Testing the Expectations Hypothesis with Yield Curve Regressions

Summary

The document introduces the Campbell–Shiller regression as a way to test the expectations hypothesis of interest rates. It describes comparing a yield spread with a subsequent change in yields, while noting that empirical estimates may differ from the theoretical benchmark. The author asks whether using daily zero-coupon-bond yields and treating 252 trading days as a year could explain a positive coefficient greater than one, despite reports of negative estimates in the literature.

The post raises a key horizon question: “one period ahead” depends on the regression’s chosen observation interval and maturity definitions. It does not provide an answer, data, or a diagnosis of the reported coefficient. Readers would need to check the exact regression specification, yield units and maturities, sampling frequency, and construction of the forward yield change before drawing conclusions. The document is therefore useful as a prompt about empirical setup, but not as a complete implementation guide.

Key ideas

• The expectations hypothesis can be examined by regressing future yield changes on yield spreads.
• The Campbell–Shiller setup relates the maturity spread to subsequent changes in yields.
• A coefficient that differs from the theoretical benchmark may reflect the specification or data construction.
• The meaning of one period ahead must match the regression horizon and sampling interval.

Tags

# Checking the expectations hypothesis of interest rates


# Checking the expectations hypothesis of interest rates












I'm trying to perform the Campbell-Shiller (1991) regressions to check the expectations hypothesis which basically says that the difference between ($s-m$) period yield and $s$ period yield should be equal to the spread between $s$ and $m$ period bond (with some constant premium).

The literature points out that the beta coefficient turns out to be negative instead of 1. I did the same but my coefficient turns out to be positive and more than one. What could I be doing wrong? I am using daily ZCB yields, and take 252 days as one year.

P.S. If I want to calculate the change in yield one period ahead, would that mean the change in yield for today to tomorrow, or the change in yield from today to next year?

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.