Monday, September 15, 2008

New paper: Is Social Security Part of the Social Safety Net?

Jeff Brown and Don Fullerton of the University of Illinois and Julia Coronado of Barclays Global Investors are authors of a new paper, "Is Social Security Part of the Social Safety Net?" released as part of the NBER's September 25th
seminar on Tax Policy and the Economy. Here's the abstract:

Early architects of the Social Security program clearly intended to "prevent destitution and dependency" of the elderly. Indeed, Social Security is the only income for many elderly individuals with no pension and no wealth. On the basis of annual income, Coronado et al (2000) show that Social Security dramatically reduces inequality. On the basis of lifetime income, however, they show that effects are ambiguous. Social Security does not clearly increase lifetime income equality, as measured by the Gini Coefficient. That paper focuses on overall inequality, using 21 years of PSID data for 1800 individuals.

Relative to that paper, we make several contributions. First, we extend the data to 26 years for 6,000 individuals. Second, we update various procedures and definitions. Third, we focus not on overall inequality, but on how Social Security affects the poor. This question is important, because a Social Security program that reduces inequality by redistributing from high income individuals to middle income individuals may do nothing to help the poor. Fourth, we look at changes over time in the extent to which Social Security helps the poor.

For this analysis, we use the PSID to develop a large sample of individuals born at different times. Second, we construct an entire lifetime earnings history for each individual, and we develop several measures by which to classify each of them from rich to poor. Third, we calculate each individual's social security taxes and retirement benefits. Fourth, we use several alternative measures of the impact of this program on "the poor".

We have four major findings. First, as we expand the definition of income to use a more comprehensive measure, we find that Social Security does less overall to help the poor. Indeed, when we use an "endowment" defined by potential labor earnings at the household level, rather than actual earnings at the individual level, we find that Social Security has virtually no effect on overall inequality. Second, we find that this result is driven largely by the lack of redistribution across the middle and upper part of the income distribution, so it masks some small positive net transfers to the bottom lifetime income quintile. Third, in cases where redistribution does occur, we find it is not efficiently targeted: many high income households receive positive net transfers, while many low income households pay net taxes. Finally, the effect of social security changes over time, and these changes depend on the income concept used to classify someone as "poor". Thus, Social Security has little effect, and changes over time are ambiguous.

I've been doing some work on the "targeting" of Social Security progressivity, which touches on some similar issues to the Brown, Fullerton, Coronado paper. I hope to preview it here in the new few weeks.

Also, Don Fullerton is the co-author an AEI book, Income Redistribution from Social Security, which is available (for free!) here.


Govt analyst said...

Their abstract is misleading: They do not analyze the "Social Security program," but only OASI. They do not include disability insurance, but they do note in footnote 2 that "Because DI is highly redistributive, including DI in the analysis would unambiguously increase the redistribution achieved by the overall system."

Andrew G. Biggs said...

That's a valid point, though I think one they could remedy easily with some altered wording. OASI and DI play different roles and there's merit in analyzing them separately.