SUMMARY OF 2025 KANSAS ECONOMETRICS WORKSHOP

                                                             SUMMARY:

The 2025 Kansas Econometrics Workshop, organized by the Department of Economics of University of Kansas (KU), was successfully held on April 26 at The KU Union on the Lawrence Campus with more than 30 attendees, including some undergraduate and Master as well as Ph.D. students. 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 as well as data science (machine learning methods). The main purpose of this series of workshops is to promote the advanced econometric research for faculty and graduate students at KU. The workshop included 12 invited speakers, including the famous econometricians and statisticians, and some young scholars. The following is the summary for each presentation.

Session I (Financial Econometrics)

Liang Peng - “Systemic Risk: CoVaR and Comovement”

Liang Peng is the Thomas P. Bowles Jr. Chair of Actuarial Science at the Maurice R. Greenberg School of Risk Science and a professor of risk management and insurance in the Robinson College of Business at Georgia State University. He is a fellow of both Institute of Mathematical Statistics and American Statistical Association. His research interests include extreme value theory, nonparametric statistics and time series analysis with various applications in finance, insurance and risk management. At the 2025 Kansas Econometrics Workshop, he presented a paper entitled “Systemic Risk: CoVaR and Comovement”. In his presentation, he introduced a novel systemic risk measure, termed CoVaRCM, which integrated information on both comovement and relevant predictors to evaluate the joint impact of two entities’ losses. Since the comovement event, conditioned on predictors, had zero probability, the study employed a three-quantile regression model to facilitate efficient statistical inference. Moreover, two comparison metrics were proposed to benchmark the performance of CoVaRCM against the more conventional CoVaR framework. The empirical analysis demonstrated the significant role of comovement in amplifying systemic risk. Finally, he discussed a statistical inference approach tailored to systemic risk-driven portfolio selection, providing further insights into the practical implications of the proposed methodology.

Viktor Todorov - “Observable versus Latent Risk Factors”

Viktor Todorov is the Harold H. Hines Jr. Professor of Risk Management and Professor of Finance at the Kellogg School of Management, Northwestern University. Professor Todorov currently serves as a Co-Editor of Journal of Econometrics and has been on the editorial board of several leading academic journals, including Econometrica. His research interests are in the areas of theoretical and empirical asset pricing, econometrics and applied probability. His recent work focuses on the robust estimation of asset pricing models using high-frequency financial data as well as on the development and application of parametric and nonparametric estimation methods for studying risks and risk pricing using derivatives markets data. At the 2025 Kansas Econometrics Workshop, he presented a paper entitled “Observable versus Latent Risk Factors”. In his presentation, he investigated the temporal stability of local linear projection coefficients that map observable risk factors to latent ones, where the latent factors were extracted from the cross-section of asset prices using PCA. The objective was to assess whether conventional linear asset pricing methods could be reliably applied over time and across different horizons. To this end, he proposed a nonparametric test based on the idea that, under the null hypothesis, residuals from global linear projections should remain locally uncorrelated with the PCA-derived factors. The asymptotic properties of the test were established under a joint in-fill and large cross-section asymptotic framework. Through empirical analysis, the study demonstrated that a stable linear relationship between the market volatility factor and the latent systematic risk factors embedded in stock returns persisted only over short intervals, specifically over a horizon of one trading day. This finding suggested important limitations for the use of static asset pricing techniques in dynamic financial environments.

Zhipeng Liao – “Testing for the Minimum Mean-Variance Spanning Set”

Zhipeng Liao is a Professor in Economics at UCLA. He currently serving as an associate editor for Journal of Econometrics. His research develops statistical methods to evaluate different economic models, and to make inference on time series models with nonstationary data, and to make robust inferences from semi/nonparametric models. At the 2025 Kansas Econometrics Workshop, he presented a paper entitled “Testing for the Minimum Mean-Variance Spanning Set”. In his presentation, he addressed the problem of determining the minimum spanning set (MSS) — the smallest collection of risky assets capable of reproducing the full mean-variance efficient frontier. He formally established the conditions under which the MSS was identified and proposed a new estimation and inference framework tailored for this purpose. The theoretical results guaranteed that the MSS estimator asymptotically recovered the true MSS with high confidence, ensuring reliable statistical inference. Through extensive Monte Carlo experiments, the method was shown to exhibit strong finite-sample performance. Applying the approach to empirical data, he examined the relative significance of individual stock momentum, factor momentum, and other well-known stock return factors. The findings emphasized the pivotal role of factor momentum, along with several stock-specific factors, in driving mean-variance efficiency. Moreover, the analysis shed light on the distinct contributions of each factor and offered a systematic ranking, providing new perspectives for understanding their importance in asset pricing and portfolio construction.

