CME Span 2 Migration Completed

CME SPAN 2 is now live for CME Energy and by all accounts it has been a successful migration. It’s been a nearly 25 year wait since Filtered Historical VaR was first suggested as a replacement for SPAN in 1999 but this has now been achieved for a major market. However, there is still a long way to go before all CME markets switch to the new SPAN 2 algorithm – and that is without considering the other exchanges still using SPAN.

So now we are at this point, what are the things that firms should be considering as we move forward into the next stages of these migrations.

What Will Be The Next Market?

The dates for go live for the various upgrades from SPAN to VaR have changed often over the years, but these are the latest dates for CME:

As for the other CCPs, ICE has already gone live with its own VaR based algorithm, IRM 2, for US Equity Index futures. Their plan is that the next market should be Energy on ICE Clear Europe in June 2024, with parallel runs being available from the beginning of the year. Other markets will follow after that.

LME plans to move onto their own VaR algorithm, that they use internally for risk purposes, have been delayed following the issues surrounding Nickel, but it is expected that they will start announcing dates for the implementation soon.

And there are other markets that are looking to move away from SPAN. JSCC are well underway with their implementation and have started publishing test files alongside VaR margin calculation software to allow firms to prepare for go live. They are also providing a Web Simulation Environment (developed in partnership with OpenGamma) that allows firms to compare margin requirements under SPAN and VaR.

What Has Been The Impact On Margin Requirements?

As a simple comparison, we calculated the requirements for a single position in CME (WTI) Light Sweet Crude Oil December 2023 future as of 15th August. The results were as follows:

This shows that the requirements under the two algorithms will be similar, but highlights one of the major differences between SPAN and the new VaR based SPAN 2. The margin for long and short positions will no longer be the same.

The differences for more complex portfolios are harder to determine. With offsets now inherent in the VaR calculation rather than calculated using specific parameters, the level of margin requirements will be entirely dependent on the portfolio.

With the concentration charge now a core component of the margin it is also likely that the margin for larger positions will be more impacted than for smaller portfolios.

What Are CME Doing Going Forward To Help Firms?

A move to VaR is a big change and the CCPs understand this. One of the changes is the additional complexity in the algorithm including the number of scenarios that are considered (up from the 16 of SPAN to the 1,000s involved in VaR). This makes it much more difficult to execute some standard use cases, for example pre-trade limit checks.

To help firms, CME will continue to publish SPAN files (SPAN 2 approximation files) that include parameters that attempt to ensure that the margin calculated using them will be as close as possible to the actual SPAN 2 margin.

Analysis has shown that the results from using these files can be very accurate or way off depending on the portfolio. The accuracy is generally better for macro portfolios and not as good on targeted portfolios. CME are working on improving this as much as possible, but it is easy to see the difficulty by considering a simple position.

Using the example above the margin for both 1 long and 1 short position in WTI would be $5887 – this is the average of the VaR margin that is different for long and short positions. So because of the limits of the SPAN methodology it would never be possible to match the SPAN 2 results for these positions. However, if you had a spread position then some of this discrepancy would be netted out.

CME is also aware of the increased size of the SPAN 2 files compared with the current SPAN files – again because of the number of scenarios involved. They currently publish a single file covering the parameters for all markets, but going forward with SPAN 2 are looking to split this into separate files of a more manageable size. They are also looking to ensure that the infrastructure in place to download the files is as efficient as possible.

This compares with Eurex whose Prisma algorithm is VaR based. They split their data into multiple files, limiting the size of each file to ensure that it does not cause technical problems. They also differentiate between contracts where there is open interest and those where there are no positions so that the amount of data downloaded can be limited.

What Have We Learnt?

Upgrading from SPAN to VaR was never going to be easy and it has been a very long time coming – the use of VaR to replace SPAN was first suggested in 1999!! However, with CME having moved Energy and ICE having done the same for their US Equity index products, it is likely that the remaining markets will not face the same difficulties. However, the work that has been completed so far does highlight some of the issues related with the move from SPAN to VaR:

• There will be winners and losers in terms of margin requirements compared with SPAN.
• Some use cases will be harder to support, for example simple pre-trade margin estimation based on the margin rate, but CCPs are aware of this and are doing what they can to assist. For CME this means continuing to publish equivalent SPAN scanning ranges.
• Performance is going to be an issue. The 16 SPAN scenarios have been replaced by 1,000s of VaR scenarios. This will impact both the margin calculation time and the download of the parameter files, which in the CME case will increase in size from a few 100 MBytes to nearly 20 GBytes once all markets have been upgraded.