In statistics we need to adjust confidence levels to significance of conclusions levels.

Sir, I refer to Tim Harford’s “Highs and lows of minimum wages”, March 28.

In it Harford writes: “A fascinating survey reported in the World Development Report showed World Bank staff some numbers and asked for an interpretation. In some cases, the staff was told that the data referred to the effectiveness of a skin cream; in other cases, they were told that the data were about whether minimum wages reduced poverty. The same numbers should lead to the same conclusions but staff had much more trouble drawing the statistically correct inference when they had been told the data were about minimum wages. It can be hard to set aside our preconceptions.”

Sorry, is that not good? Should World Bank staff not be more careful about drawing a statistically correct inference from data relating to skin cream than from data relating to something that could have such profound implications as minimum wages? In statistics, besides confidence levels, do we not need significance of conclusion levels?

For instance I sure wish that some staff somewhere, when presented data concerning the credit risk of bank borrowers, for the purpose of setting the equity requirements for banks, would have had the sufficient presence of mind to remind everyone of that what they really needed was data about what caused major bank failures… something completely different.

Had someone done so, and had someone been able to make bank regulators listen, the world would have saved itself many tears and much trouble.

@PerKurowski

PS. You might have noticed a made a slight mistake :-) I left it that way because, most probably, those who not understand what I really meant, might not be able to understand it even if I corrected it.

A bank’s expected failure going from once in 1000, to once in 200 years, does not sound like an impressive improvement :-)

Sir, Gina Chon refers to Steve Strongin, head of Goldman’s investment research division stating: “In the past the mean time for the failure of a well-capitalized bank was 41 years… Now, with increased capital standards and stress tests scrutinizing how banks would withstand a crisis, it is estimated to be about 200 years”, “Dodd-Frank rules blamed for curbing growth” July 23.

To help you understand what an unbelievable scenario for bullshit that represents, let me mention that in the Explanatory Note on the Basel II IRB Risk Weight Functions of July 2005, the confidence level is described as “fixed at 99.9%, i.e. an institution is expected to suffer losses that exceed its level of tier 1 and tier 2 capital on average once in a thousand years. This confidence level might seem rather high. However, Tier 2 does not have the loss absorbing capacity of Tier 1. The high confidence level was also chosen to protect against estimation errors that might inevitably occur from banks’ internal Probability of Default (PD), Loss Given Default (LGD) and Exposure at Default (EAD) estimation, as well as other model uncertainties.”

Chon mentions “A senior Obama administration official said banks had overreacted and argued that a person with a high credit score should be able to obtain a mortgage on decent terms, which was not happening at many banks”. That official should ask regulators to explain that the system in place is first the banks reacting to perceived credit risks with interest rates, size of exposures and other terms… and then having the regulators, for good measure, to also react to the same perceived credit risks by means of setting the capital the bank needs to hold against assets… and, of course, reacting twice to the same risk, must cause an overreaction.

In this respect the Dodd-Frank Act cannot be much blamed for curbing growth that is unless you feel, like I do, that in the home of the brave, that Act should have prohibited the odious system of risk weighing the capital requirements of banks, something which negates the fair access to bank credit to for instance all SMEs.