Social Security Income Measurement in Two Surveys

Introduction

Social Security benefits are a major source of retirement income in the United States, and they directly influence the economic well-being and poverty status of many beneficiaries. Social Security retired-worker benefits replace a portion of preretirement income. That portion is greater for low lifetime earners than for higher earners; consequently, Social Security benefits account for a greater share of retirement income for lower-income beneficiaries (SSA 2012; Butrica and others 2012). Because Social Security income influences economic well-being, it is important that household surveys measure it accurately.

The payment of Medicare premiums complicates the survey measurement of Social Security income. Most beneficiaries elect to have the Social Security Administration (SSA) deduct those premiums from their monthly Social Security benefit, so that the amount they receive reflects a net monthly benefit that is lower than their gross benefit.¹ For all Social Security beneficiaries, income tax liability, poverty status, and eligibility for means-tested federal benefit programs—such as Supplemental Security Income and the Supplemental Nutrition Assistance Program—are determined using the gross benefit before any deductions. However, because the actual cash income that most beneficiaries receive is net of Medicare deductions, they may report the net amount in household surveys as their monthly Social Security income. Analysts who use household surveys to measure income need to know whether the Social Security income recorded on these surveys reflects the gross amount or the benefit net of Medicare premiums.

This article assesses the accuracy of Social Security income as it is recorded in the Census Bureau's Survey of Income and Program Participation (SIPP) and Current Population Survey (CPS). A major goal of the SIPP is to measure income amounts by source to allow analysts to estimate how many individuals and families are eligible for government income-security programs. A major goal of the CPS is to measure total individual and family income and the percentage of the population with income below the poverty threshold. Accurately measuring Social Security income is essential to achieving those goals.

The SIPP and the CPS ask similar questions about Social Security income, but the two surveys' reference periods differ. The SIPP, as a longitudinal survey, groups its respondents in 4-year panels, in which participants are interviewed every 4 months.² The SIPP asks respondents whether they received any Social Security benefits during the 4 months prior to the month of the interview and, if so, they are asked to report the monthly amount of Social Security income they received. Those respondents are then told that “some people have what is called a ‘Medicare Part B’ premium taken out of their Social Security benefit before it reaches them,” and are asked if they had the Part B premium deducted from their check and if they know the amount deducted (Census Bureau 2010, 233).³

An annual social and economic supplement to the CPS, conducted in March, asks respondents if they received Social Security income in the year before the interview. Those who received Social Security income are asked to report their benefit as either a monthly, quarterly, or annual amount. Respondents are then asked the amount of Social Security benefits they received in the month before the interview, and whether that amount was before or after the Medicare Part B deduction (Census Bureau 2009, D-27).⁴

Previous research has demonstrated that Social Security income is often underreported in surveys (Davies and Fisher 2009). Several studies have examined SIPP data matched to Social Security administrative records (for example, Huynh, Rupp, and Sears 2002; and Sears and Rupp 2003). Koenig (2003) studied both the SIPP and the CPS for the misreporting of Supplemental Security Income as Social Security benefits. This article uses SIPP and CPS data matched to SSA records. Specifically, we compare data on Social Security benefit payments (before and after Medicare premium withholding) from SSA's Payment History Update System (PHUS) to the benefit amounts recorded on the SIPP and the CPS. With the PHUS data, we ascertain whether the Social Security benefit amounts recorded in the surveys more closely approximate gross benefits or benefits net of Medicare premiums.

Data

For the SIPP, we measure Social Security income for 2009, the first full calendar year of income covered in the 2008 panel. We sum the monthly amounts recorded for 2009 in the SIPP public-use file to obtain annual Social Security income. We weight the data using the December 2009 survey weight. For the CPS, we measure Social Security income for calendar year 2008, as collected in the March 2009 Annual Social and Economic Supplement. As we began our research, that was the most recent CPS file to which SSA data had been matched.

