From the comments - believing in Magic

Bondinvestor (whoever they may be) left a reasoned piece in the comments. It disagreed with my position on Magic outlined in this post.

I like people that disagree with me and I would love BondInvestor to get in contact...

bondinvestor said...

MTG, PMI, RDN and even ORI and GNW are all buys.

the best way to think about MI conceptually is as a bridge loan. if home prices collapse or the borrower loses a job in the first 3 yrs, the loan goes bad. otherwise, it's a solid investment.

the critical question to getting these stocks right is the level of cumulative defaults on 06/07 flow business. 05 is bad, but not catastrophic. the bulk business has already been written off. therefore, the only lingering uncertainty is what the level of defaults on 06/07 will be.

here is how to answer the question. MGIC's total claims paying resources (investments + collateralized reinsurance recoverable + the present value of installment premiums) is roughly $13B. that is 24% of the total risk in force. the subprime debacle (for which you were right to short the stock) will end up costing them about 8% of risk in force on a cash basis. that leaves 16% of risk in force (roughly $9B) available to satisfy claims on 06/07 business.

that leaves roughly 2/3rds of the claims paying ability ($9B) available to satisfy losses on flow business. the problems are concentrated in 06/07. the total risk in force on 06/07 is roughly $20B (may be high). so in order for MGIC to go bust, roughly 4/10 06/07 vintage loans must go belly up.

so how likely is that? well, if you obtain copies of MI annual statements, you can recreate loss triangles by accident year. they will show you that, over time, the cumulative default rate on a flow book is around 2%. it also demonstrates that by year 3, somewhere between 30 and 50% of lifetime defaults have been recognized.

you can use the loss triangles and historical seasoning curves to develop a rough model of what lifetime defaults will be on the 06/07 books, given the experience to date. they suggest cum defaults in the range of 10-15%. now, 06/07 only accounts for 40% of the flow book, so the total contribution to loss frequencies will probably by somewhere in the 4-6% range.

this analysis suggests that, far from being insolvent, there is actually tremendous value here that has been unrecognized by the market. the stocks are 10-20 baggers over a 5 year time frame.

that being said, i think all parts of the capital structure of these companies are interesting. for those who are intrigued by the analysis, but scared to own the equities, i'd point to RDN and PMI holding company debt. you are getting 13-25% YTM (face value 55-80% of par) for a senior interest in the holding company that owns the regulated MI subsidiary.

August 29, 2008 10:54 AM

My objections to this are that I am not sure that the bulk book is fully written off, nor am I sure that the book in 2007 will look anything like any previous year - as the last refinance of my rolling loans was onto Magic's book.

My further objection is that statutory capital deficiency happens way earlier than this - and stat capital deficiency will close the business as per TGIC.

But I remain to be convinced otherwise - and I have had a go at reconstructing Magic's loss triangles (with help from a reader) and we got not very far (probably because we gave up early).

If Bond Investor would like to share with me I would much appreciate it.

Thanks.

John

Ripping off two charts – Moodys mid year residential mortgage backed security performance update

Moodys has published a short report entitled U.S. RMBS Mid-Year Performance Update: H1 2008. This post will shamelessly rip off a two charts from it. The entire report is absolutely consistent with the data summary that I have been giving on this blog. Its really annoying to do all this work eyeballing hundreds of MBS pools and then find a really neat 8 page paper that does it all for you…

I would quote the report more broadly except that I do believe that they have some copyright issues. You might find someone to kindly send you it. Or you can buy it on the web here at an outrageous price. [If the price were one tenth the price listed here I would recommend it…]

But given that I do not own the research I will limit myself to a review.

The review goes through different types of collateral and different loss curves. I will not go through all of them for copyright reasons.

The first picture I will extract is cumulative losses on Subprime RMBS by year of origination. This is one of those “no hope” charts because losses are trending upwards fast with no obvious signs of stopping.

The picture makes it look like 2005 looked nasty and stabilised (something I have been saying for a while) but that the 2006 and 2007 pools have no hope. The 2007 pool is worse than the 2006 pool. [This should sound familiar to regular readers of this blog.]

