Credit Benchmark revolutionizes internal models


Sid Verma
Published on:

Pools data to improve accuracy and transparency; could disrupt ratings industry.

The internal-ratings based approach for banks to quantify capital for credit risk – a framework deployed by over 100 banks, from Europe to China and Australia – is in crisis. While the Fed has been consistently sceptical of it, European regulators at the Basel level have adopted an ambivalent posture by first encouraging lenders to adopt it, only to then sound the alarm over inconsistences in risk weights.

In 2013, the European Union adopted the Capital Requirements Directive, which sought to reduce systemic reliance on credit ratings by encouraging banks to calculate their own ratings; and to make bank-capital more risk-sensitive, letting lenders use these calculations in their own risk-management or economic capital models. So while the IRB approach was developed for large, internationally active banks by Basel in the early 2000s, the CRD opened it up for use by all banks in Europe.

We can build models for IRB-regulatory purposes and for economic capital models, which will help banks to manage their business

Elly Hardwick,
Credit Benchmark

However, regulators now reckon the IRB framework gives lenders too much flexibility, shown by the variance between different banks’ assessments of the creditworthiness of the same individual borrowers or groups of borrowers and, by implication, the capital requirements set aside for exposure to them. The Basel committee in December 2014 proposed a blunt floor on the capital relief provided by internal models relative to the standardized approach, while European supervisors are also now intervening in the IRB-modelling process.

Bankers hope to soften this regulatory push by explaining the variance between model-driven and standardized risk weights. Many bankers argue internal models are more accurate, granular and are appropriately calibrated across diverse market scenarios than the standardized approach. They also argue internal models provide risk managers with more personal accountability, giving them an incentive to take ownership of risk mitigants, such as collateral, and to hedge exposures, rather than outsourcing this to regulators.

Suspicious of being gamed, regulators decline to clarify the tipping point whereby legitimate differences of opinion on credit risk no longer represent a healthy form of systemic diversity, but, instead, are unjustified and may constitute a form of capital arbitrage.


There is a lack of transparency in the data to assess the validity of arguments from either side. And this is where Credit Benchmark, a London-based start-up provider of anonymous, bank-submitted credit data, hopes to help.

The idea is simple: Credit Benchmark, founded in 2012, acts as non-intrusive intermediary that collects banks’ IRB credit assessments of their borrowers’ risk and publishes the average. Participating lenders then see consensus estimates on the probability of default and loss-given default metrics on sovereigns, corporates, banks, hedge funds and institutional firms. In effect, this credit-sharing effort is similar to Bloomberg consensus forecasts for economic data and Interactive Brokers’ consensus estimates for equities, and Euromoney Country Risk, which aggregates risk scores from over 400 economists around the world.

In some cases, a bank might have data on up to 250,000 names, most of which are typically unrated by the traditional big three – Fitch, Moody’s and Standard & Poor’s – because of the size of underdeveloped local capital markets for corporate issuers in many emerging markets, and the size of the SME sector globally. For example, there are 8,000 rated entities compared with 50,000 listed companies across the world, says Credit Benchmark chief executive Elly Hardwick.

CB’s proposition, therefore, could help break the impasse between bankers and regulators over the IRB regime and transform how lenders make their business decisions.

For the first time, bankers will be able to know the consensus estimates of the credit quality of a particular borrower, as determined by the risk teams of other global lenders with skin in the game, using these scores to determine the right risk-capital requirement. Participating banks will be able to see where they stand relative to the market view, while also identifying inconsistences.

At a business level, if a bank is planning to expand into a new market, it will have an objective benchmark on how rivals are pricing risk on average for a slew of borrowers. In theory, that boosts the quality of decision-making.

At a macro level, Hardwick says CB’s back testing proves IRB banks’ sovereign estimates are good indicators of future creditworthiness relative to agency ratings and CDS prices. Hardwick says it has up to 20 committed banks for the project but signing up more is an arduous process – the record for the shortest-completion time is four months – because of legal, compliance and technology challenges.

The company raised $20 million in a fresh round of funding in April, led by venture capital firm Balderton Capital, for its US expansion and to recruit data scientists. Founders Mark Faulkner and Donal Smoth have established relationships with financials and boast a strong track record in monetizing anonymous user-submitted data: they previously set up Data Explorers, a securities lending data-provider, which was bought by financial information company Markit in 2012.

