New loan provisioning standards and procyclicality

Panel remarks by Mr Claudio Borio, Head of the Monetary and Economic Department of the BIS, at the High-level conference on "The new bank provisioning standards: implementation challenges and financial stability implications", jointly organised by the Bank of Spain, the Centro de Estudios Monetarios y Financieros (CEMFI) and the BIS Financial Stability Institute (FSI), Madrid, Spain, 18-19 October 2018.

BIS speech  | 
24 October 2018


The adoption of the new expected credit loss provisioning standard - IFRS 9 - is a landmark. What are its implications for financial stability? While the new standard is likely to mitigate the procyclicality of the financial system to some extent relative to the previous, incurred loss model, it falls short by a significant margin of what one would like from a financial stability perspective. This points to broader inevitable tensions between accounting and prudential regulation, and calls for the active use of backstops (or so-called prudential filters) to preserve stability. Experience with the operation of the alternative dynamic (countercyclical) credit loss provisioning scheme adopted by the Bank of Spain points to some strengths and weaknesses in the broader macroprudential frameworks in which such arrangements are embedded.

Full speech

It is a great pleasure for me to participate in this panel. I always look forward to coming to Spain, and to the Bank of Spain in particular - not just because I can see again so many friends, but because it brings back fond memories of my first job outside academia. My very first day of employment, at the OECD, was actually here in Madrid, when I joined my new bosses on an economic mission to the country, without even passing through head office! But, much as I would like to continue, let me stop reminiscing and turn to the topic of the panel: loan loss provisions, procyclicality and financial stability. 

The adoption of the new expected credit loss (ECL) provisioning standard - IFRS 9 - is a landmark. It represents the end-point of a long - in some respects extraordinary - journey that started around 2000, with the emergence of more systematic concerns about the "procyclicality" of the financial system. It was these concerns that prompted the development of the conceptual underpinnings of macroprudential frameworks and the subsequent implementation of those frameworks post-Great Financial Crisis (GFC). To be sure, the adoption of the new accounting standard was not intended to address procyclicality per se; rather, it aimed to align the approach with the more forward-looking nature of fair value accounting generally. Even so, the change did follow an explicit request by the G20 and the Financial Stability Board in the context of how to deal with procyclicality. 

The Bank of Spain has been a pioneer in this area, with its early adoption in 2000 of countercyclical (or dynamic) provisions (Saurina and Trucharte (2017)). 

In what follows, I would like to broaden the focus a bit and address three questions. First, how far do the new provisioning standards address the procyclicality in credit loss provisioning? Second, what does this tell us about the tensions between accounting and prudential regulation and about potential remedies? And finally, have macroprudential frameworks fulfilled the expectations of their advocates, of whom I have been one (Borio (2003))? 

Let me anticipate the three answers. First, the new standards are likely to mitigate procyclicality to some extent relative to the previous, incurred loss model, but from a financial stability perspective they fall short by a significant margin of what one would like to see. Second, the tensions between accounting and prudential regulation are inevitable, calling for the active use of backstops (or so-called prudential filters). Finally, we need to be realistic about what macroprudential frameworks can do on their own: they are more effective in strengthening the financial system's resilience than in taming procyclicality - or the financial cycle. 

Let me take each point in turn. 

I. New ECL provisioning standards and procyclicality 

To understand how far the new ECL provisioning standard addresses procyclicality, it is useful to say a few words about the nature of the problem and how the Bank of Spain's dynamic provisions tackled it. One can then compare the new standard with that benchmark. 

Procyclicality denotes the financial system's tendency to generate financial booms and busts and, more specifically, those mechanisms that feed onto themselves to amplify financial fluctuations (Borio et al (2001)). At the core of those mechanisms is the self-reinforcing interaction between funding constraints, asset prices and risk-taking. For instance, during expansions funding constraints become looser, asset prices soar, risk-taking increases, triggering a vicious circle until the process becomes unsustainable and, at some point, that risk-taking reverses. As a result, booms generate busts. 

Procyclicality arises for two reasons. One is that incentives to take on risk are procyclical. Think, for instance, of herding, just to mention one. The other is that, above all, measures of risk are procyclical, because the inputs are. During booms, asset prices soar, inflating collateral values, credit spreads narrow, volatility declines (it is inversely related to asset prices), correlations decline, reducing the volatility of portfolio returns and profits and free cash flows increase. During busts, these relationships reverse. 

Put differently, procyclicality fundamentally changes the conception of risk (Crockett (2001)). Risk is not low in booms and high in busts - the previous conception; but it builds up in booms and materialises in busts. The bust is a consequence of the boom that precedes it. 

Of course, some procyclicality is inevitable and inherent in economic activity. But, unless restrained, procyclicality can give rise to outsize financial fluctuations, or financial cycles, that are typically at the heart of financial instability. 

The previous incurred loss model of credit provisioning - IAS 39 - was clearly procyclical (Borio and Lowe (2001)). In general, provisions could be made only when a loss impairment event or events had taken place. In the former terminology, they could be taken only when risk materialised. As a result, losses over the life of the credit exposure were underestimated during the boom. The scheme did not recognise those embedded in the portfolio. 

