PrefBlog

The fascinating Chart 1.18 in the April 2008 BoE Financial Stability Report led me to a paper in their 2007Q4 Bulletin by Lewis Webber & Rohan Churm, that seeks to decompose corporate bond spreads.

They use a Merton Model with a default boundary to estimate intrinsic default probabilities, which is similar to the Bank of Canada CDS Pricing research mentioned briefly on March 20. Further, the intrinsic default risk is differentiated from the premium demanded due to uncertainty regarding this intrinsic default risk by:

It is assumed that, in practice, corporate bond investors demand compensation for bearing both expected and unexpected default losses. The sum of these two components is calculated using the model by assuming that investors recognise the uncertainty surrounding the firm’s asset value growth rate. They therefore discount the future cash flows they expect in practice at a risky rate of return to reflect the possibility of default occurring looking forward. To isolate the compensation demanded for expected default losses, it is assumed that investors continue to expect risky rates of return, but instead discount expected cash flows at the default risk-free rate. Compensation for bearing the risk of unexpected default losses can then be obtained as the difference between these two values.

Equivalently, the total compensation investors demand for bearing expected and unexpected default losses is calculated in the model using risk-neutral valuation methods. This involves calculating the expected default frequency used in equation (5) under the risk-neutral probability measure. Compensation for expected default losses is isolated by calculating the expected default frequency used in equation (5) under the real-world probability measure.

Finally, they assign the residual between calculated and market yields – mostly a liquidity premium:

In addition, there may be a residual part of observed corporate bond spreads that the model cannot explain. This contains compensation for all non-credit factors, including a premium for the relative illiquidity of the corporate bond market compared to the government bond market. This gives three contributions to observed corporate spreads: the compensation investors demand for expected default losses; compensation for uncertainty about default losses; and a non-credit related residual.

There are many fascinating graphs!

The Bank of England has released its April, 2008, Financial Stability Report and it’s excellent. There’s way too much good stuff in it to make a series of extracts of managable size, but I will point out that they have highlighted the folly of relying on market prices to estimate ultimate credit losses:

But credit losses from the turmoil are unlikely to ever rise to levels implied by current market prices unless there is a significant deterioration in fundamentals, well beyond the slowdown currently anticipated. That is because prices are likely to reflect substantial discounts for illiquidity and uncertainty that have emerged as markets have adjusted but which should ease over time. While market-based estimates and the write-downs announced by firms may be unduly pessimistic, if such concerns persist there is a risk they could become self-fulfilling.
…
This suggests that spreads on senior tranches of structured credit products have been dominated by illiquidity and uncertainty premia and a large relative fall in demand for AAA tranches rather than credit risk over the recent period. This is consistent with the analysis in Box 1, which suggests that the largest gap between mark-to-market and cash-flow based estimates of losses on sub-prime securities is in the AAA tranches.

If this were the case, long-term and unleveraged investors could potentially profit by holding these AAA tranches to maturity. But Chart 1.22 shows that, until very recently when AAA tranches have risen in price, an investor following this strategy would have suffered a string of negative month-on-month returns over the past year. And as discussed in Section 3, many long-term investors often face implicit short-term performance targets and increasingly have to mark their portfolios to market, even when they have no intention of selling securities.

The BoE contention that market prices – especially marks implied by such things as the ABX indices – are stupid is something that I’ve been saying for quite some time (in my review of the estimates by Greenlaw, et al. and in the seemingly very closely related IMF report). These reservations were echoed by BoC Governor Carney in remarks reported May 1.

Update, 2008-5-12: Willem Buiter doesn’t believe the Bank’s analysis:

But why oh why did the FSR use American subprime mortgages to make the point that marked-to-market estimates of credit losses may well exaggerate the likely eventual magnitude of these losses? I will quote the Bank’s reasoning at length, so as to be sure I don’t misrepresent it:

“Future credit losses can be estimated by extrapolating forward delinquency rates. In particular, it is assumed that serious delinquency rates of US sub-prime mortgages of different issuance ‘vintages’ continue to rise at their average rates to date until the mortgages are four years old, at which point the rate is assumed to plateau. This is a stylised representation of the way that serious delinquency rates of older sub-prime mortgages have evolved. This method results in peak delinquency rates of 34% for mortgages issued before 2006 H1, rising to 42% for mortgages issued in 2007 H2. Upon becoming seriously delinquent, mortgages are assumed to default with at least 75% probability after one year, and to have a loss given default (LGD) rate of 50%. Chart A (not shown, WHB) shows the resulting projection, in which credit losses eventually reach around US$170 billion. AAA-rated securities do not incur losses in this projection. But there is sufficient uncertainty that even these top-rated securities could conceivably bear some losses.”

This Bank estimate is then contrasted with marked-to-market estimate (or rather marked-to-model estimates using some market inputs from ABX markets to obtain estimates of credit risk on home equity loan asset-backed securities) of US$380bn. Other estimates from the wilder reaches of Wall Street (think of a number and double it) get up to US$500bn or close to US$1 trillion.
…
The Bank’s analysis, quoted above, is nevertheless likely to turn out to be complete bollocks because it uses assumed delinquency rates that are based on a “stylised representation of the way that serious delinquency rates of older sub-prime mortgages have evolved.” The burst of subprime lending and borrowing between 2003 and early 2007 was, however, unlikely anything ever seen before. A whole new population of subprime borrowers entered the market for the first time. Mortgage borrowers with these characteristics were not in the older subprime population.
…
So as far as I am concerned, we have no reliable information on which to base an estimate of the value of the subprime mortgages issued since 2003.

Well … I fully agree with that last line taken from Dr. Buiter’s analysis! But investors (and policy makers) have to come up with something.

All in all, I am much more favourably disposed to the Bank’s “bottom-up” methodology of taking observed delinquency rates and extrapolating them than to the IMF’s “top-down” approach of marking-to-ABX-market. One just has to remember it’s a forecast … no better and no worse than any other forecast of financial markets or the weather.

Update #2, 2008-5-12: Research Recap has extracted one of the charts of interest:

The Economist has acknowledged the report but warns:

The report also suggests that market valuations of losses on America’s subprime mortgages—to which British and European as well as American banks are exposed—may prove exaggerated. Eventual losses could turn out to be around $170 billion (£86 billion) rather than the market-based figure of $380 billion.

Another reason for greater confidence is that British banks are making more realistic assessments of their bad debts and raising capital. RBS led the way on April 22nd with a writedown of almost £6 billion and tapped shareholders for £12 billion. A week later HBOS announced a rights issue of £4 billion.

Yet even if the report is right in suggesting that the self-inflicted financial wounds may gradually be starting to heal, the worry now is the damage that has already been done to the economy. The bigger that turns out to be, the greater the potential for a second round of financial pain through defaults arising from a slowdown or recession.

More extracts with light commentary are provided by FTAlphaville.