Why a 25-Basis-Point Surprise Moves Markets by Only Five
The New York Fed’s study of the FOMC dot plot asks a focused empirical question: when the median SEP projection differs from what market participants expected before publication, how much does market pricing move? The answer helps distinguish an informative projection from a binding promise. Across the study’s 2015–26 sample, markets respond systematically to a dot-plot surprise, but the initial response is only partial. That result is evidence of attention and uncertainty at the same time.
The event-study result: markets listen, then discount
The cleanest way to understand the estimate is to begin with two counterfactuals. If investors treated the median dot as a firm commitment, a surprise of 25 basis points in the projected federal funds rate should produce approximately a 25-basis-point move in the corresponding market expectation. That would be one-for-one pass-through. At the other extreme, if the dots contained no useful information, an unexpected change in the median should produce no systematic movement in market pricing. The New York Fed’s estimate falls between these cases.
The authors compare the median SEP projection with prerelease survey expectations at the same horizon, then examine the change in market-based expectations around the SEP release. Their baseline measure is the change in an OIS forward rate. The regression relates that market movement to the size of the dot-plot surprise, while controlling for the change in market expectations between the survey response date and the day before publication. The coefficient answers a specific question: on average, how many basis points does the market expectation move when the SEP surprise is one basis point?
In the pooled sample, the coefficient is about 0.22 with daily OIS data and about 0.20 with intraday data. The practical translation is the study’s most useful number: if the median SEP path is 25 basis points higher than market participants expected, the corresponding market pricing initially rises by about 5 basis points on average. A basis point is one hundredth of a percentage point, so the market moves in the same direction but only about one-fifth as far in the initial window.
That is not a measure of how far the federal funds rate will eventually move. It is not an estimate of the probability that the FOMC will implement the dot path. It is an average initial pass-through from an unexpected projection to a market-implied future rate. Those distinctions matter because the market instrument is responding to the information content of the release, not mechanically copying the number printed in the chart.
The daily and intraday estimates tell a consistent story while serving different measurement purposes. Daily data cover the change in market rates across the release day. They are easier to obtain across the sample, but they can contain the effect of the post-meeting press conference and other news that arrives during the day. Intraday data look at a 30-minute window around the release. The narrower window produces a closer approximation to the immediate response and, in the study, similar coefficient estimates with smaller standard errors. It still does not isolate the dot plot perfectly, because the FOMC statement is released at the same time.
The evidence is therefore stronger than a casual visual comparison of two charts, but narrower than a claim about the entire public’s expectations. The study observes professional market pricing at specified horizons. It does not ask every household or business whether a dot changed its plans. Nor does it imply that a 0.20 coefficient applies unchanged to every meeting, every economic regime, or every maturity. It describes the average response in a defined event-study design.
The response also varies by projection horizon. In the table reported by the authors, the daily coefficients are approximately 0.19 for the end of the current year, 0.34 for the next year, 0.20 for the following year, and 0.10 for the year after that. The intraday estimates are approximately 0.24, 0.26, 0.13, and unavailable for the longest horizon because the study lacks intraday data beyond three years. The estimates are positive at most horizons, but their precision is not uniform. The longest daily estimate is based on only 21 observations and is not statistically distinguishable from zero in the reported table, unlike most of the shorter-horizon estimates.
This variation does not overturn the pooled result. It explains why a single conversion rule would be misleading. A projection for the end of the current year and one for a date several years away do not carry the same amount of uncertainty, and the market may use information about them differently. The study’s sample contains 150 observations for the pooled daily specification and 135 for the pooled intraday specification, while the horizon-specific samples are smaller. The coefficients summarize those observations; they do not establish a law of market behavior.
The diagram captures the logic without turning it into a mechanical calculator. The surprise is measured first. The market then reprices the corresponding future rate. Macroeconomic projections provide additional information about the outlook, but the FFR-projection effect remains in the combined analysis. The observed response is positive and discounted, which is exactly what one would expect from a conditional signal that carries information but is exposed to future change.
A coefficient below one can arise for several compatible reasons. Investors may believe that the projection reveals something about how policymakers are thinking while assigning a probability to a different economic path. They may expect new data to arrive before the projected date. They may also distinguish between a participant’s current assessment and the Committee’s future decision, which will be made in a later state of the economy. The New York Fed paper’s interpretation is that the partial adjustment is consistent with the public understanding the dots as conditional forecasts containing substantial uncertainty.
The important negative inference is equally clear. Partial pass-through does not show that the dots are ignored. The response has the expected sign and is statistically meaningful across the pooled specifications. A market that ignores the dots would not, on average, move in their direction when they surprise it. The market is listening; it is simply not surrendering its own assessment of future conditions.
