How did Financial Reporting Contribute to the Financial Crisis?
by Mary E. Barth & Wayne R. Landsman
Source: Taylor & Francis: European Accounting Review
We scrutinize the role financial reporting for fair values, asset securitizations, derivatives and loan loss provisioning played in the Financial Crisis. Because banks were at the center of the Financial Crisis, we focus our discussion and analysis on the effects of financial reporting by banks. We conclude fair value accounting played little or no role in the Financial Crisis. However, transparency of information associated with asset securitizations and derivatives likely was insufficient for investors to assess properly the values and riskiness of bank assets and liabilities. Although the FASB and IASB have taken laudable steps to improve disclosures relating to asset securitizations, in our view, the approach for accounting for securitizations in the IASB's Exposure Draft that would require banks to recognize whatever assets and liabilities they have after the securitization is executed better reflects the underlying economics of the securitization transaction. Regarding derivatives, we recommend disclosure of more disaggregated information, disclosure of the sensitivity of derivatives' fair values to changes in market risk variables, and implementing a risk-equivalence approach to enable investors to understand better the leverage inherent in derivatives. We also conclude that because the objectives of bank regulation and financial reporting differ, changes in financial reporting needed to improve transparency of information provided to the capital markets likely will not be identical to changes in bank regulations needed to strengthen the stability of the banking sector. We discuss how loan loss provisioning may have contributed to the Financial Crisis through its effects on procyclicality and on the effectiveness of market discipline. Accounting standard setters and bank regulators should find some common ground. However, it is the responsibility of bank regulators, not accounting standard setters, to ensure the stability of the financial system.
The Financial Crisis that began midway through 2007 and continued through the end of 2008 resulted in the collapse of numerous commercial and investment banks, including several high profile institutions such as Bear Stearns, Lehman Brothers, Merrill Lynch and Wachovia. The crisis resulted in a near systemic collapse of the banking sector, on which commercial lending activity depends. Much has been written about the causes of the crisis, as policy-makers, bankers and academics offer their somewhat differing insights and perspectives. Most agree the crisis started following the bursting of the housing bubble in the USA. The near collapse of the financial sector has resulted in the greatest economic contraction that the USA and Europe have seen since the end of the Second World War. As policy-makers work to deal with the problem, pressure is being placed on elected officials to make legislative and regulatory changes to address the underlying causes of the housing bubble and the Financial Crisis.
Calls for action to prevent a repeat of the Financial Crisis have also touched on accounting standard setting and the way in which accounting information (or lack thereof) contributed to the Financial Crisis. In this regard, fair value accounting perhaps received the most attention, with many bankers, political figures and commentators contending that fair value accounting was a major contributing factor to the procyclical decline in the value of bank assets, and hence bank stock prices, when the housing market bubble burst. However, there are other aspects of accounting information required by accounting standard setters that received less attention but nonetheless may have played a significant role in contributing to the Financial Crisis. The first, which is tied to the housing and mortgage lending markets, is the quality of information investors had relating to asset securitizations. Securitizations involve the packaging of financial assets, such as loans, credit card receivables and mortgages, and the sale of securities representing the cash flows of those assets. The second is the quality of information investors had relating to derivatives, including credit default swaps.
Our primary goal is to scrutinize the role that financial reporting relating to fair values, asset securitizations, derivatives and loan loss provisioning played in contributing to the Financial Crisis. Because banks were at the center of the Financial Crisis, we focus our discussion and analysis on the effects of financial reporting by banks. Bank regulators use information provided in bank financial statements as inputs to the calculation of regulatory capital measures and rely on capital markets that trade on the information to discipline bank behavior. As a result, despite the fact that bank regulators make a variety of adjustments to financial reporting information, financial reporting and bank regulation are often thought of as being one and the same. Therefore, we begin by discussing the differing objectives of financial reporting and bank regulation to help clarify that information standard setters require firms provide to the capital markets and information required by bank regulators for prudential supervision will not necessarily be the same. This distinction is important to understanding how financial reporting, in contrast to bank regulation, contributed to the Financial Crisis.
Our analysis of the way in which financial reporting relating to fair values, asset securitizations and derivatives potentially contributed to the Financial Crisis is structured similarly for each topic. In particular, for each we summarize the financial reporting requirements of US Generally Accepted Accounting Principles (GAAP) and International Financial Reporting Standards (IFRS), and offer insights into whether the information available to investors was sufficiently transparent to make appropriate judgments regarding the values and riskiness of affected bank assets and liabilities. We also summarize available research evidence. Where appropriate, we offer suggestions regarding potential improvements in financial reporting. Because loans comprise a large fraction of bank assets, we also discuss how loan loss provisioning may have contributed to the Financial Crisis through its effects on procyclicality and on the effectiveness of market discipline.
Our analysis leads us to conclude that contrary to what many critics of fair value contend, fair value accounting played little or no role in the Financial Crisis. This conclusion is consistent with that of other scholarly and regulatory analyses of the potential link between fair value accounting and the Financial Crisis. However, transparency of information associated with measurement and recognition of accounting amounts related to, and disclosure of information about, asset securitizations and derivatives likely were insufficient for investors to assess properly the values and riskiness of affected bank assets and liabilities. In addition, because the objectives of bank regulation differ from the objective of financial reporting, changes in financial reporting requirements to improve transparency of information provided to the capital markets likely will not be identical to the changes in bank regulations needed to strengthen the stability of the banking sector. Moreover, bank regulators have the power to require whatever information is needed to meet the objective of prudential supervision. Although it makes sense from the standpoint of efficiency for accounting standard setters and bank regulators to find some common ground, it is the responsibility of bank regulators, not accounting standard setters, to determine how best to ensure the stability of the financial system.
