Historical Data

What information do you use to calculate risk?

We retrieve the historical adjusted closing prices for all factors that are required to price a security. Sometimes a single security might require multiple factors. For example: an option on a non-USD stock will require the historical stock information, the historical foreign exchange and historical volatility information.

These closing prices (adjusted for dividends and splits when applicable) are then converted into non-overlapping returns.

In order to compute different risk aggregations, we also need to retrieve many characteristics about the security, such as: average volume traded, market capitalization, sector, country of risk to name a few.

How do you use the historical information?

We use historical closing prices as follows:

1. We collect the closing prices (OR rates) for various risk factors that affect the pricing of a security (rates, credit spread, FX, etc). We never use the historical prices of the securities themselves as their prices are not invariant (for example: as an option gets closer to its expiration the risk is very different from 6 months earlier)
2. We put more emphasis on more recent information
3. We build a sparse covariance matrix that is used to generate simulations
4. We fully reprice each security using the simulated risk factors and extract various risk statistics

Do you apply some weighting to historical data?

Yes, we want more recent data to have more weight. Events that happened in the far past have less bearing on what might happen in the near future. We apply 2 different weighting schemes: one for the volatilities of the securities and one for the correlation.

Volatilities have shorter memory: in 2 months, the importance of the information is half of today’s.

Correlations have longer memory: in 6 months the importance of the information is half.

To change these half lives in a calculation, see Exponential Decay.