« Back to Results
Marriott Marquis, Mission Hills
Hosted By:
International Association of Applied Econometrics
Advances in Panel Data Econometrics: Theory and Practice
Paper Session
Sunday, Jan. 5, 2020 10:15 AM - 12:15 PM (PDT)
- Chair: M. Hashem Pesaran, University of Southern California
Transformed Estimation for Panel Interactive Effects Models
Abstract
We propose a transformed quasi-maximum likelihood estimation (QMLE) for panel models with interactive effects. The alternative estimator doesn’t need to estimate the interactive effects in the model. It is computationally simple and is consistent and asymptotically normally distributed whether the regressors are exogenous or contain predetermined variables as long as either N or T goes to infinity. The finite sample performance of the transformed QMLE is examined through extensive simulations, and we find the alternative estimator works remarkably well in our designs, regardless of whether the model is static or dynamic, whether the common factors are stationary, cointegrated or structure changing, and whether the idiosyncratic errors are homoskedastic or heteroskedastic or weakly cross-sectionally dependent.Variational Random-Effects for Panel Data
Abstract
A popular approach in panel data models is to specify the distribution of individual heterogeneity parametrically, and to approximate the integrated likelihood using numerical methods. In this paper we explore the use of Gaussian variational approximations, developed in machine learning, to simplify implementation in these settings. We study the frequentist properties of variational estimators as both N and T tend to infinity. We illustrate their finite sample performance through simulations.Forecasting Using Cross-Section Average-Augmented Time Series Regressions
Abstract
There is a large and growing literature concerned with forecasting time series variables using factor-augmented regression models. The workhorse of this literature is a two-step approach in which the factors are first estimated by applying the principal components method to a large panel of variables, and the forecast regression is estimated conditional on the first-step factor estimates. Another stream of research that has attracted much attention is that concerned with the use of cross-section averages as common factor estimates in interactive effects panel regression models. The main justification for this second development is the simplicity and good performance of the cross-section averages when compared to estimated principal component factors. In view of this, it is quite surprising that no one has yet considered the use of cross-section averages for forecasting. Indeed, given the purpose to forecast the conditional mean, the use of the cross-section average to estimate the factors is only natural. The present paper can be seen as a reaction to this. The purpose is to investigate the asymptotic and small-sample properties of forecasts based on cross-section average-augmented regressions. In contrast to existing studies, the investigation is carried out while allowing the number of factors to be known.JEL Classifications
- C1 - Econometric and Statistical Methods and Methodology: General
- C5 - Econometric Modeling