The treated store was in a fast-growing suburb

Question

A chain remodels Store A and uses Store B as a control, estimating the remodel's effect with difference-in-differences. The estimate is positive. Then you notice that Store A sits in a booming new suburb whose population and incomes have been rising for years, while Store B is in a mature, flat neighborhood.

Explain why this makes the difference-in-differences estimate unreliable, and how you would check.

Accepted Answer

The stores were on different trajectories before the remodel. Store A is in a fast-growing area, so its sales were already climbing year after year for reasons that have nothing to do with the remodel: more residents, rising incomes, new development. Store B, in a flat neighborhood, was not growing. Difference-in-differences takes the extra rise in Store A beyond Store B and calls it the remodel effect, but a chunk of that extra rise is just the suburb's ongoing growth that would have occurred anyway. Why parallel trends fails. The assumption is that, without the remodel, Store A and Store B …

The treated store was in a fast-growing suburb

Your answer

More Statistics questions