Site Meter

Friday, April 19, 2013

Ragging on Rogoff

I can't resist.  This post has nothing to do with the recent Excel-ent work by Herndon e al at UMass Amherst.   I just want to pick on something Kenneth Rogoff wrote for Project Syndicate in 2012

He is discussing debt and growth with reference to his work with Carment Reinhart using a large data set (not the earlier work with 20 15 countries post 1946 and New Zealand post 1950.  Rogoff  wrote [I add the necessary steps which he skipped in square brackets]

Of course, there is two-way feedback between debt and growth, but normal recessions last only a year [and the only variation in growth is that due to normal recessions and that caused by high debt] and cannot explain a two-decade period of malaise [which must be due to debt as no one and especially not Robert Barro has ever found anything else which is correlatedwith growth over intervals of more than two decades].


I think my expostulation of his reasoning is entirely fair and that he absolutely is arguing that there is, for example, no pattern of convergence withing the OECD post world war II, no significant difference between growth pre and post World War II and, in sum his argument is "correlation is, of course, not causation, however oh look a shiny object."  

The next sentence

"The drag on growth is more likely to come from the eventual need for the government to raise taxes, as well as..." 

Rogoff asserts either that high taxes cause low growth or that high expected future taxes cause low growth.  He should know that this view is unsopported by available evidence.  It is an article of faith for Republicans (and here I'm sure he is sincere).  But he is discussing simple correlations in historical data and he must understand that there is a problem in arguing that the correlation between high debt and low growth is explaind by high debt causing  a high ratio of tax revenues to GDP when the ratio of tax revenues to GDP is not significantly partially correlated with low growth in standard regressions (meaning regressions which include initial GDP per capita).

Rogoff relies on simple historical correlations and ignores them in the same paragraph.

No comments: