This blog posting is the first in a series on the 2016 Bank of England stress tests. A fuller report, “No Stress III: the Flaws in the Bank of England’s 2016 Stress Tests”, will be published later in the year by the Adam Smith Institute.
Early in January this year, ITN’s Joel Hills approached me about a feature on the stress tests that he was planning to do for News at Ten, and which was broadcast on January 10th. He was going to interview Sir John Vickers on the market values vs. book values issue and he asked me if I would provide the results that showed how using the latest available market values instead of book values would have affected the results of the Bank’s 2016 stress tests. Sir John and I had been arguing for some time that the Bank should pay more attention to market values, especially when they are lower than book values: as of early January 2017, market values were about 2/3 of book values.
The choice of book vs. market values makes a big difference to the results of the stress test: if you use book values in the Bank’s stress test, then only RBS fails the test, but if you replace book values by market values and make no other changes to the test, then only Lloyds passes. So book vs. market values is a big deal.
Why should we use market values rather than book values? The reason is that market values being less than book values signals that the markets do not believe the book values: the most likely explanation is that the markets believe that there are expected losses coming through that the book values are not picking up.
Vickers had put a similar point to Carney in a letter of December 5th last year:
… market-to-book ratios for some major UK banks are well below 1. That indicates market doubt about the accuracy of book measures. To the extent that such doubts are correct, stress tests based on book values are undermined.
The Bank appears to take the view that low market-to-book ratios are down to dimmed prospects of future profitability rather than problems with current asset books. But such a view is hard to sustain for banks with [price-to-book] ratios below 1. There is, at the very least, a serious possibility that low market-to-book ratios are signalling underlying problems with book values. This certainly cannot be dismissed, especially when one is examining the ability of the system to bear stress – an exercise that calls for prudence. 
To me this statement is self-evidently correct, so I was surprised that in his reply letter Governor Carney sought to challenge it: he continued to defend the Bank’s earlier position that low market-to-book is due to low future profitability and dismissed Vickers’ concerns about the possibility that markets might be signalling deeper issues with the book values.
I have to ask myself how the Bank of England can be so sure (and prudently so!) that its interpretation is correct and that Vickers’ is not.
Vickers’ March 3rd response to Carney’s dismissal of his analysis is unanswerable:
The regulation of banks is based on accounting measures of capital. A major source of risk to financial stability is that capital is mis-measured by the accounting standards used in regulation. In that case, bank regulation that allows high (e.g. 25 times) leverage relative to accounting (or ‘book’) measures of capital is more fragile than may appear.
An instance of this point is that stress tests based on book values are themselves vulnerable to erroneous measurement of capital, because those measurements are their starting point. Furthermore, bank regulation nowadays counts convertible debt instruments such as CoCos as akin to equity capital, but the conditions in which they convert to common equity (or are written down) are also dependent on accounting measures of capital. In short, a lot is riding on book values being reasonably accurate. …
None of this is to say that markets necessarily value assets accurately. Rather, the point is that low price-to-book ratios, especially when below one, signal a serious possibility that book values are inaccurate, and hence that the basis for regulation (not just in stress tests) is open to question.
Market values are not always reliable, but
when [market values] are low, systematic attention should be paid to them, and transparently so. (My italics)
More clutching at straws: further BoE arguments against market values
Let’s consider another objection made by the Bank against it using market values:
Low market valuations can reflect a number of things, all of which lead to weak expected profitability. But, crucially, different reasons for weak profitability can have quite different implications for a bank’s resilience. This is because they have different impacts on the value of the bank’s assets if it needed to sell them to pay for losses elsewhere in the business. 
The Bank then illustrated this point by comparing two hypothetical banks with the same cash flows – one is efficient but has poor assets, the other is inefficient but has good assets and could sell some if need be.
The Bank’s argument is a distinction without a difference, however. Weak expected profitability – whatever the cause – is a potentially serious financial stability issue and it is as basic as that. As Vickers pointed out in his April 26th letter to Alex Brazier:
A holder of the BoE view, if I may put it that way, can however respond by noting … that the inefficient bank with good assets can sell some. If such a bank alone faced difficulties – so in the absence of systemic stress – this would be a reasonable answer.
