Model risk – daring to open up the black box. (22nd December 2015)

• Record Type: Journal Article
• Authors: Aggarwal, A.; Beck, M. B.; Cann, M.; Ford, T.; Georgescu, D.; Morjaria, N.; Smith, A.; Taylor, Y.; Tsanakas, A.; Witts, L.; Ye, I.
• Is Part Of: British actuarial journal. Volume 21:Number 2(2016)
• Page Start: 229
• Page End: 296
• Publication Date: 2015-12-22
• Subjects: Model, -- Model Risk, -- Model Error, -- Model Uncertainty, -- Risk Culture; Actuarial science -- Periodicals; Insurance, Life -- Periodicals; Periodicals; Periodicals; 368.010941
• DOI: 10.1017/S1357321715000276
• Languages: English
• ISSNs: 1357-3217

Abstract

Abstract: With the increasing use of complex quantitative models in applications throughout the financial world, model risk has become a major concern. Such risk is generated by the potential inaccuracy and inappropriate use of models in business applications, which can lead to substantial financial losses and reputational damage. In this paper, we deal with the management and measurement of model risk. First, a model risk framework is developed, adapting concepts such as risk appetite, monitoring, and mitigation to the particular case of model risk. The usefulness of such a framework for preventing losses associated with model risk is demonstrated through case studies. Second, we investigate the ways in which different ways of using and perceiving models within an organisation both lead to different model risks. We identify four distinct model cultures and argue that in conditions of deep model uncertainty, each of those cultures makes a valuable contribution to model risk governance. Thus, the space of legitimate challenges to models is expanded, such that, in addition to a technical critique, operational and commercial concerns are also addressed. Third, we discuss through the examples of proxy modelling, longevity risk, and investment advice, common methods and challenges for quantifying model risk. Difficulties arise in mapping model errors to actual financial impact. In the case of irreducible model uncertainty, it is necessary to employ a variety of measurement approaches, based on statistical inference, fitting multiple models, and stress and scenario analysis.