Winner of the New Statesman SPERI Prize in Political Economy 2016

Saturday 6 June 2015

The latest IMF paper on debt

This paper by IMF economists Jonathan Ostry, Atish Ghosh and Raphael Espinoza has attracted some media attention. (Flip Chart Rick has a good summary.) I want to talk about it because it does something that is quite rare - it talks about optimal debt policy in the longer run, rather than focusing on the shorter term issues associated with austerity. It also uses theory that I have used in a number of my own papers.

The headline result, that many will find startling, is that countries with what the IMF call ‘fiscal space’ have no need to reduce government debt at all, and should not therefore undertake fiscal consolidation to reduce debt - not today, or tomorrow. By fiscal space they effectively mean that the market is perfectly happy buying the debt, and are not suffering - and are unlikely to suffer - any kind of significant default premium that raises the interest on their debt. The paper suggests that most countries, including the UK, now have this fiscal space (see their page 4).

Many will find this message surprising, particularly coming from the IMF, because we are always hearing about why higher government debt is such a bad thing, and in particular how it imposes such a burden on future generations. However the result the paper is using is perfectly standard in the literature. To put it very simply, high debt does impose a burden, because the taxes required to pay the interest on that debt are ‘distortionary’ - for example income taxes prevent people from working as much as they should. But cutting debt also imposes a burden - taxes have to be raised to get debt down. Analogies between households and governments are misleading: while we as individuals do not live forever and therefore need to pay back debt eventually, the state can act as if it will go on forever. We therefore face a trade-off: is it worth paying higher taxes now in order to reduce government debt so we can pay lower taxes in the future? (Conceptually we can make the same point when talking about government spending cuts.)

The answer to that question depends on two key variables: the real rate of interest and the discount rate: the rate at which we discount utility in the future compared to utility today. In the standard model used by many/most macroeconomists, and used in this paper, these two variables are equal in the long run. If this is the case, it turns out that raising taxes today to cut debt and taxes tomorrow is a net cost, and it is better to leave debt where it is. That is where the headline message of the paper comes from. (Footnote for economists – [1].)

So suppose an economy suffers a positive government debt ‘shock’ - caused by a financial crisis, for example. The optimal policy is to leave debt higher, if the markets are happy to buy the debt (i.e, there is no additional default premium). The costs of getting debt back down again exceed the benefits. What that means for the UK, for example, is that we should not be going for a zero deficit, but should be happy with a deficit of a little over 3% of GDP that leaves the debt to GDP ratio constant.

Do I agree with this argument? The answer is no and yes.

No, because I think there are good reasons to believe that the real rate of interest on debt is normally a bit higher than the discount rate relevant to social welfare, although I admit this assumption is being put to the test right now (see the secular stagnation debate). If the long run real interest rate does exceed the discount rate, it becomes optimal to aim to reduce debt over time. (Footnote for economists - [2].)

Yes, because even in what I believe to be the more realistic case, debt is reduced very slowly. As some colleagues and I show in this paper, we could be talking about debt reduction over the period of a century or more. So, in terms of the current policy debate, the standard model used in this paper may be a useful starting point. (Footnote for economists - [3].)

The paper has plenty of interesting stuff in it. One point that I particularly liked addresses an argument that is often made to defend a rapid reduction in government debt in the UK: we must make room for the next crisis. On pages 12 and 13 the paper goes through a little cost benefit calculation to show why this argument is probably wrong. The paper does not argue that debt should never be reduced in countries with fiscal space. If opportunities arise to reduce debt without incurring significant distortionary costs, they should be taken. The obvious UK example where that was done in the past was the windfall from the sale of 3G spectrum licenses, which Gordon Brown used to reduce debt. The obvious example where that was not done were revenues from North Sea oil, which should have led to paying off public debt and/or building up a sovereign wealth fund as Norway has done. (Please note irony in terms of recent UK politics.)

In terms of the current policy debate, the message to politicians is clear and I suspect pretty robust. The shock of the financial crisis and Great Recession led to a large increase in debt levels in nearly all OECD countries. We should be in no hurry to try and return debt (relative to GDP) to pre-shock levels. That means that we can certainly afford to wait until interest rates have begun to rise, so that monetary policy can offset the impact of any subsequent fiscal consolidation. The case for reducing debt right now has no basis in standard macroeconomic theory.      

