Intended for

Empirical researchers with interests in using panel data, particularly panel data for firms, and especially in the estimation of production functions.

Prerequisites

Econometrics at an advanced undergraduate or basic graduate level. The required background knowledge of econometrics is at the level of W. H. Greene’s textbook Econometric Analysis (8th edition, 2018, Pearson) or J. M. Wooldridge’s textbook Introductory Econometrics: A Modern Approach (7th edition, 2019, South-Western). For example, participants are expected to be familiar with such concepts as linear regression, instrumental variables, and autoregressive models.

Overview

The methods part of the course will cover econometric methods used to estimate dynamic models from panel data with observations for many individuals or firms, but only a small number of time periods. This will include first-differenced GMM and system GMM estimators, of the kind developed in Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998); related specification tests to assess the validity and informativeness of the proposed instruments; and the implementation of these methods in STATA.

The applications part of the course will cover the use of these methods to estimate production functions from panel data on firms or plants, allowing for permanent unobserved differences between firms in the level of productivity and for a persistent residual productivity process, using the approach developed in Blundell and Bond (2000). The course will also discuss advantages and disadvantages of other approaches to estimating production functions, such as the two-stage estimators developed in Olley and Pakes (1996), Ackerberg, Caves and Frazer (2015), and Gandhi, Navarro and Rivers (2020); and introduce some recently developed specification tests to assess the validity of additional assumptions which are required by these two-stage estimators.

Topics

Methods for static panel data models: unobserved heterogeneity, error components models, fixed effects, random effects; within groups, error components GLS, and related estimators; specification tests
Methods for dynamic panel data models: autoregressive models; linear models with predetermined and endogenous explanatory variables; first-differenced GMM, system GMM, and related estimators; initial conditions; identification and unit roots; specification tests
Implementation: panel data analysis using STATA; the xtreg, xtabond2, and xtdpdgmm commands
Applications: estimation of dynamic Cobb-Douglas and translog production function specifications using system GMM; two-stage estimation methods for value added and gross output production functions, when one or more of the inputs is perfectly flexible; specification tests for the validity of first-stage regression specifications

Steve Bond has a B.A. in Economics from the University of Cambridge (1984) and a D.Phil. in Economics from the University of Oxford (1990). He is Professor of Economics at Oxford, Senior Research Fellow at Nuffield College, and Programme Director at the Oxford University Centre for Business Taxation. He is the author, with Manuel Arellano and Richard Blundell, of some of the key papers which developed the econometric methods covered in this course, and their application to the estimation of production functions. His current research interests include specification tests for production functions and the estimation of price-cost markups from production data.