How the economic calendar works

What a calendar actually is 

An economic calendar is a timetable, and almost everything interesting about it follows from the fact that the items on it are scheduled rather than announced. A statistics agency does not decide on the morning of publication whether to publish. The date is fixed months ahead, the hour is fixed by the agency's own release policy, and the same series is compiled on the same basis period after period. That predictability is why the calendar is a market document rather than an administrative one: an announcement whose timing is known is one around which quoting behaviour, staffing and liquidity provision can be arranged, by every participant at once.

Two quite different kinds of entry share the same grid. The first is a statistical release: a measurement of something that has already happened, compiled by a national statistics office or a central bank's research arm, describing a month or a quarter that has already closed. The second is a policy decision: a choice about the period ahead, taken by a committee, published at a scheduled moment together with the reasoning behind it. A release reports. A decision acts. Reading both off one grid is convenient, and it is also the most common source of confusion about what a calendar row is saying.

Who publishes the numbers 

The publisher is not a detail on the row. It determines the sampling method, the revision policy, the definition of the thing counted and whether the figure is an official statistic at all. Four kinds of publisher account for most of a calendar.

• National statistical offices. Government agencies with a statutory mandate: the Bureau of Labor Statistics and the Census Bureau in the United States, Eurostat and the national institutes that feed it in the euro area, the Office for National Statistics in the United Kingdom. Their output is an official statistic, produced under a published methodology, and typically revised on a published schedule.
• Central banks. Rate decisions, minutes, projection rounds, lending surveys and balance sheet data. A central bank is simultaneously a publisher of statistics and the actor whose decisions those statistics inform, which is why its calendar entries carry a different character from an agency's.
• Industry and private survey bodies. Purchasing managers' surveys, sentiment indices, private payroll estimates. These are not official statistics. They are surveys, they sample a panel rather than a population, and their sampling frames and weighting schemes are commercial property rather than published statute.
• Sector regulators and agencies. Energy inventory reports, agricultural stocks, trade licensing data. Narrow in scope, frequently the only regular public measurement of the thing they count.

The distinction between an official statistic and a private survey is worth holding on to. A private survey can be timelier and is usually more volatile, because a panel of a few hundred respondents carries sampling noise a census does not. Neither property makes one better than the other, and a calendar that lists them in one column flattens the difference away.

Previous, consensus, actual 

Almost every calendar renders a release as three numbers. Previous is the reading published last period, and it is frequently not the reading that was published last period, because revisions may have changed it in the meantime. Consensus is a forecast, assembled before the release by polling economists at banks and research houses and taking a central value of their submissions. Actual is the figure the agency publishes at the scheduled moment.

Consensus deserves more scepticism than its single-number presentation invites. The poll behind it is run by a news agency or data vendor, its panel differs between vendors, submissions close several days before the release, and the central value may be a median or a mean by the vendor's own convention. Two calendars can print different consensus figures for the same release without either being wrong, and the dispersion of the forecasts, which is the more informative statistic, is almost never shown.

The surprise convention, and what it assumes 

The dominant convention treats the deviation of actual from consensus, rather than the level of the actual figure, as the informative part of a release. The reasoning is that the level was already published in advance in the form of the consensus, so a figure that arrives exactly on consensus has told a market nothing it did not already have. The deviation is what is new.

The convention is a convention, and it rests on assumptions that are not always met. It assumes the polled consensus fairly summarises what participants expect, which holds only to the extent the panel is representative and current. It assumes the release is read as a single number, when many carry components practitioners weight differently from the headline. And it assumes a stable reference point, when expectations can move between the close of the poll and the release, leaving the printed consensus describing a view already abandoned.

Importance markings are editorial 

The stars, colours and high-medium-low bands that calendars attach to entries do not come from the publishing agency. They are assigned by the calendar vendor, they differ between vendors for the same release, and they usually encode something narrower than they appear to: how much price movement has historically been observed around that release, rather than how economically significant the statistic is. A structurally important series published quarterly and revised heavily can carry a lower marking than a volatile monthly survey with no revision history at all.

Revisions, and why a number has more than one value 

Most macroeconomic statistics are published before their underlying data collection is complete, then corrected as more of it arrives. A first estimate is followed by a second and a third; annual benchmark exercises re-base whole series against a fuller source such as tax records; seasonal adjustment factors are re-estimated yearly, changing previously published months with no new raw data at all. A given month therefore has a value, and a value as at a particular date, and the two are different objects. Statisticians call the second a vintage.

This matters for anyone comparing a release against history, because the history available today is the revised history, not the history that was on the screen at the time.

Timing, embargoes and the clock 

Major agencies release under embargo. Accredited journalists are admitted to a secured room shortly before publication, given the figures, and released to transmit at the scheduled instant, so the number reaches every wire simultaneously. Some agencies have restricted or withdrawn the arrangement on the grounds that any pre-release access is a fairness problem.

Release times are set in the publisher's own local time. That has a practical consequence a calendar rarely spells out: when a publishing country shifts on and off summer time and an observer's country shifts on a different date, or not at all, the hour a familiar release lands changes for several weeks each year. The scheduled hour has not moved. The relationship between two clocks has.

The calendar also carries entries that are not scheduled at all, or only loosely so: unscheduled policy meetings, ratings actions announced after a market close, statements from officials at conferences. Vendors add these as they are announced, which means a calendar is a document that changes during the week it describes, not a fixed timetable published in advance.

What a calendar leaves out 

Three omissions are structural rather than accidental. A calendar shows a headline and hides the components, so a release whose headline lands on consensus while its internals diverge sharply appears as an uneventful row. It shows a point forecast and hides the dispersion around it, so a confidently forecast release and a widely disputed one look identical. And it shows the release but not the publication that accompanies it: for a policy decision, the statement, the projections and the press conference arrive at separate moments, and one row collapses them into one.