Friday, 19 April 2013

Reframing Reinhart & Rogoff

The economics world is aghast. Two distinguished economists, Carmen Reinhart & Kenneth Rogoff, have been shown to have produced shoddy work.

I don't propose to comment on the details of the case. Suffice it to say that Reinhart & Rogoff were wounded by a paper that showed that their data was flawed. They defended themselves, and were also defended by numerous economists around the world who argued that although the data might be flawed, the economic analysis justified their conclusions. But the next day, the Rortybomb blog delivered the killer punch. Econometric analysis by Arindrajit Dube demonstrated that even with good data, the economic analysis was flawed and the conclusions unjustifiable. High public debt cannot reliably be shown to cause low growth. But low growth can reasonably reliably be shown to cause high public debt.

This is a no-brainer, actually. There are two reasons for this.

Firstly, public debt is normally quoted in relation to GDP. This is a ratio. Debt/GDP can rise either because debt increases at a faster rate than GDP, or because GDP falls faster than debt does. So when an economy is in recession, public debt INEVITABLY rises as a proportion of GDP, not because the nominal amount is increasing (though it probably is, as I shall explain shortly) but simply because GDP is shrinking. Severe and prolonged recession can make the debt/GDP ratio look simply terrible.

Secondly, in a recession public spending tends to increase. This is because of what economists term "automatic stabilisers", which replace some of the normal functions of the private sector when the private sector is retrenching. For example, unemployment benefits maintain basic incomes for the unemployed while they look for jobs, supporting demand in the economy and preventing serious hardship. In a recession, people tend to lose their jobs, so unemployment rises and with it the benefits bill. At the same time, tax revenue falls because GDP is falling. Therefore, in a recession public borrowing must increase because the gap between public spending and tax revenue increases (the fiscal deficit, roughly). The only way it can fail to increase is if public spending is drastically cut and/or taxes increased, interfering with the action of automatic stabilisers and crushing demand in the economy. I will return to this shortly.

So in a recession, assuming no direct action is taken to prevent it, the debt/GDP ratio naturally increases. Therefore the assumption should be that, unless shown otherwise, poor growth causes high debt, not the other way round. I am astounded that so many people have inferred the opposite from Reinhart & Rogoff's paper, and even more astounded that until Dube, no-one appears to have done any serious econometric analysis to confirm the direction of causation. Dube's econometrics are pretty basic. As Jonathan Portes commented on Twitter, why on earth didn't Reinhart & Rogoff do them in the first place?

The problem is is that people who worry about debt/GDP ratios see the ratio rising as GDP falls, and start to panic. And Reinhart & Rogoff's flawed analysis encourages them to do so. This has led to calls for restraint of fiscal deficits and debt-cutting programmes when economies are in deep recession. Many of the worst examples of severe and counterproductive fiscal austerity have been "justified" on grounds of public debt/GDP close to the Reinhart & Rogoff "tipping point" of 90%. That tipping point has now been shown to be completely arbitrary. Many people could be justifiably angry at the damage that has been done through misguided fiscal policy done on the basis of Reinhart & Rogoff's work. Not that they were the only ones, of course. Other economists have much to answer for too - Alesina, for example, with his notion of "expansionary fiscal contraction" which has been shown to be a very rare breed indeed.

However, I did say that I was reframing this debate. And that is what I shall now do. You see, Reinhart & Rogoff completely missed the point. Even if their analysis was right, even if the data were reliable, it would still be unhelpful. They are looking at the wrong metrics.

Debt/GDP is a pretty confusing metric, since it compares a stock and a flow. It would be more meaningful to compare fiscal deficit/GDP. But even that is not ideal, since GDP measures activity in the economy, not government income, and the government income figure is netted with spending to give the deficit. The reality is that governments do not have to pay all their debt off in one year - in fact most governments pay off little or no debt. What they have to do is service the debt. And their ability to service the debt is governed by two things: 1) the interest rates prevailing on outstanding debt stock and on new issues during that period: 2) revenue from taxation and other income.

In a recession, government income reduces as GDP falls. And as I've already noted, the nominal amount of debt tends to increase in recession due to automatic stabilisers. Therefore, in a recession, governments may have problems servicing debt unless interest rates also fall. Which in most cases they do, forced down by the actions of central banks. But note that this has nothing whatsoever to do with the debt/GDP ratio. Interest is paid on the nominal amount. And if interest rates are high enough, a debt/GDP ratio well below 90% could be unsustainable. Or, of course, if tax revenue is too low. High GDP does not necessarily imply large tax revenues - that depends on the design of tax policy and the effectiveness of collection.

For example, what we saw in the Eurozone periphery in 2010 was interest rates rising in recession-hit countries where the debt/GDP ratio was seen as unsustainable. Rising to the point where even if the country were not experiencing a massive GDP contraction and catastrophic fall in tax income, the debt would be impossible to service.

Every single one of the Eurozone sovereign bailouts has been to enable countries to SERVICE their debt, not pay it off. The IMF always comes out with some guff about "putting public debt on a sustainable path", but this is never achieved through the bailout itself. Absent some form of default or debt restructuring, the only way to reduce the nominal amount of public debt is by running a sustained fiscal surplus. And as I've already noted, reducing the nominal amount of public debt doesn't necessarily improve the debt/GDP ratio: in fact if running a fiscal surplus results in poor economic performance - which is distinctly possible, since fiscal surpluses extract money from the private sector in excess of the amount actually needed to meet current public spending commitments - the debt/GDP ratio could worsen even when the nominal amount of debt is reducing. I admit this would be unusual, but it is definitely possible. And this brings me to the severe fiscal austerity programmes that have been introduced in many countries with the intention of achieving fiscal surplus and therefore reducing nominal debt over the medium term.

Fiscal austerity tends to cause economic contraction. This again is something of a no-brainer: the more money you extract from the private sector through higher taxes and/or spending cuts, the less money the private sector will have for spending and investment. The IMF's paper on debt reduction during fiscal consolidation suggested that debt/GDP would actually rise during the first few years of fiscal consolidation. And they caution against targeting debt/GDP as a measure of the success of fiscal consolidation programmes. They comment that targeting debt/GDP means that as the target will inevitably be missed in the first few years, there will be political pressure to tighten policy even more, causing further economic contraction and driving the economy into a deflationary spiral. And they observe that permanent damage can be done to economies that suffer repeated fiscal tightening in the name of reducing debt/GDP. Yet this is exactly what is being done in a number of Eurozone countries. No wonder the Eurozone is in recession, parts of it deeply so.

So not only is debt/GDP a flawed metric, it is also dangerous when used as a policy target, and unhelpful as an indicator of a government's current ability to service its debt. However, it does serve one useful purpose - and interestingly this is not often mentioned. Just as the debt/equity ratio gives a useful indication of a company's financial fragility, so the debt/GDP ratio indicates the sensitivity of a economy to economic shocks. Countries with high debt/GDP ratios are more likely to have trouble servicing their debts when GDP falls and more likely to find it necessary to raise taxes and/or cut other spending.  And it is fair to say that nervous investors are not too happy about high debt/GDP ratios either - partly, it has to be said, because of Reinhart & Rogoff's work - so yields tend to rise with the debt/GDP ratio provided there isn't somewhere else in the world in a worse mess. At the moment there are lots of places in a mess, so debt/GDP is not a reliable indicator of investor attitude to risk. Having a functioning central bank seems to be much more important.

Debt/GDP is also sometimes suggested as an indicator of interest rate risk, but I disagree with this. As I said earlier, interest is paid on the nominal amount. It would be far better to do sensitivity analysis to identify the likely effect on government finances of changes in interest rates.

But of course the metric you use to measure debt is completely irrelevant anyway when you remember that government debt is not what you think it is. So the real problem with Reinhart & Rogoff, and indeed the mainstream economic view of debt, is that worries about the size of the stock of debt (nominal or debt/GDP) are largely unfounded when a plentiful supply of safe assets is essential to the smooth running of the monetary system. And it is completely illogical for governments to impose severe austerity to reduce fiscal deficits while encouraging central banks to purchase safe assets and create unlimited bank reserves. All this does is create imbalances and weird distortions in the monetary system: the two actions cancel each other out and the result, as we are seeing, is stagnation. When will we understand the real role of public debt in our fiat currency system?


Does High Public Debt Consistently Stifle Economic Growth? Herndon, Ash & Pollin
On Reinhart & Rogoff - Ritwik Priya
Reinhart-Rogoff recrunch the numbers - Chris Cook, FT (paywall)
Reinhart-Rogoff and Growth in a Time Before Debt: Arindrajit Dube at Rortybomb
Reinhart & Rogoff's scary red line - Azizonomics
The Challenge of Debt Reduction during Fiscal Consolidation - Eyraud & Weber, IMF (pdf)
Still missing the point on Reinhart-Rogoff - Pragmatic Capitalism
Revisiting the evidence on expansionary fiscal austerity - voxeu
Government debt is not what you think it is - Coppola Comment
When governments become banks - Coppola Comment


  1. Basically agree, bar a few quibbles as follows.

    1. Para starting “First, public debt…”: the idea that the debt/GDP ratio will change much because an economy contracts in a recession is a new one on me. First, some recessions involve no GDP contraction at all – i.e. they just involve an absence of growth. Second, even if GDP contraction IS INVOLVED, it’s normally no more than 5% or so. So the maximum deterioration in the debt/GDP ratio you’re likely to get from the above cause is about 5%, which is insignificant.

