QE is not a liability transformation – it is debt reduction
The remuneration of bank reserves has been a source of significant analytical confusion among economists. This arcane technicality matters far more than it should . The two most significant implications are: 1) public sector debt across the developed world is around 30% lower than typically estimated, and 2) central banks have limitless power to meet their inflation mandates, without veering into fiscal policy. As an aside, it also follows that the entire debate about central bank ‘equity’ is a distraction, and the widespread belief that monetary and fiscal policy are blurred is incorrect. There is a crystal clear distinction between monetary and fiscal policy, which is why we have independent central banks.
Recognition of these facts would aid policy substantially. We need to rethink the tools of monetary policy – not the objectives – and there is far more fiscal space than conventional analysis assumes.
The conventional argument now seems to be that QE simply converts fixed rate government liabilities into liabilities at overnight rates . This appears obvious to a US-centric economist because the Fed ‘pays’ the target overnight interest rate on reserves. Reserves are treated in central bank accounts as ‘liabilities’, so one liability – a government bond with a fixed coupon – has simply been swapped for another liability which pays the overnight interest rate.
It shouldn’t need repeating, but accounting convention is just that – accounting convention. And accounting convention has not been static. Financial liabilities – the obligation to make a payment at a future date or in contingent circumstances – are real, but identifying and measuring them accurately in accounting terms is hard (if in doubt, read Warren Buffett’s shareholder letters), or this shorter take.
Unsurprisingly there is a thoughtful, albeit small, literature which considers the accounting treatment of central bank reserves, most of which concludes that they do not have the simple characteristics of liabilities, and are almost certainly not liabilities at book value. This superb paper by Bank of England researcher, Michael Kumhof, and others, is probably the most insightful.
Why might we doubt the accuracy of accounting convention for reserves? The first obvious point is that notes and coins are not financial liabilities. A financial liability is the obligation to make a future payment – there is no such obligation with physical cash. Physical cash is the means of payment. Given that bank reserves are really just the electronic equivalent of notes and coins it would seem odd that the simple shift from physical to electronic form should transform something which is not a liability into a liability.
This is where the remuneration of reserves becomes paramount. But first let’s clarify the unique properties of reserves when compared to government bonds. Government bonds are traded in the open market, and their price (yield) is determined by the market exchange of bonds. When governments issue bonds the private sector is under no obligation to buy them. It does so at a price of its willing. To be clear, none of this really changes with QE. The outstanding stock of government bonds remains vast despite central bank holdings, and most developed government bond markets are therefore highly liquid, so if the private sector wants to change the yield, it can. For this reason, most estimates of the impact of QE on bond pricing are unsurprisingly negligible.
Either way, reserves are created in a completely different manner to government bonds, they are not subject to a potential private sector funding constraint. Because they are the electronic form of money between banks, and the central bank is the monopoly producer, the central bank can buy things with newly created money at will. Indeed the very process of creating reserves (buying assets or making transfers) already distinguishes reserves from bonds which are sold into a market.
To put it simply: you buy things with money, but you have to sell bonds to someone. Reserves are therefore qualitatively different to government bonds. This should be macroeconomics 101, as it is the basis of the distinction between fiscal and monetary policy, both theoretically and institutionally.
The crux of the matter therefore seems to be the remuneration of reserves. Bank reserves are not simply electronic notes and coins because central banks pay interest on them. This of course was not always the case, but it is currently. Does the payment of interest on reserves fundamentally change their nature? An immediate reason for scepticism is that the payments made to holders of bank reserves (regulated commercial banks) is not an ‘interest rate’ as we typically think of one. It is a percentage payment on the notional value of the reserves, but typically an interest rate is a necessary compensation (a price) required to create a voluntary market exchange between an issuer of a bond or loan and the buyer of a bond or provider of a loan. This is not the case with bank reserves. The central bank can remunerate however much it wants (positive or negative) without affecting the quantity outstanding. That’s because reserves are money.
So why do central banks pay ‘interest’ on reserves? Interestingly, the answer is entirely independent of all of these issues, and rests on a practical matter of how to control money market interest rates – the rates at which banks lend to each other, and which act as a benchmark for the pricing of loans and deposits across the financial system. Recent innovations in reserve management have rendered this need redundant. It is no longer the case that a positive or negative IOR is necessary to control money market rates. By paying different interest rates on required reserves and excess reserves – as a number of central banks now do – the central bank can now pay whatever ‘interest rate’ they want on reserves, independently of the money market rate they are targeting. If the central bank is targeting a negative money market rate, it pays a positive rate on required reserves, netting to zero. Conversely, if the money market rate is positive, the rate on required reserves would be negative.
