At 9:45 this morning, S&P Global publishes its flash Purchasing Managers' Index readings for August. By ten o'clock you'll see a headline that says something like "US business activity rises to 54" or "growth cools to 52," and roughly nobody reading it will know what the number means.
I want to fix that, because the PMI is one of the most useful indicators available to a private investor and one of the most consistently misread.
Here's the thing almost everybody gets wrong. A PMI reading of 53.6 does not mean the economy grew 3.6 percent. It does not mean anything grew by any percentage. The number isn't measuring size at all.
A PMI is a diffusion index. Every month, S&P Global surveys purchasing managers at roughly 650 manufacturers and 500 service providers. It asks a simple question about each business line: compared to last month, is this better, the same, or worse? Then it takes the percentage answering better, adds half the percentage answering the same, and prints the result.
Fifty means the responses were evenly split. Above fifty means more firms said things improved than said they got worse. Below fifty means the reverse.
That's it. It is a vote count.
Which means the index measures breadth, not magnitude. A reading of 53.6 tells you that expansion was reasonably widespread. It tells you nothing about how big that expansion was in dollars. Fifty-three point six is entirely compatible with an economy adding a great deal of output, and it's entirely compatible with an economy barely moving, as long as the direction was consistent across a lot of firms.
And here's the part that trips up even sophisticated readers: every respondent's vote counts the same. A company with eleven employees carries exactly the same weight as a company with forty thousand. There's no dollar weighting anywhere in the calculation.
That sounds like a flaw. It's actually the source of the indicator's greatest strength and its most important limitation, and you need to hold both.
The strength is speed and sensitivity. Small firms turn first. They have thinner buffers, less pricing power, and no ability to absorb a demand shock for two quarters while they wait for conditions to improve. When orders start slowing, a small manufacturer feels it in weeks. A large one feels it after the backlog clears. By counting them equally, a diffusion index picks up direction changes long before dollar-weighted measures like GDP register anything.
Then there's the timing, which is the real reason these numbers move markets. The July flash survey was conducted between July 9th and July 23rd and published on July 24th. One day of lag. Gross domestic product covering that same period gets its first estimate months later and then gets revised twice. Payroll data covering July arrived in August. The PMI is about as close to real time as macroeconomic data gets, and in markets, being early is most of the game.
The limitation is the flip side of the same coin. Because votes aren't dollar weighted, a diffusion index can look strong while the actual economy is being carried by a handful of enormous firms, or look weak while a few giants are doing fine. It answers "how many," never "how much." Pair it with something that measures dollars, and it becomes far more useful than either measure alone.
There's a second source of confusion worth clearing up while we're here, because it produces genuinely contradictory headlines several times a year. There isn't one American PMI. There are two, published by different organizations, and they regularly disagree.
The one landing this morning is from S&P Global. The other comes from the Institute for Supply Management, and its manufacturing report typically arrives on the first business day of the month with services following a couple of days later. They survey different panels, weight their subcomponents differently, and use different seasonal adjustments. It is entirely normal for one to print above 50 while the other prints below it in the same month.
When that happens, the financial press usually just reports whichever one supports the story it was already writing. The professional habit is to hold both, notice when they diverge, and treat a divergence as information rather than as one of them being wrong. Sustained disagreement between the two usually means the economy is genuinely split, and the split is often between large firms and small ones, which is exactly what an unweighted vote count is built to detect.
One more piece of nuance, and it's the sort of thing that separates a careful reader from a casual one. Fifty is the survey's breakeven, but it is not the economy's breakeven. Historically, a composite reading somewhere in the high forties has corresponded to roughly flat real output, because a modern economy grows a little by default through productivity and population. So a print of 51 isn't robust growth. It's barely above stall speed. And a print of 48 isn't a recession, it's a soft patch. Anchoring to 50 as if it were the line between growth and contraction is the most common error in reading these releases.
Now look at what July actually said, because it sets up what to watch this morning.
The flash composite came in at 53.6, up 1.7 points, an eight month high. Services drove all of it, jumping 2.4 points to 53.6 against expectations closer to 51.5. Manufacturing slipped slightly to 53.8, a four month low, with production, new orders, and inventories all cooling and employment the only component moving up. S&P Global's chief business economist framed the composite as consistent with roughly two percent annualized growth for the third quarter.
Then came the honest caveat, and it's the reason I'm writing this today. The survey window included the FIFA World Cup and the USA 250 bicentennial celebrations. Both pumped one-time spending into services during exactly the weeks being measured. The economist who publishes the number said so plainly.
