Monetary Policy & Inflation | US
Forecasting the Fed’s Policy Rate
- By Sam van de Schootbrugge
- Sign-up to discuss on Slack
- 7 min read
Summary
- A new BIS working paper examines how prone investors are to incorrectly forecasting US monetary policy. It does so by decomposing excess returns in fed funds futures and overnight index swaps.
- It finds excess returns in money markets over the last 30 years were predictable and can be explained by market participants consistently underestimating Fed policy rate changes.
- A large share of this predictability is concentrated in periods with large and negative macroeconomic shocks, but positive excess returns in FF futures are also found during Fed hikes.
Introduction
How well do investors forecast changes in US monetary policy? The answer is embedded in the prices of fed funds (FF) futures and overnight index swaps (OIS), according to a new BIS working paper. By looking at the term structure of money market rates, the authors can assess how accurately market participants predict future short-term interest rates.
The key is decomposing FF futures and OIS excess returns into two components: (i) a term premium and (ii) an expectations error. The prevailing view across asset classes is that investors earn positive excess returns because of risk premia, but the results from the BIS paper for money markets sharply contrast this. It finds:
- All excess returns in FF futures and OIS stem from expectations errors, implying market participants are prone to monetary policy expectation errors that do not average out over time.
- These errors are the result of ‘conservatism’ – market participants can correctly predict the direction of future interest rate changes but tend to underestimate their magnitude.
- Expectation errors, and therefore the excess returns, are higher in periods of deteriorating financial conditions – a drop in equity prices significantly predicts higher excess returns on FF futures and OIS.
Methodology and Data
The best way to understand the authors’ approach is algebraically. On the expiry of an FF contract of length n, an investor earns a fixed rate ft and pays a short rate in the future it+n, which is the average realised effective fed funds rate (EFFR). So the excess returns of an FF futures contract are given by:
- rxt+n = ft – it+n
And so, if we think about what the current fixed rate on FF futures should incorporate, we switch ft and rxt+n around to find that it contains a term premium Et[rxt+n] and an expectations component of the future EFFR short rate Et[ it+n ] .
- ft = Et[rxt=n] – Et[it+n]
If we substitute (2) into (1), excess returns are then explained by a term premium and an expectations error that is Et[ it+n ] – it+n . Under rational expectations, this would be zero (investors do not make systematic forecast errors), and changes in the excess returns would reflect changes in the term premium.
To test if this is so, the authors proxy Et[ it+n ] with interest rate forecasts from the Blue Chip Financial Forecasts survey. From the survey, they obtain fixed-horizon short rate expectations for n = 3; 6; 9; and 12 months, denoted by St. And so, the survey-based analogue of (1) then becomes:
- rxt+n = ft – St + St – it+n
Notice here that the term premium is ft – St and equals the amount by which FF futures or OIS rates deviate from expected short rates over the maturity of the contract. Also, notice that St – it+n is the ‘expectations error’, defined as the difference between expected and realised short rates.
They obtain data on historical FF futures prices and OIS quotes from Bloomberg. For FF futures, the data go back to 1990, while OIS rates are available for the US since December 2001. For both series, the sample ends in September 2021.
Excess Returns in FF Futures and OIS
First, the paper shows how mean excess returns of FF futures and OIS have varied over the last three decades (Chart 1). On average, excess returns have been positive, and the variability of those returns has fallen during the 21st century, apart from during recessions.
What does this mean? It says investors in these markets over the last three decades have earned positive excess returns by entering FF futures contracts that lock in fixed rates today while paying the realised short rate in the future.
Algebraically, it implies that rxt+n has typically been positive. From equation (3), this can be explained by positive term premia, positive expectation errors, or some combination of the two.
Term Premia and Excess Returns
Are changes in the term premia, or term spreads, responsible for changes in FF futures or OIS excess returns? Yes, according to the authors. But contrasting the prevailing literature, the term premium negatively affects excess returns! In other words, a negative term spread predicts positive excess returns.
For this to be the case, there must be large expectation errors. For example, for 12-months-ahead OIS, the authors find that a predicted 1% change in short rates is, on average, followed by a 1.44% realisation. That is, market participants systematically underestimate future short rate changes.
How can we rationalise that? Well, if the term spread is negative (i.e., markets expect a cut), the subsequent expectation error must be positive. So investors, on average, believe the rate cut will be smaller than it turns out to be.
Graphically, the following quadrant chart helps (Chart 2). The diagonal quadrants are periods when the market correctly anticipated changes in the short rate. For the result above to hold, most of the observations need to be in these quadrants, which they are (78%).
However, the magnitude of the errors also needs to be largest in these quadrants. The distance of each dot away from the 45-degree line is the size of the error. And immediately we see most of the dots are farthest away from the line in the bottom-left quadrant.
This quadrant represents the observations that correctly predicted a rate cut. The large deviations away from the 45-degree line in this quadrant imply most of the forecast errors in short rates were underestimates of rate cuts, exactly what a negative relationship between the term premia and excess returns required!
For example, for six-month-ahead FF futures, the market underestimated the magnitude of the decrease by at least 25bps in 36% of cases. Meanwhile, it only underestimated the magnitude of an increase by at least 25bps in 4% of cases.
The Fed and Excess Returns
That forecasters are more likely to make errors during periods of rate cuts than rate hikes coincides with excess returns being higher during recessions (Chart 1). This is very much linked to monetary policy uncertainty: the authors find uncertainty about future short rates (as measured by the dispersion of expectations across forecasters) is a strong predictor of expectation errors.
The link between excess returns, forecast errors, and monetary policy uncertainty also helps explain why the variation in excess returns has fallen during the 21st century. According to the paper, the Fed’s ability to transparently communicate its path for policy rates has improved significantly since the 1990s.
Other Predictors of Excess Returns
The authors also show how drops in equity prices also predict higher excess returns on FF futures and OIS. Moreover, they observe a strong asymmetry in this relation: lower stock returns predict higher excess returns on money market derivatives, whereas rising stock returns contain no predictive power.
Bottom Line
The paper shows why long positions in fed funds futures and overnight index swaps over the last three decades have on average delivered positive excess returns. Contrasting traditional explanations that link positive excess returns to a term premium, the authors show how incorrect forecasting of Fed policy drove the result.
So, with FF futures and OIS, the authors argue that viewing the term premium as compensation for risk is wrong. Instead, any term premium variation reflects the price that institutions active in the money market are willing to pay to hedge against future short rate changes.
Sam van de Schootbrugge is a Macro Research Analyst at Macro Hive, currently completing his PhD in international finance. He has a master’s degree in economic research from the University of Cambridge and has worked in research roles for over 3 years in both the public and private sector.