Bank efficiency and risk during the financial crisis: Evidence from weight restricted DEA models

The recent nancial crisis highlighted how banks' funding and investment portfolios are associated with their risk taking. In a "normal" DEA banking efficiency model this risk element, embedded in banks' business models, is indicated by possibly inappropriate mixes of funding portfolio and/or investment portfolio which, in turn, is revealed by the weights assigned by each bank to the inputs and outputs in the model. Using the crisis as a natural experiment, we defie weight restrictions to be imposed on all banks by using the banks that were not bailed out during the crisis, or with relatively higher external rating, and which are efficient in the normal DEA model, as our model banks. Thus, banks seemingly taking excessive risks by using inappropriate weights to make themselves look efficient, are restricted to only applying weights also used by efficient non-bailed-out banks. Analysing data collected from audited financial statements of around 70 of the largest EU banks from 2006 to 2009, we find a clear pattern indicating that non-bailed-out banks with relatively high external rating become significantly more efficient than the other banks once weight restrictions are applied to control for risk, even if this pattern was not clear from the models without weight restrictions. In models already incorporating some risk elements, the non-bailed-out banks are significantly more efficient even before weight restrictions are included, but the imposition of weight restrictions makes the pattern even stronger.