2. Different Objectives of Financial Reporting and Bank Regulation
The concepts statements underlying US GAAP and IFRS state that the objective of financial reporting is to provide information that is useful to present and potential investors and creditors and others in making investment, credit and similar resource allocation decisions (FASB, 2008; IASB, 2008a). This objective applies to general purpose financial reporting for all firms, regardless of industry or whether firms in a particular industry are subject to regulation that uses financial statement information as an input.
The primary objectives of bank regulation are prudential, that is, to reduce the level of risk to which bank creditors, for example, depositors, are exposed, and to mitigate systemic financial risks. Although bank regulators may choose to use general purpose financial reporting information in meeting their objectives, one should expect that bank regulators would not limit themselves to information contained in general purpose financial reports. For example, in the USA, bank regulators require a variety of additional disclosures relating to recognized assets and liabilities, e.g., non-performing loans and deposits, as well as additional information relating to bank risks. In addition, when calculating measures used as input for their supervision of banks, such as regulatory capital, regulators often make their own adjustments to recognized financial statement amounts to better suit their objectives. Regulatory capital need not be equal to financial reporting capital because bank regulators apply so-called ‘prudential filters’, e.g., specific adjustments when calculating regulatory capital, to meet their objectives of prudential supervision. Examples include neutralizing pension surpluses, e.g., recognized pension assets, and gains/losses associated with the fair value option in International Accounting Standard (IAS) 39 (CEBS, 2007). In addition, regulators can adjust the risk weights they assign to specific assets when determining required levels of capital.
Given the differing objectives of financial reporting and bank regulation, it is not surprising that the information each requires of banks differs. For example, the Financial Accounting Standards Board (FASB), which issues US GAAP, and the International Accounting Standards Board (IASB), which issues IFRS, require recognition of some unrealized gains and losses presumably because they believe doing so meets their objective of enhancing the information banks provide to investors and creditors. In contrast, bank regulators often neutralize such unrealized gains and losses in regulatory capital presumably because they believe doing so yields a measure of bank capital that is more useful for prudential supervision. As a result, whereas US GAAP and IFRS require banks to measure available-for-sale investment securities at fair value and recognize cumulative unrealized gains and losses in accumulated other comprehensive income, bank regulators in many countries, including the USA, compute Tier 1 capital after removing these cumulative unrealized gains and losses.
In light of the differing objectives of financial reporting and bank regulation, standard setters should not be surprised that bank regulators make adjustments to general purpose financial statement information for use in prudential supervision. At the same time, bank regulators should not be surprised that accounting standard setters require information that is not perfectly suited for prudential supervision. It does not make sense for accounting standard setters to issue recognition and measurement standards that meet the needs of one set of users, including bank regulators, while ignoring the informational needs of others. However, it makes sense from the standpoint of efficiency for accounting standard setters and bank regulators to find some common ground (Bushman and Landsman, forthcoming). For example, bank regulators' concerns with fair value accounting and loan loss provisioning are best addressed by making appropriate adjustments for regulatory purposes to amounts in general purpose financial statements issued by the banks they regulate. That is, if bank regulators believe prudential supervision is best achieved by making adjustments to recognized accounting amounts when determining bank capital (e.g. applying prudential filters for fair value gains and losses and for loan loss provisions), standard setters could require banks to disclose in their financial statements a reconciliation between financial reporting capital and profit or loss and regulatory capital and profit or loss.
Returning to the point that accounting standard setters need to be concerned with the information needs of the capital markets, it is important to note that bank regulators also should be concerned with those needs. Pillar 3 of Basel II states that regulators can rely on capital market disciplining forces as a tool in prudential supervision. That is, the capital markets can serve as a complementary force to direct bank supervision. The extent to which bank regulators can rely on market discipline to perform this role depends on the quality of information available to the capital markets. Thus, if accounting standard setters fail to keep the informational needs of capital markets as their first priority, an unintended consequence is that the effectiveness of market discipline as a regulatory tool could be undermined.
3. Fair Value Accounting
During the Financial Crisis, US GAAP and IFRS required recognition of some assets and liabilities – principally financial instruments – at fair value, with some changes in fair values recognized in profit or loss. Fair value is defined by the two Boards as the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between knowledgeable market participants at the measurement date.
Fair value accounting has been criticized generally and more specifically as a major contributing factor in the Financial Crisis. The general criticism of fair value is that it is not the best measurement attribute for conveying decision-useful information to financial statement users. Further, it has been suggested that some other measure, for example, modified historical cost, is more decision-useful (Penman, 2007). The criticism has been addressed by accounting researchers by comparing the value relevance of various current value measures to that of historical cost measures. In addition, critics claim that if there are no observable prices on which to base fair value estimates, the estimates can lack decision usefulness because management is afforded the opportunity to manipulate the estimates to meet their own objectives. However, SFAS 157 Fair Value Measurements (FASB, 2006) specifies how to estimate fair values, thereby limiting the scope of management's ability to manipulate the estimates. Critics further contend that even if prices are observable, fair value estimates based on observable prices will not reflect dimensions of asset values about which management has private information. However, SFAS 157 provides that observed prices need not be used if the price results from a distressed transaction, or if the price relates to an asset that has different attributes from the asset for which management is estimating fair value. In other words, management is required to adjust observed prices to reflect attributes specific to its asset.
