Expectations against actual

The number is not the news 

A price at any moment already contains a view about releases that have not happened yet. The calendar is public, the release dates are fixed months ahead, what each series measures is documented, and participants form expectations about all of it. Those expectations are not opinions held quietly. They are expressed by committing money, which is what a price is. So by the time a release is published, the expectation of it has already been paid for and is sitting in the quoted price of every instrument the release touches.

What is left to move a price, therefore, is only the part of the release that nobody had accounted for. That residual is the difference between what arrived and what was expected, and it is the only quantity in the whole exercise that carries new information. The level of the figure does not, because the level was anticipated. Its change from the previous month does not, because that change was anticipated too. A deterioration smaller than expected and an improvement smaller than expected can produce moves in opposite directions, though both describe the same direction of travel in the underlying economy, and neither reaction is a mistake.

Where the expected figure comes from 

The previous lesson established what the consensus column is: a median, at some vendors a mean, of a survey of forecasters compiled by the calendar provider up to a cut off before the release. What follows from that construction is worth drawing out, because a figure printed in a confident column beside an official statistic invites an assumption the process does not support. Participation in the survey is voluntary, submissions arrive days early, and some contributors revise after intervening data while others do not.

The consequence stated plainly is this: the consensus is not a forecast anyone is accountable for. No individual economist holds the median. No institution publishes it as its house view. No committee reviewed it, nobody is measured against it, and no consequence attaches to a contributor whose submission lands a long way from the outcome. It is an average taken over a group of people who did not coordinate, were not all answering on the same day, and were each asked for an estimate rather than for a commitment.

The compression also discards the most informative part of the survey. A panel whose forecasts sit within a narrow band and a panel whose forecasts are spread across a wide range can produce an identical median while describing completely different states of knowledge, and the dispersion behind the figure is rarely printed beside it. Where the range is available it describes how much the forecasters disagreed, which is a different question from what they concluded on average.

There is a further gap. The published consensus is not necessarily the expectation the market is actually trading against. Positions are adjusted continuously, and prices absorb information that arrived after the survey closed. Practitioners describe that gap with the term whisper number, an informal expectation circulating in the hours before a release that can sit some distance from the printed one, and where an expectation can be read directly out of prices, as the implied path of policy rates can be read out of interest rate futures, that market implied expectation is often treated as the more relevant benchmark. Nothing obliges the two to agree.

What priced in describes 

Priced in describes a price that already reflects an expectation, so that the arrival of the expected thing changes nothing. As a description of the mechanism in the first section it is exact. The difficulty is that it names a quantity nobody can look up. There is no field on any screen that reports what is priced in. It is inferred, and the inference is assembled from several partial sources.

• Instruments whose prices are a direct function of an expectation. Interest rate futures and overnight index swaps imply a path of policy rates, from which the probability the market attaches to a move at a given meeting can be derived.
• Options. The implied volatility of contracts expiring around a scheduled release describes the size of move participants are paying to cover. It describes size only, and says nothing about direction.
• Positioning data. Regulatory reports of futures commitments and dealer surveys describe what a subset of participants are holding, and they are published with a lag measured in days.
• The behaviour of prices since the previous comparable release, read as evidence of what participants have been adjusting toward.

Each of those describes something adjacent to the question rather than the question itself. Rate futures describe expectations about a committee, not about a statistics office. Options describe magnitude, not content. Positioning data covers a reported subset and arrives late. So priced in is a construction assembled from indirect evidence, held with differing confidence by different desks, and revised continuously. It is a reasonable working description. It is not a measurement, and the two are easy to confuse because the phrase is spoken with equal confidence in both senses.

Measuring the difference 

The residual has a name. The surprise is the actual figure minus the expected one, and in its simplest form it is that subtraction and nothing more. It is signed, positive when the release came in above the consensus and negative when it came in below. The sign is arithmetic rather than a verdict: whether a release above expectation is read as favourable to a currency depends entirely on what the series measures and on the mapping described in the next section, and there are releases where the arithmetic and the market convention point opposite ways.

