Heterogeneity of Bank Risk Weights in the EU
Evidence by Asset Class and Country of Counterparty Exposure1
Author: Rima Turk-Ariss

Contributor Notes

Author’s E-Mail Address: RTurk@imf.org

Concerns about excessive variability in bank risk weights have prompted their review by regulators. This paper provides prima facie evidence on the extent of risk weight heterogeneity across broad asset classes and by country of counterparty for major banks in the European Union using internal models. It also finds that corporate risk weights are sensitive to the riskiness of an average representative firm, but not to a market indicator of a firm’s probablity of default. Under plausible yet severe hypothetical scenarios for harmonized risk weights, counterfactual capital ratios would decline significantly for some banks, but they would not experience a shortfall relative to Basel III’s minimum requirements. This, however, does not preclude falling short of meeting additional national supervisory capital requirements.

Abstract

Concerns about excessive variability in bank risk weights have prompted their review by regulators. This paper provides prima facie evidence on the extent of risk weight heterogeneity across broad asset classes and by country of counterparty for major banks in the European Union using internal models. It also finds that corporate risk weights are sensitive to the riskiness of an average representative firm, but not to a market indicator of a firm’s probablity of default. Under plausible yet severe hypothetical scenarios for harmonized risk weights, counterfactual capital ratios would decline significantly for some banks, but they would not experience a shortfall relative to Basel III’s minimum requirements. This, however, does not preclude falling short of meeting additional national supervisory capital requirements.

I. Introduction

Holding capital against risk-weighted assets (RWA) rather than total assets is consistent with the greater risk sensitivity intended by the Basel framework. Such risk sensitivity moderates banks’ incentives to hold assets with high expected returns by requiring them to hold adequate capital to cover the underlying risk. Many banks use internal models to calculate risk, which rely on parameters that are largely based on historical data and previous loss experience. However, doubts regarding the resulting capital ratios have arisen and their usefulness has been questioned (Le Leslé and Avramova, 2012; Vallascas and Hagendorff, 2013; Behn, Haselmann, and Vig, 2016). Recent papers have documented a strategic underreporting of bank risk (Mariathasan and Merrouche, 2014; Begley, Purnanandam, and Zeng, forthcoming) and some supervisors went as far as alluding to regulatory arbitrage by banks.

The combined complexity and opacity of risk weights generated by each banking organization for purposes of its regulatory capital requirement create manifold risks of gaming, mistake, and monitoring difficulty.

Governor Daniel Tarullo, May 8, 2014 speech at the Federal Reserve Bank of Chicago Bank Structure Conference, Chicago, Illinois.

In the aftermath of the global financial crisis (GFC), the Basel Committee on Banking Supervision (BCBS) has been devoting great attention to strengthening the regulatory capital framework. Having improved the quantity and quality of capital that banks must hold to absorb losses, the BCBS is seeking to address “the issue of excessive variability in risk-weighted assets” to restore “market confidence in risk-based capital ratios” and “promote sound levels of capital and comparability across banks” (Bank for International Settlements (BIS), BCBS, 2016a).2

An issue at stake is that the proposed reforms to calculate bank risk (see section II) are expected to increase by a greater amount the capital that European banks must hold compared with, for example, their American peers (The Economist, December 3, 2016).3

This reflects, among others, different asset compositions on banks’ balance sheets4 and wider use of internal models in the EU relative to the U.S, which significantly reduce the ratio of RWA to total assets, the so-called RWA density. With many European banking sectors still suffering from the legacies of the GFC and the euro area crisis, there may be a reluctance to impose higher capital requirements that could increase bank costs.

