Recommendations: 1
Betting on Pru's Bonds
I live on the West coast, but I function in Eastcoast time to stay synchronous with domestic markets. That means that 8am Eastern is 5am Pacific for me and when I should be rolling out of bed to get heat going in my home office and a computer turned on. But days when I don't have to shop the bond market, I like to lie abed, snuggled under the covers, just thinking. Sometimes, maybe most times, it's 34 hours later before the problem I've been mulling reaches enough of a resolution that I can switch over to working out the details with a keyboard.
Two trains of thought held my attention this morning, both having to do with the reply I made to fulgore about the investmentworthiness of Pru's bonds. In essence, my reply was only the fivecent version of the $10 report any investor would need. (Creating that report is their problem, not mine.) But I'd like to add another two cents. So let's have another go at it.
Apparently, Smart Money is touting Pru's bonds as a buying opportunity. (I haven't read the article. I don't intend to. This isn't to say the mag might not be a good place to get hot tips.)
Having been given such a tip, the first thing a wouldbe bond investor should do for herself is to construct a yieldcurve of Pru's debt from its current offering list. If Doc Tarr is right, and if my suspicions are right, that also should be the only thing a wouldbe bond investor should have to do. That alone should be a sufficient go/nogo gage. So let's review their yieldcurve. It's flat, as I remarked. But I was so focused on making a reply to fulgore, that I didn't see what was in front of me. A flat yieldcurve for their debt? My immediate reaction should have been <PASS>.
Let me explain why, and let me run through it quickly, simplifying greatly. Yieldcurves come in three shapes: rising, flat, and falling. Let's rename those three shapes: Rising, Level, and Falling, and refer to them subsequently as R, L, F. Modeling yieldcurves as a Markov process, any state has equal probability. If, presently, we are in state L, there is equal probability of R, L, or F occurring next. Persistence of a state doesn't concern us for now. We're only interested in change. Will R or F occur next? Markov analysis favors neither. But bond markets aren't governed (though they might be modeled) by Markov processes. Thus, a probability has to be assigned to the likelihood of each state. In plain English, is the yieldcurve going to turn down or not?
If present uncertainties merely persist or actually do resolve themselves in Pru's favor, the wouldbe investor in Pru's debt is home free with a fat return. But let's consider the downside.
For our wouldbe, lotteryticket buyer to make a profit, a bunch of things have to happen. She must total Pru's outstanding debt and divide it by her estimate of the company's breakup value (properly discounted by the costs of a Chapter 11 filing). That workout result has to be subtracted from her entry price (discounted by coupons received before Pru files). If she got in at 45, but her amortized costs are 40 by the time Pru files, and if a workout price will be 10, her realized loss will be 30. Her upside is 55. But let's make some quasirealistic assumptions. If Pru (the company) doesn't go under, the price of Pru's bonds will go up. In a relatively short time, they should trend toward par. If our investor bought their 6.4's of '37, and if Pru recovers completely, those bonds will most likely trade at premium. But let's say that Pru merely gets itself out of serious trouble, and the market prices their 6.4's at 95.
At that point, our investor can close out her position, or continue to hold until maturity. Let's say she sells. Her net might be 90. A return of 90 on an investment of 45 doubles her money. A loss of (45) with a recovery of 5 from coupons and 10 from a workout reduces to (30). So those two figures describe the high and low points of the game. Let's also say the game is a coin flip. Should she make the bet?
Obviously so provided [and Dr. Tarr was shrewd to pick up on this but, then, he's a shrewd guy] she has the capital to sustain the bet. Let's explain that remark. If I have $100 and you have $1, you can't flip me for quarters. I'll break you. But we could flip for pennies, and a likely case is neither of us will walk away much richer or poorer. (Though there will be instances, however rare, when I go bust, as modeling with a Monte Carlo engine will demonstrate.) So, our investor can't be betting the farm, or the rent money, or a substantial fraction of her working capital.
Now, let's change the odds. Instead of heads wins 90 and tails loses 30, the payout is 70, the loss is (30), but the odds are 1:2. Should she bet? Obviously so, and in as many of such games as she can find, so long as she sizes her bets so she can sustain any series of losses she might encounter. And that's exactly where those idiots at Long Term Cap Mgmt, and the banksters at Bear Sterns, and anyone who believes in the mathematical garbage of Modern Portfolio Theory always go wrong. [But see my endnote] They use historical data to create probability tables for things that aren't closed systems. But having identified inefficiencies to which they have assigned precise probabilities, they then also do the even more stupid thing of levering to the max. Inevitably, they blow themselves up which, as Talib explains, is a very precise term. To “blow up you account” doesn't mean to lose a lot of money. It means to lose so much money, so quickly, that you get yourself thrown out of the game.
So, gaming Pru's debt is a very simple matter. Plot a yieldcurve. Of you've done this lots of times for lots of bonds, one glance tells you whether you want to go further. Pru's debt is currently rated A/A3, but it isn't. Simplistic scans, such as the financial mags run, will always float these kinds of “specialsituations investments” to the top of their sort. But rather than reporting the results of the scan in a page or two, they ought to devote the whole issue to how to do specialsituations investing generally, and let readers find their own hot tips.
What would be the heart of the tutorial they should offer? Ah, grasshopper, I'm glad you asked.
IMHO ('natch”) a wouldbe buyer of Pru's lottery tickets needs to do no more than the following:
#1, construct a yieldcurve. #2, pull their SEC filings and use Marty Whitman's methods to do an analysis of the company. (This is nothing more than classic, deepvalue investing. Also, it becomes a method of assigning guesstimates about probabilities of states.) #3, do whatever quantitative analysis you find useful, using nothing more elaborate than a spreadsheet, data about their bonds (chiefly the current offering list, plus T&S), and the ideas in a book comparable to Barnhill's on highyields. (Again, just more attempts to guesstimate probabilities.) #4, size your position properly if you do decide to execute.
That's it. Draw a picture, answer some questions for yourself about its details, and then stand aside or go forward.
Charlie
 Endnote: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1142785
You'll be able to access Nassim Talib's working paper (at least I'm assuming it's a working paper, because the editing is so bad) on "Finiteness of Variance is Irrelevant in the Practice of Quantitative Finance
If your math skills aren't good and mine aren't this paper will bust your butt. It's tough, tough reading. But his central point, at least I'm assuming that's his central point, is although making Gaussian assumptions about market can get you into a lot of trouble, more trouble than you could ever have wanted to deal with, the market practitioner has to have methods of dealing with those unpredictable uncertainties.
My take? Though MPT is demonstrably flawed doesn't mean that a lowlevel investor has to try to work within what is known to be an immanentlybetter theory. Commonsense, kitchentable methods of dealing with risk are available, and they might even prove more practically useful than more powerful theories.
Tentatively, I've called such an approach Ptolemaic Finance. I need to work out the details, then I'll be describing it. Yieldcurves are central to it and simple probability theory (and, maybe Monte Carlo engines too, because I want to explore them and Talib values their utility over more conventional methods of dealing with quantification).



Announcements
