Working Paper: NBER ID: w22345
Authors: Laurence Levin; Matthew S. Lewis; Frank A. Wolak
Abstract: Daily city-level expenditures and prices are used to estimate the price responsiveness of gasoline demand in the U.S. Using a frequency of purchase model that explicitly acknowledges the distinction between gasoline demand and gasoline expenditures, we consistently find the price elasticity of demand to be an order of magnitude larger than estimates from recent studies using more aggregated data. We demonstrate directly that higher levels of spatial and temporal aggregation generate increasingly inelastic demand estimates, and then perform a decomposition to examine the relative importance of several different sources of bias likely to arise in more aggregated studies.
Keywords: gasoline demand; price elasticity; high-frequency data; panel data
JEL Codes: L91
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| high-frequency data (C55) | clearer picture of demand responsiveness (R22) |
| policy measures (gasoline taxes) (R48) | larger impact on consumption (F62) |
| price change (D41) | stronger consumer response (D18) |
| stronger consumer response (D18) | dissipates after 4 to 5 days (Y40) |
| higher daily gasoline prices (Q31) | decrease in gasoline demand (Q47) |
| 10% increase in gasoline prices (R48) | decrease in consumption by approximately 27% to 35% (D12) |