Skip to main content

You are here

Advertisement

Improving How Retirement Income Is Measured

Practice Management

Data on income during retirement can be fraught with errors in gauging income and poverty levels among the elderly, says a recent study that also suggests a way to improve that measurement. 

“Improving the Measurement of Retirement Income of the Aged Population,” ORES Working Paper No. 116 released in January 2021, Irena Dushi and Brad Trenkamp, respectively an economist and policy analyst with the Social Security Administration (SSA), study how that data may be made more accurate. They use data from the U.S. Census’ 2016 Current Population Survey (CPS) Annual Social and Economic Supplement (ASEC), IRS tax records and Social Security earnings and benefit records to examine whether and to what extent their data improves income estimates, and compare those estimates with public-use data from the University of Michigan’s 2016 Health and Retirement Study (HRS). 

“Accurate measurement of retirement income in national surveys is a challenge,” Dushi and Trenkamp write. They continue that the CPS “is no exception,” and add that “recent research has reinvigorated a debate about the adequacy of its retirement income measures” and that the SSA is very interested in the matter.

Dushi and Trenkamp note that they are not alone, writing that other researchers also have criticized the CPS and have emphasized that the CPS inadequately measures income from assets and retirement accounts such as 401(k)s and IRAs, and that the result is understating their importance and overstating the importance of Social Security. 

On the other hand, Dushi and Trenkamp call the HRS “the most comprehensive national longitudinal survey of Americans aged 51 or older.” They write that an advantage of the HRS is that it asks for respondents’ consent to link their survey information with earnings and benefits information from Social Security records. They add that the HRS is more systematic than the CPS in collecting information on pensions, retirement-plan account balances, and their distributions, and note that earlier research showed that HRS-reported household income amounts among people aged 51 or older were substantially higher than amounts the CPS reported. 

Findings

Dushi and Trenkamp supplemented CPS data from March 2016 with data from the Social Security Administration and the IRS and compared the matched data files with the original data. The result, they said, showed that even though the CPS redesigned its questionnaire, retirement income was still underreported in its data.

The ORES study and CPS also differ on the significance of Social Security as a source of retirement income. 

 

Data Source Primary Source of Aggregate Income % of People > Age 65 for Whom Social Security is > 90% of Family Income Estimated Poverty Rate Among People > Age 65
CPS Social Security                 25.6%    8.8%
HRS       --                 21.2%    6.6%
ORES Study Pensions, including IRA Withdrawals                13.8%   7.1%

 


The ORES study, Dushi and Trenkamp write, provides “strong evidence in support of supplementing survey data with administrative data to describe the income of older Americans.” 

Dushi and Trenkamp found that for the population aged 65 or older, supplementing the CPS ASEC with IRS and Social Security administrative data yields the following results in: 

  • a higher estimate of the share of aggregate income comprised by pension income; 
  • there is less reliance on Social Security; 
  • the estimated poverty rate is lower; and 
  • HRS provides better estimates of the income of the aged population than the public-use CPS data.

Dushi and Trenkamp  add a caveat. They warn that using blended data poses some risks; foremost among them is that disclosures could result in individuals being identified. They add that protecting data can cause statistics to be not released in a timely manner.