Earnings of Black and Nonblack Workers Who Died or Became Disabled in 1996 and 1997

Social Security solvency proposals may affect blacks as a group differently than those of other races because of differences in earnings, mortality, and rates of disability.¹ To provide some background for understanding this issue, this note examines the earnings of workers by age and race, comparing those who recently died or became entitled to Social Security disability benefits with those still alive. It does not analyze any specific proposal for changing benefits.

Findings

Comparing black workers with those of other racial and ethnic backgrounds, we find that:

• Black workers in every age group shown are more likely to die or become eligible for Social Security disability benefits.
• For the most part, black workers in 1996 had lower average lifetime Social Security taxable earnings by age group and vital status, even though they worked a similar number of years.
• Workers in 1996 who had died by the end of 1997 generally had lower average annual earnings per years worked than those still alive. That pattern was more pronounced for black workers under age 50.
• Black workers in 1996 who became entitled to disability benefits in 1997 had somewhat lower average annual earnings than those who did not become entitled. Workers of other racial groups who became entitled to disability benefits had much lower earnings.

Sources of Data

Our analysis is based on an extract file of all workers with Social Security taxable earnings in 1996 taken from the Social Security Administration's (SSA's) 1 percent Continuous Work History Sample (CWHS), a longitudinal sample of earnings, benefit, and demographic data. We updated information on death in our extract file with data from other SSA data files.

Tables 1–8 show the number of workers, average lifetime taxable earnings, average number of years with taxable earnings, and average annual taxable earnings by age, race, and vital status (alive or deceased) at the end of 1997. Because there is more than one way to measure the incidence of disability and death, we tabulated the data in two different manners:

• Tables 1–4 use data for all workers with Social Security taxable earnings in 1996, including retired and disabled workers. Workers classified as deceased as of December 31, 1997, could have died anytime from early 1996 through 1997. Persons receiving Social Security retirement or disability benefits in 1996 are included as long as they had taxable earnings in 1996.
• Tables 5–8 use data for workers with Social Security taxable earnings in 1996, excluding workers who retired, died, or became disabled before January 1, 1997. That narrower selection reduced the number of workers by about 6.8 million.

Our sample is incomplete for workers who became entitled to disability benefits in 1997. The data include benefit awards posted in our file through June 1998, but awards are sometimes not made for several years after filing because of the extended appeals process. We estimate that about half of the workers becoming eligible for disability benefits in 1997 did not make the cutoff for our data. Nevertheless, even though the data are incomplete, they are useful for comparative purposes.

Classification by Race and Age

SSA collects data on race and ethnicity from Form SS-5 (Application for Social Security Card). Responses to the question on race and ethnicity are voluntary. From 1936 to about 1980, SSA limited the racial/ethnic designations to white, black, and other. After that, SSA revised the form to have five racial/ethnic designations in response to Office of Management and Budget (OMB) Directive No. 15 (May 12, 1977), "Race and Ethnic Standards for Federal Statistics and Administrative Reporting." Those designations are white, not of Hispanic origin; black, not of Hispanic origin; Hispanic; Asian or Pacific Islander; and American Indian or Alaska Native. OMB has recently promulgated still another scheme for classifying race and ethnicity.²

Because most of the racial data in SSA's records are based on the pre-1980 reporting scheme, we have combined the data into two groups: black and all other. The "all other" category includes a small number of blacks recorded in SSA's data as Hispanic. We also categorized workers by 10-year age groups: under 30, 30–39, 40–49, 50–59, and 60 or older.

The data show that black workers in each age bracket are more likely to die or become eligible for Social Security disability benefits than are those of other racial groups. For example, of workers aged 50–59 with earnings in 1996, 0.84 percent of blacks but only 0.51 percent of nonblacks had died by the end of 1997 (Table 1). Of workers in the same age bracket who had earnings in 1996 but were not retired, disabled, or deceased at the end of that year, 1.09 percent of blacks but only 0.59 percent of nonblacks started to receive disability benefits in 1997 (Table 5).

Lifetime Earnings

To compute lifetime earnings, we indexed and summed earnings over the 46 years from 1951 to 1996. We did not include earnings from 1937 to 1950 because they are available in the CWHS only as an aggregate for each individual and are not easily indexed. To put the earnings in 1996 terms, we indexed earnings using the same factors that Social Security uses to index earnings used in calculating benefits.³ Note that the measure of lifetime earnings used in this analysis is not the same as the one used in determining Social Security benefits.

Comparison of lifetime earnings between age groups by race seems straightforward. For the most part, blacks had lower average lifetime Social Security taxable earnings by age group and vital status (alive, nondisabled; deceased; or disabled) than nonblacks (Tables 2 and 6). Caution is needed, however, when making comparisons between alive and deceased persons within age categories.

At first glance, a comparison between 10-year age groups shows higher lifetime earnings for workers who died than for those who were still alive (Tables 2 and 6). That comparison is misleading, however, because the two groups are not of the same age. Within each age bracket, those who die tend to be older than those who survive. The deceased therefore have more years of work and higher lifetime earnings—a result that is most pronounced for the oldest age group (Tables 3 and 7).

Average Annual Earnings

To solve the problem created by the age brackets, we considered using individual years of age, but some of the data cells in the CWHS would then be too small to produce reliable estimates. Instead, we divided lifetime earnings by the number of years worked to produce a measure of average annual earnings (Tables 4 and 8). That measure reduces the impact of the number of years worked within age brackets and allows a better comparison of those who died with those still alive.

Looking at average annual earnings, we see that workers in 1996 who had died by the end of 1997 generally had lower indexed lifetime average annual earnings per years worked than those who were still alive. That finding was more pronounced for black workers under age 50 (Table 4). Also, black workers in 1996 who became entitled to disability benefits in 1997 had somewhat lower indexed lifetime average annual earnings per years worked across age groups than those who did not become entitled (Table 8). Workers of other racial groups who became entitled to disability benefits had much lower average annual earnings.

Appendix - Sampling Variability

The tables in this note present data that are from a 1 percent sample file drawn from the administrative records of the Social Security Administration.

Because of sampling variability, estimates based on sample data may differ from the figures that would have been obtained had we used all of the records rather than specified samples. The standard error is a measure of sampling variability. About 68 percent of all possible probability samples selected with the same specifications will give estimates within one standard error of the figure obtained from the compilation of all records. Similarly, about 95 percent will give estimates within two standard errors, and about 99 percent will give estimates within two and one-half standard errors. The standard error of an estimate depends on the design element, such as the method of sampling, sample size, and the estimation process.

Because of the large number of data cells tabulated from the sample files, it is not practical to calculate the standard error for every possible cell. However, we do estimate standard errors for a large number of cells. We used those estimates to fit regression curves to provide estimates of approximate standard errors associated with tabulated counts and proportions.

Tables A-1 and A-2 show the sampling variability and provide a general order of magnitude for similar estimates from the various sample files. Table A-1 presents approximate standard errors for the estimated number of persons from the 1 percent sample file. The reliability of an estimated percentage depends on the size of both the percentage and the total on which the percentage is based. Data in Table A-2 provide approximations of the standard errors of the estimated percentage of persons in the 1 percent sample file. The standard errors are expressed in percentage points, and the bases shown are expressed in terms of the estimated total population.