The Bank of Spain's dynamic provisioning scheme tackled this problem head-on (Saurina and Trucharte (2017)). To simplify: at inception, the provisions on a loan would be equal to the average loss made on similar loans during previous recessions. Those provisions would be released automatically as losses materialised. 

One could, of course, take issue with some aspects of the scheme. For instance, it did not account explicitly for loan pricing, which should already incorporate an expected-loss element (Borio and Lowe (2001)). And there was some inevitable arbitrariness in the selection of benchmarks for the size of the provisions. In fact, the chosen recession year - 1993 - turned out to underestimate by a very large margin the losses during the GFC. But the scheme had the great merit of being truly countercyclical, of being simple and, in particular, of having an automatic release mechanism. The importance of this last feature should not be underestimated. One should recall that it has proved exceedingly difficult to design a similar automatic release trigger for the countercyclical capital buffer. I was intimately involved in the process and, believe me, we did try! In the end, the Basel Committee on Banking Supervision could only produce general guidelines. This left plenty of room for discretion, making life harder for supervisors. 

Against this benchmark, the new ECL forward-looking scheme falls short by a significant margin. Granted, the scheme correctly seeks to identify the losses embedded in the portfolio in good times: this is an important step forward. But its impact on procyclicality is much weaker. For one, the scheme leaves ample room for firms' discretion. They will still have a strong incentive to underprovision, especially in good times, and it will not be easy for auditors to correct this - just as it has not been easy for prudential authorities to address the biases embedded in banks' internal risk models. In addition, and above all, the scheme remains point-in-time. That is, it does not have the in-built look-back, mean-reverting element at the core of the Bank of Spain's dynamic provisioning scheme. Firms are simply asked to forecast losses over a particular horizon given available information, without the restriction of using the average or stress loss incurred over past cycles. As a result, provisions are still subject to the typically strong procyclicality of risk assessments. 

Thus, compared with the incurred loss standard, the most we can expect from the new one is that it will bring forward some of the provisioning. Work done at the BIS, published in our Quarterly Review, confirms this intuition (Cohen and Edwards (2017)). Better loss recognition in good times is very welcome. And if the scheme is properly implemented, recognition of higher losses in good times means recognition of smaller losses in bad times, ie less procyclicality. The extent, though, is to be seen and deserves close study. It will clearly depend, among other things, on the implementation details, not least the models used to forecast losses. 

II. Accounting and prudential regulation: uncomfortable bedfellows 

This naturally takes me to my second point - the uneasy relationship between accounting, on the one hand, and prudential regulation, on the other. We can call them two "uncomfortable bedfellows" (Borio and Tsatsaronis (2004)). And in fact, the same is true of the relationship between accounting and sound risk management (Borio and Tsatsaronis (2006)). 

The tensions between accounting and prudential regulation started to become irreconcilable once accounting shifted away from the "prudence" principle in order to provide a "true and fair" picture of a firm's condition. We could have a long discussion about what "true and fair" really means and about how far the principles really do that. Think, for instance, of the well known debate around "income smoothing" in the context of dynamic provisions (Borio and Lowe (2001)). But there is little doubt that accounting standards are not always consistent with the requirements of financial stability. The incurred loss model example, and the acute procyclicality induced by fair value accounting more generally, are testimony to this. 

A similar tension arises between accounting and sound risk management. Let me just quote from a famous firm's internal risk management manual: 

"Reported earnings follow the rules and principles of accounting. The results do not always create measures consistent with underlying economics. However, corporate management's performance is generally measured by accounting income, not underlying economics. Risk management strategies are therefore directed at accounting rather than economic performance." 

This quotation happens to be from Enron's operating manual - and we all know what happened to that firm! But I suspect it could equally come from that of any other firm. Whether we like it or not, accounting drives incentives and hence behaviour - in fact, it is designed to do precisely that. A kind of Heisenberg's uncertainty principle is at work here: how we measure valuations ends up influencing valuations, as agents respond to them - again, think of procyclicality. It is here that micro meets macro: what firms take as given in the small, such as market prices, is influenced by their collective behaviour. Measurement cannot be neutral: it affects what is being measured (Borio and Tsatsaronis (2006)). Accounting doesn't just record facts, it alters those facts. 

How can the tension between accounting and prudential regulation best be managed? I think it would be unrealistic to have accounting standard setters take financial stability into account: that is not their objective. Nor do I think they could be persuaded to re-adopt the principle of prudence. Except where bank supervisors have authority over accounting standards for banks and are prepared to override accounting standard setters, this is not a feasible option. That, of course, is what the Bank of Spain did with its dynamic provisions. And it is what a number of prudential authorities in Asia and Latin America are still doing (Restoy and Zamil (2017)). The cost of doing so, however, is to clash head-on with the accounting profession, as it would get in the way of what they want and what they are mandated to achieve. 

Short of that, prudential authorities have three options. 

One is to argue with accounting standard setters on their own terms. Forward-looking provisioning principles are one such example. But, as noted, they do not go far enough from a financial stability perspective. 

