Haoge Chang

Welcome!

I am a Ph.D. candidate in Economics at Yale University, specializing in econometrics.

I will be on the job market 2022-2023.

Here is a list of projects I worked on or have been working on.

Publications

  1. “Completion of TP and TN border patterns” (2020) with Charles R. Johnson, Linear and Multilinear Algebra, Volume 68, 1753-1766.

Working Paper

  1. Design-based Estimation Theory for Complex Experiments (Job Market Paper). [Link]

  2. Exact Bias Correction for Linear Adjustment of Randomized Controlled Trials (with Joel A. Middleton and P. M. Aronow, R&R Econometrica). [Link to arXiv]

  3. Approximating Choice Data by Discrete Choice Models (with Yusuke Narita and Kota Saito). [Link to arXiv]

  4. Fast Permutation Inference for Randomized Experiments with Binary Outcomes (with P. M. Aronow and Patrick Lopatto). [Link]

Works in Progress

  1. Semiparametric Analysis of Estimating Average Derivatives with a Generated Regressor and an application to Marginal Policy Relevant Treatment Effects.


Abstract: We study the problem of estimating average derivatives with a generated regressor and apply our results to estimate the Marginal Policy Relevant Effect (MPRTE) parameters. In this problem, the support points of the function are not observed directly and are estimated from the data instead. This problem is motivated by the Marginal Treatment Effects literature (Heckman and Vytlactil, 2005) where the MTE functionals are identified and estimated by differentiating the outcome function with respect to an estimated propensity score. We characterize the semiparametric efficiency bounds for this problem, both with and without an index structure, and propose estimators to achieve them. We point out an important geometric regularity condition for the parameters to be well-defined and the bounds to be nonsingular.


  1. Minimax Estimation in Regression Discontinuity Designs in the Sobolev Space.


Abstract: This paper studies the problem of minimax point estimation for functions in the Sobolev classes. Following Armstrong and Kolesar (2018), Imbens and Wager (2018) and Armstrong and Kolesar (2020), we construct a finite sample minimax linear estimator and provide an honest confidence interval that guarantees correct uniform coverage. Using Reproducing Kernel Hilbert Space (RKHS) techniques and under a bound on the L_2 norm of the curvature of the conditional response functions, we show the optimization problem in Imbens and Wager (2018) can be solved directly without being approximated on a grid. We apply our results to Regression Discontinuity Designs and Regression Kink Designs.


  1. Employee Attrition and Collaboration Disruption (with Jonathan Larson, Sida Peng, Mengting Wan, Longqi Yang)


Abstract: We study the effect of employee attrition on the collaboration patterns of the remaining employees. We study the case of triad collaborations. Using a large telemetry data set and a diff-and-diff design, we document a sizable reduction in collaboration efforts of the remaining two employees when one of the three employees in the triad exits the institution.