AVERAGE WAGES FOR 1985-90 FOR INDEXING UNDER THE
SOCIAL SECURITY ACT

Introduction

Under the Old-Age, Survivors, and Disability Insurance (OASDI) program, all persons newly eligible for benefits after 1978 have their benefits computed using a procedure which first indexes their earnings taxable under Social Security. This indexation is designed to reflect the changes in levels of average wages in the economy over a person's working lifetime. Accordingly, the Social Security Act requires an annual determination of the national average wage¹ to update the series of average wage amounts used for this indexing purpose. The law also requires that the series be used to index the earnings intervals (or "bend points") in the formulas for computing Primary Insurance Amounts and maximum family benefit amounts, as well as to index other program amounts such as the OASDI contribution and benefit base. See the appendix to this note for more information on wage-indexed program amounts.

This note documents the determination of the average wage amounts for 1985-90. Documentation of the series for years 1951-84 may be found in Actuarial Notes 103, 112, 115, 119, 124, and 126.

Percentage increases in the average wage series are used for all indexing purposes. The nominal amounts in the series are, therefore, relatively unimportant. Changes in data sources can be made provided that a single data source is used to measure a particular annual percentage increase in the average wage.

Beginning with the national average wage for 1978, wages used in the determination of the average have been defined as those subject to Federal income taxes. Initially, data on such wages came from income tax returns processed by the Internal Revenue Service (IRS). Later, the data came from Form W-2's processed by the Social Security Administration (SSA). This note will discuss the transition and will compare the wage data from these two sources. To provide perspective on the transition, the note will include a brief summary of relevant information from earlier Actuarial Notes.

The "Omnibus Budget Reconciliation Act of 1989" amended the method used to determine the average wage amounts for 1991 and later by including contributions to certain deferred compensation plans. (Such contributions are essentially income-tax-deferred wages.) The amendment also provided a transitional rule which required determination of special "deemed average wage" amounts for each year 1988-90. These deemed average wages were designed to increase the levels of the OASDI contribution and benefit base to what they would have been if such deferred compensation contributions had always been used in calculating the average wage series. These changes to the Social Security Act will be discussed in detail toward the end of this note. 

Transition From IRS To SSA Data

Legislation designed to reduce the wage reporting burden on employers was enacted in 1976 and revised in 1977. The legislation eliminated the then existing requirement that employers report wages for each employee on a quarterly basis. It also permitted IRS and SSA "to enter into an agreement for cooperative processing of a revised annual wage reporting form (i.e. Form W-2) ... "² Prior to 1978, the average wage had comparatively limited uses as an index and had been based on first quarter wage reports. Recognizing the need for a transition from quarterly to annual wage reporting, which became effective for wages earned in 1978, the Senate Committee on Finance wrote that for "1977 and 1978, form 1040 data would be used and after 1978, forms W-2 data would be used."³ The "Code of Federal Regulations" (see §404.211(c)) thus defines the average wage, for years after 1977, to mean "all remuneration reported as wages on Form W-2 to the Internal Revenue Service for all employees for income tax purposes, divided by the number of wage earners ... "

Following the legislative guidelines, SSA contracted with IRS for wage data from income tax returns for 1977 and 1978. SSA decided to extend its contract with the IRS due to SSA's difficulty in processing the large volume of W-2's and the need to announce the average wage by a statutory deadline. SSA was finally able to process the annual W-2 data with a sufficient degree of completeness, and in time to meet statutory deadlines for announcing the average wage, in 1986 for wages earned in 1985.

Average wage determinations for 1978-84 were based on wage data collected by IRS during their processing of annual tax returns for 1977-84. Under the contract, IRS recorded the amount of wages subject to Federal income taxes (described on the tax return as "Wages, salaries, tips, etc.") from each return. This amount generally equals the total of corresponding wage amounts from attached Form W-2's. IRS, however, did not check for discrepant wage amounts in all cases.

Also under the contract, IRS examined the W-2 forms attached to each joint tax return to distinguish those returns with one wage earner from those with two wage earners. This distinction was essential to providing the total number of wage earners. The "raw" average wage was then computed as the aggregate amount of wages subject to Federal income tax divided by the total number of wage earners. The ratio of the raw average for the latest year's data to the corresponding average for the prior year's data was multiplied by the last official average wage to produce the next official indexing series amount.

