April 18, 2013
Former students and admirers of Harvard professor Martin Feldstein, who was also the chief economist in the Reagan administration, were undoubtedly outraged to see him excluded from the NYT’s Economix blog list of top blunderers in economics. Professor Feldstein’s claim to fame in this category stems in large part from a 1974 article which purports to show that Social Security led to a reduction in private savings.
This article received considerable attention and played an important role in advancing Feldstein’s reputation as one of the top economists of his generation. However, it turned out that the result was attributable to a programming error. This error was eventually uncovered by Dean Leimer and Selig Lesnoy, two researchers at the Social Security Administration. When the error was corrected, the relationship between Social Security wealth and private saving turned out to be statistically insignificant.
Feldstein actually revisited the topic again in the 1990s and claimed that with two decades of additional data he was able to establish that he had been right all along. This one turned out not to be quite right either. My colleague David Rosnick and I tried to replicate his results without success. After repeated efforts to contact Professor Feldstein to better ascertain his methodology, he eventually gave us enough information to determine that we were running our regressions correctly.
Feldstein also added that it did not surprise him that we could not replicate his results, since saving data are subject to large revisions. This is true, but then it leads one to wonder why anyone would make major policy pronouncements based on results using the pre-revision data.
I am also reminded of this $12 trillion mistake back in 2003 by Michael Boskin, who had been the chief economist in the administration of the first President Bush. He was briefly convinced that money being withdrawn from tax sheltered accounts like 401(k)s would lead to huge budget surpluses. (Thanks to Charles McMillion for this one.)