DP13068 Structural Interpretation of Vector Autoregressions with Incomplete Information: Revisiting the Role of Oil Supply and Demand Shocks: Comment
|Author(s):||Lutz Kilian, Xiaoqing Zhou|
|Publication Date:||July 2018|
|Keyword(s):||Oil market models, oil supply elasticity, structural VAR. identification|
|JEL(s):||C32, E32, Q43|
|Programme Areas:||International Macroeconomics and Finance|
|Link to this Page:||cepr.org/active/publications/discussion_papers/dp.php?dpno=13068|
Recently, Baumeister and Hamilton (henceforth: BH) have argued that existing studies of the global oil market fail to account for uncertainty about their identifying assumptions. They recommend an alternative econometric approach intended to address this concern by formulating priors on the structural model parameters. We demonstrate that in practice BH are unable to parameterize identification uncertainty without falling back on ad hoc prior specifications. They are also unable to show that earlier studies did not impose all relevant identifying information. In fact, to the extent that BH's substantive conclusions differ from earlier studies, these differences do not reflect their use of a superior econometric methodology, but mainly the imposition of a highly unrealistic prior for the global impact price elasticity of oil supply. Once identification uncertainty about the global price elasticity of oil supply is accounted for by specifying a prior more in line with extraneous evidence and economic theory, the substantive results of earlier oil market studies are reaffirmed. We also refute BH's claim that existing oil market studies are invalid or not robust. Finally, we explain why the BH method is not a generalization of all existing methods. It is, in fact, not designed to be applied to state-or-the-art oil market models because key assumptions of the proposed approach are not met in these models.