How to Update a Macro View When New Data Arrive

Updating a view means moving a stated position by an amount that new information justifies. Moving it that way is only possible if the position and the conditions that would change it were written down before the information arrived. Everything after that is marking: each condition comes back met, unmet or unaddressed, and a release that meets none produces a written line saying nothing changed.

Three things behind the method are set out separately and taken as given. A published figure travels some distance after it is first put out, and that distance is what shows how small a gap really is. The distance between a forecast and a print usually turns out to be small. The reading sequences for the two releases most people watch are set out separately rather than repeated.

What does it mean to update a view rather than replace one?

Updating means the sheet being written on already has something on it. A position, in one sentence. Beside it, the two or three things that would change that position, each with a number attached. When a release lands the question is not what the analyst now thinks. The question is which of the written things the release spoke to, and what it said. An update moves an existing position by a measured amount; a replacement throws the position away and writes a fresh one, then calls the result an update because the timing looks the same.

Think of a household that writes two lines on the fridge in April, before the school fee letter for the year arrives. Line one: the holiday stays booked. Line two: it comes off the calendar if the fee crosses Rs 92,000/- for the year, or if the deposit on the flat is not returned by the end of May. In June the letter arrives at Rs 88,000/-. Nothing comes off the calendar, and the household has still done something: it has read the letter against what it wrote and recorded that the plan stands. The household is running the sequence below. Nothing about the sequence is clever, and that is rather the point.

The sequence has seven steps. Each one is an action with an output that can be pointed at afterwards. Growth itself, what a price index measures, and why a figure gets restated later are covered separately and used here as inputs.

Step one: what goes on the sheet before any data arrive?

Two things go on the sheet, and both are short. The stated position, in one sentence with a number in it. Then the conditions, each naming a threshold and naming which release, and which vintageWhich round of publication a figure belongs to. The same period can be published more than once, and the value can differ between rounds without anybody having made a mistake. of it, the threshold applies to. A condition that does not name a threshold cannot be marked. A condition that does not name the vintage will be marked against the wrong number.

The sheet used for the rest of this walkthrough belongs to the Republic of Sankhya, an invented economy, and was written before any print of its growth series existed. The position: Sankhya growth is running near 6.40 per cent, and it stays there. Three conditions, all dated before the data: a vintage published after the first print reading at or above 6.75 per cent or at or below 6.05 per cent; a gap against the 6.40 per cent consensusThe figure a group of forecasters collectively stands behind before a release, usually the middle value of what they submitted. A consensus is a forecast, not a measurement. larger than the 0.60 point revision bandHow far a series is known to move between one round of publication and a later one. Covered separately; here it is a number the conditions are written against. for this series; and the Sankhya industrial volume series turning down while the growth series holds.

A condition written after the data have landed is not a condition, it is a rationalisation, and it is the one thing in the sequence that cannot be repaired later. Steps two through seven can all be done slowly, done again, or done by somebody else. Step one is the only step whose value comes entirely from when it happened. Write it on the tenth and it is a test. Write it on the twentieth, after the release, and it is an explanation, and an explanation can be made to fit whatever arrived.

Step two: how does the release itself get read?

Step two is a handover, and the shortest step here. The release is read with the reading sequence built for that release. For output that means walking the national accountsThe set of official statements that add up what a country produced, earned and spent over a period. Built and taught separately; used here only as a name. release in its own order. For prices it means separating the headlineThe single summary figure a release leads with, before the parts underneath it are separated out. Which figure counts as the headline is set by the release, not by the reader. figure from what is driving it. Both of those orders are covered separately. The step routes by name, takes the outputs those sequences produce, and stops there.

Step two adds nothing of its own. A reader who starts forming an opinion while reading the release has skipped step three entirely. The temptation is real, because reading a release feels like the moment when thinking should happen. It is not. Reading is the moment when a number is extracted correctly. Step three decides what becomes of the number, and step three has a written sheet to answer to.

Step three: what is the release compared against?

