The driver bonus that pulled cars away from control drivers

Question

A ride-hailing company randomly gives some drivers a bonus for working busy zones and keeps others as controls. Treated drivers earn more; control drivers earn less. The team concludes the bonus is a big earnings booster and plans to give it to everyone.

Explain why the A/B comparison misleads here, and how you would test the bonus properly.

Accepted Answer

Drivers compete for the same riders. The number of ride requests in a city at a given moment is roughly fixed. When treated drivers rush to the busy zones, they scoop up trips that control drivers would otherwise have taken. So treated drivers earn more partly by taking business from control drivers, whose earnings fall not because they got worse but because the treated group crowded them out. This is interference: one driver's treatment changes another driver's outcome through a shared, limited pool of demand. Why it overstates the bonus. The A/B gap between treated and control is inflated f…

The driver bonus that pulled cars away from control drivers

Your answer

More Statistics questions