The light turned red on a margin of 0.03%

Where we are

The market-risk light reads RED. It turned on October 14 and has held since, under the rule that needs two weekly readings in a row. The light is red when two things are true together: households are gloomy (the University of Michigan consumer survey below 55; the latest reading is 48.1, published September 25, 6.9 points below the line) and the money the world's four biggest central banks hold, counted in dollars, is shrinking. Last week's edition said one more shrinking reading would do it. The reading came.

How far past the line? The money total is $24.90 trillion, $7 billion below where it was twelve weeks ago, a change of 0.03%. Over four weeks it is down 0.49%. Down on both counts is the definition of shrinking. The rule has no tolerance band, by design: a reading 0.03% below zero counts the same as one 3% below. What it has instead is the two-readings rule and a visible margin, printed here. If next Monday's twelve-week change is back above zero, the raw reading turns 'flat'; the light would then need a second flat reading before it goes back to YELLOW.

The light has been RED before in this twelve-week window: for a week from September 1, and for a day on September 24 before the confirmation rule took effect. Before August 3 it was RED for six weeks. The difference this time is that it got there under the stricter rule, so it is a confirmed state, not a flicker. The difference the other way is the margin.

The cost and availability of borrowing reads 'neutral', heading tighter, for the third weekly reading in a row (October 5, 12 and 19). For the eight weeks before that it read 'loose'.

The money: red by seven billion dollars

Split the twelve-week change, as this brief does every week. The total fell $7 billion. The central banks' own balance sheets, each measured in its own currency, shrank by $114 billion; currency moves added $107 billion, because the European, Japanese and Chinese balance sheets are worth more in dollars than they were in late July. Neither side dominates. They nearly cancel, and the light sits on the remainder.

Over four weeks the picture is different, and worth stating plainly. The total fell $122 billion, but the banks' own books grew by $60 billion. A stronger dollar took $182 billion off the dollar value of the foreign balance sheets. So the one-month drop that keeps this week's reading below zero is an exchange-rate move; the three-month drop is the banks themselves, cushioned by the exchange rate. Both are true. The reading does not know the difference, which is why the split is printed.

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 $5.87 trillion, reading flat: down $44 billion over twelve weeks, up $3 billion over four. The Treasury's account is $883 billion and the overnight cash is zero. 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 $17.50 trillion on July 27 to $17.36 trillion, down 0.8%. The Fed is flat ($6.75 to $6.76 trillion). The Bank of Japan's leg is flat on the window ($3.94 to $3.95 trillion), with the yen at 158.3 to the dollar. The European Central Bank's leg is down from $6.81 to $6.65 trillion, and that is the whole of the three-bank drain. Last week's caveat is cleared: the ECB weekly figure arrived on October 13, describing October 9, and the leg barely moved on it ($6.64 to $6.65 trillion).

China's September statement arrived on October 14: ¥50.5 trillion, up 1.29% on the month, a new high for 2026, reversing August's 0.7% dip. Three rises in four months, all driven by the same line, lending to banks, up ¥487 billion in September after running off ¥508 billion in August. In dollars the China leg is $7.55 trillion, up $110 billion on the week as the new statement replaced the old one; that is a month of change landing on one day, not a move that happened on that day. One honesty note on method: the shrinking rule carries the latest Chinese statement at both ends of its comparison, so China's rise shows up in the level of the total, not in the twelve-week change that decides the light. One country adding money at a 2026 high while the light counts the world's money as shrinking is a tension in the method, printed here rather than hidden.

Borrowing: riskier companies pay more again

Riskier companies pay 3.23% more than the US government to borrow this week, up from 3.15% last week and 0.46 points above where it was twelve weeks ago. The reading is 'widening' for the third week running. Safer companies pay 0.84% extra, up from 0.82%, still calm. This is the first-named of the three drivers behind the 'neutral, heading tighter' reading; the other two are the money total and the real cost of borrowing.

The real cost of borrowing for ten years, the ten-year US government bond yield minus the inflation the market expects, is 2.78%: the bond at 5.10%, the inflation forecast at 2.32%. It was 2.87% last week and 2.88% the week before, so two weekly readings down from the peak, though still 0.33 points above late July. The fall is in the bond yield, not the forecast: the yield dropped from 5.22% to 5.10%, the inflation forecast from 2.35% to 2.32%.

