Econometrics and Policy Analysis: A Critique and an Alternative

Econometric modelling is at the core of economic policy analysis. Such models have long been used to forecast the effects of various national economic policy measures or, at least, to inform policy analysts about likely trends such as inflation or employment rates in the absence of policy interventions. Econometric analysis has long been closely associated with equilibrium economic theory. The dominance of rational [Muth, 1961, Lucas, 1972] and then consistent expectations theory [Walters, 1971] starting in the 1970s had its scientific basis in equilibrium theory. More recently, autoregressive conditional heteroskedastic (ARCH) models [Engle, 1982] have been devised to reconcile the experience of unpredictable volatile episodes in the financial markets with rational expectations theory and therefore equilibrium economics. Alongside these developments, co-integration techniques [Engle and Granger, 1987] have been devised on the presumption that there is an unseen data generating mechanism with long run properties supporting a long run solution to econometric models. Recently, Barker et al. [2006, p. 144] have explicitly rejected equilibrium in favour of a “history” approach to the representation of economic growth. The authors evidently believe that the dismissal of economic theory1 leaves econometric technique in tact. Barker et al. use co-integration techniques, thereby to raise the question of whether a long run solution is substantively different from a long run equilibrium. The core argument of this paper is that econometric technique stands and falls with equilibrium economics so that inter alia it does not make much difference whether or not a long run solution is a long run equilibrium. This argument has considerable implications for the ways in which models can and should be used for policy analysis in general. Analysis of policies to address climate change issues provides an edifying example of the nature and importance of these implications. The argument of the paper is developed in the following sequence. First in section 1, I consider a formally based objection to the use of equilibrium in social analysis. This consideration is based on a significant and longstanding body of evidence and well tested implications of that evidence in both social and natural science. These implications, moreover, have led to unexpected simulation results that were subsequently validated with respect to real social statistics. Then, in section 2, I point out the implications of this result for econometrics. That these results render econometrics (and probability-based techniques more generally) inapplicable has been known (though ignored by economists and econometricians) for more than 40 years now. Here, I build on this argument to show why the impossibility of equilibrium implies the otiosity of econometrics. A further problem for econometric technique is the nature of the evidence on which econometricians rely (section 3) and what this implies for the rôle of modelling in policy analysis. The example used to make the relevant points here is the representation by Barker et al. [2006] of endogenous technical change. In the final substantive section 5, I apply the preceding results to the Stern Review and the work used and reviewed therein. It would appear that that whole literature has nothing to offer the analysis of climate change, its social impacts and policies to mitigate both the change and its impacts.