Why removing an incentive rarely restores the earlier behaviour

A nursery introduces a fine for late collection. Late pickups rise, which is already surprising. The fine is then withdrawn, and late pickups stay high — they do not return to the pre-fine level. The experiment ran in both directions and only one direction behaved as expected.

What everyone sees

Incentives are treated as reversible instruments, switched on when needed and off when not, with behaviour tracking the switch. That model assumes the incentive acted on a stable underlying preference. It acted on the interpretation of the situation, and interpretations do not revert when a payment stops.

What is actually happening

Gneezy and Rustichini ran exactly this design in Israeli nurseries and reported that late collection increased after a fine was introduced and remained at the higher level after the fine was removed. Their reading is that the fine reclassified lateness — from a breach of an unpriced obligation into a service with a known price — and that removing the price left the reclassification in place, now with the price implicitly set to zero. Bowles and Polanía-Reyes survey a broad set of studies finding similar non-reversal, while noting that the phenomenon is heterogeneous across settings and that some interventions do revert cleanly. The asymmetry, where it occurs, is informational rather than motivational: the participants learned something about the relationship that the withdrawal does not unteach.

Why it stays hidden

The irreversibility hides because the introduction and the withdrawal are usually studied separately, months apart, by people asking different questions. Nobody runs the reversal as a test. And when behaviour fails to revert, the explanation reached for is drift, staffing or season — all of which are true of the period and none of which are the mechanism.