[1] This is sometimes called the steady state random walk debt result.

[2] In this case the optimal long run level of debt becomes negative. So what is called the ‘random walk steady state debt’ result from the paper is a knife edge result: even an epsilon increase in the long run real interest rate leads to a radically different optimal long run debt level. To see why we get this apparent knife edge, see the next footnote.

[3] The general result is that debt should be reduced, but slowly. The closer the discount rate gets to the real rate of interest, the slower the adjustment should be. When they are equal, adjustment should be infinitely slow - that is where the ‘steady state random walk debt’ result comes from. 


  1. "for example income taxes prevent people from working as much as they should"

    I know things like that tend to be stated as uncontroversial facts, yet they are in violent contradiction to my behaviour or that or pretty much anyone I know.
    Drop my income tax and I will either find a way of working part time or retire much earlier (OK, you can be productive without being paid so that might actually increase my social utility, but only because the wrong type of activities seem to be well remunerated), because I would not need the extra money.

    Yes, there is a small but very visible minority who are almost never sated, but even for them there is no linear increment to "satisfaction" - it's pretty much logarithmic. And with a logarithmic satisfaction curve, tax rate should have no impact on how much they work -the slope will be the same whichever the rate (if the rate was fixed, admittedly).

    1. There's nothing wrong with your THEORY here Cyrille - the income effect could dominate the substitution effect on leisure so that heavy tax would make you work more, not less. But empirically it usually just aint so - as innumerable microeconomic studies of the question over the last 60 years have shown.

      Mind you, those same studies tell you that the net effect (ie how much less you work if you are taxed) is typically smaller than the 'small government" people would have you believe - probably because of either diminishing marginal utility (your point) or rank utility. Just don't make their mistake of believing what it is personally agreeable for you to believe though.

  2. Simon, as ever, one needs to emphasize clearly in these pieces the distinction between IMF staff notes and IMF policy positions.

  3. Is "fiscal space" what we once used to call "savings" a long time ago?

  4. Is "fiscal space" what we once used to call "savings" a long time ago?

    1. The only "fiscal space" is real resources. As Japan has proved for two decades now.
      The concept of a currency issuing government saving is inapplicable.

    2. Fiscal space is a nonsense concept as explained recently by Bill Mitchell, and as I’ve pointed out for years on my own blog. For Mitchell, see:

    Bill Mitchell's view.

  6. IMHO it's incorrect to suppose, for a given date to the present, that the discount rate is unique. It can even inverse based on whether the utility is positive or negative. For instance, the person who is willing to take his or her lumps now is often (usually?) the same person who is willing to delay gratification. That is, a person can prefer pain in the present vs the future but prefer pleasure in the future vs the present. On the level of nations, it is entirely possible that the nation which does not want to get into debt, wants to get out of debt as quickly as possible. While I appreciate the economist's desire to arrive at value-free conclusions, this does not actually happen when one makes assumptions where particular values are assumed - and assumptions about the discount rate are value-laden.

    1. Actually that seems to be a lot of tosh, and confused about positive and negative. But making assumptions about the discount rate *are* assumptions about values, and if one parades the conclusions one gets as value-free, this seems IMHO to be a serious mistake. Whatever the long-term real rates, one country may rationally prefer not to be in debt, while another prefers to stay there.

  7. It's like driving a racing car. You don't take the bend late and as fast as you can, but you already adjust the wheel before that and accept having to use the shoulder. This way, the risk of a crash or of the car spinning out of control is minimized, there is no need to push the brake too harshly and overall speed (welfare) is highest.

  8. What result do we get if we set a discount rate of zero? For example, suppose we feel that we ought not discount future utility at all (even though we do), and we are interested in setting a normative ideal for long-term debt.

    1. Good question. That makes it much more likely that we have a real interest rate above the discount rate (=0) - not inevitable, because the real rate needs to be growth corrected, so its that r-g thing again. So positive long run debt is likely to be sub-optimal, but the optimal path from present levels is still likely to be very slow (see my paper that I linked to).

    2. This comment has been removed by the author.

    3. Interesting discussion: regarding the intertemporal solvency condition a zero discount rate (compared to a greater one) means a clear preference for a backloading fiscal consolidation path over a frontloading one which is usually considered by IMF as a stronger solvency requirement.

  9. When economists talk about adjusting public sector debt burdens- whether up , down , or sideways - without mentioning the effect on private sector debt burdens and on the combination of the two , I know they cannot possibly have any idea about how the real economy functions.


  10. This is one of those issues that annoys me - it seems like some leaders can't chew gum and walk at the same time. Over the long run, yes we want to lower debt. But we we talk about fiscal policy at ZLB, we are in a very rare situation.

    And when we look at employment, growth, output gaps, etc, all of these need to inform us on how to act. We should also look at the effects of austerity around the world and how effective it has been at reducing debt to GDP (not so much - see the Sen article, excellent).

    There's a minor kerfuffle about economists and ideology - of course it involves Krugman. Russ Roberts says economists, esp. Krugman and even himself, view economics through ideological priors. Keynesians always want big big government!

    No mention ZLB, liquidity trap, etc. Just small government good, big govt. bad. I find this cheaply cynical. I'm going to be ideologue immune to empirical evidence b/c everyone is.

    Mom, everybody's doing it!

    Anyway, when we're near full employment, interest rates and inflation are up, we can move on the debt.

  11. Speaking as an MMTer, watching the IMF struggling with the question as to what the optimum amount of debt is, is like watching a child grow up: the IMF is gradually tumbling to ideas that are second nature to MMTers.

    First, it’s good to see the word “optimum” being used. That concept is beyond the grasp of most of the World’s so called “intelligentsia”. But if the IMF is now thinking about what the optimum debt is, that’s a step forward.

    Second, the standard MMT view is, first that the state must provide the private sector with what MMTers call the “private sector net financial assets” that the private sector wants, else the private sector will try to save so as to acquire those assets and we get Keynsian paradox of thrift unemployment. (Incidentally PSNFA equals debt plus base money.) A second point often made by MMTers is that a country that issues its own currency can pay any rate of interest on its debt that it likes.

    As to what the optimum rate is, that’s a difficult question. Leading MMTer Warren Mosler (in common with Milton Friedman) advocated a ZERO rate: i.e. those two advocate/d that the only liability the state should issue should be base money.

    Personally I’d advocate any old rate as long as it’s less than inflation. That way the state profits at the expense of its creditors.

  12. It is beyond irritating that professors of economics still talk about reducing public debt when discussing sovereign, fiat currency issuing nations. Such a nation, having an unlimited supply of money, needs never to pay off its debt, it doesn't need to borrow for economic reasons at all. And the GDP equation shows quite clearly that public debt generates private income. And history shows that public surpluses eventuate in recessions. Why does this message not penetrate?

  13. RGC - Public debt directly generate private WEALTH - it means taxes have been stopped in favor of borrowing the cash instead, by way of transferring an asset to this class.

    The UK, Japan, the EZ, and the US (in particular) are wealthy.

    Let us raise taxation to recapture the unearned windfalls of poor public policy and distorted labor markets of the last 30 years so borrowing can be lowered. This should in fact keep the borrowing costs of the public's govt lower (principal and its interest costs) - allowing the public a lower cost way to finance public goods (real things and systems) that have benefits that also occur in the future - the classical public finance reason why states issue 30 year bonds to finance big bridges, etc. - it makes no sense to burden the current period with all the costs, which are significant, when many if not most of the benefits occur over 50 years.

    What is "optimum" cannot be just a sense of 12-month period flows.

    This notional goal should consider the net wealth of residents within the polity (particularly if a period or periods of economic policy caused major distortions in markets and wealth-distributions).


  14. "But we we talk about fiscal policy at ZLB, we are in a very rare situation."
    It does not have to be. Warren Mosler advocates permanent zirp and just using fiscal policy.