The general tenor of the fair value criticisms is that fair value information, particularly in the context of the Financial Crisis, lacks sufficient quality to be informative to investors and other financial statement users. There is a substantial body of accounting research that addresses this criticism using a variety approaches, particularly value relevance. Value relevance is a particularly applicable approach to address the relevance and reliability of accounting information because an accounting amount is value relevant only if it is relevant to investors' equity valuation decisions and sufficiently reliable to be reflected in share prices (Barth et al., 2001). Landsman (2007) provides a survey of extant value relevance research relating to fair value accounting (see also Barth and Landsman, 1995; Barth, 2004, 2006). Studies that focus particularly on the value relevance of fair values for banks include Barth (1994), Bernard et al.(1995), Barth et al. (1996), Beatty et al. (1996), Eccher et al. (1996) and Nelson (1996). Taken together, the fair value literature, including the studies that focus on banks, provides rather substantial evidence that recognized and disclosed fair values are relevant to investors and reliable enough to be reflected in share prices.
Critics of the quality of fair value information also contend that including unrealized gains and losses in earnings makes earnings ‘too volatile’. Barth (2004) makes the observation that there are three primary sources of ‘extra’ volatility associated with fair value-based accounting amounts relative to those determined using modified historical cost. The first is true underlying economic volatility that is reflected in changes in assets' and liabilities' fair values. For earnings to be informative to investors it needs to reflect this volatility. The second, induced volatility arising from using a mixed-measurement accounting model (i.e. volatility arising from measuring some assets and liabilities at fair value and others at modified historical cost), essentially vanishes if banks measure all financial instruments at fair value as permitted by IAS 39 and SFAS 159. The third is volatility induced by measurement error in estimates of fair value changes. Barth (1994) provides evidence of this in the context of banks' investment securities. Nonetheless, it is important to note that this source of volatility exists for all accounting measures that depend on estimates, including those based on modified historical cost. In addition, the importance of the source of volatility arising from including fair value gains and losses in earnings depends on the informational benefits arising from the relevance of fair values vs. any costs arising from this source of volatility. Evidence from the value relevance literature discussed above suggests that the benefits of fair values outweigh their costs.
Critics of fair value contend that although fair values might be value relevant during times of relative market stability, they lack relevance and reliability during times of relative instability. Because of the recency of the Financial Crisis, evidence from academic research addressing the value relevance of fair values during this time of market instability is not yet available. Thus, the extent to which this criticism is valid remains to be seen. Because there is nothing inherent in value relevance research that depends on market stability, it is premature to conclude that fair values determined during times of market instability lack relevance and reliability. However, during times of market instability, it may be more difficult to estimate fair value, which could reduce the combined relevance and reliability of fair value. Nonetheless, we cannot think of a better alternative to fair value during times of market instability. For example, modified historical cost lacks relevance and other estimates of current value lack the discipline on the estimation process and objectivity associated with fair value measurement.
The Financial Crisis began with the collapse of the housing market in the USA and quickly spread to other markets worldwide. The bursting of the housing market bubble resulted in the collapse in prices of loans and other financial instruments whose values are tied to housing prices. The decrease in bank asset values necessitated recognition of impairments in banks' financial statements (Shaffer, 2010). Banks are required by regulators to maintain specified capital ratios to avoid the risk of regulatory intervention. Because bank capital ratios are calculated based on financial statement amounts, the recognized asset impairments could have caused banks to sell impaired assets to generate cash, which could have been used to repay debt to maintain required capital ratios.
This process of deleveraging by banks could have had two macroeconomic effects related to the Financial Crisis. First, banks essentially were unable to originate new loans – that is, extend credit to homebuyers and commercial businesses – thereby beginning the credit crisis and contraction of the economy (Ivashina and Scharfstein, 2009). Second, as the supply of assets being sold likely increased, prices likely declined further. Both of these effects are procyclical, which means that these actions taken by banks could have caused further decline in bank asset prices and contraction of the economy. Procyclicality is a natural consequence of an economic downturn, particularly one caused by the bursting of an asset market bubble. For example, in the case of the bursting housing bubble, homeowners reduced consumption, causing retail and wholesale businesses reliant on consumer spending to scale back operations, and so on. More generally, procyclicality naturally occurs throughout the business cycle. However, the concern of policy-makers is that institutional features of financial reporting and the regulatory system could amplify natural procyclicality. For example, requiring banks to hold more capital during an economic downturn because bank assets have become riskier could result in less lending than otherwise would be the case, thus amplifying the downturn. Hereafter, we distinguish between natural procyclicality and amplified procyclicality potentially arising from financial reporting and the regulatory system.
Some have asserted that procyclical effects, natural or amplified, of asset impairments derive from fair value accounting. The basic argument tying fair value accounting to amplified procyclicality is that auditors required banks to write down affected assets to unrealistically low values as reflected by ABX index prices. Because the Financial Crisis caused a drop in liquidity, ABX index prices allegedly reflected distressed prices rather than prices from an orderly market. Bank managers contended that ABX prices, being artificially low relative to the bank managers' perceived asset values, caused unnecessarily large impairment charges. That is, impairment charges would have been lower had bank managers been permitted to use their personal assessments of value, and the economy would have suffered a less severe downturn.
Although, in principle, an ‘excessive’ fair value-related impairment charge could have amplified procyclicality of bank asset prices, we believe this is unlikely for two reasons. First, this claim can only apply to those bank assets either that were measured at fair value or for which fair values apply when determining impairment. The proportion of bank assets for which this is the case is limited. Laux and Leuz (2010) report that during the 2004–2006 period banks held approximately 50% of their assets in loans and leases, which are not subject to fair value accounting and are not impaired to fair value. Although, for the 14 largest US commercial banks, Shaffer (2010) reports the decline in Tier 1 capital during the Financial Crisis arising from impairments of loans averaged 15.6%, those impairments were based on an incurred loss model and not on fair value. Therefore, as discussed in Section 6, although impairments of loans, that is, loan loss provisioning, during the Financial Crisis likely had procyclical effects, fair value accounting played no role relating to loans.
