How are rate expectations measured?

How much a rate decision matters depends largely on how far it departs from expectations. There are two common ways to measure those expectations: market pricing and surveys. They measure different things and sometimes disagree.

Market pricing

Some financial instruments pay off according to the future overnight rate: interest rate futures, OIS (overnight index swaps) and options on them. Their prices today reflect what the market expects the average overnight rate to be over the coming months. That implied path can be split at meeting dates and turned into the change expected at each meeting.

The expected change can also be turned into a probability, but only under an assumption: if the only options are "no change" or "25 basis points", a priced +13 basis points means roughly a 50% chance of a hike. In reality, the possibility of a larger move, term premia and the smoothing of the curve all affect the figure, so the result should be read as approximate.

The site uses this method for the BoE: the daily OIS curve published by the Bank of England, read against that day's actual overnight rate (SONIA). The Atlanta Fed probabilities shown for the Fed are something else: they do not refer to a single meeting but to which band the average rate over a three-month period will fall into. They read as "the probability that the average rate over September–December is above today's range is X%", not "the probability of a hike at the September meeting is X%".

Surveys

Surveys ask economists and market participants directly. The site uses three:

• CBRT Survey of Market Participants: published monthly; gives the average expected policy rate for the next meetings and the end of the year.
• ECB Survey of Monetary Analysts: run before each monetary policy meeting, results published after it; includes the median expectation and the odds analysts give to each size of change.
• New York Fed Survey of Market Expectations: run before each FOMC meeting, results published a few weeks later; gives meeting-by-meeting expectations and probability distributions.

Surveys take the pulse of opinion, but they are infrequent and follow a publication schedule; data or remarks that arrive after the survey are not reflected. The average and the median can also tell different stories: if most participants expect a hold but a few expect a big cut, the average is pulled down.

Why can the two disagree?

Market prices change every trading day and represent a forecast backed by real money, with a risk premium built in. A survey is a set of views collected on a given date. If important data or a speech arrives after the survey, market pricing can quickly move away from it. That is why the site always states the type and date of each expectation.

How is a surprise measured?

On the decisions page, each decision is compared with the latest expectation formed after the previous decision and before this one: for the BoE, the curve from the last business day before the decision; for surveys, the latest round whose responses were collected before the decision. Older expectations are not used, because they would carry the surprise of the decision in between into the next one.

• Gap = decision − expected rate, in basis points. A positive gap is a decision tighter than expected (hawkish), a negative gap one looser than expected (dovish).
• A gap below 10 basis points counts as in line with expectations.
• Between 10 and 25 basis points is shown as "slightly more hawkish/dovish than expected", and 25 or more as a surprise.

Current expectations are on the rate probabilities page and source details on the methodology page. To read a decision from start to finish, see how to read a rate decision.