The release is compared against the conditions, and not against the position. Step three is the one most readers get wrong, and the reason is that comparing a release with a position is easy and feels like work. A print of 6.20 per cent against a position of 6.40 per cent produces a feeling of disagreement, and a feeling of disagreement is not a mark. A feeling cannot be written down as met or unmet, it cannot be checked by anybody else, and next month it will quietly have become whatever the reader thinks by then.

Marking allows three words and only three. Met, when the release crossed the threshold the condition named. Unmet, when the release spoke to the condition and fell short of the threshold. Unaddressed, when the release says nothing about the condition either way. Most releases address no condition at all. Recognising an unaddressed condition is therefore the main skill in the sequence.

Run the Sankhya sheet against the first print of 6.20 per cent and the marks come out like this. Condition one was written about a vintage published after the first print, and the first print is not one, so condition one is unaddressed. The gap to the 6.40 per cent consensus is 0.20 points and the condition asked for more than 0.60, so condition two is unmet. The growth release carries no industrial volume seriesA record built by counting quantities produced rather than by asking anybody how things are going. Which activities such a record reaches, and which sit outside its edge, is treated separately. at all, so it is silent on the question in both directions, and condition three is unaddressed. Two unaddressed and one unmet: nothing met.

Step four: how far does one release have to move to count?

Step four is one comparison with both sides stated in points. On one side, the gap between the release and the figure that stood before it. On the other side, how far this series is known to travel between one round of publication and a later one. Both of those numbers were built elsewhere and are used here as inputs: the sizes of gaps against forecasts, and the distance a figure covers after it is first published. Route to them; do not rebuild them.

For Sankhya the comparison is 0.20 points against 0.60 points. The first print reads 6.20 per cent, the consensus standing before it was 6.40 per cent, so the gap is a miss of 0.20 points. The band, first print to later vintage, is 0.60 points. The series moves three times as far on its own as the gap anybody was reacting to, so a gap of that size is not information about the economy, it is information about how early in the publication round the reading sits.

Step four is not a mark. Step four does not overrule step three, and it cannot make a condition met that step three marked unmet. The comparison is a sanity check sitting beside the marking, and its use is negative: it shows when a gap is too small to be worth anything, whichever way it points. Whether a gap can also change sign between vintages is a separate matter, treated elsewhere; step four needs only the size.

Step five: if a condition was met, what gets written down?

Two things get written down: the new stated position, and how far it moved, in points. Run the same Sankhya sheet against the later vintage of 6.80 per cent. A reading of 6.80 per cent arrived as a vintage published after the first print, so condition one is now addressed. The reading sits at or above the 6.75 per cent edge the condition named, so the condition is met. The sheet already says what to do: move the position to the reading that met the condition. The position goes from 6.40 per cent to 6.80 per cent, and the move is 0.40 points. The size of the move goes on the sheet next to the new position.

The size of the move was settled at step one, so a view that moves by less than the condition specified was never really conditioned. Quietly shrinking the move afterwards is the same fault as writing the condition afterwards. Watch for the half move. A reader whose sheet said move to the reading, who then moves a fifth of the way there because the full move feels uncomfortable, has kept the appearance of a conditioned view and thrown away the part that made it testable.

Note also what step five did not need. The gap between 6.80 per cent and the consensus is 0.40 points, still inside the 0.60 point band, so step four still reports that nothing about the gap moves anything. Condition one, met on a level at a named vintage, is what moved the position. The gap test and the level test are two different tests, written for two different things, and running the sequence keeps them apart when instinct would run them together.

Step six: what is written when nothing was met?

One line is written, and it says that nothing changed. The line names the release that was read, it names the marks, it restates the position unchanged, and it carries the date on which the reading was done. For the first print run on the Sankhya sheet the line reads: read the first print at 6.20 per cent, nothing met, two conditions unaddressed and one unmet, position unchanged at 6.40 per cent, move of 0.00 points, read on 19 August 2026.

The line saying nothing changed is a real output and not a failure. A rule that only ever fires in one direction has not been tested by anything, so a sequence that never produces that line is not a sequence at all. This is the step that gets skipped, and skipping it is invisible at the time. Nobody notices the absence of a line. A year later they notice a record in which every single release produced a change. Such a record shows somebody responding to whatever arrived, and no method at all.

