The money gauge turned friendly. Leave out one old statement from China and it is still shrinking.

Where we are

The market-risk light reads YELLOW. It read YELLOW on every recorded day from September 7 to September 13, and again on this morning's read. Last week's edition warned that it could flip back to RED within a fortnight, because the number holding it yellow sat one billion dollars from its line. It did not flip, and it is no longer that close.

The rule is published and simple. The light turns RED only when two things are true at once: consumer sentiment is below 55, and the world's central-bank money is classified as draining, meaning it fell over the last four weeks and also over the last twelve. Sentiment is still 51.7, from the University of Michigan survey of August 28, and the next reading is due September 25. So the first condition is still met. The second is now well away from being met: over four weeks central-bank money is up $95 billion, and over twelve weeks up $117 billion. Last week the same two numbers were plus $1 billion and minus $172 billion.

The same rule means this week's friendlier number cannot turn the light GREEN either; green also needs sentiment at 70 or above. The business cycle has not moved: SUMMER, the season where the economy grows and prices rise together, confirmed on 21 consecutive daily reads since August 24, with all three growth measures improving and all three inflation measures rising. Inflation is still the dominant risk.

What turned the gauge friendly

The central-bank money gauge adds up what four central banks hold — the US Federal Reserve, the European Central Bank, the Bank of Japan and the People's Bank of China — and counts it all in dollars. On September 7 and 8 it read neutral. By September 10 it read supportive, the friendliest of its three labels. Two separate things happened in those three days, and neither of them was a central bank putting out new money.

The first was exchange rates. On September 9 the total rose from $25.09 trillion to $25.22 trillion in a single read: $130 billion. The Fed's part did not move at all, $6.74 trillion before and after. The other three moved by what their currencies did against the dollar. The yen strengthened from 159.97 to 156.11 per dollar, so each yen was worth 2.5% more dollars, and the Bank of Japan's part rose from $4.03 trillion to $4.13 trillion — also 2.5%. The euro and the yuan did the same on a smaller scale. Converted back into their own currencies, the three foreign balance sheets were the same size as the day before, as far as the stored figures show. The exchange rates doing this are dated September 4, the latest on file.

The second was the calendar. On September 9, even after that step, the gauge still read neutral: up $128 billion over four weeks, but down $54 billion over twelve. On September 10 it read supportive: up $136 billion over four weeks and up $62 billion over twelve. The total that day was exactly the same, $25.22 trillion, and so was every one of its four parts. What moved was the thing it is compared against. The rule looks up the reading from 84 days earlier; as that date slid forward by one day, the lookup landed on a figure $116 billion lower, and the twelve-week change crossed zero. The label changed on a day the money did not.

A small disclosure in the same spirit as last week's: the weekly history tool dates this change to September 13, which is simply its sample date. The daily record says September 10. Same sampling behaviour as last week, smaller consequence.

Leave China out and it is still shrinking

This brief always reports the trend on three banks measured the same way at both ends of the window, so that a bank entering or leaving the count cannot fake a move. On that basis the Fed, the European Central Bank and the Bank of Japan together went from $17.85 trillion on June 22 to $17.74 trillion now: down $110 billion, or 0.62%, over twelve weeks. The Fed's part is flat at $6.74 trillion at both ends. The European part fell from $7.00 trillion to $6.87 trillion. The Japanese part is about level, $4.11 trillion then and $4.13 trillion now. The June 22 row is rebuilt from data as it was published at the time; every later row was recorded live.

Add China and the same twelve weeks run from $25.03 trillion to $25.22 trillion: up $190 billion. The whole difference is China's part, which rose from $7.18 trillion to $7.48 trillion. That rise is real. China's central bank grew its balance sheet in June and again in July — by 1.56% in July, to ¥50.207 trillion, a second monthly rise in a row. But it reached this gauge in two steps, on the days the statements were published, July 31 and August 15, adding about $160 billion and $120 billion on those days rather than when the money actually moved.

The most recent statement is for July, and it is 75 days old. Every Jawz reading of this total now carries that age in its own label, but the number itself will not change until the August statement is out, which is due between September 15 and October 5. Until then China's part moves only when the yuan moves against the dollar.

So the honest reading has three parts. The three banks with current figures have, on balance, taken money out over the quarter. China was adding money as of July. A weaker dollar has made all of it look bigger this month. The friendly label is the sum of those three facts, and the largest of them is 75 days old.

The price of borrowing went the other way

In the same week the money gauge turned friendly, borrowing got more expensive. The ten-year US government bond yield minus the inflation the market expects — the real cost of borrowing money for ten years — rose from 2.42% to 2.55%, the highest of the thirteen weekly readings on file. It was 2.21% on June 22. Most of this week's rise was in the borrowing cost itself rather than in expected inflation: the plain ten-year yield went from 4.77% to 4.95%, while the market's inflation forecast went from 2.35% to 2.40%.

Other market prices leaned the same way, mildly. The overall direction of financial conditions changed from easing to tightening on September 11. Expected stock-market volatility rose from 14.3 to 17.8, still classified calm. Riskier companies pay 2.70% extra to borrow, up from 2.65%; safer companies pay 0.80%. The dollar fell 0.57% on the week and is down 2.05% over twelve weeks.

This is the fifth edition in a row naming the real cost of borrowing as the one measure that keeps moving in the same direction. This week it moved faster, and it moved against the label that got the attention.

What to watch

• China's August statement, due between September 15 and October 5. On today's numbers, a move the size of July's — about $110 billion either way — would by itself take the four-week change to roughly zero or roughly double it. A fall that size would not turn the light red on its own, because the twelve-week change would still sit just above zero. Just above zero is exactly where last week's edition started. If it lands there, this brief will say so.
• The consumer survey on September 25. It is the first switch of the warning light, and one of the three growth measures holding the cycle in SUMMER.
• The real cost of borrowing at 2.55%: the one measure that has kept going the same way all quarter, set by market prices rather than by a threshold.
• The yen. This month's friendliest number came mostly from a 2.5% move in one currency. A move back would take it away again, and that would not be news about central banks either.
• Data quality is clean: every source current, nothing stale or unavailable, the collector healthy. Twelve of the thirteen weekly rows behind this edition were recorded live; the oldest is rebuilt from data as published at the time. Three instrument issues stay open and are repeated here until they close. The weekly regime history still shows no colour change since August 3, although the daily record holds six RED days from September 1 to 6. The September 3 edition dates that turn to September 2 where the daily record says September 1; the correction decision is still with Pedro. And the August 17 conditions-direction discrepancy is unchanged for a fifth week.

The frame, not the trade

The way to be wrong this week is to read 'supportive' as the world's central banks opening the taps. The week's step came from exchange rates. The label's crossing came from its comparison date sliding forward. Most of the twelve-week gain is a Chinese statement from July. Meanwhile the three banks with current figures are down over the quarter, and the real cost of borrowing just posted its highest weekly reading on file.

The general lesson is about what a change is measured in. A number reported in one currency moves when that currency moves, even if nothing underneath changed. A number measured 'over twelve weeks' moves when its start date moves, even if today did not. Both are legitimate ways to measure. Neither, by itself, tells you the thing changed.

So the question for your own book: take the last number in your records that improved — a holding's value, a return, a balance — and split it three ways. How much is the thing itself? How much is the currency you count it in? How much is the date you chose to measure from? If most of the improvement lives in the second and third, you have learned something about your ruler, not about your position.

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 friendly label taken apart into the pieces that made it.