“A range of studies found an unacceptably wide variation in risk-weighted assets across banks that cannot be explained solely by differences in the riskiness of portfolios. The unwarranted variation makes it difficult to compare capital ratios across banks and undermines confidence in capital ratios.” Basel Committee on Banking Supervisions, 7 December 2017.
After nearly a decade of negotiations, last month, the Basel Committee on Banking Supervision completed the Basel III post-crisis reforms to capital regulation. The final standards include refinements to: credit risk measurement and the computation of risk-weighted assets; the calculation of off-balance-sheet exposures and of the requirements to address operational risk; and the leverage ratio requirement for global systemically important banks (G-SIBs).
In this post, we focus on revisions to the way in which banks compute risk-weighted assets. To foreshadow our conclusion: the new approach adds unnecessarily to regulatory complexity. If the concern is that current risk-based requirements result in insufficient capital, it would be better simply to raise the requirements.
Before getting to the changes, it is worth recalling a bit about capital regulation. Capital is a form of self-insurance. Common sense dictates that a risky driver, for example, is more likely to cause accidents, so they should pay more for their insurance—in the form of higher deductibles, copays and the like. Much as this self-insurance prompts a car’s driver to accept more of the risks they take, higher levels of capital compel a bank’s owners to bear more of the systemic risk the bank creates, and gives them an incentive to limit that risk. Put differently, a sufficiently large capital buffer prompts a bank to internalize the externality arising from its distress. It follows that a bank engaging in risky activity should have more capital as a buffer against what are likely to be more frequent and larger losses. In practice, this leads regulators to state capital requirements as a ratio of a bank’s equity to its risk-weighted assets (RWA).
Now, it is difficult to measure risk. For nearly all of the banks in the world, it is too costly to create their own models to do the job. Instead, they rely on the standardized approach where, based on the Basel Committee’s standards, national authorities set the risk weights for different types of assets. The largest banks in the world employ their own risk models and use the internal-ratings based (IRB) approach. Based on the Basel Committee’s most recent monitoring report, we estimate that 100 or so worldwide go to the cost and trouble of building their own risk models.
As we discussed in an earlier post, the complexities of modeling risk make it unsurprising that different banks come up with different appraisals even for the same individual asset or portfolio. However, as the chart below highlights, these estimates vary by more than seems reasonable when applied by different banks to common hypothetical portfolios of held-to-maturity (banking book) assets. The chart shows the range of estimated risk densities (risk-weighted assets as a percent of unweighted assets) by 30 banks that use their internal models to assess the risk of four hypothetical loan portfolios. In each case, the box represents the interquartile range (25th to 75th percentile) of the estimates, the line extends from the minimum to the maximum, and the centerline with the black diamond is the median.
Average Risk Densities for Corporate and Retail Portfolios (Risk-weighted assets as a percent of unweighted assets): Median, interquartile range, minimum and maximum
Source: Basel Committee on Banking Supervisions, Regulatory consistency assessment programme (RCAP) – Analysis of risk-weighted assets for credit risk in the banking book, April 2016, Table 1.
Even if we assume banks are not trying to game the system using their models (a dubious assumption at best), we would expect some divergence in their estimates of the riskiness of particular types of loans. But the results of this exercise suggest that the variation in estimates is, as the quote at the beginning suggests, unacceptably wide. For large corporate exposures, the risk densities range from 25% to 61% (with the interquartile range running from 38% to 50%). Mortgage portfolios exhibit even wider variation: the interquartile range is from 14% to 31%, with the minimum at an unbelievably low 5.2% (!) and the maximum at a very conservative 80.2%.
To be sure, we have seen similar results before in the Basel Committee’s original 2013 study of risk-weighted assets (see here.) Moreover, in a recent examination of the practices in more than 100 institutions in 17 EU countries, the European Banking Association found even greater variation (see here).
How should regulators address the inevitable imprecision in measuring risk? Prior to the December 2017 revisions, the Basel Committee had two backup strategies for addressing the unavoidable uncertainty in risk measurement: an unweighted leverage ratio and stress testing. Now, it has added a third: a floor on the risk weights of assets when using internal models.
