The excel error in Rogoff-Reinhart

By Matt Nolan 18/04/2013


I see the Rogoff-Reinhart figures regarding the correlation between GDP growth and the size of the debt stock are currently under attack – due largely to an unfortunate excel error discovered reported by Mike Konczal.  Here are the list of posts about it at the moment:

All very nice.  Depending on the message people were trying to sell they either said the result was meaningless, or central, so I don’t think this makes any actual difference.  Honestly, without clear causal drivers there just were not good evidence based claims for actual policy adjustments – a lot of people were actually just saying we need to do X based on their preconceptions.  And they found this correlation either something that supports that (somehow) or something they need to rule out.

Over the last few years I have seen authors, at different times, use R-R as central to their argument on one thing, and then dismiss it for arguments regarding other issues (I’m not going to name names).  For example, you can’t use this result to say we need more savings policy because the stock of debt is to high, then complain that the study is flawed when you want more government borrowing … just focus on the actual core elements of your frikken argument instead!

My problem with the result isn’t the excel errors, or anything R-R appear to have said – it is the way it has been used as an inconsistent marketing tool by people for selling their own unrelated ideological policies.  I’m just hoping that this shuts that up.

As a side note, here are my feelings on twitter:

People who think the R&R result caused austerity overestimate the impact evidence has on government policy.