Journal of Risk Model Validation

The conference season is, for me at least, well-defined as the three months covering September to November, and I have attended, and in some cases presented at, five conferences during this period this year. (I believe some might call me "a conference junkie".) I was struck by how little was said about risk model validation during these conferences, especially by vendors; hopefully this will change in time.

The papers in this issue mainly concern value-at-risk (VaR): a fairly hardy perennial for this journal.

Our first paper, "Backtesting value-at-risk: a comparison between filtered bootstrap and historical simulation" by Dario Brandolini and Stefano Colucci, compares ex ante VaR estimation produced by two risk models: historical simulation and Monte Carlo filtered bootstrap. The authors carry out a number of tests using a number of data sets. Their results show a clear difference between the two procedures and further analysis is carried out and implications for risk validation discussed.

The second paper in the issue, "Scientific stochastic volatility models for the European energy market: forecasting and extracting conditional volatility" by Kai Erik Dahlen and Per Bjarte Solibakke, involves the construction of a multifactor stochastic volatility model for the latent (and unobservable) volatility of the baseload and peakload forward contracts at the European Energy Exchange AG. Sophisticated estimation procedures are used to take account of the problems caused by volatility being a latent variable. Forecasts (filtering) of one step-ahead conditional volatility are fundamental to the model validation process. The authors stress the importance of valid models, which in a broad sense is what validation is concerned with.

The third paper in the issue, "Capturing value-at-risk in futures markets: a revised filtered historical simulation approach" by Chang-Cheng Changchien, Chu-Hsiung Lin and Wei-Shun Kao, modifies the filtered historical simulation developed by Barone-Adesi et al, using a general power weighted moving average estimator simulation to forecast VaR. The authors claim that their proposed approach is relatively simple and computationally straightforward compared with those of McNeil and Frey (2000) and Barone-Adesi et al (1998). Using backtesting of historical daily return series of five futures prices, they show that their proposed method would provide an improvement in the precision of VaR forecasts in times when credit is doing badly. It would be excellent if the original authors cited above would submit a "counterblast" to these claims so that we can generate a debate.

The last paper in the issue, "Assessing the performance of generalized autoregressive conditional heteroskedasticity-based value-at-risk models: a case of frontier markets" by Dany Ng Cheong Vee, Preethee Nunkoo Gonpot and Noor Sookia, assesses the performance of twelve generalized autoregressive conditional heteroskedasticity (GARCH)-type models for modeling 99% VaR for indexes from countries classified as frontier economies: namely, Mauritius, Tunisia, Sri Lanka, Pakistan, Kazakhstan and Croatia. Leverage effects and long memory in volatility are considered by making use of exponential GARCH (EGARCH), GJR (Glosten, Jagannathan and Runkle) and integrated GARCH (IGARCH) models. The backtesting process is twofold, making use of recognized hypothesis testing for assessing coverage and loss functions for accuracy. Models are estimated using pre-financial crisis data and backtested over 500 days (within the 2008 crisis) of increased volatility. Despite returns exhibiting high kurtosis, the assumption of normality for innovations in the GARCH models sometimes provides better VaR estimates. The standard GARCH model also proves more useful than asymmetric models or the IGARCH in several cases. This study indicates that one model cannot be singled out for all the indexes and that correct VaR estimation requires different model specifications for each market. This conclusion is particularly interesting as the editor has, over a long working life, worked for global managers who insist on a homogeneous process across all markets.

Finally, as indicated above, I would welcome submissions that challenge our published papers.

REFERENCES

Barone-Adesi, G., Boutgoin, F., and Giannopoulos, K. (1998). Don't look back. Risk 11(8), 31-58.

McNeil, A., and Frey, R. (2000). Estimation of tail-related risk heteroscedastic financial time series: an extreme value. Journal of Empirical Finance 7(3), 271-300.