The Impact of Omitting Promotion Variables on Simulation Experiments

Using store-level scanner data, elasticity matrices are estimated using a twotiered demand system. Two basic models are estimated, one with promotion variables and one without. Differences between the estimates across the two models are statistically significant. However, when the elasticities are used as 'inputs' into several simulation exercises, there are only small differences in merger effects and patent damage estimates.The results suggest that the differences are not 'economically' significant.