Americans Underestimate Social Security Income by $1,900: Study

Older adults generally have accurate expectations about their Social Security claiming age, but they underestimate their annual Social Security income by approximately $1,900, or 11.5%, on average, according to a new paper published by the National Bureau of Economic Research.

The analysis, written by contributing NBER researchers Grant Seiter and Sita Slavov, utilizes panel data from the University of Michigan’s Health and Retirement Study survey series, which collects data from about 20,000 U.S. household members ages 50 and older every two years.

Specifically, Seiter and Slavov compare HRS respondents’ observed Social Security claiming ages and benefits with subjective expectations provided during their 50s and early 60s, finding that, while older adults generally have accurate expectations about their claiming age, they commonly underestimate their annual Social Security income. Both forecast accuracy and precision increase with age, however, such that the average forecast error for people in their early 60s is not statistically different from zero.

The researchers go on to use “plausibly exogenous variation” in the mailing of Social Security statements, which contain personalized information about future benefits, to show that the provision of additional information to late-career workers can help reduce forecast errors in annual income.

Ultimately, the researchers conclude that the provision of better information through Social Security statements would have a statistically significant effect on reducing forecast errors, thereby allowing Americans to plan for their retirements more accurately.

Running the Analysis

As Seiter and Slavov point out, the University of Michigan’s ongoing Health and Retirement Study provides detailed demographic, financial and expectations data on a nationally representative set of individuals over the age of 50 and their spouses or partners. Crucially, the researchers explain, the HRS collects information about respondents’ expected and actual Social Security claiming ages and benefit amounts.

In this case, Seiter and Slavov use data from the HRS “waves” 1-14, covering 1992 to 2018. The survey questions and responses analyzed include the following:

• Do you expect to receive Social Security benefits at some time in the future?
• At what age do you expect to start collecting these benefits?
• If you start collecting Social Security benefits then, about how much do you expect the payments to be in today’s dollars?

According to Seiter and Slavov, the responses collected generally indicate whether the frequency of a given respondent’s expected income is per week, bi-weekly, monthly, annually or a lump sum. For the purposes of their analysis, the NBER researchers ignore “lump sum” and “other” frequencies and convert the remaining expectations to annual values. They also set expected income to zero for those who report not expecting to receive Social Security benefits.

The researchers then converted these monetary values to 2021 dollars using the Retroactive Consumer Price Index for All Urban Consumers, and to inflate expected income, they use the index value for the calendar year and month in which the respondent’s interview ended. Finally, for observed income, which is reported for the last calendar year, they use the index value for December of the previous year.

The results of this analysis, according to Seiter and Slavov, suggest that approximately 10% of respondents — all of whom eventually collect Social Security — do not expect any benefits. This fraction is smaller than estimates from surveys that include people of all ages, the researchers note.

Overall, the mean Social Security income forecast is $1,897 lower than the actual benefit that is due, which represents an 11.5% underestimate relative to the mean observed benefit. Notably, there is considerable variance, with 25% of older Americans underestimating their benefit by $5,167 or more, and 10% overestimating by $5,319 or more.

Seiter and Slavov further find that, as those who claim benefits later receive an actuarial adjustment resulting in larger monthly benefits, people tend to underestimate their benefit at the time of claiming if they have underestimated their claiming age.

However, the data also suggests that people have fairly accurate expectations of their claiming age: They underestimate by less than one month on average, and more than half the forecasts are within six months of the observed claiming age.

According to the authors, the mean and standard deviation of the income and claiming age forecast errors, respectively, evolve with age. Put simply, people get better at forecasting their benefit amount as they grow older.

The Effects of Better Information Sharing

As Seiter and Slavov point out, in fiscal year 1995, the Social Security Administration initiated a program to automatically mail personalized statements to individuals. These statements include the recipient’s earnings history and projected benefits payable at different claiming ages. They also provide information on disability, spousal and survivor benefits payable on the recipient’s record.

Statement mailings were rolled out by age, the NBER researchers note, starting with those ages 60 and above. Younger age groups received statements in subsequent years, and starting in 2000, all adults 25 and older with covered earnings received annual statements.

However, statement mailings were suspended in March 2011 because of budgetary constraints. A more limited reintroduction of statement mailings began in September 2014, Seiter and Slavov explain, with statements sent to those ages 60 and above, as well as individuals ages 25 and above who reached an age that is a multiple of five. Since 2017, only individuals ages 60 and older have received statements.

Ultimately, following the availability of online statements to all adults in 2012, only individuals who had not yet created an online account with the Social Security Administration received statements after the 2014 reintroduction of mailings. Using this timeline, Seiter and Slavov tabulate the number of statements an individual should have received between consecutive HRS interviews.

According to the researchers, receiving one or more statements increases the benefit prediction by $344, or 2.1%, relative to the mean. Stated another way, the data suggests that those who did not receive a statement underestimate benefits by $1,949, while those who received one or more statements underestimate benefits by only $1,605.

Seiter and Slavov conclude that the coefficients on receiving two or three statements are “insignificant and imprecisely estimated,” but there is some reason to suspect that receipt of one or more statements reduces the relative forecast error.

(Image: Shutterstock)