A Look at Who Pays the Payroll Tax

Report Social Security

A Look at Who Pays the Payroll Tax

August 30, 2001 Over an hour read
D. Mark Wilson
Director

The interim report of President Bush's Commission to Strengthen Social Security clearly identifies the enormous financial problems facing Social Security.2 Without significant changes in the way the program is financed, substantial tax increases, benefit reductions, cuts in other government programs, or increases in publicly held debt will have to occur. According to the Commission, payroll taxes for a couple earning $50,000 per year would have to increase by 37 percent in 2040 in order to cover promised benefits.3

Although Americans are increasingly aware of Social Security's financial problems, they are far less knowledgeable about who pays for, and who benefits from, the Old-Age and Survivors and Disability Insurance (OASDI) programs. Many workers do not realize that their pay stubs report only half of the Social Security taxes they actually pay. Far fewer Americans know that different groups of workers pay different average effective OASDI tax rates depending on their age, income, and other demographic characteristics.4 It is particularly important for policymakers to understand these differences in order to make knowledgeable decisions regarding Social Security reform.

Since 1935, Social Security has been portrayed as a national pension fund to which Americans contribute when they are working and from which they receive benefits when they retire. The reality is different.5 Social Security is managed as a pay-as-you-go income transfer program.6 In any given year, workers pay taxes on their wage and salary income and that tax revenue is immediately paid out to retired beneficiaries.7 Therefore, it is important to analyze who pays Social Security taxes so that policymakers and the public understand the consequences of Social Security reform on the average effective tax rates that various groups would pay.

For example, if policymakers consider resolving the system's financial challenges by increasing payroll taxes, they should know exactly who will be affected by that decision: the income, age, ethnic, educational, occupational, and family characteristics of those workers upon whom the new taxes would be levied. Increasing taxes may further reduce the capability of some covered workers to save, invest, and build wealth for their families. Unfortunately, the current debate on reforming Social Security is filled with conflicting and confusing information about who pays the payroll tax.

Policymakers also should know how tax and benefits changes would affect the income dynamics of the Social Security system. In any given year, Social Security transfers tens of billions of dollars between different groups of individuals and families. If policymakers decide to increase OASDI taxes in order to maintain benefits, they should know what impact that decision would have on the redistribution of income.8

This Report is a reissue of "Who Pays the Payroll Tax?" published by the Heritage Foundation Center for Data Analysis on March 3, 2000. It is being republished to contribute to the deliberations of the President's Commission to Strengthen Social Security and the current Social Security debate. The Report is designed to clarify many of the tax, benefit, and income redistribution aspects of Social Security for workers, families, and all individuals.9 It presents tabulations of the number of Americans who paid taxes and received benefits in 1997 under the OASDI programs by a variety of income and demographic characteristics.10 The data presented here should be employed along with rates of return as a benchmark for evaluating Social Security reform proposals.

A Summary of Findings

The Heritage analysis shows that average effective tax rates and benefit rates (or roughly the taxes paid and benefits received divided by total income) vary significantly across demographic and income groups:

  • Workers with pre-OASDI incomes from $17,813 to $27,732 have the highest average effective social security tax rate (10.3 percent).11 Workers in the highest income group have the lowest tax rate (6.7 percent). Young workers who are 27 to 30 years of age have the highest tax rate of any age group (10.1 percent); and workers with a high school diploma or less have the highest tax rates of any educational group (9.9 percent).
  • As a group, married families with children have the highest average effective payroll tax rate (8.9 percent) and pay the largest share of OASDI taxes (38.3 percent). Married families without children and single persons without children pay the lowest average effective tax rates.
  • Families in the middle to upper income groups pay larger shares of OASDI taxes than do families with lower incomes. Families in the lowest pre-OASDI income decile ($2,867 or less) pay less than 0.1 percent of all OASDI taxes, while families in the highest income decile ($88,206 or more) pay 28.2 percent of all OASDI taxes.12 In 1997, families with incomes of $29,100 or more paid $261.9 billion more in payroll taxes than they received in benefits while families with incomes below that amount received $201.1 billion more in benefits than they paid in taxes.13
  • Social Security tax rates on pre-OASDI income for workers grouped by income are fairly steady as income rises except for the top income group, while the tax rates for all families and individuals grouped by income generally rise as income increases except for the top income group.
  • For all individuals, women have a higher Social Security tax rate than men. On average, women pay an OASDI tax rate of 8.4 percent, or $1,472 per year, while men pay a rate of 8.2 percent, or $2,642. In 1997, the Social Security program transferred $23.9 billion from men to women.14
  • For all individuals, Hispanics and black Americans have higher OASDI tax rates than do whites and Americans of other races.15 Whites and Americans of other races pay tax rates of 8.1 percent and 8.5 percent, respectively, compared with 9 percent for blacks and 9.3 percent for Hispanics. Hispanics and Americans of other races on average paid significantly more in OASDI taxes than those groups received in benefits in 1997.16
  • Social Security tax rates vary significantly by state. Indiana, Mississippi, Michigan, South Carolina, Alabama, and Iowa have the highest tax rates while the District of Columbia, Alaska, Colorado, Louisiana, and Virginia have the lowest tax rates. In all, residents in 10 states receive more in Social Security benefits than they pay in taxes, while residents of 40 states and the District of Columbia pay more in taxes than they receive in benefits.

Overview of the Analysis

The analysis in this Report is based on data from the March 1998 Current Population Survey of the U.S. Census Bureau. A number of adjustments were made to the Census data to reconcile and calibrate them to Social Security Administration data (see Appendix A).17 The Report focuses on the average effective OASDI tax rate and benefit patterns for just one year (1997)--as such, it presents a snapshot in time. Social Security also has important effects on the redistribution of lifetime income. However, the annual data utilized in this Report are of limited use for the analysis of Social Security over time (see sidebar).18

This Report first presents an examination of Social Security's tax and benefit distributions for workers. It then looks at all families and all individuals. The Report includes a methodological note (Appendix A) that describes the database constructed by the Center for Data Analysis for this analysis. All data tables referenced in the Report are contained in Appendix B.

THE PAYROLL TAX BURDEN ON WORKERS

Not all workers are covered by Social Security and have to pay Old-Age and Survivors and Disability Insurance taxes. Workers excluded from coverage fall into five major categories: federal civilian employees hired before January 1, 1984; certain state and local employees who are covered under another retirement system; railroad workers who are covered under the railroad retirement system; household and farm workers whose earnings do not meet certain minimum requirements; and persons with very low net earnings from self-employment.19 In 1997, 6.6 million, or 4.3 percent, of workers were not covered by the OASDI programs.20

Income and Tax Rates. The Social Security system is funded by a payroll tax of 12.4 percent on the earned income (wages, salaries, and self-employment income) of covered workers.21 However, the average effective Social Security tax rate can vary for two reasons: (1) the share of earned income to total income can differ among covered workers; and (2) not all earned income is subject to OASDI taxes.22 Income from other sources--such as public assistance, Social Security and private pension benefits, and interest, rent, and dividend income--substantially affects the share of earned income to total income between workers. Any earned income over the maximum taxable amount--$65,400 in 1997 and $76,200 in 2000--is not subject to the OASDI tax.23

Compared to all individuals, tax rates for workers are higher for all income groups and benefits are much smaller for lower income groups. Tax rates are higher for workers than for non-workers because a much larger share of their total income consists of earned income and is subject to OASDI taxes. Lower income groups of workers receive fewer benefits than all Americans because focusing the analysis on workers excludes most retirees who, on average, have substantially lower incomes and receive significantly higher Social Security benefits.

Average effective Social Security tax rates vary between different groups of workers.

In 1997, workers paid a total of $398.9 billion in OASDI taxes and received $55.3 billion in benefits (see Table B-1).24 The $55.3 billion in benefits is attributed to OASDI beneficiaries over the age of 61 and under the age of 18 who also work; persons who worked for part of the year and were retired or disabled for the rest of the year; and adults who have been assigned the Social Security benefits that their children under the age of 15 receive.25 Over 80 percent of the OASDI benefits that workers received in 1997 went to those over the age of 61.