This begs an obvious question: how much should banks be compensated for holding reserves? As is always the case when we think of a general equilibrium model with competition, it may not matter very much. If banks are ‘penalised’ for holding reserves, capital should leave banking and margins should rise – i.e. the rest of the economy will pay more for banking services, so everything nets to zero. There may be truth in this framing, but it feels unsatisfactory and clearer guidance would seem helpful.
There are very strong grounds for paying a zero interest rate on reserves. The most obvious is that banks are taking zero credit risk. Reserves are electronic money. Base money has no credit risk, but is required for the functioning of the banking system, and is created under a monopoly by the central bank. What is true is that holders of base money – either in its electronic form (bank reserves) or its physical form (notes and coins) – are exposed to inflation risk. If we impose the costs of inflation on banks’ holdings of base money, this cost will presumably be passed on to some extent through the bank system. The challenge with targeting a zero real return on reserves, although it may be ideal, is that inflation is imperfectly measured and we are almost certainly overstating it. In a world where central banks are targeting less than 2% inflation, zero nominal remuneration seems reasonable. After all, central banks are providing a services to the commercial banking system – reserves on demand – which it seems fair to charge commercial banks something for. (Interestingly, if the Fed was to be consistent with its own philosophy on this it would average a rate of interest equal to inflation. It is intriguing that following the shift to IOER targeting, no US economists have argued that the Fed should offset the effects of the transfer to banks with a change in the IORR).
A competing rule worth considering is that the central bank should target zero net profitability. Again, this trivial feature of central bank accounting, which results in transfer payments of “profits” to governments has confused many economists, not just opponents to the ECB’s bond purchases, that there is a blurred line between monetary and fiscal policy. An easy way to do this under the current accounting regime of course would simply be to provision away any net income surplus. There would be obvious distributional consequences within the financial sector in targeting zero net income, which is another reason to abandon asset purchases as the main means of trying (failing) to stimulate demand. Direct transfers to households are hands down more efficient, with no impact on bank profits or fiscal positions under a regime of zero interest on reserves.
My suggestion is that monetary policy can, and should, be implemented by providing a net remuneration of zero at any point in time to commercial banks holding reserves at the central bank. At its simplest, central banks could operate with two tiers, required reserves and excess reserves. Excess reserves – which can be loaned to money markets would be remunerated at an interest rate consistent with policy objectives – and the interest rate on required reserves would be calibrated so the net compensation is zero.
This is not only appropriate given banks are taking no credit risk and cannot control the stock of reserves. It helps in both directions: when central banks are targeting negative interest rates in money markets commercial banks are not penalised for holding a quantity of reserves determined by the central bank. Similarly there is no ‘gift’ to banks when interest rates are positive.
But the major benefit it seems is to clarify a dangerous but widespread error – public sector debt is not high, and central banks can do a lot more. Why does this follow? First, we can simply net off holdings of public sector debt by central banks, while being cognisant that this could reverse in future. Secondly, central banks can make transfers via dual rates and perpetual loans which permit them to meet their inflation targets without any impact on accounting equity or institutional transfers to the fiscal authorities.
Analytical confusions dissipates, central banks can do their jobs, and fiscal policy is liberated to focusing on much needed (and high returning) capital investment.
 See this crystal clear Eco Notepad from the Banque de France. “There is nothing magic in central bank money.”
 This paper by Robert Hall and Riccardo Reis is a good example, it even describes money printing through reserve creation as ‘borrowing from commercial banks”, by the same linguistic rule you would say central banks are ‘borrowing from the general public when then issue notes and coins’ – which doesn’t even make sense. This is a good example of the problems economics faces in using natural language to describe objective diverse phenomena. As occurs frequently in natural and unscientific languages, economists have a terrible habit of using the same terms to identifying distinct phenomena. It often starts as a metaphor – signalled by inverted commas – and morphs seamlessly into a (false) identity. Base money is the purest form means of payment, which is created at will by the central bank and defines monetary policy – no one does any borrowing or lending, as this short note will make clear.
 This UK economics commentary from Robert Peston illustrates the political risks associated with these category errors.