The Federal Open Market Committee meets today and tomorrow, and this is one of the four meetings a year that publishes a Summary of Economic Projections. Within that document sits the dot plot — the chart that will dominate coverage, and the one most reliably misread.
What it is
Each participant submits their view of the appropriate federal funds rate at the end of each of the next several years, and over the longer run. Those submissions are plotted as anonymous dots.
Four features of the construction do most of the work in explaining why it misleads.
- It includes non-voters. All participants submit dots, including the regional presidents who do not hold a vote in the current rotation. The chart therefore describes a different population from the one that will actually decide.
- It is anonymous. There is no way to know which dot belongs to whom, so a dot from a permanent voting member counts the same as one from a participant who will not vote for years.
- It is conditional. Each participant submits the rate they consider appropriate given their own forecast for growth, unemployment and inflation. Two participants with identical policy preferences will submit different dots if they hold different forecasts.
- It is not a commitment. No participant is bound by it, and no vote has been taken on it.
The median is not a plan
Coverage reduces the chart to a single number — the median dot — and reports it as the Fed’s intention for the year.
It is not an intention. It is the middle observation of a set of individually held, forecast-conditional opinions that were never aggregated by any decision process. Nobody voted for the median. It is a statistic computed over a scatter, and it can move because one participant shifted position without anything changing about the Committee’s centre of gravity.
The distribution carries far more information than the median. A tight cluster indicates genuine agreement. A bimodal scatter — dots grouped at two separated levels — indicates a Committee split into camps, and that is a materially different situation with identical medians.
The forecasting record
Over horizons beyond a couple of quarters, the dot plot has been a poor predictor of the actual policy rate. This is well documented, acknowledged by officials, and unsurprising.
The reason is not that participants forecast badly relative to anyone else. It is the conditionality. A dot says: if the economy evolves as I expect, this is the appropriate rate. The economy routinely does not evolve as expected, and when it diverges the appropriate rate diverges with it. The dot was never a prediction of the rate; it was a prediction of the rate conditional on a forecast that turned out to be wrong.
Read correctly, the dot plot is a reaction function expressed in shorthand — information about how the Committee would respond to a particular path, not about which path will occur.
Why it exists at all
Given the misreading, it is fair to ask why the Fed publishes it. The answer is that expectations about the future path of rates matter more for the economy than the current rate does.
A ten-year yield reflects the market’s expectation of short rates over a decade. Mortgage and corporate borrowing costs price off those longer yields. If the central bank can shape expectations about the path, it influences borrowing costs across the economy without changing the current rate at all.
The dot plot is an instrument for doing that. Its imprecision is partly deliberate — a specific commitment would be more powerful and also more dangerous, because breaking it would damage credibility more than vagueness does.
What to look at instead of the median
- The dispersion. How wide is the scatter, and is it widening or narrowing against the previous submission? Widening dispersion signals genuine disagreement.
- The shape. Clustered or bimodal. Two camps is a different committee from one distribution with a long tail.
- The longer-run dot. Participants’ view of the neutral rate — the level that neither stimulates nor restrains. Movement here is the most consequential change the chart can show, because it repositions the entire framework rather than the near-term path.
- The accompanying forecasts. The rate dots are conditional on the growth, unemployment and inflation projections published alongside them. If the inflation forecast moved and the dots did not, the reaction function has changed — which is more significant than a shift in the dots themselves.
The trap
Comparing today’s median against the previous one and reporting the difference as a policy change is the standard error. The dots are not a decision, the participants are not all voters, and the projections they are conditioned on will change before the dates in question arrive.
The vote recorded in the statement is a fact. The dot plot is a survey. When they appear to conflict, the vote is the one that happened — a point worth holding onto whenever the two are reported with equal weight.
