Speech by Adair Turner, Chairman, FSA:

In an article in the Financial Times on 15 December 2009, Martin Taylor, former Chief Executive of Barclays, argued that the core cause of the financial crisis was that ‘the system was brought down because bankers could not count’ and ‘because there was no measure of cash flow to tell them that they were idiots’ in their assessments of income.

The article was reported as an attack on innumerate bankers. But in fact it was an attack on accounting standards and on the way in which they are applied. And, indeed, there have always been bankers concerned about existing accounting approaches. Ahead of the crisis, there were both some bankers and some boards of banks worried about how low their commercial loan loss provisions had to be to comply with accounting standards. They were low because few commercial customers were behind with payments, so that there were very few observable facts to suggest potential loan impairment. But after a big boom and with warning signs beginning to flash across the world economy, judgement suggested that future loan losses might well be higher. But the accounting standards are designed to reflect today’s already observable facts – and to limit the role of judgement as to future possible events.

Whether that should be the case is now subject to intense debate, with two very different points of view.

• Among bank prudential regulators and central banks there is a belief that existing bank accounting standards were among the factors contributing to the crisis, inducing procyclicality in credit provision and pricing. And there is a demand that bank accounting standards must reflect the concerns of prudential regulators. There is a belief that banks are different, and that accounting standards need to recognise this.
• Among many securities analysts and investors, however, and among some accounting standards setters, there is a belief that accounts are for investors and not for regulators, that they must tell the ‘truth’ as it exists at one particular point in time, and that any influence of prudential regulators on bank accounting standards could be a Trojan horse for a wider politicisation.

This tension exists even within the regulatory community. Around the table of the international Financial Stability Board, the prudential regulators and central banks are the most convinced that banks are different and that the accounting standards setters must listen to us. The pure securities regulators, conversely, tend to be more sympathetic to the ‘accounts are for investors’ philosophy. And indeed the tension exists within the accounting standards-setting bodies, complicating any progress towards the convergence of international accounting standards. The International Accounting Standards Board (IASB), under David Tweedie’s leadership, has been sympathetic to the idea that it must be involved in close dialogue with the prudential regulators. The Financial Accounting Standards Board (FASB) has been more wedded to the ‘accounts are for investors only’ philosophy, and to the philosophy that banks, in their accounting, should be treated no differently from anybody else.

So are banks different in ways relevant to accounting standards? In what ways, and what should we do about it?

Obviously in one crucial way banks are different from – say – retailers, hoteliers, manufacturers, mining companies, airlines, and we have seen that difference illustrated dramatically over the last two years. Bank system failures – the simultaneous failure of many banks – can topple the whole world economy into recession. Bank failures can be a key cause of economic recession rather than, as with the failure of companies in other sectors, a consequence. That is for two reasons, the first general to all financial markets and to all financial firms which act as principals rather than intermediaries; the second specific to banks and bank-like institutions as a sub-set of the financial services industry.

Financial instruments link the present to the future; they have value in markets which are inherently intertemporal. Bananas are worth what they sell for today: their market clears today, balancing today’s suppliers and today’s buyers. But a loan or an equity contract has value today determined by events still to happen in the future and by a changing set of other opportunities to trade future for present value via other financial instruments. That complexity means that market imperfection problems are more severe in the market for financial contracts than in other markets. There are far greater asymmetries of information between buyers and sellers, and a greater influence of imperfect principal/agent relationships. And it means that perceptions of value can be highly volatile. Indeed, it introduces not just risk but, in the terms in which Frank Knight and Keynes¹ used it, inherent uncertainty as to the value of, say, an equity claim on the future. At any instant, there may be a price at which an equity will sell in small quantity, but that price can change radically and rapidly in the face of self-fulfilling changes in perception. And it will be a quite different price if many people simultaneously decide to sell, influenced by one another.

This inherent uncertainty over how to value long-term and contingent assets and liabilities introduces specific problems and complexities into the regulation and the accounting of all financial institutions, for insurance companies as much as banks. But banks are also crucially different even from other financial institutions, in both their riskiness and their importance, in two key respects.

• First because of the maturity transformation function they perform, lending at longer term than their liabilities. It is an important function of considerable social value: it results in a term structure of interest rates more favourable to long-term investment than would otherwise pertain; and it enables householders to enjoy the financial planning benefits and assurance of owing long term liabilities (typically mortgages) but owning short term assets (deposits). But it is an extremely risky activity, and one where the risks are inherently collective rather than individual.²
• Second, the fact that banks extend credit to the real economy, and that volatility in the extension of credit to the real economy (in both quantity and price) is one of the most crucial drivers of macroeconomic volatility. Banks are different because swings in credit supply – irrationally exuberant supply and over capacity in the upswing and then sudden withdrawal in the downswing – matter far more to the macroeconomy than irrationally exuberant over supply and contraction in, say, airline capacity or internet entrepreneurship.