A raw difference is not comparable across releases, because the same numeric gap means different things on different series. A gap of a tenth of a percentage point on an inflation rate is an ordinary week; the same gap on an unemployment rate that has barely moved in a year is not. So the difference is conventionally standardised, divided by a measure of how far that particular release typically lands from expectation, either the dispersion of the current forecast panel or the historical variability of past surprises on the same series. The result expresses the surprise in units of typical, which is what makes one release comparable with another.

Reading the two cases side by side is the whole lesson in one object. Nothing about the economy differs between them. The published figure is identical, its movement from the previous month is identical, and any description of the number in isolation would be word for word the same. What differs is the figure it is being measured against, and that is the thing the price was already holding.

Why the direction is frequently counterintuitive 

Even when the surprise is unambiguous, the direction of the move often is not, and there are four separate mechanisms behind that. The first is the reaction function, the mapping from an economic outcome to what a central bank is expected to do about it. A macro release reaches a currency mostly through that mapping, and the mapping is not fixed. Where a committee has said its concern is inflation, an activity release above expectation implies rates held higher for longer. Where the same committee has turned to a slowing economy, the same release implies less urgency to reduce them. Where the concern has moved to stress in the financial system, the mapping can invert again. The number is stable; what it is understood to imply is not.

The second is composition. A release is a document rather than a number, and the headline figure at the top is one line of it. An inflation release is conventionally read on a core measure that excludes food and energy, on the reasoning that those components are volatile for reasons unrelated to domestic price pressure, and that convention is contested precisely because food and energy are what households actually pay. An employment release is read alongside earnings growth, hours worked and participation. A headline above expectation sitting on top of a core measure below it is an ordinary occurrence, and a price that moves against the headline is frequently moving with the detail.

The third is that a release usually restates the past as well as reporting the present. Prior periods are revised in the same publication, and a strong current month accompanied by downward revisions to the two before it can leave the level of the series lower than it was understood to be before the release opened. Revisions are the subject of the next lesson in this module, so they are only flagged here.

The fourth is positioning. Where a large share of participants already hold the same position going into a release, the outcome that confirms it can move the price very little, because there is little left to be added, while the outcome that contradicts it can move the price a long way as those positions are closed. That mechanism is symmetrical and it accounts for a good share of the moves that look backwards, including the case in the opening paragraph. Positioning is treated in full in the last lesson of this module.

Two structural points close the list. Several releases frequently publish at the same instant, so attributing a move to one of them is an assumption rather than an observation, and the assumption is usually made in the same sentence that reports the move. And an exchange rate operating under a peg does not express a surprise at all: the arrangement fixes the rate, so the adjustment appears instead in local money market rates and in the policy rate the central bank sets, which is the structure the earlier lesson on the regional central banks described.

The minutes after a release 

The market in the moments after a scheduled release is not the market that existed in the moments before it. A large amount of repricing is concentrated into a very short window, liquidity around the current price thins, quoted spreads widen, and the price at which an instruction executes can sit a meaningful distance from the price that was displayed when it was sent. The initial move is also frequently unwound within minutes as the detail beneath the headline is read, which is why practitioners disagree about whether the first print of a reaction or the level it settles at an hour later is the reaction at all.

Volatility around scheduled events is treated in this curriculum as a risk topic rather than as a schedule of opportunity, and it has a lesson of its own later in this module. The longer treatment of a single release, its components, its publication schedule and the vocabulary its statistics office uses, sits in the guide to the US consumer price index.

Where practitioners disagree 

The first disagreement is about the standardised surprise itself. Aggregated across many releases it produces the economic surprise indices carried by most data vendors, and one tradition treats those as a compact description of whether an economy is running above or below what forecasters had assumed. The objection is that the divisor is a choice, that the aggregation mixes releases of very unequal consequence, and that a rising index therefore describes a forecasting panel revising itself upward rather than anything about an economy or a price. Both sides agree the index measures forecasters. They disagree about whether measuring forecasters is informative.

The second is whether priced in is a claim at all, for the reason set out in the caveat above, and the argument does not resolve because the sceptical position and the ordinary usage are both defensible descriptions of the same untestable sentence. The third is the benchmark question. One camp holds that the survey consensus is the right reference, because it is published, visible to every participant at once, and the number every commentary reports the release against. The other holds that a survey closes days early and carries no money behind it, while an expectation implied by prices is a position somebody is actually holding. The two do not converge, because they answer different questions: what forecasters said, and what positions imply.