Journal of Risk: Non-parametric Value-at-Risk forecasts

29 July 2012

Whitepaper

New model for Value-at-Risk forecastsimage sanddunes

This paper proposes a new model for computing Value-at-Risk forecasts. To summarize, it incorporates information about the market’s perceived uncertainty about the future from volatility indices. By comparing the model primarily used in the banking sector to our new model, the authors found that a financial institution using our model has on average a lower market induced capital requirement (MCR). However, during the time period leading up to the financial crises our model gives a 40% higher MCR.

Value-at-Risk model

KYOS has incorporated Value-at-Risk (VaR) into its platform of analytical models. Value-at-Risk helps risks managers and traders to manage market risk on a portfolio of positions. It is the standard risk concept in most trading organisations. It gives insight in potential future losses and helps to take the right measures. Furthermore, it is possible to adjust positions, execute new transactions (hedges), or employ more capital as a buffer.

The standard VaR model in the KYOS Analytical Platform is based on the variance-covariance matrix. Other words to describe the same method are parametric VaR, normal VaR or varcovar VaR. It is actually the most easy to use and interpret results. As an input it requires the volatilities and correlations of all monthly prices. The model calculates these parameters then automatically, and base them on historical market prices. Additionally, the user may overwrite the volatility estimates, either to test sensitivities or because he has more accurate estimates from option markets.

As alternative methodologies, KYOS offers the Monte Carlo simulation approach and the historical simulation approach. Especially when the market price returns do not have a very nice Normal distribution, this may be more accurate. For the Monte Carlo simulation approach, the price simulations from KySim are used. For the historical simulation approach, the model takes the variations in market prices directly from the history, without any distribution assumption; that is therefore the non-parametric VaR approach.

Published in Journal of Risk, Vol. 16, No. 4, 2014

Authors: Marcus Nossman (KYOS) and Anders Vilhelmsson

click here to download

Contact KYOS

KYOS
Nieuwe Gracht 49
2011 ND Haarlem
The Netherlands

E-mail: info@kyos.com
Tel: +31 (0)23 551 02 21

Follow us

Contact us via the form below. We will reply within 48 hours.

Request a demo via the form below. We will reply within 48 hours.

    We will send you a maximum of 1 newsletter/mailing per month. You can unsubscribe at any time. Our privacy policy is available under the Contact page.
  • This field is for validation purposes and should be left unchanged.

Book a training course via the form below.

  • (If available)
    We will send you a maximum of 1 newsletter/mailing per month. You can unsubscribe at any time. Our privacy policy is available under the Contact page.
  • This field is for validation purposes and should be left unchanged.

Request a training course via the form below. We will reply within 48 hours.

    We will send you a maximum of 1 newsletter/mailing per month. You can unsubscribe at any time. Our privacy policy is available under the Contact page.
  • This field is for validation purposes and should be left unchanged.

Plan a meeting via the form below. We will reply within 48 hours.

    We will send you a maximum of 1 newsletter/mailing per month. You can unsubscribe at any time. Our privacy policy is available under the Contact page.
  • This field is for validation purposes and should be left unchanged.