Home About RSS

Stanford Securities Litigation Analytics
Stanford Securities Litigation Analytics

In defense of economics

Three years ago, Harvard economists Ken Rogoff and Carmen Reinhart published a paper that became  one of the most talked about studies in the media, politics and field of economics.  However, just this past week a University of Maryland economics grad student discovered errors in the mathematical formulas used, as well as underlying problems within the data itself.  This prompted a recent podcast by NPR’s Planet Money titled “How Much Should We Trust Economists?” published on April 19, 2013.

Justin Wolfers, an economist at the University of Michigan, states that in his analysis of economics papers regarding the death penalty, he found errors of various kinds in 90% of published studies anywhere between one month and 10 years after their publication.  In some cases, errors were in methodologies, and others the errors occurred within the underlying math.

It’s understandable why such common errors could create a of mistrust of economics.  Macro economists, as Planet Money points out, suffer a data problem.  We’re unable to run controlled experiments where specific variables are easily isolated and explained.  The study of securities litigation is especially susceptible to this problem.  The data itself is extremely difficult to reliably find, and even more difficult to accurately collect.  These issues are compounded by the relatively small universe of cases relative to other fields of study – small errors can have a big impact on the accuracy of any analysis.  We’re acutely aware of this problem at SLA, and take painstaking steps to minimize the risks and impacts of errors when counseling practitioners in our industry; we’re confident that our dataset is the most carefully collected one of its kind.

However, quality data collection is only half the battle of for quality analytics.  The other major hurdle that economists face is the challenge of teasing out, and measuring the influence of single variables against the entire dataset.  For instance, we find that class actions with a parallel SEC action are highly correlated to cases in which a restatement occurs.  Likewise, our data shows that cases with the largest damages will settle for the most amount of money.  All of this is obvious on the surface, but presents a challenge in understanding the true dynamic among various factors in a class action and its ultimate outcome.  We are continuing to refine our models every day to provide the most useful analytics to practitioners.  If you are interested in helping in our development, please send us a note.

As Mr. Wolfers points out, “The best defense of economics is not that we’re good, it’s that everything else is terrible.”

Leave a Reply