Second, bank regulators in many countries apply a prudential filter when calculating Tier 1 capital that neutralizes some fair value gains and losses. Thus, temporary changes in fair value to which prudential filters apply do not affect Tier 1 capital. Except for trading assets, generally only impairments to fair value relating to available-for-sale and held-to-maturity securities that are deemed other than temporary affect Tier 1 capital. As Shaffer (2010) notes, few contended that fair value was inappropriate for trading assets; concerns focused primarily on impairments to fair value of available-for-sale and held-to-maturity assets. Shaffer (2010) provides evidence that the decline in Tier 1 capital arising from impairment of available-for-sale and held-to-maturity assets for the 14 largest US banks during the Financial Crisis averaged only 2.1%. Because there was undoubtedly some decline in value of all bank assets, 2.1% is the upper bound of the potential effect on Tier 1 capital arising from recognition of impairments of these assets using so-called ‘artificially low’ ABX prices.
Thus, because prudential filters neutralized the effect on Tier 1 capital of some fair value losses and the larger effect on Tier 1 capital arose from loan losses that were not determined using fair value, critics' assertions that fair value accounting played a significant contributing role in causing procyclicality of bank asset prices during the Financial Crisis appear to be overstated. However, it is not possible to determine the extent to which fair value accounting would have contributed to amplified procyclicality had the prudential filter relating to fair value losses not been applied. Perhaps a motivation underlying criticism of fair value accounting is the concern that bank regulators might remove this prudential filter.
Although application of the prudential filter relating to fair value gains and losses on investments could mitigate procyclical effects of fair value accounting, application of some prudential filters relating to fair value can have the opposite effect. Notably, the prudential filter that neutralizes the effects of fair value gains and losses on liabilities arising from changes in a bank's own credit risk amplifies procyclicality. This is because when a bank's credit risk increases (decreases), the value of its liabilities decreases (increases), which results in a gain (loss) and a commensurate increase (decrease) in equity. Neutralizing this effect in times of economic decline (growth) results in decreases (increases) in equity and Tier 1 capital, which amplify procyclicality.
Regardless of any role that fair value accounting played in the Financial Crisis, it is important to recall that it is the responsibility of bank regulators, not accounting standard setters, to determine how best to mitigate the effects of procyclicality on the stability of the banking system. To meet their objectives of prudential supervision, bank regulators have many tools at their disposal, including application of prudential filters (as illustrated by the filter for fair value losses on available-for-sale assets), relaxation of regulatory capital ratios during economic downturns, for example, by altering risk-weighting of specific assets, and use of counter-cyclical measures in loan loss provisioning for regulatory purposes. Moreover, as noted earlier, the effectiveness of market discipline as a regulatory tool could be undermined if investors' informational needs were hindered by not having impairments for bank assets measured at fair value as required by IFRS or US GAAP.
4. Asset Securitizations
At the heart of the credit crisis of 2007–2008 is asset securitizations, a Wall Street innovation that became a large source of financing beginning in the early 1990s. By the end of 2007, the outstanding securitization market was valued at $9.3 trillion, making it over twice the size of the market for US treasuries, which was valued at $4.5 trillion. Asset securitizations enable banks – as well as other sponsor-originators of securitizations – to obtain cash for assets transferred to another entity. In a typical securitization, the bank that originates loans sets up a special purpose entity, or SPE, that issues securities to third parties. The SPE uses the proceeds from its securities issuance to purchase the loans from the bank. Securitizations can provide benefits to banks by enabling them to diversify their asset holdings, by extending their capital base to wholesale money markets, and by enabling them to increase regulatory capital by keeping off-balance sheet assets and debt that otherwise would be on-balance sheet. Banks usually retain an interest in the SPE representing a subordinated interest in the SPE's assets to mitigate the information asymmetry regarding the riskiness of the transferred assets. Because SPEs are typically passive trusts, banks also usually manage the assets for their SPEs.
Growth in the securitization market was a major contributing factor to the growth in the credit markets that enabled the housing boom that went bust for two reasons. First, asset securitizations permitted banks to take advantage of regulatory arbitrage that enabled them to originate more and riskier loans than would have been the case otherwise, that is, to engage in excess lending. Regulatory arbitrage results when banks exploit regulatory capital rules to reduce the amount of capital they are required to hold against their assets, e.g., by transferring financial assets to SPEs (Acharya and Richardson, 2009a).
Second, bank investors and other SPE investors may not have exercised appropriate market discipline when banks originated loans and after the loans were securitized. Regarding bank investors, they had less incentive to monitor the quality of loans banks originated that would be securitized because securitizations transferred much of the loans' risk to SPE investors. In addition, bank investors had limited information about the quality of the bank's retained interest in the securitized loans. Regarding SPE investors, which include banks that purchased or retained interest in SPEs, although they had some information about the quality of loans at the date of the securitization, they had limited information subsequently. That is, bank and SPE investors not only had difficulty evaluating the quality of loans banks originated and securitized, but also had difficulty evaluating the fair value and risk of the securitized assets after the securitization transaction. The latter difficulty was more pronounced the farther the SPE's assets were removed from the assets underlying the securitizations. For example, consider four SPEs, SPE1 through SPE4. Assume SPE4 holds securities issued by SPE3, which in turn holds securities issued by SPE1 and SPE2. Assume SPE1 holds securitized loans and SPE2 holds credit card receivables. Because SPE4 is further removed from the underlying securitized assets held by SPE1 and SPE2, it is more difficult for investors in SPE4 to determine the value and riskiness of their investment than it is for investors in SPE3.
An additional challenge bank investors face is that for reputational or other reasons, the bank may assume more risk related to the SPE's assets than is required by the securitization agreement. If this occurs, then the bank's risk of loss exceeds the value of its retained interest. As an example of a bank assuming such additional risk, in December 2007 Citigroup voluntarily reacquired assets with a value $49 billion from several of its SPEs by assuming liabilities of approximately $87 billion (Amiram et al., 2010).