A growing body of literature has examined issues related to risk weight heterogeneity. One feature common to all previous studies is the use of the RWA density as basis for their analysis, simply because disaggregated data at the bank portfolio level were previously lacking. Another common feature of existing research is the coverage of large listed banks across the world. For instance, Beltratti and Paladino (2016) use the RWA density to find that banks use internal models to optimize their financial structure, under the hypothesis that a larger cost of equity capital induces banks to reduce the share of equity financing of their assets. Other studies have looked at issues arising from the use of internal models other than regulatory arbitrage. Using a large sample of 246 international listed banks where less than 60 banks are from Europe, Vallascas and Hagendorff (2013) also use the RWA density to show that risk-weighted assets are ill-calibrated to portfolio risk so that banks under-report portfolio risk, which undermines their ability to withstand shocks. Mariathasan and Merrouche (2014) report that the RWA density across 115 banks from 21 OECD countries declines considerably once regulatory approval for using the internal ratings-based (IRB) methods is granted. Such a drop in risk weights is particularly pronounced for weakly capitalized banks, where supervision is weak, and in countries where supervisors are overseeing many IRB banks. More recent evidence by Begley, Purnanandam, and Zeng (forthcoming) using 41 banks in the U.S., Canada, and Europe, indicates that bank risk measures become less informative precisely when banks are approaching financial distress.

Other studies analyze banks’ risk assessment in a single country context. For the U.S., Barakova and Palvia (2014) find that internally generated risk weights are determined mostly by portfolio risk. But Plosser and Sanos (2014), who estimate bank biases at the credit level, report that low-capital banks have low risk estimates. As for evidence from Europe, Behn, Haselmann, and Vig (2016) use loan level data from Germany to show that internal models systematically under-predict actual default rates, that defaults and losses are higher for loans originated under the model-based approach and carrying low risk weights, and that banks had priced those loans in accordance with their higher actual risk. In contrast, Fraisse, Lé, and Thesmar (2015) do not find much support for corporate risk weight manipulation via internal models in France.

This paper differs from the literature in three important aspects. First, instead of using RWA density as proxy for banks’ risk assessment, risk weights are evaluated at the portfolio level and across country of counterparty exposure. Second, whereas existing papers have generally looked for evidence of gaming by banks to minimize holdings of equity capital, this study aims to document the extent of risk weight heterogeneity across different dimensions, assess the sensitivity of corporate risk weights to fundamentals, and analyze the implications from applying less heterogeneous risk weights on bank capital positions under hypothetical counterfactual scenarios. Third, the data used cover a large sample of listed and non-listed European banks, compared with existing evidence on international listed banks or in a single country. Such a focus on Europe is important given concerns about the expected effects from risk weight harmonization rules on the capital required for European banks.

The novelty of the paper derives from using data published as part of the transparency exercise by the European Banking Authority (EBA), which allow for a better understanding of the extent of risk-weight heterogeneity across Europe.5 Cross-country research is often handicapped by incomplete datasets on risk weights for private and publicly-listed banks or by availability of aggregate information at the institution level only. In contrast, the EBA data, for the first time, disclose three important dimensions of portfolio information that allow for the calculation of risk weights at a more granular level.6 First, the data provide the distribution of bank portfolios across the standardized and internal risk assessment methods. Second, the data allow for the calculation of bank risk weights across major asset classes including corporate, retail, and mortgage portfolios.7 Third, the data include the largest exposures by country of counterparty for each bank and each asset class.8 This detailed level of disclosure allows for prima facie evidence on variations in risk weights at the portfolio level, by asset class, and by country of counterparty exposure for the major banks using internal models in Europe.

In addition, the study investigates possible determinants of corporate risk weights, assessing their sensitivity to indicators of firm fundamentals and corporate default. It also presents hypothetical counterfactual capital ratios if more harmonized risk weights were applied to bank portfolios, using as benchmarks other banks’ risk weights for the same asset class (corporate/ retail/ mortgage) and country of counterparty exposure. The findings could inform policy discussions by regulatory bodies, which are currently seeking to reduce the complexity of internal models and improve their comparability, as well as addressing excessive heterogeneity in credit risk assessment methods.

The rest of the paper is organized as follows. Section II provides a brief background on the review of bank risk weights by international regulatory bodies. Section III describes variations in risk weights across bank portfolios and section IV presents their variations by SA/IRB approach, asset class, and country of counterparty exposure. Section V investigates the sensitivity of corporate risk weights to firm fundamentals using both accounting data and expected default frequencies. Section VI derives counterfactual capital ratios from hypothetical scenarios using alternative risk weights for banks’ exposures at default. Section VII concludes.

