In my last post, I pointed out that QE does not work when the transmission mechanism for monetary policy is impaired because of a damaged and risk-averse financial sector. This caused some confusion among those who think that throwing money at banks automatically makes them lend, so I attempted to explain it on twitter. Predictably, I ended up in an extended discussion first with David Beckworth and then with Andrew Lilico, in the course of which it became clear - to me, at any rate - that not only does QE fail when damaged banks aren't lending normally, but it actually impairs the transmission mechanism itself. This might explain why QE seems to become less effective the more of it you do. It's like hard water. It gradually clogs up its own pipes.
To explain this, let me first go through the money creation process in our fiat money system and the ways in which QE influences that process.
The monetary base, M0, is created by the central bank. It consists of notes & coins, and bank reserves - the money that banks use to settle payments. M0 makes up maybe 10-15% of the total money in circulation. The rest - "broad money" - is created in the course of lending both by banks and by non-banks that do bank-like things. It should be remembered that non-banks are the customers of banks: cash held by non-banks always finds its way into banks.
Bank reserves never leave the banking system. They are not "lent out", as is often claimed. When a bank lends, it creates a deposit "from nothing", which is placed in the customer's demand deposit account. When that loan is drawn down, the bank must obtain reserves to settle that payment - but the payment simply goes to another bank (or even the same one), either directly through an interbank settlement process, or indirectly via cash withdrawal and subsequent deposit. The total amount of reserves in the system DOES NOT CHANGE as a consequence of bank lending. Only the central bank can change the total amount of reserves in the system. This is usually done by means of "open market operations" - buying and selling securities in return for cash.
Because banks create deposits from nothing when they lend, the availability of reserves at the time of loan creation is not a constraint on lending. When the financial system is functioning normally, banks borrow reserves from each other to settle payment requests, and if there is a shortage of reserves in the system the central bank will create more: alternatively if there are more reserves in the system than are needed to settle payments, the central bank would normally drain them by selling securities back into the market. Some central banks impose a "reserve requirement" of, say, 10% of eligible deposits: this is a liquidity buffer to ensure that banks can meet most payment requests without having to borrow from each other or from the central bank. If there is a positive reserve requirement, it is the central bank's responsibility to ensure that there are sufficient reserves in the system to enable all banks to meet the reserve requirement.
It should be apparent from this that the monetary base RESPONDS TO lending demand. It does not drive it. This is borne out by evidence that if anything, M0 creation lags broad money creation.
That's the basic mechanics of the system. However, it is not quite that simple. Monetary policy may influence lending demand by means of reserve adjustments. If the central bank decides to reduce the total amount of reserves in the system, scarcity of reserves pushes up the interest rate at which banks will lend to each other: conversely, increasing the amount of reserves pushes down the lending rate. The increased cost of reserves is supposed to act as a brake on lending. Unfortunately, when banks are chasing market share instead of margin - as they were in the early to mid-2000s - increasing the cost of reserves is not a particularly effective brake on lending unless the increase is very large. Increasing market share can compensate for loss of margin to quite an extent (this is why low-margin retail lending is only really viable as a high-volume business). And in the end, no central bank is going to allow payments to fail because of scarcity of reserves. Indeed in Europe, reserve creation by the Eurosystem to facilitate payments is automated.
QE can be regarded to an extent as large-scale open market operations. The central bank buys securities in return for newly-created cash. If the securities are bought directly from banks, then the banks simply replace riskier and less liquid assets with cash. If the securities are bought from institutional investors or individuals, the cash still ends up in banks in the form of deposits. Either way, though, the total amount of reserves in the system increases.