On average, workers paid $2,715 in taxes, or 9.2 percent of their total income, and received $376 in Social Security benefits, or 1.3 percent of their total income.26 Grouping workers into 10 total income deciles, however, reveals that workers in most total income groups have average effective tax rates above 10 percent (see Table B-1, column 12). 27

Workers in most total income groups have average effective tax rates above 10 percent.

The two groups of workers with total incomes from $4,548 to $9,316 and $56,988 or more have the lowest average effective Social Security tax rates because a smaller share of their total income is subject to the OASDI tax. Workers with total incomes from $4,548 to $9,316 have a tax rate of 9.7 percent because they receive a larger share of their total income as a group from non-labor income sources such as public assistance and Social Security.28 Workers with total incomes of $56,988 or more have a tax rate of 6.8 percent because of Social Security's cap on taxable wages.29

While the previous analysis of tax rates by total income is informative, a more accurate picture of the burden of Social Security taxes and the income dynamics of the OASDI program is gained by classifying workers according to their pre-OASDI income. Pre-OASDI income is the amount of workers' income before any Social Security benefits are received or any Social Security taxes are paid (see Appendix A). Grouping workers, families, or individuals by pre-OASDI income removes the effect that the Social Security program has on the distribution of total income and results in a more appropriate representation of the distribution of OASDI taxes, as well as the receipt and transfer of benefits by income.

Grouping workers into pre-OASDI income deciles decreases the average effective Social Security tax rate for all income groups but does not significantly change the distribution of taxes paid.30 The distribution of Social Security benefits, however, does substantially change. The four lowest pre-OASDI income groups of workers receive 60.6 percent of the Social Security benefits (see Table B-2, column 8) compared with 29.5 percent for the four lowest total income groups (see Table B-1, column 8).31 The small amount of OASDI benefits in the two lowest total income deciles compared with the same pre-OASDI income groups is attributable to the entitlement feature of Social Security benefits. Because entitlement to Social Security tends to lift workers out of the bottom total income decile, the workers that remain in the bottom total income decile are less likely to receive any Social Security benefits at all.

Except for the highest income group, the average effective Social Security tax rate for workers is fairly flat as income rises (see Chart 1). The OASDI tax rate is highest (10.3 percent) for workers with pre-OASDI income from $17,813 to $27,732, and lowest (6.7 percent) for workers with incomes above $59,580 (see Table B-2, column 12). Yet for workers within the broad range of income from $8,908 to $42,889, the pre-OASDI tax rate only varies slightly, between 10 percent and 10.3 percent. Despite some modest differences by income group, compared with individual income tax rates the pre-OASDI tax rates for workers are reasonably flat.

The average effective Social Security tax rate is highest (10.3 percent) for workers with pre-OASDI income from $17,813 to $27,732.

Workers in the two lowest pre-OASDI income deciles ($8,907 or less) pay just 3 percent of all OASDI taxes, while workers in the two highest income deciles ($42,890 or more) pay 44.4 percent of all taxes (see Chart 2). Workers with pre-OASDI incomes from $8,908 to $59,580, however, pay a significantly larger share of OASDI taxes (71 percent) than they receive in income (62.6 percent). (See Table B-2, columns 6 and 7). On average, workers in the lowest income group paid $191 in Social Security taxes in 1997 while workers in the fifth income decile paid $2,062 and workers in the highest income group paid $7,072 (see Table B-2, column 9).

Workers in the lowest income group receive 31.5 percent of their total pre-OASDI income from Social Security, suggesting that these benefits are important for some low-income workers (see Chart 1).32 Social Security benefits as a percent of workers' income quickly falls as income rises, which indicates that benefits are inversely graduated against income. For workers with incomes above $22,470, Social Security benefits account for less than 1 percent of their income.

Net OASDI Benefits.
In 1997, workers in all but the two lowest income groups paid more in taxes than they received in benefits (see Table B-2, column 5). This reflects one of Social Security's major principles: Individuals pay taxes into the system when they are working and collect benefits when they retire.33 In 1997, workers with pre-OASDI incomes of $8,908 or more paid $351.2 billion more in payroll taxes than they received in benefits, while workers with incomes below that amount received $7.3 billion more in benefits than they paid in taxes. Of the $398.9 billion that workers paid into Social Security in 1997, $336.5 billion was used for OASDI benefits to individuals (see Table B-8, column 4) and $62.4 billion went to pay for other federal government programs and to reduce publicly held federal debt.34

Workers in all but the two lowest income groups paid more in taxes than they received in benefits.

As a group, the Social Security program hits hardest those workers who have pre-OASDI incomes of $34,370 to $42,889. In 1997, they paid a larger share of OASDI taxes than their share of income (see Table B-2, columns 6 and 7). They also had the second highest average effective OASDI tax rate (10.2 percent) and the lowest net benefit rate (-9.6 percent) of any income group (see Table B-2, columns 12 and 14).

Age and Tax Rates.
Much of the distribution of OASDI taxes and benefits by income can be attributed to the pattern of income by age. Grouping workers into 10 age groups (deciles) with a roughly equal number of workers in each group reveals that Social Security tax rates vary by age (see Chart 3).35 Workers 27 to 30 years of age pay the highest effective tax rate (10.1 percent), while people over the age of 57 pay the lowest rate (7.3 percent). Tax rates first rise and then fall with age because earned income as a percentage of pre-OASDI income also rises and then falls with age. The oldest age group of workers is the only one that receives more in benefits than they pay in taxes in any particular year.

Average income and OASDI taxes generally rise through age 57 and then fall to much lower levels for workers over the age of 57 (see Table B-3, columns 9 and 10). Average benefits for workers, on the other hand, are fairly low until age 62 when individuals become eligible for Social Security benefits.36 The graduated character of OASDI taxes and benefits with respect to pre-OASDI income can be attributed to the fact that incomes are, on average, higher during ages when OASDI taxes are paid and lower during the ages when benefits are received.

Other Demographic Characteristics and Tax Rates.
Average effective Social Security tax and benefit rates for workers vary by gender, race, marital status, and citizenship (see Chart 4, and Table B-4, columns 12 and 14).

  • As a group, women workers have a higher OASDI tax rate (9.3 percent) than men (8.7 percent), although men and women have nearly the same net benefit rate.37
  • By race, Hispanic workers have the highest OASDI tax rate (9.8 percent) followed by blacks (9.6 percent). Americans of all other races have nearly the same Social Security tax rate as whites (8.9 percent and 8.8 percent, respectively). The net benefit rate by race also varies significantly.38
  • Single workers have a higher OASDI tax rate (9.2 percent) than married workers (8.8 percent), although single and married workers have nearly the same net benefit rate.39
  • Non-citizens who work have a tax rate of 9.7 percent, while workers who are naturalized citizens or who were born in the United States pay 8.8 percent and 8.9 percent of their income, respectively.

Education and Tax Rates.
OASDI tax rates for workers also vary by education (see Chart 4). High school graduates (with no college) and workers who did not graduate from high school have the highest tax rate (9.9 percent). High school graduates, however, pay on average significantly more OASDI taxes than those without high school diplomas, $2,368 compared with $1,380. Workers with some college or a two-year college degree pay an average tax rate of 9.4 percent, or $2,601, compared with 7.9 percent ($4,043) for college graduates (see Table B-4, columns 9 and 12). College graduates have a lower tax rate because a larger share of their earnings is above the taxable wage cap. College graduates do, however, pay an average of $4,043 in taxes compared with $2,368 for high school graduates.

Labor Factors and Tax Rates.
Social Security tax rates for workers vary by labor force sector (see Chart 5). The five major labor force groups are the private sector, the federal government, state government, local government, and the self-employed.40 OASDI tax rates are highest for workers in the private sector (9.5 percent) and lowest for workers in the federal government (6.9 percent).41 Tax rates vary by labor force sector because many federal, state, and local government workers are not covered by Social Security and do not pay OASDI taxes.