II. Regulatory Review of Bank Risk Weights

Having substantially strengthened the banking system’s regulatory framework, the Committee’s attention is now turning to the framework’s complexity and the comparability of capital adequacy ratios across banks and jurisdictions.

BIS, BCBS, The regulatory framework: balancing risk sensitivity, simplicity and comparability, July 2013.

Post-crisis reforms of the Basel capital accord have first focused on the numerator of the capital adequacy ratio (CAR), increasing both the quantity and quality of capital that banks must hold to increase their loss absorption capacity. More recently, the denominator of the CAR—RWA—has received more attention, with the issue of risk weight heterogeneity being at the forefront of discussions on the harmonization of banking rules.

RWA are derived using risk weights, which are expected to reflect the varying intrinsic risk characteristics of each asset, so that banks hold appropriate amounts of capital against them as a cushion to absorb future unexpected losses.9 Hence, some variation in risk weights across portfolios is to be expected given the differences in the financial profile of counterparties, domestic conditions, specific business policy by the banks etc. The Basel framework allows banks to use a range of methods to measure portfolio risk subject to supervisory approval. One option is to follow a standardized approach (SA) and the alternative is to use a bank’s own internal models subject to explicit supervisory approval.10 Whereas the SA uses prescribed risk weights to assess bank portfolio risk, the IRB approaches align risk weights more closely to sophisticated quantitative risk assessment techniques in the financial industry.11

Heterogeneity in bank risk weights also arises among internal models. IRB models require a number of key parameters, including probability of default (PD), loss given default (LGD), exposure at default (EAD), and effective maturity. These parameters are not always available and require calibration or the exercise of judgement, thereby generating different risk weights for the same assets. Adding to that, differences in risk weights produced by IRB models can arise from the exercise of supervisory judgment across jurisdictions.12 In sum, regulation and supervision allow for risk weights to differ across banks, which translates into different levels of capital ratios across those institutions. To the extent that such variations in risk weights do not reflect differences in risk exposure, the consequence is reduced comparability of capital ratios across institutions.

In Europe, use of internal models results in significant variation in capital ratios across countries, varying from 12.5 percent for Portuguese banks up to as high as 28 and 25 percent, respectively, for the Netherlands and Sweden (see Figure 1). Whereas a number of studies have analyzed how much bank capital is enough, Dagher et al. (2016) find that CAR of 15–23 percent in advanced economies would have avoided creditor losses in the past. Such a range is also in line with the 16–20 percent estimate by the Financial Stability Board (FSB) for global systemic banks (FSB, 2014) and the US Federal Reserve’s proposal of more than 18 percent in total loss-absorbing capacity (Board of Governors of the Federal Reserve System, 2015).

Figure 1.
Figure 1.

Capital Ratios for Banks in the EU (in percent), June 2015

Citation: IMF Working Papers 2017, 137; 10.5089/9781484302958.001.A001

Source: European Banking Authority.

In order to mitigate regulatory and model uncertainties from risk-based assessments the leverage ratio was introduced by Basel III as backstop to risk-sensitive capital.13 Some even contend that adopting a leverage ratio would “induce truthful risk reporting” (Blum, 2008).

The remarkable dispersion in RWA has prompted a review of their measurement by international regulatory bodies. As part of its regulatory consistency assessment programme (RCAP), the BIS has published two reports analyzing RWA for credit risk in the banking book. In 2013, it conducted a Hypothetical Portfolio Exercise for more than 100 major banks and 32 large banking groups in 13 jurisdictions to investigate the level and variation of risk weights and identify some of the primary drivers of this variation (BIS, 2013). This study also used surveys to consider differences in the practices of national supervisory authorities, including areas of national discretion permitted in the Basel framework, and differences in the internal estimation practices of banks. Its main conclusion was that observed variations in risk weights are driven by a mix of differences in underlying risk and differences in banking and supervisory practices.