The UK, US and Japan have all done extensive QE, and as a consequence the banking system is now awash with US dollar, sterling and yen reserves far in excess of the amount needed to settle payments. Increasing the amount of reserves in the system was supposed to encourage banks to lend. But the financial sector is badly damaged in all three countries: bank balance sheets are full of non-performing loans that tie up capital and are not easy to unwind. Additionally, there is regulatory pressure on banks to reduce their risks and shrink their balance sheets. Shrinking a bank balance sheet means selling or unwinding unwanted loan portfolios. If a lot of banks are doing this all at once, the effect must be a reduction in overall lending volume. Remember I said that money is created when banks lend? When loans are paid off - or written off - money is destroyed. So general deleveraging in the banking sector, as we have been seeing for the last five years in the US and UK and for the last fifteen years in Japan, means that broad money supply is likely to be stagnant or actually falling. Now, it may be that QE encourages some banks to maintain higher levels of lending than they would otherwise have done, because in theory it reduces their funding costs (though that may not actually be true in practice, as I shall discuss shortly). We simply don't know. But what is clear is that the size of the monetary base has nothing whatsoever to do with broad money supply. When banks are deleveraging, broad money may still fall even when the monetary base is increasing.
To be fair, the Bank of England and the Fed both noted that damaged banks were not likely to increase lending and QE should achieve its effects mainly in other ways. But what they both missed was the damaging effects of QE on the flow of money through the financial system and the consequences for monetary policy transmission.
QE increases the amount of reserves in the system and reduces the amount of other forms of safe security, particularly various forms of government debt. Since the financial crisis, borrowing and lending between banks and non-banks has become more-or-less completely collateralised, with the debt of highly-rated sovereigns being the preferred choice of collateral. There is also a scarcity of collateral due to collapse of US MBS issuance, increasing shortages of high-quality sovereign debt due to sovereign downgrades, and regulatory changes encouraging buildup of safe asset reserves and hoarding of collateral. One of QE's effects is to reduce even further availability of safe collateral and therefore increase its price. THIS IS DELIBERATE. The stated intention of QE is to depress government bond yields to make them less attractive to investors and therefore nudge those investors towards riskier assets. Unfortunately this also increases the cost of the collateral needed by banks and non-banks to obtain funding, including from central banks. It could be argued that whatever encouragement increased reserves give to banks to lend is offset by the increasing cost and scarcity of the collateral needed to obtain the funds to settle lending. It's a wash.
Which brings me to the problems with monetary policy. The first thing to note is that as the increased availability of funds is balanced by increased cost and scarcity of collateral needed to obtain funds, QE makes no difference to liquidity in the financial system. This point has been brilliantly (and repeatedly) made by Peter Stella and the IMF's Manmohan Singh, but it seems no-one is listening. The extra reserves provided by QE are in no sense expansionary. If anything, QE is contractionary, because it reduces the velocity of money in the financial system. When collateral is scarce, funding flows are impeded. There may be more actual funds available, but if they aren't moving, they aren't any use.
The second point concerns the means by which central banks influence the behaviour of banks. Because the fundamental driver of lending is the risk versus return profile for both lenders and borrowers, lending is intrinsically cyclical. Central banks attempt to dampen the cyclicality of lending by means of macroprudential regulation and monetary policy. Of these, the second is arguably more important: macroprudential regulation historically has had limited success. But large-scale QE fundamentally changes the way monetary policy is transmitted. When the system is awash with excess reserves, central banks cannot use reserve scarcity to drive up the cost of funding. The "funds" rate (Fed Funds in the US, "bank" rate in the UK) becomes useless as a policy measure. When reserves are excessive, therefore, policy must be transmitted via the deposit rate.
