This paper investigates the unintended cognitive consequences of a major retirement savings policy and their implications for intervention design. While previous studies recognize that various heuristics and biases hinder optimal retirement decisions, policy efforts have largely focused on increasing savings rather than addressing the cognitive processes behind them. We take a step in this direction by examining how a nationwide matching contributions policy introduced in Türkiye’s Individual Pension System affected the reinforcement learning (RL) heuristic—individuals’ tendency to over-extrapolate from past returns. Analyzing a large administrative dataset spanning both pre- and post-policy periods, we find that participants’ responsiveness to contemporaneous and lagged returns increases dramatically—by more than five-fold—after the policy is implemented. These results persist after testing alternative explanations. Rather than curbing cognitive distortions, the policy unintentionally amplifies them. These findings highlight the importance of evaluating interventions not just by their impact on overall savings but also by their influence on the cognitive foundations of decision-making.

Authors
Sadettin Haluk Çitçi
Full Professor in Economics, Gebze Technical University

Authors
Halit Yanikkaya
Full Professor, Department of Economics, Gebze Technical...

Authors
Yunis Dede
Ph.D. Student, Department of Economics, Gebze Technical...