OASDI tax rates also vary by the number of hours worked (full-time or part-time work) and the number of weeks worked per year (see Chart 5). Americans who work full-time 48 to 52 weeks per year pay the highest Social Security tax rate (9.1 percent). Americans who work less than half the year pay the lowest tax rates; 7.4 percent if they work full-time and 5.1 percent if they work part-time. Part-year and part-time workers pay the lowest rate because earned income is a relatively small share of their total income. Part-time, part-year workers (one to 26 and 27 to 47 weeks per year) were the only two groups that received more in Social Security benefits than they paid in taxes in 1997 (see Table B-4, column 5).42

THE PAYROLL TAX BURDEN ON FAMILIES

Although few currently employed workers receive OASDI benefits, it is more often the case that significant benefits flow to a non-working family member. Also, many families, particularly those over the age of 61, receive all or most of their income from Social Security. Other families pay significant amounts of OASDI taxes and receive no benefits.

In 1997, families paid a total of $397.9 billion in taxes and received $337.1 billion in benefits (see Table B-5).43 On average, families paid $3,397 in taxes, or 8.2 percent of their pre-OASDI income, and received $2,878 in Social Security benefits, or 7 percent of their income.44 Grouping families into 10 pre-OASDI income deciles,45 however, reveals that the average effective Social Security tax rates for families varies considerably more than they do for workers (see Tables B-2 and B-5, column 12).46

Income and Tax Rates.
Families with lower pre-OASDI incomes have lower average effective Social Security tax rates than do families with higher incomes (see Chart 6). Families with incomes from $48,589 to $88,205 pay the highest OASDI tax rate (9.6 percent), while families in the lowest pre-OASDI income group of $2,867 or less pay the lowest rate (3.2 percent).47 (See Table B-5, column 12.) The average effective Social Security tax rate for families increases steadily as income rises except for the top income group. Average taxes paid in 1997 vary from $26 per family in the lowest income group to $9,582 per family in the highest income group (see Table B-5, column 9).

Families in middle- to upper-income groups pay a larger share of Social Security taxes than do families with lower incomes (see Chart 7). Families in the lowest income group ($2,867 or less) pay less than 0.1 percent of all OASDI taxes, while families in the highest income group ($88,206 or more) pay 28.2 percent of all taxes.

This substantial rise in the share of payroll taxes paid by income is attributable to the increasing share of total income that goes into different deciles as income rises and the strong association between earnings and income. As income rises, earnings account for a greater share of total income except in the highest income decile.48 In the highest income decile, earned income as a share of total income declines because of an increase in the share of that income from interest, rent, and dividend income.49

Families with pre-OASDI incomes from $14,844 to $88,205 pay a larger share of Social Security taxes than their share of income (see Table B-5, columns 6 and 7). This results from the fact that earned income accounts for a greater share of the total income for these families compared with other income groups. Families with incomes below $14,844 and those in the highest family income group ($88,206 and above) pay a smaller share of OASDI taxes than income for two different reasons. A significant portion of earned income in the top family income decile is above the tax cap ($65,400 in 1997) and is not subject to the OASDI tax. For low-income families, public assistance and Social Security benefits combine to reduce the portion of income subject to payroll taxes and the amount of taxes paid on that income.

Families in the lowest income group pay less than 0.1 percent of all OASDI taxes while families in the highest income group pay 28.2 percent of all taxes.

Families in the lowest income groups receive the largest shares of OASDI benefits (see Chart 7). Families with pre-OASDI income of $2,867 or less receive 29.6 percent of all OASDI benefits and families with pre-OASDI income of $2,868 to $8,577 receive 16.4 percent of all OASDI benefits. The high OASDI benefit shares in the two lowest income groups reflects the dependency many Americans have on the Social Security program as their major source of retirement income.

Net OASDI Benefits.
Taking into account both benefits and taxes, or net benefits, reveals significant differences between income groups. As a group, families with low pre-OASDI incomes received significantly more Social Security benefits in 1997 than they paid in taxes, while families in middle- to upper-income groups paid substantially more in taxes than they received in benefits.50 As a group, families with incomes below $8,578 received $151.7 billion more in benefits than they paid in taxes while families with incomes of $29,100 or more paid $261.9 billion more in taxes than they received in benefits (see Table B-5, column 5).51 This reflects the fact that pre-OASDI family income is likely to be significantly lower for elderly retirees than for middle-aged working families. In 1997, the tax and benefit structure of the Social Security program transferred $201.1 billion from middle- and upper-income families to lower-income families. The ratio of combined OASDI taxes and benefits, or net benefits, to family income indicates that, on balance, the Social Security program is very progressive (see Table B-5, column 14).52

Poverty Levels and Taxes.
An alternative way to analyze the distribution of OASDI taxes and benefits by income is to group families by poverty level.53 Families with incomes 2.00 to 3.99 times the poverty level pay the highest average tax rate (9.4 percent), while people living in poor families pay an average tax rate of 6.9 percent (see Table B-6, column 12). As a group, poor families pay very little (1.4 percent) of the OASDI taxes (see Table B-6, column 7). Moreover, near-poor families pay only 6.9 percent of all OASDI taxes, while families with incomes four times the poverty level pay 64.2 percent of all OASDI taxes (see Table B-6,
column 7).

Poor families have the lowest average OASDI tax rate, 6.9 percent.

Families with Children.
Still another way to analyze the distribution of OASDI taxes and benefits is to group families by marital status and the presence of children. As a group, married families with children pay the largest share of OASDI taxes (38.3 percent) and have the highest average payroll tax rate of 8.9 percent (see Chart 8 and Table B-7, columns 7 and 12). This higher payroll tax rate arises because married couples with children are more likely to have a member in the labor force compared with other family groups. They also are likely to be in age groups where earnings are the highest. Many married families also have more than one wage earner.

As a group, married families with children pay the largest share of OASDI taxes.

Married families without children pay 32 percent of all OASDI taxes and have the lowest average OASDI tax rate of 7.7 percent. Single persons without children pay 24.3 percent of all OASDI taxes and have an average Social Security tax rate of 8 percent. These two household groups also receive the largest shares of OASDI benefits because many of these individuals are of retirement age with grown children (who have their own families). Single persons and married families without children were the only two groups in 1997 to receive more in Social Security benefits than they paid in OASDI taxes.

THE PAYROLL TAX BURDEN ON INDIVIDUALS

To examine the pattern of Social Security taxes and benefits for various demographic groups, it also is useful to shift the analysis from workers and families to all persons. The use of "person income" instead of "family income" significantly increases the dispersion of the income distribution relative to families.54 Expanding the analysis from workers to all individuals facilitates the analysis of all OASDI benefits instead of just the share of benefits received by workers. The analysis of individuals, like workers and families, focuses on pre-OASDI income tax rates.

In 1997, Americans paid $398.9 billion in OASDI taxes and received $336.5 billion in benefits (see Table B-8). On average, individuals paid $2,047 in taxes, or 8.3 percent of their pre-OASDI income, and received $1,727 in benefits, or 7 percent of their income.55 Grouping individuals into pre-OASDI income deciles reveals that average effective Social Security tax rates vary significantly more for all individuals than for just workers.

Income and Tax Rates. The average effective OASDI tax rate on individual pre-OASDI income is fairly graduated (see Chart 9). Individuals with lower incomes have a lower Social Security tax rate than do people with higher incomes. The tax rate starts at 5 percent for Americans with pre-OASDI incomes of less than $3,423 and rises to 9.6 percent for persons with incomes from $29,198 to $53,609 before falling to 6.8 percent for the highest income group ($53,610 or more). Average OASDI taxes paid per year vary from $5 per person in the lowest income group to $6,452 per person in the highest income group. (See Table B-8, column 12 and column 9).

