Although automatic enrollment and automatic escalation have prompted more defined contribution participants to save more for retirement, their impact has been less than originally expected.
The improvement is "smaller than we thought," said David Laibson, a Harvard University professor of economics and prominent researcher on behavioral economics, on Nov. 13 at the Academic Forum in New York sponsored by the Defined Contribution Institutional Investment Association.
Laibson described recent research that he and co-authors conducted showing that initial expectations of retirement savings improvement are eroded by time, participants changing jobs and participants withdrawing their retirement account funds when they leave their employers.
Although auto features "are a net positive," Laibson said they "are not working as well as we thought." Lower-income participants are the least affected, he added.
One "small step forward in the right direction" would be for DC plans to set initial default rates at 6%, 8% or 10% rather than the more common 3%," he said. "But this doesn't work well for the lowest-income workers."
As a result, auto features are providing opportunities for "midlife consumption" rather than long-term wealth accumulation, he said, referring to research published Nov. 5 by Laibson and four colleagues from Harvard, Yale University and the University of California Berkeley for the National Bureau of Economic Research.
"When we jointly incorporate several medium- and long-run dynamic factors, we find that the magnitude of the savings effect is substantially diminished relative to previous estimates," the research report said.
"Automatic savings policies have a positive statistically significant impact on measured savings," the report said. "The effects we estimate are modest in magnitude because of the increase in savings over time by those who are not subject to automatic policies, high employee turnover rates, vesting requirements with respect to employer matching contributions, the high rate of cash leakage upon job separation, and the low acceptance of auto-escalation defaults."
The research is based on a review of nine employers that introduced auto enrollment, auto escalation or both between 2005 and 2011. It compared 62,340 employees in the year after auto features were introduced vs. 55,937 employees who were hired the year before.
"Our sample of firms is small, which means that our estimated net effect of automatic policies could differ substantially from an estimate in a larger, more representative sample," the authors conceded. "Nevertheless, statistics from other large samples suggest that the erosion of automatic policy treatment effects through the additional channels we consider is also sizable in the general population hired the year before auto features were offered."
The authors assumed 85% of participants would auto-enroll. After one year, the research indicated a 2.2-percentage-point increase in retirement savings. By five years, however, the gain was 1.8 percentage points.
When authors introduced an incomplete-vesting component, the gain fell to 1.5 percentage points. When they incorporated withdrawals, the overall gain was down to 0.6 percentage points.
Auto escalation produced a similar trend. After one-year, the increase was 0.8 percentage points. After five years, it was down to 0.7 percentage points. Incomplete vesting cut the gain to 0.6 percentage points. Withdrawals reduced the overall gain to 0.3 percentage points.
When combining auto enrollment and auto escalation, the authors reported that the initial increase of 3.1 percentage points after one year had fallen to 0.8 percentage points when the various factors were measured.