Jonathan B. Hill

HOME PAGE

Home Page of Jonathan B. Hill

Associate Professor of Economics

University of North Carolina – Chapel Hill

CV (pdf)

LINKS

CONTACT INFORMATION

Dept. of Economics

Gardner Hall 208B

University of North Carolina

Chapel Hill, NC 27599-3305

jbhill at email dot unc dot edu.

RECENT RESEARCH INTERESTS

Heavy-Tail Robust Estimation and Inference

This line of research involves M-Estimation, GMM, and GEL when tails are heavy, leading to Gaussian asymptotics. Simultaneously I develop robust methods of hypothesis testing for tests of functional form, white noise, volatility spillover, etc. I also look at robust estimation of coherent risk measures like the Expected Shortfall, robust estimation of Average Treatment Effects when there may be limited overlap in the covariate distribution, and robust variance targeting for heavy tailed GARCH models.

Co-authors: S. Chaudhuri, M. Aguilar, A. Prokhorov, E. Renault

Mixed Frequency Data

I work on problems of causation and indirect inference for multivariate mixed frequency models. The deeper problems are identification of high frequency relationships when some variables are not observe at a high frequency (e.g. GDP), dimensionality problems that arise for stock or flow variables, and asymptotic theory under minimal assumptions.

Co-authors: E. Ghysels, K. Montegi

Unified Inference for (Random Coefficient) Autoregressions

This line of work involves methods and theory for unified estimation for AR and AR/RCA models. The error may be weakly dependent and may have an infinite variance, and the parameter may be random or not for a robust and unified test of a possibly stochastic unit root.

Co-authors: L. Peng

Extreme Value Theory

My primary contributions here are asymptotic theory for tail index and tail dependence estimation under possibly non-standard situations involving dependence and heterogeneity. I develop a general notion of tail dependence, a general weak limit theory for tail arrays, a functional limit theory for a tail index estimator for dependent data, and consistent HAC estimators for tail arrays allowing for non-parametric tail inference. I also work on problems of tail index estimation for a latent or unobserved process where a plug-in estimator is required and may be less than root-n convergent. I have extended my work to first price auctions and to problems that arise with average treatment effects.

Co-authors: A. Shneyerov

Model Specification Tests and Tests when there are Nuisance Parameters

In this area I have interests in testing for functional form where power is asymptotically one against any deviation from the null. I develop new classes of test weights and heavy tail robust tests. I also develop a general occupation time p-value test that can handle the nuisance parameter in a fast and convenient way since a bootstrap step is not required, and the test statistic itself is its own p-value. The method is profoundly faster to compute than existing bootstrap p-values, and extends to tests when there are nuisance parameters under either hypothesis. Examples include tests of function form where the nuisance parameter may arise from nonlinearity and/or tail-trimming for heavy tail robustness; GARCH effects; white noise robust to heavy tails; moment existence based on a tail-index estimator process; and causation in mixed frequency models.