  15. Simon considers only changes in taxes to change debt. I think that most people who wish to reduce debt wish to do so by reducing certain expenditures that they believe have NPV<0 at the borrowing rate (e.g., transfers from high income taxpayers who reduce their output to low income non-payers who may also reduce their output). I has little to do with multipliers and interest rates and discount rates.

  16. Is this paper historically well-informed?

    Or does it rely on macro-theory?

    When you are talking about the impact of debt on the political and economic stability of a country, you really should look over almost a century. I would hope, if this is serious, that there is a lot of good qualitative and quantitative historical documentation.

  17. Does anyone have a view on what the equivalent IMF report on private debt levels would look like and how much fiscal space (whatever that's meant to mean) exists in each country? What is the endogenous view on debt sustainability? My bet is that this graph would be the mirror of the other and allow us to frame this question optimally.

  18. OK, if government is issuing debt it means on the other side of the fence the private sector is saving more then they are borrowing. The "debt" (which could be base money) will be paid off when the private sector stops doing this. Currency issuing government borrowing is borrowing in the same way a bank borrows when you get paid and your deposits increase. Trying to "pay off" this is ridiculous.

    Government spends by crediting the reserve accounts of private banks, who then credit their deposit accounts and taxes by debiting reserve accounts at banks who debit deposits. Taxes are not actually paid in deposits. When you pay tax, you have to stump up real money otherwise the tax is not settled.
    Debt does not "finance" spending. The govt always spends by "printing money." Taxes "unprint money." The govt just borrows back what it spends to allow the central bank to hit its overnight rate.

    "MMT primarily describes the policy benefits of a particular system – one where the private banks are locked in a fixed exchange rate system for their own liabilities with the liabilities of one or more central banks under a legally enforced arrangement, but where that central bank floats their own liabilities on the currency markets.

    It also describes the follies that ensue when the central banks fix the exchange rate of their liabilities with each other or with a supranational central bank.

    To pay your taxes you need to present the government’s own bank’s liabilities to settle that debt. If you don’t then you haven’t settled them.

    Therefore for the taxes to be credited to the National Loan Fund Account or Consolidated Fund Account (using the UK as an example), the banks making the payment have to have access to central bank liabilities.

    It is the function of the intermediary Government Banking Service to translate whatever liabilities the punter provides into central bank liabilities which can be transferred to the Tax accounts at the central bank.

    And that is the key driver in the system. The operation of the tax account at the central bank is the way that central bank liabilities are given to and taken from the private banks rather than lent to them.

    And that matters – due to the fixed exchange rate between the central bank and the private banks.

    I’ll use Treasury rather than Government. Government owns or controls both the central bank and the Treasury in modern states. MMT generally consolidates the balance sheets of Treasury and Central Bank to avoid the confusion of the internal transactions.

    The Treasury can’t create central bank money. If it credits an account, one has to be debited as well. The tri-party transaction involves the Treasury, the central bank and the private bank.

    What the Treasury is doing functionally is purchasing new private bank liabilities with central bank liabilities via the internal fixed exchange rate mechanism.

    So an amount of central bank liabilities are transferred to the private bank. That causes the private bank to increase its own assets and liabilities via the fixed exchange rate structure. Those private bank liabilities are then transferred to the target of Treasury’s beneficence.

    You’ll note that mechanism is functionally the same as the private bank issuing a loan. Assets and liabilities are increased and the private bank receives an income from the assets (interest on reserves). It is forced private bank money creation.

    Similarly in reverse. The overburdened tax payer presents some private bank liabilities to the Government Banking Service at a private bank. The private bank transfers some central bank liabilities to the Treasury central bank account and the private bank extinguishes an amount of its own assets and liabilities under the fixed exchange rate mechanism to maintain parity with the central bank.

    Again functionally the same as paying off a private loan. Assets and Liabilities go down and the private bank loses the income from the assets. In other words forced private money destruction."

  20. This was a hilarious read because I'm also reading Stiglitz's book detailing his exasperation with his job at the World Bank.

    It doesn't matter what IMF economists finally discover to be simple truth. What matters is the clueless opinion of the plutocrats running the IMF.


Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time. In addition, I cannot publish comments with links to websites because it takes too much time to check whether these sites are legitimate.