The second chart I think makes the story a little clearer:

The 2005 pool had a delinquency hiccup – but is now back in the fold. I think we can safely estimate losses on 2005 subprime pools now and they are not that bad. The 2006 pools looked like they were stabilizing – and I blogged to that effect. The data in the last month however has given me some pause – hence my depressing thoughts post. Any light at the end of the tunnel in the 2007 pool looks like an oncoming train.

The same trends exist for Alt-A generally – but I am not going to extract all the data from the Moody’s report as I respect their copyright. I should say with Alt A the 2007 pools are not modestly worse then 2006 pools they are immodestly worse.

This is consistent with my rolling loans thesis. The loans could never be repaid but they could be refinanced. They kept getting refinanced to later pools and the last pool contained the detritus.

The 2006 Alt A delinquencies also looked to be stabilising but with a hiccup in the last month. The hiccup in the last month was worse than the subprime pools. 2005 is right back in the fold as with subprime.

Moodys then break the Alt-A into conventional adjustable rate mortgages (ARMs) and Option Arms. It seems there is almost no last-month kick up in delinquency for conventional arms but a big kick-up for option ARMS. This suggests the problem is Option ARM resets and confirms the analysis of many of the people who have commented on my blog. [Dear readers – a lot of you are very smart - indeed the great joy of this blog is what it delivers to my inbox.]

They then do cumulative losses on jumbos. Similar trends – but there is a kick down in last month delinquency. Does anyone know why jumbo credit looks to have got better very quickly. I have no idea.

They then do subprime Closed End Seconds and HELOCs (something I have studied at great detail). There is MUCH more stabilisation in CES – even the 2007 pools appear to be stabilising, however HELOCs, which are generally more prime than CES are not showing the same stabilisation.

Also with PRIME CES the 2005 pools look awful, with subprime CES the 2005 pools look good. In general 2005 mortgages are only bad if they are prime. [That should confuse most of you - but I think its because in bad poosl the really bad loans rolled out into 2006 and 2007 pools but prime people did not roll their loans.]

I would go further than the Moodys report and note that more generally the better the credit the worse the trend, or the worse the credit the better the trend. But nonetheless the Moody’s report is accurate, neat, short and should be got by anyone with decent access. The expensive source I link to is probably too expensive – but that is your choice.

If you are really maniac you can take this data and recalibrate Bill Ackman's open-source model of the bond insurers. I have done that - but my estimates are probably not much better than anyone else who does it with some integrity.

J

Reducing post frequency

I started doing this blog with the idea of two posts per week, but it somehow morphed into one per day. I had a lot to say and somehow had found an outlet.

I was however chewing considerable human capital - having developed ideas over a long period of time. [I do not generate ideas as fast as I have put out posts.]

So it is time to face reality and drop the post frequency to one per week - somewhere near the level that I consider sustainable.

If given a choice between quantity and quality I will chose quality. I hope my readers are not too disappointed.

J

Counting and double counting – a comment on losses and accounting standards esp FAS 159

Warning: Wonky accounting alert. Don’t bother reading this if you are not interested in wonky accounting issues

Last post I asked if anyone had a decent list of the losses so far realised. The problem was double counting – which is pervasive if you do this sort of thing. Aaron of the HF implode meter (who is usually acute about these things) made a comment which misses the point entirely – but it required a little thinking – and I admit I missed it at first. Here is his comment –

Then again, some gains aren't really gains, like the gains counted on the financial co's own collapsing bonds. Sure, they could buy those back, if of course they weren't already in hock for cash...

I will get back to Aaron’s comment later…

The gains from debt forgiveness in tax

In the US if you have a debt forgiven that is income for ordinary tax purposes. It sounds quirky – but its not.

If a bank lends company A $100 and they haven’t paid it back then the bank has clearly lost $100. Company A has lost $100 too – but that loss ultimately did not come out of Company A’s pocket – it came out of the bank’s pocket.

Unless you are very careful though the tax law is going to allow both company A and the bank to claim $100 in tax losses. That is a good way for the tax system to collapse because Company A could lend to Company B who could lend to Company C and so on and a single $100 loss could theoretically produce thousands of dollars in tax losses. Those losses could wipe out the tax base of any respectable country.

