Crash of 2008/Tutorials: Difference between revisions
imported>Nick Gardner |
imported>Nick Gardner |
||
(20 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
{{subpages}} | {{subpages}} | ||
== | ==Sources of risk-management errors== | ||
===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. <ref>[http://www.riskworx.com/insights/theory/theory.pdf] Barry du Toit ''Risk, theory, reflection: Limitations of the stochastic model of uncertainty in financial risk analysis'' Riskworx June 2004</ref>. The assessment were thus inapplicable to risks due to policy errors. | |||
<ref>[http:// | ===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" <ref>[http://faculty.chicagogsb.edu/steven.davis/research/Interpreting_the_Great_Moderation,_NBER_version.pdf Stephen Davis and James Kahn: ''Interpreting the Great Moderation'', National Bureau of Economic Research Working Paper 14048, May 2008]</ref>, 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 <ref>[http://oversight.house.gov/images/stories/Hearings/Committee_on_Oversight/RaiterTestimony.pdf Frank Raiter: written statement before the Hearing on Credit Rating Agencies and the Financial Crisis by the House of Representatives Committee on Oversight and Government Reform, October 22 2008]</ref>. Later - improved and updated - models had been developed, but were not implemented because of budgetary constraints.) | |||
<ref>[http:// | |||
* 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 <ref>[http://oversight.house.gov/documents/20081022124926.pdf Statement by the President and CEO of Fitch Inc to the Hearing on Credit Rating Agencies and the Financial Crisis by the House of Representatives Committee on Oversight and Government Reform, October 22 2008]</ref>.) | |||
===Tail risk=== | ===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) <ref>[http://www.world-economics-journal.com/Contents/ArticleViewer.aspx?ID=341 Andrew Haldane: "Risk-Pricing and the Sub-Prime Crisis", ''World Economics'' July-September 2008]</ref> suggests that they arose from investors' and rating agencies' use of linear models based upon the CAPM ''(Capital Asset Pricing Model)'' <ref> See paragraph 2.3 of [[Financial economics]]</ref>. 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. | 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) <ref>[http://www.world-economics-journal.com/Contents/ArticleViewer.aspx?ID=341 Andrew Haldane: "Risk-Pricing and the Sub-Prime Crisis", ''World Economics'' July-September 2008]</ref> suggests that they arose from investors' and rating agencies' use of linear models based upon the CAPM ''(Capital Asset Pricing Model)'' <ref> See paragraph 2.3 of [[Financial economics]]</ref>. 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. | ||
<ref>[http://faculty.chicagogsb.edu/raghuram.rajan/research/finrisk.pdf Raghuram Rajan: ''Has Financial Development Made the World Riskier?'' , Working Paper No 11728, National Bureau of Economic Research September 2005]</ref> . 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. | <ref>[http://faculty.chicagogsb.edu/raghuram.rajan/research/finrisk.pdf Raghuram Rajan: ''Has Financial Development Made the World Riskier?'' , Working Paper No 11728, National Bureau of Economic Research September 2005]</ref> . 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 | ===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 | |||
<ref>[http://oversight.house.gov/documents/20081022111050.pdf Confidential Presentation to Moody's Board of Directors, October 2007]</ref>. | |||
<ref>[http://oversight.house.gov/documents/20081022111050.pdf Confidential Presentation to Moody's Board of Directors, October 2007 | |||
==References== | ==References== | ||
<references/> | <references/> |
Latest revision as of 08:31, 11 January 2010
![](http://s9.addthis.com/button1-share.gif)
Sources of risk-management errors
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. [1]. 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" [2], 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 [3]. 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 [4].)
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) [5] suggests that they arose from investors' and rating agencies' use of linear models based upon the CAPM (Capital Asset Pricing Model) [6]. 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. [7] . 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 [8].
References
- ↑ [1] Barry du Toit Risk, theory, reflection: Limitations of the stochastic model of uncertainty in financial risk analysis Riskworx June 2004
- ↑ Stephen Davis and James Kahn: Interpreting the Great Moderation, National Bureau of Economic Research Working Paper 14048, May 2008
- ↑ Frank Raiter: written statement before the Hearing on Credit Rating Agencies and the Financial Crisis by the House of Representatives Committee on Oversight and Government Reform, October 22 2008
- ↑ Statement by the President and CEO of Fitch Inc to the Hearing on Credit Rating Agencies and the Financial Crisis by the House of Representatives Committee on Oversight and Government Reform, October 22 2008
- ↑ Andrew Haldane: "Risk-Pricing and the Sub-Prime Crisis", World Economics July-September 2008
- ↑ See paragraph 2.3 of Financial economics
- ↑ Raghuram Rajan: Has Financial Development Made the World Riskier? , Working Paper No 11728, National Bureau of Economic Research September 2005
- ↑ Confidential Presentation to Moody's Board of Directors, October 2007