Applied Micro Seminar - Adeline Delavande (UTS)

Title: Talking to a Wall or to a Mirror? Within-Couple Expectations and Information Sharing (with Gizem Kosar and Basit Zafar)

Abstract:  The decision of when to retire, how much to save, or when to claim Social Security benefits are made jointly at the household level. Perceptions of own social security as well as those of the spouse are likely important determinants of these decisions. However, little is known about how knowledge about Social Security is distributed within couples. Our innovation is to put together a sample of spouses, where individuals within the couple are surveyed sequentially. We then embed a randomized information experiment to examine how informed individuals are about their own and spouse’s Social Security benefits, how these frictions affect their expected claiming and retirement behavior, and the potential for light-touch interventions to correct bias. We find that: (1) There are large errors in both own and spouse’s expected benefits. These errors are correlated at the individual level, such that individuals who underestimate their own benefits tend to also underestimate their spouse’s benefits. The correlations for own and spouse's expected benefit tend to be larger for those who are less financially literate and those who are further away from retirement. (2) These errors are also correlated within couples, compounding the potential effects on expected retirement behavior and making some couples particularly vulnerable. This suggests a lack of specialization in Social Security knowledge within the household. (3) Providing individuals with information about their future benefits reduces errors in their own and their spouse's expected benefits, with even greater reductions when information is provided about both. The treatment also leads to a better alignment of retirement plans within the couple. (4) There is information spillover within couples, as respondents whose spouse received the information treatment had more accurate expectations about future benefits. The level of information-sharing is similar whether the index spouse who received the treatment is male or female. However, there is no information sharing among couples whose members disagree about their marital satisfaction.