Fortunately (or unfortunately depending on your point of view) the people who draft tax law in most countries are not that stupid. [I started my career doing this stuff at the Australian Treasury – so it was once an area of expertise for me. We had terrible trouble in those days with cascading tax losses in trusts…]

What most countries do is that they have provisions against “loss cascading” or the like. They fix this up ultimately by taxing gains from debt forgiveness.

When it is finally clear that Company A in my above example will not have to pay back the said $100 it gets assessed on a “gain from debt forgiveness”. The idea is that if you borrow $100 and don’t have to pay it back that is a gain of $100. It offsets the loss of $100 above.

Gains from debt forgiveness in GAAP

The accounting rules do the same thing. When Conseco (a company I knew extremely well) completed its bankruptcy it compromised a few billion in debt (mostly preferred shares). The $2 billion or so it did not have to repay was a gain for accounting purposes – and the “New Conseco” published accounts reflecting that gain.

There were similar adjustments for the removal of Conseco Finance from the balance sheet.

Done properly these gains will remove double counting of losses from the accounting system. If you add up the losses made by Conseco and the losses made by the preferred stockholders without assessing the gain from debt forgiveness you get double counting.

Accountants rightly get concerned about double counting. But the accounting answers questions in a way that might be misleading to investors.

Typical question:

How much did company X lose doing that stupid lending:

Two possible answers:

(a). $2 billion dollars

(b). $500 million dollars [but we have ignored the $1.5 billion more borne by debt holders but ultimately forgiven].

The accounting answer is technically correct but doesn’t really get to the guts of why the lending was so stupid.

Now accountants are often pedants and investors should be practical people. And I have never seen a place of more vociferous disagreement than the disagreement over FAS 159.

FAS 159 says (roughly and in this context) that if a company is using mark-to-market accounting on its assets and its liabilities it should also use mark-to-market accounting on liabilities whose market value is affected by their own credit worthiness.

To take a real example, Ambac is a bond insurer which is showing in its accounts about 7 billion in derivative losses. During the last quarter they booked about a billion in gains because the market assessment of their credit worthiness went down and so the market value of Ambac’s derivative liabilities went down. Had the accounts been measured a month later Ambac would have booked over a billion in losses on the same transactions. The result was that during the last quarter Ambac’s profit was approximately equal to its market cap. It will reverse that profit this quarter.

That profit of course was meaningless for an investor assessing Ambac stock. [I own Ambac stock - purchased at $1.30.] I don’t think the FAS159 movements make any difference in assessing Ambac. I ignore them. I think Ambac’s results were not great but they showed stabilisation which was enough given the stock price.

David Einhorn – a man considerably smarter than me - put it this way in a speech (“FAS 159: Profiting From Your Own Demise”)

If your own credit spread widening counts as revenue...

... and you pay compensation as a percent of revenue ...

... the most profitable and lucrative day in the history of your firm will be...

THE DAY YOU GO BANKRUPT!

Ok – so let’s ignore FAS 159 assessing Ambac, investment banks and the like. And the results are much worse than the headlines. The underlying of the investment banking industry sucks.

When you shouldn’t ignore FAS159

In once sense what FAS159 does is it brings forward the gains from debt forgiveness. If the price of repurchasing Lehman’s debt falls then Lehman has made a gain which it is hard for shareholders to profit from. If Lehman goes bust and has its debts forgiven Lehman has also made a gain – but it’s the same gain recognised on a mark-to-market basis as its credit worthiness collapses. [Its also hard for the shareholders to benefit from this gain.]

This is of course a problem with mark-to-market accounting generally. Mark to market accounting brings forward profits. It also brings forward losses. It’s a darn stupid way to run an investment bank when you pay cash on profits that are not realised but just bought forward. It’s a darn stupid way to invest when the profits being bought forward are the profits from total collapse and debt forgiveness.

But if you are adding up losses then the gains from debt forgiveness are a necessary part of ensuring that you don’t double count.

Likewise – in a mark-to-market world if you are adding up losses then the gains on FAS159 are a good thing to include.

Now let me bring back Aaron’s comment:

Then again, some gains aren't really gains, like the gains counted on the financial co's own collapsing bonds. Sure, they could buy those back, if of course they weren't already in hock for cash...

