Fast Poisson estimation with high-dimensional fixed effects

Abstract

In this paper we present ppmlhdfe, a new Stata command for estimation of (pseudo) Poisson regression models with multiple high-dimensional fixed effects (HDFE). Estimation is implemented using a modified version of the iteratively reweighted least-squares (IRLS) algorithm that allows for fast estimation in the presence of HDFE. Because the code is built around the reghdfe package, it has similar syntax, supports many of the same functionalities, and benefits from reghdfe's fast convergence properties for computing high-dimensional least squares problems. Performance is further enhanced by some new techniques we introduce for accelerating HDFE-IRLS estimation specifically. ppmlhdfe also implements a novel and more robust approach to check for the existence of (pseudo) maximum likelihood estimates.

EMU and trade: A PPML re‐assessment with intra‐national trade flows

“…All the estimations have been carried out using the Poisson pseudo-maximum likelihood (PPML) estimator with the ppmlhdfe Stata command recently developed by Correia et al (2019).…”

Trade creation and trade diversion of economic integration agreements revisited: a constrained panel pseudo-maximum likelihood approach

“…Poissonnier (2019) proposes the usage of the RAS-algorithm used in the input-output literature to solve the system multilateral resistances efficiently. Lastly, the present paper contributes to the evolving literature proposing fast zig-zag algorithms for the estimation of high dimensional panel models based pseudo-demeaning or separate up-dating of fixed effects (see Correia et al 2020;Larch et al 2019;Stammann 2018).…”

Dealing with the Log of Zero in Regression Models

“…For instance, in Stata, the package ppml(Santos Silva and Tenreyro (2015)) is straightforward to use and can account for fixed effects when they are not too many of them. The package ppmlhdfe(Correia et al (2019)) allows one to deal with high-dimensional fixed effects. A recent example of application can be found inHead and Mayer (2019).…”