将当前的人口调查与所得税数据进行比较（pdf）

Peter J. BradySteven BassInvestment Company Institute*1401 H St. NWWashington, DC 20005pbrady@ici.org March 17, 2021 Abstract This paper compares measures of 2010 annual income from the Current Population Survey(CPS) with those derived from IRS Statistics of Income Division (SOI) administrative tax dataand finds that the CPS vastly understates the income of the elderly. We focus on four types ofincome found to be accurately reported on income tax returns and which represent the vastmajority of income for most Americans: wages and salaries, Social Security benefits, retirementincome (distributions from pensions, annuities, and IRAs) and investment income (taxableinterest, tax-exempt interest, and dividends). The CPS appears to accurately measure wage andsalary and Social Security income, but it misses more than half of retirement income and morethan one-third of investment income. Differences exist regardless of age and, within age groups,persist even after accounting for nonfilers and controlling for other income. These differencestranslate into larger differences in more comprehensive measures of income—in both amountand composition—for older individuals, who get a much larger share of their income fromretirement and investment income. Despite the fact that nearly one-in-four individuals age 70 orolder do not file a tax return, aggregate income is nearly one-third higher in the SOI. The taxdata also show that, as a group, individuals age 70 or older get more income from pensions,annuities, and IRAs than they get in Social Security benefits. The tax data only allow us todefinitively identify two broad categories of retirement income: IRA distributions and incomefrom pensions and annuities. Combining SOI-derived estimates with aggregated plan-level dataon pension distributions, we estimate that traditional private-sector DB pension plansaccounted for, at most, 19 percent of pension and annuity income in 2010—or about 14 percentof all retirement income, inclusive of IRA distributions. Government employee DB pensionsaccounted for a much larger share: up to 45 percent of pension and annuity income—or nearlyone-third of all retirement income. *This research was conducted as part of the Statistics of Income Joint Research Program. Views presentedare those of the authors and do not necessarily represent the views of the Internal Revenue Service or theviews of the Investment Company Institute or its members. We thank Kevin Pierce for his assistance withthis project. Comparing the Current Population Survey to Income Tax Data 1. Introduction This paper compares income measures derived from household survey data collected bythe U.S. Census Bureau’s Current Population Survey (CPS) withincome measures derived fromadministrative tax data compiled by the Internal Revenue Service Statistics of Income Division(SOI). We find that the CPS substantially undercounts retirement income (distributions frompensions, annuities, and IRAs) and investment income (taxable interest, tax-exempt interest,and dividends). Although differences exist regardless of age, they translate into largerdifferences in more comprehensive measures of income for older individuals. For example, forindividuals age 70 or older in 2010, retirement income and investment income reported on taxreturns are both more than twice the amounts reported in the CPS. Comparing a broadermeasure of income for this age group that also includes Social Security benefits and wages,taxpayers report nearly one-third more income in the SOI than is reported by all individuals—including both filers and nonfilers—in the CPS. This study relates to the broader literature which compares household surveys to otherdata sources. The US Census Bureau periodically assesses the quality of the income dataobtained from the CPS by comparing the estimates to other data sources.1Both the US CensusBureau and the US Social Security Administration have matched survey data to administrativerecords and these data have been used to assess survey quality for certain types of income,including wage and salary income,2Social Security benefits,3transfer payments,4and electivedeferrals in retirement plans.5 The fact that the CPS underreports income has long been noted in the literature. Forexample, Rector, Johnson, and Youssef (1999) compares the CPS to the Bureau of Economics(BEA) National Income and Product Accounts (NIPA) and concludes that the CPS understatesincome and overstates poverty. Comparing household survey data to administrative data on 10 Comparing the Current Population Survey to Income Tax Data transfer programs, Meyer, Mok, and Sullivan (2009) finds the CPS suffers from misclassificationof income, underreporting of income, and underreporting of enrollment in means-testedgovernment programs. Underreporting of retirement income was noted as far back as Schieber(1995), which compares CPS data to both NIPA estimates and published tabulations of SOI dataand found that retirement income