It is therefore a combination and interaction of three factors which make banks different:

• Inherent uncertainty of the present value of future financial claims.
• The maturity transformation function.
• And the fundamental role which credit extension plays as a driver of macroeconomic volatility.

Insurance companies have the first of these features, but not all three. They have complex and difficult to interpret accounts: they need to be regulated for customer protection reasons, but the collective action of insurance companies together is not a key driver of macroeconomic instability.

For these reasons we have prudential regulation of financial firms, but of banks in particular. For these reasons prudential regulation of banks (unlike that of insurance companies) requires a close coordination with the ‘lender of last resort’ functions of the central banks, and in the future will need to entail a joint central bank/prudential regulator approach to macro-prudential through-the-cycle regulation focused not just on the vulnerability on individual institutions, but on the total system.³

This need for a systemic macro-prudential approach lies behind the regulatory policies now being developed by the international Financial Stability Board and the Basel Committee.

These include:

• Higher capital and liquidity requirements overall, designed not merely to make individual institutions sound, but to improve the resilience of the overall system.
• And, crucially, counter-cyclical capital requirements, building up capital buffers in good years so that they are available to absorb losses during recessions.

Such policies are quite different from those which we apply in other sectors of the economy. In non-financial sectors we do not have prudential regulation at all: and in insurance, we have prudential regulation focused on the sustainability of the individual institution, but not on the macroeconomic impact of all insurance companies together.

The key question therefore is not whether banks are different. They are and we are already designing prudential regulations better to reflect that fact. Instead the question for today is whether this difference has implications for bank accounting or whether it should be reflected solely in prudential regulation.

Two aspects of bank accounting in particular could be relevant to the macro-prudential and macroeconomic concerns which make banks different.

• First, the treatment of loan losses within the banking book, the way in which we capture, or fail to capture, present or future potential loan losses arising from credit default.
• Second, the valuation approach in the trading book (and other items which are marked to market), the recognition of unrealised gains or losses in general, but in particular in more illiquid securities.

In both these areas, there is a strong case that the present accounting treatment contributes to the problem of procyclicality.

On the banking book side, the current IASB accounting treatment requires banks to recognise the implications for potential loan losses of events which have already occurred, such as failures to make interest or principal payments; but also requires them only to recognise such known events, not to anticipate possible or probable future events. This necessarily implies that loan loss provisions will vary dramatically through the economic cycle, and means that in good years income will be declared which does not reflect the average future loan losses likely to arise from loans being put on the books.

As a result, this accounting treatment can contribute to a cycle of self-reinforcing responses which tends to exacerbate the volatility of credit extension and of the economic cycle, both on the way up and the way down.

In the upswing, (Exhibit 1 – see the slides on the right-hand side of this web page) the incurred loss model produces low figures for loan loss, boosting measured bank profits and flattering lending margins through three transmission mechanisms that can stimulate the procyclical effect of an increased quantity and lower price of credit:

• high bank profits bolstering capital ratios and thus increasing lending capacity;
• high apparent margins on lending generating competition for lending volume, and declining loan spreads; and
• high banker bonuses, paid out on the basis of high measured profits, reinforcing banker confidence and desire to go further to earn equivalent bonuses in the subsequent year.

Together, this can drive further credit extension at fine spreads, initially improving economic prospects and firm finances, and further reducing apparent indicators of potential loan loss.

But with the excessive lending that that generates, inevitably producing a set-back (Exhibit 2), and with each of these factors operating in reverse in the downswing:

• bank capital and lending capacity is depleted; and
• lending at existing prices is apparently suddenly less profitable, driving reduced willingness to lend and increased lending spreads.

On the trading book side, meanwhile, the accounting approach requires banks to value assets at ‘fair value’.⁴ Wherever possible this means a ‘mark-to-market’ approach – the assets valued at what they could be sold for in a market on the day on which the accounts are struck. When this is not possible, management and auditors are essentially trying to infer what the market price would be if it existed, through the use of indirect information inputs such as reference to other relevant prices, or to models which use other relevant prices as inputs.

The rationale of this approach is that it reflects a set of facts – the facts, or the nearest approximation of the facts, of the price for which the assets could be sold if sold at the balance sheet date. In some cases this approach is unavoidable; in some ways valuable, as I’ll come back to shortly. But it clearly creates dangers of procyclicality both on the way up and the way down, but with the effects even stronger and more disruptive than in the case of the banking book, as a result of the particularly procyclical tendencies of securitised credit extension.

