Identifying heterogeneity using recursive partitioning: evidence from SMS nudges encouraging voluntary retirement savings in Mexico

Abstract

Individuals regularly struggle to save for retirement. Using a largescale field experiment (N=97,149) in Mexico, we test the effectiveness of several behavioral interventions relative to existing policy and each other geared towards improving voluntary retirement savings contributions. We find that an intervention framing savings as a way to secure one’s family future significantly improves contribution rates. We leverage recursive partitioning techniques and identify that the overall positive treatment effect masks sub-populations where the treatment is even more effective and other groups where the treatment has a significant negative effect, decreasing contribution rates. Accounting for this variation is significant for theoretical and policy development as well as firm profitability. Our work also provides a methodological framework for how to better design, scale, and deploy behavioral interventions to maximize their effectiveness.

Field testing the transferability of behavioural science knowledge on promoting vaccinations

“…Further, investigating moderators of intervention effectiveness can help identify actionable insights that can be translated across contexts 20 , 54 . Ongoing research has recognized that nuanced differences in measurement 51 and population 33 , 51 , 55 – 58 , as well as contextual moderators and individuals’ sense-making processes, may contribute to differences in intervention effect sizes or even cause interventions that seem promising in prior work to backfire unexpectedly 59 , 60 . This paper aims to extend these endeavours by assessing the transferability of evidence from various sources to real-world scenarios.…”

Field Tested: Assessing the Transferability of Behavioral Interventions

“…Also, behavioral scientists should provide policymakers with tools for identifying actionable research insights that can be translated across contexts 4,37 . Ongoing efforts in this area have recognized that nuanced differences in measurement 33 and population 33,38,39,40,41,42 , as well as contextual moderators and individuals' sense-making processes may contribute to differences in intervention effect sizes or even cause interventions that seem promising in prior work to backfire unexpectedly 43,44,45 . This paper aims to extend these endeavors by assessing the transferability of evidence about predictions or behavioral intentions to actual behavior in the field, as well as the transferability of findings from one field setting at one point in time to another.…”