But it is harder to see how asset sales could be a satisfactory response in conditions of systemic stress, a typical feature of which is precisely the inability of banks to sell assets except at distressed prices. This is the well-known ‘fire sale’ problem …
The gist of this problem that a bank that suffers a large loss might be forced to reduce its asset holdings by selling assets at fire-sale prices. If other banks must revalue their assets at these temporarily low market values, then the first sale can set off a cascade of fire sales that inflicts losses on many institutions and thereby create asystemic problem.
This kind of risk, I suggest, should be central to thinking about financial stability, and to stress tests. Financial stability policy should take a prudent approach as a general matter. In particular, it should not place reliance on banks being able to sell assets in crises at good prices. While that might cope with an idiosyncratic shock affecting one bank, it will not do in a systemic crisis. But systemic crisis risk is the principal risk that regulation should guard against. The prudent stress test question, then, is whether the bank can meet its obligations without resorting to asset sales. It is not whether it can do so on the assumption that assets can be sold at good prices.
In sum, low market valuations imply less resilience even when the possibility of asset sales is allowed for. Tests of resilience that rely on resort to asset sales are flawed because, as experience shows, in a systemic crisis it may well be impossible to realise full value from asset sales.
Tim Bush also offers a powerful rebuttal:
Essentially, from the perspective of a shareholder providing capital, the Bank’s second example (good current balance sheet, poor future returns) is really an admission that a bank as a whole is one big impaired asset. Nothing resilient about that. Particularly, no incentive to refinance it if it incurs unexpected losses for example. New investment won't achieve an appropriate return.
The Bank’s line is a bit like saying British Leyland was resilient if the factories were brand new. 
Another objection to the use of market values was made by Alex Brazier in his evidence to the Treasury Committee on January 11th 2017:
…if you had [relied on market cap values] before the crisis, you would have been led completely astray … You would have been led to the conclusion that the British banking system was remarkably resilient, and, as forecasting errors go, that would have been quite a good one. 
Really? Consider this chart, which shows how the price-to-book (P2B) ratios of international banks fell the before crisis. The P2B ratios for UK banks are similar.
Then consider the next chart, which shows the ratios of market capitalisation to the book value of equity for two sets of international banks, the “crisis” ones that failed, required assistance or were taken over in distressed conditions, and the “non-crisis” ones that weathered the storm.
It is, thus, clear that markets were signalling problems with the banks and they correctly identified the weakest banks too. In the UK case, they also correctly identified in advance the two biggest UK problem banks, HBOS and RBS. 
Mr. Brazier omits to mention that the Bank was relying on Basel model-based book values that completely missed the impending meltdown and he does not offer any alternative that would have credibly worked better.
He also omits to mention the Bank’s own record on this issue. The ‘British banking system is resilient’ is exactly the message that the Bank itself was putting out before the Global Financial Crisis (GFC). Not only did the Bank itself have no inkling of the GFC before it hit, but in the early stages of the GFC and even after the run on Northern Rock, it was still reassuring us that there was little to worry about and that the UK banking system was more than adequately capitalised. These reassurances proved to be as wrong as wrong can be.
The charts above are evidence that market values did provide some warning and there is further evidence too. To quote the Bank’s own chief economist, Andy Haldane:
market-based measures of capital offered clear advance signals of impending distress. … Replacing the book value of capital with its market value lowers errors by a half, often much more. Market values provide both fewer false positives and more reliable advance warnings of future banking distress.
… market-based solvency metrics perform creditably against first principles: they appear to offer the potential for simple, timely and robust control of a complex financial web. 
It is also helpful to compare their respective track records at predicting subsequently realised bank failures: markets have sometimes got it right and sometimes got it wrong, but bank regulators have always got it wrong. Their failure prediction rate is exactly zero percent. Even chicken entrails would have had a better success rate than whatever model or crystal ball regulators anywhere use to peer into the future and no rational person would ever believe the forecasts of a group of forecasters with a zero percent success rate.
The former President and CEO of BB&T Bank, John Allison, confirms this point and explains why:
One observation in my 40-year career at BB&T: I don’t know a single time when federal regulators—primarily the FDIC—actually identified a significant bank failure in advance. Regulators are always the last ones to the party after everybody in the market (the other bankers) know something is going on. … regulators have a 100 percent failure rate. Indeed, in my experience, whenever they get involved with a bank that is struggling, they always make it worse—because they don’t know how to run a bank. 
