How to read the economic calendar

One row, read across 

A row describes one scheduled publication, and its columns run in much the same order on every calendar in use. A time, stated to the minute. An economy, shown as a flag or a currency code, identifying whose statistics these are rather than which instrument is involved. The name of the release. The period the release describes, printed beside the name and easy to miss. A rating of expected importance, usually as one, two or three marks. Then the numbers: the previous figure, a consensus, and an actual column that stays empty until the stated minute passes. Some calendars add a revised previous, a link to the primary source, or a chart of the series.

One structural fact about the object explains several of its oddities. A calendar is not published by the statistical agencies whose work appears on it. It is assembled by a data provider that gathers publication schedules from national statistics offices, central banks, industry bodies and exchanges, prints them in one place, and adds material those primary sources never produced: the importance rating and, in most cases, the consensus. Two of the columns are therefore the provider's own work rather than anybody's official figure, and that is why two calendars showing the same release can disagree with each other while neither of them is wrong. A calendar of this kind sits in the tools rail as the economic calendar.

The release, and the period it describes 

The two most confusable columns sit next to each other. The name identifies which statistic is being published. The period identifies which stretch of time that statistic measures, and it is almost always a period that has already finished. A monthly series describes a month that ended some weeks before, because the agency has to collect, clean and seasonally adjust the returns first. A quarterly series describes a quarter that ended a month or more before. The date on the row is the moment the number becomes public, not the moment the activity happened, and the two can sit a long way apart.

That gap is why the same name appears on the calendar more than once for a single period. Many statistical systems publish a series in successive estimates, each drawing on more complete returns than the last, labelled flash or preliminary, then second, then final. The qualifier beside the name is what distinguishes them, and a row read without it looks like a repeat of a number that was already published. Why the estimates differ from one another, and what the practice of restating them implies about any first figure, is the subject of a later lesson in this module.

The unit belongs to the same cluster of easily missed detail. A calendar prints one headline figure from a release that may run to dozens of pages, and the basis of that figure differs from row to row: a change on the previous month, a change on the same month a year earlier, an annualised quarterly rate, a level, a count of people, a quantity of barrels, or an amount of money. Two rows carrying the same statistic on different bases are not comparable, and the basis is stated in the column header rather than in the number itself.

The timing column 

Most statistical releases are published at a fixed local minute that has been set for years and is announced far ahead, and the discipline around that minute is one of the more rigorous things in public administration: journalists and analysts receive the material in advance under embargo in secured rooms, and it becomes public at the stated second. A calendar displays that instant in whichever timezone is selected, which introduces an arithmetic step that trips readers up twice a year.

Not every row carries a minute. Some entries are marked tentative, meaning the publication is expected within a window rather than at an instant, which is common for fiscal figures, credit rating reviews and certain central bank communications. Some are marked as running all day, which is how multi day summits, policy meetings without a fixed announcement time and speaking engagements are usually shown. A row without a minute is a row whose timing is genuinely not known, and treating it as though it were scheduled to the second misreads what the provider has printed.

One more property of the timing column is worth reading deliberately. Agencies frequently publish a bundle of series at a single timestamp, so what looks like one row is often several releases arriving together, and a filtered calendar understates what lands in that minute. The sessions module already established that the same clock time falls in a different liquidity picture depending on which centres are open, and a release minute inherits the conditions of whichever session it falls in.

Previous, consensus, actual 

The three number columns look alike and come from three entirely different places, which is the single most useful thing to hold on to about them.

• Previous is the figure the agency published for the preceding period, on the same basis as the figure about to be released. It is an official number and it is already public. It is also not necessarily fixed: agencies restate earlier periods when fuller returns arrive, so a previous value can differ from the number that was printed in that column when it was itself the actual.
• Consensus is not an official number at all. The provider surveys a panel of forecasters, typically economists at banks and research houses, and prints the median or the mean of their estimates as of a cut off before the release. It is a summary of a panel's opinion on a particular day, and nothing published by any statistical agency or central bank.
• Actual is the figure as published, appearing at the stated minute. It is the only one of the three that is new information at the moment the row completes, and on most calendars it arrives alongside any restatement of the previous period, which is why two of the three columns can change in the same second.

Because consensus is a survey, its composition matters and is rarely visible. Panels differ in size and membership, cut off dates differ, and a forecaster who revises an estimate afterwards does not change the printed figure. Two calendars can therefore show different consensus values for the identical release, and a row showing no consensus usually means the provider surveys no panel for that series rather than that nobody has a view.

