(a)

that the institution carefully considers any form of linear or non-linear dependence, relating to all the data, either to the body or to the tail, across two or more operational risk categories or within an operational risk category;

(b)

that the institution supports its correlation assumptions, to the greatest extent possible, on an appropriate combination of empirical data analysis and expert judgement;

(c)

that losses within each operational risk category are independent of each other;

(d)

that where the condition of point (c) is not met, dependent losses are aggregated together;

(e)

that, only where neither of the conditions of points (c) or (d) can be met, dependence within the operational risk categories is appropriately modelled;

(f)

that the institution carefully considers dependence between tail events;

(g)

that the institution does not base the dependence structure on Gaussian or Normal-like distributions;

(h)

that all assumptions regarding dependence used by the institution are conservative given the uncertainties relating to dependence modelling for operational risk, and that the degree of conservatism used by the institution increases as the rigour of the dependence assumptions and the reliability of the resulting own funds requirements decrease;

(i)

that the institution properly justifies the dependence assumptions it uses and that it regularly performs sensitivity analyses with the view to assessing the effect of the dependence assumptions on its AMA own funds requirements.