The IFRS option to reclassify financial assets out of fair value in 2008: the roles played by regulatory capital and too-important-to-fail status

The policy fixes included a call for increasing bank capital ratio requirements, restrictions on investments and trading, particularly derivatives, and changes in disclosure requirements.

There is a third category of investments recognized at fair value, namely those investments that a firm elects to recognize at fair value based on application of the fair value option (FVO). As with HFT investments, fair value gains and losses are recognized in income. An important difference between investments in this third category and HFT investments is that the former are precluded by IFRS from being reclassified into AFS or LAR/HTM. Although investments recognized at fair value based on application of the FVO cannot be reclassified, the extent to which a bank makes use of the FVO can, in principle, affect our inferences. We address this issue in section 6.1.

During our sample period and in all our sample countries, the Tier 1 capital ratio had to be at least 4%. In practice, banks typically maintain a target capital position above the required minimum as protection against economic shocks that would reduce their Tier 1 capital below the 4% level. The Financial Crisis represented an economic shock of a magnitude unprecedented in modern times. In such a situation, with heightened volatility and fall in asset prices, bank managers likely changed their beliefs as to what constituted an adequate Tier 1 capital cushion.

A similar situation occurred in the United States in the 1990s, when pressure from technology firms, primarily in California’s Silicon Valley, resulted in the Financial Accounting Standards Board (FASB) changing the requirement to recognize employee stock expense in income. The resulting standard, SFAS 123 (FASB 1995), which required only disclosure of earnings adjusted for the effects of employee stock option expense, was applicable to all firms, not just those in the technology sector.

Although IAS 39 does not define “foreseeable future,” it is reasonable to assume that it would have constrained banks that reclassified assets from making sales out of the LAR/HTM in early 2009. We have examined the 2009 financial statements of our sample banks that reclassified investments and found very few appeared to sell investments in 2009. For example, by the middle of 2009, only four sample banks had sold or settled investments from any category. Therefore, hindsight does not provide evidence that banks regarded the “foreseeable future” selling constraint as being without force.

Such bailouts took place in many countries. In the United States, Congress passed a $1 trillion bank bailout that resulted in taxpayer funds being used to prop up America’s largest banks including Bank of America, Citibank, and Goldman Sachs. In the United Kingdom, Northern Rock was nationalized, and the Royal Bank of Scotland received bailout funds. In many other European countries, notably Ireland, private bank debt was essentially guaranteed by the government.

For example, in the United States, TITF banks gambled with taxpayers’ money by originating loans to risky borrowers and purchasing asset-backed securities secured by subprime loans (e.g., Gorton and Metrick 2012; Gorton and Souleles 2006; Greenspan 1998).

In the United States, such rule changes include recognition of derivatives at fair value (SFAS 133, FASB 1998), a measurement standard for fair value (SFAS 157, FASB 2006), and a standard giving firms the option to measure certain assets at fair value (SFAS 159, FASB 2007). With minor variations, the IASB largely followed suit by issuing standards.

Regulators in some countries—in which distressed banks previously had not been allowed to fail—nevertheless let some banks fail during the Financial Crisis. For example, this was true in Russia. The critical question is what rational beliefs bank managers were likely to have held when deciding whether to take advantage of the option to reclassify investments. It is thus important to avoid introducing ex post bias by classifying a bank as TITF based on bank failures after the reclassification decision.

We assume that bank managers act on behalf of equity investors when making these decisions. That is, with regard to the investment classification decision, we assume there is no substantive goal incongruence between the two groups. Our rationale for this assumption is based on the fact that the compensation of bank managers was largely comprised of bonuses and equity-based instruments (Becher et al. 2005; Chen et al. 2006; Fahlenbrach and Stulz 2011). Rather, bank managers and their equity investors are potentially in conflict with regulators and taxpayers.

Inferences do not change when using a logit model. We also considered, but disregarded, using a linear probability model, because we encountered the difficulty of obtaining fitted probability estimates outside of the [0,1] interval. In particular, although there are no fitted estimates in excess of 1, 7% are less than 0, with a largest negative value of −26%. For this reason, all econometric textbooks caution against the use of ordinary least squares estimation when the dependent variable is discrete (e.g., Wooldridge 2009, 574–575).

We considered three additional cutoff points: 0%, 2.5%, and 10% of the ex ante HFT and AFS portfolios. Untabulated statistics from analyses based on these other cutoff points result in similar inferences to those based on tabulated findings.

The magnitude of the interaction effect in nonlinear models does not equal the marginal effect of the interaction term. Therefore we use the Ai and Norton (2003) correction to estimate the magnitude and standard errors of the interaction, TITF×Pre_Tier1.

We also used two cutoff points of 5%, which reduces the number of such banks from 19 to 10, and 15%, which increases the number to 33. Estimations of equation (1) based on these alternative cutoff points yield similar inferences to those based on tabulated findings.

We re-estimated equation (1) eliminating banks in countries where banks were subsequently permitted to fail (e.g., Russian banks). We also re-estimated equation (1) defining TITF banks as either TBTF or domiciled in a no-fail country in which no bank was subject to government intervention in 2008. Redefining TITF banks in this way allows for the possibility that bank managers in no -fail countries might have revised their expectations regarding the likelihood of regulatory forbearance when another bank in their country was subject to government intervention in 2008 (e.g., the United Kingdom and Switzerland). Untabulated findings from both analyses reveal no change in our inferences relative to those based on tabulated findings.

Reclassifications occurred in the second half of 2008. This suggests that the best we can do is measure the determinants in June 2008 by using mid-year financial statements. However, doing so would result in a substantial reduction in sample size because either mid-year financial statements are not available or Tier 1 capital disclosures are not included in mid-year financial statements for a substantial number of sample banks. Nonetheless, as a sensitivity analysis, we re-estimated Eq. (1) using Tier 1 capital as of June 2008, which reduces the sample from 160 to 134 observations. Untabulated findings yield inferences similar to those based on tabulated findings.

