4. The data that moves the market: why it reacts to the surprise and not to the number

Macro 3: The yield curve and the business cycle

Every investor has seen the scene: a dreadful inflation print comes out and the stock market rallies hard. An excellent employment print comes out and the market sinks. The easy conclusion is that the market is irrational. The correct conclusion is that the price today already contained the part of the number that everybody expected, and the only thing capable of moving it is the part nobody expected.

The price already contains the consensus

Before every release there is a collective forecast. The price of a bond, a currency or an index in the minute before the number lands is a summary of that forecast: whoever believed inflation would come down has already bought, and whoever believed the opposite has already sold. When the number appears, only the unanticipated part is new information, and only that part has any reason to move anything.

That collective expectation can be measured in two ways, and they do not always agree:

• The survey. Agencies collect the forecasts of dozens of economists and publish the median. That is the number the press calls the consensus.
• The prices. Futures on official rates, swaps and options tell you what real money is discounting. When the survey and the market disagree, what you have to beat in order to move the price is the market, not the survey.

What a surprise actually is

A surprise is the difference between the published figure and what was expected, but in raw form it is no use for comparing different series: missing by a tenth of a point on inflation and by fifty thousand jobs on payrolls are not surprises of the same size. That is why it gets normalized, dividing the difference by the historical standard deviation of the surprises in that same series. That way you know whether the miss was large or small for that indicator.

Out of that idea come the economic surprise indexes, which aggregate the recent deviations of a whole country. It is worth being clear about what they measure: they do not say whether the economy is doing well, but whether it is doing better or worse than analysts believed. That is why they tend to drift back to zero on their own: when the data surprises to the upside for long enough, analysts raise their forecasts and the surprise disappears even though the economy is just as strong as before.

Which numbers really move the market

Not all of them carry the same weight, and the ranking shifts with whatever the market fears at the time:

• Employment and prices. US non-farm payrolls and the CPI are, by some distance, the two that generate the most movement.
• The central bank meeting, with its statement and its projections, more than the rate change itself, which is usually already priced in.
• Activity surveys and weekly jobless claims, valued because they are fast: they arrive ahead of the definitive numbers.
• GDP, which moves little despite being the most important figure in theory, because it arrives late and has largely been deduced already from the preceding monthly data.

The same number, two opposite reactions

Here is the nuance that baffles everybody. A strong employment print is good news for corporate earnings and, at the same time, bad news for interest rates, because it pushes the central bank to keep money dear for longer. Which of the two effects dominates depends on what the market is afraid of. If it fears a recession, good economic news is good news. If it fears inflation, that same good news reads as higher rates for longer and equities fall. It is not incoherence: they are two separate channels, and which one rules changes with the context.

There is also a second act within the same session. The first move, in the seconds after the release, is executed by automated systems reading the headline. The correction comes minutes or hours later, when someone looks at the composition of the number: which sectors created the jobs, whether the previous month has been revised, which components pushed the price index up. It is not unusual for the market to turn round completely before the close.

Revisions: the number you thought you had

Almost no macro figure is final. A first estimate is published on the basis of an incomplete sample and corrected afterwards, sometimes heavily. In August 2024, the preliminary benchmark revision to US employment cut around 818,000 jobs from the estimated level of employment through March of that year. For months, the market had been reacting to a picture that turned out to be a different one.

Hence a practical rule: any system built on the latest published figure of a series that gets rewritten is operating on provisional information, and testing it backwards with the already-revised data produces results that were impossible to obtain at the time.

A real case: two inflation prints from 2022

On September 13, 2022 the US CPI for August was released. The market was expecting relief because gasoline had come down, and the core reading came in at 0.6% month on month against the 0.3% forecast. The S&P 500 shed around 4.3% on the session, its worst day in more than two years. Inflation that month, in level terms, had been slightly better than the month before.

On November 10 the same indicator came in at 7.7% year on year against the 8.0% expected. The S&P 500 rose more than 5%, its best session in more than two years. Inflation of 7.7%, almost four times the central bank target, produced the best day of the year in equities.

Two months, the same series, opposite reactions and in both cases a dreadful level of inflation. What changed was not the economy: it was the distance between the number and what the price already contained. If you keep one sentence from this level, make it this one: the market does not price the number, it prices the difference.