Data shortage hits margin models for Asia banks

Thin trade volumes in local derivatives threaten to undermine key tests for initial margin models

Asia data

Asian banks preparing for the next wave of initial margin rules are scrambling to gather the necessary data that will allow them to test their margining models.

Under the initial margin regime, a wide range of derivatives counterparties will soon have to post funds against uncleared transactions. Although dealers are able to use industry-standard models to calculate margin, local regulators often require these models to be validated. This process involves calibration by backtesting against three years of historical data and a one-year stress period.

Banks including Singapore’s DBS and OCBC have raised concerns that derivatives denominated in emerging market currencies, such as Chinese renminbi, lack sufficient trading history to provide data for backtesting. Any discrepancies in model outputs between counterparties could result in disputes over margin, and incur the ire of regulators who are keen to ensure models are robust.

“We’ve been running our value-at-risk engine for 13 to 14 years, so capturing historical rates scenarios is not a challenge,” says Brian Lo, group head of market and liquidity risk at DBS. “The more critical point is to capture the new rates and curves. In particular, the major market instruments that emerged in the last five years are the renminbi products and the curves related to it, such as swaps and foreign exchange forwards. This, in our view, would more likely be an area that might pose challenges.”

Banks typically use 2008 as the year for the stress period, which means that any products not traded in that year are unable to feature in the stress calculations. For products whose growth is recent, data across the three-year historical period is spotty, too.

Banks acknowledge there is no easy solution and are seeking help from market data vendors, or trying to gather proxy data to plug the gaps.

“These gaps in data are certainly an issue for emerging market institutions subject to backtesting regulation – regardless of which phase they are in,” says London-based Thomas Griffiths, co-head of triCalculate, a tech vendor.

“Their portfolio composition often consists of illiquid underlyings and currencies which is where the problem of missing data arises. Proxy-filling market data in Asia is far more challenging as there are more emerging markets than most other places. Typically banks have so far used proxy-filling as a stopgap but in Asia it may become a longer-term trend.”

However banks using proxy data need armies of quants to make the calculations work, and such personnel are in short supply in Asia.

Not so Simm-ple

Initial margin requirements on bilateral swaps exposures are part of a post-crisis regime intended to make the non-cleared derivatives market safer. The rules arrive in five phases, between 2016 and 2020, and require firms to collect initial margin from other in-scope entities if they exceed a notional threshold that drops each year.

Currently, around 50 of the largest dealers, plus Brevan Howard, must post initial margin on their non-cleared derivatives following the first three implementation phases, which kicked off in 2016 with a $3 trillion threshold for aggregate average notional amount of non-cleared derivatives outstanding. The threshold drops to $750 billion in 2019 for phase 4 of the rollout, bringing an additional estimated 30 to 50 counterparties into scope. The threshold is scheduled to plummet from $750 billion to just $8 billion – or local market equivalent – for phase 5.

According to the International Swaps and Derivatives Association, those in-scope firms represent an estimated 9,500 counterparty relationships for which new documentation must be negotiated; 19,000 segregated custody accounts would also need to be created.

Most Asian banks, including the Singapore-based regional lenders, will be caught by the final phase of margining in September 2020. In the meantime, firms intending to use Isda’s standard initial margin model (Simm) must provide backtesting results to regulators. The alternative is a more punitive standardised grid approach. In some countries banks may have to notify their regulator about which margining approach they intend to use at least six months in advance.

To calculate initial margin, the Simm requires firms to work out their respective sensitivities – for example, the change in value of an interest rate derivative for a basis point change in the relevant yield curves, known as PV01 – and to multiply these sensitivities by the size of a shock.

Trades are slotted into one of four product classes – credit, equity, commodity, and interest rates and foreign exchange – with each subject to risk from six different risk classes. The end goal, specified by the regulators, is to cover counterparty exposure over a 10-day holding period to a 99% confidence level.

A dearth of data for initial margin backtesting is going to be an issue for most of the Asian banks and the buy side in phase 4 and phase 5

Frederick Shen, OCBC Bank

Regulators vary in the kind of information they require from firms regarding their Simm models. In Singapore, banks must notify the regulator of their state of preparation, while in Hong Kong and Australia, banks must submit their Simm models to authorities for approval – including the backtesting results – before they go live.

For many banks in the region, emerging markets derivatives exposures will be significant, having grown over the last 10 years. In China, for instance, over-the-counter forex derivatives linked to the renminbi saw a 13-fold increase in net-net daily average turnover, moving from $15 billion in 2007 to $120 billion in 2013, and $202 billion in 2016, according to the most recent Bank for International Settlements triennial derivatives survey. While some types of forex derivatives are exempt from the initial margin regime, the volume increase gives an indication of the effect of the new margin requirements on many regional banks.

