The authors are with the Office of Retirement Policy, Office of Retirement and Disability Policy, Social Security Administration.
The findings and conclusions presented in this paper are those of the authors and do not necessarily represent the views of the Social Security Administration.
Since 1999, the Social Security Administration (SSA) has mailed an annual Social Security Statement to individuals aged 25 or older showing their reported earnings history. The Statement also provides workers with estimates of benefit amounts available under the Social Security programs. This note focuses on estimates of retirement benefits in the annual Social Security Statement.
Estimated benefits are adjusted for economy-wide average wage growth from about the time of Statement issuance to about the time of retirement. In addition, the Social Security Statement uses certain assumptions about current and future individual earnings to estimate retirement benefits. This note evaluates whether the Statement's assumptions produce accurate estimates, using data from a sample of recently eligible retired workers in SSA's Modeling Income in the Near Term (MINT) microsimulation model.
The Primary Insurance Amount Formula
The primary insurance amount (PIA) is the basis for calculating monthly Social Security benefits. The PIA formula is used both in computing actual benefits and in preparing future benefit estimates on the Statement.
Consider the computation for a retirement benefit-eligible worker turning 62 in 2008. The PIA formula's earnings input is the worker's highest 35 years of Social Security taxable earnings, as determined by his or her average indexed monthly earnings (AIME). Indexed earnings have been adjusted to account for wage growth in the economy, as measured by SSA's average wage index (AWI) from a given year of earnings to the year the worker attains age 60. The PIA for this worker would be computed as follows.
2008 PIA formula (equation 1):
(a) 90 percent of the first $711 of AIME plus
(b) 32 percent of AIME over $711 and up to $4,288 plus
(c) 15 percent of AIME over $4,288.
The dollar amounts in the initial PIA formula, referred to as bend points, are adjusted annually for economy-wide wage growth via the AWI. Bend points increase over time, as illustrated by the lower bend points in the 2001 PIA formula.
2001 PIA formula (equation 2):
(a) 90 percent of the first $561 of AIME plus
(b) 32 percent of AIME over $561 and up to $3,381 plus
(c) 15 percent of AIME over $3,381.
The bend points in equation 1 are about 27 percent higher than those in equation 2 because the AWI grew by about 27 percent over the 7-year period from 1999 (2 years before the Statement) to 2006 (2 years before reaching age 62—the earliest age at which retirement benefits can be received). These 2-year offsets reflect the lag in updating average wage data.1
Benefit estimates in the Statement rely on the PIA formula in effect for the year the Statement is actually sent. To illustrate, assume an eligible worker retires at age 62 in 2008. The monthly benefit at that age would be based on the PIA computed from equation 1. However, the Statement sent to the same worker in 2001 (when he or she was age 55) would have a monthly benefit estimate for age 62 based on the PIA computed using equation 2. The monthly benefit estimated in the Statement for age 62 can thus be interpreted as the future benefit amount deflated for economy-wide wage growth that occurs from 2 years before the Statement to 2 years before the earliest age of retirement benefit eligibility (age 60).
As noted above, benefits are estimated using the PIA formula for the year the Statement is sent. When computing the AIME that is used in the PIA formula, current-year and all future preretirement earnings are assumed to be equal to observed earnings in the latest full year. For example, if the worker earned $40,000 last year, it is assumed he or she nominally earns $40,000 in the current year and will continue to nominally earn $40,000 in all future years until the relevant retirement age.2 If last year's earnings were zero, then earnings from 2 years ago are assumed to equal the individual's earnings from the current year until retirement. Zero earnings in the last 2 years result in the assumption of no current and future earnings.
Earnings for more than 2 years before the Statement are indexed for economy-wide growth in wages from the time of the earnings to the second year prior to the Statement. For example, if a 55-year-old worker received a Statement in 2001, the earnings from 1980 would be increased by a factor that reflects wage growth in the economy from 1980 to 1999. The AIME used in the PIA formula that is in effect the year the Statement is sent is based on these indexed prior earnings and on assumed current and future earnings.
Research Goals, Data, and Methods
The goal of this research is to assess the accuracy of Statement PIAs in predicting actual PIAs at age 62, after accounting for economy-wide wage growth, using earnings histories and other data for recent retirees. The results will test the validity of the Statement assumptions outlined in the preceding section.
The data in this note come from MINT,3 a microsimulation model that matches SSA earnings and benefit records to data from the Census Bureau's Survey of Income and Program Participation (SIPP). The current version of MINT is benchmarked to the intermediate assumptions of the 2004 OASDI Trustees Report. Program values such as the AWI are historical through 2002, and projected after 2002 using the intermediate assumptions. Individual earnings histories through 2002 are taken directly from SSA records; earnings after 2002 are simulated using MINT.
The overall sample for this note comprises individuals born from 1940 through 1946 who are eligible for Social Security benefits. Depending on the birth cohort and the year of analysis, MINT contains historical information, projections, or a combination of the two. For the earliest birth cohort (1940), the data are historical. For the latest birth cohort (1946), earnings data through age 56 are historical and those for ages 57 through 62 are simulated. We present results for the overall sample and for each cohort.
Our analysis measures the accuracy of the Social Security Statement benefit estimates using variables including:
- Retirement PIA, as calculated at age 62 using earnings through that age. For example, for persons in our data who were born in 1946 and attain age 62 in 2008, we calculate the retirement PIA according to equation 1.
- Statement PIA, as estimated in Social Security Statements. For example, for persons in our data who were born in 1946 and attained age 55 in 2001, we calculate the PIA for Statement age 55 according to equation 2 and Statement assumptions.
- Deflated retirement PIA, derived by dividing the retirement PIA by the ratio of the AWI in the year the person turns 60 to the AWI 2 years prior to the Statement.
- Dollar and percent differences between the Statement PIA and the deflated retirement PIA. For example, if a person in the sample had a Statement PIA of $500 and a deflated retirement PIA of $600, the dollar difference would be minus $100 and the percent difference would be minus 16.7 percent. We calculate differences for individuals and present median and other values in the tables accompanying this note.4
Subsamples from MINT are used for various analyses. As examples, we show some results by Statement age and birth cohort. Note that the official Social Security Statement does not provide estimates unless an individual meets eligibility requirements (that is, 40 quarters of covered earnings). Thus, relatively few persons who receive a Statement at a young age (such as age 25) have benefit estimates displayed. Such persons would not be included in our tabulations for Statement age 25, but would be included in the tabulations for a later age if 40 quarters of coverage were earned by that later age.5
The analysis is historical by design: Our goal is to use completed earnings histories (through age 62) to assess whether Statement assumptions are valid. Although Social Security Statements were not actually received by members of the MINT sample in all years of their working lives, we can replicate the PIAs that would have been used under current Statement methodology. And while the current structure of the PIA formula was first applied in 1979, we can use the AWI to adjust the PIA bend points for years prior to 1979.
A hypothetical example will help illustrate. Consider a worker who was 40 years old in 1980. Using actual earnings through 1979 and the Statement assumptions about earnings after 1979, the PIA using the 1980 formula is $400. Suppose the hypothetical individual is observed (through MINT) to have an actual PIA at age 62 (in 2002) of $1,500. Between 1978 and 2000, average wages rose from $10,556.03 to $32,154.82, thus the AWI increased by a factor of 3.046. So the 2002 $1,500 PIA would be deflated to approximately $492 ($492 = $1,500/3.046). In this example, the Statement PIA would underestimate the deflated retirement PIA by $92, or 18.7 percent ($400 versus $492), because the individual's wages grew at a faster rate than the economy average in the post-Statement period. By using a representative sample from MINT, it will be possible to determine whether the Statement PIAs typically overestimate or underestimate PIAs based on observed earnings.
The main results are presented in Tables 1A–17A. Table 1A provides a summary of the results using the combined cohort sample (persons born from 1940 through 1946). This summary is shown by Statement age—note that earlier Statement ages have smaller sample sizes because fewer individuals were insured for retirement benefits at those ages. Findings include:
- Under current Statement assumptions, the median Statement PIA for individuals in these cohorts at age 25 would have been $174, while the median deflated retirement PIA is $218. At Statement age 40, the respective figures are $495 and $504. For Statements issued at age 55, the figures are $870 and $895.
- Examining statistics based on differences measured at the individual level, we find the median differences are highest at the earlier Statement ages. For example, the median percentage difference between Statement PIAs and deflated retirement PIAs is minus 16 percent at Statement age 25, but narrows considerably at later ages (around minus 1 percent to minus 2 percent at Statement ages 40 or older). Note that these are median differences, not differences in medians.
- A majority of sample members (57 percent) would have received Statements at age 55 with benefit estimates based on PIAs very close (plus/minus 5 percent) to their deflated retirement PIAs. Over three-quarters (78 percent) had values that were within 10 percent. Considering a 10 percent difference to be "fairly close," we note that by Statement age 40 a majority of persons have Statement PIAs fairly close to their deflated retirement PIAs.