Under the laws in effect for tax years 1983-86, married couples filing joint tax returns could reduce their taxable income if they filed a Schedule W with their tax return to take advantage of the so-called two-earner deduction. For 1985 tax returns, the IRS decided to replace the examination of W-2's attached to joint tax returns with an examination of Schedule W's attached to tax returns. (Examination of the Schedule W's, rather than a simple count of the number of such forms, was needed to distinguish wage earners from those who were self-employed.)

Unfortunately, the decision to distinguish the number of wage earners on joint returns by examining Schedule W's heavily skewed the apparent number of wage earners because not all couples eligible for the two-earner deduction elected to complete Schedule W's. Thus, the number of joint returns with 2 wage earners was underrepresented in IRS's tabulation of 1985 wage data, causing an understatement of the number of wage earners and a corresponding substantial overstatement of the average wage.

Because the IRS data for 1985 lacked the proper split of joint returns between one-wage-earner couples and two-wage-earner couples, an estimate of the split was derived from the Taxpayer Usage Study (TPUS). The TPUS is a small, detailed sample of returns maintained by the IRS. Using the estimated proportion of two-wage-earner returns to total joint returns from the TPUS led to an estimated average wage of $16,861.31 -- a 4.50-percent increase over the official 1984 average wage.

If SSA had not been steadily improving its own processing of wages from Form W-2's, the estimated average wage for 1985 given above would have become the official average wage for that year. Owing to improved processing speed, however, SSA had processed nearly all of the W-2's for 1985 by the time a choice had to made between using the estimated average wage based on IRS data and an average derived from SSA's own data. Given the uncertainty of the estimate using IRS data, the decision to use SSA's data was easily made.

Comparison of SSA and IRS Data

As noted, beginning with the average wage for 1978, the average has been based on wages subject to Federal income taxes. Such wages were first captured from income tax returns processed by IRS. Later, the wages were from Form W-2's processed by SSA. Each data source has advantages and disadvantages.

With successive tabulations of tax returns for a given year, more and more wage earners are included in the average. With each additional wage earner included in the average, his or her complete wages for that year are also added. Thus, successive tabulations of a year's wages have relatively little trend effect on the average wage.

SSA data, on the other hand, is based on employer-filed Form W-2's. An individual may work for more than one employer during a year, either by holding more than one job simultaneously or by changing jobs during the year. For this and other reasons, the total number of W-2's filed by employers will exceed the total number of employees. At a given point in time, SSA may not have received and processed all the W-2's associated with the individuals identified as having at least one W-2 for the year. Thus, unless virtually all W-2's have been received, this process will tend to understate the average amount of wages per worker. As a result, the average wage increases with each successive tabulation of a given year's wages. (Table 1 in the next section illustrates this phenomenon.)

Theoretically, because of different filing deadlines, SSA should be able to process a higher proportion of the W-2 forms filed compared to tax returns processed by IRS. Employers must file their earnings statements with SSA by the February 28 following the year in which wages were earned. Tax returns, on the other hand, can be filed as late as April 15, and extensions of 4 months or more are granted in certain circumstances. The statutory deadline for promulgating the average wage for a given year is November 1 of the following year. To meet this deadline, SSA needs to have the raw data by early October. In the case of the W-2's processed by SSA, SSA should have virtually all the necessary data available by early October. For IRS, on the other hand, having complete data by early October is much less likely.

Another difference is that SSA data is more representative of the entire working population. All employers must distribute W-2 earnings statements to their employees and to SSA. In contrast, IRS data reflect only those employees who file income tax returns. People with wages below specified minimums are not required to file, thus raising the calculated average. Furthermore, minimum-filing requirements can change over time, thereby affecting the average wage from one year to another. A factor that may lower the calculated average is the expectation that people with high incomes are more likely to have complicated returns and file for extensions. Their tax returns are therefore less likely to be processed by October when the official average wage must be determined. The size of these two groups and their effect on the average wage series for 1978-84 is unknown.

As noted earlier, IRS did not check wages on tax returns against wages reported on attached W-2's. If an individual's W-2's were missing from a joint tax return with two wage earners, the person would not be counted as a wage earner even if the individual reported his or her wages on the return. Thus, IRS only provided an estimate of the number of wage earners, and the estimate was probably somewhat low.

