200 likes | 281 Views
How Automatic Enrollment Affects the Likelihood and Distribution of 401(k) Contributions: Evidence from a National Survey. Barbara Butrica and Nadia Karamcheva Urban Institute August 7, 2014. Background and Question.
E N D
How Automatic Enrollment Affects the Likelihood and Distribution of 401(k) Contributions: Evidence from a National Survey Barbara Butrica and Nadia Karamcheva Urban Institute August 7, 2014
Background and Question • Automatic enrollment has been shown to increase participation in 401k plans. • Empirical evidence from three main sources: • Firm-level case studies: may not generalize to larger population of workers • Proprietary plan-level data: mostly large plans and may not represent all covered workers • Firm-level national data: lacks demographic information to analyze participant behavior • We use individual-level data from a nationally representative survey to look at how autoenrollment is related to the likelihood and distribution of 401(k) contributions.
Data: Health and Retirement Study • Nationally representative survey of adults ages 51 and older. • Tracks information on personal characteristics, employment, earnings, income, financial assets, and pensions. • In 2006, the HRS began asking household respondents about automatic enrollment. • Because these questions changed between the 2006 and 2008 waves, our analysis uses pooled data from only the 2008 and 2010 waves. • Analysis includes workers ages 55 to 69 who are not self-employed.
New hires and low earners are most likely to be offered a DC plan with autoenrollment
Autoenrolled workers are more likely to be included in DC plans—especially new hires and low earners
Autoenrolled workers are less likely to participate in DC plans—except new hires and low earners
Median contributions are lower among autoenrolled workers than opt-in workers
Among participants, however, there is no statistical difference in contribution amounts between autoenrolled and opt-in workers
Median contribution rates are lower among autoenrolled workers than opt-in workers
Again, for those participating there is no statistical difference in contribution rates between autoenrolled and opt-in workers
Controlling for other factors, autoenrollment increases the likelihood of being included in a DC plan—especially for new hires
However, autoenrollment is associated with a decline in DC participation—except for new hires and low earners
Furthermore, autoenrolled workers contribute less to DC plans than opt-in workers
Among DC participants, contributions of those autoenrolled are not statistically different from those who opt in
Autoenrolled workers also have lower DC contribution rates than opt-in workers—no differences by tenure or earnings
Among workers who participate in DC plans, contribution rates of those autoenrolled are not statistically different from those who opt in
Conclusions • Relationship between automatic enrollment and DC contributions may be more ambiguous than originally expected. • Autoenrollment increases likelihood of being included in a DC plan for all workers, but increases likelihood of participating only for new hires and low earners. • Reduces probability of participation for old hires.
Conclusions (continued) • Automatic enrollment is associated with lower contributions and contribution rates • Large share of autoenrolled workers does not contribute to their plans. • So controlling for positive contributions, there is no statistically significant difference between workers. • Autoenrollment has no effect on the contributions of new hires and is associated with a reduction in contributions among old hires.
Policy Recommendations • Autoenrollment appears effective at increasing participation among new hires and low earners. • Needs to do better job of keeping old hires and boosting overall contribution levels among participants. • Possible ways to achieve this might be: • Autoescalation • More generous employer match
Future Research • Examine how autoenrollment affects workers’ participation and contributions, controlling for various life events. • Estimate quintile regressions to examine whether the effect of automatic enrollment changes throughout the distribution. • Use information on employer contributions to derive a more complete measure of DC savings and analyze the relationship between total worker and employer contributions and autoenrollment.