This thesis contributes to advance the fields of Finance and Financial Risk Management. The first chapter introduces a new risk measurement technique called Capital at Risk (CaR) for the evaluation of economic capital requirements of banks and other financial institutions. CaR does not rely on the choice of a specific quantile and contemplates the entire loss distribution. It has the least number of breaches with respect to 99%-Value at Risk and 97.5%-Expected Shortfall. CaR is subadditive also in the case of extreme heavy-tails. We recommend the adoption of CaR for the evaluation of economic capital requirements as complement to the risk measures currently in use.

The second chapter extends the understanding of the link between Operational Risk (OR) losses and macro-economic factors. Our results confirm the connection with GDP postulated by previous literature. As novelty element, the Governance Indicators are the macro-factors with the largest number of links made with the Event Types losses. The adoption of a jackknife bias correction provides estimates with a lower bias and mean squared error (MSE). The findings of this study greatly re-encourage the adoption of macro-economic factors for the internal risk management processes of risk assessment and mitigation and dispel the myth of operational risk being exclusively a bank specific risk.

In the third chapter, we revisit the ubiquitous practice of creating portfolios by sorting financial returns according to a given variable. The sorting is usually done brute force and ignores the estimation error present in the measurement of the sorting variable. Also the estimation error of the quantile is ignored. We propose a procedure to control for this and show that ignoring this error may produce a substantial classification error. The importance of portfolio sorts is not only acknowledged in Finance but also in Risk Management.