THE IMPACT OF THE BEFORE-AFTER ERROR TERM CORRELATION ON WELFARE MEASUREMENT IN LOGIT

We consider random utility models with independent and identical type I extreme value distribution of the error terms. To compute the expectation of the compensating variation it is necessary to consider the correlation of the error terms between the state before the price and quality change and the state after. We investigate the impact of the before-after correlation of the error terms on the expectation of the compensating variation. We consider each error term to be correlated between the before state and the after state independently and identically across alternatives. We prove the theoretical property that in the case without income effect the logsum formula holds for any assumption on the before-after correlation. We use numerical evidence to show that in the case with income effect the variability of the expectation of the compensating variation with the assumption on the before-after correlation increases with the size of the income effect.