The vegetable seller who decides on Monday that she switches supplier if the wholesale rate for onions crosses Rs 34/- a kilo has done step one. When Thursday's rate comes in at Rs 31/- she does not switch, and if she is running the sequence she also says out loud that she is not switching, and why. Saying it is the whole difference between a rule and a mood. The seller who says nothing on Thursday will, by the end of the month, be unable to say whether she ever had a rule.

Step seven: what are the conditions for the next release?

Step seven writes two or three conditions again, dated again, before the next release. Where a condition was met, its threshold is now stale and gets rewritten around the new position. After the Sankhya position moved to 6.80 per cent, the corridor of condition one is rewritten as 6.45 to 7.15 per cent. The half width is still 0.35 points, recentred on where the position now sits. Nothing has happened to a condition that was never addressed by anything, so it is carried across untouched.

Condition two needs a different repair, and it is worth seeing. The threshold was a gap against a consensus of 6.40 per cent, and that consensus was formed before the vintage that moved the position. A condition written against a reference figure has to be rewritten whenever the reference figure goes stale, otherwise the condition will be marked against a comparison nobody is making any more. So condition two is rewritten to say that no gap can be marked until a fresh consensus figure has been recorded on the sheet.

When does the update stop?

The update stops when four things exist on paper. Every condition carries a mark. The gap has been set beside the band, in points, once. Either a change has been stated with its size, or the absence of a change has been recorded with its date. And the conditions for the next release have been written. The four together are the stopping point, and reaching them is the whole job.

The sequence does not stop with a forecast, and the reason is that no step above produces one. Nothing in the sequence asks what the next release will show. Steps one and seven ask what would change the position. Step three asks what the release said about those things. Nowhere does the method reach for a prediction, and a reader who finishes the seven steps and then writes one has added something the sequence neither produced nor checked. The habit of writing what would change a position, rather than what will happen, is the one this whole sequence exists to build. Statisticians following Thomas Bayes call the position written before the evidence a priorIn statistics, the belief held before the new evidence is looked at. Named after Thomas Bayes. Borrowed here only as a word for what sits on the sheet at step one., and the vocabulary is worth borrowing even where none of the arithmetic is.

What does the sequence look like run twice on the same sheet?

Run it twice against the Sankhya sheet, once on the first print and once on the later vintage, with the sheet written once and never touched between the two runs. The whole point of the exercise is that step one is identical in both runs. Only the release changes, and the two runs come out in different places, so the reader is watching the method rather than the answer.

Read the table down the two columns rather than across the rows and the shape appears. Everything above the marks is identical in kind. Everything below them differs. The vintage row does the work. The same reader with the same sheet reached two different steps on the strength of which round of publication the number arrived in. Two runs from one sheet, one producing a written line saying nothing changed and one producing a 0.40 point move, is what a working method looks like from the outside.

What separates an update from a reaction?

An update names in advance what would change it. A reaction explains afterwards why it changed. The difference is the whole of it, and it is invisible at the moment the change happens. Both produce the same sentence on the same day with the same number in it. A month later the two are not remotely the same thing. Only one of them can be scored.

The difference matters most to whoever has to answer for the record. An analyst who keeps a sheet per view can be asked, at the end of a year, how many readings produced a change and how many produced a line saying nothing changed, and the ratio is a fact about their process rather than an opinion about their judgement. A credit officer reviewing a borrower can name in advance the two things in the next set of accounts that would change the internal grade, and then the grade either moved for a written reason or it did not. A tenant can write down, before the landlord names a figure, the rent at which they start looking at other flats, and be spared the conversation with themselves afterwards about whether the new figure is really so bad.

The reaction is not wrong because it is fast, it is wrong because nothing on it could have failed, and a record in which nothing could have failed teaches its author nothing at all. The tell is easy to check. Counting the entries in a set of notes where the number did not move settles it: if there are none, the notes are reactions with dates on them.