Interpreting Fed Funds Futures Probabilities and Forecast Errors
Summary
The question compares a probability shown by the CME FedWatch Tool with a calculation based on a fed funds futures price, the effective rate for part of the month, and the number of days affected by a rate decision. The author finds that the calculation implies a substantially different chance of a rate increase and asks why the displayed probability appears unchanged when the futures price changes. The document does not resolve how the tool converts futures prices into probabilities, so it cannot establish whether the calculation or displayed figure is correct.
A separate response argues that FedWatch predictions made weeks before several 2022 meetings were wrong, citing instances where the predicted chance of a change was zero or where the predicted size differed from the eventual move. These examples raise a useful distinction between market-implied probabilities and reliable forecasts, but they are an anecdotal critique rather than a systematic accuracy study. The document gives no evaluation method, full sample, or treatment of changing probabilities over time.
Key ideas
- Fed funds futures prices can be used to infer expectations about policy rates, but translating prices into meeting probabilities requires assumptions.
- The document highlights a discrepancy between a hand calculation and the probability displayed by FedWatch.
- Historical examples are offered as a critique of FedWatch forecasts, but they do not constitute a systematic accuracy test.
- Market-implied probabilities should not be treated as certain predictions of future policy decisions.
Tags
Full text
# Question regarding the accuracy of CME FedWatch Tool # Question regarding the accuracy of CME FedWatch Tool I just accessed the CME FedWatch Tool at 10:27am ET. It reported mid price of ZQZ2 as 95.8788 and a corresponding probability of 74.7% of 50bp hike at Dec FOMC meting (vs 75bp). For December, as this question explains, there are 14 days of effective Fed funds rate at 3.83% and 17 days of EFFR to be decided at the Dec 14 FOMC meeting. The implied probability should be (futures MID - weighted EFFR)/(size of hike * num of days after hike/total), which, for 50bp, is (95.8788 - 95.6216)/(0.5 * 17/31) and that's 93.8%. Why is there such a large difference between what should make intuitive sense and the CME FedWatch Tool number? Btw, when I accessed the page earlier, it reported the same probability for a different ZQZ2 mid of 95.8738. How can that be? ## Answer by Daxx (score 1) https://quant.stackexchange.com/a/76663 My analysis shows in the last 2 years and roughly 9 rate hikes, the the "CME FedWatch Tool" was completely wrong 4 times! That is HUGE (approaching 50% wrong)! The following dates reflecting the FOMC meeting for which the CME FedWatch Tool made predictions a few weeks prior. July 27, 2022 0% prediction and the fed DID raise rates. Sept 21, 2022 at 54% predicting no change and they raised rates. Nov 2, 2022 > while the probability of a hate hike was correct at 70% .. all 70% was signalling a 50 bases point hike and 0% of a 25% which is actually what happened. Dec 14, 2022 got it completely wrote at 0% predicted any change, but the feds did raise rates. My point is that I can't see why SO MANY PEOPOLE (cites it as some sort of proof) when the results of the CME FedWatch Tool predictions a few weeks before the FOMC meeting have such a history of being COMPLETELY wrong (it's not like it said 25% and that was wrong - it sometimes read 0% chance and that is wrong ).
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.