Looking forward to back testing

With increasing challenges to measure value-at-risk and meet high regulatory requirements, the focus has turned to back testing as a way of assuring models' adequacy. Carsten S Wehn proposes a new regime of back testing, combining state-of-the-art methodologies in a decision tree to ameliorate the market risk model

To measure market risk, banks usually predict the distribution of the future (and therefore unknown) gains and losses G_t of a portfolio. Modelling gains and losses is achieved by taking into account all relevant information given by observed risk factor prices or returns, respectively. Thus, the relevant conditional univariate one-step-ahead predictive distribution can be denoted by F_t: = F_t(G_t/R_t