Why can good news be bad news?

A blockbuster jobs report lands — the economy added far more jobs than anyone expected. Stocks... fall. Welcome to the most confusing spectacle in markets, and the lesson that finally makes the news make sense.

The mechanics of surprise

Course 2 taught that prices carry expectations; the central-bank unit showed decision days trading on surprise. Now the general law: before every release, a consensus forecast exists — and the market has already positioned for it. The reaction runs on the GAP between the printed number and that consensus:

• Number ≈ forecast → often barely a ripple, however "good" or "bad" the level sounds.
• Number far from forecast → the surprise reprices everything touched by it.

The units of market reaction are units of surprise, not units of news.

The good-news-is-bad-news channel

Here's the twist that explains the falling stocks: strong economic data, in a period when the central bank is fighting inflation, means more pressure to keep rates high — stronger gravity ahead. The chain: hot economy → sticky inflation risk → higher-for-longer rates → heavier discount on every asset. So a "great" jobs number can sink stocks, and a soft one can spark a rally ("the Fed can ease sooner!"). Whether good news is good depends entirely on WHICH fear currently runs the market — growth fear or inflation fear. Same number, opposite meanings, different seasons.

Reading reactions like a practitioner

The professional's questions, in order: What was consensus? What printed? Which fear is in charge this season? Only then does a reaction make sense — and quite often the market's first move reverses by lunch as positioning unwinds (Course 3's after-hours caution applies to macro minutes too).

In the data

The gap this lesson runs on is two numbers on one calendar row: the actual against the estimate, with the previous reading as a third reference point. Here is one release:

Two things to know before you trust a row. The estimate is missing far more often than newcomers expect — whole categories of release carry an actual and a previous reading but no consensus at all, so surprise is undefined for them. And the change a calendar prints beside a release is measured against the previous reading, never against the estimate: it tells you how the economy moved, not how far the number missed.

Try it now

1. Translate this real headline pattern: "stocks rally on weak retail sales."
2. Compute a surprise, then count how often you cannot. Subtract the estimate from the actual on each row below — seven US releases from the week of 31 August 2026, measured 28 September 2026 — because that difference is the only part of a release the market trades.

Then count the rows with no estimate. Over the two US weeks 31 August to 13 September 2026 the calendar held 167 rows: 76 with both an actual and an estimate, and 91 with no estimate at all. Work out that share. It is larger than anybody expects, and for those releases surprise is not small: it is undefined, and a spreadsheet that treats the blank as zero will happily report it as "no surprise". 3. Check what the change rows actually measure, on the inflation release above: recompute the change once against the previous reading and once against the estimate, and see which one the printed change matches. One of them tells you how the economy moved; the other tells you how far the forecasters missed, and only the second one moves prices. 4. This season, which fear runs YOUR market — growth or inflation? What's the evidence? 5. One sentence: why is consensus, not zero, the baseline every number is judged against?