Just an interesting observation that might be seed for an interesting discussion.
I’ve been tasked recently on some macro questions around fixed income that have forced me to do more research on the yield curve, bootstrapping spot rates, figuring out forward rates, using forward rates to figure out what future yield curves are predicted by the market.
It uses stuff I haven’t seen since L1 and L2, though I’ve been doing it with logarithmic transformations to make it easier, whereas L1/L2 skip that and therefore make the math more cumbersome (but easier to explain to candidates).
It’s really interesting, and quite different than thinking about equities and macro issues. I am glad I’ve gotten the chance to do this in greater depth.
One thing that I’ve known for a while is that macro investing tends to more about interest rates than any other thing (though growth and political risks are also important). Part of it is because interest rates are both the driver and the result of macro events, part of it is because the fixed income market is so much larger than the equity market, and part of it is because currency ideas are generally implemented with fixed income instruments (or derivatives based on them).
Sounds like fun stuff. I didn’t see a question in there, but I feel that FI is way overlooked. Each issue has it’s own implications whereas equity is just that, same old vanilla share.
Too many candidates want to be a rockstar stock picker and overlook the skilled ability it takes to build a solid FI portfolio generating yield and preserving capital.
Bill Gross used to be a professional card counter before he honed his investment acumen.
Well, I figured there may be more experienced FI types that can expand on what it’s like to do FI analysis.
In the past, I’ve mostly done analysis of pressures on central banks and how that is likely to affect nominal and real interest rates.
What I find interesting is just how much information is potentially embedded in the yield curve. For example, the current yield curve contains an estimate about what the yield curve is likely to look like next year and the year after. There are shocks of course, and a lot can happen in a year or 5 or 10, but it’s interesting that the estimates are there in the first place.
And what I’m realizing is that one of the huge issues in fixed income is that - given that the yield curve is estimated by observations at only specific points along it - how you fill in the missing areas is really where most of the art is. There’s all this mathematics that looks reliable, but the fact that you can get very different numbers depending on how you interpolate between opservations seems to open up a whole can of worms.
There are other issues, of course, such as the fact that interest rate risk should presumably cary a risk premium, so you somehow have to subtract that from spot yields before doing any estimation of yield curve expectations, and then there’s the fact that different maturities have different liquidities. There’s actually a lot more room for errors than one thinks in the beginning, and all of those errors are used to figure out whether an extra 20 bps of yield is worth it. Scary!
That’s a link that I’ve found fairly interesting re: yield curve. Pretty neat to be able to see how the yield curve changed at different points in time and what that information is telling us.
I’ve built a script in R that downloads treasury yield histories from FRED and then plots the yield curve changing over time. With a little more work, I could probably turn it into a video, but it’s still pretty interesting to watch.
Yea, that’s actually the one I had seen originally, but it wasn’t working on the CPU I was on at the time (Java wasn’t working) so I wasn’t positive that was it.
I spent six years at PIMCO analyzing mortgage-backeds. Most of my work was development of prepayment models (though there was also a lot of work streamlining their analysis software: improving OAS algorithms, rewriting all of the code to run efficiently on multiprocessor machines, and so on). Messing about with complex tranches is fun.
At work I tend to use a simple approach as that is what I need to explain to bosses, but my own stuff usually devolves to a can of worms. Splines are common as are parsimonious models like Nelson-Siegel. I also like to apply transformations that make it easier to work on the zero lower bound.
I gotta admit, I rarely really think interest rate risk premium issues. I don’t manage any bond portfolios, but I don’t really have much understanding of why it is needed.
For instance, suppose I’m managing a U.S. government bond portfolio. I model the yield curve as a Nelson-Siegel decomposition. I estimate a VAR model for the Nelson-Siegel factors (maybe plus inflation), project them to the future with Monte Carlo simulations, and back out the future yield curve in each scenario. I can take the joint distribution of the yield curve in the future and price every bond I can invest in and use the future prices to get returns. I would then have means and standard deviations, and covariances for use in optimization.
It seems like any interest rate or inflation risk premia should get bundled up into the factors, which you can model as functions of inflation or something. Maybe if you’re focused on bond pricing or writing options on bonds they’re improtant. I don’t really know.
