The Macro Investing Tool

The US added 29,000 jobs in September against the 84,000 expected; the previous two months were revised down by 60,000 between them, and unemployment ticked up to 4.2%. In the same week, the Atlanta Fed had the third quarter growing at 3.7% in real terms, the 30-year hit 5.62%, its highest since 2002, and the Nasdaq made another all-time high. The press read Friday’s number as the economy cracking under higher rates, but with output running close to 4% and hours barely moving, I read it as productivity, i.e. the same amount of work, and more, being done by the same number of people.

Start with the maths, because it’s simple. Output is what GDP measures and hours are what the jobs report measures, and when output grows and hours don’t, the difference is productivity by definition. In the second quarter, output grew 1.7% while hours worked grew 0.3%, so productivity is up 2.2% on the year, and if the Atlanta Fed is anywhere near right about this quarter then the gap has widened again, which is what growth without jobs looks like while it’s happening and what you’d expect as AI gets absorbed into the economy.

Look at where the jobs actually are, because healthcare added 17,000, construction added 11,000, and manufacturing added 9,000, while government shed 17,000 and most of the rest of the private economy was flat. So the hiring is in the things you can’t automate, i.e., looking after an ageing population and physically building the data centres and the power to run them, and it’s absent everywhere a machine can now do the work. I’d expect that pattern to persist for a long time, with low employment growth overall and healthcare, construction and a few others carrying it.

Wages tell the same story. Average hourly earnings are up 3.0% on the year while unit labor costs rose only 1.2% annualized in the second quarter, and that gap is productivity absorbing the wage bill, so businesses can pay people a bit more without passing it through to prices. The unemployment rate rising isn’t firings either, because the labor force grew by 485,000 in the month and participation rose to 61.8%, its highest since May, so people are coming back into the workforce faster than they’re being hired, and with initial claims at only 197,000 there’s no wave of layoffs sitting behind the number.

So what does this mean for bond yields?

It doesn’t change last week’s read, because the bond market prices output, not headcount. The 10-year is at 5.26%, and breakevens are unchanged at 2.36%, so the real yield is now 2.9%, and real yields sit around trend real growth over time. An economy growing 3% to 4% in real terms with core inflation at 2% should have a 10-year at 5% or a bit more, and the fact that it got there in a week where payrolls printed 29,000 tells you the market has worked out the same thing, so I think yields have to reflect that growth even though it isn’t showing up in the jobs data.

You can see the same thing at the company level. The five big hyperscalers have had flat headcount since 2021, a little over two million people between them, while their net income has gone from $231bn to $524bn, i.e., the same pattern at company scale, with the output going up and the people not.

Which brings me to the flip side, and it’s the important bit...

This is not a backdrop for more rate hikes, because a Fed can’t keep hiking into 29,000 payrolls, 3% wage growth, and core inflation running at 2% annualized over three months, whatever the long end is doing on supply, and the market has started to work that out too. The odds of a hike at the October meeting fell to 17% from 36% a week earlier, the 2-year came down from 4.93% on Tuesday to 4.80% by Friday, and December is still priced at about 75%, which I think comes out as well over time.

Remember Greenspan hiked once too. In March 1997, the Fed raised rates by 25 basis points because demand looked too strong, and then it stopped, because Greenspan suspected that productivity was running faster than the models could measure and that the inflation everyone was waiting for wasn’t coming. He was right: productivity picked up sharply in 1998, inflation kept falling with growth above 4%, and the next move, in the autumn of 1998, was a cut.

And a Fed that stops is what steepens the curve. The 45 basis point level on 2s10s, where the top of the channel met the 200-day average, went on Thursday with a close at 46, up from 20 basis points on the 21st of September, and this week the steepening came from both ends, with the 10-year rising on growth and the 2-year falling as the hikes came out. As the front end keeps coming down while the long end holds up on growth and supply, the curve should keep steepening.

The banks make money on the gap between what they pay for deposits and what they earn on Treasuries and loans, so a steeper curve is what gets them to use the roughly trillion dollars of balance sheet room they were handed earlier this year, and that’s the liquidity that funds the buildout without the Fed having to expand its balance sheet.

The piece that doesn’t fit yet is the dollar. It’s at 102.3, up three points in five weeks, and gold fell 4.5% last week largely because of it, so for now the dollar has joined the long end as a second leg squeezing financial conditions, just as oil dropped out by falling back to $90. I wanted it back under 100, and it’s gone the other way, so that’s still uncomfortable, and until the front end coming down pulls the dollar with it, that’s the thing holding global liquidity back.

Which brings me on to crypto...

Bitcoin is at about $86,000, and it held through a five-week bond sell-off that took gold down and the 30-year to a 24-year high, which tells you the marginal seller is gone. But nothing has changed in the liquidity picture since last week, i.e., the excess liquidity measure that drives it is still below zero, and the ignition for the next leg is the same steeper curve and the same weaker dollar I’ve described above, so until those move, there’s nothing in the liquidity data to chase.

Where am I wrong?

The other reading of Friday’s number is that the tightening in financial conditions since February is arriving in the economy early, rather than productivity doing the work, and one weak payroll print can’t tell the two apart. The tells are claims, which are at 197,000 and would need to rise through about 230,000, and ISM new orders, which are at 55.3 and above the headline, and neither has moved. The other risk is the long end itself, because the 30-year at 5.62% is now well above where the Treasury started buying it back, and if the November refunding doesn’t bring a bigger response, then the state’s grip on the long end is weaker than I think it is.

Into next week, I’m keeping a close eye on the ISM services print and whether the 61.7 holds, the 2-year and whether the December pricing starts to come out, 2s10s holding above 45 basis points, the dollar against 102 — and whether gold stabilizes with it, and the 30-year ahead of the November refunding.

Raoul