Journal of Risk Model Validation

This issue of The Journal of Risk Model Validation looks at a good spread of issues and, curiously, contains two single-author papers: a most unusual occurrence. Financial research has become a team activity and the solitary scholar is now very uncommon.

Our first paper, “Shrunk volatility value-at-risk: an application on US balanced portfolios” by Stefano Colucci, tests the naive model for forecasting exante value-at-risk (VaR) using a shrinkage estimator between realized volatility estimated on past return time series and implied volatility quoted on the market. Colucci argues that the VaR prediction strategy uses information about both the expected future risk and the past estimated risk and extends these ideas by modeling nonnormality. He also validates the models under loss function backtests, and his results confirm the efficacy of the implied volatility indexes as inputs for a VaR calculation. While some of our papers have, at times, had weak associations with validation, it is good to see a paper in which validation is central to the exposition.

The issue’s second paper, “Evaluating the risk performance of online peer-to-peer lending platforms in China” by Chong Wu, Dong Zhang and Ying Wang, is a welcome contribution because we have published little or nothing on peer-to-peer (P2P) platforms previously, and the new issues that these structures raise in terms of validation and backtesting require careful thought. To begin with, the whole nature of counterparty risk probably needs to be rethought. The author shows that the testing results for the proposed risk assessment for P2P platforms in China are, in the author’s words, “rational and efficient, giving us unique insight into how to evaluate the risk performance of P2P platforms, and thus can be used by regulators to better supervise these platforms in China”. It is quite possible that the techniques used here are applicable in other P2P markets.

The third paper in this issue, “Smoothing algorithms by constrained maximum likelihood: methodologies and implementations for Comprehensive Capital Analysis and Review stress testing and International Financial Reporting Standard 9 expected credit loss estimation” by Bill Huajian Yang, addresses issues of nonmonotonicity between higher risk grades and higher default loss. This is an important issue because, as every applied statistician knows, unless there are very few grades and a lot of data, unconstrained estimates do not always exhibit these desirable properties. The author proposes smoothing algorithms for rating level probability of default and rating migration probability. The smoothed estimates that are used are optimal in terms of constrained maximum likelihood, with a risk scale determined by constrained maximum likelihood, leading to more robust credit loss estimation – which, it is claimed, is straightforward to calculate. It is good to see a technically rigorous paper that clearly has practical application.