Individuals with higher incomes pay a larger share of OASDI taxes than do persons with lower incomes (see Chart 10). Americans with pre-OASDI incomes below $530 paid less than 0.1 percent of all OASDI taxes while individuals in the highest income group ($53,610 or more) paid 31.5 percent of all Social Security taxes. People with pre-OASDI incomes from $11,875 to $53,609 paid a larger share of Social Security taxes in 1997 than their share of income (see Table B-8, columns 6 and 7). Americans with pre-OASDI incomes below $11,875 and those in the highest income group paid a smaller share of OASDI taxes than their share of income.56

Individuals in the lowest income groups receive the largest shares of Social Security benefits (see Chart 10). Persons with pre-OASDI incomes below $530 received 31.4 percent of all OASDI benefits and individuals with pre-OASDI incomes from $530 to $3,422 received 17.4 percent of all Social Security benefits. The high OASDI benefit shares in the two lowest income groups reflects the fact that for 18 percent of Americans over the age of 64, Social Security is their only source of income.57 Despite the relatively high share of OASDI benefits that go to groups of low-income individuals, economists have observed that Social Security's rate of return for low-wage individuals is below the rate of return that is available from other retirement savings options.58

Net OASDI Benefits.
Taking both taxes and benefits, or net benefits, into account reveals important differences between income groups. Persons with low pre-OASDI incomes received considerably more Social Security benefits than they paid in taxes, while people above the median pre-OASDI income paid substantially more in taxes than they received in benefits.59 Individuals with incomes below $16,713 received $227.1 billion more in benefits than they paid in taxes, while persons with incomes above $16,712 paid $289.4 billion more in taxes than they received in benefits (see Table B-8, column 5).60

Social Security benefits as a percent of income falls rapidly when moving from the lowest to highest pre-OASDI income group, indicating that OASDI benefits are very progressive (see Table B-8, column 13). The ratio of combined OASDI taxes and benefits, or net benefits, to individual pre-OASDI income also indicates that the Social Security program progressively redistributes income in any particular year (see Table B-8, column 14).61

Age and Tax Rates.
Much of the distribution of OASDI taxes and benefits by income can be attributed to the pattern of income by age. Grouping individuals into ten age groups (deciles) with a roughly equal number of persons in each group reveals that the young (ages 15 to 22) and the elderly (ages 62 and over) pay substantially smaller shares of OASDI taxes than do Americans ages 23 to 61 (see Table B-9).62 Average income and OASDI taxes generally rise through age 52, decline slightly for the 53- to 61-year-old age group, and then fall to much lower levels for individuals over the age of 61 (see Table B-9, columns 9 and 10). Average benefits, on the other hand, start low and rise continually with age.63 The graduated character of OASDI taxes and benefits with respect to pre-OASDI income can be attributed to the fact that incomes are, on average, higher during ages when OASDI taxes are paid and lower during the ages when benefits are received. In 1997, the Social Security program transferred $272.9 billion from Americans under the age of 62 to individuals over the age of 61.

Social Security tax rates for all individuals vary considerably by age. Tax rates first rise and then fall with age because earned income as a percentage of total income rises and then falls with age. OASDI tax rates rise with age until age 28 and then decline steadily (see Table B-9, column 13). Americans 23 to 28 years old pay the highest tax rate (10 percent), while people over the age of 71 pay the lowest rate (1.5 percent). (See Chart 11.) Moreover, the percentage of individuals with any earned income is high from age 23 to age 53 and then falls as labor force participation declines for persons over the age of 53.

By age, Americans 23 to 28 years old pay the highest tax rate.

Other Demographic Characteristics and Tax Rates.
The Social Security effective tax rate is higher for woman than for men (see Chart 12), partly due to a higher percentage of men with earnings above the Social Security taxable wage cap. On average, women pay an OASDI tax rate of 8.4 percent while men pay a rate of 8.2 percent (see Table B-10, column 12). Women pay on average $1,472 per year in OASDI taxes and receive $1,713 in benefits for an average net benefit of $241 (see Table B-10, column 11). Men, on the other hand, pay an average of $2,642 per year in OASDI taxes and receive $1,741 in benefits for an average net loss of $902. In 1997, the Social Security program transferred $23.9 billion from men to women primarily because men have a higher employment rate and women live longer.64

Women have a higher Social Security average effective tax rate than do men, but they also receive more in benefits. Men pay 63.4 percent of all OASDI taxes, while women pay just 36.6 percent.

Men also pay a substantially larger share of OASDI taxes than do women. Men pay 63.4 percent of all OASDI taxes, while women pay just 36.6 percent. This results from the fact that men have higher employment rates and median earnings than women do. Even so, economic studies show that rates of return from Social Security fall below 3 percent for nearly all women who work--well below the rate of return for short-term certificates of deposit available at most banks.65

Compared with other races, Hispanics have the highest OASDI tax rate.

Hispanics have higher average effective OASDI tax rates than do white, black, and Americans of other races (see Chart 12). Hispanics pay an average 9.3 percent compared with 9 percent for blacks, 8.5 percent for other races, and 8.1 percent for whites (see Table B-10, column 12). On average, Hispanics and Americans of other races pay significantly more OASDI taxes than they receive in benefits compared with white or black Americans.66 Americans of other races have a net tax payment equal to 5.7 percent of their income, and Hispanics have a net tax payment equal to 4.7 percent of their income, compared with 0.7 percent for whites and 1.6 percent for blacks (see Table B-10, column 14).

Married men and women have nearly the same effective tax rate as single Americans. The tax rate for married persons is 8.2 percent compared with 8.3 percent for single individuals. Single individuals, however, receive an average net benefit from Social Security equal to 1.4 percent of their total pre-OASDI income, or $261 per year, while married people have a net tax payment equal to 2.7 percent of their income, or $801 (see Table B-10, columns 11 and 14).67 This difference arises partly because in any given year retired individuals are more likely to be single, while married individuals are more likely to be in higher income groups.

Non-citizens have a higher OASDI tax rate than do U.S. born or naturalized citizens (see Chart 12). Non-citizens have an average effective tax rate of 9.3 percent compared with 8.2 percent for U.S. born and naturalized citizens (see Table B-10, column 12). Non-citizens on average pay significantly more in OASDI taxes than they receive in benefits compared with U.S. citizens.68 Non-citizens pay on average $1,820 per year in OASDI taxes and receive $484 in benefits for an average net tax of $1,336. U.S.-born citizens, on the other hand, pay an average net tax of just $241.

By education, high school graduates have the highest OASDI effective tax rate.

Education and Tax Rates.
High school graduates (with no college education) have the highest OASDI effective tax rate (see Chart 12). High school graduates pay an average tax rate of 9 percent, compared with 8.2 percent for people who did not graduate from high school, 8.7 percent for individuals with some college, and 7.5 percent for college graduates (see Table B-10, column 12). Americans who did not finish high school have a lower tax rate because a larger share of them are retired or are receiving public assistance. College graduates have a lower tax rate because a larger share of their earnings is above the taxable wage cap. College graduates do, however, pay an average of $3,452 in taxes compared with $1,751 for high school graduates.

For all individuals, the burden of Social Security taxes falls hardest on Hispanic women, ages 23 to 28, with just a high school diploma and pre-OASDI incomes of between $29,198 and $53,609.

High school graduates (with no college) and Americans who did not finish high school receive the largest shares of Social Security benefits (see Table B-10, column 8). Both groups account for 66.3 percent of all OASDI benefits primarily because the educational attainment of today's retirees is significantly lower than that of Americans born after 1945.69

State Tax Rates.
Social Security tax rates vary significantly by state (see Map 1). The effective tax rates range from a high of 9.3 percent in Indiana to a low of 6.8 percent in the District of Columbia (see Table B-11, column 12). Indiana, Mississippi, Michigan, South Carolina, Alabama, and Iowa have the six highest tax rates, while the District of Columbia, Alaska, Colorado, Louisiana, and Virginia have the lowest tax rates. West Virginia, on the other hand, receives the largest average net benefit from Social Security benefits, while Alaska pays the largest average net tax (see Table B-12, column 11). In all, Americans in 10 states receive more in Social Security benefits than they pay in taxes, while those in 40 states and the District of Columbia pay more in taxes than they receive in benefits.

CONCLUSION

The tax and benefit dynamics of the Social Security program reflects the policy objectives of the current Old-Age, Survivors Insurance, and Disability Insurance programs. When this analysis of who pays the payroll tax and who receives OASDI benefits is placed alongside the analysis of Social Security's rate of return by income and demographic groups, the true challenge facing Social Security becomes clear. Policymakers attempting to reform Social Security may find it difficult, if not impossible, to do two things simultaneously: maintain Social Security's current structure while improving the system's rate of return.