In addition to the use of Size as control variable, we allow for the possibility that bank size is nonlinearly related to RECLASS. We re-estimate Eq. (1) by replacing TITF with an indicator variable for whether a bank’s assets are above or below the sample median. Findings from this additional test in Section 6.1 suggest that our inferences are not driven by bank size per se.

Excluding the country indicator variables does not affect our inferences.

IFRS are required to be applied by European entities when preparing consolidated financial statements. In principle, financial statements of a holding company’s subsidiaries can be prepared using domestic accounting standards of the countries in which they operate. This raises a question of whether accounting choices by a bank’s subsidiaries can differ from those of the holding company and, if so, whether this can affect inferences from our tests as they relate to the effect of prudential filter regulations on accounting choices. Although bank subsidiaries are subject to the bank regulations of the country in which they operate, holding companies are subject to the financial reporting and prudential filter rules only of the country in which their consolidated financial statements are filed. A subsidiary bank operating in a country with different prudential filter rules than those its parent company faces could lead to accounting choices at the subsidiary level differing from those the consolidated level. However, capital adequacy only matters at the consolidated level because any actions taken by the holding company to address capital needs at the subsidiary level are eliminated during the consolidation process (IAS 27, paragraph 25, IASB 2005). Therefore, our TITF and prudential filter classifications are unaffected by whether bank holding companies’ subsidiaries operate in countries other than the one in which the holding company files financial statements.

In several European countries (e.g., Austria, Denmark, Germany, Hungary, and Norway), banks have the option to use domestic accounting standards instead of IFRS for regulatory reporting. However, none of the banks in our sample uses this option.

The large relative size of the banking sector in several TITF countries raises the question whether the countries could afford to rescue their banks in the event of a systemic failure of the sector. A maintained assumption in this study is that, at the time TITF bank managers were making the reclassification choices, they assumed that they would be rescued in the event of failure. If this maintained assumption does not apply to bank managers in particular countries, predictions regarding the effects of TITF status are less likely to be borne out in our tests.

The banking sector in many countries also includes nonlisted banks, which are not included in our sample.

As noted earlier, banks typically maintain Tier 1 capital ratios in excess of the 4% minimum. The fact that both TITF and non-TITF banks typically had pre-reclassification ratios above 4% does not imply that bank managers regarded the cushions to be sufficient in the volatile conditions of the Financial Crisis.

Throughout the paper, we use a 5% significance level under a one-sided alternative when we have a signed prediction and under a two-sided alternative otherwise.

Ai and Norton (2003) notes that interpretation of coefficients on interaction variables in nonlinear models is problematic. The marginal effects reported in the table are averages of marginal effects evaluated at each data point. The study suggests focusing instead on evaluating marginal effects at points on the distribution, for example, at the mean or median. Marginal effects evaluated at these two points yield the same inferences as those based on the tabulated averages. Specifically, marginal effects calculated at the median (mean) are 3.79 (5.57), with t-values of 2.45 (2.66).

Another issue relating to the interaction of TITF and Pre_Tier1 concerns the possibility that the marginal response of a bank when making its reclassification choice is likely to differ substantially if its regulatory capital is above or below a critical threshold. We therefore estimate an alternative version of Eq. (1) in which we replace the continuous Pre_Tier1 variable with a dichotomous, LOW_Pre_Tier1, that equals one if Pre_Tier 1 capital is below the sample median and zero otherwise. Untabulated findings reveal that the LOW_Pre_Tier1 coefficient is significantly positive (coefficient = 1.279; z-statistic = 4.63), and the marginal effect, 0.350, is significantly positive (z-statistic = 5.09). This finding suggests that non-TITF banks with low regulatory capital are 35% more likely to reclassify than non-TITF banks with high regulatory capital. In addition, the coefficient on the interaction of TITF and LOW_Pre_Tier1 is significantly positive (coefficient = 1.254; z-statistic = 2.97), and the marginal effect, −0.343, is significantly negative (z-statistic = −2.92). A test for the sum of the coefficients on LOW_Pre_Tier1 and the interaction of TITF and LOW_Pre_Tier1 (+1.279–1.254) is insignificantly different from zero (χ²-value = 0.00; p-value = 0.94). This finding suggests that TITF banks with low regulatory capital are not more likely to reclassify than TITF banks with high regulatory capital.

Another identification issue that could affect our inferences regarding the effects of TITF status is the possibility that TITF banks are more likely than non-TITF banks to have unrealized fair value losses from the second to fourth quarter of 2008 that would be directly affected by retroactive reclassifications. Data limitations prevent us from directly addressing this issue because such information is not in the public domain. However, we re-estimated Eq. (1) including unrealized fair value gains or losses for the full fiscal year 2008. Untabulated findings indicate that the additional variable’s coefficient is insignificant and, more importantly, inferences regarding TITF status are the same as those based on the Table 5 findings.

Eleven percent of our sample banks are domiciled in the common law countries of Cyprus, Ireland, the Netherlands, and the United Kingdom.

For example, untabulated statistics reveal the 66 TITF banks that are not TBTF are significantly better capitalized, smaller, and more profitable than the 19 TBTF banks. The fact that the 66 banks are better capitalized and more profitable suggests that they have even less incentive to reclassify investments than the 19 TBTF banks.

Furthermore, untabulated statistics relating to total loan loss provisions, that is, the sum of discretionary and nondiscretionary components, reveal the similar inference that TITF banks did not make smaller loan loss provisions in 2008.