With annual expenditures of over $600 Billion, Social Security is the largest government program in the U.S. Social Security is also the single largest source of income for the elderly, accounting for 40 percent of all income going to individuals age 65 and above and over 80 percent of income for the poorest quintile of families.
While Social Security is clearly an important source of income for poor families, it is less clear whether the program redistributes income from more well-off to poorer families. Although Social Security has a progressive benefit formula, various features of the program reduce the extent of redistribution in the system.
This question is relevant for the debate over Social Security reform, since most reform proposals intend to maintain Social Security as a progressive program and yet have the potential to alter the extent of redistribution. Knowing how much redistribution exists in the current system, and which aspects of the program's design influence this, is useful in order for policy makers to assess the distributional effects of any program change.
In Is Social Security Part of the Social Safety Net? (NBER Working Paper 15070), researchers Jeffrey Brown, Julia Coronado, and Don Fullerton examine the extent of redistribution in the Social Security system. To do so, the authors build a model that categorizes individuals by their lifetime resources and calculates the taxes they pay and benefits they receive from Social Security. Importantly, the authors use a number of different definitions of income and of redistribution, in order to obtain a more complete understanding of the issue.
The authors use 26 years of data from the Panel Study of Income Dynamics (PSID) to construct complete lifetime earning histories for individuals in their sample. The use of actual (rather than simulated) data allows the authors to incorporate real events and phenomena, such as spells of unemployment or a correlation between earnings and marital status, which may be important for the analysis.
The authors begin by calculating the lifetime net Social Security tax rate for each individual in their sample. This is the present value of Social Security tax payments minus the present value of Social Security benefits divided by the present value of the individual's lifetime income. They then use this tax rate to calculate three different measures of redistribution. The first is a measure of how the Gini coefficient (a well-known gauge of income inequality) changes when Social Security is included. This is useful for understanding the overall impact of Social Security on inequality. The other measures are the average net tax rate in each quintile of the income distribution and the fraction of individuals in each quintile that receive positive net transfers from Social Security. The latter two measures are useful for assessing whether the program does indeed benefit those at the bottom of the income distribution.
In addition to the multiple definitions of redistribution, the authors also employ multiple definitions of income. The first is an individual's actual lifetime earnings. The second is an individual's potential lifetime earnings (their earnings if they had worked full-time throughout their adult lives, minus any periods of unemployment). Potential earnings measures the individual's ability to earn, regardless of what he or she actually chooses to do. The third pools the earnings of married couples, since their economic well-being depends on total household resources rather than individual earnings.
The authors have several major findings. First, when a more comprehensive income definition is used (potential or household income), Social Security has virtually no impact on the overall distribution of lifetime economic resources, as measured by the Gini coefficient. Second, however, as the authors note, "while Social Security is not particularly good at flattening the overall income distribution, it nonetheless is at least mildly successful at transferring resources, on average, to the lifetime poor." Over 85 percent of individuals in the lowest quintile receive positive net transfers from Social Security when the narrowest definition of income is used (individual actual earnings). As the income definition is broadened, this share falls, but some individuals still receive positive transfers.
Third, transfers through the Social Security system are imperfectly targeted; some high-income individuals receive positive net transfers, particularly when income is defined at the household level, and some low-income individuals do not. One reason, for example, is that spousal benefits provide more generous payouts to the spouses of high earners. Finally, the authors examine whether the extent of redistribution has changed over time and find that the program has become somewhat more progressive on balance, but the direction and extent of the changes depend on the income definition used.
As the authors note, their research suggests several avenues for future work. Their analysis explicitly ignores behavioral responses to Social Security, such as changes in labor supply or savings, which might have an impact on inequality. Second, their analysis does not incorporate the value of Social Security as an insurance program, for example in providing protection against the risk of disability or unexpectedly long life. Despite these limitations, the framework the authors develop could be used to explore the distributional consequences of Social Security reform.
The authors acknowledge funding from the Social Security Administration through grant #10-P-98363-1-03 to the National Bureau of Economic Research as part of the SSA Retirement Research Consortium.