Aaron bought this up in the context of double counting. But this is the one context when those gains really are gains. Aaron is thinking like an investor – and I love him for it. But just for once I wanted someone thinking like an accountant.

J

Does anyone have a good estimate of losses recognised to date?

In this post I talked about the Bloomberg list of losses recognised to date. I even gave a spreadsheet here.

The problem with the Bloomberg list is that it doesn't include lots of entities that have booked losses. It excludes for instance insurance companies (AIG does not appear and they are almost exhibit A in big companies gone mad). It also excludes GSEs, mortgage insurers, bond insurers, hedge funds that have lost money and the like. It does not include Residential Finanial Corp (GMAC), or any of the imploded subprime issuers.

Even then recognised losses come in above 500 billion.

Does anyone have a complete list?

Even better - does anyone have a complete list with spreadsheets speculating on obvious problems of double-counting. For instance if someone is mark-to-market on a bond insured by MBIA then they are probably showing a loss. MBIA is showing losses. There is of necessity some double-counting here. [Such an adjustment will be rough...]

Hoping for help.

---

The reason is that I want to work out how far down this credit crisis really is. The Bloomberg list is misleading for that task.


Thanks.


John Hempton

This week's bank collapse

FDIC Friday - and another bank was taken over by the FDIC. The number so far is relatively small - because the bad mortgages are not in the banks. (The bad mortgages were sold to the securitisation market as this blog has made clear several times.)

Anyway this bank collapsed due to commercial real estate. That matters as the CR loans are in regional banks.

Calculated risk has a nice post observing that this collapse - whilst small - is probably going to be representative of this cycle. I have short positions in a few CRE exposed banks.


J

Berkshire risk aversion - a follow up

Thanks for all the people who followed up on this. I think we can safely conclude that the factual background in the original post was correct. Mostly GEICO sells household insurance as an AGENT for Travelers. It may or may not take on umbrella policies through other Berkshire subsidiaries.

I really appreciate the confirm from my readers. Disagreements about facts should be easily sorted [provided of course I remain adequately circumspect about what constitutes knowledge]. I expect disagreements about interpretation to be much more difficult.

This is a follow up from my Berkshire risk aversion post.

I have received an email from a finance guy who has a homeowners insurance policy attached to his GEICO auto policy. In some states (notably Florida but also possibly other states) this policy may actually be written by Berkshire - but mostly it is a policy whereby GEICO acts as an agent.

Can people with umbrella and homeowners insurance policies from GEICO tell me who the underwriters are for the homeowners component - and what State they are in. Also useful would be the number of years that they were policy holders.

Thanks.


J

Some depressing thoughts

The position of this blog has been stated several times. Subprime credit is a pig-in-the-python. It got bad very quickly – and it will get better very quickly. The losses on some pools (particularly loans originated in early 2007) will look implausibly large – but 2005 pools will be much tamer.

Prime credit by contrast is deteriorating at an increasing rate and there is no obvious end in sight. It is not able to be modelled. It is very scary.

The thinking behind that is detailed here and here.

Wish it were so simple. The prime credit is looking every bit as bad as all that. But the trend in subprime credit is also less clear than a month ago. In some pools (not all) the trends are less good. Short dated delinquency is rising in some places. Losses look a little larger than they might have looked a month ago. The improvement trend was three months old. And one month does not a cycle make – but my level of comfort is falling.

Risk aversion – Berkshire style

Chatting with a well known blogger yesterday who made the (common) assertion that all insurance companies eventually go to zero. The world – it is argued – is full of surprises – and if you take even a small chance of blowing up then eventually your number will be up and you will blow up.

Then our protagonist argues all insurance companies take a small chance of blowing up – and eventually there number is up.

This I think is too simplistic. I think some types of insurance have realistic (albeit small chances) of blowing up. Others (rarer) do not.

Take for example household versus auto insurance.

Household insurance is subject to super-catastrophes. Big earthquakes are theoretically possible anywhere (although vanishingly unlikely in some locations). Hurricanes happen as well and the 250 year storm seems to happen fairly regularly.