What remains after controlling for the economic outlook
A higher dot path can mean at least two things. It can say that FOMC participants expect stronger growth or higher inflation, conditions that would themselves make higher future rates more likely. Or it can say that, given a broadly similar outlook, the Committee would respond more firmly to that outlook than markets had assumed. These channels are related in the SEP, but they are not identical. The second is information about the policy reaction function: the way the FOMC maps economic conditions and risks into policy decisions.
The study tests the distinction by adding surprises in SEP projections for GDP and headline inflation measured by the Personal Consumption Expenditures Price Index. These macroeconomic surprises are measured relative to the Bloomberg survey. Because the relevant survey data begin in 2017, the sample becomes smaller. The authors therefore focus on pooled horizons rather than maintaining all of the horizon-specific specifications.
The baseline FFR-surprise coefficient in that reduced sample is about 0.23. After GDP and headline PCE inflation surprises are included together with the FFR surprise, the coefficient remains about 0.20. The macroeconomic variables matter: in the macro-only specification, the GDP-surprise coefficient is about 0.06 and the headline PCE inflation coefficient about 0.21. In the combined specification, the corresponding coefficients are about 0.07 and 0.13. The direction is consistent with the idea that stronger growth or higher inflation can lead markets to price a higher future policy rate.
But the central FFR coefficient does not collapse toward zero when those variables are added. That persistence is why the authors conclude that much of the information the market extracts from the SEP concerns how the FOMC is likely to behave conditional on the economic outlook, not only what the outlook is. The dot plot can therefore communicate a policy reaction before the policy rate itself changes. It tells markets something about the decision rule or judgment they should expect to encounter as the economy develops.
This is an interpretation supported by the pattern of coefficients, not a complete causal decomposition. The FOMC statement arrives with the SEP and may contain related information. The GDP and PCE projections do not capture every aspect of policymakers’ assessments, such as uncertainty around their forecasts or variables not represented in the survey. The authors note that surprises in unemployment and core PCE inflation were less explanatory in their additional tests, but that does not mean those variables are irrelevant to policy. It means they add less explanatory power in the specification and sample examined.
The distinction between forecast information and reaction-function information is useful because it changes how a reader interprets a revised path. If a higher dot simply reflected a higher expected inflation path, the market could learn mainly about the economy. If the FFR projection remains informative after macro surprises are controlled for, the market is also learning how policymakers may respond to the outlook. That does not make the response deterministic. A reaction function is not necessarily a fixed formula, and the paper does not claim that the median dot reveals every judgment inside the Committee.
The alternative explanations set a useful standard for what future evidence would change the conclusion. If controlling for the relevant macroeconomic information consistently removed the FFR-projection effect, the case for a distinct reaction-function signal would weaken. If a different measure of expectations or a narrower release window produced no systematic response, confidence in the market-attention finding would also decline. In the reported results, however, the FFR coefficient remains positive and close to the baseline estimate. The evidence therefore favors a mixed interpretation: markets learn about both the outlook and the policy response, with the reaction-function component remaining material.
That mixed interpretation also clarifies why partial adjustment is better described as a conditional discount than as confusion. A market participant can take the dot plot seriously and still ask what would happen if inflation were lower, demand weaker, or financial conditions tighter than assumed. The participant’s price then reflects a weighted assessment across possible states rather than an instruction to place the full 25 basis points into one future rate. The resulting 5-basis-point initial move is not a claim that the other 20 basis points disappear. It is the amount of immediate repricing observed in the study’s market measure.
The market can later incorporate more of the signal as the projection date approaches, or the SEP projection can move toward the market’s view. That later movement is outside the immediate coefficient’s scope. The event study establishes the first step: an unexpected dot-plot change moves expectations in the same direction, but less than one-for-one. The fact that the initial response is incomplete is precisely what leaves room for a time dimension in which both forecasts can be revised.
A careful reading therefore keeps four statements separate. First, the released median was different from what the selected survey expected. Second, the corresponding market-implied rate moved after the release. Third, the size of that initial move was positive but partial. Fourth, controlling for selected macroeconomic surprises left a meaningful FFR-projection effect. Together these statements support an informative, conditional signal. They do not support the stronger claims that the dots dictate future rates, that the market ignores them, or that the study has isolated a single pure policy channel.
The title’s five-basis-point figure is consequently a clue about the status of the dot plot, not a promise about the next rate decision. It shows that a published projection can change beliefs without replacing judgment. To understand the full implication, one must connect the initial repricing to the way the market forecast and the SEP projection evolve after the release, while keeping the study’s sample, instruments, and uncertainty in view.