In 2016, a second report from the BIS’s RCAP (BIS, 2016b) evaluated regulatory outcomes by examining variability in RWA (for loans to retail customers and small and medium-sized firms) and in exposure at default (across the entire banking book).14 The analysis compares PD, LGD, and EAD estimates (E) to actual (A) default and loss outcomes or the A/E ratio in the form of a “back-testing” exercise. The findings indicate that, on average, there is a close alignment of actual PD outcomes and IRB estimates but not for LGD and loss rates— suggesting that differences in RWA are based more on differences in risk rather than varying estimation practices. The dispersion of all A/E outcomes (for PD, LGD, and EAD) across banks, however, is similar. The report also describes sound practices for the independent model validation of banks, highlighting the potential to either reduce practice-based RWA variation or to simplify the IRB capital framework and increase its comparability.

Prior to that in December 2015, the BCBS had engaged in a review of standardized approaches to credit risk (BIS, 2015). The review sought to reduce differences in the way risk weights were calculated under the SA, which had implications for real estate exposures. In addition, the review suggested removing certain types of exposures (large corporates and financial institutions) from the IRB approach.

In parallel, the EBA has similarly committed to increasing the robustness of the risk-based capital framework for banks.15 In 2014, it performed a review of risk weights for residential mortgages to better understand risk weight sensitivity to key model parameters (EBA, 2014). The Single Supervisory Mechanism is also planning to review 7,000 IRB bank models over four years to ensure that internal models are “solid, credible, and consistent” (Nouy, 2015; European Central Bank (ECB), 2016).

In addition to the review of risk weights by the BIS and European authorities, the Basel Committee’s oversight body has agreed in January 2016 to complete its work on addressing the problem of excessive variability in RWA. In this vein, the BCBS started in March 2016 a consultation process for the setting of additional constraints on IRB models for credit risk, in particular through the use of floors. The proposal aims to (1) reduce complexity and improve comparability of IRB approaches, and (2) address excessive variability in capital requirements for credit risk (BIS, 2016c).16 The floor is meant to mitigate model risk measurement error from using IRB modeling, thereby enhancing the comparability and transparency of bank capital and ensuring its level does not fall below a certain level.17 The new proposed constraints on the IRB approaches would complement the design of a capital floor based on the SA, non-IRB approaches (BIS, 2014).18

The proposed regulatory capital changes may have a significant effect on European banks, which have been using internal models since they were first developed. Figure 2 panel A shows the extent of variation in RWA density across Europe and, from panel B, the average RWA density is considerably lower for European (35 percent) than for U.S. (58 percent) Global Systemically Important Banks (G-SIBs), although this large difference partly reflects higher shares of mortgages and government securities in European banks.19

Figure 2.
Figure 2.

Risk Weighted Assets Density (RWA/TA, in percent)

Citation: IMF Working Papers 2017, 137; 10.5089/9781484302958.001.A001

More recently in February 2017, the ECB began the implementation of the Targeted Review of Internal Models (TRIM) to assess whether internal models comply with regulatory requirements, and whether they are reliable and comparable (ECB, 2017a). A major objective of TRIM, which involves on-site missions to 68 banks in 15 countries stretching over 2019, is to reduce inconsistencies and unwarranted variability in risk weights. Whereas increases in RWA are not the intention, TRIM could either raise or lower the capital requirements for individual banks (ECB, 2017b).20

In March 2017, the EBA also published a report on the consistency of RWA for “high default portfolios”—which include residential mortgages, SME retail, SME corporate, and corporate-other portfolios—covering 114 institutions across 17 EU countries (EBA, 2017). The report, which calls for a cautious interpretation of the results, finds that RWA variability can be explained to a large extent by portfolio features, including the proportion of defaulted exposures in the portfolio, the country of the counterparty, and the portfolio mix. The remaining variability is likely attributed to idiosyncratic features, modeling assumptions, and risk management and supervisory practices.