The dollar, against a broad basket of currencies, is at 121.6, up from 121.4: up 1.84% over four weeks and 0.69% over twelve. The whole of the three-month rise came in the last month, which is the four-week money split seen from the other side. Expected stock-market volatility is 16.5, up from 15.4, calm. The dollar and exchange-rate figures are as of October 9, the latest the source has published.

The economy: the jobs vote went flat

The economy label is unchanged on paper: running hot, our label SUMMER, growth still going and prices still rising, on every reading since December 4, 81 in a row. What changed is underneath. Growth is a vote of three inputs, two of three decide, and this week only one of them says growing.

Factory output: flat, 0.0% over three months in the September figure published October 16. Jobs, read from new claims for unemployment benefits: the four-week average is 199,000, 4.6% below where it was twelve weeks earlier. The jobs input votes growing only when that gap is wider than 5%. It was 9.1% in late September, then 8.8%, then 7.8% last week; on the October 15 claims release it crossed to 4.6%, and jobs went from growing to flat. Investment, business spending on computers, equipment and software after inflation: up 16.8% on a year earlier in the April to June figure, still growing. One growing, two flat.

A one-of-three vote is not decisive, and the rule says what happens then: keep the direction of the last decisive vote, looking back up to eight readings. The last decisive vote was October 14, two of three growing, so the label keeps 'growing' and the season stays SUMMER. That is the rule as written on September 30, applied as written. It also means the label now rests on one input and a memory. It moves the other way only when two of the three say stalling (for jobs, claims more than 5% above twelve weeks earlier; they are 4.6% below) and two readings in a row agree. Last week's edition named the jobs margin as the one to watch. It closed.

Prices: the Fed's preferred inflation measure, excluding food and energy, is 3.0% over a year in the August figure, unchanged in the October 14 reading. The direction still reads rising: the shorter trends in that measure and in the consumer-price version both point up, and the market's five-year inflation forecast, which pointed down last week, now reads stable. Inflation remains the dominant risk.

The AI build-out, which is most of what the investment input measures, reads 'booming' on the separate scoreboard: data-centre construction at $85 billion a year in August, chip imports at $5.5 billion in August, the four biggest builders' spending on equipment and buildings at $165 billion in April to June, and utility output (electricity and gas) up 8.7% on a year earlier in the September figure, from 6.4% in August. The next investment figure arrives with the next quarterly GDP release. It is the input the economy label now leans on.

What to watch

• Next Monday's money reading. The deciding number is 0.03% below zero over twelve weeks, and the dollar alone moved the total by $182 billion in the last four weeks. A reading back above zero does not turn the light YELLOW by itself; it starts the two-reading count the other way.
• What riskier companies pay to borrow: 3.23% extra, widening for three weeks and the first-named driver of 'heading tighter'. A fourth week up would be the longest run in this window.
• The growth vote. Both factory output and jobs would have to say stalling, two readings running, to move the economy label; neither does yet, and both say flat. The next claims release is Thursday.
• The consumer survey's next final reading, October 30. At 48.1 it is 6.9 points below the 55 line; a reading above 55 would make the mood input 'fine', and the light changes when two readings in a row agree.
• China's October statement, due between November 15 and December 5.
• Data quality: no source stale, the collector healthy (last refresh October 18), all thirteen weekly rows in each history recorded live. The dollar and exchange-rate figures are as of October 9. The China statement is 48 days old and counts as current.

The frame, not the trade

A red light is a summary of two numbers, and this week both read bad by the rules written before the fact. One of them reads bad by $7 billion out of $24.9 trillion. The rules have no 'barely' setting, on purpose: a tolerance band would be a judgment dressed up as data. So the honest reading is both things at once. The light is RED, confirmed, and the margin is about as thin as it could be.

The useful question is the one last week's edition asked you to prepare for. The light went red mostly on a currency move, with the central banks themselves adding money over the last month. If you wrote down what you would do if it turned red on the exchange rate rather than on a balance-sheet cut, this is the first case. The second question belongs to the economy paragraph: which holdings in your book assume the jobs market keeps improving? That input stopped saying so this week, quietly, while the label stayed the same.

None of this is advice and none of it is a forecast. It is a week's data, the reasoning applied to it, and the rule that decides next week printed in advance.