We changed how we read the economy. The market's real move this week was somewhere else.

Where we are

The market-risk light reads YELLOW, as it has since September 25. It turns RED only when two things are true together: households are gloomy (the University of Michigan consumer survey below 55) and the money the world's four biggest central banks hold, counted in dollars, is shrinking. The first is true, at 48.1. The second is not, but it is close: the money total is already down on the month (−0.67% over four weeks), and over twelve weeks it is up just 0.52%. If the twelve-week figure also turns negative, the reading becomes 'shrinking', and two weekly readings like that turn the light red.

Last week's reading leaned toward 'growing'. It did not get the second reading it needed: this Monday's came in flat, so the light holds the 'flat' it has held for three weeks.

The economy label changed, and this edition has to say plainly why. Last week we published 'growth stalling, prices rising' (label FALL). Today it reads 'economy running hot' (label SUMMER). That is our rule change of September 30, not the economy turning.

What changed: our rule, not the economy

Until September 29, Jawz judged whether the economy is growing from one input: factory output over the last three months. It is flat (−0.1%), so the label said growth had stalled.

From September 30, growth is a vote of three inputs, and two of three decide. Factory output: flat. Jobs, read from new claims for unemployment benefits: the four-week average is 200,000, 8.8% lower than twelve weeks ago, so jobs vote growing. Investment, read from business spending on computers, equipment and software, after inflation: up 16.8% on a year earlier (April to June), so it votes growing. Two of three say growing, prices are still rising, and that combination is 'running hot'. Recomputed under the new rule, the label would have read 'running hot' all along.

Why the change: factories alone no longer describe this economy. The money being spent is going into the AI build-out — data-centre construction is running at $85 billion a year, up from $73 billion in June's figure, and the four biggest builders (Microsoft, Alphabet, Meta, Amazon) spent $165 billion on equipment and buildings in April to June, up from $119 billion at the end of last year. A rule that ignored that was measuring the wrong thing.

What would turn it back: investment falling below where it was a year ago, joined by one other input turning down. The rule's blind spot is the other side of the same choice — factories and jobs could both go flat and the label would hold while investment keeps booming. So watch the investment number itself. It is still booming, but it is slowing: +21.1% a year at the end of 2025, +20.3% in January to March, +16.8% now.

Prices: the Fed's preferred inflation measure, excluding food and energy, eased from 3.3% to 3.0% over a year in the August figure published September 30. The tool still reads inflation's direction as rising, because the shorter-term trends in that measure and the consumer-price version both point up. Inflation remains the dominant risk.

Borrowing got riskier

Last week's brief said to watch whether riskier companies' borrowing costs would follow the rate up. They did. Riskier companies now pay 3.24% more than the US government to borrow, up from 2.80% in the reading last week's edition used — the largest weekly move in the thirteen readings on file, and enough for the tool to change that measure from 'calm' to 'widening'. Over twelve weeks it is up 0.54 points, nearly all of it in the last month. Safer companies pay 0.86% extra, up from 0.79%, still calm.

The real cost of borrowing for ten years — the ten-year US government bond yield minus the inflation the market expects — rose again, from 2.85% to 2.88%, its eighth edition in a row moving the same way. It was 2.30% on July 13. This week the bond yield (5.24%) and the inflation forecast (2.36%) both edged up.

Put together, the tool's overall reading of borrowing conditions moved from 'loose' to 'neutral', heading tighter. Expected stock-market volatility rose from 14.2 to 16.4, still calm. The pattern of the summer — money getting dearer while credit markets stayed relaxed — broke this week: credit noticed.

The money: the dollar giveth, the dollar taketh

Over twelve weeks the four-bank total rose $130 billion, to $25.00 trillion. Split it and the parts point opposite ways: the central banks' own balance sheets shrank by $74 billion, and currency moves added $204 billion by making the European, Japanese and Chinese balance sheets worth more in dollars. The currency part is the bigger one.

Over the last four weeks the same split runs the other way: the total fell $168 billion, and $157 billion of that — 93% — is the dollar strengthening, not any central bank shrinking. The balance sheets themselves moved by $11 billion. The light is near its line mostly because of the exchange rate.

The US-only measure — the Fed's balance sheet minus the Treasury's own account at the Fed minus cash parked overnight at the Fed — is $5.79 trillion, reading flat over twelve weeks, down $168 billion over that span but up $24 billion over four. The Treasury's account stands at $949 billion. This number explains the backdrop; it is not part of the light.

The trend this brief trusts most — the Fed, the European Central Bank and the Bank of Japan, measured the same way at both ends — went from $17.56 trillion on July 13 to $17.56 trillion now: exactly flat over twelve weeks. China's statement on file is for August, 65 days old; September's is due between October 16 and November 5. The Bank of Japan figure is monthly and still August's, so its weekly moves are the yen, not policy.

What to watch

• The money light's margin: up 0.52% over twelve weeks. Two weekly readings below zero turn the light RED, and on recent weeks the dollar, not the central banks, decides which side of zero it lands on.
• Whether riskier companies' borrowing costs keep widening from 3.24%, or this week was a single jump.
• Investment, the input now carrying the economy label. The next quarterly figure arrives with the GDP release; it has slowed three quarters running while staying well above the 10% line that makes it vote 'growing'.
• China's September statement, due October 16 to November 5, and the consumer survey's next final reading on October 30.
• Data quality is clean: every source current, nothing stale or unavailable, the collector healthy, all thirteen weekly rows in each history recorded live. The dollar and exchange-rate figures are as of September 25, the latest the source has published. The history keeps 'growth stalling' on the rows we published under the old rule; only the label from September 30 on uses the new one.

The frame, not the trade

Two weeks in a row now, a Jawz label moved because we changed the rule. Both times we said so first, because the way to be wrong is to read a changed label as a changed world. This week the world did change, just not where the label is: in what riskier borrowers have to pay.

The general lesson: a label is a summary of a few inputs, and choosing the inputs is a judgement. We chose to let investment vote because that is where the money is going. That same choice means the label now leans on one booming number, and we have written down what would turn it.

So the question for your own book: is there a holding whose case rests on 'the economy is fine'? Which single number would you have to see turn — jobs, factories, investment, or what risky companies pay to borrow — before you stopped believing it? If you can't name it, you might write it down now, while nothing is forcing the answer.

None of this is advice and none of it is a forecast. It is a week's data, the reasoning applied to it, and a change to our own instrument printed where you can see it.