But auto insurance is not generally subject to super-catastrophes – cars drive away from hurricanes and bounce around in earthquakes. The only catastrophes that regularly affect cars are hailstorms – and the biggest hailstorm claim in insurance history happened in my home town (check this out about 3 inch hailstones).

Has anyone noticed that Berkshire does auto insurance (GEICO) but does next to no household insurance (it writes a few policies in Florida which it tries regularly to cancel – but regulatory rules force it to renew).

If you write household insurance there is the mega-storm that could bankrupt you. Buffett has said that storms substantially bigger than Katrina are possible. Buffett won’t do it even though GEICO could probably – if it chose to be – become the most profitable household insurer in the US.

Also take life insurance

Any student of human history knows that plagues – real devastating plagues which wipe out very large numbers of young healthy people happen irregularly but often enough to be a real risk.

The insurance industry have gotten quite complacent about them – because the only new plague of our lifetime did not affect life insurance companies. HIV is – within the West anyway – primarily a disease of gay men, IV drug users and haemophiliacs. None of these people were regular buyers of life insurance. Had HIV been a disease of middle age men who visit prostitutes because they don’t get it at home it would have selectively targeted the core market for life insurance companies – and every life insurance company in the Western World would have gone bust. In Thailand HIV morphed into such a disease – but surprisingly or not life companies have not been heavily impacted in the West.

Note that if you think plagues are inevitably going to happen then most life insurance companies have a finite life.

Notice that Berkshire does not insure mortality at a primary level and does almost no mortality reinsurance. Swiss Re is the dominant player in that business. Buffett has the capital and the expertise and he does not play.

By contrast annuities are not subject to plagues or shock risk – they can cost you money as life extends due to medicine – but that looks like a continuous process – not a shock process like plagues. Berkshire will write you an annuity online (see http://www.brkdirect.com/)

Again a small risk of blow-up and Buffett will not play.

The Gen Re debacle

The terrorist attacks of 2001 were a true shock to Buffett. General Re was taking a chance of complete wipe-out and had the attack been nuclear there is a fair chance that Gen Re would have folded. Buffett wrote about that in a special letter to shareholders – see this quarterly report which differs from other Berkshire quarterlies as it contains a shareholder letter...

Buffett criticises Gen Re for writing risks that could have been lethal. His company broke the first Buffett rule of risk aversion. [Despite the joke that the rule is don’t lose money – the rule really is don’t blow up. Buffett is quite prepared to lose 5 billion on a single policy.]

Note that almost none of this involves mathematical models. It involves stress tests. And the stress is not a selected 99% probability stress test – it’s a complete implausible debacle stress test.

And when most people do stress tests the process is “well what happens if property prices drop 30%?” and someone says “what if they drop 40%?” and the response is – they can’t go that far.

The Buffett question for a risk manager is “what if they drop 70%?” You might say it can’t happen. One word: Japan.

I know how you live by the Buffett stress test with a diversified portfolio – when I open a fund it will live by that test. (No intolerable concentrations of long only banks or life insurance companies.) But I have no idea how you run a life company by the Buffett stress test. I don’t think you can. That doesn’t mean the businesses shouldn’t exist – but they probably never be considered completely safe. And a portfolio investor should never be totally weighted there even if that is their field of expertise.

John

Market puzzles - Fannie Mae and Freddie Mac

Crank your mind back to 2006. You are a run-of-the-mill hyper bear (as I appeared at the time). Someone told you that in 2008 there would be a crisis at Fannie Mae and Freddie Mac. The Secretary Treasury would ask for and get an essentially unlimited pool of government money to ensure that the senior debt of those institutions did not fail. [He said nothing about the preference shares or common.]

What would you have thought would happen?

To me it seemed obvious. The senior debt would now be as safe as Treasuries and the spread on the senior debt would come in to maybe 30bps from crisis levels.

But the whole scenario would be very bad for the US dollar because the implied debt of the Federal Government just rose an awful lot.

So what did happen?

Well the spread on Fannie and Freddie securities widened not narrowed, and the US dollar got strong.

The widening of the spreads on Fannie is either irrational or a rational bet that the US Government cannot be trusted to honour its promises. But if it is the latter why is the USD strong?

What is happening now is every bit as weird as what happened in the bubble.