A sceptic's view of QE, an initially irrational policy, the controversies over which were whitewashed
I am, and always have been, dubious about the usefulness of QE, and regret the process by which the Bank and the MPC decided to embark on it, and gave overly confident accounts of how it worked and its impact. This post elaborates, putting the sceptical side of the QE story that the MPC has sidestepped. None of these sceptical arguments are original, and all have been made before by people higher up the food chain than I. I don't go through the arguments in favour in great detail, as this side of the story has already been told.
There are two main pieces of evidence in favour of QE having been the right . One is the accumulated 'event study' analysis of the effects of tranches of QE on gilt yields, not just in the UK, but also in the US and Japan. Event studies measure yields just before a QE announcement is made, and then just after. By timing the measurements tightly around the announcements, they try to isolate the effects of QE from all the other market news on yields. These effects are extensively documented in BoE work, and also good work by the New York Fed and the BoJ. The other piece of evidence is more circumstancial, but powerful nonetheless - we have had high and relatively stable inflation for several years despite interest rates hitting their natural floor.
However.
1. It is possible, even likely, that this effect on yields was not persistent. Temporary effects on yields could be explained by changes in beliefs by particpants in gilt markets about resting places for prices and premia; and changes in what they inferred from the QE operations about the likely path of future rates. Other commentators have referred to this as a 'signalling' channel; QE simply signalled the authorities determination to keep rates low, or revealed private information the authorities had about just how bad the economy was (and therefore how unlikely interest rates were to rise). The temporariness of the effect on yields would not itself be a cast iron argument for not doing QE; there are circumstances where you would proceed anyway, expanding QE by ever more. (Perhaps this is behind the Fed's constant rate of asset purchases policy, rather than MPC's constant level policy). However, if QE is really about signalling the path of future interest rates, it raises the possibility that what the MPC should have done was, well, explain their path for future interest rates, rather than do it so indirectly. Signalling directly would be i) less confusing, ii) lead to less uncertainty about future rates, a good thing in itself, iii) help with any credibility issues, since the progress of actual rates can be monitored. Another explanation for the temporary effect on yields is that for a while people thought 'perhaps this will work', but then later realised it would not. This is one explanation for why the effect seems stronger for earlier QE announcements than later ones, (an explanation carefully avoided in the BoE studies), though there are others, more supportive of QE, namely, that the early announcements were more of a surprise, while the later ones were largely expected. (That they were expected doesn't mean they had any less of an effect, as if QE had not been implemented as expected, yields could have been higher).
2. Even if QE did affect yields, it doesn't follow that it lowered the rates anyone in the private sector pays to borrow, other things equal. Vissing-Jorgensen and Krishnamurthy dissect the effects of QE and argues that purchases of US Treasuries simply widened the spread between those and other riskier assets. On the contrary, purchases of the debt of the mortgage backing agencies did lower the yields on those assets. Michael Woodford makes this point too in his Jackson Hole paper.
3. Even if QE did have an effect on bond yields, and even if that was passed on into yields on other assets, it doesn't follow from this that it was socially beneficial. The MPC have many times told a story that runs something like this; we make gilts more scarce, this pushes previous holders of gilts to hold something else more risky, like equities or coporate bonds. The prices of those assets rise, and that lowers the cost of funding for corporates. This boosts investment spending, employment, output, and everyone is better off. So it might. But, it also deprives the holders of gilts of something they wanted. If the government were to buy up the stock of oranges, orange-lovers would move into grapefruit, and the price of grapefruit would rise, and grapefruit growers would enjoy a boom, but orange lovers would be poorer for it, no longer able to enjoy that sweet taste so cheaply. Some of those desparate for citrus of any kind would buy grapefruit and eat them, and we would see their pain and disappointment written into their sour faces. This point is made forcefully (without the silly fruit analogy) in work by Krishnamurthy and Vissing-Jorgensen. The logical corollary of this argument is that we need not QE but negative QE, with the proceeds invested in a risky asset. We need to expand the supply of safe assets, by over-financing government borrowing needs, and take risk, notwithstanding worries about public sector credit-worthiness, onto the government balance sheet. Pressurising those in the private sector already struggling to bear risk to bear more of it makes them worse off. This said, the mere existence of these costs doesn't imply that they dominate potential benefits of QE. But they might.
4. A sceptical argument that I anticipate will alienate some readers, and must nonplus BoE policymakers by their revealed preference for QE, is that there is no sound theoretical account of why QE should work. By sound, I mean one that meets the standards of modern macroeconomic theory, which places a premium on assuming the minimum, making explicit what agents are doing in the economy and why, and works out the effects by adding up the behaviours of all the agents in the economy. On the contrary, all the models of monetary policy that we do have suggest that QE should be impotent. Michael Woodford explains this point, in his Jackson Hole paper, and in several before that. These models are probably wrong too, and not to be taken literally. But they at least provide a cautionary note. Non-economists might find it amusing to hear the old joke about economists (that they are people who see whether something that works in pratise also works in theory) turning out to be true. Isn't it specious to insist on theory when you have empirical and common sense insights on your side? Robert Lucas got his nobel prize, in part, for his explanation for why this is important. If you rely solely on data-based insights, and stop trying to understand what the people in the model you have are doing, you might find the correlations they embody crumbling when you try to exploit them. Also, if you fail to try to formalise your stories, you run the risk of being fooled by a story that sounds good but is just internally inconsistent or wrong. In defence of this philosophy, many MPC prounouncements in the past have followed this kind of reasoning, and the same notion lies behind the construction of the Bank's new model COMPASS.
