Skip to content

Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations

Why this mattered

Arellano and Bond made dynamic panel models practical for settings where the outcome depends on its own past and where unobserved unit-specific effects are correlated with regressors. Before this paper, researchers faced a difficult tradeoff: fixed-effects estimators were biased in short panels with lagged dependent variables, while standard instrumental-variable approaches were often ad hoc or inefficient. The paper showed how generalized method of moments could systematically use lagged levels as instruments for differenced equations, turning assumptions about no serial correlation in the idiosyncratic errors into a usable set of moment restrictions.

The paradigm shift was not just the estimator, but the specification-testing framework around it. By proposing residual-based serial-correlation tests and placing them alongside Sargan over-identification tests and Hausman comparisons, Arellano and Bond gave applied researchers a disciplined workflow for estimating and diagnosing dynamic panel models. This made it newly feasible to study persistence, adjustment costs, employment dynamics, investment, growth, and policy effects in panels with many individuals and relatively few time periods, without requiring strictly exogenous regressors.

Its influence also set the stage for later breakthroughs in dynamic panel econometrics. The Arellano-Bond difference GMM framework became the reference point for extensions such as system GMM, especially Arellano-Bover and Blundell-Bond, which addressed weak-instrument problems when variables are highly persistent. The paper’s broader legacy is that it helped move panel-data econometrics from static control-for-effects models toward a mature toolkit for causal and structural questions involving dynamics, endogeneity, and unobserved heterogeneity.

Abstract

This paper presents specification tests that are applicable after estimating a dynamic model from panel data by the generalized method of moments, and studies the practical performance of these procedures using both generated and real data. The authors' generalized method of moments estimator optimally exploits all the linear moment restrictions that follow from the assumption of no serial correlation in the errors in an equation which contains individual effects, lagged dependent variables, and no strictly exogenous variables. They propose a test of serial correlation based on the generalized method of moments residuals and compare this with Sargan tests of over-identifying restrictions and Hausman specification tests.

Sources