As far as errors on W-2's are concerned, the W-2's that employees attach to their income tax returns are checked to some extent by the employees themselves. (One is not likely to attach a W-2 to his or her income tax return if the W-2 shows $2,500,000 in wages and actual wages were only $25,000!) The W-2's employers send to SSA are not necessarily exact duplicates of those they give to their employees. Thus, it's likely that wages reported by employees to IRS will be more accurate than those reported by employers to SSA.

The wages processed by IRS are manually transcribed -- first by the person preparing the tax return and then by IRS in preparing an electronic record. Manual transcription is a common source of error. In contrast, any employer with 250 or more employees must report W-2 data to SSA for each employee in computer-generated files. Therefore, the more automated data processing used by SSA, compared to the IRS processing, would tend to offset any advantage IRS data might have in terms of accuracy. On balance, then, no firm conclusions on the relative accuracy of the data can be drawn without further study.

The advantages and disadvantages of SSA data compared to IRS data can now be summarized as follows:

• SSA data is more complete than IRS data; i.e., a higher percentage of all workers will have their wages for a year processed by the time the average wage is determined
• SSA data is more representative of all employees
• SSA data is unaffected by changes in income tax filing requirements
• A more accurate count of the number of wage earners is possible with SSA data

• The average wage from SSA data is more sensitive to the degree of completeness and therefore needs to be as close as possible to 100 percent complete to ensure accurate year-to-year increases in the average wage

SSA Average Wage Data

Because SSA maintains earnings records for benefit computation purposes, SSA must collect data on wages subject to Social Security taxes. Recall that such wages differ from those used for indexing purposes. Because the wages used for indexing are those subject to Federal income taxes, such wages include wages in employment not covered by Social Security and covered wages in excess of the maximum annual amounts subject to taxation under the OASDI or Hospital Insurance programs. These wages may also exclude compensation, such as contributions to certain deferred compensation plans, that is taxable under Social Security. Throughout the remainder of this section, the term wages will be used to mean wages subject to Federal income tax -- not wages covered or taxable under Social Security.

For each year after 1977, SSA has produced reports that show the number of workers and the aggregate amount of their wages. Normally, a report is produced four times a year.

The Social Security Act requires that the average wage be promulgated in the Federal Register by November 1 of the year following the year in which the wages were earned. Table 1 shows the number of workers and their wages from the last report prior to the November 1 deadline and corresponding data from the final⁴ report for a year, produced in the second following year. The data are shown for calendar years 1981-90. Data for earlier years are not shown because SSA had not processed the data fast enough to produce any reports by November 1.

1. Completeness and accuracy

Ideally, 100 percent of the wage data for a year should be processed and tabulated so that the average wages for the year can be accurately promulgated in the Federal Register by the statutory deadline--November 1 of the following year. How close actual processing comes to this ideal, based on the available reports, can be measured in two ways. One measure is the ratio of the average wage from a pre-final report to that from the corresponding final report. The closer this average wage measure is to 100 percent, the more complete the processing. Another measure is the ratio of the number of workers from a pre-final report to the final number. Table 1 shows both such measures.

Table 1 indicates dramatic improvement in the processing of 1984 wage data, in terms of the percentage of workers whose wages were processed, and a smaller improvement in processing the 1985 data. Using the average wage measure of completeness, however, the improvement in processing 1985 wage data was even more significant than the prior year's improvement. In the 1985 data, nearly 98 percent of the workers represented in the final report were represented in the October⁵ report. The October average wage, as a percentage of the final average, jumped to 99.66 percent -- up from 93.74 percent for the corresponding 1984 data.

As shown in table 1, the average wage amount for any given year always increases from the October report to the final report for that year. In general, the higher the ratio of workers represented in the October report to those in the final report, the closer the October average wage is to the final average wage. In fact, for wages earned in 1985 and in 1987-90, when the former ratio was greater than about 97.5 percent, the October average wage was within 0.5 percent of the final average wage. As noted previously, this increase in the average wage by report date occurs because additional W-2 forms are processed for those people with multiple employers, increasing the amount of wages for these people without increasing their number.

Table 2 shows the year-to-year percentage increases in the average wages from October tabulations (column 2). Fluctuations in the percent increases are much greater than those for the corresponding averages from final reports (column 4) in the early 1980's, but after 1985 the two columns appear much more alike. The large fluctuations reflect the lack of completeness, and resulting understatement of the average wage, in the October series during the early 1980's. Generally speaking, the more complete the processing of W-2's, the more accurate the average wage.

Accuracy of Social Security's wage-indexed calculations requires that year-to-year changes in the average wage indexing series should be correct. The absolute level of the series is immaterial. As noted before, data from IRS was used to determine the official wage-indexing series for 1978-84. Table 2 shows that the year-to-year increases in average wages from 1981 to 1984, based on SSA's W-2 data tabulated in the final reports for those years (column 5), was very close to the corresponding increase in the official wage-indexing series (column 7). Thus, the final SSA data for the four years 1981-84 is believed to be as accurate as the accepted data from IRS. The percentage increase based on SSA's data from October reports, on the other hand, was clearly unreliable during that period.

2. Determining the 1984 base average wage

As discussed earlier, SSA decided to use its own wage data for 1985,rather than an estimate based on IRS data. SSA then had to decide which 1984 wage tabulation to use in the determination of the annual wage increase. Three possible choices were the SSA wage data tabulated in September 1985, November 1985, and March 1986, the last being the final report of 1984 wages. The average wages associated with these choices and the resulting increases to the $15,900.51 raw average wage for 1985 (October 1986 report) are summarized below:

The September 1985 data were clearly incomplete, resulting in a major overstatement in the percentage increase. Based simply on the timing of the reports, both the September and November 1985 reports would appear to be fairly comparable to the October 1986 report. Such appearance, however, does not reflect the significant increase in processing speed that occurred in 1986. (For example, over 10 million more workers with processed wages were reported at the end of August 1986 than one year earlier.)

The 1984 raw average wage from the November 1985 report ($15,077.33), as a percentage of that from the final March 1986 report ($15,250.75), was 98.86 percent. With the increase in processing speed, the corresponding ratio of 1985 raw averages (from the October 1986 report and the final report) was expected to be closer to 100 percent than to 98.86 percent.⁶ This lent support to preferring the higher $15,250.75 average in the calculation of the 1985 average wage index.

The increase in average wages from 1984 to 1985, using the $15,250.75 raw average as a base, was 4.26 percent. This increase was much closer to the increase that had been anticipated by economists at the time. For example, in July 1986 SSA economists estimated that the increase in average wages from 1984 to 1985 was 4.03 percent. In contrast, the increase from the November 1985 reported average, $15,077.33, to the October 1986 reported average, $15,900.51, was significantly higher -- 5.46 percent.

As described earlier, an estimated 4.50-percent increase in average wages from 1984 to 1985 was derived from IRS data. This estimate also suggested that the better choice for the 1984-85 increase was 4.26 percent. Thus, the available evidence supported a decision to use the final, March 1986 report of 1984 average wages as a base for measuring the increase in average wages. Multiplying the resultant 4.26-percent increase in raw average wages from 1984 to 1985 by the previously determined official 1984 average wage, $16,135.07, gave the official 1985 average wage of $16,822.51.

As shown in table 2, the increase in average wages from 1984 to 1985, based on final reports for both years, was 4.62 percent. While this information was unavailable at the time of the 1985 average wage determination, the 4.62-percent figure substantiated the decision to use the final 1984 average wage report as a base for computing the 1984-85 increase. The comparison with the increase based on final reports shows that the 4.26-percent increase that was used to determine the official 1985 average wage was somewhat too low, but was still a better choice than the 5.46-percent alternative. It also indicates that the estimate based on IRS data proved to be quite accurate -- at the time, however, there was considerable reluctance to rely on unproven sample data.

As just noted, the 4.26-percent increase in average wages from 1984 to 1985 was somewhat too low. Increases in the processing speed of SSA wage data in following years, however, tended to bring the ratio of average wages from October reports to final reports closer to 100 percent. The improvement toward 100-percent complete processing, in turn, gave somewhat larger increases in average wages than would otherwise have occurred, thereby compensating for the slightly low 1984-85 increase.

Table 1 indicates that wage processing for 1986 was somewhat less complete than in 1985 or 1987-90. Table 1 shows that the number of workers reported with wages in 1986 increased only slightly from the corresponding figures for 1985 in both the October report and the final report. Table 1 also shows that the ratio of the October reported average wage for 1986 to the final average (98.97 percent) was significantly less than corresponding ratios for both the year before (99.66 percent) and the year after (99.77 percent). As shown in table 2, these factors combined to cause a low increase -- 2.97 percent -- in the official average wage. This increase was significantly lower than that based on final data (3.69 percent), which itself was probably too low due to incomplete processing.