5. On the contrary, the monetary models we do have suggest that QE does not work. What happens is this: at the zero bound, swapping money for bonds is swapping one default-risk-free, zero-interest-rate asset for another, and not to be expected to have any effect. (Using the fruit analogy above, QE is swapping oranges for oranges !) Money's usefulness derives from it offering a cheap, quick and easy way to transform it into something you can eat or use. But if you have enough of it on hand anyway, more money isn't going to be useful. You have enough to see you through. Hence money is just like a gilt at this point. It's just another safe way to store your wealth. Why should swapping something almost like a gilt for a gilt change anything? Robert Lucas, Godfather of modern macro, and Nobel laureate, explained the very same argument in a recent WSJ article. [He's in favour of the Fed QE program, but because they are buying illiquid agency bonds]. In these models, QE will work to the extent that markets think the operation will not be reversed and that that will impy lower interest rates than would be expected given likely rules for setting interest rates. (Because rates would have to be higher, in order for markets to want the extra money). This is why some think that the empirical evidence showing QE does something indicates a 'signalling' effect. And why if this was the motive, explaining precisely what the implication was for future interest rates would have been a better policy. Perhaps we will one day find a sound theoretical explanation for why QE works. But if we do, we can't guarantee even that it will tell us even that we should do more of it in a recession, not less. And that same theory might overturn all that policymakers use to inform the setting of their conventional interest rate instrument, or even that interest rates should be the normal policy insrument.
6. On many occasions, MPC emphasised the importance of expanding the money supply, precisely the mechanism that the models we have says is ineffective. The first mention is in the March 2009 MPC minutes, where we read: 'the Committee had previously chosen to influence the amount of nominal spending in the economy by varying the price at which it supplied central bank money in exchange for assets held by the private sector.... [now]...the Committee would instead be focusing more directly on the quantity of money it supplied in exchange for assets held by the private sector... By increasing the supply of money in the ec onomy, these operations should, over time, cause nominal spending to rise.' [Ref 1 linked to below]. That's ok then: the message is: monetary policy is simply about varying the supply of money, which before we did by changing the price, now we do by changing the quantity when the price is pinned at zero. The argument appears in later documents too. For example: 'The purpose of the purchases was and is to inject money directly into the economy in order to boost nominal demand', [Ref 2]. 'Increases in money should eventually lead to a rise in prices', [Ref 3], which states the fallacy more explicitly, neglecting to note that this relationship is probably true, except at the zero lower bound to interest rates. What monetary economists would have found astonishing is that nowhere in these early deliberations and communications is there an explanation of, or even an attempt to evaluate and discard the account of what happens at the zero bound in monetary macroeconomic models. The MPC might have worried that by disclosing that an unpleasant feature of its own monetary model, (in fact all monetary models), a feature that implied that QE would not work, this would make it less likely that QE would work, panicking the markets. But there was a contending risk, that MPC would look like it did not understand monetary economics. Whatever, MPC seem to have got away with it.
7. The Bank's presenation of the evidence is accompanied by verbal accounts of why and how QE should work which are inappropriately confident, and no airtime is given to more sceptical, contrary thinking. The Bank's empirical work forms an impressive body. But in some senses it is a whitewash; it gives inadequate attention to sceptical readings of the evidence, and presents verbal accounts of the intuition for why these effects should be taken at face value [ie as persistent, and to be regarded as beneficial] with a confidence that is inappropriate (given the practice at large in the profession of not doing this unless there is consensus and or confirmatory theoretical accounts of the same effects seen in the data). Imaginative conjecture is recounted as conclusive, almost formal analysis. To give some examples: 'Purchases of financial assets financed by central bank money should initially increase broad money holdings, push up asset prices and stimulate expenditure by lowering borrowing costs and increasing wealth. Asset purchases may also have a stimulatory impact through their broader effects on expectations and by influencing bank lending, though this channel would not be expected to be material during times of financial crisis.' [Ref: 2] (It sounds good, and makes you think we know the answer, but we don't. We can't even check that it holds water theoretically.)