That is the single most important discipline in reading these surveys: the commentary is worth more than the headline. The number is a summary statistic. The accompanying text tells you whether it's describing a trend or a party.
Which makes this morning's August print unusually informative. The World Cup is over. The bicentennial is over. If services holds near 53, the July strength was real. If it drops back toward 51, July was a sugar high and the underlying trend never changed. One number, and it resolves a genuine question about the second half of the year.
The subindexes matter more than the headline anyway, and this is where you can actually get an edge over the coverage.
New orders is the forward-looking one. Output tells you what's being produced right now, which reflects decisions made weeks ago. New orders tells you what's been ordered but not yet produced. When output is holding up while new orders are rolling over, a firm is working through backlog and the slowdown is already baked. That divergence showed up in July manufacturing, and it's the sort of thing that never makes a headline.
Prices charged is the one the Fed is watching. In July, input cost inflation hit a fourteen month high and selling price inflation reached its steepest pace since August 2022. That combination is the reason a September hike is still a live question despite an inflation rate that ticked down to 3.4 percent. Consumer prices are a lagging record of what already happened. Prices charged in a business survey is what firms are doing right now, this month, before it reaches any consumer index.
Employment is the one that reaches your household. A firm that's still expanding output but has stopped adding people is a firm that isn't confident. The employment subindex has turned ahead of payroll data in most recent cycles.
Supplier delivery times is the odd one. Longer delivery times normally signal strength, because it means suppliers are swamped. But it can also signal disruption, and with energy and shipping affected by the Middle East this year, that reading needs context rather than a reflex.
Here's where I want to take this, because there's a lesson in the methodology that applies directly to your own finances.
Most people evaluate their financial situation with a single number. Net worth. Account balance. This month's income. One number, checked occasionally, usually when something prompts anxiety.
That's a magnitude measure, and magnitude measures are noisy. Your net worth moved four percent last month. Was that you, or was that the market? A single number can't tell you.
Diffusion can. Build yourself the household version of what S&P Global builds for the economy: instead of tracking one figure, track direction across a set of lines and count how many are improving.
Pick eight. Take-home income. Fixed monthly obligations. Variable spending. Total debt balance. Cash buffer measured in months of expenses. Monthly investment contributions. Insurance and premium costs. Any income outside your primary job. Once a month, mark each one better, same, or worse than the prior month. Then run the same arithmetic the professionals do: percentage better, plus half the percentage unchanged.
Six of eight improving, one flat, one worse, and you're at 81. That's an expansion, and a broad one. Three improving, two flat, three worse, and you're at 50. Flat, regardless of what your account balance did, because a market rally can lift your net worth in a month when everything you actually control is deteriorating.
That's the entire point. Breadth catches deterioration that magnitude hides. A bonus can mask five months of drift. A good quarter in the market can paper over rising fixed costs, a shrinking buffer, and creeping debt, right up until it can't. Counting direction across categories surfaces the drift while it's still small enough to fix.
The reason people don't do this isn't that it's hard. It's that gathering eight numbers by hand every month means logging into six places, and by month three most people quit.
Which is a workflow problem, not a discipline problem, and workflow problems have workflow solutions. Make.com is an automation platform that connects the services you already use and moves data between them without any code. A scenario that pulls your balances into a single sheet on the first of every month, timestamps them, and computes the month-over-month direction takes about an hour to build once and then runs forever. The free tier handles this comfortably. You show up, look at eight arrows, and the score calculates itself.
Build it once and you'll never wonder how you're actually doing again. You'll know, the same way an analyst knows: not from one number, but from how many things are pointing the right way.
The August print lands at 9:45 this morning. Read the commentary, check new orders against output, and look at prices charged before you look at the headline. That's the professional read, and it takes about four minutes.
I put together the full framework as a one-page dashboard. It includes the eight household categories with definitions so you're measuring the same thing each month, the diffusion calculation with worked examples, a scoring band that tells you what different readings should trigger, and the PMI subindex checklist for reading the official release like an analyst instead of a headline reader.
Reply to this email with the word BREADTH and I'll send you The Breadth Dashboard.
Own the system,
Taylor Voss
Money Systems Lab
DISCLOSURE: This newsletter is educational content and does not constitute financial, investment, tax, or legal advice. Money Systems Lab is not a registered investment advisor. All figures cited are drawn from public sources as of the date of publication and are subject to change. This issue contains affiliate links, which means Money Systems Lab may earn a commission at no additional cost to you if you open an account through them. We only reference tools we would use ourselves. Always do your own research and consult a qualified professional before making financial decisions.