Oct 20, 2014

Pitfalls and Best Practices - Navigating Initial Margin Methodologies for OTC Derivatives


As a follow-up to our previous initial margin discussion, “Navigating the Murky Waters of Initial Margin for OTC Derivatives regarding how to navigate the complexities of non-centrally cleared OTC derivatives, Part II of our analysis will outline and break down some of the standard methodologies most popularly used today. 

Watch the related webinar which provides a detailed analysis of initial margin and its impact on the derivative markets, “Primer on Initial Margin and its Impact on Derivatives Markets.

In our analysis, we describe what we consider to be the four most widely used, industry-accepted IM based methodologies. These include: Historical VaR (industry standard), Variance Covariance VaR, Monte Carlo VaR and Scenario-based methods (industry standard). To begin our analysis, we will first outline the steps involved in these methodologies:


Historical VaR (industry standard)

 The Historical VaR approach includes the following steps:

  • Step 1: Gathering of historical raw market data and market data cleansing, such as the back-filling of any missing data and adjustment for corporate actions
  • Step 2: Choosing appropriate risk factors, such as quotes, yield curve and implied risk factors
  • Step 3: Construct time series/time series adjustments, such as volatility scaling
  • Step 4: Scenario generation
  • Step 5: Calculate exposures/exposure distribution using some pricing model
  • Step 6: VaR or expected loss calculation (either one can be used)


Variance Covariance VaR

 The Variance Covariance VaR approach involves the following steps:

  • Step 1: Decomposing portfolio instruments into a set of simple standardized instruments, such as stocks and zero coupon bonds
  • Step 2: Gathering of historical raw market data
  • Step 3: Market data cleansing , such as back filling any missing data and adjustment for corporate actions
  • Step 4: Derive model parameters, using :
  • Step 5: Compute VaR
    • Mean
    • Standard deviation
    • Covariance matrix



Monte Carlo VaR

 The Monte Carlo VaR method would clearly be a more computationally intensive approach, involving the following steps:

  • Step 1: Choosing appropriate risk factors and associating them with simulation models, thereby assuming some kind of distribution for each risk factor
  • Step 2: Producing model parameterizations, either from historical data (requires the same steps as for historical VaR) and forward implied data
  • Step 3: Generating a covariance matrix for all risk factors
  • Scenario generation using Monte Carlo framework
  • Step 4: Exposure distribution using some pricing model
  • Step 5: VaR or expected loss calculation


Scenario Based Approach (industry standard)

 This approach is most suitable for exchange traded products, and very popular today. It relies on the following steps:

  • Step 1: Choice of a pricing model for each instrument type
  • Step 2: Scenario definitions for each risk factor associated with a pricing model
  • Step 3: Produce portfolio distribution associated with each scenario
  • Step 4: Additional charges to allow for correlation adjustments between products, e.g. calendar spreads

Potential Issues and Best Practices

Given the various IM methodologies outlined above, how can today’s OTC derivative practitioners decide what is the best methodology to use and when? Equally important, what are some potential pitfalls to look out for when utilizing these methodologies, which could result in significantly different IM calculations? 

Clearly, one potential issue involves differences used during the market data preparation step—given different choices of market data sources and different choices of market data cleansing methodologies. A second potential issue could occur during the time series derivation and adjustment steps. Depending on the choice of the methodology, risk factors driving the simulations will differ. For example, would we be observing the changes in the movements of direct quotes (e.g. 10 year swap rate) or derived factors—such as zero rates or PCA-based factors? Furthermore, construction of more complex risk factors such as curves, volatility surfaces or cubes and associated interpolation/extrapolation techniques could introduce further complications.

In addition, when it comes to scenario generation and generation of portfolio price distribution for each scenario, some methodologies could result in different scenarios being generated (all else being equal) due to the choice of simulation models, random number generation and variance reduction techniques.  Moreover, portfolio price distribution could differ for more complex derivatives, depending on the choice of pricing model.

Due to the potential issues described above, very different IM calculations can result. However, regardless of the choice of methodology,we can observe similar patterns across them— and that is why we recommend, as a best practice, keeping the following steps together:

Step 1: Market data preparation, including:

  • Gathering of historical (Raw) Market Data
  • Market data cleansing
    • back filling any missing data        
    • Adjustment for corporate actions
    • Etc.

Step 2: Time series derivation, risk factor adjustments and scenario generation, including:

  • Choice of appropriate risk factors for time series sampling
  • Derivation of risk factors (from step 2 in the Histortical VaR Methodology above)
  • Differencing  of risk factors  and time series adjustments

Step 3: Generation of portfolio price distribution for each scenario

Step 4: Calculation of Initial Margin

Conclusion: Managing the Ambiguity Surrounding Initial Margin Rules

Given the methodologies we’ve touched upon here, and the issues associated with them, we could envision the following potential developments taking place in the future to help alleviate collateral disputes around IM:

  • The construction of a central market data repository
  • The movement toward one methodology, most likely Historical VaR
  • The publishing of scenario definitions by regulators that would be uniformly utilized to calculate portfolio distribution

While the new regulations for OTC derivatives are an important step forward towards making our financial system a safer place, at this point there is still a lot of ambiguity around implementation of these rules and their cumulative impact.

For now, we believe institutions should be looking to minimize the costs of doing business by developing robust systems to help them measure the amount of initial margin and variation margin to be posted on the daily basis. We also recommend forecasting these measures into the future to help manage liquidity needs and to optimize collateral usage.

Need Assistance?

Want More From Numerix? Subscribe to our mailing list to stay current on what we're doing and thinking

Want More from Numerix?

Subscribe to our mailing list to stay current on what we're doing and thinking at Numerix

Subscribe Today!