Friday, April 26, 2013

A Glimpse into Mainstream Economic Number Massaging

I have already discussed the methodological problems with the Reinhart-Rogoff study and the current battle among empiricists over their work.

This morning, I just want to point out part of some of the defense that Reinhart-Rogoff have published in today's NYT. It is quite an insightful window into how much number massaging goes into these type studies. In other words, even if one were willing to buy into the empirical testing method for economics, something Austrian economists don't, it might be difficult to accept the data used in these studies.

Check out the amount of number massaging that Reinhart-Rogoff admit to in their study and justify as seemingly standard practice (my highlights):
As for the accusations of selective omission of data, there is little appreciation that this is archival research, involving constant judgments at every step. The New Zealand data we used was part of the problem that Herndon and his colleagues allude to biasing the results in favor of lower growth at higher levels of debt. We have since incorporated the correct data in our Journal of Economic Perspectives paper.

Oddly, Herndon and his colleagues do not mention another data omission. This one was intentional on our part. Back in 2010, we were still sorting inconsistencies in Spanish G.D.P. data from the 1960s from three different sources. Our primary source for real G.D.P. growth was the work of the economic historian Angus Madison. But we also checked his data and, where inconsistencies appeared, refrained from using it. Other sources, including the I.M.F. and Spain’s monumental and scholarly historical statistics, had very different numbers. In our 2010 paper, we omitted Spain for the 1960s entirely. Had we included these observations, it would have strengthened our results, since Spain had very low public debt in the 1960s (under 30 percent of G.D.P.), and yet enjoyed very fast average G.D.P. growth (over 6 percent) over that period. We later reconciled this problem for our 2012 paper. This is just an example of what our archival research involves; it is not simply a matter of filling in cells on an Excel spreadsheet from sanitized, easy-to-use databases.

1 comment:

  1. In contrast, here is a page using the whole of international public data sets in an easy to replicate fashion that shows that *most* of the time, the faster government spending grows, the slower the private economy grows, and vice versa:

    although it doesn't isolate cause and effect (critiquing some aspects studies that claim to in the realm of "stimulus" to point out its difficulty with existing sparse and incomplete data).