Abstract

In Mexico, as in many other countries, retirement savings levels are low. The situation is worse for informal workers and the unemployed, who cannot rely on employer contributions to help build their nest eggs. To investigate ways to increase voluntary retirement contributions in Mexico, Innovations for Poverty Action collaborated with the Inter-American Development Bank and CONSAR, the government regulator in charge of overseeing the Mexican retirement system, to conduct three waves of rapid-fire randomized evaluations that will measure the impact of text messages on retirement savings behavior. Researchers conducted related research in Colombia on the impact of different messaging strategies on savings contributions to retirement savings. Preliminary results suggest that the text messages did not increase the number of transactions or savings amounts of account holders.

Policy Issue

While saving for retirement is essential for the future well-being of working age people, saving for the future is difficult and voluntary contribution levels tend to be low across the world. Research has shown that barriers to saving are wide-ranging: they include transaction costs (both monetary and non-monetary), a lack of information or knowledge, social constraints, a lack of trust in the financial system, ill-designed regulation, as well as human tendencies that hinder good decision-making (“behavioral biases”).¹

Behavioral biases are now considered a significant, but often overlooked, contributor to the problem of under-saving. Research suggests that even those with access to convenient and trusted financial products and a thorough understanding of financial concepts still face significant behavioral barriers to following savings plans. In spite of their best intentions, individuals tend to spend on day to day consumption, giving in to temptation or more immediate needs, rather than prioritizing saving for future needs that are less salient.² In developed countries, interventions aimed at overcoming these behavioral biases have been found to be successful in increasing employee savings for long-term retirement fund products.³ However, there is a need for more evidence on the use of behaviorally-informed tools to promote retirement savings in the developing world, and in Latin America in particular.

Context of the Evaluation

In Mexico, retirement savings levels are low. Although 6.5 percent of an employee’s salary is automatically deposited into long-term savings, that amount is too low to ensure a comfortable retirement: workers can expect to receive about 40 percent, or less, of their salary upon retiring.⁴ Meanwhile, those who are employed in the informal sector or who are unemployed cannot rely on employer contributions to help build their nest eggs, putting elderly Mexicans at risk of poverty. In this context, voluntary retirement contributions are vital to increase savings levels for workers in all sectors and for those who are unemployed.

This research builds upon the previous evidence of the impact SMS reminders can have on savings (here) and on a pilot study that IPA conducted with Colpensiones, the public defined-benefit provider in Colombia, in late 2016, which examined the impact of using text messages to increase savings behavior. While results from these interventions indicated that messages can improve savings behaviors, IPA worked with CONSAR and the IDB’s Labor Markets Unit to further investigate the impact of a variety of messaging strategies on the retirement savings behavior of a sample of inactive accounts (accounts that have not been making voluntary contributions).

Details of the Intervention

Researchers partnered with CONSAR and the IDB’s Labor Markets Unit to conduct a series of randomized evaluations on the impact of various messaging strategies on retirement savings behavior. Researchers conducted three rapid fire randomized evaluations over one year, each lasting approximately four months, which each addressed different behavioral- and information-based barriers to saving for retirement compared to a standard savings message. This iterative research design enabled researchers to optimize the messages tested and quickly inform CONSAR on the most effective and lasting strategies for improving savings behavior. During each wave, individuals enrolled in AFORE received a total of eight SMS messages, sent every other week. Researchers randomly assigned users to three equally-sized groups, and respectively sent each group messages the day before payday, the day of payday, or the day after payday.

Behavioral messages: The first wave tested the effectiveness of three different behavioral-based messaging strategies. Among the 1.4 million eligible account holders, from ten of the eleven Mexican AFORES, 400,000 users were randomly sampled to participate in the study.  From the study sample, 200,000 users were randomly selected to receive one of the messaging strategies while 200,000 users did not receive any messages and served as a comparison group. The specific language used in the messages varied depending on the employment status of the recipient (formal vs. informal). Furthermore, each message included a different behavioral framing, which built upon a framework identified by Ideas42⁵: making retirement feel vivid and present, making retirement visible and common, and reducing the sense of loss associated with saving.  

Savings goal messages: The second wave tested the effectiveness of setting different levels of annual savings goals for users. High, medium, and low saving goals were assigned based on an individual’s savings behavior the previous year. After establishing the goal, subsequent messages informed users about their progress towards accomplishing their goal. Users were assigned to either receive instructions to make a one-time savings deposit manually at a physical location or to set up automatic savings contributions through an internet link. Among the 200,000 users in the behavioral messages treatment group, 125,000 users were randomly selected to receive savings goal messages in the second wave. The remaining 75,000 users stopped receiving any messages.

Information messages: The third wave tested the effectiveness of providing users with additional information about how to add money into their retirement savings accounts. Users continued to receive savings goal messages, and half of the sample received an additional hyperlink to a CONSAR webpage that provided information about voluntary savings. Researchers included links to pages describing the process, the benefits, and the expected returns of saving voluntarily through the AFORE. During this wave, among the 125,000 users that kept receiving SMS during wave 250,000 users were randomly selected to continue receiving messages, and the remaining 75,000 users stopped receiving messages after the second wave.

Auxiliary study: Expanding the study to a population with more active savings behavior

Results of the related study in Colombia demonstrated that individuals who were actively saving before receiving messages increased their savings amounts in response to messages at a higher rate than people who were not actively saving prior to receiving messages. In response to these results, researchers decided to relax the original eligibility criteria of the study, which had largely restricted the sample to non-savers. They randomly assigned 30,000 individuals with higher incomes who were making voluntary savings contributions to their retirement accounts prior to receiving messages to receive the savings goal and additional information messages for four months.

The research team collected transaction-level administrative data to measure the impact of these different messaging strategies on the number of savings transactions and savings account balances of the AFORE account holders. 

Results and Policy Lessons

Researchers are still analyzing data for a complete analysis — all results reported below are preliminary.

Overall, researchers did not find evidence that any of the account holders who received the messages increased the number of transactions or savings amounts. While these results have disappointing implications for the potential of simple text messages to promote savings behavior, they largely align with findings from the parallel study that researchers conducted in Colombia with Colpensiones. In Colombia, the messages improved the number and amount of voluntary retirement savings contributions among individuals who were actively saving prior to receiving the messages, but impacts among those who were not saving were much smaller in magnitude. Given that the main study sample in Mexico was composed of individuals who had not made voluntary savings contributions prior to the study, it makes sense that in this context, the messaging program was less effective overall.

To investigate whether the messaging program was more effective for individuals in Mexico who were more actively using their retirement accounts, researchers examined the results from the auxiliary study of 30,000 active savers. The results demonstrated that sending messages to these individuals did not positively affect their savings behavior in terms of number of deposits or amount saved, which was consistent across the four months and the different messaging content and timing tested.

Sources