When credit is extended in a securitised form (Exhibit 3), with the market price of credit clearly visible from trading in credit securities, there is a inherent risk that credit supply and pricing can be subject to self-reinforcing herd effects, with originators of and investors in credit treating the market level of credit or credit default swap (CDS) spreads as indicators of credit risk and thus of appropriate credit pricing. In the upswing this feeds a rising price of credit securities, falling spreads, increased origination and a self-reinforcing willingness to invest in credit securities; or indeed to lend on balance sheet.

That cycle in turn can be reinforced, however, by the use of fair value approaches to the valuation of assets in the trading books of banks and to the calculation of profit and loss. Rising values generate apparently high profits and swell capital basis, which can support either more trading activity or increased on balance sheet lending. And those effects may be reinforced by behavioural and confidence effects, the ‘animal spirits’ (to use Keynes’ words) of bankers paid high bonuses on unrealised and potentially illusory profits.

(Exhibit 4) A set of interlocking cycles which can then switch dramatically into reverse once confidence is lost.

• Falling security prices exacerbated by illiquidity and lack of confidence.
• Bank balance sheets impaired by trading losses which may well reflect an over-shoot of rational equilibrium values.
• And rising credit and CDS spreads feeding through to dramatically increased assumptions on the fair price of credit, whether extended in a securitised or on-balance sheet form.

It is therefore, I believe, clear that the current accounting treatment in both banking and trading books can contribute to a harmful procyclicality in profits and in credit extension in both its securitised and unsecuritised form. The issue therefore is not whether there are problems – there are – but whether changes to accounting standards can help address them, or whether we simply need to recognise the problems and offset them through other levers such as countercyclical and liquidity standards. It may be that there are problems here, exacerbated by accounting treatment, but that no feasible alternative accounting treatment exists which would help solve these problems without creating others.

In the trading books in particular, there are many instruments for which there is no feasible alternative to a fair value approach. It is almost impossible for derivative contracts to be dealt with in any other fashion. Concepts of historic cost, nominal value or incurred loss cannot help us gauge the economic substance of the risks inherent in a derivative contract.

And there is certainly also a case for being a bit cynical about some of the criticisms of fair value which broke out in the period of market downturn, but only in that period. As credit securities prices fell in 2008, many bankers expressed concern about unrealistically low market values in illiquid markets, and demanded greater freedom to shift assets from a fair value to a historic cost/hold to maturity/incurred loss accounting basis. But these concerns would have had greater credibility if the same bankers had complained about the potential adverse consequences of the over-valuation of assets in conditions of irrational exuberance in 2006 and early 2007. And any changes which make it easier for banks to shift assets between categories in the face of changing circumstances will feed concerns that problems are being hidden. In late 2008, uncertainty about the value of ‘toxic assets’, and a lack of transparency about how values had been measured by different banks, were major factors eroding market confidence.

But equally, the idea that all problems would go away if only there were total transparency and rigorous mark-to-market accounting, extending, according to the credo of true believers, to banking book loans as well as trading book securities, is I think quite wrong and dangerous. Belief in this credo is sometimes based on a fallacy of composition, a failure to distinguish between the impact of a policy on the competitive advantage of one bank relative to others, and the systemic impact of that policy universally applied. It is for instance asserted, and it may well be the case, that banks which applied a very strong daily mark-to-market philosophy in their internal risk management were better placed than others to navigate the crisis, more rapidly closing out loss-making positions, rather than holding on in the hope that something would turn up. But it is quite possible for that to be true, and for it also to be true that a system of universally applied and totally transparent mark-to-market accounting would increase the volatility of prices, and increase the volatility of credit extension. What is good for one bank seeking to compete with others can be harmful in its systemic effect.

For the fundamental problem we face is that there are no definitive ‘facts’ about value – but that value in financial markets is contingent on specific circumstances and on the action of all other participants. For an individual bank selling slices of its individual portfolio in conditions where the actions of other banks can be considered as independent, mark-to-market accounting provides meaningful facts and a useful management discipline. But if all banks simultaneously try to sell all or a significant proportion of their assets, the facts become quite different. And a fully transparent system of across the board mark-to-market accounting could simply increase the speed with which self-reinforcing assumptions about appropriate value generate cycles of irrational exuberance and then despair. Market prices can be subject to self-reinforcing momentum and herd effects as much in highly liquid, technically efficient and transparently accounted markets, as in inefficient and untransparent ones. And herd and momentum effects in credit markets (whether loans or credit securities) will cause more harm to the macroeconomy than similar irrational movements in equity prices.⁵

So in summary on the trading books:

• There is no alternative to mark-to-market accounting for some instruments. And mark-to-market valuation provides information which shareholders should logically value. And freedom to switch accounting approaches to hide problems can be dangerous.
• But conversely there can be systemic problems of increased volatility if mark-to-market accounting is applied widely.

