Let me highlight a passage from the "Understanding Our Adversaries" evolution-of-economists'-views talk that I started giving three months ago, a passage based on work by Owen Zidar summarized by the graph above:
The argument [for fiscal contraction and against fiscal expansion in the short run] is now: never mind why, the costs of debt accumulation are very high. This is the argument made by Reinhart, Reinhart, and Rogoff: when your debt to annual GDP ratio rises above 90%, your growth tends to be slow.
This is the most live argument today. So let me nibble away at it. And let me start by presenting the RRR case in the form of Owen Zidar's graph.
First: note well: no cliff at 90%.
Second, RRR present a correlation--not a causal mechanism, and not a properly-instrumented regression. There argument is a claim that high debt-to-GDP and slow subsequent growth go together, without answering the question of which way causation runs. Let us answer that question.
The third thing to note is how small the correlation is. Suppose that we consider two cases: a multiplier of 1.5 and a multiplier of 2.5, both with a marginal tax share of 1/3. Suppose the growth-depressing effect lasts for 10 years. Suppose that all of the correlation is causation running from high debt to slower future growth. And suppose that we boost government spending by 2% of GDP this year in the first case. Output this year then goes up by 3% of GDP. Debt goes up by 1% of GDP taking account of higher tax collections. This higher debt then reduces growth by... wait for it... 0.006% points per year. After 10 years GDP is lower than it would otherwise have been by 0.06%. 3% higher GDP this year and slower growth that leads to GDP lower by 0.06% in a decade. And this is supposed to be an argument against expansionary fiscal policy right now?
The 2.5 multiplier case is more so. Spend 2% of GDP over each of the next three years. Collect 15% of a year's extra output in the short run. Taking account of higher tax revenues, debt goes up by 1% of GDP and we have the same ten-year depressing effect of 0.06% of GDP. 15% now. -0.06% in a decade. The first would be temporary, the second is permanent, but even so the costs are much less than the benefits as long as the economy is still at the zero lower bound.
And this isn’t the graph that you were looking for. You want the causal graph. That, worldwide, growth is slow for other reasons when debt is high for other reasons or where debt is high for other reasons is in this graph, and should not be. Control for country and era effects and Owen reports that the -0.06% becomes -0.03%. As Larry Summers never tires of pointing out, (a) debt-to-annual-GDP ratio has a numerator and a denominator, and (b) sometimes high-debt comes with high interest rates and we expect that to slow growth but that is not relevant to the North Atlantic right now. If the ratio is high because of the denominator, causation is already running the other way. We want to focus on cases of high debt and low interest rates. Do those two things and we are down to a -0.01% coefficient.
We are supposed to be scared of a government-spending program of between 2% and 6% of a year's GDP because we see a causal mechanism at work that would also lower GDP in a decade by 0.01% of GDP? That does not seem to me to compute.
Now tonight I have been nibbling the RRR result down. Presumably they are trying to see if it can legitimately be pushed up. This will be interesting to watch over the next several years, because RRR is the heart of the pro-austerity case right now.