The Impact of Social Security on Private Savings

Martin Feldstein, who is on the short list to win a Nobel Prize someday, published a truly seminal paper on social security in the early 1970s. In it he argued that forced savings programs like social security might have a negative impact on private saving and hence capital accumulation. The idea is that if the government is going to force you to save for your own retirement, you will have less of an incentive to save on your own.

You might respond by saying, "So what? Why does it matter how you finance your retirement, through your own efforts or the government's forced savings program?" Feldstein's answer was that since the social security program for most of its history is a pay-as-you-go, or currently funded, program, the government is not actually saving through social security. It simply taxes young workers and immediately transfers most of those tax revenues to current retirees. It is not actually investing any of the money being transferred. If you as an individual respond to social security by saving less, and everyone else saves less, and the government is taking the social security taxes and simply transferring them to retirees and not saving them, total capital accumulation falls and this can hurt domestic productivity, and growth in productivity in the future.

Feldstein provided strong empirical support for this hypothesis. He regressed aggregate saving on a variety of variables including social security wealth. The regression coefficient on the social security variable was negative and large in magnitude meaning that social security reduced private savings during the time frame of the data available at the time of his study, 1929 - 1971. In fact, he forecast a huge drop in saving as a result of the social security program, between 30 and 50%! (The coefficient of the social security wealth variable ranged from 0.03 to 0.07 in a consumption function.)

Needless to say (so why say it?), this was very controversial at the time it was published and created a firestorm of controversy. Robert Barro argued that the issue was very similar to the issue of whether government bonds were net wealth for a society or not. If you raise taxes now, the government needs to borrow less and can reduce taxes in the future. This shifts taxes from the future to the present, or the other way around if there is a tax cut today, much the same way social security does. (Think of the benefit received in the future under social security as a negative tax.)

Others, including Barro, tried testing the proposition using other data sources, other time periods, and various statistical techniques. Two analysts at the social security administration, Leimer and Lesnoy, for example, tried to replicate Feldstein's work so they could extend it. One problem they ran into immediately was that social security wealth is a constructed variable. Net social security wealth is the individual's lifetime social security benefits minus the individual's lifetime taxes. However, calculating the average person's social security wealth is a daunting task since it depends on their labor earnings and their future benefit depends in a complicated way on their earnings history. Leimer and Lesnoy made a good faith effort to replicate Feldstein's variable but couldn't. The plot thickens.

Feldstein then discovered a programming error in his social security wealth variable. (How embarrassing! The plot thickens even more.) He corrected it, re-estimated, and got the same result as in his earlier paper. Leimer and Lesnoy, however, couldn't replicate his result, no matter what they tried. Feldstein (National Tax Journal, 1996) updated the calculation of the social security wealth data and re-estimated his earlier equation and some refinements of it. Once again he found support for his own hypothesis. (This is a bit odd since social security changed into a partially funded system in the early 1980s. It should have had less of an impact on savings since the government was creating a large trust fund which was earning interest.

Now you can investigate this issue yourself using Feldstein's data on social security. Follow the instructions to do the project.