Abstract: Financial risks are usually analysed by type and by activity using different assumptions and methodologies as may seem appropriate in each case. This approach makes it very difficult to ascertain the degree of diversification between various activities and to obtain a proper estimate of global risk. We show that different risk aggregation methodologies should be used depending on the purpose of the exercise. In particular, if it is to promote an efficient allocation of resources, a short term, normal circumstances view should be adopted, but if it is to ensure a high degree of financial soundness over the long term, then extreme circumstances and contingency plans should be explored. We propose a simple linear risk factor model in the first case but suggest that a full business model is required for the second. Finally, financial regulators raise an intermediate question that is almost impossible to answer, namely, what is the minimum level of capital consistent with a probability of default of the firm of 0.1% over one year, that is consistent with a single ?A? rating. We suggest that an extension of our normal risk factor model to estimate ?tail? effects could give a better approximation than the current regulatory rules.