-frequency data. A Monte Carlo study demonstrates the theoretical results performs good and an empirical application to S&P 500 index futures data confirms that the detectors is effective in capturing such no-arbitrage violation phenomena.
Qiwei Yao is from Department of Statistics, London school of Economics. Professor Yao's recent research has mainly focused on analyzing complex and high-dimensional time series in the sense that the observation recorded at each time is more complex than a single scalar or a short vector. He has also published in the areas of nonlinear time series, functional time series, econometrics, extreme values of dependent data, non/semi-parametric estimation for conditional distributions, non/semi-parametric regression. In this workshop, Professor Yao talked about “Autoregressive networks”. An AR(1) network model is proposed to depict the dynamic changes explicitly. It also facilitates simple and efficient statistical inference. The model diagnostic checking can be carried out easily using a permutation test. Besides, AR(1) stochastic block model has been investigated in depth as an illustration. Finally, the authors applied three real data sets to illustrate that the models are both relevant and useful.
John Keating is from Department of Economics, University of Kansas. Professor Keating’s research interest involves macroeconomics, time series econometrics, monetary theory and policy, and DSGE modeling. In the workshop, Professor Keating discussed “Asymmetric vector autoregressive moving average models”. A new type of reduced-form model AVARMA is proposed. The reduced model permits each endogenous variable to have a unique lag length common across equations, and each equation’s error to be a univariate moving average process. And it is proved that the AVARMA models have advantages over both traditional VARs and VARMA approaches to multivariate time series modeling.
Dacheng Xiu is a professor of Econometrics and Statistics at Booth School of Business at University of Chicago. His research focuses on developing machine learning solutions to big-data problems in empirical asset pricing. At the 2023 Kansas Econometrics Workshop, he presented a paper entitled “Prediction when factors are weak”. In his presentation, he proposed a new approach based on the supervised PCA (select a subset of predictors that are correlated with the prediction target before PCA is applied). The availability of a large number of predictors also allows people to discard predictors that are not sufficiently informative. They built a new model which involves an additional projection step, and a subsequent iterative procedure over selection, PCA, and projection to extract latent factors. The procedure is justified in an asymptotic scheme where both the sample size and the cross-sectional dimension increases but at potentially different rates. The procedure can consistently handle a whole range of latent factor strength. Those latent factors can play an important role in predicting some macroeconomic index, such as inflation and changes in unemployment.
Qiankun Zhou is an associate professor of Economics with tenure at the Louisiana State University. His research interests are theoretical and applied econometrics, with a focus on panel data econometrics. At the 2023 Kansas Econometrics Workshop, he gave a talk on the paper “Confidence intervals of treatment effects in panel data models with interactive fixed effects”. In his presentation, he introduced an idea that he and his co-author used the factor-based matrix completion technique for panel models to estimate the treatment effects, and then used bootstrap method to create confidence intervals of the treatment effects for treated units at each post-treatment period. This construction doesn’t need specific distribution assumptions on the error terms, or large number of post-treatment periods. These confidence intervals are demonstrated to have asymptotically correct coverage probabilities. He also mentioned that they conducted simulation studies to show that these confidence intervals have satisfactory finite sample performances. The proposed bootstrap procedure works well for constructing confidence intervals for estimated treatments. Finally, the empirical studies lead to treatment effect estimates of similar magnitude and reliable confidence intervals.
Ben Sherwood is a Howard Mid Career associate professor in Business Analytics at the University of Kansas School of Business. His research interests include developing flexible and robust models for high-dimensional data. At the 2023 Kansas Econometrics Workshop, he gave a presentation about “Penalized quantile regression”. In his speech, he proposed a new package to solve the high-dimensional problem that the traditional approach (frame it as a linear programming, similar to what is doe with unpenalized quantile regression). The model is estimated by minimizing a penalized objective function with Huber loss and the group lasso penalty. He provided coordinate descent algorithms and algorithms based on approximating the quantile loss with a Huber-type approximation. Rates of convergence are provided which demonstrate the potential advantages of using the group penalty and that bias from the Huber-type approximation vanishes asymptotically. Simulation and empirical results are provided to demonstrate the effectiveness of the proposed algorithm and support the theoretical results.
Karen Yan is an assistant professor at the School of Economics at the Georgia Institute of Technology. Her research interests include econometrics, applied econometrics, and industrial organization. At the 2023 Kansas Econometrics Workshop, she presented a paper entitled "A New Semiparametric Method for Estimating Conditional Quantile Treatment Effects under Ignorability". In her presentation, she introduced a new assumption, Ignorability, which is essential in studying the effects of conditional quantile treatment. This assumption implies that the treatment is independent of the potential outcome. Karen and her co-authors proposed a CQTE estimator that is straightforward to compute. They also conducted Monte Carlo simulations, which revealed that their estimator performs well and yields estimation MSE that is comparable to the existing check-function-based estimator. Furthermore, they applied their method to examine the quantile effects of maternal smoking on infant birth weight, and their results were illustrated through an empirical application. Overall, Karen's research provides valuable insights into the estimation of conditional quantile treatment effects, and her work promises to contribute significantly to the field of econometrics.
Brantly Callaway is an Assistant Professor at the Economics Department at the University of Georgia. His research focuses on microeconometrics and labor economics, with a particular emphasis on the causal effects of economic policies. He gave a talk on the paper "Difference in Difference with Time-Varying Covariates" at the 2023 Kansas Econometrics Workshop. In his speech, he presented new empirical strategies for two cases: time-varying covariates that evolve exogenously and time-varying covariates that may be affected by the treatment. The two proposed strategies are "providing a doubly robust estimand" and "providing a regression adjustment." These approaches are more robust compared to TWFE due to its weaknesses, such as hidden linearity, linearity conditions on the propensity score, and the problem of "weight reversal." Overall, Brantly's work has significant implications for the field of microeconometrics and labor Economics, and his research promises to advance our understanding of the causal effects of economic policies.
Shahnaz Parsaeian is an Assistant Professor of Economics at the University of Kansas, with research interests in econometrics, both theoretical and applied areas, and macroeconomic forecasting. At the Econometrics Workshop, she presented a talk on her new working paper entitled "Time-Varying Panel Data Models with Latent Group Structures". In her speech, she introduced a framework to jointly estimate and identify latent group structures in time-varying panel data by using a pairwise adaptive group fused Lasso penalty (PAGFL). This approach does not require prior knowledge about the number of groups or the group structures. Moreover, she and her co-author developed a simple and fast iterative algorithm (ADMM) that guarantees convergence to the unique global minimizer. Their study showed that the penalized sieve estimator enjoys the oracle property, and it can consistently estimate the true group structure and the regression functional coefficients. Overall, Shahnaz's research provides valuable insights into the estimation of latent group structures in time-varying panel data, and her work promises to contribute significantly to the field of econometrics.