Current research on financial risk management applications of econometrics centers on the accurate assessment of individual market and credit risks with relatively little theoretical or applied econometric research on other types of risk, aggregation risk, data incompleteness, and optimal risk control. We argue that consideration of the model risk arising from crude aggregation rules and inadequate data could lead to a new class of reduced-form Bayesian risk assessment models. Logically, these models should be set within a common factor framework that allows proper risk aggregation methods to be developed. We explain how such a framework could also provide the essential links between risk control, risk assessments, and the optimal allocation of resources. Copyright 2005, Oxford University Press.