Tables 2A–8A display additional results for each Statement age, including breakouts by sex and earnings quintile. To illustrate some of the findings, consider the results for Statement age 40 (table 5A):
- The Statement PIAs are closer to deflated retirement PIAs for men. The median percent difference is 0 percent and 63 percent of men have Statement PIAs within 10 percent of their deflated retirement PIAs. The median value for women is minus 6 percent; only 39 percent have Statement PIAs within 10 percent of their deflated retirement PIAs, and 30 percent have a difference greater than 25 percent.
- Those in the top AIME quintile have Statement PIAs that are close to deflated retirement PIAs.6 Eighty-eight percent of these high earners have their two PIAs valued within 10 percent of each other.
- Persons with zero earnings in the 2 years prior to the Statement tend not to have accurate Statement PIAs. Half of these individuals have Statement PIAs that differ from deflated retirement PIAs by more than 25 percent. If these individuals return to the workforce, they may ultimately receive a more accurate Statement estimate because the Statement assumptions use recent earnings. We also note that at Statement age 55 (table 8A) those with zero earnings have relatively accurate estimates.
Tables 9A–17A display results by single-year birth cohort. Illustrative findings include:
- The overall pattern of increasing Statement PIA accuracy as the Statement age increases generally holds across birth cohorts. For each cohort, the median percent difference is less at age 55 than at ages 25 and 30 (table 9A).
- For each cohort, with only a few exceptions, the Statement PIAs coming to within successively smaller percentage point differences from deflated retirement PIAs increases as the Statement age increases (table 10A).
- For later Statement ages, the Statement PIAs tend to be more accurate among the earlier birth cohorts. At Statement age 55, 71 percent of Statement PIAs for the 1940 cohort fall within 5 percent of the deflated retirement PIAs, but only 42 percent of Statement PIAs for the 1946 cohort come within 5 percent of the deflated retirement PIAs (figure 1). These differences need to be interpreted with caution because the results for the 1940 cohort are based completely on historical data whereas some outcomes for the 1946 cohort are simulated and projected.
- Gender and lifetime earnings effects by cohort (tables 11A–17A) are largely similar to those for the sample as a whole.
Additional tables provide further perspectives:
- Tables 1B–17B repeat tables 1A–17A except that dollar amounts are based on the inflated Statement PIA and the retirement PIA (see footnote 4).
- Tables 1C–8C repeat tables 1A–8A except that cohorts are restricted to a narrower range (1940–1943).
- Tables 1D–8D are also restricted to the 1940–1943 cohorts, but dollar amounts are based on the inflated Statement PIA and the retirement PIA (see footnote 4).
1. Values for the AWI can be found at http://www.ssa.gov/OACT/COLA/awiseries.html. Regarding the lag in reported earnings, please see Considerations for Potential Proposals to Change the Earliest Eligibility Age for Retirement at http://www.socialsecurity.gov/policy/docs/policybriefs/pb2007-01.html.
2. Benefits are estimated for three retirement ages: 62, full retirement age (which varies depending on year of birth), and 70. The PIA will differ by retirement age if earnings after age 62 affect the individual’s AIME. In this note, however, all PIA computations include earnings only through age 62
3. For a description of the MINT model, see Toder and others (2002).
4. For some tables, we also calculate the dollar difference between an “inflated” Statement PIA and the retirement PIA. The inflated Statement PIA is derived by multiplying the Statement PIA by the ratio of the AWI in the year the person turns 60 to the AWI 2 years prior to the Statement. The percentage difference between the inflated Statement PIA and the retirement PIA is the same as the percentage difference between the Statement PIA and the deflated retirement PIA.
5. The official Statement includes retirement benefit estimates for some special cases in which a period of disability has occurred but the person does not have 40 quarters of coverage. In this note, we simply use the 40 quarters of coverage criteria.
6. The AIME quintiles, which are based on nominal AIMEs at age 62, are useful as analytical variables but have an important limitation when cohorts are combined: Members of earlier birth cohorts will have lower nominal AIMEs and will be more likely to inhabit the lower quintiles. Also, when cohorts are combined, the nominal dollar cutoffs cannot be deflated to a single year. Thus, for some combined cohort tables in this note, we simply label the quintiles from lowest to highest and do not show dollar amounts.
Toder, Eric, Lawrence H. Thompson, Melissa Favreault, Richard W. Johnson, Kevin Perese, Caroline Ratcliffe, Karen E. Smith and others. 2002. Modeling Income in the Near Term: Revised projections of retirement income through 2020 for the 1931–1960 birth cohorts. Washington, DC: Urban Institute Press.