Why The Rogoff-Reinhart Ratio Is So Destructive
Krugman recently wrote about the problems associated with the Rogoff-Reinhart ratio, which says that when a country’s debt-to-GDP ratio crosses above the 90-percent threshold, growth starts to slow. Krugman pointed out the obvious issue that correlation does not imply causation: does growth slow because of the high debt load, or is the debt load high because growth slowed?
In addition, Dean Baker notes that the debt-to-GDP ratio is probably not the best measure to even consider, because the ratio can be arbitrarily reduced or increased whenever interest rates rise or fall. If the yield on the 10-year rises over the next decade according to the CBO’s latest projections, then the Treasury will be able to reduce the debt-to-GDP ratio with just the click of a mouse by buying back debt issued today. (If rates rise as anticipated, then the debt issued today will come at a sharp discount.) According to Dean, the interest burden (as a percentage of GDP) is a better measure to consider, since it would not change in the event that the Treasury did decide to engage in such mouse-clicking shenanigans.
I think Krugman and Dean are basically right; but there is a broader philosophical point that needs to be made. That is, in a world in which the social scientist is not just a describer but also a prescriber, the Rogoff-Reinhart findings become not just wrong, but also destructive.
The issue of performativity is often underestimated by economists, who see themselves as objective scientists discovering all of the various relationships that exist in the macroeconomy. But what if it’s the case that by exposing such relationships, economists actually end up influencing the way in which policy is constructed?
This is, after all, exactly what happened when economists first introduced the Phillips curve in the 1950s, named after the late William Phillips. What the Phillips curve showed, in its original form, is that throughout history, there was always a tight negative correlation between inflation and unemployment. (The higher the inflation rate, the lower the unemployment rate, and vice versa.) However, once this correlation was exposed – once the social scientist made what he thought was an objective description – the world changed, as policymakers attempted to exploit the apparent trade-off between inflation and unemployment. Moreover, as policymakers sought to favor a higher inflation rate in order keep unemployment low, the original relationship that Phillips exposed actually broke down: once higher inflation expectations became ingrained into society — as they did in the late-1960s and 1970s — the trade-off between higher actual inflation and lower unemployment no longer held. The verdict among economists today is that there is only a trade-off between actual inflation and unemployment in the short run, before inflation expectations have the chance to become hardwired into society.
The point is, the same sort of interplay between the economist’s findings and the reactions of policymakers could easily occur with the Rogoff-Reinhart ratio. The risk is that even if the ratio is incorrect as an accurate threshold for fiscal sustainability, it may actually change the way in which policymakers perceive fiscal sustainability. The obvious danger is that it may cause governments to prematurely implement austerity policies at precisely the time when expansionary policies are needed. In fact, we may actually be seeing this in the US, as anyone who has followed the debate in Washington closely over these past few years knows that the Rogoff-Reinhart ratio has been used as a lever by the deficit scolds.
There are two ways forward. Either we need to make mainstream media more scientifically rigorous – why do reporters continue to cite the Rogoff-Reinhart ratio as important when it has such obvious flaws? – or we need more economists to understand the performative aspects of their research. On the latter point, it seems to me that if Rogoff and Reinhart truly understood the policy implications of their research, then they would have been much more careful in their analysis.