Before and during the Financial Crisis, financial reporting for asset securitizations in the USA was specified in Statement of Financial Accounting Standards (SFAS) No. 140 Accounting for Transfers and Servicing of Financial Assets and Extinguishments of Liabilities and Financial Interpretation 46R (FIN 46R). SFAS 140 provides that an asset transfer be accounted for as a sale – with assets derecognized from the originating bank's statement of financial position – if a set of guidelines is met, including legal isolation of transferred assets and no explicit recourse provisions. If the transaction fails to meet this set of guidelines, the bank recognizes the proceeds received as debt and separately presents the assets as pledged as security for the debt. FIN 46R provides guidance on whether the SPE's financial statements are to be consolidated with those of the bank. To obtain the benefits of securitizations described earlier, banks structured most securitizations to obtain sale accounting treatment and to avoid consolidation of the SPE.
SFAS 140 required banks that transfer assets to SPEs via securitizations that qualify as sales to disclose separately, among other things, the total principal amount of financial assets that the bank managed on behalf of its SPEs, gains from securitizations during the year and the carrying amount of any retained interest. SFAS 140 did not require disclosure of the fair value of retained interest, the fair value of assets and liabilities of its SPEs, or information to enable bank investors to assess the risk related to the bank's retained interest. For example, the disclosures did not include the extent to which the retained interest was subordinated to other SPE investors' claims to the SPE assets. Further, SFAS 140 disclosures contained no information regarding implicit agreements between the bank and its SPEs for the bank to guarantee performance of the securitized assets because implicit recourse plays no role in SFAS 140's sale accounting requirements. Thus, SFAS 140 disclosures were of limited usefulness to bank investors to assess any risk related to the SPE's assets in addition to risk the securitization agreement required the bank to bear.
Despite the lack of complete transparency of securitizations provided by accounting disclosures, empirical research provides evidence that investors include the information provided in the disclosures when assessing the value and risk of securities issued by banks engaging in asset securitizations. Using managed assets and liabilities as proxies for SPE assets and liabilities, Landsman et al. (2008) provide evidence consistent with equity investors viewing SPE assets and liabilities as belonging to the sponsor-originators, including banks. Niu and Richardson (2006) provide evidence that off-balance sheet debt related to asset securitizations has the same risk relevance as recognized debt for explaining market measures of equity risk, that is, the Capital Asset Pricing Model beta. Chen et al. (2008) show that equity investors consider the characteristics of the assets securitized when assessing the extent to which banks retain equity risk associated with securitized assets. Cheng et al. (2008) find that banks that undertake securitizations have higher information uncertainty, that is, higher stock price bid–ask spreads and analyst forecast dispersion, compared to banks without such transactions, and that information uncertainty increases with the amount of the securitized assets. Barth et al. (2009) provide evidence that the bond market views the risk of the assets securitized by the banks as residing with the banks, which is consistent with much of the equity market evidence. However, in marked contrast, the study provides evidence that when assessing bank credit risk it appears that credit rating agencies largely ignore asset securitizations, that is, the study finds an association between asset securitizations and credit ratings only for retained interest in mortgage loans.
Although the research shows that investors include the information provided in the disclosures when assessing the value and risk of securities issued by banks engaging in asset securitizations, the question remains whether their assessments would differ if they had more complete information. For example, although Barth et al. (2009) find the bond market views the risk of the assets securitized by the banks as residing with the banks, it is possible that the risk of such assets resides with the banks' SPEs. More generally, the extent to which the risk of such assets resides with the banks may differ from the bond market's assessment, which is based on information that is not fully transparent. It is also important to note that even if investors properly assessed the risks that securitized assets posed to banks, this does not imply that market discipline was adequate to prevent excessive lending by banks. For example, investors could have understood correctly that banks would not bear the risk of potential non-performance of assets transferred to SPEs because third parties – for example, taxpayers – would make up for any cash flow shortfalls, and therefore could have concluded that, from their perspective as investors, the amount of lending by banks was not excessive. Despite the fact that investors could have been exercising market discipline as it related to their investments, systemic excessive lending could have occurred because of the effects of other incentives banks faced.
It is possible that more transparent financial reporting by banks would have mitigated the extent of the housing boom and subsequent bust that precipitated the Financial Crisis. In the wake of the Financial Crisis, both the FASB and IASB have taken steps to improve financial reporting requirements relating to asset securitizations. In June 2009, the FASB issued SFAS 166 Accounting for Transfers of Financial Assets, which amends SFAS 140, and SFAS 167Amendments to Financial Interpretation (FIN) No. 46(R). As a result of these two standards, one would expect that most SPE assets and liabilities would now be consolidated with those of the originating bank, bringing the frequency of consolidation of SPEs under US GAAP more in line with that under IFRS. In addition, US banks are now required to provide enhanced disclosures to investors, including fair values of their SPEs' assets and liabilities. The required disclosures also include qualitative and quantitative information regarding the bank's continuing involvement with the securitized assets that provides investors with information to assess the reasons for the continuing involvement and the risks related to the assets to which the bank continues to be exposed.
The IASB has also taken steps to improve the financial reporting requirements relating to asset securitizations. In March 2009, the IASB issued an Exposure Draft, Derecognition, Proposed Amendments to IAS 39 and IFRS 7 (IASB,2009a). The Exposure Draft proposes to retain the basic approach to derecognition in IAS 39, which has elements in common with SFAS 166. The Exposure Draft also offers an alternative approach, which would require banks and other transferors to derecognize the transferred assets and recognize as assets and liabilities all the rights and obligations either retained or obtained in the transfer, including forward contracts, puts, calls, guarantees or disproportionate involvement with respect to the transferred cash flows, at their fair values. The approach would result in recognizing these rights and obligations as if the transferred assets had not previously been owned. The subsequent accounting for these newly recognized assets and liabilities would be determined by their nature – for example, derivatives obtained in the transfer would be measured subsequently at fair value.
