Identifying Value in Sovereign Bonds: Understanding the Yield Curve
Chapter 5 brings us to the classic fixed income question: which bonds are cheap and which are rich? But the authors remind us right away that cheap-rich analysis is just one part of relative value. The bigger question remains: what is the best way to express a particular view?
This chapter focuses on sovereign bonds and breaks into three parts: yield curve fundamentals, spread measures, and applied techniques for finding value. This post covers the first part. The spreads and applications get their own post.
Macro vs. Micro Relative Value
The authors make a useful distinction. Micro relative value is the textbook version: buy bond A because it is underpriced, sell bond B because it is overpriced. But most micro trades have a macro story underneath them.
For example, trading the asset swap spread between two sovereign issuers is technically a micro trade. But asset swap spreads move with the market direction. You cannot fully separate the micro from the macro. Keep that in mind whenever someone tells you their trade is “pure relative value.” It probably has a directional element hiding in there.
They also point out that relative value trades are not risk-free arbitrage. A lot of CDOs were marketed as “arbitrage” vehicles before 2007. They just redistributed credit risk without removing it. True arbitrage is incredibly rare and gets closed almost immediately.
Yield Curve Theories
There are four main theories about why the yield curve has the shape it does. The authors walk through each one, and I think this is actually one of the most useful sections in the book.
Expectations Hypothesis
This is the cleanest theory. The yield curve reflects the market’s expectations of future interest rates. If the 1-year rate is 5% and the market expects the 1-year rate next year to be 4%, then the 2-year rate should be 4.5%. You can extend this logic to build the entire curve from a series of expected short-term rates.
The powerful implication: if this theory holds, you can back out the market’s rate expectations from the observable yield curve. When financial commentators say “the market expects rates to rise,” they are doing exactly this calculation.
But here is the catch. If other factors besides expectations also influence the curve, then the forward rates you calculate are not pure expectations. They are forward rates, not expected rates. That distinction is really important and most people mix them up.
Liquidity Preference
People generally want to keep their money accessible. To get them to lock up funds for longer periods, you need to pay them more. This creates a natural upward slope to the yield curve even if rate expectations are flat.
Preferred Habitat
Different investor types cluster around certain maturities. Pension funds want 10-30 year bonds. Money market funds want 0-1 year paper. Mortgage banks want 5-year instruments. Each group has a “home” on the curve. Getting them to move to a different maturity requires a bribe in the form of higher yield.
Market Segmentation
Similar to preferred habitat, but stricter. Some investors have legal or regulatory constraints that prevent them from investing outside certain maturity buckets. Money market funds cannot buy 10-year bonds. Period.
How They All Work Together
In reality, the curve reflects all four factors at once. The short end might be shaped by market segmentation (money market funds piling in). The middle of the curve might reflect rate expectations. The long end might show liquidity preference and preferred habitat effects from pension funds.
The authors make a good cynical point: realistically, the market cannot formulate reasonable rate expectations much past 5 years. Maybe not even past 1 year. So pretending that the 30-year yield tells you anything specific about rate expectations is probably fiction. At best, a certain rate profile is “discounted” in the curve.
They also note something I find fascinating. If yield curves truly reflected the economic cycle, they should show a cyclical pattern of rising and falling rates. But they almost never do. Curves tend to discount rates rising and then flattening out. That means longer-dated yields are generally too high relative to what actually happens. Investors in long bonds earn excess returns over time. Borrowers who lock in long-term funding generally pay more than they need to.
How the Yield Curve Moves
The authors reference Antti Ilmanen’s work, which identifies three factors that influence curve shape:
Market expectations of future rate changes. If yields are expected to rise, the curve steepens so that the higher initial yield offsets the expected capital loss. If yields are expected to fall, the curve flattens or inverts.
Risk premium differences across maturities. The pure expectations hypothesis says all bonds of the same credit quality should have the same holding period return. But risk premia exist and vary across maturities. Ilmanen found that shorter-dated bonds generally carry higher risk premia, though this changes over time.
Convexity bias. Convexity is valuable, especially in volatile markets. Investors are willing to accept lower yields on highly convex bonds because they offer better return profiles when rates swing. This tends to push down yields at the long end of the curve.
Types of Curve Movements
The standard vocabulary: parallel shifts (all maturities move the same amount), steepening, flattening, inversion, normalization, and changes in curvature.
A steepening curve can be bullish (driven by falling short rates) or bearish (driven by rising long rates). A flattening curve can also be bullish (falling long rates) or bearish (rising short rates). You need both the direction and the driver to describe what is actually happening.
Principal Component Analysis
PCA is the standard tool for understanding how yield curves actually move in practice. The technique looks at historical yield changes across many maturities and identifies the independent components that explain the variation.
For most government bond markets, three factors explain over 95% of all yield curve movements:
Factor 1: Parallel shift. This is the dominant factor. All yields move roughly the same direction by roughly the same amount. This explains the largest share of total variation.
Factor 2: Steepening/flattening. Short rates move in one direction while long rates move in the other. The rotation point tends to be around the 3-year maturity.
Factor 3: Curvature change. The curve becomes more or less humped. Short and long rates move one way while the middle of the curve moves the other.
The authors show PCA results for US Treasuries from 2006 to 2009. The factor values make intuitive sense. Factor 1 values are similar for the 2-year through 30-year maturities, confirming the parallel shift interpretation. Factor 2 values decline with maturity and turn negative around the 3-year point, confirming the steepening/flattening interpretation.
Here is the practical takeaway: if you can predict Central Bank activity, you can explain most of what the yield curve will do. When central banks cut rates, curves rally, steepen, and become more concave. When they hike, curves sell off, flatten, and become more convex. Shorter-term rates display greater volatility than longer-term rates because they are more directly linked to policy rates.
Yield Curve Modelling
Building a yield curve from actual bond data requires some choices. Which bonds do you include? How do you connect the dots?
For the bond selection, you want issues with the same credit rating. Within government bonds, you typically prefer “on the run” issues (recent, liquid) over “off the run” issues (older, less liquid). High-coupon bonds trading at a big premium might be excluded because of “pull to par” effects. Callable bonds get excluded because the embedded option distorts the yield.
For connecting the observations, the most popular technique uses cubic splines. You fit cubic equations between pairs of maturity points, then enforce smoothness by ensuring the slope and curvature match at each joining point. This gives you a continuous, smooth curve.
Every institution will have its own preferred method and its own bond selection criteria. That means no two institutions will have exactly the same yield curve. This is actually one reason why some analysts prefer using the swap curve as a benchmark. Swap rates are directly observable and do not require the same degree of subjective modelling.
My Take
The yield curve theories section is something every finance person should read carefully at least once. Most people learn the expectations hypothesis in school and stop there. But the reality is messier and more interesting.
I keep coming back to the authors’ point that forward rates are not expected rates. It seems like a small distinction, but it changes how you interpret market signals. When Bloomberg shows you “market implied” rate hikes, remember that number includes term premia, liquidity preferences, and all sorts of other stuff. It is not a clean forecast.
The PCA section is also great for building intuition. Three factors explain 95% of curve moves. That is a lot of complexity reduced to three simple stories: level, slope, and curvature. If you get those right, you get most of the trade right.