In his 1982 book, Economic Effects of Social Security, Aaron called the results of the controversy surrounding Feldstein's work a "statistical cacophony." Refering to work by several economists in addition to Lesnoy and Leimer, Aaron wrote that, "almost all participants have concluded that essentially nothing can be learned about the effects of social security on saving from time series analysis."
Feldstein's critics argue that time series studies cannot prove a relationship between savings data and projections of social security wealth.
Aaron concludes, "It is clear that Feldstein's theoretical analysis has been vigorously challenged by scholars engaged in research on social security, and his empirical findings have been discredited."
Several Twists
The debate over Feldstein's work has taken several twists over the last decade. The most notable one was in 1980 when Leimer and Lesnoy discovered an error in Feldstein's computer program that had overestimated by 37 percent the value of future social security benefits to which people are entitled.
Feldstein responded by reworking his data and including new changes in social security benefit payments which were enacted in 1972--a 20 percent increase in benefits and automatic indexing of benefits to inflation. Based on these changes, which according to Lesnoy and Leimer did not consider all of the effects to future benefits in the 1972 legislation, Feldstein's corrected data supported his original view that social security benefits reduce private savings.
Feldstein's 1980 defense of his original conclusions prompted new criticisms by Lesnoy, Leimer, and others. Articles by economists from such diverse groups as the liberal Brookings Institution and the conservative American Enterprise Institute questioned several of the assumptions in Feldstein's model. While some criticisms were based on theoretical disagreements and challenges to the life-cycle model, others were based simply on Feldstein's choice of data.
Lesnoy and Leimer released a series of papers criticizing Feldstein's conclusions after they discovered the programming error in his original work.
Feldstein used data from 1930 to 1971 in his corrected estimates of the reductions in savings due to social security. Lesnoy and Leimer claim, however, that if only data from 1947 to 1971 are used, Feldstein would have concluded that social security payments actually cause a huge increase in private saving and that savings would actually have decreased were it not for social security.
Lesnoy and Leimer also argue that the data from before World War II are of questionable relevance to Feldstein's model to begin with. The first social security taxes were not collected until 1937, and the first benefits were not paid out until 1940.
They note that most saving during the decade before the war was not done with the expectation of receiving benefits upon retirement. These expectations of retirement benefits are a fundamental assumption of Feldstein's "life cycle" model. Moreover, they say, the Great Depression and the war represent abnormal periods for private saving in this country.
Lesnoy and Leimer also argue that minor revisions in Feldstein's equations--made to take account of differences in the way individuals plan their retirement savings--lead to radically different conclusions about the effect of social security on private saving. These modified assumptions lead to contradictory conclusions, they say.
Feldstein, however, stands by his original conclusion that social security decreases private saving and insists that his original statistical estimates of a 50 percent decrease in savings are on target.
He argues that the statistical reliability of his model can be measured in different ways and argues that Lesnoy and Leimer have not shown his results to be questionable.
Regarding Lesnoy and Leimer's assertion that his data sample is inadequate to make strong statistical inferences, Feldstein says, "All data sets have problems," adding, "Leaving out the early years and the pre-social security world you throw away sample points and throw away variability and just make the standard error larger."
Beyond the squabbles over the use and alleged misuse of data, there is one thing on which all economists agree: there is little in economics that cannot be viewed and used in a number of different ways. Says MIT's Edwin Kuh, "While the ostensible rules of the game are different, an economist is much like a lawyer filing a brief."