This alternative approach differs from the approaches in IAS 39 and SFAS 166 in that those standards require transferors to make a determination as to whether the securitized assets have been sold. In contrast, there is no sale notion in the Exposure Draft's alternative approach. In our view, because the alternative approach results in banks recognizing the assets and liabilities they have after the transfer, the approach better reflects the underlying economics of the securitization transaction and therefore is likely to result in more transparent financial reporting. The sale or secured borrowing alternatives in IAS 39 and SFAS 166 rarely would capture the economics of these transactions because the transactions are rarely pure sales or pure borrowings. Regarding disclosures, in December 2008 the IASB issued an Exposure Draft, Consolidated Financial Statements (IASB, 2008b), which together with theDerecognition Exposure Draft, proposes expanded disclosures about asset securitizations that are similar to those in SFAS 167.
Although accounting standard setters are in the process of improving the transparency of information relating to assets securitizations, thereby enhancing the effectiveness of the market discipline pillar of prudential supervision, bank regulators can also take steps to improve transparency. In addition, to ensure securitizations do not expose banks to excessive risk, bank regulators can alter the ways in which securitized assets are used to calculate regulatory capital and in their supervisory review of banks. Going forward, it is important for bank regulators to work with accounting standard setters to help ensure information relating to asset securitizations provided in bank financial statements contributes to market discipline by complementing the information necessary for prudential bank supervision.
Derivatives were also at the center of the Financial Crisis because they magnified the risk to which investors in derivatives and equities of firms holding derivatives were exposed. The credit default swap (CDS) contracts sold by AIG are a case in point. A CDS is essentially an insurance contract on a bond. For example, if a bank buys corporate debt, it can enter into an agreement with a counterparty, e.g., AIG, to receive payment of the principle in the event of default. In the years leading up to the Financial Crisis, hedge funds and investment banks expanded their demand for insurance to cover their holdings in asset-backed debt securities issued by SPEs arising from the growth in the securitization market. In response to this demand, AIG wrote CDS contracts without offsetting the risk by buying CDS protection for itself from another counterparty, that is, AIG wrote ‘naked’ CDSs. Credit default swaps written by AIG covered more than $440 billion in bonds, most of which was related to debt issued by Lehman Brothers. Because AIG was unable to honor its CDS contracts when Lehman Brothers defaulted on its debt, the US government gave financial support to AIG. The government did so because of concern that if AIG defaulted on its CDSs with investment banks, commercial banks and hedge funds, the entire financial system would collapse. Had AIG defaulted, each bank/hedge fund would have had to sell other assets to meet its own obligations, thereby fueling a market-wide downward pressure on asset prices.
The key issue for us is not whether AIG should have been permitted to do this, or more generally whether derivatives markets should be regulated. This is another debate. Rather, we are interested in the question of whether and to what extent lack of transparency of accounting information contributed to the risk to which the financial system was exposed by derivatives. In other words, did investors have sufficient information to provide adequate systemic market discipline by pricing appropriately the derivatives and equities of firms engaging in derivatives contracts. More specifically, did investors understand the risks CDSs posed to AIG and its counterparties?
IFRS and US GAAP have similar accounting standards relating to derivatives. The two primary standards relating to recognition and measurement are IAS 39 Financial Instruments: Recognition and Measurement and SFAS 133 Accounting for Derivative Instruments and Hedging Activities. These standards require recognition of most derivatives at fair value on the statement of financial position. Whether gains and losses on derivatives are recognized in profit or loss or as part of other comprehensive income depends on whether and how the derivatives are used to hedge financial risks. The two primary standards that were in effect at the outset of the Financial Crisis relating to disclosure of derivatives were IAS 32 Financial Instruments: Disclosure and Presentation and SFAS 133. Although there are differences between the standards, each requires disclosure relating to derivatives' gains and losses, fair values at the end of the reporting period and the purposes for which the derivatives are held. Effective for calendar year 2007, IFRS 7 Financial Instruments: Disclosures replaced the disclosure requirements in IAS 32 and requires disclosure of qualitative information relating to financial instruments' liquidity, credit and market risks.
We believe that the recognition and measurement standards for derivatives greatly improved the transparency of banks' financial statements. Findings in Ahmed et al. (2006) provide empirical evidence of such improvement. Under prior standards, it was difficult if not impossible for investors to assess the value of firms' derivatives positions. For example, because many derivatives have negligible cost, basing recognition and measurement on a modified cost approach would provide investors with little to no information about the value or the risk of a bank's derivative positions. However, although in our view fair value is the most relevant single summary measure of derivatives, it is impossible for a single estimate of a single measure to provide all the information investors need to assess the value of such instruments and the risk the instruments pose to banks. The disclosures required by IFRS and US GAAP are meant to address the needs of investors by providing additional information about derivatives' fair value estimates and risks.
Although the derivatives disclosures likely are informative to investors, their usefulness is limited by the extent to which the information is aggregated. For example, AIG's 2006 annual report provides a table listing aggregate notional amounts for interest rate swaps, credit default swaps, currency swaps and other derivatives, disaggregated by maturity. Because the amounts are aggregated, financial statement users cannot determine which side of each contract AIG holds or who the counterparties are and therefore the risk the contracts pose. In addition, notional amounts are simply the basis on which derivatives are contracted. Notional amounts bear no direct relation to the value or risk of derivatives.
