Maybe the Bayesian school can offer us some insight in operational risk by defining management causality that cannot be measured adequately using VaR, RAROC or EVA (economic value-added). Bayesian networks constitute a branch of Bayesian conditional probability theory, e.g.
Prob ( A) = Prob( B ) given Prob(C )
Where A, B, C are discrete events.
Prob (company defaulting) = Prob (20 % share price fall ) given Prob (bad CEO)
This has some potential for creating deductive causal links in the loss database.
The use of Bayesian probability has uses in VaR in that we can build conditional VaR modelling.

There are additional techniques to deal with the risk spectre.
Scenario analysis
Scenario analysis lets us think wider to encapsulate more dramatic risk events. These include the Exxon Valdez and Hurricance Andrew extreme risk events that drove many companies to the brink of bankruptcy. There cannot have been a more damaging and unthinkable recent disaster than September 11th.
The limits of thinking and use of business imagination are widened under scenario analysis, consistent with a set of given possible future events for brainstorming.
Scenario testing has been applied in preference to other modelling techniques in some cases because it is easier to comprehend. For example, the pension funds around the world have used simple what-if scenarios at times to determine the money left to cover the payment of pensions.
Let us say that scenarios of 4 %, 6 %, 8 % and 10 % rates of return were taken up by some banks. These were somewhat conservative given the double-digit real returns on the stock markets in the 1990s. Yet, as we have seen, investor behaviour can be irrational and place too much weight upon recent experiences. This is unrealistic against the evidence of a probability density function (PDF) for percentage returns. A comparison of market values against PDF numbers gives a more objective view than being caught in a buying mania.
Unfortunately, pension funds should have foreseen the impending difficulties in the value of stock markets, and made provisions accordingly under more pessimistic scenario analysis. A 4–6 % annual return would have kept the pension funds more secure, but many got carried away in the optimism of the stock market and have not kept provisions in reserve for future bad debts.37 Life insurance companies also guaranteeing 8 % pay-outs to their policy-holders was a disastrous move when stock market returns fell to record low levels.
For pensions and life annuities, things have generally gone from bad to worse. The UK pensions mis-selling episode was estimated to have cost around £12 billion to future pensioners. The UK disaster over shortfalls in endowment mortgages has been variously projected to cost many times more than this figure. A benevolent summary of this situation is a business-like patch over what could prove a very messy and expensive mistake. Some would call this compensation exercise a fudge, but it has given a limited redress for victims, and it has put in place damage-limitation procedures.
Scenario analysis does not complicate the risk models, but simplifies their application. Well-designed scenarios can bring to light weaknesses of risk management systems, including modelling vulnerability. Results help to identify critical procedures necessary for reducing risk and conserving capital, returns, market position, core competencies and reputation. Scenario analysis should lead to changes in the way to allocate capital, or planning contingency procedures. Scenario analysis suffers from being less standardised than the more established VaR, and it makes little use of complex mathematical modelling. There is less in a way of a formal methodology or modelling tool, but it rests upon widely accepted engineering and actuarial techniques. It is often a difficult technique to sell internally within the company.