What therefore should we do about trading book risk and trading book and related accounting? The answer must, the FSA believes, be a combination of regulatory and accounting reform, each aiming to ensure that a distinctive trading book approach is used only for activities where it is clearly appropriate.

• On the regulatory side, this entails higher capital requirements against trading books, and some changes to the trading book capital regime have already been agreed by the Basel Committee, and will be implemented by 1 January 2011.⁶ In addition, among the Basel Committee’s most important agenda items for the next year is a fundamental review of the trading book capital regime going back to the basic question of why and in relation to what specific activities, is it justifiable to have a different and lighter capital requirement against an asset held for trading as against one held to maturity? The result is likely to be much higher capital requirements for any assets which are, or under conditions of stress might become, illiquid.
• Meanwhile on the accounting side, the way forward requires a parallel look at the appropriate coverage of fair value approaches to the definition of profit or loss, ensuring that fair value gains or losses affect profit and loss (P&L) only where instruments are liquidly traded, and are likely to be liquidly tradable in almost all circumstances.

In the trading book, the crucial issue is not therefore ‘fair value’ yes or no, or mark-to-market yes or no, but rather what should be included in trading books whether for regulatory or accounting purposes.

Turning to the banking books, the problem, as Martin Taylor defined it in his article, is that we have ‘spreads that take no account of default probabilities’ and thus interest income recognised with inadequate allowance for the loss probability always inherent whenever a loan is placed on the books. The fundamental issues are therefore: how to ensure that banks adequately anticipate the probable and even the improbable but possible future in their pricing decisions and capital adequacy levels? And should this anticipation also be reflected in published accounts?

One possible approach would be to leave the accounting as it is, and concentrate entirely on prudential regulation and in particular on two changes.

• First, ensuring that Basel II is implemented in a way which does indeed anticipate the future, rather than reflecting only the recent past in a procyclical fashion. In principle that is what it was always meant to do. Capital requirements within the advanced IRB approach are meant to anticipate potential economic losses, and therefore to reflect analysis of the economic losses which might arise for different categories of loans, looking forward into an uncertain future and informed by the history of the losses which have arisen for that category in the past, in bad times as well as in good. In reality, however, Basel II’s implementation has often been focused on procyclical point-in-time estimates, in part because some banks have lacked adequate databases of past historic loss experience. We need to shift to a through-the-cycle approach, and the FSA is working with firms to achieve that.
• But second, we also need to add a clearly countercyclical regulatory capital requirement, building up capital buffers in good times to draw down in bad, and that too is a key priority within the redesign of the global capital adequacy system now being debated by the Basel Committee.

We could decide to make only these regulatory changes and to leave the accounting unchanged. But the FSA believes that there would also be considerable merit in changing the current accounting approach to loan loss provisioning – seeking both to address some of the behavioural implications arising from unrealistically high declared profits, and seeking indeed to provide investors with a more realistic picture of underlying profit.

The case in favour of a step in that direction has already been accepted by the IASB, which is now consulting on a new version of its impairment (provisioning) requirements which would require loans on balance sheet to bear an expected loss provision throughout their life, rather than recognising losses solely according to the existing incurred loss approach. In principle this approach has merit, but the devil is very much in the detail, and in particular in the detail of how ‘expected loss’ will be calculated.

• If it is calculated by reference to current market expectations of future losses, there is a danger that the new approach could actually be more procyclical than the past. In extremis, indeed, if ‘expected losses’ are calculated by reference to the market prices and spreads of traded credit securities, then an expected loss approach to the banking books becomes a form of mark-to-market by another name, potentially increasing rather than reducing procyclicality.
• Conversely, however, if, as prudential regulators would generally prefer, expected loss is calculated by reference to judgements about future possible losses informed by past experience or by formulae which link provisions to broad indicators of likely future credit problems (such as the pace of credit growth on which the Spanish dynamic provisioning approach is based) some investors might have concerns about whether these judgements, whether made by the management or by the regulator, are based on fact and are transparently understandable.

So in the detailed design of the expected loss approach, the IASB may still face the inherent trade-off between the divergent and incompatible demands placed on accounts, the investor concern that they should reflect currently verifiable facts, and the prudential regulators’ concern that they should inform the market about underlying economic reality and possibility, and should not contribute to procyclical credit extension.