We assess the consistency of the Social Security income in both surveys with the amounts recorded in SSA data. SSA's records include both the gross benefit amounts and the net amounts paid to beneficiaries after Medicare premiums have been deducted. This allows us to compare Social Security income in the surveys to both gross and net Social Security benefits. Because we focus on the income of the aged, we restrict our analysis to persons aged 60 or older in the survey year. We include in our sample all survey respondents who were recorded in the Social Security Master Beneficiary Record (MBR) as receiving Social Security benefits as disabled workers, retired workers, survivors, or spouses. In the SIPP, approximately 90 percent of the records for our sample were matched to Social Security administrative records. The match rate for the CPS was approximately 87 percent.

SIPP Results

For most SIPP respondents, the pattern of benefit receipt is consistent between the SIPP and the PHUS. Table 1 shows SIPP respondents aged 60 or older who were beneficiaries in December 2009 according to the MBR. About 88 percent of the SIPP respondents have a Social Security benefit recorded in both the SIPP and the PHUS. About 10 percent have a false negative on the SIPP: A benefit payment appears in the PHUS, but no Social Security income is recorded on the SIPP. Only 0.2 percent of SIPP respondents aged 60 or older had a false positive, in that they reported Social Security income in the SIPP, although the PHUS indicates no Social Security benefit. Slightly more than 1 percent received no Social Security income according to both the SIPP and the PHUS. In the latter two groups, the presence of beneficiaries for whom PHUS records indicate no benefit payments largely reflects the small portion of the MBR population that is not in current-pay status at a given time.

The differences between SIPP and PHUS records of Social Security income receipt vary by age and type of benefit. False negatives, in which the SIPP records indicate no Social Security income although payments appear in the PHUS, are more common among respondents aged 60–64 (about 15–17 percent) than among those aged 65 or older (9.2 percent). False negatives also are more common among disabled workers (15.2 percent) than among retired workers (9.8 percent), spouses (8.8 percent), or survivors (11.1 percent). Among the SIPP respondents who have a false positive report of Social Security income, there is little difference by age or type of benefit.

PHUS data for mean annual gross Social Security benefits in 2009 were about $1,000 higher than the mean annual Social Security income recorded in the SIPP (Table 2).⁵ The mean gross PHUS benefit was $13,548, which was $960 (or almost 8 percent) higher than the SIPP mean of $12,588. The median annual gross PHUS benefit, $13,595, was $1,176 (9.5 percent) higher than the median benefit, $12,419.

Mean gross Social Security benefit amounts in the PHUS exceeded the mean SIPP benefit amounts in all categories, but the extent of the differences varied by age and type of benefit. By age, the largest difference occurred among beneficiaries aged 60–61, and the smallest occurred among those aged 65 or older. By type of benefit, the largest difference occurred among disabled-worker beneficiaries who were younger than the Social Security full retirement age (66) throughout 2009.⁶ Disabled-worker beneficiaries become eligible for Medicare (and thus subject to premium deductions) after 2 years on the Social Security rolls. Most beneficiaries aged 60–61 are disabled workers, and for them, the amount by which mean gross Social Security income in the PHUS exceeded the SIPP amount ($2,079) is more than twice as high as the amounts for retired-worker, spouse, and widow beneficiaries ($942, $117, and $1,015, respectively). The difference between median gross Social Security income in the PHUS and the median SIPP benefit also was greater among disabled workers than among retired-worker, survivor, or spouse beneficiaries. The smallest difference between median gross benefits in the PHUS and median benefits in the SIPP occurred among spouse beneficiaries not entitled to worker benefits ($677).

Another way to measure differences between the data sources is to compare the benefit amounts for individual respondents. This approach finds the mean (or median) of the differences between the SIPP and PHUS record values, as opposed to the difference in the aggregate means (or medians) discussed above. To calculate the individual difference, we subtract the SIPP benefit from the gross PHUS benefit for each SIPP respondent with a Social Security benefit. If the SIPP respondent reported a benefit amount equal to the gross benefit recorded in the PHUS, the difference is zero. Overall, the mean and median differences between the individual Social Security income values reported in PHUS and the SIPP were $989 and $497, respectively. The median difference indicates that one-half of the SIPP records contained benefit amounts that were at least $497 less than the gross amount in that individual's PHUS record.

