In which FT Alphaville asks, should you really take CDS default probabilities at face value?
It will come as no surprise to readers that the clairvoyance of CDS is open to question. CDS spreads are more likely to be steered by frightened bond portfolio managers or CVA desks dealing with volatility than by the invisible hand.
Given the fickle animal spirits of this particular derivative market, you would be right to be sceptical of the default probabilities implied from CDS. These are constructed from models (hence “implied”) with various simplifying assumptions, which we detail below.
CDS and fire insurance
Stepping back from CDS for just a moment, let’s consider fire insurance to illustrate some of what’s going on here. Insurance companies use datasets comprised of many customers’ characteristics to calculate the odds they’ll have to make a single payout should a given customer’s house burn down.
Insurance companies are also businesses, so they’ll add whatever margin they can get away with to the fair value of the contract. The “fair value” cost to a consumer, in a world where insurance companies were not-for-profit organisations, with no costs themselves, would be:
[The value of your home] * [Probability of it burning down]
Now imagine a world occupied by worrywarts. Seriously anxious types. Chances are the insurance companies could get away with charging even higher premiums. (And pssst! Chances are that this isn’t too different from the world we live in.) The insurance companies will consider this premium to be their margin over fixed costs.
If you observed prices in this worrywart market and tried to back out the probability of your house burning down by using a formula that failed to account for the worrywart premium/margin, you’d likely never sleep at home again for fear of impending fire.
In other words, the price of the insurance doesn’t just reflect the real world risk, it’s also reflective of how worried you are, hence how much an insurance company can charge you. It’s the price of certainty. The certainty is that you’ll either have a home or enough money from your insurance company to rebuild it. That people prefer certain outcomes and will pay premiums to get them is a well known result of psychological experiments.
The default of a sovereign nation
What holds for fire insurance holds for default insurance (to the extent that CDS can be called that). That is to say that CDS also exist in the worrywart world – one occupied by bond portfolio managers and CVA desks willing to pay a premium for certainty.
This is why CDS spreads are not a reliable source of real world default probabilities. The size of the premium matters and liquidity of a given contract will also influence accuracy, as some CDS contracts are far more actively traded than others. A contract that only trades a few times a week is likely to contain little information compared to one that trades upwards of twenty times a day.
The simple formula that is used to back out probabilities of a house/country burning down/defaulting doesn’t back out the margin/premium and even has additional assumptions baked in, like recovery rates.
On to some numbers then. Let’s look at a table from CMA Datavision’s Q3 Sovereign Debt Credit Risk Report.
(We suspect that the “Q4″ is a typo – no financial instrument is that good at predicting!)
The Kingdom of No Debt has CDS
Here we can see that Saudi Arabia has a 8.4 per cent chance of defaulting over the next five years. Funny thing about CDS… they trigger upon a default event and that necessitates there actually being something for there to be a default on. Saudi Arabia. Has. No. Debt.
Yes, CDS markets are crazy enough to allow contracts written on Saudi Arabia anyway. And even on Norway. As with any market, it is driven by supply and demand, and when supply and demand meet, trades are done. Simple as that.
Maybe a fund manager put Saudi Arabia CDS on as a hedge against the prices of Middle East assets he or she holds (and hence also implicitly bet on the correlation between the spread and the state of the Middle East more generally – assume correlation is constant at your peril).
And other small CDS that make big predictions
In any case, here’s another table to ponder. This time, we take some of the above data, then add net notional amounts on CDS outstanding as recorded in DTCC’s Trade Information Warehouse, which almost covers the entire market.
Switzerland, Pakistan, and Hong Kong don’t even list among the Top 1000 single-name CDS, hence they’ve been left blank. Given this, how relevant is it that Switzerland has a 6.2 per cent probability of default over the next five years or that Pakistan has a 51.8 per cent probability of default over the same horizon?
It’s not that CDS spreads have no information at all. They clearly do. If participants in this market go from being willing to pay 123bps at the end of Q3 to 300bps at the end of Q4, on whatever credit, then that is informative.
The bizarre part comes when you start unreservedly backing out a probability of default. We guess you could look at how the probability evolves over time, but why not just check at how liquid the credit is and then track its spread, if it is indeed liquid enough?
It could be so much better than this
Alternatively, you could get rather fancy about dividing the probability of default implied from CDS into more meaningful component parts, as some clever chaps at the IMF attempted to do. They took CDS spreads and divvied up the probability of default that:
(a) is attributed to fundamentals;
(b) is attributed to risk aversion, i.e. worrywartness;
(c) comes from contagion, i.e. the risk that the sovereign will default if some other sovereign defaults first.
Interesting. If anyone wants to design a service that takes CDS spreads and backs these elements out, please do that and put FT Alphaville on your distribution list immediately.
Given all of the above, one last question: why do probabilities of default implied from CDS get bandied about so often?
Might the answer lie in how easy it is to write headlines out of them? One to contemplate anyhow…
Related link:
Put (CDS) trading activity where your mouth is – FT Alphaville