Panel and Network Data
Saturday, Jan. 6, 2018 10:15 AM - 12:15 PM
Network and Panel Quantile Effects Via Distribution Regression
AbstractThis paper provides a method to construct simultaneous confidence bands for quantile functions and quantile effects in nonlinear network and panel models with unobserved two-way effects, strictly exogenous covariates, and possibly discrete outcome variables. The method is based upon projection of simultaneous confidence bands for distribution functions constructed from fixed effects distribution regression estimators. These fixed effects estimators are bias corrected to deal with the incidental parameter problem. Under asymptotic sequences where both dimensions of the data set grow at the same rate, the confidence bands for the quantile functions and effects have correct joint coverage in large samples. An empirical application to gravity models of trade illustrates the applicability of the methods to network data.
Nonlinear Panel Data Correlated Random Coefficient Models
AbstractIn this paper we consider identification of the distribution of marginal effects in a model with potentially high dimensional random coefficients in a general nonlinear and nonseparable specification. In this setup, the time invariant random coefficients can be arbitrarily correlated with the covariates of interest and are allowed to enter the structural relationship in any way possible. Our main result shows that the distribution of marginal effects is constructively point identified for the subpopulation of stayers. Due to the arbitrary nonlinearity, this approach generalizes both the existing Panel literature in random coefficients, and allows for models to be identified which are not identifiable in a cross section setup. Building on the identification result, we provide the asymptotic theory for a sample counterparts estimator, analyze the small sample behavior through a Monte Carlo experiment, and provide an application to consumer demand.
Minimizing Sensitivity to Model Misspecification
AbstractWe propose a framework to compute predictions based on an economic model when the model may be misspecified. Our approach relies on minimizing sensitivity of the estimates to the type of misspecification that is most influential for the parameter of interest. We rely on local asymptotic approach where the size of the misspecification is indexed on the sample size, which results in simple rules to adjust the predictions from the reference model. We calibrate the size of misspecification using a detection error probability approach, which allows us to perform systematic sensitivity analysis in both point-identified and partially-identified settings. We study three examples: demand analysis, treatment effects estimation under selection on observables, and panel data models where the distribution of individual effects may be misspecified and the number of time periods is small.
- C14 - Semiparametric and Nonparametric Methods: General
- C23 - Panel Data Models; Spatio-temporal Models