The pattern of the means of individual differences generally resembles that of the previously discussed differences of aggregate means. The exception is the spouse benefit, which averages $22 in individual differences, substantially lower than the $117 difference in aggregate means. That result is attributable mainly to negative values for the 2.0 percent of respondents who have benefit payments recorded in the SIPP but not recorded in the PHUS. By contrast, most of the median values of individual differences are substantially lower than the differences of medians. That result could indicate that differences between the SIPP and the PHUS records are less common among respondents with smaller benefit payments, as well as the effect of including negative values for respondents with benefit payments recorded in the SIPP but not in the PHUS.

Table 2 shows that the median gross PHUS benefit amounts, like the mean amounts, exceed those reported in the SIPP for all beneficiary categories. For at least some beneficiaries, the SIPP amounts might represent the net benefit after deducting Medicare premiums, which could explain the difference between the sources. Benefit amounts in the SIPP generally approximate the PHUS net benefit more closely than they approximate the PHUS gross benefit for SIPP respondents with positive Social Security amounts. Table 3 shows that the mean annual net PHUS benefit was only $81 lower than the mean SIPP benefit, and that the median annual net PHUS benefit was $121 higher than the SIPP benefit. Those differences are substantially smaller than the differences between the gross PHUS benefits and the SIPP benefits.

As with the gross benefit amounts, the differences between the SIPP benefits and the net PHUS benefits vary by age and type of benefit. The mean annual net PHUS benefit was $201 lower than the mean SIPP benefit among those aged 65 or older. At ages 60–61 and 62–64, the mean net PHUS benefits were $1,115 and $987 higher, respectively, than the SIPP means. The mean net PHUS benefit was much larger than the mean SIPP benefit among disabled workers, who become eligible for Medicare benefits (and subject to premium deductions) after 2 years on the Social Security rolls. Among SIPP respondents with retired-worker or survivor benefits, the mean net PHUS benefits are quite similar to mean SIPP benefits, differing by less than $100. The mean net PHUS benefit for spouse beneficiaries is substantially lower than the mean SIPP benefit, perhaps reflecting that category's higher share of false positives (respondents with zero benefits in the PHUS and positive benefits in the SIPP; see Table 1). Differences between median net PHUS benefits and median SIPP benefits generally followed patterns similar to those for mean benefits.

As we did for gross benefits, we also compare the net benefit payments for individual respondents in the PHUS with those recorded on the SIPP. As before, we subtract the SIPP benefit from the net PHUS benefit for each SIPP respondent with a Social Security benefit. The overall mean individual difference of -$31 is similar in magnitude to the overall difference in means of -$81. If the benefit recorded in the SIPP were equal to the gross benefit, the median individual difference for Medicare-eligible beneficiaries would equal the Medicare premium. In 2009, the Part B premium was $96.40 per month, or $1,157 for the year, so we would expect to see a median difference of about -$1,157. The median individual difference for beneficiaries aged 65 or older of -$543 indicates that for many Medicare-eligible beneficiaries, the SIPP benefit is substantially less than the actual gross benefit.

CPS Results

Patterns of benefit receipt in the CPS were similar to those in the SIPP. Table 4 shows CPS respondents aged 60 or older who were beneficiaries in December 2008 according to the MBR. About 90 percent received Social Security income at some time in 2008 according to both the CPS and the PHUS. About 9 percent of CPS respondents aged 60 or older had false negatives, with the PHUS indicating that they received Social Security income not reported in the CPS. About 1 percent of CPS respondents aged 60 or older received no Social Security income according to both the CPS and the PHUS, and another 0.2 percent reported Social Security income in the CPS although the PHUS indicates no benefits. As noted earlier, instances of beneficiaries having no benefits recorded in the PHUS occur mainly because a relatively small number of people on the MBR are not in current-pay status at a given time.

As was true with the SIPP, the extent to which the CPS records on beneficiary status differed from the PHUS varied by age and type of benefit. False negatives, in which no Social Security income was recorded in the CPS but benefit payments appeared in the PHUS, were more common among respondents aged 60–61 (26.2 percent) and 62–64 (20.0 percent) than those aged 65 or older (7.1 percent). False negatives in the CPS were more common among disabled workers (19.9 percent) than among retired workers (7.8 percent), spouses (8.8 percent), and survivors (7.6 percent). False positives were relatively rare at 0 to 2 percent, and they varied little by age or type of beneficiary.

