Skip to main content

Investing for retirement: The moderating effect of fund assortment size on the 1/n heuristic (Morrin et al. 2012)

Review Guidelines

Citation

Morrin, M., Inman, J. J., Broniarczyk, S. M., Nenkov, G., & Reuter, J. (2012). Investing for retirement: The moderating effect of fund assortment size on the 1/n heuristic. Fox School of Business Research Paper No. 14-009, 1–38.

Highlights

Experiments using U.S. households

  • The study’s objective was to examine whether increases in the number of funds offered in a retirement plan caused investors to allocate their contributions more evenly across all available investment options.
  • The authors administered one survey to a nationally representative sample of U.S. households and two surveys to people from consumer panels. The surveys contained information on hypothetical savings plans that were manipulated to include either a small or large fund assortment size. Participants were instructed to choose the plans in which they would invest and indicate the percentage they would invest in each of the chosen funds.
  • The study found that participants were more likely to allocate hypothetical investments across all available funds when offered a small versus large assortment size.
  • The quality of the causal evidence presented in this portion of the study is moderate because, although the authors controlled for observable characteristics of samples members and there was no evidence that survey versions were systematically assigned to sample members, the Clearinghouse for Labor Evaluation and Research (CLEAR) could not confirm that survey versions were randomly assigned.

Analysis of employees’ data 

  • The study’s objective was to determine whether the number of fund options offered by a 401(k) plan influenced the investment behavior of plan members.
  • The authors compared investment behaviors among employees offered 10 or 19 fund options for investment, controlling for employees’ characteristics. The analysis used administrative data on contributions to the defined contribution plan of the Oregon University System. The study used an interrupted time series (ITS) design that analyzed changes in outcomes before and after an increase in the number of available investment options.
  • The analysis demonstrated that offering a larger fund assortment size was associated with a statistically insignificant decrease in the tendency of employees to invest in all available funds and spread invested dollars evenly among chosen alternatives.
  • The quality of causal evidence presented in this study is low. This means that we are not confident that the estimated effects are solely attributable to fund assortment size; other factors are likely to have contributed.

Intervention Examined

Informational Interventions for Households and Workers

Findings

Experiments using U.S. households

  • Participants were significantly more likely to invest in all available funds when offered a small fund assortment size compared with a large assortment size.
  • Increases in fund assortment size were associated with significant increases in the propensity to allocate investments evenly across chosen funds only when participants faced a high cognitive load.

Analysis of employees’ data

  • Participants offered 19 funds invested in a significantly larger number of funds than those offered 10 options (5.30 compared with 3.74 funds, on average). However, increasing the number of available funds for investment did not significantly change the probability that participants would invest in all available funds or that the allocation of money would be evenly spread across funds.

Considerations for Interpreting the Findings

In the studies using survey data, the authors used ordinary least squares regression and controlled adequately for observable characteristics that might be associated with the outcomes of interest, such as age, gender, income, and 401(k) enrollment. There is no reason to believe that the treatment and comparison groups differed systematically; however, CLEAR could not determine how participants were allocated to different versions of the survey. Thus, these studies cannot be considered randomized controlled trials.

Low response rates in study 1a also suggest that survey respondents might differ somewhat from the population as a whole. Additionally, participants in studies 1b and 2 were members of consumer panels, which might not represent the entire population and could lead to less generalizable results. It should also be noted that the studies of survey data asked participants to make hypothetical decisions; these decisions might not represent actual behavior.

In the final study, the comparison and treatment groups fall along a chronological continuum; thus, effects are estimated in an interrupted time series framework. The authors conducted only a single demonstration of the intervention and did not provide any evidence to suggest that its timing was not confounded with prior trends, so it is possible that the intervention coincided with other events that mask its true effect.

Causal Evidence Rating

The quality of the causal evidence presented in each of the household survey-based studies detailed in this study is moderate. This means we have some confidence that the estimated effects are attributable to the interventions studied, although other factors also could have contributed.

The quality of the causal evidence presented in the study using employees is low. This means we are not confident that the estimated effects are attributable to the number of funds offered by the defined contribution plan. Other factors are likely to have contributed. To provide more convincing evidence of the intervention’s effects that satisfies CLEAR criteria, the authors should consider implementing the program at other firms and using more pre- and post-intervention observations in their analyses. If other implementations demonstrated similar effects and more data were used to show such effects, we could be more confident that changes in outcomes were driven by the intervention, and not some other concurrent change.

Reviewed by CLEAR

April 2015