Social Security and personal saving: 1971 and beyond

Feldstein (1996, 1974) reported that Social Security in the U.S.A. reduced personal saving (“saving”) in 1992 (1971) by $416 ($61) billion. I reestimate his life-cycle consumption specification using data from the latest NIPA revision, correct his calculations, and find that the implied reduction in 1992 (1971) saving is now $280 ($22) billion, 48% (16%) of actual net private saving, with a standard error of $114 ($14) billion. If structural breaks around WWII and the 1972 Social Security amendments (which raised real per capita SSW by 22%) are allowed, and the market value of Treasury debt included in the specification, the reduction in 1971 and 1992 saving attributable to Social Security is at most 0.55 times its standard error, and 12% of net private saving. I then reestimate the preferred specification of Coates and Humphreys (1999), allowing for these structural breaks and relaxing other restrictions. The implied effect of Social Security on saving is again statistically zero. First version received: September 2000/Final version received: September 2001 RID="*" ID="*"  I thank Les Oxley for pointing out that correcting for AR(1) residuals is not a categorical imperative but a cultural relative, in which case common factor restrictions are crucial.