But I digress.
So what it comes down to is that if the Bank does not use market values for the stress tests, then it should have a good reason not to. In terms of a concrete operating criterion, the natural answer is provided by the Principle of Prudence, which suggests that it should value using the lesser of book values and market values – and central bankers are famously prudent.
Whilst on the subject of prudence, wouldn’t it be wise for the Bank to acknowledge at least the possibility that outsiders – not just Vickers and I, but also Anat Admati, Tim Bush, James Ferguson and Gordon Kerr, to name a few, and even Mervyn King, who have pointedly failed to endorse the stress tests – might be right or that we might at least have a point?
So answer me this, Bank of England: you say that your stress tests show that the UK banking system is sound. But how can you be confident in such assertions, when your stress tests are based on book-value numbers and when the markets are clearly signalling that something is wrong with those book values?
To cut to the chase, how can you expect the public to believe your narrative when the markets don’t?
The Vickers proposal for parallel market and book value tests
So let me endorse Sir John’s suggestion for a compromise as set out in his December 5th letter. The Bank should present both sets of results and let readers make up their own minds. As he wrote:
[My] proposal is not that market-based tests for such banks should replace tests of the kind that the Bank has run. The request is merely that the Bank supplements its results with market-based results.
That would inform public debate on a matter of great importance for economic policy, and it would enhance the transparency and accountability of the Bank.
Yet the Bank still insists that it should not publish any such results because – to quote Governor Carney in his December 19th reply to Vickers’ letter – to do so might confuse
the Bank’s communication around its stress tests. If we publish two sets of results that give different messages, people might struggle to understand what we are trying to say about the resilience of the banking system.
But as Vickers responded:
A stress test is primarily a test of the resilience of the banks, not a communications exercise. …. Considerations of transparency and accountability should therefore far outweigh the regulator’s communications agenda. 
A related problem is that Dr. Carney takes the Bank’s credibility for granted and then focusses on making the message simple for the audience. Such reasoning puts the cart before the horse. Instead, the key to effective communication is credibility and credibility must be earned and maintained, not presumed.
The Bank does not help its own credibility by brushing aside good outside advice, however politely. Publishing market-based results could allay any possible concerns that it might be trying to window-dress the banking system and itself in the best possible light. The Bank would still be able to give its own commentary explaining why it thinks that the book-value results are more credible than the market-value results.
It is also a mistake for the Bank to under-estimate its intended audience, who should be presumed to be capable of making up their own minds when presented with the evidence and should be treated with appropriate respect.
The Bank repeatedly makes the mistake of ‘oversimplifying’ its message and then making claims that turn out later to have been way off the mark, thereby undermining its own credibility again and again. It made that mistake when it reassured us before the financial crisis that the banking system was strong. It made that mistake when it told us during the Brexit referendum campaign that a Leave vote could trigger a recession and that Brexit was the biggest single risk facing the UK economy, and it is making the same mistake again with the stress tests.
To paraphrase Hubert Humphrey on propaganda, a perhaps not entirely unrelated subject: the Bank of England message, to be effective, must be believed. To be believed, it must be credible. At the moment, it is not.
 “Supplementary market-based stress test results,” letter from Sir John Vickers to Governor Mark Carney, December 5th 2016.
 Sir John Vickers, “Response to the Treasury Select Committee’s Capital Inquiry: Recovery and Resolution,” March 3rd 2017, pp. 7, 8 and 12.
 Quoted from Sir John Vickers’ letter to Alex Brazier, April 26th 2017, copies of which are available on request from Sir John.
 Personal correspondence.
 Treasury Committee “Oral evidence: Bank of England Financial Stability Reports,” HC 549, Wednesday 11 January 2017, answer to Q173.
 See, e.g., Chart 2.73 on p. 153 of the FCA/PRA report The Failure of HBOS plc.
 A. G. Haldane, “Capital discipline,” speech given at the American Economic Association, Denver, January 9th 2011, p. 8.
 J. Allison, “Market discipline beats regulatory discipline,” Cato Journal, 24(2), Spring-Summer 2014, p. 345.
 Quoted in Vickers Capital Enquiry testimony.