The CBOK says that OAS does not include default risk - a concept with which I am struggling.
Apparently, I am not the only one as I found this post from 2011 - http://www.analystforum.com/forums/cfa-forums/cfa-level-iii-forum/91255488?page=1
I read through that thread but still don’t quite understand.
Let’s start with this question - is the default risk included in the Z-spread?
OAS does not account for default risk, so it has little or no use with respect to analyzing risky bonds generally, and speculative grade bonds in particular., the z-spread removes the option risk.OAS is measure of embedded option which is option spread above benchmark. it only reflects (liquidity, maturity) but not default. but bonds do have default risk and its not reflected in OAS.
OAS modeling ASSUMES that the bond in question is NOT going to be defaulted (at least not for the assumed interest rate volatility). Remember the binary tree used in level II: none of the nodes has default as value. Therefore OAS is good for comparison of (similar) investment grade bonds since those bonds are unlikely to default in standard volatility scenarios.
Agency MBS has implicit guarantee of the US gov thus assumed to be default-free (or at least investment grade) thus appropriate for OAS modeling.
Junk bond has high default risk thus not appropriate for OAS modeling.
OAS is good for comparison between option-free and option-embedded bonds with similar investment graded credit rankings (and between option-embedded bonds even though you’d get same conclusion using z spread) since the OAS ‘removes’ the extra spread one pays for the option (option price) from the z-spread thus make it comparable between bonds with options and bonds without.
The confusion is due to the definition of ‘exclude’
OAS excludes option risk: OAS removes the option price. All other risks remain (model risk, interest rate risk,…), even though it is not appropriate to address default risk (see below).
OAS excludes default risk: It actually means the OAS methodology excludes the possibility of default. OAS modeling assumes no default thus does not take default into account. OAS is not appropriate to be used for high risk option-embedded bonds. The OAS number DOES include the default risk, but it is meaningless in the same way as you use DCF model for a start up firm using risk free rate as discount rate (or one may argue that it is inappropriate to use a DCF model for a start up at all).
With respect, I believe that you have this all wrong. OAS does not assume that bond is not going to default.
If any of the nodes had a default value, then the OAS would exclude default, just as nodes exhibiting exercise of the option result in excluding the option value.
If there is a default risk it is included in the market price of the bond. Nothing in the OAS methodology removes that default risk.
Agreed here magic.
My thought process went like this …
In a different part of the CBOK, swaptions are used to add or remove the call option on a bond. (A receiver swaption IS a call option on a bond.) Therefore, IF the option on a bond is a call, for example, and the market provides a price for that call in the swaption market (the premium) then one could derive the OAS simply by manipulating the swaption premium into a spread value that can be subtracted from the z-spread.
Anyway, I suppose for the exam we consider the OAS to exclude default risk?
OAS option-valuation models do not incorporate default risk; therefore, OAS valuation is not appropriate to use for the higher-risk markets.
I don’t get this sentence. Do you say, that if a node had a default value, then the calculated OAS would exclude default risk? Turn the reasoning around: as there is no default assumption in any of the nodes, the calculated OAS includes default risk?
Correct.
OK. Then, I have to admit that I do not understand this line of reasoning. 
OAS is the worst, I am prepared to pray and guess and allocate that time elsewhere… can’t win them all right?
