An interesting rebuke to the oft-quoted 2010 Reinhart-Rogoff paper “Growth in a Time of Debt” was announced today. The paper identifies a magic 90% debt-to-GDP level above which median economic growth falls one percent and average growth falls even more.
The paper was not without its critiques — many argued that the causation was backwards (i.e. slow growth leads to higher debt-to-GDP ratios). Nevertheless, it was highly quoted in political and economic policy discussions. Not until today has someone actually argued that the data itself was wrong.
That’s what Thomas Herndon, Michael Ash, and Robert Pollin do in their new paper “Does High Public Debt Consistently Stifle Economic Growth? A Critique of Reinhart and Rogoff”. Reinhart and Rogoff shared their original Excel spreadsheet and lo-and-behold, Herndon, Ash and Pollin found three major issues:
- Reinhart and Rogoff selectively excluded years of high debt and average growth
- Reinhart and Rogoff used a debatable method of weighting the data
- There appears to be a coding error that excluded high-debt and average-growth countries
For more detailed information on each of these critiques, see a well-written breakdown here.
The results? High debt-to-GDP resulted in 2.2% growth, not -0.1% growth. Potentially more interesting: the team could not identify a break point after which economic growth fell significantly.
There are a few takeaways here for those of us in the quantitative research field. Primarily, it reminds us that just because a paper has been published and peer-reviewed, it does not mean the data or the code has been peer-reviewed. Relying on someone else’s research as a key component of your own models, without first verifying the results, is equivalent to relying on a rumor. That’s why we find it important to do all our own research and develop all of our own models. While we constantly read new academic and industry papers, we always attempt to replicate before incorporating any results.