The October reported average wage for 1987 was 99.77 percent of the final average, indicating a return to the more complete processing that had marked the 1985 data. Because the 1985-86 increase was too low and because processing speed had returned to, or even slightly exceeded, the speed at which 1985 data was processed, the 1986-87 increase was too high. The cumulative increase from 1985 to 1987, however, was satisfactory.

Estimates of increases in average covered wages, based on a one-percent edited sample from SSA's Master Earnings File, tend to confirm the above analysis.⁷ This sample indicates an approximate 4.8-percent 1985-86 increase and a 4.4-percent 1986-87 increase. The cumulative 1985-87 increase of 9.45 percent falls between the 9.53-percent cumulative increase based on the official average wage indexing series and the 9.41-percent cumulative increase based on final reports for 1985 and 1987.

There are factors affecting the average wage over which SSA has only limited control or capability to correct. One significant factor is reporting errors by employers and by service bureaus engaged by employers to report the wages of their employees. Some of these errors can be detected by noting inconsistencies with other wage or tax data reported. If an employer report is so flawed, it will be sent back to the employer for correction. Nevertheless, some incorrect employee wage records do get posted to SSA's earnings file.

Consistency checks are applied to the data from the earnings file as the data are compiled for average wage reports. (This does not alter or affect the records in the earnings file.) Checks are added or modified as different types of errors are discovered.

As noted earlier, incomplete posting of wage reports can distort the average wage series. With regard to 1990 wage data, for example, processing of W-2 forms submitted on paper (as opposed to magnetic media) was slowed because the 1990 W-2 forms were larger and could not be electronically scanned as quickly as prior-year forms. Although virtually all such paper forms were processed in time for the 1990 average wage determination, the deadline was just barely met. Without improved capability in processing paper W-2's, slow processing could cause a degree of incompleteness and hence distortion in future average wage determinations.

Finally, as noted above, employer reporting errors can cause incomplete posting of wage records. When the final report on 1990 wages was tabulated this year, the Department of Health and Human Services (HHS), which includes SSA, had still not filed a corrected 1990 wage report with SSA. So for 1990, the wages of all HHS employees were left out of both the official determination and the final report. 

Recent Legislation Affecting Average Wages And The OASDI Taxable Maximum

As mentioned earlier, the transitional rule was designed to increase the OASDI contribution and benefit base. The following table indicates that the bases for 1990 and 1991, as a result of the 2-percent additive component of the deemed average wage, were $900 higher than they otherwise would have been. The table also indicates that this difference dropped to $600 for the 1992 base, reflecting the usage of actual deferred compensation data.

The new legislation, as noted earlier, requires that the average wage index for 1991 and later reflect the combined growth of both wages and contributions to deferred compensation plans. The average wage index for 1991, in particular, will be equal to the product of (1) the average wage index for 1990, times (2) the ratio of the raw average wage for 1991 to that for 1990,¹¹ where each of the raw averages reflects inclusion of deferred compensation plan contributions. This process will prevent any further widening of the gap between average covered wages and the average wage indexing series. It does not, however, place the indexing series at the level it would have been if deferred compensation had always been included in the definition of wages. As noted previously, the deemed average wage series was designed on the latter basis in order to increase the maximum contribution base by a greater amount, relatively, than the other wage-indexed program amounts.

APPENDIX

The following amounts are determined each year on the basis of increases in the average wage indexing series:

• the OASDI contribution and benefit base--the maximum amount of earnings subject to OASDI taxes and creditable towards benefits;
• the contribution and benefit base that would have resulted if the "Social Security Amendments of 1977" had not been enacted--referred to as the "old-law" base;
• the Hospital Insurance (HI) contribution base--the maximum amount of earnings subject to HI taxes;
• the retirement earnings test exempt amounts;
• the amount of earnings required for a quarter of coverage;
• the dollar amounts ("bend points") in the formula for the Primary Insurance Amount; and
• the dollar amounts ("bend points") in the maximum family benefit formula.

The table below shows the average wage indexing series for 1976-90 and the wage-indexed amounts for 1976-92.