Design a switchback and choose block length

A switchback experiment here should assign pricing treatment at the airport-time block level, not at the rider level. The analytical unit should be a small time block, because the marketplace has temporal spillovers and strongly autocorrelated demand. Block length and rotation schedule Use a primary block length of $$ L = 120 \text{ minutes}. $$ The empirical demand ACF falls below 0.1 at about 75 minutes, and the median trip duration is 22 minutes. A two-hour block gives a…