CB’s business model is capital-efficient since it doesn’t need to employ expensive credit analysts itself.

Over the past two years, banks have stepped up their collaborative efforts to share sensitive data – including new high-profile bond-liquidity platforms – to meet regulatory- and market-driven changes. In late August, JPMorgan, Goldman Sachs and Morgan Stanley created a company with a mandate to clean securities data to boost pricing accuracy and meet regulators’ calls to standardize data sets. The CB project consists of collaboration at a much-higher level than the heads of markets business, since the sharing of confidential IRB data and associated anti-trust concerns require the involvement of chief risk officers and, ultimately, CEOs.

Risk maps

Greater transparency on data should help root out fat-finger errors and, in theory, make benchmark-manipulation tougher. Hardwick says the team has already detected an instance of the former, while regular bank-submitted updates to CB might help lenders prepare for the projected jump in IRB stress testing from supervisors.

As Euromoney reported last month, regulators are also seeking to beef up systemic-risk data. Contributed-data platforms should help regulators map out cross-border risk by complementing existing market knowledge – such as aggregate Bank for International Settlements data on banks’ consolidated foreign claims – with lender-reported shifts in the average-weighted creditworthiness of their exposures.

Even if regulators kill off the IRB approach entirely, CB is spoilt by choice for business opportunities, especially in risk management, because lenders need to determine consistent collateral values and calibrate credit valuation adjustments in their trading books, Hardwick says. “We are building relationships with banks to get them to entrust us with highly confidential data. We can build models for IRB-regulatory purposes and for economic capital models, which will help banks to manage their business, especially those outside the top 20.”

CB might combine IRB and economic-capital data sets to generate comprehensive exposure models to cover all SMEs in a large domestic market. Hardwick adds: “Clearing institutions, regulators, agencies, buy-side firms and large corporates all have diverse data sets not standardized by supervisors, so that is also an opportunity.”

 Aside from improving risk and capital management, it’s easy to foresee a large array of trading opportunities from the data. Nevertheless, at this stage, no policy has been set on whether or not non-participants will be able consume the data. Hardwick says eventually Credit Benchmark will look to pool in banks’ data on a given borrower’s multiple fixed-income securities, where available. IRB models mandate yearly updates to credit assessments but, in practice, banks do this several times a year. Credit Benchmark receives monthly updates. Hardwick hopes this will become even more frequent.

Cyber security is emerging as banks biggest gripe after regulation. One senior banker at a US broker-dealer told Euromoney the bank is attacked at least 1,000 times daily. Clearly a breach as serious as the one at dating site Ashley Madison in August would be a disaster for CB’s business. CB receives bank-submitted data through a secured, encrypted file-transfer protocol and hosts the information in physical, guarded data-centres in London and Frankfurt, rather than using cloud-based storage solutions.


Promoting the use of consensus estimates will inevitably unleash two common criticisms of empirical models. Firstly, greater transparency might lead to confirmation-bias, when the risk community implicitly censors any outlier opinions. Secondly, IRB-derived probabilities go beyond the ordinal nature of agency-credit ratings by calculating risk in specific basis-points terms. Such granularity can provide an illusion of precision when risk officers make what are usually subjective judgments.

So the emergence of CB comes at a crucial moment.

Regulators remain wary of rating agencies, citing their issuer-based payment models, poor sub-prime credit decisions and pro-cyclical downgrades during the eurozone sovereign debt crisis. The 2013 failure to establish a new European-based rating agency was a sobering reminder of the natural oligarchy of the rating agency business. The need for a good reputation before winning clients and large human-resource requirements to rate complex entities means new entrants face a costly, uphill battle to gain market share.

By contrast, Credit Benchmark represents a classic form of fintech-led disruption since it deploys a nimble, technology-led platform to generate smart data and to scale up.

At the same time, regulators admit they don’t want the financial system to rely solely on standardized ratings in case they are accused of concentrating credit risk in the next banking crisis. The IRB framework, therefore, should have a role to play. In Credit Benchmark, aside from helping lenders make business decisions, bankers might have found a cheap way of preserving its integrity.