There have been a few broker notes out suggesting that the GSE preferred stock is going to zero. The preferred stock itself has been dreadful lately – retreating almost to our original purchase price.
I think the broker notes are wrong – but lets do this formally because if you look at the assumptions in my model and the assumptions in the broker notes you can make up your own mind. [I will lay out their assumptions and my assumptions clearly – you decide.]
The first “GSEs are zero” broker note was produced by Keefe Bruyette & Woods (one of the few brokers left covering the stocks). I have reproduced the note here (and claim fair comment use for doing so).
The core assumption is that the GSEs are closed – and that they are put into very rapid run off – and that they do not earn much money during this run off period. Here is the revenue model for Freddie Mac.
(You will need to click all the tables in this note for details.)
There are implicitly a lot of assumptions here.
The first core assumption is that the net interest yield (after hedging costs) on the retained portfolio will be about 1 percent over the long run. I agree. In the bad-old-days Fannie Mae used to report about 120bps, Freddie Mac used to report about 80bps. When they restated their results Fannie restated the results down and Freddie restated them up. The right number was about half way between the Freddie and Fannie numbers – so 100bps is as good an estimate as any.
The second core assumption is that the short run hedged interest margin is also 1%. This is flat wrong. Fannie and Freddie are getting absolutely record interest spreads at the moment – absolutely shooting the lights out. I detailed this here. This model assumes that Freddie has net interest income of $8 billion this year – which is rather difficult because they are currently getting over $4 billion per quarter. The high current net interest margin is a function of three things:
- Firstly – and most obviously – the lack of competition in the mortgage market. That is not going away in the short term – and it would be crazy to assume that net interest margins compress to 1 percent rapidly.
- Secondly the Fed is being more than generous with the shape of the yield curve. That is going to end – but possibly not that rapidly.
- Thirdly – and this is important – there were several charge offs of derivatives which were used to hedge the net interest margin when the businesses went into conservatorship. Those hedges are still there (but they have been written off up front rather than amortised over the life of the product). As a result reported net interest margins will be higher in the short term.
All up I would expect the net interest margin over the first two years to be maybe 12 billion dollars cumulative higher than this model. Indeed as those numbers are currently being reported it is perverse to argue otherwise.
The third core assumption is that the company is put into massive and sudden runoff. This is a political decision that – as far as I can tell – has not been made. You can see this in the numbers because the owned portfolio (and for that matter the guaranteed portfolio) is assumed to drop 20 percent per year from now. This is a far more aggressive assumption than the government is currently indicating for Fannie or Freddie. Indeed last month Freddie’s owned portfolio actually rose a little. Moreover no government official has so far indicated that Fannie or Freddie will have to get out of the guarantee business – and this model assumes that they must leave the guarantee business.
In my long series I made it clear that the value in the preference shares depended critically on the companies being allowed to stay in business – at least for a few years. That is true of the value of almost every bank in America – in that the whole sector is dependent on the pre-tax, pre-provision profits from their current business to cover their credit losses. If it were not for pre-tax, pre-provision profits even big (and sacred) companies like GE would not really be viable.
If we just assume the portfolio remains flat for three years we can add another 10 billion to Freddie’s pre-tax, pre-provision profits.
Now it is possible that the Government might choose to put the GSEs into rapid run-off (and there are Wall Street firms who crave the interest rate hedging business and who would like it) but if the GSEs are put into rapid run off it would have profound (and negative) effects for the price of conventional mortgages and for any housing recovery. I think it is reasonable to assume that they are not insane – so I think three extra years income is a reasonable assumption. However again I am just exposing the assumptions.
The fourth issue is simple double counting.
Freddie in the KBW model is assumed to write no new business. If it writes no new business it can incur no credit losses on that business.
However KBW assumes 5bps of credit losses and 5bps of credit-associated costs.
They are going to have credit losses on the old business (but they count them below). Counting additional credit losses is double counting.
It would (of course) be reasonable to assume credit losses would be incurred on new business – but KBW is asserting that there will be no new business.
The extra credit costs in the KBW model add up to another $6 billion over ten years.
I think on reasonable assumptions – including a rapid run off of the book after three years the pre-tax earnings of Freddie are thus 28 billion higher than in the KBW note. Nonetheless I am just reporting the assumptions implicit in the argument that Fannie and Freddie are permanently impaired.
Credit losses in the KBW note
There is no real model of credit losses for the existing book in the KBW note. However they do give a chart with base case, stress case and best case.
Note the cumulative loss in the best case is $33.7 billion at Freddie Mac. My model was a little worse than the best case - $37.6 billion of losses still to incur (so over 40 billion cumulative losses on the book). Since I wrote that, there has been a solid bounce in the demand for houses around the $200-300 thousand dollar mark (that is largely GSE foreclosures) in the key bubble states. This video from Jim the Realtor (who is usually a dour bear) explains just how strongly the San Diego area has bounced.