III. Variation in Risk Weights Across Bank Portfolios

This section uses the detailed data from the EBA to describe heterogeneity in risk weights across EU banks. As part of its commitment to enhance transparency in the banking sector, the EBA published in November 2015 bank-by-bank information on capital positions and risk exposure amounts, using December 2014 and June 2015 as reference dates. This EBA EU-wide transparency exercise aims at making regulatory capital ratios a more transparent metric to assess banks’ financial strength. It provides detailed and comparable bank-level data for 105 banks across 21 European countries (representing around 70 percent of EU banking assets) both at the group level and for the largest ten countries of counterparty credit exposures.21 The granularity of the EBA data allows investigating bank risk weights along three dimensions: portfolio type (IRB and SA methods), asset class (corporate, retail, and mortgage exposures), and country of counterparty exposure (across 60 reported countries).

Appendix A explains in more detail how risk weights are inferred from the EBA data. They are calculated as the ratio of what the EBA terminology labels as “Risk exposure amount” (RWA using the BIS lexicon) to “Exposure value” (“exposures at default” under the BIS lexicon). Table 1 summarizes risk weights as of June 2015 across portfolio type (IRB/SA) and major three asset classes (corporate, retail, and mortgage loans), and Figure 3 presents basic charts on their distribution averaged at the country level.

Figure 3.
Figure 3.

RWA, IRB/SA Portfolio Decomposition, and Risk Weights

Citation: IMF Working Papers 2017, 137; 10.5089/9781484302958.001.A001

Table 1.

Median Risk Weights across IRB/SA Portfolios of EU Banks, June 2015

Significant differences in risk weights across IRB/SA portfolios and credit exposures.

article image
Sources: EBA and Fund staff calculations

At 85 percent of total RWA, credit risk is its largest bank risk component, followed by operational risk which amounts to 10 percent of RWA (see Figure 3). Except for banks in Denmark, Germany, Hungary, Sweden, and the U.K., the market risk share of RWA is less than 5 percent for banks in all other EU countries. Since credit risk is the dominant source of bank risk, the focus of the analysis in the rest of the paper is the credit portfolio split by corporate, retail, and mortgage credit exposures across both IRB and SA portfolios.22

Banks in the EBA sample from Cyprus, Hungary, Latvia, Malta, Poland, and Slovenia rely on the SA approach to assess credit risk. For EBA banks from other EU countries, the share of the credit portfolio assessed using the IRB method is lowest in Portugal at one-half, whereas in Finland and Sweden the IRB portfolio share is highest at 97 percent.

Other notable differences across bank credit portfolios are credit risk weights that, on average, are twice as high for SA than for IRB portfolios.23 The median IRB risk weight for banks in the EU is 34 percent as of June 2015, significantly below the SA median risk weight of 75 percent. Looking at country-level averages, IRB credit risk weights range from 22 percent in Sweden to close to 50 percent in Austria. For the SA, credit portfolio risk weights also exhibit some dispersion, varying between slightly less than 60 percent in Malta to 90 percent in Latvia.

Figure 4 presents the average risk weight by type of credit exposure across IRB/SA portfolios. For the corporate credit portfolio, IRB risk weights exhibit significantly more variability than SA RW. Banks in Denmark and Sweden apply the lowest corporate risk weights (on average 33 and 34 percent, respectively, of EAD) for their IRB corporate asset class, whereas average risk weight in Ireland and Portugal are highest at 69 and 80 percent, respectively.24 In contrast, SA corporate risk weights vary between 76 and 103 percent, respectively, for banks in France and Hungary.

Figure 4.
Figure 4.

IRB/SA Average Risk Weights by Credit Exposure

Citation: IMF Working Papers 2017, 137; 10.5089/9781484302958.001.A001

Cross-country heterogeneity in risk weights is also greater for the IRB than for the SA retail portfolios. In their IRB risk-based framework for capital adequacy, banks in Luxembourg use an average risk weight of 11 percent for their retail portfolio, whereas banks in Spain apply a 46 percent risk weight, although these differences again likely in part reflect differences in loan performance. In contrast, there is much less dispersion in risk weights under the SA across the EU at large, where the median risk weight for retail exposures is at 72 percent.