Before explaining the idea of a floor, let’s recall the rationale for the first two supplementary mechanisms. The leverage ratio, which treats all exposures (both on- and off-balance sheet) equally, reflects regulators’ own doubts about risk weights—both standardized and internal-ratings based—and their concerns that the weights may unintentionally encourage systemic risk-taking. Whether the risk-weighted or leverage requirement binds depends on a bank’s risk density. The following chart, which shows the average capital ratios (the gray and red bars) for various regions of the world, reveals the enormous regional variation in risk density, shown as blue diamonds (and measured on the right scale). By our calculations, the Basel III minimum leverage ratio is now calibrated so that it (rather than the risk-weighted ratio) will bind on those G-SIBS with a risk density of less than about 44%—roughly the average in the euro area (42%).
Average Common Equity Tier 1 Risk-based Capital Ratio, Leverage Ratio and Risk Density, 2015
Source: BIS Annual Report 2016, Chart VI.1.
Stress tests supplement risk-weighted capital measures by revealing whether the resulting capital is adequate to survive a simulated crisis. We have discussed our strong support for stress testing on numerous occasions (see here and here). Very briefly, exercises like the Federal Reserve’s Comprehensive Capital Analysis and Review (CCAR) and the ECB’s Comprehensive Assessment examine whether systemically risky banks’ capital levels are sufficient to withstand an adverse macroeconomic scenario that combines a severe recession, equity and property price crashes, dramatic increases in risk spreads, and potentially a variety of additional calamities all simultaneously. Stress tests take seriously the fact that when a large common shock hits, there is no one to whom these banks can sell assets or anyone from whom they can raise capital. Ensuring that each systemic intermediary can withstand significant stress raises the likelihood that the system can survive.
To summarize, the two outstanding mechanisms for addressing uncertainty over risk assessments take what are almost diametrically opposite strategies. The leverage ratio ignores risk sensitivity altogether. By contrast, as noted recently by Greenwood, Hanson, Stein and Sunderam, stress tests effectively scale up existing risk weights by amounts related to worst-case scenarios.
Turning to the most recent innovation, output floors limit the banks’ scope to lower risk weights based on their internal models. The details are complex, but the principle is straightforward: for a bank using its own internal models, total risk-weighted assets cannot be less than 72.5% of the level computed using the standardized approach.
To see the implication of this, consider a simple case where a bank’s assets are solely composed of the large corporate loan portfolio in the 2016 Basel study. Furthermore, assume that the maximum risk weight for this portfolio shown in the chart (61%) comes from using the standardized approach. (A bank can always revert to the standardized approach, so the number should never go higher than that.) Now, the new rule means that this bank will not be allowed to have risk-weighted assets of less than
0.725 × 61% = 44% of its total assets (which are only corporate bonds). This new minimum is only a bit below the median of the distribution (46.5%).
Will this change improve the credibility of the Basel III framework? More importantly, will limiting how banks can use their internal models in computing their risk-based capital requirement improve the resilience of the financial system? The standards are so complex, and compliance and monitoring is so time consuming and difficult, that we suspect no one really knows the answer.
Where does this leave us? At the risk of repetition, our answer continues to be the same: more capital. If regulators’ concern is that uncertainties about risk measurement, gaming by banks, or both are reducing the resilience of the system, it would be better to raise the risk-based requirement than to rely on backup mechanisms that sacrifice risk sensitivity or, like the new risk-based floor, increase complexity. The more often that these backup mechanisms bind, the more likely it is that the risk-based requirements are set too low.
Alternatively, if the worry is that one or more banks using the IRB approach routinely games the requirements in an abusive way, regulators could detect this by requiring banks regularly to provide their risk-weighted asset estimates for a range of hypothetical portfolios. Those banks that report measures that are consistently well below the norm would be subject to close review of their internal models. Compared to imposing a floor on risk weights, this approach is more likely to lead to the dissemination of best practices in risk management.
Finally, why not take a page from the playbook of Sweden’s financial supervisory authority, Finansinspektionen? Using their discretionary authority, Swedish supervisors have set capital requirements for the four large Swedish banks (common equity tier 1 relative to risk-weighted assets) at between 19.2% and 23.9%―twice the Basel III minimum. We suggest that, as they contemplate further refinements to international standards, the Basel Committee seriously consider following suit.