Bond Relative Value Analysis with Yield Curves and Duration
Summary
The document discusses ways to compare non-callable government bonds from the same issuer when coupons and maturities differ. It raises the problem that raw yield comparisons can mix relative value with coupon effects and curve exposure, and that duration matching does not remove every difference. The questioner also notes limitations of par asset-swap spreads and zero-volatility spreads for identifying directly tradable opportunities.
The response suggests fitting a Nelson–Siegel–Svensson curve to observed yields as an initial way to assess bonds against a fitted government curve, assuming no credit spread or other structural features. It cautions that duration itself reflects coupon characteristics and suggests comparing by maturity, then considering the relative duration risk when evaluating a signal. The exchange offers no worked example or empirical evidence, and it deliberately leaves funding, convexity, liquidity, and other sources of richness or cheapness outside the simplified analysis.
Key ideas
- A fitted Nelson–Siegel–Svensson curve can provide an initial benchmark for government bond relative value.
- Observed yield differences can reflect coupon effects and curve exposure as well as mispricing.
- Duration matching does not eliminate all differences between bonds that may be fairly valued.
- Maturity may be a cleaner comparison axis, with relative duration considered separately when judging yield opportunities.
- The discussion omits funding, convexity, liquidity, and other potential sources of bond richness or cheapness.
Tags
Full text
# Bond rv metrics/measures # Bond rv metrics/measures I'm looking at some bond yields, varying coupons and maturities, same issuer (g3 govies) and non-callable. I have compared these by maturity and duration to get a sense of some rv opportunities and vs swaps. Focusing mainly on single currency curves in isolation at the moment. My aim is to really be able to find or clean the data as much as possible so as to only be seeing interesting opportunities. I.e. stripping out all obvious idiosyncrasies in the data in a systematic way. - For the maturity comparisons I think I need to adjust the yields for the bonds for coupon effects. The remainder of the yield spread to par should be a reasonable indicator of 'rv', correct ? (I am deliberately ignoring funding/convexity and other sources of richeness/cheapness here to dumb things down). In the absence of being able to estimate a fair value curve yield yield spreads seem useful but for steep/flat curves they're polluted by implicit curve positions and as not durn matched sometimes make different coupons seem rich/cheap. - How can I see a clearer picture of yields for similar duration bonds ? I have tried many swap related spreads, but each has their own shortcoming. For eg, par par asw, typically make the high coupons look rich, also theyre not flat outright risk. Z spreads are useful and stable but are not tradable directly (assume no liquid enough strips). - Lastly, looking at yields by duration (modified) is a lot better than maturity, but, there is still a reasonable difference in yields for similar duration securities which are not mispriced, i.e. are fair value. Are there better axis to look at to compare yields ? If not, what kind of adjustments are made to factor the noise out. An example for any (preferrably all) would be very helpful (also happy to post some toy examples to illustrate my points). Pointers or resources which address these issues in detail also very welcome. Thank you. ## Answer by fixedincome94 (score 0) https://quant.stackexchange.com/a/81846 I can't contribute to the last few points of your question however, what I can say regarding your first point is the suggestion of using an NSS curve which you have likely explored. I say this as it solves for the coupon effect of observed YTMs - other curve fitting requires par yields whilst par yields are the output of the NSS curve itself once fitted to the observed. Notwithstanding any other structural features of the bonds, using a fitted NSS curve plotted through observed yields will give you a 'first' step regarding government RV (assuming the absence of a credit spread). Regarding your latter points, I would likely caution against using duration as the comparability measure on the X axis considering this contains an inherent 'coupon' effect - maturity here seems to be cleaner with the option of then adjusting that RV signal for the relative durations of the security i.e two 5y govvies trading cheap / cheap - one likely has a better yield breakeven than the other etc.. As for resources I'd recommend Fixed Income Relative Value Analysis by Huggins and Schaller - am currently making my way through it, has some good practical examples.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.