The traditional "fixes" for Social Security's funding crisis (higher taxes, lower benefits, and increases in the retirement age) all would likely result in increasing the average effective tax rates at each income decile. Put another way, these approaches to resolving Social Security's financial challenge would wo

Appendix A
Methodology

The tabulations in this Report have been developed from the March 1998 Current Population Survey (CPS) conducted by the U.S. Bureau of the Census. The March 1998 CPS contains information for the preceding year on earnings, Social Security benefits, and other demographic characteristics for each person in the survey population aged 15 or older.70 The CPS is used, rather than Social Security Administration (SSA) data, because publicly available SSA data do not contain information on family income or demographics for taxpayers and beneficiaries of the Old-Age, Survivors, and Disability Insurance (OASDI) programs.71

There are numerous conceptual and methodological differences between the March 1998 CPS and the SSA data.72 These differences result in different population, employment, and beneficiary counts, as well as different total OASDI tax and benefit amounts from the two data sources. For example, the SSA's population estimates include federal employees working overseas and their dependents, other U.S. citizens working overseas, and population estimates for Puerto Rico, Virgin Islands, Guam, American Samoa, Palau, and the Northern Mariana Islands (also known as the outlying areas); the March 1998 CPS does not. The SSA data include, but the March 1998 CPS does not, those Americans who were alive during 1997 but who died before the survey was conducted, as well as those who were institutionalized when the survey was taken in March 1998.73

For this Report, a number of adjustments were made in both the SSA data for 1997 and the 1998 March CPS to reconcile the two different sources of information and enable the CPS data to produce estimates that are consistent with the aggregate SSA data. First, the SSA estimates for population, employment, and the number of OASDI beneficiaries were reduced to account for coverage differences between the SSA estimates and the Census Bureau estimates (see Table A-1).74 Similar adjustments also were made in SSA data for the dollar value of covered earnings and Social Security benefits (see Table A-2).

All tabulation weights in the March 1998 CPS file were scaled up by 1.99 percent to calibrate the CPS counts to the adjusted SSA estimates for total employment and the number of Social Security beneficiaries. Increasing all March 1998 CPS weights by a uniform factor results in a slight overestimation of total employment and a small underestimation of Social Security beneficiaries compared with the adjusted SSA data. Specifically, total CPS employment is 147.460 million, compared with 147.359 million from the SSA data--a difference of 101,000 or 0.07 percent (see Table A-3). The total number of CPS Social Security beneficiaries is 38.494 million compared with 38.547 million from the SSA data--a difference of 53,000 or 0.14 percent (see Table A-3).75 This uniform scaling of weights has no effect on averages or on percent distributions calculated from the CPS data.76

Tabulating who pays the payroll tax in the CPS requires identifying those workers who are covered by Social Security. At any point in time, the OASDI program does not cover all workers. In 1997, 95.7 percent of all workers were covered by Social Security.77 Most uncovered workers were federal, state, and local workers covered by other retirement programs; workers covered by the Railroad Retirement program, or students employed by colleges and universities.78 Identifying private-sector covered and uncovered workers in the CPS is fairly straightforward. Data on occupation, industry, earnings, Social Security benefits, and school enrollment were used to identify uncovered private sector workers.79

Identifying covered and uncovered federal, state, and local government workers in the CPS requires a different approach.80 Federal employees under the age of 30 were assumed to be covered by Social Security.81 The CPS records for federal employees over the age of 29 were duplicated into two separate datasets (one representing covered federal workers, the other representing uncovered federal workers). The individual weights in one dataset were reduced to match the total number of federal workers in the dataset with the count of covered federal employees by the U.S. Office of Personal Management. The individual weights in the second dataset were reduced to match the total number of federal workers in the dataset with the number of uncovered federal employees (total federal workers minus covered federal workers). The two datasets were then recombined.82 A similar method was used to identify covered and uncovered state and local workers.83 Total CPS covered employment after all adjustments were made for federal, state, and local workers is 140.317 million, compared with 141.477 million from the SSA data--a difference of 1.2 million or 0.8 percent (see Table A-3).

Once the CPS data for total employment, covered employment, and the number of Social Security beneficiaries were reconciled and calibrated to SSA data, additional adjustments were made to the dollar amount of covered earnings and Social Security benefits in the CPS. First, all wage and salary income in the CPS was multiplied by .9331 and all self-employment income was multiplied by .8268 to calibrate the dollar value of total CPS covered earnings to aggregate SSA data. An additional adjustment was made to covered wage and salary income and covered self-employment income to calibrate taxable earned income to aggregate SSA data on taxable income. Specifically, covered wage and salary income was multiplied by .9752 and capped at $65,400. Self-employment income was multiplied by .9468 and capped at $65,400 minus any taxable wage and salary income a worker might have.84 These two adjustments lower average earned income by 5.7 percent but have no significant effect on the distribution of earned income in the CPS data.

Two significant adjustments were made to the dollar value of Social Security benefits in the CPS to calibrate it to SSA data. First, because of a conflict between the Census Bureau-imposed maximum cap (top-code) on individual Social Security benefits in the CPS ($50,000) and data from the Social Security Administration that reports a maximum annual benefit for an individual retired worker of $15,919, the dollar value of Social Security benefits on the CPS was limited to $29,999 (the old Census top-code).85 Second, all Social Security benefit income was multiplied by 1.0594 to calibrate the total amount of Social Security benefits in the CPS to aggregate SSA data. These two adjustments raise the average Social Security benefit by 4.7 percent and result in a small change to the distribution of benefits in the CPS data.86 Even with these adjustments, differences in survey design and data collection processes between the CPS and SSA result in slight variations of Social Security benefits by age and income.87

Unlike Social Security benefits and other types of income, OASDI taxes are not directly obtained in the CPS. OASDI taxes for each worker were estimated from survey information on wages and self-employment earnings that was reconciled and calibrated to match SSA aggregate data on taxable earnings. Using Social Security rules that were in effect in 1997, for maximum taxable earnings and tax rates, a tax was estimated for each earner in the CPS. For workers who have only self-employment income, the OASDI tax equals their taxable self-employment income (subject to the 1997 wage cap of $65,400) multiplied by the Self-Employment Contributions Act (SECA) tax rate of 12.4 percent. For workers who have only wage and salary income, the OASDI tax equals their taxable wage and salary income (subject to the 1997 wage cap of $65,400) times the combined employer and employee Federal Insurance Contributions Act (FICA) tax rate of 12.4 percent.88 For workers with both wage and salary and self-employment income, the OASDI tax equals their taxable wage and salary income multiplied by the combined employer and employee FICA tax rate of 12.4 percent plus their taxable self-employment income (subject to a cap of $65,400 minus their taxable wage and salary income) times the SECA tax rate of 12.4 percent.

Total income for persons in the CPS was then recomputed using the reconciled and calibrated wage and salary income, self-employment income, and Social Security benefit estimates. Table A-3 shows the final results of the reconciliation and calibration process for individuals. Total family income was recomputed in the CPS by summing personal income across families and assigning this value to the family head. A final adjustment was made to the family weights to calibrate the value of Social Security benefits and OASDI taxes summed across families, with the value of benefits and taxes summed across individuals in the CPS.89

The tabulations in this Report utilize two different income concepts: total income and pre-OASDI income. "Total income" refers to the official Census money income concept that has been calibrated to aggregate SSA data as described above. It includes cash income from wages and salaries; net self-employment income; Social Security and railroad retirement benefits; other private and public pension income; supplemental security income (SSI); public assistance or welfare payments; unemployment insurance and workers' compensation benefits; income from dividends, interest, rents, and trusts; veterans' payments; alimony or child support; regular contributions from persons not living in the household; and other periodic income.

A more useful picture of the distribution and impact of Social Security taxes and benefits can be gained by classifying families and individuals by "pre-OASDI" income. Pre-OASDI income is equal to total income minus Social Security benefits plus the employer share of OASDI payroll taxes.90 Pre-OASDI income is the amount of income prior to receiving any Social Security benefits or paying any Social Security taxes. Grouping families or individuals by pre-OASDI income removes the effect that the Social Security program has on the distribution of total income and results in a more appropriate representation of the distribution of OASDI taxes, and the receipt of benefits by income.