Extant research on the relevance of derivatives disclosures is not extensive (and largely predates SFAS 133), perhaps in large part because, as described above, it is difficult to use the disclosures to operationalize measures of risk for determining whether the disclosures contain risk-relevant information. Wong (2000) attributes the inability of foreign exchange risk disclosures to help investors assess the sensitivity of equity returns to currency fluctuations to inadequacy of the disclosures, particularly because of the level of aggregation of notional amounts. Despite these inadequacies of the disclosures, for a sample of US banks, Venkatachalam (1996) shows net notional amounts of derivatives are value relevant incremental to other bank assets and liabilities.
To understand better the risks posed by derivatives to banks, investors need to know the nature of the risk relating to the underlying instrument that drives the value and risk of the derivative. Certainly, investors can better assess these risks if they know which side of each contract a bank holds and who the counterparties are. IFRS 7 moves in the direction of helping investors assess the risk of derivatives by requiring sensitivity of the fair values of financial instruments, including derivatives, to reasonably possible changes in relevant market risk variables. For example, in the case of an interest rate swap, a bank would provide information regarding the effect of interest rate changes on changes in the value of the swaps.
Another potentially informative disclosure would be to show what the effects on the statement of financial position would be of implementing a ‘risk-equivalence’ approach to accounting for some classes of derivatives. Under a risk-equivalence approach, the implied gross assets and liabilities associated with derivatives contracts that, on net, have risk equivalent to the derivatives are separately identified. This permits investors and other financial statement users to gain insight into the risk arising from the leverage inherent in derivatives.
For example, consider a bank that sells a five-year credit default swap on a five-year B-rated loan with value of €2 billion. This CDS provides that the bank receives compensation for the credit enhancement it provides to the buyer of the CDS. The enhancement is the difference between the B credit rating of the underlying loan and the credit rating of a loan with the same terms issued by the bank, which presumably is higher than B. Assuming the CDS guarantees the full value of the loan and is properly priced, the amount of cash the bank receives from the CDS buyer would be the difference between the interest on the B-rated loan and the interest the bank would have to pay on a €2 billion loan with the same terms. The risk of the CDS to the bank is equivalent to it borrowing €2 billion for five years and buying the underlying B-rated loan. To see this, note that if the bank entered into these two transactions, the net of the interest cash flows it would receive exactly equals the cash flow it would receive under the CDS contract. In addition, the principal payments exactly offset. If the B-rated loan is repaid, under the CDS contract the bank will make no payments to the CDS holder. Under these risk-equivalent transactions, that is, the bank buying the B-rated loan and borrowing the same amount, the bank would use the proceeds from the B-rated loan to repay its own debt of the same amount. If the B-rated issuer defaults and the loan is worthless, the bank would pay the principal of the B-rated loan to the CDS holder; under the risk-equivalence approach, the bank's B-rated loan asset would be worthless but the bank would be required to repay the principal of its own borrowing.
An advantage of a risk-equivalence approach is that it permits financial statement users to understand the implied economic leverage position of the bank. To see this, let us return to the CDS example and assume that credit deterioration of the issuer of the B-rated loan guaranteed by the bank resulted in the loan decreasing in value to €1.5 billion at the end of the year of issuance. Assuming no change in credit worthiness of the bank, this implies the CDS has a fair value of €0.5 billion, which would appear as a liability on the bank's statement of financial position. However, without additional disclosure, financial statement users cannot assess fully the economic leverage position of the bank, which is equivalent to holding an asset – the underlying loan – worth €1.5 billion and having debt of €2 billion. That is, the €0.5 billion recognized CDS liability suggests that the bank's exposure is €0.5 billion, when in reality its risk of loss is €2 billion.
Although a risk-equivalence approach provides useful information to financial statement users in assessing the risk associated with CDSs and other derivatives, we do not advocate recognizing the risk-equivalent assets and liabilities. This is because such assets and liabilities are not literally assets and liabilities of the bank and, consequently, do not meet the definitions of assets and liabilities in US GAAP and IFRS. We do, however, advocate disclosing this information in the financial statements. Moreover, as the example illustrates, to permit financial statement users to assess fully the riskiness of the bank's leverage position, it is important to disclose the fair values of the risk-equivalent assets and liabilities at each financial statement date. Simply reporting the notional amounts of the derivatives is not adequate. Note that our suggestion to provide disclosure of the fair value of risk-equivalent assets and liabilities applies to all derivatives, including those arising from asset securitizations as described in Section 4.
As with asset securitizations, although accounting standard setters are in the process of improving the transparency of information relating to derivatives, bank regulators can also take steps to improve transparency as well as incorporate the effects of derivatives in their calculation of regulatory capital. For example, regulators can calculate regulatory capital using risk-equivalent gross assets and liabilities, rather than the derivatives themselves, if the regulators believe doing so is more useful for bank supervision.
6. Loan Loss Provisioning
As noted above, on average, loans comprise a significant proportion of bank assets, and therefore banks' financial reporting for loans, particularly loan loss provisioning, was also central to the Financial Crisis. Loan loss provisioning may have contributed to the Financial Crisis through its effects on procyclicality and on the effectiveness of market discipline. Recognizing losses is naturally procyclical and provides information to markets about loan values. The extent to which loan loss provisioning is procyclical, natural or amplified, and provides information depends on how provisions are determined.
During the Financial Crisis, loan loss provisioning under US GAAP and IFRS was based on an incurred loss model. Under an incurred loss model, banks do not recognize a provision for a loan loss until there is objective evidence the loan has been impaired. As a result, a bank would not necessarily recognize losses based on external indicators of economic loss in the value if its loans, for example, the bursting of the housing bubble, even though such indicators suggest that a substantial number of borrowers will default on their loans. The incurred loss model is not as procyclical as other loss models discussed below because it delays loss recognition. In addition, it can only be procyclical during economic downturns because under this model loans are only written down, not up. However, to the extent that financial markets rely on financial reporting information when making capital allocation decisions, such delayed and asymmetric recognition of losses potentially deprives the markets of timely information regarding the value of bank assets. Thus, the incurred loss model can reduce the effectiveness of market discipline.