Banks can achieve smaller margin payments using Simm rather than the grid methodology. But they need historical data on these trades to be able to backtest the model, and this data is not easy to come by.

Dealers report particular trouble sourcing data for the stressed year, which in most cases defaults to 2008 – a time when trading in many emerging markets derivatives instruments was thin.

For renminbi forex derivatives, volumes were far thinner back in 2008 compared to now. In some asset classes there was no activity at all; the BIS survey shows OTC interest rate derivatives denominated in renminbi were not traded in 2007, but reached $10 billion in 2016.

“A dearth of data for initial margin backtesting is going to be an issue for most of the Asian banks and the buy side in phase 4 and phase 5,” says Frederick Shen, head of global treasury business management at OCBC Bank in Singapore. “We do collect data but we may not have the data relevant for the stress period at hand.”

Griffiths of triCalculate says: “One of the biggest challenges for Asian banks is finding the right volatility data and the numbers for the stressed year, 2008, when some instruments were just not trading.”

Patchy trading activity is also hindering the collection of data for the three-year historical period. While daily average turnover in products like renminbi forex derivatives has grown, much of it has been in the OTC market: three times as many trades were executed OTC than on an exchange in 2016, according to a BIS research paper on emerging market derivatives.

When banks only have ready access to trade volumes in their own books, this data might not represent the wider market. This raises the likelihood that regulators receive differing backtest results between banks. Also, if margin models are calibrated to significantly different data, mismatched outputs could result in margin disputes between counterparties.

Bank sources highlight specific difficulties with longer-duration non-deliverable forwards linked to currencies such as the Malaysian ringgit, Thai baht and Philippine peso.

Dealers caught in the first three phases of initial margin haven’t faced the same level of challenges as derivatives denominated in Asian emerging market currencies are a small part of their portfolio. They have either received regulatory approval to exclude these trades from their model validation efforts, or have used their quant capability to build proxy curves for some trades, two dealers in the region say.

Delving for data

Asian regional banks coming into scope for initial margin in 2020 are already looking at a number of options. The most straightforward one is to lean on market data vendors such as Refinitiv and Bloomberg, which aggregate data points such as price and volatility from a variety of sources. This data doesn’t come cheap, though: banks are said to be charged hundreds of thousands of dollars for access, and the data is still said to be incomplete for certain emerging market currencies.

OCBC’s Shen says the bank is considering using proxy data to plug gaps for products where there is little data, especially for the stressed one-year period. Sources at three Asian banks say they are working with quants to come up with an acceptable proxy for transactions in renminbi products and non-deliverable forwards. For offshore renminbi trades, banks can use the renminbi-denominated Hibor fixing, an interbank lending rate, a source at one market data vendor says.

Some are advocating a plan where lenders in the region share common best practices, pool data or conduct polls to see which base inputs, cutoff times, or valuation and volatility points they should use to ensure backtest numbers match for counterparties.

This has its own drawbacks though: some banks might be reluctant to share their data, for instance. Quant teams will also need to work out the best ways to use any proxy data to fill in the gaps. Another option is a hybrid model where dealers use market data supplemented by proxy information, for both the stress year and the three-year historical period. But again this requires quant resources.

“Market data vendors can give the base input for proxy-filling but banks need quants to then validate and compute the data for the testing,” says Sydney-based Arun Prasad, a senior executive at Refinitiv, which was formerly the Thomson Reuters financial and risk unit.

Another option is changing the stressed year. OCBC’s Shen argues that market structures have changed post-crisis – for instance, central clearing of interest rate swaps and non-deliverable forwards has increased dramatically, making data in that period less relevant to today.

Prasad agrees, saying dealers might be better off picking a different stress period if they don’t have the data.

“You can look at other years where we have had, for instance, taper tantrums in bonds,” he says, referring to the 2013 market panic that sent 10-year US yields surging by 140 basis points between May and early September that year after the Federal Reserve hinted at a reduction in stimulus.

“The year 2008 was very unique because we had stress across credit, rates, underlying assets, cashflows, loans – everything together and that is why it was a global crisis. For individual asset classes you can pick stressed scenarios since then.”

Regulators may even allow banks to exclude certain trades for model validation, if the data is too thin to support the model’s results. But that wouldn’t solve the problem of mismatched margin outputs between counterparties.

Editing by Alex Krohn

  • LinkedIn  
  • Save this article
  • Print this page  

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: