Personalizing or reminding? How to better incentivize savings among underbanked individuals

We use a large-scale randomized control trial to evaluate information interventions to motivate savings in a mostly unbanked population. We find positive impacts of personalized SMSs on savings during the messaging period and larger than existing studies. However, they fade away after messaging stops, like in other experiments. Compared to the existing literature, we can better identify the sources of that effect. Personalized SMSs are not more likely to be received or more memorable. Highlighting a specific monetary goal appears to be more effective than tangibility. Also, salience or personalization separately do little to increase the intensive margin of savings. Heterogeneous effects are consistent with limited attention co-existing with other behavioral biases; for groups where reactance and frustration play a role, those are overcome only when personalized SMSs are employed.