Capital and Risk: New Evidence on Implications of Large Operational Losses

Abstract

Operational risk is currently receiving significant media attention, as financial scandals have appeared regularly and multiple events have exceeded one billion dollars in impact. Regulators have also been devoting attention to this risk and are finalizing proposals that would require banks to hold capital for potential operational losses. This paper uses newly available loss data to model operational risk at internationally active banks. Our results suggest that the amount of capital held for operational risk will often exceed capital held for market risk and that the largest banks could choose to allocate several billion dollars in capital to operational risk.

Expected Shortfall Estimation for Apparently Infinite-Mean Models of Operational Risk

“…As shown in Moscadelli (2004), de Fontnouvelle (2005), de Fontnouvelle et al (2003de Fontnouvelle et al ( , 2005, and further discussed in Fiordelisi et al (2014), Neslehová et al (2006) and Peters and Shevchenko (2015), a peculiar characteristic of operational risk is that the distribution of losses is extremely heavy-tailed, showing a clear Paretian behavior for the upper tail, when we consider losses as positive amounts. Following the standard division of banks' activities into business lines, as required by the so-called standardized and advanced measurement approaches (BCBS 2011b(BCBS , 2014, Moscadelli (2004) has even shown that for corporate finance, trading and sales, and payment and settlement, the loss distribution has a right tail thick enough to prevent it from having a finite mean (as the shape parameter ξ > 1, see Subsection 2.2 for more details).…”

The cost of operational risk loss insurance

“…For comparison with the Basel II (Basel Committee on Banking Supervision 2005) classification of oprisk events and the related empirical studies which also use this classification (see de Fontnouvelle et al 2004de Fontnouvelle et al , 2005Chernobai et al 2008), we use the Basel II partitioning of the oprisk events in our subsequent estimation. It is possible, however, to group oprisk events into those that are either agency cost or system based (see Jarrow 2008, p.…”

“…The existing literature on operational risk studies two issues: one, the estimation of operational risk loss processes using either extreme value theory or Cox processes, see Chavez-Demoulin et al (2006), Coleman (2003), de Fontnouvelle et al (2004), de Fontnouvelle et al (2005, (2001), Embrechts and Puccetti (2006), Jang (2004), Moscadelli (2004), Lindskog and McNeil (2003), Neslehova et al (2006), and Chernobai et al (2008); and two, the application of these estimates to the determination of economic capital, see de Fontnouvelle et al (2004de Fontnouvelle et al ( , 2005 and Moscadelli (2004). In the determination of economic capital required for operational risk, the estimates appear to be at least as large as that necessary to cover market risk.…”

A naive uncertainty model for measuring operational risks faced by financial institutions

“…de Fontnouvelle et al (2003) fit loss distributions to data on operational loss events, and conclude that capital requirements for operational loss events often exceed capital requirements for market risk at large US banks, and Allen and Bali (2007) provide quantitative evidence on the magnitude of operational losses by analyzing monthly stock price data. Both studies show that OR is considerable.…”