This unfortunate policy [of sweeping aside worries about QE's effectiveness] can even be found in the official response to the request to investigate the 'distributional effects of asset purchases', in particular claims that (eg) pensioners were being adversely affected. That response did not appraise the reader that the assertion that QE had worked was highly contentious. An alternative reading of QE is that it might well have lowered yields on government bonds on which many savers live, and entirely in vain, for it may not have had any effect on anything else. No weight is put on that argument in the Bank's official response, nor on the argument that there is even a case, as explained above, for engaging in negative QE. Thus, in the introduction to 'The distributional effects of asset purchases' we read : 'Without the Bank’s asset purchases, most people in the United Kingdom would have been worse off. Economic growth would have been lower. Unemployment would have been higher. Many more companies would have gone out of business.' Well, maybe, but maybe not! Perhaps we are about as well off as if they didn't happen?!
It's not true that uncerainty is ignored entirely in BoE material. For example, one can find several references like this in the Bank's survey of the evidence: 'It is difficult to measure directly the effects of policy measures such as QE and so estimates of those effects are highly uncertain' [Ref 4]. But, the overriding impression we are directed to, not least since the MPC's QE (and QE only) actions (and MPC minutes) speak just as loudly, is that it works, and so much better than any other option, that other policies are not even worth considering. This quote is illustrative: 'evidence suggests that the policy had economically significant effects - equivalent to a cut of 150 to 300 basis points in Bank Rate - but there is considerable uncertainty around the precise magnitude of the impact.' Which I read to mean; ok, the precise magnitude is not certain, but we are pretty sure it worked, big-time!
This failure to provide a more balanced account of QE might have been motivated by a desire to instill confidence. But ultimately, it undermines the authority with which the Bank's voice speaks on technical matters, and will increase the chance that future such documents are not taken at face value, but as energetic and adversarial attempts to justify a particular course of action.
8. QE was not a rational policy at the outset, when the (for some) supportive event study analysis was not yet available. Suppose you adopt the position that history has proved the MPC right, and QE effective. And you dismiss all the contrary arguments set out above. Still, at the time the MPC first embarked on it, we did not have the supportive event study analysis that QE might lower yields. [The introduction of the corporate paper facility can be set aside as a detail, since it was never intended, and never evolved into a way to intervene on a large scale]. The examples we did have were the old evidence on operation twist in the 60s, which at that time showed very little effect of changing maturities in the hands of the private secor [effects revised up since in work by Swanson], and the evidence of Japan, which was singularly unencouraging. There were other possible reasons for the failure in Japan, but at least one possibility was that QE had little or no effect. The only thing in favour of QE was the belief that although our monetary models said that the power of money injections to generate inflation was undermined by the zero bound, there was something different about the real world which meant that monetary injections still had potency. So, faced with many possible options at the zero boud, the MPC chose to place almost all its eggs in an extremely unpromising basket. One can argue that empirical evidence has proven them right, but that does not rationalise the decision made initially, before this evidence existed.
9. Other more promising alternatives were inappropriately discarded. There were two other promising alternatives to QE at the time: forward guidance on interest rates, and credit easing, or large scale purchases of private assets. Both were discarded. In the March 2009 minutes, I can't see any discussion of forward guidance at all, though from Spencer Dale's speech later that year, we can guess at what the consensus view might have been. As for private sector assets, there is some discussion of this, but the decision about what to buy is described as secondary to the decision about how much to buy (something that is a corrollary of believing that it's not the nature of the assets the central bank takes on its balance sheet that matters, but the quantity of monetary liabilities it creates, something I view as a mistake, as explained earlier).
10. The BoE executive railroaded the external MPC members over who does what, based on the faulty logic that only the size of the balance sheet mattered. Judging from the March 2009 minutes, and subsequent votes on asset purchases, it's clear that the practice was settled on whereby the MPC decide on the quantum of QE, and the executive decide what to buy with the proceeds. The MPC appear to have submitted to this when there was no unambiuous, legal reason in the Bank of England Act for them to do so. It was also illogical, as though the quantum was all that needed to be decided in order to best hit the MPCs remit, and the asset purchases themselves were a technical detail best left to market experts. This reasoning is unsound. Not only because the quantity of monetary liabilities created might well have been irrelevant. But also because purchases of private sector assets could have substantial beneficial supply side effects, boosting credit supply, supply itself, and perhaps even lowering inflation. [The opposite perhaps of the constriction to credit that constituted the financial crisis the MPC were trying to combat]. The matter of whether one embarks on a trade-off improving policy intervention (credit easing) as against a trade-off preserving one would seem to be exactly what the MPC should be deciding on collectively, and not an implementation matter for the exeuctive of the BoE only. I recall that this operating procedure (as well as the focus on gilts) was questioned in public by Adam Posen, but others appear to have gone along with it. It's a shame that HMT's review of the MPCs remit, published in March this year, did not take the chance to codify these responsiblities properly.
Ref 1: MPC minutes, 4,5 March 2009, paras 30/31.
Ref 2: Quantitative easing explained
Ref 3: 'The UK's quantitative easing policy: design, operation and impact', p201
Ref 4: 'The distributional effects of asset purchases', p1.
Ref 5: Same as Ref 3, but p204.