How to Pick Government Bonds Systematically Using Value, Momentum, and Carry

Book: Systematic Fixed Income: An Investor’s Guide Author: Scott A. Richardson, Ph.D. ISBN: 9781119900139 Publisher: John Wiley & Sons, 2022

Chapter 5 is where we shift from analyzing what active managers do wrong to exploring what systematic investors can do right. And it starts with government bonds, the largest and most liquid part of fixed income markets.

The Investment Universe Is Huge (But Not Really)

The ICE/BAML Global Government Index tracks about $35 trillion in government bonds across 24 sovereign issuers. That’s 1,087 individual bonds as of December 2020. The US and Japan alone account for about 64% of the total.

So do you need to forecast returns for all 1,087 bonds? No. And this is where it gets clever.

Principal Component Analysis: Simplifying Everything

Richardson introduces principal component analysis (PCA) to show that you don’t need to model every bond individually. PCA is basically a math technique that finds the common patterns in data.

Using US zero-coupon yields from 1971 to 2009, the first three principal components explain 99% of all variation in yields:

• PC1 (Level): 96.7% of variation. This is just the general level of interest rates going up or down together.
• PC2 (Slope): 2.9% of variation. This captures when short-term and long-term rates move differently.
• PC3 (Curvature): The remaining bit. This captures the “hump” in the middle of the yield curve.

So instead of modeling 1,087 bonds, you can model just three “assets” per country: short (1-5 year), medium (5-10 year), and long (10-30 year) maturity buckets. With about 13 countries that matter, your forecasting challenge goes from 1,087 items down to roughly 39. Much more manageable.

Building Zero-Coupon Yields (A Fun Exercise)

Richardson walks through a neat classroom example of how to create synthetic zero-coupon bonds from regular coupon bonds. You take three bonds with different coupons and maturities, set up a system of equations, and solve for the combination that gives you a pure single-payment bond. Matrix algebra makes this pretty straightforward.

The key insight: zero-coupon yields give you a clean way to compare bonds across different maturities and issuers. They strip out the noise from different coupon structures.

The Three Investment Signals

Now for the actual investing part. Richardson applies three familiar themes to government bond selection across countries:

Value

The idea is simple: find bonds where yields are out of line with fundamentals. The measure used here is the real bond yield, which is just the nominal yield minus expected inflation. Countries where real yields are high relative to fundamentals are “cheap.”

A good value measure should predict future yield changes. You want to see that the gap between market yields and your fair value estimate closes over time.

Momentum

Recent performance tends to continue. The measure here is the 12-month average of government bond excess returns (skipping the most recent month to avoid microstructure noise). Countries with better recent performance are expected to keep outperforming.

But own price momentum is just one option. Changes in economic indicators, inflation expectations, and prices of related assets (currencies, equities) could all feed into a broader momentum signal.

Carry

If nothing changes, carry is the return you get just from holding the bond. The measure is the term spread: the difference between long-term and short-term government bond yields. Countries with steeper yield curves have higher carry.

One consideration: carry returns can be linked to episodic crashes, so you might want to scale your exposure based on recent volatility.

The Results: Nearly a Century of Data

Using data going back to 1926 across up to 28 countries, Richardson shows the performance of each signal and their combination:

Value and carry have the most attractive return profiles. Momentum is weaker for government bonds (consistent with the difficulty of timing bond markets we saw earlier). But here’s the important part: the three signals have low correlations with each other, so combining them produces much better risk-adjusted returns than any signal alone.

The combination portfolio has a Sharpe ratio of 0.54 and, critically, the returns are diversifying with respect to traditional market risk premiums. The regression intercept (alpha) is 1.46% with a strong statistical significance.

More Recent Evidence (1995-2020)

Using JP Morgan Government Bond Index data across 13 countries, the results hold up in the more recent period too. Value has a Sharpe ratio of 0.67, carry is 0.57, and even momentum improves to 0.31. The combination reaches a Sharpe of 0.76 with an information ratio of 0.76.

The negative correlation between momentum and both value and carry is especially useful. Even though momentum is less attractive on its own, it’s a powerful diversifier when combined with the other signals.

Slope and Curvature Add More

Beyond just picking countries (the “level” asset), you can also trade the slope and curvature of yield curves across countries. The combination Sharpe ratios for slope and curvature assets are 0.84 and 0.87 respectively. And these returns barely correlate with traditional risk premiums.

Some Practical Considerations

A few things Richardson highlights:

Which bond to actually trade? When your model says “buy Japanese 5-10 year bonds,” you need to pick a specific bond. Focus on the most liquid ones with attractive carry profiles.

Europe is special. Countries sharing monetary policy (Euro members) need careful treatment. Germany, France, Italy, and Spain all share the same central bank, so their yield differences are really about credit/spread risk, not rate risk.

Emerging markets work differently. The relationship between growth and rates can flip for emerging markets. Better economic conditions can actually push yields down through tightening spreads, which offsets the usual rate-hiking channel. Don’t just copy-paste your developed market model.

Market-cap weighting isn’t broken. Some people complain that the US and Japan dominate bond indices. But market cap weights reflect market prices, and markets are reasonably efficient. If you think some issuers have too much debt, bet on that directly instead of switching to equal weights.

The bottom line: systematic security selection in government bonds works. It produces attractive, diversifying returns across nearly a century of data. That’s a strong foundation for what comes next.

Previous: Do Active Fixed Income Managers Actually Beat the Benchmark? | Next: Credit Bond Selection Fundamentals