The OECD has published a paper by Adrian Blundell-Wignall, The Subprime Crisis: Size, Deleveraging and Some Policy Options … it was actually published in April, but I missed it … the numbers weren’t scary enough, I suppose, so it was ignored by bloggers and the media.
The abstract reads:
The paper revises our previous USD 300 bn estimate for mortgage related losses to a range of USD 350-420 bn. In doing this the paper explicitly rejects the previous approach based on implied defaults from ABX pricing, because these prices are affected by illiquidity and extreme volatility; they will likely lead to misleading estimates of losses. Instead it builds a proper default model approach and allows for recovery of collateral via house sales over time. The paper separates out the losses due to commercial banks in the US, and goes on to look at the implied deleveraging required to meet capital standards. It could take 6-12 months for banks to offset losses via earnings alone, depending on Fed rate cuts and the dividend policy of banks. Since even more capital than this is required if banks were to expand their balance sheets, the paper looks at possibilities for capital injections from groups like sovereign wealth funds; and it also looks at a novel plan for the use of public money with an RTC-style approach and the issue of zero coupon bonds. Finally the paper looks at the issues of moral hazard, the likely size of the impact in Europe and Asia and non-bank corporate leverage.
The author points out that mark-to-market estimates are more than just a little suspicious:
The ABX estimates are shown in Table 1. The prices for each tranche/vintage are shown in the top section of the table. Thus in the first row, for ABX 06(1), the 14 March price 86 implies that 14% losses are discounted for AAA.5 The weights by vintage and tranche (not shown) are applied and, the weighted expected loss is shown in the bottom row of the table. This number is applied to the stock of US RMBS. Using the September 7 numbers, USD 292 bn is the implied loss (the main basis of the work last year). But as can be seen, over time the implied size of the losses seems to get ever larger. On the 14th of March, a staggering USD 887 bn loss is implied.
A similar picture emerges from our naïve equity market-cap-loss approach in Table 2. Far from the USD 308 bn published in the last FMT, the market cap losses for levered financial institutions most affected by mortgages is now a staggering USD 702 bn, very much showing the same pattern as the ABX approach.
Both approaches are undermined by recent market panic and problems with price discovery. If it is agreed that these are features of recent experience, then it follows that these estimates of losses are way too high.
This estimate of ultimate losses may be compared with
- Fitch, USD 400-billion
- Greenlaw, USD 400-billion
- IMF, USD 565-billion
- Bank of England, USD 175-billion
Blundell-Wignall does not give a lot of details regarding his calculation. Essentially, he’s fitting into the formula
Total losses = (total outstanding) x (delinquency rate) x (foreclosures / delinquencies) x (loss given foreclosure)
There’s not a lot of information: I have tried and failed to find details of the parameterization of this equation in either the Bank of England model or the OECD model. The best I can do is state that the BoE assumes loss given foreclosure of 50%, while the OECD varies this in a range of 40%-60%.
Additionally, the BoE examined 1,400-billion in sub-prime, while the OECD is looking at 1,300-billion sub-prime and 1,000-billion “Alt-A, etc.”.
[…] last review of loss estimates was the discussion of the OECD paper; there is not really enough data in the BIS note to put it on the board as an estimate … but […]
[…] loss estimates during the crisis ranged from $175-billion to $565-billion. It’s a pity Kirabaeva didn’t footnote his $500-billion figure. The Brunnermeier paper […]