A second option is to persuade accounting authorities to require that firms disclose more risk information, notably information about the degree of uncertainty surrounding valuations. I argued for this many years ago (Borio and Tsatsaronis (2004, 2006)), and I am glad to see that standards have moved in that direction, in particular IFRS 7. But again, this is not enough for financial stability: the investors and depositors that should enforce market discipline are subject to the same measurement and incentive problems of the institutions they are supposed to restrain. 

A third option is to adopt prudential backstops or filters to offset accounting valuations - a practice that has been in place for some time now. These measures can compensate for some of the shortcomings of accounting provisions by adjusting regulatory capital, possibly complemented with restrictions on dividend payments. To my mind, these filters are indeed indispensable. 

The real issue is how to calibrate and structure them. There is a wide range of possibilities, from simply deriving adjustments based on information that contains a mean-reverting element, such as financial cycle indicators, to adopting the same type of adjustment embedded in the Bank of Spain's dynamic provisions - in effect, a simple through-the-cycle filter. With colleagues, we plan to look into this issue in more depth. 

The advantage of such prudential filters is that they decouple accounting from regulatory valuations, allowing each authority to pursue its preferred objective (Borio and Tsatsaronis (2004)). This advantage should not be underestimated. The disadvantage is the other side of the coin, ie the filters are less effective in enforcing market discipline on banks than changing accounting standards themselves. This is because they do not affect the bottom line earnings figures analysts and markets focus on. 

III. Macroprudential frameworks 

Finally, some reflections on macroprudential frameworks, of which backstops for ECLs are just one element. 

As an early strong advocate of such frameworks, I was very happy when they were adopted following the GFC. This has represented real progress. 

But now the pendulum may have swung too far. There is a widespread belief that macroprudential frameworks are the solution to procyclicality. My personal assessment is that they are part of the solution, but not the whole solution. They unquestionably strengthen the financial system's resilience, as they reinforce its defences to face financial cycle busts. But they are less effective in restraining financial booms in the first place. 

There is considerable evidence to that effect. In particular, the active deployment of macroprudential tools in some countries, mainly emerging market economies, has not prevented the emergence of the familiar signs of the build-up of dangerous financial imbalances - typically, cumulative credit growth and asset price increases, notably property prices, in excess of historical benchmarks. These elements are the basis of the early warning indicators that worked pretty well pre-GFC (Borio and Drehmann (2009)). The Spanish experience with dynamic provisions confirms this assessment: according to the Bank of Spain's own analysis, dynamic provisions have not succeeded in restraining credit growth significantly (Saurina and Trucharte (2017)). Admittedly, other measures, such as maximum loan-to-value ratios and debt service-to-income ratios, have a larger impact. But this does not alter the overall conclusion. 

In my view, tackling the financial cycle requires a more holistic policy framework, which in addition to sound prudential standards also involves monetary and fiscal policy, and even structural policies. This is what we at the BIS call a "macro-financial stability framework". We discussed these issues at some length in the latest BIS Annual Economic Report (BIS (2018)), and I examined them further in my remarks at our last Annual General Meeting (Borio (2018)). 

In other words, there is a material risk that unrealistic expectations of what macroprudential frameworks can deliver on their own stands in the way of desirable adjustments in monetary and fiscal policies. I think we have seen signs of this danger materialising. 

Of course, my assessment could be overly pessimistic. Only time will tell. Let's hope it will not take another crisis to find out.  


Bank for International Settlements (2018), BIS Annual Economic Report 2018, June. 

Borio, C (2003): "Towards a macroprudential framework for financial supervision and regulation?", CESifo Economic Studies, vol 49, no 2/2003, pp 181-215. Also available as BIS Working Papers, no 128, February; as well as (revised and updated) in F Milne and E Neave (eds), Current directions in financial regulation, May. 

--- (2018): "Macroprudential frameworks: experience, prospects and a way forward", speech on the occasion of the Bank's Annual General Meeting, Basel, 24 June. 

Borio, C and M Drehmann (2009): "Assessing the risk of banking crises - revisited", BIS Quarterly Review, March, pp 29-46. 

Borio, C, C Furfine and P Lowe (2001): "Procyclicality of the financial system and financial stability: issues and policy options", BIS Papers, no 1, March, pp 1-57. 

Borio, C and P Lowe (2001): "To provision or not to provision", BIS Quarterly Review, September, pp 36-48. 

Borio, C and K Tsatsaronis (2004): "Accounting and prudential regulation: from uncomfortable bedfellows to perfect partners?", Journal of Financial Stability, vol 1, issue 1, September, pp 111-35. 

--- (2006): "Risk in financial reporting: status, challenges and suggested directions", BIS Working Papers, no 213, August. 

Cohen, B and G Edwards Jr (2017): "The new era of expected credit loss provisioning", BIS Quarterly Review, March, pp 39-56. 

Crockett, A (2001): "Marrying the micro- and macroprudential dimensions of financial stability", BIS Speeches, 21 September. 

Restoy, F and R Zamil (2017): "Prudential policy considerations under expected loss provisioning: lessons from Asia", FSI Insights, no 5, October. 

Saurina, J and C Trucharte (2017): The countercyclical provisions of the Banco de España (2000-2016), Bank of Spain.