Estimated responses of real oil prices and US gross domestic product (GDP) to oil supply disruptions vary widely. We show that most variation is attributable to differences in identification assumptions and in the model specification. Models that allow for a large short-run price elasticity of oil supply imply a larger response of oil prices and a larger, longer lived contraction in US real GDP. We find that, if we condition on a range of supply elasticity values supported by microeconomic estimates, the differences in the oil price responses diminishes. We also examine the role of lag length, of using pre-1973 data, alternative measures of real economic activity and using the median response function instead of the modal structural model.