In sum, cocaine consumption appears to respond to cocaine price changes, but the range of elasticity estimates is wide, and all except those from the Grossman and Chaloupka (1998) models that include individual fixed effects come from models that fail to control for geographic fixed effects.
Omitting geographic fixed effects entirely doubles the cocaine price elasticity and triples the marijuana arrest rate elasticity while further decreasing the cocaine arrest rate elasticity.
(23) These models exclude geographic fixed effects because state effects are not identified when the state-level price measure is included.
First, caution is advised when interpreting results of similar analyses that do not control for geographic fixed effects.
(3.) As long as respondents do not move, geographic fixed effects drop out of an individual fixed effects model because they are time-invariant.