    2. “even more astounded that until Dube, no-one appears to have done any serious econometric analysis to confirm the direction of causation..” I’m not sure about that: it was blindingly obvious to large numbers of us long ago (MMTers in particular) that Rogoff and like-minded persons have no grasp of economic theory. As to “econometrics”, I don’t attach much importance to including maths in economics. Rogoff employed plenty of maths – and look at the result.

    3. One point missing from the above article is that a deficit does not necessarily involve any increase in the debt at all. As Keynes pointed out, a deficit can be funded either by increased debt or simply by printing money. Milton Friedman also advocated a system under which there is never any government debt. Warren Mosler advocates likewise. I agree with them.

    1. 1) The standard definition of a recession is two successive quarters of negative growth. Negative is not zero.

      2) "Econometric" and "economic" are not the same thing. Many people may have questioned R&R's economics. But there don't seem to have been any econometrics. Which is what I said.

      3) I refer you to my final paragraph, and the associated link, where I point out that govt debt and money are pretty much the same thing. Deficits always require an increase in something, and that something is either debt or money. That is basic maths.

  2. Perhaps I'm stupid but in above quibble 3., it is stated "a deficit does not necessarily involve any increase in the debt at all." and then it is stated "a deficit can be funded either by increased debt or simply printing money." hmm... Aren't both ways an increase in debt? And wouldn't the latter method cause inflation to rise? Someone please explain. We go from the clear blog to the obscure comment! Scratches head....

    1. thanks. I have replied to Ralph's comment. I hope this clarifies things for you.

    2. Anon,

      You ask whether printing money “wouldn't . . . cause inflation to rise?” The answer is that debt and money are very similar, so there is not much difference inflation-wise between having government print money rather than borrow (or “print” debt). But to the extent that money printing is more stimulatory and inflationary, the solution when going for a particular amount of stimulus and adopting the print option rather than the borrow option is to print a bit less than whatever amount would have been borrowed, had one gone for the borrow option.

  3. Of course anyone with half a brain has been saying since the crisis that what is required is growth.
    The 2.2% growth the paper links with 90% debt/GDP ratio would be very welcome!

    As Danny Blanchflower says SMEs are crucial to this, and he isn't wrong.
    The Panacea is linking banker and bank taxes to the performance of the economy.
    The easiest link would be unemployment level.

    Dynamically set banker's top-rate income tax to ten times the unemployment rate would have seen it rise from ~50% in 2008 to around 80%-83% recently.
    In a bid to reduce their income tax, I think that SMEs would see honestly priced finance; this would lead to an immediate and sustained drop in unemployment and increase in GDP. A virtuous circle then forms with the financial sector then tied to the real economy like a sheep savaging dog tied to a burly ram.

    The bank-levy? Simple: the cost to the state of Jobseekers Allowance, about £6bn a year just now.

    1. I like that idea. Make the bankers pay the dole.
      It's worth discussion at the very least.

  4. Thank you Frances, crystal clear as usual :-)

  5. "This again is something of a no-brainer: the more money you extract from the private sector through higher taxes and/or spending cuts, the less money the private sector will have for spending and investment. "

    That depends if the private sector was intent on saving that money in excess of investment.

    In which case taxing away that extra saving may have little effect.

    Removing an amount of tax relief on pension contributions would be one example.

  6. An elegant debunking as usual. Frances for Chancellor!

    But what about the original referees and publishers (National Bureau of Economic Research - a private foundation)? Are they not jointly culpable? The schoolboy/girl Excel error is unpardonable but that is only partly responsible for the casual misinterpretation, which is economics 101 level.

  7. "The economics world is aghast"

    But not it seems, MMT scholars. Please see Wray & Nersisyan's 2010 paper:

    "Does Excessive Sovereign Debt Really Hurt Growth? A Critique of This Time Is Different, by Reinhart and Rogoff"

    Referring back to that earlier critique in his blog on R&R this week, Randy Wray reminded us(without recourse to any econometrics at all!) that their thesis is, in essence, simply illogical garbage ...

    "Here’s the bigger problem highlighted by Yeva and Me: they do not know what they are talking about. Sovereign countries that issue their own floating currency cannot be forced into involuntary default no matter what the debt ratio."

  8. Some rough and ready thought on the metrics:

    Would not interest cover, ie annual tax revenue/annual interest payments, be a more meaningful metric to assess a government's "solvency"?

    Secondly, would it not make more sense to use the ratio of govt net borrowings in the year to government spending in the year ?

    Both these metrics compare a flow to a flow

    1. I absolutely agree. Those are far more meaningful metrics. Particularly the interest cover.

  9. I understand the argument against austerity, but what alternative strategy do you propose? An increase in short/medium term public expenditure? Can sustainable growth be manufactured like this? Bowles-Simpson report in US also pointed towards austerity as the answer. Should their work be disregarded too?