Session II (Applied Econometrics)

Yuya Sasaki - “Genuinely Robust Inference for Clustered Data”Yuya Sasaki is the Brian and Charlotte Grove Chair and a professor at Vanderbilt University, whose research focuses on econometrics, particularly in machine learning, big data analysis, cluster-robust inference, and panel data analysis. He has also served in editorial roles for several prestigious journals and currently, Professor Sasaki is serving as the Editor-in-Chief of Econometric Reviews. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Genuinely Robust Inference for Clustered Data”. In his presentation, he shows that conventional cluster-robust inference methods are often inconsistent when clusters are unignorably large. A necessary and sufficient condition for consistent inference is derived, but it is commonly violated in practice — for instance, in 77% of empirical studies published in American Economic Review and Econometrica (2020–2021). To overcome this diffculty, Professor Sasaki proposes two alternative methods: (i) score subsampling, which preserves the original estimator while achieving robustness, and (ii) size-adjusted reweighting, which slightly modifies the estimator but is easily implementable in software and remains valid even under heavy-tailed cluster size distributions like Zipf’s law. Simulation evidence confirms the reliability and broad applicability of these methods.

Iván Fernández-Val - “Conditional Rank-Rank Regression”

Iván Fernández-Val is a Professor of Economics at Boston University, whose research focuses on econometrics and labor economics, with recent work on nonlinear panel data, distributional and causal methods, and the application of machine learning to causal inference. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Conditional Rank-Rank Regression”. During the presentation, he introduces conditional rank-rank regression to better measure within-group persistence when studying relationships between economic variables. Unlike standard rank-rank regression, which uses unconditional ranks, the conditional version uses ranks adjusted for covariates, making within-group interpretations clearer. The difference between conditional and unconditional coefficients captures between-group persistence. The paper develops a flexible estimation method using distribution regression and establishes large-sample inference results. An application to Swiss intergenerational income mobility shows that persistence is stronger between fathers and sons than fathers and daughters, with within-group factors explaining 62% of persistence for sons and 52% for daughters. Greater persistence is also found in smaller families and families with highly educated fathers.

Li Gan – “Tiered Migration: Education Quality, Consumption Variety and Family Strategies in Development”

Li Gan is the Clifford A. Taylor, Jr. Professor in Liberal Arts and a Professor of Economics at Texas A&M University and the Director of the Survey and Research Center for China Household Finance at Southwestern University of Finance and Economics in Chengdu. His research primarily focuses on applied microeconomics. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Tiered Migration: Education Quality, Consumption Variety and Family Strategies in Development”. He examines rural migration patterns in China and finds that migration typically occurs in stages, with small towns acting as crucial intermediate destinations rather than direct moves from villages to cities. Using night light data and household surveys, the study shows that settlements of around 4,000 residents experienced the fastest growth from 2010 to 2020, while smaller villages declined. A dynamic model captures household decisions balancing income, consumption variety, education quality, and family unity, revealing that towns offer higher effective education quality than cities due to urban access barriers. The analysis highlights that improving education quality in towns, rather than cities, would lead to greater welfare gains and more sustainable development, and emphasizes the need for family-centered policies to mitigate the negative effects of split-family living arrangements.

Session III (Time Series Econometrics)

John Chao - “Consistent Estimation, Variable Selection, and Forecasting in Factor-Augmented VAR Models”

John Chao is a professor ofeEconomics at the University of Maryland, whose research focuses on IV regressions with many instruments, Bayesian econometrics, and model selection methods for nonstationary time series analysis.He also serves as an Associate Editor of the Econometrics Journal. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Consistent Estimation, Variable Selection, and Forecasting in Factor-Augmented VAR Models.” In his presentation, he introduced a consistent variable selection method for high-dimensional datasets within a framework where latent factors are estimated for dimension reduction. This method is particularly valuable when many irrelevant predictors are present, as it provides consistent estimates of both the factors and the number of irrelevant variables. When the constructed factors are used in an estimated FAVAR model, the approach enables consistent estimation of conditional mean forecasting functions. Monte Carlo simulations and a real-time FAVAR forecasting exercise demonstrate that this method often outperforms approaches that use all variables or apply simple thresholding methods, even in small samples. These findings suggest that the proposed method offers a useful complement to existing variable selection techniques.

Leland E. Farmer – “Disagreement About the Term Structure of Inflation Expectations”

Leland E. Farmer is an associate professor of economics at the University of Virginia, specializing in macroeconomics, asset pricing, and econometrics. His most recent research focuses on understanding how models of learning can explain apparent anomalies in data from surveys of professional forecasters. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Disagreement About the Term Structure of Inflation Expectations.” In this work, he proposes a parsimonious framework for analyzing individual inflation expectations across different horizons. The model captures expectations using two key factors — level and slope — and decomposes them into contributions from long-term beliefs, public information, and private information.Using individual-level data from the Survey of Professional Forecasters, the study finds that during normal times, long-horizon disagreement stems mainly from long-term beliefs, while short-horizon disagreement is driven by private information. However, during economic downturns, heterogeneity in responses to public information becomes the dominant source of disagreement at all horizons. When forecasters interpret public information differently, monetary policy responses are delayed and a price puzzle emerges, highlighting the critical role of clear and effective monetary policy communication in anchoring inflation expectations. 

