I’ve been a bit mystified by Excel-gate (see this good, balanced summary by Gavyn Davies). Bravo for Thomas Herndon, the graduate student who uncovered the error in the now-notorious paper by Carmen Reinhardt and Kenneth Rogoff; his job prospects will be rightly enhanced by this episode.
But the glee with which anti-Austerians pounced on this episode to ‘prove’ that austerity doesn’t work seems to involve an assumption that the original Reinhardt-Rogoff paper of 2010 ‘proved’ anything to the contrary in the first place. There are lots of papers about the impact of debt/GDP ratios on growth, and they demonstrate all kinds of different things – see for example this BIS paper by Cecchetti and others, or this IMF paper (pdf) from last year on the Caribbean economies, or this Fed paper published in December (pdf), or this much-cited 2010 paper by Koehler-Geib and others, or for that matter the new paper debunking Reinhardt and Rogoff’s 90% as it too finds the same correlation albeit with different numbers.
Well, you get the idea. Taking these together, we ‘know’ there might be a threshold for sovereign debt, but it varies over time and across countries, it’s a correlation whose causal direction and mechanism is unclear, and there isn’t enough data for any estimates to be robust (because history only runs once). All of which only goes to underline how little is known about the macroeconomy, not to mention how hard any macroeconomists and their camp followers find it to resist claiming certainty where there is none.
No doubt Reinhardt and Rogoff were tempted into over-claiming for their work by the politicisation of the debt threshold issue. But the underlying message of their big 2009 book, [amazon_link id=”0691152640″ target=”_blank” ]This Time is Different[/amazon_link], is unscathed: unlike the later paper, it makes it absolutely clear that debt ‘thresholds’ above which increasing borrowing is correlated with lower growth vary widely in different countries and at different times (no magic 90% here); and that the historical record indicates it generally takes a long time for growth to recover after banking crises involving debt overhangs.
[amazon_image id=”0691152640″ link=”true” target=”_blank” size=”medium” ]This Time Is Different: Eight Centuries of Financial Folly[/amazon_image]
At a more basic level, I have a theory. OK, Oh dear, I call it the “Wile E Coyote Syndrome” as if you do not realise that you have run off the cliff until you are already into space and there is only one way to go, that is down.
Given that many of our leaders (e.g. Olli Rehn in Europe, various members of Congress in the USA) took to citing R&R’s 90% threshold as fact – bolstered in no small part by embarrassing (for academics) newspaper columns written by R&R, it seems a little shortsighted to claim that debunking this “line in the sand” isn’t important.
I don’t really agree with you because so much of the debunking is taking the form ‘the 90% threshold isn’t correct therefore the right number is enormous and deficits and debt ratios don’t matter’. Looking at a fairly random but respectable bunch of papers on this yesterday, the most plausible conclusion is that there *is* a correlation between higher debt ratios and lower growth rates most of the time but we don’t know the level in any instance, could be 150%, but could be as low as 50%. I object to the spurious certainty expressed in both directions!
Everyone appears to agree that there is a correlation. The discussion is more about which way the causation runs (does debt reduce growth, or does low growth increase debt?). So far the data *and* predictions from economic theory appear to indicate the latter.
The idea that there is a specific “level” at which debt becomes problematic was introduced by R&R, but was generally disbelieved at the time, and much more so now. There’s no real reason to think any such level exists.
Seems to me theory would point to two-way causation, and evidence indicating context very much matters is consistent with that.
It doesn’t seem that unreasonable to me that some threshold exists for each country beyond which a death spiral begins. I’d propose the following mechanism:
1) Country experiences low growth
2) Automatic stabilisers kick in (which is fine, it’s what they’re for)
3) Public spending increases
4) Surplus is reduced or turned to deficit, or deficit increased
5) Public debt increases
6) Markets lose confidence in state being able to repay its debts to expected schedule, or at expected rates
7) Markets demand higher yields from government bonds
8) Public spending increases further due to increased cost of borrowing
[…]
The key step is step 6, and this depends on a state’s track record for effective public spending and repaying its debts as planned. So each country presumably has its own threshold; some below 90%, some well above.
People in general, I would say, rather than markets. If that point is ever reached in practice, a non-linearity or ‘threshold’ seems entirely plausible to me. And to repeat the post, there are other papers finding apparent threshold effects or rather non-linearities, albeit at different debt/GDP ratios.
‘Unscathed’? The threshold is now a ‘threshold’ and a clear correlation is now one that varies ‘widely in different countries and at different times.’ At best, RR may still be citable by academics on their lesser point of a general correlation, but it has been rendered completely useless to policy makers (except perhaps as a punchline).
To underline the point of the post, the book and the subsequent paper make different arguments…..
Many thanks for the comments on this post.
Pingback: Tunnelvisie | All out of bubblegum...