For each income concept, 10 income deciles are calculated. Total income deciles for families are estimated by ranking family heads by income and determining the nine income percentiles that divide the number of families into 10 equal-numbered groups. Family pre-OASDI income deciles are determined the same way. For the person-based total income and pre-OASDI income deciles, an identical procedure is followed, except that persons rather than families are ranked by income.

For the worker total income percentile calculations and tabulations by decile, workers with zero or negative total incomes are excluded. There are 542,000 workers, or 0.4 percent of all workers, with zero or negative total income.91 For the worker pre-OASDI income percentile calculations and tabulations by decile, workers with zero or negative total incomes are excluded and a very small number of workers with negative pre-OASDI income are excluded. There are 36,000 workers, or less than 0.1 percent of all workers, with a positive total income but negative pre-OASDI income who are excluded.92

For the family pre-OASDI income percentile calculations and tabulations by decile, families with zero or negative total incomes are excluded and a very small number of families with negative pre-OASDI income are excluded. There are 2.6 million families (2.1 percent of the total) with zero or negative family total income and negative pre-OASDI income.93

For individual pre-OASDI income percentile calculations, those individuals whose total incomes are either zero or negative and their pre-OASDI incomes are negative are not included in the analysis, as are children under age 15.94 Out of a total of 213.5 million individuals over the age of 14, 18.6 million people, or 8.7 percent, are excluded. Over two-thirds of these individuals are non-working spouses or children. The median family income of all excluded individuals is $25,060. Including the individuals with zero or negative incomes in the analysis significantly reduces the income decile breakpoints and creates a bottom income decile with negative total income and zero OASDI taxes and benefits (see Table A-4). Aside from the bottom income group, the distribution of OASDI taxes and benefits by income is not qualitatively different from the distribution that excludes individuals with zero or negative income.95

Appendix B


1. This Report is a reissue of "Who Pays the Payroll Tax?" Heritage Foundation Center for Data Analysis Report No. 00-03, March 3, 2000. The introduction has been updated but the analysis is identical to that in the original Report. The author thanks Center for Data Analysis economists William W. Beach, Philippe Lacoude, and Ralph Rector for their contributions to the original Report.

2. President's Commission to Strengthen Social Security, Interim Report, August 2001, at www.commtostrengthensocsec.gov/reports/Report-Final.pdf (August 23, 2001).

3. Ibid., p. 20.

4. The average effective OASDI tax rate is total OASDI taxes divided by total income. This differs from the OASDI statutory tax rate of 12.4 percent on earned income below the taxable wage cap ($80,400 in 2001). Because the share of earned income to total income can differ from worker to worker and not all workers are subject to the OASDI tax, the average effective Social Security tax rate can vary significantly. Other sources of income, besides wages and salaries, include public assistance, Social Security, private pensions, alimony and child support, interest, rent, and dividends.

5. President's Commission to Strengthen Social Security, Interim Report, p. 10.

6. Ibid.

7. Surplus OASDI revenue is spent by other government programs or used to pay down publicly held federal debt.

8. The net benefit data presented in this Report cover just one year. The redistribution of income from Social Security over the lifetime of any cohort of Americans could be substantially different from the redistribution of income in any particular year.

9. The methodology used in this Report is similar to the methodology used by the Office of Research and Statistics at the Social Security Administration. See David Pattison, "The Distribution of OASDI Taxes and Benefits by Income Decile," Social Security Bulletin, Vol. 58, No. 2 (Summer 1995).

10. The Old-Age and Survivors Insurance program (OASI) and the Disability Insurance program (DI) are combined in this Report only because OASI income and DI income are combined in the Current Population Survey data and cannot be separated accurately.

11. The tabulations in this Report utilize two different income concepts--total income and pre-OASDI income. See Appendix A for an explanation of these income concepts. Taxes, tax rates, and benefits always refer to Social Security payroll taxes, tax rates, and benefits. The tax rates presented in this Report are average effective tax rates for pre-OASDI income unless otherwise noted.

12. Deciles are calculated by dividing the number of families into 10 equally sized groups classified by income.

13. Readers should use caution when interpreting the net benefit data presented in this Report because they cover just one year. The redistribution of income from Social Security over the lifetime of any cohort of Americans could be substantially different from the redistribution of income in any particular year. Other research suggests that there are lifetime transfers from high-income to low-income individuals. See Constantijn Pansi and Lee Lillard, "Socioeconomic Differentials in the Returns to Social Security," RAND Corporation Working Paper Series No. 96-05, February 1996.

14. Research suggests that there are also sizable lifetime transfers from men to women individuals. See note 13.

15. All race categories used in this Report are mutually exclusive. Specifically, white means whites only with no Hispanic persons; black means blacks only with no Hispanic persons; Hispanic includes Hispanics of all races; other races means all other races with no Hispanic persons.

16. Readers should use caution when interpreting the net benefit data presented in this Report because they cover just one year. The redistribution of income from Social Security over the lifetime of any cohort of Americans could be substantially different from the redistribution of income in any particular year.

17. Slight differences remain between the aggregate Census data and the Social Security Administration data, but these differences do not qualitatively effect the distributional analysis presented in this Report. See Table A-3 in Appendix A.

18. Measures of annual income for individuals are likely to vary considerably from their average lifetime income. For example, net self-employment income can be negative one year and positive the next, and high taxes may be paid by lifetime-poor individuals who happen to be at their peak earnings year. The lifetime distributional effects of taxes are likely to be muted compared with annual effects. See Don Fullerton and Diane Lim Rogers, Who Bears the Lifetime Tax Burden? (Washington, D.C.: The Brookings Institution, 1993), p. 17.

19. Social Security Administration, Annual Statistical Supplement to the Social Security Bulletin, December 1997.

20. 1998 Green Book, WMCP 105-7, Committee on Ways and Means, U.S. House of Representatives, 105th Cong., 2nd Sess., 1998, p. 6.

21. The OASDI payroll tax for self-employed workers is 12.4 percent of their earned income up to the taxable wage cap ($65,400 in 1997). Wage and salary workers pay 6.2 percent and their employers pay 6.2 percent, for a combined rate of 12.4 percent of their earned income up to the same taxable wage cap. Studies indicate that, on average, over 70 percent of the cost of all employer-paid payroll taxes is shifted to workers in the form of lower real wages. See Patricia M. Anderson and Bruce D. Meyer, "The Effects of Firm Specific Taxes and Government Mandates with an Application to the U.S. Unemployment Insurance Program," Journal of Public Economics, August 1997, and Jonathan Gruber and Alan B. Krueger, "The Incidence of Mandated Employer-Provided Insurance: Lessons from Workers Compensation Insurance," Tax Policy and Economy, 1991. In this Report, it is assumed that workers pay 100 percent of OASDI taxes.

22. The number of uncovered workers in any particular group also can affect the average effective OASDI tax rate of the group. All else being equal, if one group of workers includes a larger share of uncovered workers than another group, its average effective tax rate will be lower.

23. The maximum taxable amount of earned income, or the tax cap, for both self-employed and wage and salaried workers is indexed to change annually by the rate of growth in average wages. The tax cap has existed since the inception of the Social Security system in 1937, when Congress decided that, because there would be a limit on the amount of Social Security benefits Americans could receive, there should also be a limit on the amount of taxes workers pay into the program. See Gareth G. Davis and D. Mark Wilson, "The Impact of Removing Social Security's Tax Cap on Wages," Heritage Foundation Center for Data Analysis Report No. CDA99-01, January 19, 1999.

24. This does not include the $7.9 billion in taxes paid on Social Security benefits. Excluding the OASDI benefit tax results in a slight underestimation of the Social Security tax rate, particularly for workers and individuals with incomes above $25,000.

25. In the Current Population Survey data, any Social Security (survivor) benefits for a deceased worker's children under the age of 15 are assigned to the adult record in the household (see Appendix A). This has the effect of slightly increasing the amount of benefits received by middle-aged adults but does not qualitatively change the results.

26. The analysis of workers does not include the 542,000 workers, or 0.4 percent of all workers, who have zero or negative total income. Average taxes and benefits are therefore slightly higher than if the averages were estimated for all workers.

27. Deciles are calculated by dividing the number of workers into 10 roughly equal-sized groups classified by income. The number of workers in each income group are not exactly equal because there is a clustering of workers around specific dollar amounts of reported income such as $10,000, $15,000, etc. The upper limit of the fifth income group from the bottom represents the median total income for all workers or $21,600 (see Table B-1).

28. Workers in the lowest total income decile have a higher tax rate (10.2 percent) than workers in the second lowest decile (9.7 percent) because a larger share of their total income is taxable. For example, workers in the lowest total income decile are more likely to be teenagers living at home than are workers in the second lowest decile. Workers in the second lowest income decile are more likely to receive Social Security benefits and other forms of public assistance that moves them out of the lowest income group and reduces their share of total income subject to OASDI taxes.

29. In 1997, earnings greater than the maximum taxable amount (the tax cap) of $65,400 were not subject to the OASDI tax. See note 23.

30. The average reduction of 0.5 percentage points is the result of adding the employer's share of OASDI taxes to the income of wage and salary workers. See Appendix A.

31. The income breaks for the 10 total income and pre-OASDI income groups differ primarily because Social Security benefits are counted as part of total income but are not included in pre-OASDI income (see Appendix A). The total amount of income and benefits in Table B-2 differs slightly from the total in Table B-1 because the analysis of workers by pre-OASDI income excludes 36,000 workers with negative pre-OASDI income, some of whom also receive OASDI benefits. This has the effect of slightly increasing income and decreasing OASDI benefits.

32. Over 270,000 workers in the lowest income group are 15 to 21 years of age and also receive OASDI survivors benefits. Another 725,000 are over the age of 61 and receive OASDI retirement benefits. These two groups account for 6.8 percent of all workers in the lowest pre-OASDI income group.

33. Although this is generally true, a number of OASDI beneficiaries over the age of 61continue to work. The $55.3 billion in benefits is attributed to OASDI beneficiaries over the age of 61 and under the age of 18 who also work; persons who worked for part of the year and were retired or disabled for the rest of the year; and adults who have been assigned the Social Security benefits their children under the age of 15 receive. In the Current Population Survey data, any Social Security (survivor) benefits for a deceased worker's children under the age of 15 are assigned to the adult record in the household (see Appendix A).

34. Social Security tax revenue that is not used to pay benefits is converted into federal government bonds and spent as general revenue or used to reduce publicly held federal debt if there is a unified federal budget surplus.

35. The number of workers in each age group is not exactly equal because of the discrete nature of the age variable in the Current Population Survey. If workers' ages were reported in days instead of years, the number of workers in each age group would be nearly the same.

36. Average benefits are relatively high for the lowest age group primarily because of OASDI survivors benefits. OASDI benefits for workers between the ages of 21 and 62 are the result of workers becoming disabled during the year and the CPS methodology that assigns the OASDI benefits of children under the age of 15 to the record of the adult head of the family.

37. Readers should use caution when interpreting the net benefit data presented in this Report because they cover just one year. The lifetime net benefit of Social Security for any given person or family, or groups of people or families, could be substantially different from the net benefit for any particular year. Research suggests that there are sizable lifetime transfers from men to women. See Pansi and Lillard, "Socioeconomic Differentials in the Returns to Social Security."

38. Research suggests that there is a small but significant lifetime transfer from blacks to whites. See note 37.

39. Research suggests that there are sizable lifetime transfers from single individuals to married couples. See note 37.

40. There are also a small number of unpaid family members. The Census Bureau assigns each worker to a group based on the longest job they held in 1997.

41. This ignores the very small number of unpaid family workers.

42. Workers 15 to 21 years of age and those 58 or older are more likely to receive Social Security benefits and account for 58.8 percent of the group of workers who work part-time one to 26 weeks per year, and 42.3 percent of the group of workers who work part-time 27 to 47 weeks per year. The low incomes and low taxes paid by these workers also contribute to the positive net benefit result. A portion of those workers 58 or older may have worked for only part of the year before retiring completely.

43. The total amount of taxes in Table B-5 differs slightly from the total in Table B-2 because of small differences between an individual's population weight and family weight that were introduced in the calibration process. See Appendix A.

44. The analysis of families does not include those with zero or negative total income. There are 2.5 million families, or 2.1 percent of all families, with zero or negative total income. Average taxes and benefits are therefore slightly higher than if the averages were estimated for all families.

45. Like workers, a more accurate picture of the burden of Social Security taxes and the distribution of benefits of the OASDI program is gained by classifying families according to their pre-OASDI income. Grouping families by pre-OASDI income removes the effect the Social Security program has on the distribution of total income and results in a more appropriate representation of the distribution of OASDI taxes and the receipt of benefits by income.

46. There is approximately an equal number of families in each income group or decile. The numbers of families in each income group are not exactly equal because there is a clustering of families around specific dollar amounts of income, such as $20,000, $25,000, etc. The upper limit of the fifth group from the bottom represents the median income for all families.

47. Taxes as a percent of family income are lower in the bottom and top deciles for two different reasons. A significant portion of earned income in the top decile is above the tax cap of $65,400 (in 1997) and is not subject to the OASDI tax. Substantial portions of the income in the bottom decile come from non-labor sources such as public assistance and Social Security and therefore are not subject to payroll taxes.

48. As income rises, the share of non-labor income such as public assistance and Social Security declines.

49. Families in the highest income group also pay a smaller share of OASDI taxes because of the cap on taxable earnings.

50. The reader should use caution when interpreting the net benefit data presented in this Report. Research suggests that there are lifetime transfers from high-income to low-income individuals, and this is likely to be the case for families as well.

51. Note that $60.8 billion of the $261.9 billion was used to fund other federal government programs or to reduce publicly held federal debt. Social Security tax revenue that is not used to pay benefits is converted into federal government bonds and spent as general revenue.

52. The apparent progressiveness of Social Security is also due to the fact that it has become the only source of income for many elderly families.

53. The poverty level is based on total family income and family size, as defined by the Bureau of the Census for 1997, rather than pre-OASDI income. The difference between the Census measure of a family's poverty level and the poverty level determined by pre-OASDI income does not qualitatively effect the results.

54. Identifying spouses and teenagers separately from the heads of working families pushes some individuals down the income distribution relative to the family head. Also, a substantial fraction of the elderly who live with their adult children will have much lower incomes than they would if they are counted with the rest of their family. For example, if a widow aged 70 with an income of $8,000 lives with her married daughter who earns $4,000 and her husband who earns $50,000, the three would be classified as one family with an income of $62,000; but under the person definition, they are three separate individuals with incomes of $8,000, $4,000, and $50,000.

55. The averages are taken for all individuals in each decile, including those who do not pay taxes or receive benefits. Average taxes and benefits are therefore lower than if the averages were taken only among individuals with non-zero taxes or benefits.

56. Individuals in the lowest and highest income groups pay smaller shares of OASDI taxes than their income share for two different reasons. A significant portion of earned income in the top decile is above the tax cap of $65,400 and not subject to the OASDI tax (see note 23). Substantial portions of the income in the bottom decile comes from non-labor sources such as public assistance and Social Security and therefore are not subject to payroll taxes.

57. Social Security Administration, Fast Facts & Figures About Social Security 1998, July 1998, p. 7.

58. See note 5.

59. Readers should use caution when interpreting the net benefit data presented in this Report because they cover just one year. Research suggests that there are lifetime transfers from high-income to low-income individuals. See Pansi and Lillard, "Socioeconomic Differentials in the Returns to Social Security."

60. $62.4 billion of the $289.4 billion was used to fund other federal government programs or to reduce publicly held federal debt.

61. The lifetime distributional effects of the Social Security program are likely to be muted compared with the annual effects presented here. See note 18.

62. The age decile breaks for all individuals and workers differ as a result of the inclusion of non-working adults. The number of workers in each age group is not exactly equal because of the discrete nature of the age variable in the Current Population Survey. If workers' ages were reported in days instead of years, the number of workers in each age group would be nearly the same.

63. Average benefits are relatively high for the lowest age group primarily because of OASDI survivor benefits.

64. In 1997, 71.3 percent of men were employed compared to 56.8 percent of women, and women made up 57.1 percent of the population over the age of 61compared to just 50.7 percent of the population under the age of 62. Other research suggests that there are also sizable lifetime transfers from men to women individuals. See Pansi and Lillard, "Socioeconomic Differentials in the Returns to Social Security."

