Journal of Risk Model Validation

Welcome to the first issue of the fourteenth volume of The Journal of Risk Model Validation.

It has been the case that, on occasion, some of our papers have been a little off-piste, with some connection to model validation, but perhaps not as much as desired. I am happy to say that all four papers in this issue directly address topics highly relevant in this area.

The issue’s first paper, “Measuring economic cycles in data” by Joseph L. Breeden, addresses some interesting questions: essentially, how much data do you need to estimate a loan risk model, and what metric can you use to measure this amount? The author proposes a data-driven metric that measures the number of economic cycles weighted by the severity of those cycles, utilizing a state space reconstruction approach. This measure can be employed to describe the amount of useful structure in historic data from a modelling perspective. Data length is a potential risk and can be a component of model risk scoring for corporate model inventories. In my own professional work, I spend a great deal of time reshaping data in various ways to assess its usefulness, so I can endorse the relevance of this topic.

“International Financial Reporting Standard 9 expected credit loss estimation: advanced models for estimating portfolio loss and weighting scenario losses”, our second paper, is by five authors: Bill Huajian Yang, Biao Wu, Kaijie Cui, Zunwei Du and Glenn Fei. Its title gives away what the paper is about. Here, the authors propose a model to estimate the expected portfolio losses brought about by recession risk as well as a quantitative approach to determine the scenario weights. Both model and approach are validated by an empirical example, wherein the authors apply stress testing to the portfolio expected loss using recession risk and calculate the scenario weights accordingly.

The third paper in the issue addresses a topic that I feel I have rather neglected: I refer to internet search as a means of forecasting returns and risk, or behavior more generally. This is clearly a growth area, although a commercially sensitive one, I suspect, because of either profitability or invasion of privacy (or both). In this paper, “Volatility forecasting: the role of internet search activity and implied volatility”, Arabinda Basistha, Alexander Kurov and Marketa Halova Wolfe build on an existing literature that shows that internet search activity is closely associated with volatility prediction in financial and commodity markets. The authors search for a benchmark model with available market-based predictors to evaluate the net contribution of internet search activity data in forecasting volatility. They find that the predictive power of internet search activity data disappears in financial markets and substantially diminishes in commodity markets once the model includes implied volatility. They also find that implied volatility has additional predictive information that is not contained in internet search activity data. This and related research will play an important role in the future, and we expect to see more papers in this area.

Peter Mitic, James Cooper and Nicholas Bloxham propose a novel method for estimating future operational risk capital in “Incremental value-at-risk”, our final paper. Incremental value-at-risk, or IVaR, can be used either for predicting VaR in the short term or as a “sense check” for a capital value that has already been calculated. Its foundation is the difference in data between one capital calculation using an established procedure and the next. It is interesting and, in my view, entirely correct that model validation should involve some form of comparison with a benchmark model, as the last two papers demonstrate.

Steve Satchell
Trinity College, University of Cambridge