Most central banks pay interest on excess reserves placed with them by banks: normally that rate is some distance below the interbank lending rate, to encourage banks to lend excess reserves to each other instead of parking them at the central bank. But as funding rates crash to zero, the deposit rate suddenly becomes far more important. Positive interest on excess reserves is currently used by both the US and the UK to prevent repo rates turning negative and prop up the short end of the Treasury yield curve. But this means that parking funds at the central bank becomes an increasingly attractive proposition for damaged banks that don't want to lend. To "nudge" banks towards productive lending, policy makers are now thinking about cutting central bank deposit rates to zero or even below. The consequences of negative rates are not fully understood, but even the brief look that I had a while ago suggested that they might not be quite what policy makers anticipate. And the problem with relying on deposit rates as the primary means of monetary policy transmission is that they are not underpinned by coherent macroeconomic modelling and their effects are not well understood. Some might argue that this is true of funds rates too, but it's far worse with deposit rates because they have never been used in this way before. Policymakers are making it up as they go along. And the economics profession is not exactly helping. The extent of disagreement among economists about how monetary policy works under these exceptional circumstances is eye-watering. It is telling that many of the most useful contributions to the debate about how best to conduct monetary policy at present have come from the financial blogosphere, not from the economics profession.
In short, monetary policy transmission is weirdly distorted by the effects of excess reserves and it has become extremely difficult for central banks to influence bank behaviour. Because monetary policy is hampered and there is reluctance to use fiscal policy as a complementary toolset, some politicians have been looking to macroprudential regulation to nudge banks towards more productive lending. This is madness. It is not the job of prudential regulators to repair damaged economies by, for example, watering down bank capital requirements intended to reduce the likelihood of catastrophic bank failures. I would rather see acceptance that, as Pozsar and McCulley (among others) have suggested, when interest rates are very low and monetary policy transmission is impeded there is a need for complementary fiscal policies.
There is much that I haven't covered in this post, and even what I have described here is controversial because it rests on an unconventional view of how the monetary system works (although perhaps not that unconventional now, since it is consistent with recent papers by both the Fed and BIS). You may disagree with much of what I have written, and I welcome constructive comments. A huge topic that I have not yet discussed is the whole question of expectations management, not only in relation to QE and its effects but in the transmission of monetary policy. I shall return to this in another post. In the meantime, Woodford is well worth reading on this matter.
Inflation, deflation and QE - Coppola Comment
Does the Federal Reserve fully control the money supply? - John Aziz (The Week) (with very cool charts!)
The whole question of collateral shortage, monetary policy transmission and (potentially) negative rates has been extensively covered in the financial blogosphere. Here are some of the best posts.
When safe assets return - FT Alphaville
The decline of safe assets - FT Alphaville
Pledged collateral in an IS/LM framework (part 1) - Manmohan Singh at FT Alphaville
Pledged collateral in an IS/LM framework (part 2) - Manmohan Singh at FT Alphaville
A confederacy of dorks - Interfluidity
(and all the links in this post, plus Interfluidity's previous posts on the "floor" system. A fascinating and important debate)
The roving cavaliers of credit - Steve Keen
When governments become banks - Coppola Comment
The strange world of negative interest rates - Coppola Comment
Central bank reserve creation in the era of negative money multipliers - Stella & Singh (Vox)
The recommendation of this post, as others I have written, is for fiscal & monetary coordination as suggested in this paper:
Helicopter money - Pozsar & McCulley
Implicit in this post is the idea (as spelled out in the Vox link above) that the traditional "money multiplier" does not exist. This was recently confirmed in these papers from the Fed and BIS:
Money, reserves and the transmission of monetary policy - Federal Reserve
The bank lending channel revisited - Disyatat, BIS
This paper from the Hungarian Central Bank documents the lack of connection between size of central bank balance sheet (i.e. reserve expansion) and broad money growth:
The effect of the monetary base on money supply - Andras Komaromi, MNB
For a good description of how endogenous money creation works in a fiat money system, read Cullen Roche's paper:
Understanding the modern monetary system - Roche
This paper has a US focus and it is likely that things work slightly differently in other countries: for example, the European banking model relies much less on disintermediated "shadow" banking. Perhaps we need more papers describing how money creation works in the UK, Europe and Japan.
On expectations, a subject that I haven't addressed at all yet but discussed extensively with David Beckworth - here's Woodford's paper on the importance of forward guidance:
Methods of policy accomodation at the interest-rate lower bound - Woodford