Do Savers Learn from Experience? Evidence from Pension Contributions

We examine whether households’ voluntary retirement saving decisions are influenced by reinforcement learning (RL), a behavioral heuristic where recent outcomes disproportionately shape future choices. Using eight years of universe-wide administrative data from Türkiye’s Individual Pension System, we show that savers over-weight recent return experiences. Specifically, individuals experiencing higher returns in one year substantially increase their voluntary contributions in the following year, and past returns continue to affect contributions with a diminished but persistent impact. The implied one-year learning weight is moderate, closely mirroring laboratory estimates. Alternative explanations such as inertia, skill learning, or asset rebalancing do not explain these observed behaviors