The mean annual Social Security benefit before deductions in 2008 was, on average, $202 lower in the PHUS record than in the CPS (Table 5). The mean annual gross PHUS benefit was $12,685, compared to $12,888 in the CPS. The relationship of the median CPS values to the gross PHUS medians reversed that of the means. The median gross PHUS benefit was $204 (1.6 percent) higher than the median CPS benefit, $12,761 versus $12,557.

The differences between the benefit amounts recorded in the PHUS and the CPS were generally smaller than the differences between the PHUS and the SIPP. Moreover, unlike the SIPP benefits, the mean CPS benefit amounts were often greater than the gross means in the PHUS. With respect to age, the largest difference in mean benefits was among beneficiaries aged 62–64. For that group, the mean CPS benefit exceeded the gross PHUS benefit by $477. Among beneficiary groups, the biggest difference in mean benefits occurred among spouse beneficiaries, whose mean annual CPS benefit was $968 higher than the gross PHUS benefit. Only disabled workers had a mean CPS benefit that was lower (by $638) than the mean gross PHUS benefit.

In all three age categories, the median annual CPS benefit differed from the median gross PHUS benefit by less than 2 percent. The smallest difference between median gross PHUS benefits and the median CPS benefit occurred among disabled-worker beneficiaries, and the largest occurred among retired workers. For retired-worker and survivor beneficiaries, the median gross PHUS benefit was about 4 percent higher than the median CPS benefit, while for disabled-worker and spouse beneficiaries, the difference between the PHUS and the CPS was less than 2 percent.

As we did for the SIPP, we examined the mean and the median of the differences between benefit amounts for individual beneficiaries by subtracting the benefit reported in the CPS from the benefit in the PHUS for each CPS respondent who reported receiving Social Security. Taking the mean of the differences, the gross PHUS benefits are $209 lower than the amounts in the CPS, almost the same as the difference in means ($202). The median individual difference was less than $1, substantially less than the difference in medians ($204). The median difference of almost zero indicates that a large proportion of CPS records had benefit amounts equal to the gross PHUS benefit amount.

The means of individual differences are similar to the differences of means. By contrast, some of the medians of individual differences differ substantially from the differences in medians. For example, among people aged 65 or older, the median annual CPS benefit was $226 lower than the median gross PHUS benefit, but the median of the individual differences was less than $1. Likewise, among retired-worker and survivor beneficiaries, the median CPS benefits were $583 and $484 lower, respectively, than the median gross PHUS benefits, but the medians of the individual differences were both less than $1.

We also assessed whether CPS benefit amounts approximate net Social Security benefits. In general, they do not. The amounts reported in the CPS approximate the gross PHUS benefit amounts more closely than they approximate the net PHUS amounts. The mean annual gross PHUS benefit was $202 less than the mean annual CPS benefit, while the mean annual net PHUS benefit was $1,190 lower than the mean annual CPS benefit (Table 6). The median net PHUS benefit was $833 lower than the CPS median.

The differences between net Social Security benefits in the PHUS and the benefits reported in the CPS varied by beneficiary age and type. However, in all three age groups and in three of four benefit categories, the mean CPS benefit amounts were higher than the mean net PHUS benefits, and the median CPS benefit exceeded the median net PHUS benefit in all categories. Overall, the results provide evidence that the Social Security income reported in the CPS approximates gross Social Security benefits.

Also in Table 6, we compare the mean and median net benefit payments to individual respondents in the PHUS and the CPS. The overall mean individual difference of -$1,199 is similar to the overall difference in means of -$1,190. If the Social Security benefit recorded on the CPS is equal to the gross benefit, the median individual difference between the PHUS net benefit and the amount in the CPS for Medicare-eligible beneficiaries will equal the Medicare premium. In 2008, the Part B premium was $96.40 per month, or $1,157 in total for the year, so we would expect to see a median individual difference of about -$1,157. The median individual difference of -$1,155 for beneficiaries aged 65 or older in Table 6 suggests that for nearly all Medicare-eligible beneficiaries, the benefit recorded on the CPS is equal to the gross Social Security benefit.