A completed row contains two comparisons rather than one, and they answer different questions. The actual against the consensus measures the release against what a panel of forecasters expected. The actual against the previous measures the release against the preceding period, which is a question about the economy rather than about anybody's expectations. The two can point in opposite directions, and a reader who collapses them into a single impression of a good or a bad number has thrown away half the row.

What the impact rating claims 

The marks in the importance column are the provider's classification of a release type, assigned in advance and applied to every instance of it. The judgement behind it is editorial: this kind of publication has historically drawn a lot of attention, this kind has drawn some, this kind is followed mainly by specialists. Read as that, the column is useful, because it compresses into a glance a working knowledge of which of several hundred monthly publications the market has habitually watched.

Read as anything more, it claims things it cannot measure, and the list of them is worth being precise about.

• It does not measure the size of the surprise. The distance between the actual and the consensus is unknown until the release, and the rating was assigned long before, so it cannot describe the one variable most associated with movement.
• It does not reference an instrument. A rating attaches to a release, not to a pair, an index or a metal, and it says nothing about whether the release has any connection to a given instrument at all.
• It does not describe the conditions in the minute it lands. What is open, how thin the book is, whether other releases share the timestamp and whether a policy meeting is imminent are all outside anything the rating knows.
• It does not know what is already reflected in a price. A release can be widely anticipated and thoroughly discussed for weeks before it appears, and the rating is identical either way.
• It is static while attention rotates. Which series a market watches most closely changes with the regime: labour data can dominate in one stretch and inflation data in another, and a rating fixed in advance does not move with that rotation.

Providers also disagree with one another, visibly. The same release carries three marks on one calendar and two on another, because the classification is a house judgement rather than a measurement, and there is no authority that adjudicates between them. The practical consequence is narrow and checkable: a rating is a reasonable way to sort a long list into a shorter one, and it is not evidence about any particular instance of any particular release.

The calendar as a schedule of risk 

There is one question a calendar answers factually and completely, and it is not a question about what a release will do. It is this: which scheduled publications attached to the economies an open position is exposed to fall inside the window in which that position is open. That is a matter of record, it can be checked in a few seconds, and getting it wrong is one of the more avoidable ways of being surprised by something that was published on a fixed schedule announced a year in advance.

Working out which economies a position is exposed to is the part that takes care, because the exposure is frequently wider than the instrument's name suggests. A currency pair has two sides, so a calendar filtered to one country omits half of what is attached to it. An index is exposed to the statistics of its own economy and also to whatever moves the rate at which future earnings are discounted, which is often published somewhere else entirely. A commodity quoted in a currency carries the schedule of that currency in addition to its own inventory and production releases. And where a currency is administered against another rather than floating, the schedule that matters to it is substantially the anchor currency's schedule, which is a structure this region lives inside and which has its own lesson later in this module.

This is also where the regional institutions belong on the same footing as the large ones. The publication calendars of the Central Bank of the UAE and of the Saudi Central Bank, the regional purchasing managers surveys, the oil inventory and production releases, and the meeting schedules of the producer groups are all scheduled information in exactly the sense this lesson has been describing, and they sit on the same grid as the Federal Reserve and the European Central Bank rather than beneath them.

The most important limit of the object sits at the end of this section rather than the beginning, because it only makes sense once the rest is understood. A calendar lists scheduled information and nothing else. Unscheduled events, which include most of the ones that have moved markets furthest, are absent from it by construction: they have no publication time to print. An empty afternoon on a calendar is a statement that nothing is scheduled, and it is not a statement about how quiet that afternoon will be.

Where practitioners disagree 

The first argument is about what the consensus column represents. One tradition treats the survey median as the market's expectation, on the grounds that it is the only published, comparable, timestamped estimate available and that it is what commentary refers to afterwards. Another holds that the expectation which matters is whatever is already reflected in prices and in positioning, that this is not directly observable, and that a median of forecasts submitted before a cut off is at best a stale proxy for it. Both traditions read the same row. The disagreement is about whether a survey of forecasters and the expectation embedded in a price are the same object, and it has not resolved.

The second is about filtering. One view holds that filtering a calendar to the highest rated rows is the only way to keep a schedule legible, since a working day can carry dozens of entries across the economies a portfolio touches. Another points out that the rating is a third party's judgement made in advance, so filtering by it delegates the definition of relevance to a provider, and that the rows most often hidden are the ones attached to the second currency of a pair. Both positions are held by people who look at a calendar every day, and the disagreement is really about whether legibility or completeness is the property worth preserving.