Finally, at 26 percentage points, the gap between the IRB and SA risk weights for mortgage exposures is, on average, narrower than for other types of credit exposures. Using IRB models, less than 10 percent of EAD in Finland and Sweden are subject to regulatory capital, whereas in Austria and Ireland risk weights are, on average, 25 and 32 percent, respectively.25 For the retail SA portfolio, banks in Latvia apply a risk weight of 35 percent, on average, whereas the highest risk weight is for Polish banks at 85 percent of risk exposure amounts.

To summarize, Figure 5 depicts the IRB average risk weights that are applied by banks for each of their corporate, retail, and mortgage portfolios, averaged at the country level. It is such significant heterogeneity in IRB bank risk weights that has prompted their regulatory review as well as concern that internal models do not “strike the right balance between simplicity, comparability and risk sensitivity” (BIS, 2016a).

Figure 5.
Figure 5.

IRB Average Risk Weights in the EU (in percent), June 2015

Citation: IMF Working Papers 2017, 137; 10.5089/9781484302958.001.A001

Sources: EBA and Author’s calculations

IV. Variation in Risk Weights by Country of Counterparty and Asset Class

In addition to portfolio type (SA/IRB method) and asset class (corporate/ retail/ mortgage), the EBA transparency exercise provides an important third dimension in the data, which is the breakdown of total exposures at default by country of counterparty. Each bank in the EBA sample reports its counterparty exposures to the largest ten countries. This granularity in the data at the bank level allows for a comparison of risk weights that are used by banks for exposures to the same country for a particular asset class.

Since loan quality affects RW, the analysis focuses to the extent possible on good quality portfolios to better analyze risk weight comparability across banks. Indeed, risk weights may be skewed in a bank portfolio that carries, say, defaulted mortgages, in comparison with a bank for which the mortgage portfolio is not impaired. As explained in Appendix A, defaulted loans are not included in the calculation of risk weights to ensure that variations in the share of defaulted loans do not undermine the comparability of portfolios, although some variation in the quality of portfolios will remain.

The full set of average risk weights by portfolio type, asset class, and counterparty is presented in matrix format in Appendix B. Table B1 Panel A lists the average IRB corporate risk weight matrix for banks in the EU. For banks in each of the countries displayed in columns, the IRB risk weights applied in the country of their counterparty exposure are listed in rows. To illustrate, consider Germany as the country of counterparty in the IRB corporate credit exposures. For these exposures to German corporates, banks in Austria implement, on average a 59 percent risk weight to calculate their risk exposure amount, but this same ratio is as low as 9 percent at Danish banks and as high as 137 percent at Irish banks, whereas German banks use a 46 percent risk weight for their corporate risk exposures to Germany. Panels B and C of Table B1 likewise report risk weights by country of counterparty for the IRB retail and mortgage credit exposures, respectively. Table B2 displays similar statistics for the SA portfolios.26

Table 2 summarizes the findings of Tables B1 and B2. Panels A, B, and C display descriptive statistics for risk weights by country of counterparty for IRB/SA corporate, retail, and mortgage credit portfolios, respectively.27 To interpret, consider IRB corporate exposures to the Netherlands row in Panel A from which the following facts emerge. First, banks from 10 different countries have IRB corporate credit exposures to the Netherlands. Second, the average corporate risk weight for these exposures across all banks and countries is 61 percent of exposures at default, whereas the median is lower at 47 percent indicating positive skewness in the distribution of risk weights to the Netherlands. Third, heterogeneity in risk weights is also reflected by the high standard deviation of 31 percent, where the minimum and maximum IRB corporate risk weights for exposures to the Netherlands are 41 and 137 percent, respectively. Briefly, in Panel A, the greatest variability in IRB and SA corporate risk weights is for counterparty exposures to Luxembourg and Poland; in Panel B, the largest dispersion in retail IRB and SA risk weights is in Ireland and Spain; and in Panel C, IRB and SA mortgage risk weights differ mostly for exposures to the U.S.

Table 2.

Descriptive Statistics for Bank Risk Weights by Country of Counterparty Exposure

Panel A. Corporate Credit (in Percent, June 2015)

article image

Panel B. Retail Credit (in Percent, June 2015)

article image

Panel C. Mortgage Credit (in Percent, June 2015)

article image
Sources: EBA and Author’s calculations.