It is pretty clear from stories like this (and there are many more) that it is much easier to clear inventory. My model assumed that it was going to get much harder and that severity on the book would rise from the current 43 percent to about 50 percent.
Now this bottom-end housing bounce could be “the mother of all head fakes” – but again – like the interest margin – I am only reporting what is happening now. The housing market could take another big swan dive and then my model will be wrong. That is the bet I spelt out in the long series.
Anyway we should look at the current losses – and projected forward losses. The last Freddie Mac quarterly credit supplement gave the credit losses, provisions and reserves by quarter.
In the second quarter the cash losses on the Freddie guaranteed mortgage book were $1.907 billion. Cumulative cash losses have been a bit over 5.8 billion dollars.
My model assumes that the future losses will be roughly 6.4 times losses booked to date. That is my estimate is consistent with the housing market getting dramatically worse. The evidence for that is thin.
Not only is the anecdotal evidence (such as Jim the Realtor) pointing the other way, but foreclosures are up only 5 percent from summer to fall. Housing bears treated that news as evidence of crisis – and I thought it was a remarkably good number. The foreclosure moratoriums have expired and we are not swamped by foreclosures. My estimate that cash losses on the existing book are likely to be about 6.4 times the so-far-recorded losses does not seem low.
That said – the base case in the KBW note have cash losses being 10.1 times already booked losses. That seems unreasonably high with the strong evidence of a turn in the housing market. The stress case (which are the Congressional Budget Office numbers) are for end losses to be 24 times the amount of losses already recorded. If you believe that then depressive illness is probably the best diagnosis. Surely you should not be doing stock analysis – and it puzzles me why the CBO should be putting out such patently ridiculous estimates.
That said – KBW uses their base case in their model (59 billion of cumulative losses) and I use my base case (37 billion). There is a further 22 billion difference between my assumptions and theirs.
I note that they do not justify their 59 billion number – and I went to great lengths to justify mine. However if the housing market takes another massive turn downwards theirs (not mine) will be right. If the housing market continues to bounce (as it has) then we will both be wrong – losses will be lower than either of our estimates.
The non-mention of write-backs on the private label securities
The KBW note does not make any mention of the possibility of write-backs on the private label securities. I went to some lengths to show why – at least in Freddie Mac’s case – those write-backs were likely. I produced an estimate of about $10 billion. These write backs are reflected in part in current market prices for these securities.
This adds another 10 billion difference between my model and the model used by KBW. The total difference is thus $60 billion.
You can do a little bit better than that too – because Freddie will earn some return on the 60 billion it does not lose – but lets ignore that.
KBW’s solvency model
KBW then presents a solvency model for Fannie and Freddie. I reproduce it here:
There are some nuanced differences between my model and their. My model of losses is a model of losses not yet recognised whereas they provide an estimate of end-cumulative losses. That differs by the losses recognised to date ($5.8 billion though KBW state incorrectly that they are 8 billion). These add to the difference between the KBW model and my model.
Also Freddie Mac currently has a 7 billion dollar positive capital position (remember it made a profit last quarter and it had write backs of the private label securities). KBW has ignored that (something I consider another pure date-input error). So you could add another 15 billion benefit to Freddie on my assumptions over KBW.
Nonetheless my model of Frannie is – on fairly easily justifiable assumptions – 60 billion better than the KBW model.
Add the 60 billion to the net capital position as estimated by KBW (the –39 billion at the bottom of the table) and it is pretty clear that Freddie can repay the shortfall and make the preference shares whole. Add in the remaining 15 billion (being the chargeoffs to date and the current net worth of Freddie after profits last quarter) and it repays easily.
A plan for Obama
Reform of the GSEs is quite tricky at the moment. The jury is still out on their end losses. Moreover the ether is full of self-interested lobbyists who want to take the good bit of their business (mostly interest rate risk management) and leave the bad bit (credit risk management) with the government.
Winding down the GSEs right now runs the risk of killing the nascent recovery in the housing market.
The sensible course of action is to just wait. This is policy that can be delayed without any real additional risk to the government. (The government is already on the hook for the losses.)
If my math is right – and I think it is – then the GSEs will appear solvent in time for the 2012 election. The government can demand (and receive) almost 100 billion in capital to be repaid from them (which will make the budget look good and undermine the only viable Republican argument that the Democrats are irresponsible). It will make the government look like good conservators of key institutions. It will make Obama look like safe hands for running America.
The anti-GSE lobby knows this is a possibility and they are determined to capture as much GSE business as possible right now – so they are vociferous in their claims. Sensible people should ignore them.
John