One approach to mitigate the effects of delayed and asymmetric recognition of losses associated with the incurred loss model is to implement an expected loss model for loans. Under this model, a bank would reflect in its loan loss provisions all changes in expected future cash collections from its loans, including expected increases as well as decreases. Such a model is currently under consideration by the IASB (IASB, 2009b). A beneficial effect of the model is more timely and symmetric loss recognition, which can enhance market discipline because markets have more timely information about loan values. However, an effect of more timely information is an increase in natural procyclicality relative to that associated with the incurred loss model. A shortcoming of the expected loss model is that it does not change the discount rate used in calculating the present value of the expected cash flows to reflect changes in interest rates. As a result, it does not fully reflect the value of expected future cash collections, which makes the information provided to financial markets about the value of bank loans incomplete.
The fair value model overcomes this shortcoming of the expected loss model because a bank would reflect in its loan loss provisions not only changes in expected future cash collections, but also changes in the discount rate the market would apply to those cash flows. Changes in discount rates arise from changes in general market rates of interest, changes in credit risk and changes in the price of credit. Thus, relative to the incurred loss and expected loss models, the fair value model is the most effective from the standpoint of market discipline because it provides financial markets with the best information about loan values. As with the expected loss model, an effect of timely and symmetric loss recognition is an increase in natural procyclicality relative to that associated with the incurred loss model. Whether the fair value model results in a greater procyclicality relative to the expected loss model is difficult to determine because ignoring changes in discount rates when determining loan loss provisions as in the expected loss model can result in recognized loan amounts that are higher or lower than loan values. More importantly, it is not possible to determine the extent to which measuring loans at fair value would have amplified procyclicality during the Financial Crisis. First, as noted in Section 3, loans were not measured at fair value for financial reporting purposes. Second, it is unknown what prudential filters bank regulators would have applied to the fair value amounts.
Regardless of the measurement model used for loan loss provisioning, some have advocated – notably the Bank of Spain – overlaying a ‘through-the-cycle’ adjustment to the provision amounts. This overlay, also referred to as ‘dynamic provisioning’, would effectively increase loan loss provisions in economic upturns, and decrease them in economic downturns. The goal of dynamic provisioning is to mitigate natural procyclicality by creating a capital cushion for banks during economic upturns that can be used to buffer capital declines during economic downturns. To the extent that this adjustment creates or uses a capital cushion that is not reflective of loan values, dynamic provisioning is unlikely to be acceptable for financial reporting purposes. However, such an adjustment could be an effective regulatory tool to address amplified procyclicality.
We scrutinize the role that financial reporting for fair values, asset securitizations, derivatives and loan loss provisioning played in contributing to the Financial Crisis. Because banks were at the center of the Financial Crisis, we focus our discussion and analysis on the effects of financial reporting by banks. We begin by discussing the objectives of financial reporting and bank regulation to help clarify that information standard setters require firms provide to the capital markets and information required by bank regulators for prudential supervision will not necessarily be the same. This distinction is important to understanding why financial reporting played a limited role in contributing to the Financial Crisis.
We analyze the way in which financial reporting for fair values, asset securitizations and derivatives potentially contributed to the Financial Crisis. For each topic we summarize the financial reporting requirements of US GAAP and IFRS, offer insights into whether the information available to investors was sufficiently transparent to make appropriate judgments regarding the values and riskiness of affected bank assets and liabilities, and summarize available research evidence. Because loans comprise a large fraction of bank assets, we also discuss how loan loss provisioning may have contributed to the Financial Crisis through its effects on procyclicality and on the effectiveness of market discipline.
Our analysis leads us to conclude, as have others, that contrary to what many critics of fair value contend, fair value accounting played little or no role in the Financial Crisis. However, transparency of information associated with measurement and recognition of accounting amounts relating to, and disclosure of information about, asset securitizations and derivatives likely was insufficient for investors to assess properly the values and riskiness of affected bank assets and liabilities. The FASB and IASB have taken laudable steps to improve disclosures relating to asset securitizations by requiring enhanced disclosures, including fair values of securitized assets and liabilities and qualitative and quantitative information regarding a bank's continuing involvement with the securitized assets, to enable market participants to assess the risks related to the assets to which the bank is exposed. However, in our view, the alternative approach for accounting for securitizations in the IASB's Exposure Draft that would require banks to recognize whatever assets and liabilities they have after the securitization is executed better reflects the underlying economics of the securitization transaction and therefore, coupled with enhanced disclosures about derivatives, is likely to result in more transparent financial reporting.
For derivatives, our recommendations include disclosure of disaggregated information to permit investors to know which side of the contracts a bank holds and who the counterparties are, and disclosure of the sensitivity of the fair values of derivatives – as well as other financial instruments measured at fair value – to changes in relevant market risk variables. We also recommend implementing a risk-equivalence approach to enhance disclosures relating to the leverage inherent in derivatives.
Finally, we conclude that because the objectives of bank regulation differ from the objective of financial reporting, changes in financial reporting requirements to improve transparency of information provided to the capital markets likely will not be identical to the changes in bank regulations needed to strengthen the stability of the banking sector. Moreover, bank regulators have the power to require whatever information is needed to meet the objective of prudential supervision. We conclude that it makes sense from the standpoint of efficiency for accounting standard setters and bank regulators to find some common ground. However, it is the responsibility of bank regulators, not accounting standard setters, to determine how best to ensure the stability of the financial system.