It is evident from Table 2 that there is significant heterogeneity in bank risk weights across portfolios for the same country of counterparty. When banks lend across borders, they are subject to more informational opacity relative to extending credit domestically because they are in a less favorable position to collect borrower information and closely monitor debtor performance in a foreign country. In turn, uncertainty regarding debtor performance could translate into higher risk weights by banks to capture riskier portfolios abroad. Alternatively, banks may only be willing to lend abroad because they want to cater to the needs of their own domestic clients who branch out to other countries, in which case they are likely to apply lower risk weights to such credit portfolios abroad. Regardless, given that bank portfolios are cleaned of defaulted loans and disaggregated by loan type and counterparty country, they should in principle control for some of the major factors shifting the RWA density. Yet, they still show major differences in risk weights, adding to challenges of the comparability of risk weights.

V. IRB Corporate Risk Weights and Firm Fundamentals

As the use of internal risk models provides room for maneuver, banks are often alleged to adjust IRB model parameters with a view to reduce their risk weights and, thereby, inflate their capital ratios (Vallascas and Hagendorff, 2013; Behn, Haselmann, and Vig, 2016).28 We investigate the extent to which bank risk weights reflect asset risk by focusing on the corporate portfolio using the following baseline regression:

IRBCorporateRWi,c,t=α1+β1Fundamentalsc,t+β2Zit+β3Xct+Ci+εit(1)

IRB Corporate RWi,c,t denotes bank i’s risk weight for its corporate portfolio in country c at time t.29 It is retrieved for each bank included in the EBA transparency exercise that uses the IRB method and reports its ten largest corporate exposures across countries.30 Fundamentals, our main variable of interest, is a vector of firm characteristics detailed further below.

Z is a vector of bank controls that includes the share of the corporate portfolio in total bank loans (Corporate portfolio) and pretax return on assets (Pretax ROA). Higher corporate loan concentration could imply higher risk exposure and thereby associate positively with corporate risk weights, or it could convey more expertise using IRB methods for risk management potentially associating with lower risk weights. More profitable banks are likely to favor higher risk weights because their charter value is higher, but if their lending strategy is aggressive they could also apply lower corporate risk weights.31 X represents the growth rate in real GDP in the country of counterparty exposure to control for domestic lending conditions32, C is a vector of fixed effects for bank i’s country (controlling for the parent lender’s domestic conditions)33, and is a random error term. Since the IRB corporate risk weights vary by bank and by country of counterparty exposure, we run regressions at the bank level with robust standard errors clustered by period and country of counterparty exposure.

Fundamentals include both accounting-based indicators of firm risk and expected default frequencies. Accounting-based indicators are retrieved from the International Monetary Fund’s Corporate Vulnerability Utility (CVU) for 2013 and 2014 (i.e. they are lagged relative to the periods of measurement of corporate risk weights, December 2014 and June 2015).34 They include five main risk corporate risk indicators: Leverage (ratio of debt to assets), interest coverage (ratio of earnings before interest and taxes to interest expense), liquidity (ratio of current assets to total assets), stability (z-score)35, and market value (ratio of market-to-book value).36 The indicators from the CVU are based on firm-level data from annual reports of publicly traded companies so that they may not necessarily mirror bank i’s spectrum of corporate borrowers. Yet, we assess the sensitivity of corporate risk weights to each of these indicators evaluated at the median and the 75th percentile of their risk distribution. If corporate risk weights are risk sensitive, they would be expected to positively correlate with higher leverage, lower interest coverage, lower liquidity, reduced stability, or lower market value.

Table 3 shows the correlation of corporate risk weights with each firm indicator evaluated at the median (Panel A) and at the 75th percentile (Panel B) of its distribution.37 Model 1 considers firm fundamentals and fixed effects for bank i’s country of origin; Model 2 additionally controls for bank i’s share of corporate loans and its profitability, as well as the growth rate in the country of counterparty exposure.

Table 3.

IRB Corporate Risk Weights and Firm Fundamentals

Panel A. Firm fundamentals evaluated at the median.

article image
Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1