Faced with that trade off between divergent aims, the FSA’s ideal preference would be to provide not one but two separate lines of account information on loan loss provisions.

• The existing line, based as now, on the facts of already incurred credit impairment events.
• And a separate line, based either on a formula, as in Spain, or on the judgements of management, challenged by regulators, and with the details, basis and rationale for that judgement extensively disclosed.

Two separate lines and extensive disclosure would, we believe, provide better information to investors than either the current incurred loss line or any one ‘expected loss’ based line could ever provide. If we do instead proceed with one ‘expected loss’ line, the disclosure of supporting information to enable investors to understand the assumptions made will be important.

At least in the US, however, it is often reported that ‘investors’ are strongly opposed to any divergence from existing standards, and indeed that they continue to be attracted to the extension of ‘fair value’ accounting to which FASB still appears to be devoted. At first sight this attachment to fair value accounting appears strange. It is not clear why it was in the interest of investors to be told in spring 2007 that the credit securities held by the US or UK banking system were worth hugely more than they appeared to be worth only a year later. But looked at another way, the logic of this ‘investor’ perspective, or at least the perspective of the agents who make decisions on behalf of investors (i.e. hedge funds, pension fund managers etc) is clearer. Because the fund managers acting for end investors (if not the investors themselves) may be logically more interested in precise and non-judgemental information as to the relative economic position of different firms, than in judgemental information, which, while capturing important aspects of aggregate economic reality, is likely to create more opportunities for cloudiness – non-transparency – in inter-firm comparison.

But that clearly illustrates that the interest of investors (or at least of fund managers) and of prudential regulators (and of sound macroeconomic management) may diverge. And that we need to find ways of serving both interests.

And indeed it illustrates again the fundamental problem with which we are struggling in accounting for financial services companies in general but for banks or near banks in particular: which is that under the conditions of inherent uncertainty which govern financial contracts – contracts which link the present to the future – there is no one ‘truth’, no one set of ‘facts’ relevant to all decisions and decision makers, but several truths, several facts.

• It is true that the value at which one individual firm can liquidate part of its portfolio at a particular point in time is best given by the mark-to-market value.
• But it is also true that if all firms try to liquidate a large proportion of their portfolio simultaneously, the mark-to-market value will not be obtainable. And that a transparent system of mark-to-market accounting, with fluctuating and transparent prices, may tend to induce the very collective behaviour which makes the measured mark-to-market value cease to be true.

In formal economic terms, there are multiple and unstable equilibria, and the extent to which the equilibria are unstable is itself influenced by the nature of information flows about market values.

Similarly, and turning to the banking books:

• It was true that at the end of 2006 and even 2007 that the number of business borrowers behind with their payments in many banks’ commercial loan portfolios was very low, and that was a fact which investors might legitimately want to note, providing information both about the idiosyncratic position of each bank relative to other banks, and about the point in the cycle.
• But it is also true that when a bank puts on loans during periods of rapid economic and credit growth and general market confidence, that the measured loan losses then observed will have a systematic tendency to understate the average subsequent resulting loss.

The fundamental challenge of bank accounting is therefore to provide good information in an environment where the idea that there is one measure of value, one ‘economic reality’, is a chimera, a sub-set of the wider intellectual delusion that has attempted to construct economic theory and policy on the assumption that markets are always rationally equilibrating and that market prices are by definition ‘correct’.

And that does make banks different – not of course in absolute terms but to such a degree that different considerations must influence the development of bank accounting standards than are taken into account in other sectors of the economy. Clearly, for other companies as well as banks, asset values can vary between normal and distressed times: if all retailers went bankrupt simultaneously the liquidation value of their stocks would be quite different from that observed in the accounts. But in reality retailers do not all get into trouble simultaneously; and no other sector of the economy is remotely comparable with banking in its capacity to be a driver of economic volatility rather than a victim of it. And there is no other sector about which Martin Taylor would have argued, reasonably, that accounting played an important contributory role in provoking general sectoral collapse and macroeconomic recession.

Banks are different because they matter more, because they can do more harm. That’s why we regulate and supervise their business but do not regulate the business of retailers, hoteliers or manufacturers. That’s why there is a special relationship between the banking system and central banks as lenders of last resort. That’s why we worry a lot about ‘too big to fail’ considerations. And that’s why prudential regulators, central banks and economic policymakers have a vital interest in the decisions of accounting standard setters on bank accounting standards, which does not apply between regulators and accounting bodies in any other sector of the economy.