Zeqin Liu – “Dual-Pillar Regulation and Growth-at-Risk: Insights from Dynamic Quantile Treatment Effects”

Zeqin Liu is an associate professor at the School of Statistics, Shanxi University of Finance and Economics, and a visiting scholar at the University of Kansas. Her primary research interests lie in macroeconomic policy evaluation and econometrics. At the 2025 Kansas Econometrics Workshop, she presented the paper entitled “Dual-Pillar Regulation and Growth-at-Risk: Insights from Dynamic Quantile Treatment Effects.” The study examines the role of the dual-pillar policy framework — monetary and macroprudential policies — in promoting macroeconomic and financial stability. Using a dynamic quantile treatment effects approach and focusing on China's experience from 2007 to 2023, the paper analyzes how policy combinations impact the distribution of economic growth. The study finds asymmetric policy effects across different quantiles: expansionary monetary policy paired with neutral or expansionary macroprudential policy mitigates recession risks but may fuel overheating during booms; expansionary macroprudential policy cushions downturns particularly at low quantiles. Tightening monetary policy, especially when combined with contractionary macroprudential policy, tends to amplify recession risks but helps control overheating. Notably, expansionary monetary policy combined with contractionary macroprudential policy achieves a balanced effect, easing recession risks while containing financial risks. These results highlight the need for flexible and coordinated application of the dual-pillar framework to achieve stable growth and enhance financial resilience.

Session IV (Theoretical Econometrics)

Peter R. Hansen – “Moments by Integrating the Moment-Generating Function”

Peter R. Hansen is the Henry A. Latané Distinguished Professor at the University of North Carolina, Chapel Hill. His research focuses on forecasting and volatility modeling. His contributions include the realized kernel estimator, the realized GARCH framework, and the model confidence set. He has been recognized among the world's most influential scientific minds. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Moments by Integrating the Moment-Generating Function”. In his presentation, he introduced a novel approach for computing integer, fractional, and complex moments using a complex extension of the moment-generating function (MGF). The method, called CMGF, avoids the use of MGF derivatives by expressing moments as integrals over complex contours. This framework is especially useful in dynamic models where standard analytical tools are limited. Applications include the normal-inverse Gaussian distribution, Heston-Nandi GARCH, and autoregressive models for volatility and jump processes. The method is shown to be significantly faster and more accurate than simulation-based alternatives. Additionally, the approach yields a new integral representation of the reciprocal gamma function, offering broader mathematical insights beyond econometrics.

Boris Hanin – “Bayesian Inference with Deep and Wide Neural Networks”

Boris Hanin is an Assistant Professor at Princeton University in the Department of Operations Research and Financial Engineering. His research focuses on the mathematical theory of deep learning, particularly in understanding how neural networks behave and generalize in highly overparameterized regimes. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Bayesian Inference with Deep and Wide Neural Networks”. In his presentation, he presented joint work with Alexander Zlokapa that offers a solvable model of feature learning in deep and wide fully connected neural networks under Bayesian inference. He introduced the concept of effective depth—defined as the number of samples times the depth-to-width ratio—as the key parameter governing posterior behavior. In the infinite-width limit, the model recovers kernel regression behavior. However, when the effective depth is strictly positive, the posterior corresponds to a data-dependent kernel method. This kernel is explicitly computable and reflects a learned feature map, thereby providing the first model where learning can be fully characterized as depth, width, input dimension, and sample size all diverge. The work also includes a perturbative analysis for weakly nonlinear networks, showing that the same parameter controls their departure from Gaussian process behavior.

Xinwei Ma – “Covariate Adjusted Response Adaptive Design with Delayed Outcomes”

Xinwei Ma is an assistant professor of economics at the University of California, San Diego. His research focuses on econometric theory, particularly in causal inference, nonparametric and semiparametric methods, and adaptive experimental design under complex data environments. At the 2025 Kansas Econometrics Workshop, he presented the paper entitled “Covariate Adjusted Response Adaptive Design with Delayed Outcomes”. In his presentation, he introduced a new approach to Covariate-Adjusted Response Adaptive (CARA) designs that addresses the challenge of delayed outcome data. Traditional CARA designs assume immediate responses, but delayed observations can undermine treatment reassignment and statistical efficiency. He proposed a fully forward-looking design that dynamically updates treatment assignments as the delay mechanism is progressively estimated during the experiment. Using a novel semiparametric efficiency framework, he showed that the classic Neyman allocation is no longer optimal when delays are present. His method involves sequential estimation of auxiliary quantities and conservative extrapolation for unobserved data, ensuring robust and efficient inference. Theoretical results and simulation studies demonstrate that this approach significantly improves both statistical power and participant welfare in multi-stage adaptive experiments with delayed outcomes.