65. William W. Beach and Gareth E. Davis, "Social Security's Rate of Return," Heritage Foundation Center for Data Analysis Report No. CDA98-01, January 15, 1998.

66. Readers should use caution when interpreting the net benefit data presented in this Report because they cover just one year. The redistribution of income from Social Security over the lifetime of any cohort of Americans could be substantially different from the redistribution of income in any particular year. Research suggests that there are sizable lifetime transfers from black to white individuals. See Pansi and Lillard, "Socioeconomic Differentials in the Returns to Social Security."

67. Research suggests that there are sizable lifetime transfers from single individuals to married couples--the opposite result of this one-year estimate. See note 66.

68. This one year net benefit estimate results primarily from the fact that just 11.2 percent of non-citizens are of retirement age, compared with 18.5 percent of U.S. citizens. See note 66.

69. For many Americans over the age of 65, a high school degree was the end of their formal education.

70. The annual March CPS is the source for official statistics on poverty and family and individual income for 1997.

71. In fact, no SSA microdata on beneficiaries and taxpayers are currently available to non-government analysts. Access to public data, in the form of SSA's Continuous Work History Sample, was restricted to federal government researchers in 1974. SSA and the Treasury Department have severely limited access to these data over the past 25 years.

72. SSA's Continuous Work History Sample.

73. The institutionalized population includes those Americans living in penal and mental facilities, as well as homes for the aged. One study estimates that the March 1992 CPS data underestimate elderly taxpaying units by as much as 9 percent because of death and institutionalization. See David Pattison and David E Harrington, "Proposals to Modify the Taxation of Social Security Benefits: Options and Distributional Effects," Social Security Bulletin, Vol. 56, No. 2 (Summer 1993).

74. This adjustment required obtaining a wide variety of data from a number of sources including the Census Bureau, SSA, and the Department of Justice, and the Statistical Abstract of the United States. Some of the data were not available for 1997 and had to be estimated from prior year trends. Data on the number of deceased were estimated using age-adjusted fatality rates.

75. Scaling up the March 1998 CPS weights also has the effect of increasing the dollar amount of earned income and Social Security benefits reported from the CPS by 1.99 percent.

76. The uniform scaling of weights also results in a CPS population total of 213.456 million, compared with 210.850 million from the SSA data--a difference of 1.2 percent. This population overestimate does not affect the results presented in this Report because it stems from CPS records that are not included in the analysis. These CPS records represent teenagers and spouses that are not working or receiving Social Security benefits and have no individual income.

77. 1998 Green Book, p. 6.

78. Ibid. There are also a small number of self-employed and private household workers with very low earnings who are not covered. See Social Security Administration, Annual Statistical Supplement to the Social Security Bulletin, December 1997, p. 29.

79. Any worker who also reported receiving Social Security benefits in 1997 was assumed to be working in covered employment. In 1997, 4.3 percent of all workers also received some Social Security benefit.

80. This approach assumes that the earnings and demographic profiles of covered and uncovered federal, state, and local workers are identical.

81. Federal employees hired after January 1, 1984, are covered by Social Security. A worker who was 29 years old in 1997 would have been 16 years old in 1984. Age 29 was chosen because there probably were no Americans younger than 16 working for the federal government in 1984.

82. CPS records representing 2.2 million federal workers were duplicated, forming two datasets. All individual weights in the first dataset were multiplied by .4607, and all individual weights in the second dataset were multiplied by .5393. The resulting two datasets represent the 1.2 million uncovered federal workers and 1 million covered federal workers. When re-combined, the two datasets continue to represent the original count of 2.2 million federal workers. All family weights in both datasets were divided by 2.

83. CPS records representing 15.6 million state and local workers were duplicated into two datasets. All individual weights in the first dataset were multiplied by a unique factor for each state, and all individual weights in the second dataset were multiplied by 1.0 minus the state specific factor. The state factor adjustments were derived from 1998 Green Book, p. 10. The resulting two datasets represent the 4.3 million uncovered state and local workers and 11.3 million covered state and local workers. When re-combined, the two datasets continue to represent the original count of 15.6 million state and local workers. All family weights in both datasets were divided by 2.

84. For example, if a covered worker had wage and salary income of $40,000 and self-employment income of $40,000, the OASDI taxable income would only be $65,400. The full $40,000 of wage and salary income would be taxable, but only a portion ($25,400) of their self-employment income would be taxable.

85. During the course of this analysis, it was discovered from CPS records that respondents representing 156,000 people reported receiving $50,000 in Social Security benefits in 1997 ($50,000 is the Census-imposed maximum), and survey respondents representing 15,000 people reported living in families that received over $98,000 in Social Security benefits. Conversations with SSA analysts revealed that the CPS data are not consistent with the benefit limitations of the Social Security program. The apparent reporting or coding error in the Census data is significant. The 156,000 people with $50,000 in Social Security benefits account for $9.7 billion, or 3 percent, of all reported Social Security benefits in the March 1998 CPS.

86. These two adjustments decrease the share of Social Security benefits going to the highest and lowest income percentiles in the CPS. There are also small declines in the share of Social Security benefits going to the second, third, and fourth percentiles, and corresponding increases in the share of benefits going to the fifth, sixth, seventh, and eighth income percentiles.

87. While SSA data include information on the Social Security benefits going directly to children under the age of 15 and the elderly who die during the year being studied, the CPS has no information on the deceased elderly (who often have low incomes) and the value of Social Security benefits for children under the age of 15 that are assigned to the adult head of family.

88. Studies indicate that, on average, over 70 percent of the cost of all employer-paid payroll taxes is shifted to workers in the form of lower real wages. See Patricia M. Anderson and Bruce D. Meyer, "The Effects of Firm Specific Taxes and Government Mandates with an Application to the U.S. Unemployment Insurance Program," Journal of Public Economics, August 1997, and Jonathan Gruber and Alan B. Krueger, "The Incidence of Mandated Employer-Provided Insurance: Lessons from Workers Compensation Insurance," Tax Policy and Economy, 1991. In this Report, it is assumed that workers pay 100 percent of OASDI taxes.

89. Specifically, the family weights for both single and family heads were multiplied by 1.0318. This adjustment increased the number of families in the CPS from 116.5 million to 119.7 million.

90. For example, for the Social Security beneficiaries who do not work and those who are only self-employed, pre-OASDI income is total income minus OASDI benefits. For wage and salary workers with no benefits, pre-OASDI income is total income plus the employer share of OASDI taxes.

91. Workers' total income can be negative if net self-employment income is negative (gross earnings minus expenses). Including these workers lowers the average income from $29,572 to $29,451, or 0.4 percent, but has little impact on the income decile breakpoints or the distribution of income, taxes, and benefits.

92. Workers' total income can be positive while their pre-OASDI income is negative in a given year if net self-employment income is low enough to reduce total income to a level that is below any Social Security benefits they may have also received but not negative enough to cause total income to be negative. Including these workers has very little impact on the income decile breakpoints and no impact on the distribution of income, taxes, and benefits.

93. Family income for any one year can be negative if net self-employment income is negative (gross earnings minus expenses). Total family income can be positive while their pre-OASDI income is negative if net self-employment income is low enough to reduce total income to a level that is below any Social Security benefits the family may also have received but not low enough to cause total income to be negative. Including these families lowers the average income from $45,523 to $41,628, or 8.6 percent, and the income decile breakpoints by an average of $677, but has little impact on the distribution of income, taxes, or benefits.

94. The CPS contains no income data on individuals under age 15. Any Social Security benefits received by children under age 15 are assigned to the adult in the family.

95. The CPS weight adjustment was chosen to optimize the reconciliation of total employment and OASDI beneficiary counts with Social Security data. This adjustment results in an over-weighting of the non-working and non-OADSI beneficiary records in the CPS and results in a total population count that is 2.6 million higher than the adjusted Social Security data (see Table A-3). Because of this and the fact that including or excluding individuals with zero or negative incomes does not qualitatively change the results (see Table A-4 and Table A-5), these individuals have been excluded from this analysis.

Authors

D. Mark Wilson

Director