Conclusion

Social Security is a major source of income for retired persons in the United States, and Social Security income strongly affects beneficiaries' economic security and poverty status. Because the Social Security benefit formula replaces a higher proportion of lifetime average wages for lower earners and also provides benefits to disabled workers and survivors, it provides a safety net for people in the lower half of the income distribution. It is important for household surveys of income to measure Social Security benefits accurately in order to generate reliable estimates of household income and the percentage of persons in poverty. To correctly estimate the income of the elderly using data from household surveys, researchers need to know whether the amount of Social Security income reported by respondents represents the gross amount of the Social Security benefit or the amount net of deductions for Medicare premiums. The difference is not trivial. For most Medicare participants, the Part B premium in 2013 is $105 per month, which amounts to about 8 percent of the mean monthly retired-worker benefit. Additionally, both members of a married couple must pay Medicare premiums. Moreover, federal income tax liability, the official poverty rate, and income limits for federal programs such as Supplemental Security Income, the Supplemental Nutrition Assistance Program, and the Low-Income Home Energy Assistance Program are based on income before Medicare premiums have been deducted.

Our analysis suggests that among SIPP respondents aged 60 or older in 2009, the mean annual gross Social Security benefit in SSA records exceeded the mean benefit reported in the SIPP by 7.6 percent, and the median gross benefit in SSA's records exceeded the SIPP median benefit by 9.5 percent. Among beneficiaries aged 65 or older, the mean gross benefit in administrative records exceeded the SIPP benefit by 7.3 percent, and the median gross benefit in SSA records exceeded the median SIPP benefit by 8.6 percent.

Although survey documentation indicates that both the CPS and SIPP record the gross Social Security benefit amounts, our results suggest that the benefit amounts recorded in the SIPP reflect net benefits more closely than they reflect gross benefits. Much of the difference in the amounts recorded in the PHUS and the SIPP could reflect the omission by SIPP respondents of the Medicare premiums that SSA deducts from most Social Security benefit payments. However, some of the difference may also reflect response errors from about 10 percent of SIPP respondents who received payments from SSA, but reported no payment in the SIPP. As a result, interpreting the Social Security benefit amounts in the SIPP as representing total Social Security income could lead analysts to underestimate total income and to overestimate the proportion of Social Security beneficiaries who are in poverty or are eligible for federal income support programs.

In contrast to the SIPP, the mean and median Social Security benefit amounts recorded in the CPS approximate the gross benefit amounts. In fact, the mean 2008 benefit reported in the CPS was $202 (or 1.6 percent) higher than the mean gross benefit recorded in the PHUS for CPS respondents. The median CPS benefit was about $204 (1.6 percent) lower than the median gross PHUS benefit.

Finally, it is interesting to note the differences in Social Security income recorded in the PHUS between SIPP participants and CPS participants. The mean and median Social Security benefits in 2009, as recorded in the SIPP for beneficiaries aged 65 or older, were $12,823 and $12,720, respectively (Table 2). The comparable values for 2008 in the CPS were $13,012 and $12,677, respectively (Table 5). Even though the SIPP reflects benefits in 2009 and the CPS reflects benefits in 2008, those figures are very similar. The means differ by less than 2 percent and the medians differ by less than 0.5 percent.

We have documented that the mean and median benefits recorded in the PHUS for CPS respondents differ very little from the benefit amounts for 2008 recorded in the CPS, which in turn are similar to the mean and median benefits recorded in the SIPP for 2009. It is somewhat surprising that the mean and median benefit amounts recorded in the PHUS for SIPP respondents are greater than the amounts reported in the SIPP—and are also greater than the amounts in the CPS and in the PHUS for CPS respondents. A number of factors, including sample selection, sample weighting, and imputation procedures could contribute to this discrepancy. Although the cause of the difference is beyond the scope of this analysis, the question warrants further investigation.