Credit rating agency/Addendum

Estimates of implied default probabilities

(January 2002–June 2006 in basis points)

Moody's

	Credit rating	Sovereign	Corporate
Investment grade	Aaa-AA	0	0
	A	5.4	6.1
	Baa	48.8	40.6
Speculative grade	Ba	64.0	139.1
	B	123.5	280.5

Standard and Poor

	Credit rating	Sovereign	Corporate
Investment grade	AAA-AA	0	0
	A	4.5	9.4
	BBB	45.6	41.7
Speculative grade	BB	111.9	139.3
	B	266.4	315.7

Fitch

	Credit rating	Sovereign	Corporate
Investment grade	AAA-AA	0	0
	A	13.2	12.0
	BBB	58.8	39.6
Speculative grade	BB	92.8	90.9
	B	142	125.3

(Source: Bank for International Settlements Quarterly Review, March 2007[2])

Risk assessment errors

Introduction

Reports of inquiries into the crash of 2008 revealed the existence of errors made by some credit rating agencies, arising from the use of unrealistic assumptions, unrepresentative data, and bias.

The fundamental assumption

A judgement-free assessment of risk is possible only on the basis of an assumption about the probability distribution of the relevant variables, and the assumption made in the risk-management models is that the future distribution is the same as the distribution observed during a past period. It has been pointed out that a favourable assessment from a risk model embodying that assumption amounts to saying that "all will be well unless something goes wrong".

The Stochastic assumption

The underlying assumption of the portfolio theory upon which the risk assessments were based is that investment risks are stochastic rather than deterministic - that is to say, the assumption that they arise from the existence of random fluctuations, and not as a consequence of human behaviour. . The assessment were thus inapplicable to risks due to policy errors.

Data limitations

The data used to estimate risk probabilities were subject to several limitations:

• They were taken from the period of historically low economic volatility that started in the early 1980s and came to be known as the "great moderation" , and as such were applicable only on the assumption that such low volatility would continue. The fact that volatility did, in fact increase was recognised but not acted upon. (According to the former Managing Director and Head of Residential Mortgage-backed Securities Rating at Standard and Poor's, his company had assessed default probabilities using a model based upon the analysis of 900,000 mortgages that had been implemented in 1996, and which did not, therefore, capture the changes in performance brought about by the subsequent increase in the numbers of subprime mortgages . Later - improved and updated - models had been developed, but were not implemented because of budgetary constraints.)

• They were contaminated by the fact that rescue action had averted some downside risks and thus embodied the assumption that similar action would take place in the future. (The president of another credit rating agency said the their ratings had been made on the assumption that there would be government financial support for any major company that ran into in difficulties .)

Tail risk

An explanation for risk-management errors that has been put forward by Andrew Haldane (Head of the Bank of England's Systemic Risk Assessment Department) suggests that they arose from investors' and rating agencies' use of linear models based upon the CAPM (Capital Asset Pricing Model) . Such models assume that risks can be represented by the symmetrical bell-shaped normal distribution, and can give inaccurate results if the true distribution has a "fat tail", as a result of which there is a significant additional tail risk. Earlier work by Raghuram Rajan (Director of Research at the International Monetary Fund) suggested that securitised assets may be expected to involve significant tail risks. . Since the events involving such risks are by definition rare, they cannot be expected to be picked up by models based upon a five or six years' run of data. The errors in pricing the riskiest tranches of mortgage-based derivatives were estimated to have amountad to as much as an additional 9 per cent per annum.

Credit rating agency bias

The fact that the income of the credit rating agencies came from the issuers of the securities may have introduced some bias to their risk ratings. A 2007 report to the Board of Moody's spoke of a conflict between ratings quality and the defence of market share, and of the danger that ratings of securities might be influenced by pressure from their issuers .