An Examination of Empirical Tests of Social Security and Savings

The effect of social security and other forms of government debt on national savings is one of the most widely debated policy questions in economics today. Some estimates suggest that social security has reduced U.S. savings by almost forty percent. This paper examines recent cross-section and time series empirical tests of the social security-savings question and argues that, given current data, neither type of test has much potential for settling the controversy. In particular, there are a number of specification problems relating to social security time series regressions that can easily lead to highly unstable coefficients and to rejection of the hypothesis that social security reduces savings, even if it is actually true. These points are demonstrated by running regressions on hypothetical data generated by a perfect foresight life-cycle growth model developed previously by the authors. While the data is obtained from a model in which social security reduces the nation's capital stock by almost twenty percent, time series social security regression coefficients vary enormously depending on the specified level of the program, the preferences of hypothetical households, the level of concommitant government policies, and the time interval of the data.