Banca d’Italia paper looks at factors behind RWA calculations, why they differ so much

In a Securities Finance Monitor post just the other day on the BIS and the push for simplicity in capital rules, we noted that one argument in favor of simpler rules was the inconsistency of applying current RWA rules. Banks with similar exposures can report different results based on differing interpretations. A September 2012 paper from the Banca d’Italia, “Inside the labyrinth of Basel risk-weighted assets: how not to get lost,” by Francesco Cannata, Simone Casellina and Gregorio Guidi examines those differences. Here are the main points:

Differences in RWA between banks are not, unto itself, an indicator that banks are gaming the rules. No two banks are the same. From the paper,

“…Investment banks are mostly exposed to market risk, retail banks to credit risk, universal banks to both types of risk. Given the different prudential treatment of these risks, different groups of institutions are likely to have – other things being equal – different RWA…”

Bank asset mix differs across regions and this can cause substantial RWA variation.

“…RWA/Asset ratios are quite similar between, say, JPMorgan and Deutsche Bank in the investment banking business, but they differ significantly in the retail exposures. In particular, US banks have on average a higher share of mortgages and consumer credit as well as a lower share of corporate loans than European institutions; in addition, they rely more on securitisation for mortgage exposures and tend to remove high-quality (and low-RW) assets, therefore making the average risk weight of their portfolio greater than that of the European banks. Also, their greater credit card business, with its higher regulatory risk weight, contributes to the divergence…”

But different institutions are subject to different rules – notably accounting rules – and this creates RWA divergence.

“…Accounting rules have a direct impact on prudential metrics: balance-sheet items and accounting valuations are fully part of the mechanisms behind the calculation of capital requirements for prudential purposes. The issue comprises a twofold problem: the interaction between accounting and prudential frameworks, and the current lack of convergence amongst accounting regimes (IFRS in Europe, GAAP in the U.S.). This lack of convergence is a major explanation for RWA divergences, especially between European and US institutions…”


“…Looking at the summary balance sheet of Deutsche Bank, published under both US and European accounting regimes, the conversion from IFRS to GAAP reduces total assets by 37 per cent. In particular, the different treatment of netting shrinks the derivatives book by 90 per cent. This conversion, when applied in the same fashion also to other banking groups, seems to suggest that with a few adjustments for key differences in accounting principles and business mix, the ratio of RWA to total assets is not significantly dissimilar…”

The paper finds that the sheer complexity of the rules, especially for risk management, contributes to the divergence in RWA calculations. 

“…Indeed, the influence of risk management practices on RWA calculation is a direct consequence of the key role that risk measurement plays in the Basel framework. As Haldane (2011) notes, a variety of modelling choices is likely to determine possible differences across banks, even when other things are equal: “to determine the regulatory capital ratio of a bank, the number of calculations has risen from single figures to over 200 million”. The quality and quantity of the data used for the estimation of risk parameters, the estimation methodology, the combination of quantitative and qualitative variables, the degree of freedom for qualitative override, and the calibration of risk parameters are amongst the major choices a risk manager has to make…”

An unintended consequence of the ability to choose either standardized models or internal models was to create inconsistency across banks. Compounding the problem was the variation in internal models – no two were the same. Indeed, it appears that internal models provided an opportunity to manage (read: game) RWA.

“…in approaching the new capital targets, manoeuvres on RWA do represent a possible option for banks – especially those which already use or plan to use internal models – to increase their capital ratios, casting doubt on the role of supervisory rules and practices in determining capital requirements (first of all, by validating banks’ internal models)…”

While markets change and approaches are constantly being fine tuned, banks don’t help themselves when their models are constantly evolving. According to the paper, banks have responded to criticisms that RWA looks like a moving target by saying they have been undergoing a process of data cleansing (RBS), that there were model changes (Lloyds), they were optimizing (Santander, Unicredit, Nordea), and there were parameter updates (Danske and Commerzbank). Interesting and valid points, but this makes year-to-year comparisons of the same bank all but impossible.

The inconsistent approaches – some which look unavoidable, others may have been used to create better optics – have a negative impact on the broader market and come back to bite the banks.

“…What is clear enough is that market distrust of the RWA reported by banks is likely to have consequences, such as: re-calculation by banks themselves of capital ratios and disregard of regulatory ratios; greater reliance on measures similar to leverage ratios; higher capital ratios requested by the market to compensate for the perceived unreliability of the denominator…”


“…Notwithstanding the methodological differences, most of them argue that banks’ RWA measurements contain an excessive degree of subjectivity and are therefore not readily comparable across institutions; in addition, some of them claim that risk-weighted assets do not even properly reflect the actual risk in banks’ balance sheets…”

But what about the downside of over-simplification? The paper acknowledges that

“…risk-weighted assets constitute a complex phenomenon, and in a risk-sensitive prudential framework this complexity derives largely from the complexity of the underlying financial business. Therefore, given the very large number of variables in play in the prudential rules, measuring banks’ riskiness with over simple proxies is not always easy or even desirable: proper understanding of a complex phenomenon requires the right glasses, to avoid mistaken or misleading conclusions…


“…the ultimate objective of risk-based capital regulation is more accurate allocation of capital among institutions, penalising those with low quality portfolios and offering adequate capital incentives to those with high-quality assets. Thus, to some extent divergences in banks’ RWA represent an intended, desired effect of regulation…”

But – and this is a big but – “…The problem is the unintended, undesired differences, which are likely to jeopardise a true level playing field among banks and jurisdictions…”

It is very difficult to understand the differences between banks when it is impossible to differentiate between the impact of interpretation, model, and asset allocation effects. When information is so hard to process, markets inevitably assume the worst. In a stressed environment, analysts are forced to use worst-case assumptions to paint all institutions – both the good and the bad. While simplification may paper over some differences, if it can give confidence to the market that comparisons are “apples to apples” then the market is certainly better off.

A link to the Banca d’Italia paper is here.

A link to the Securities Finance Monitor post on BIS simplification is here.

Related Posts

Previous Post
Finadium: Large OTC Derivatives End-Users on Clearing and Collateral (free via Calypso)
Next Post
Do asset managers think that securities lending is opaque or transparent, and why does it matter? (Finadium subscribers only)

Related Posts

Fill out this field
Fill out this field
Please enter a valid email address.


Reset password

Create an account