Why Agnostic Sign Restrictions Are Not Enough: Understanding the Dynamics of Oil Market VAR Models

Working Paper: CEPR ID: DP7471

Authors: Lutz Kilian; Daniel P. Murphy

Abstract: Sign restrictions on the responses generated by structural vector autoregressive models have been proposed as an alternative approach to the use of exclusion restrictions on the impact multiplier matrix. In recent years such models have been increasingly used to identify demand and supply shocks in the market for crude oil. We demonstrate that sign restrictions alone are insufficient to infer the responses of the real price of oil to such shocks. Moreover, the conventional assumption that all admissible models are equally likely is routinely violated in oil market models, calling into question the use of median responses to characterize the responses to structural shocks. When combining sign restrictions with additional empirically plausible bounds on the magnitude of the short-run oil supply elasticity and on the impact response of real activity, however, it is possible to reduce the set of admissible model solutions to a small number of qualitatively similar estimates. The resulting model estimates are broadly consistent with earlier results regarding the relative importance of demand and supply shocks for the real price of oil based on structural VAR models identified by exclusion restrictions, but imply very different dynamics from the median responses in VAR models based on sign restrictions only.

Keywords: demand shocks; identification; median response; oil market; sign restriction; supply shocks; vector autoregression

JEL Codes: C68; E31; E32; Q43


Causal Claims Network Graph

Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.


Causal Claims

CauseEffect
sign restrictions alone (C29)relative importance of oil demand and supply shocks (Q43)
median responses derived from posterior distribution of admissible models (C51)misleading results (C52)
additional identifying information (Y90)improvement of structural model identification (C51)
fluctuations in real price of oil (Q31)driven by oil demand shocks (Q43)

Back to index