Monetary policy in econometric models: primacy of theory over empirics.

Summary Based on the limits of econometrics highlighted in the debates on monetary policy since the 1960s, this thesis aims to show the primacy of theory over empirics and that econometrics cannot be decisive in challenging theory. We adopt an approach based on epistemological arguments to show that these debates go beyond the theory/empiry divide and include a difference of vision as to the usefulness of an empirical model. The research program of the Cowles Commission was built around a particular articulation of three fundamental elements. A theoretical frame of reference derived from Keynes's General Theory, a formal model based on the relative consensus around the IS-LM scheme and econometric techniques for estimating the parameters of this model. It is the nature and degree of interdependence between the above three elements that are questioned by monetarists and VAR modellers. While Keynesians make a clear distinction between the theoretical model and the estimated model, for monetarists this distinction is not clear and does not seem relevant. Sims (1980) criticizes the Cowles Commission's structural models for containing too many theoretical assumptions that have not been empirically tested. He suggests that the exogeneity assumptions be subjected to direct and precise econometric tests. However, the empirical indeterminacy of causality in a VAR model, linked to the problem of observational equivalence (Basmann, 1965), imposes the adoption of an identification scheme based on theoretical a priori to identify monetary policy shocks. This is an extreme case of the problem of the underdetermination of theory by data raised by the Duhem-Quine thesis (Duhem, 1906, Quine, 1951). In addition, Hoover (2009) notes that the analysis of impulse responses in a VAR provides a good example of what Cartwright (2007) refers to as the ''imposter counterfactual.'' The development of Error Correction Models and cointegrated VAR models has allowed for a renewed analysis of monetarist propositions. However, the links between cointegration proposals and the notions of long-run equilibrium and short-run disequilibrium are rarely interpreted within the framework of a rigorous and fully specified theoretical model. For Faust and Leeper (1994), identifying a model by imposing constraints may not be fruitful when economic theory does not clearly distinguish between short- and long-run dynamics. Faust and Whiteman (1997) note the absence of an arbitration criterion in these approaches when there is a conflict between the theoretical principle and the fit to the data, if not a subordination of the theory to econometrics. Alongside the problem of identification, Lucas's (1976) critique constitutes the second fundamental criticism of econometric models. Lucas (1980, 1986) adopts a new epistemological stance by considering the theoretical model as a "fiction" and no longer as a set of propositions about the behaviour of a real economy. He defends the idea of explaining the cycle in terms of the discipline of equilibrium (Lucas, 1977). The DSGE models, which constitute the basic models of the New Synthesis, are strongly influenced by Lucasian methodology and are in line with the RBC models (Taouil, 2011). Benati and Surico (2009) established the superiority of DSGEs over structural VARs (SVARs). This failure of SVARs is a direct consequence of the inter-equation restrictions imposed by the rational expectations assumption, as initially raised by the critique of Sargent (1979).