ISM vs S&P Global PMI: why can two surveys of the same month disagree?
ISM and S&P Global both survey US purchasing managers, but they weight the answers differently and can disagree. September 2026 showed it, and an early estimate moved EUR/USD most.
Two surveys of US purchasing managers come out every month, and in September 2026 they pointed in opposite directions. S&P Global's services index rose to 58.8, while the one from the Institute for Supply Management (ISM) fell to 54.9. In our price data, the biggest EUR/USD move came after neither survey's monthly headline but after S&P Global's early "flash" estimate of 23 September, which beat forecasts by about three points.
Why are there two PMIs for the same economy?
Because two different organizations run their own panels and publish their own indexes. Both are called a purchasing managers' index (PMI), and both read above 50 as growth and below 50 as contraction.
| Survey | Published by | Headline built from | Release time |
|---|---|---|---|
| ISM Manufacturing PMI | Institute for Supply Management (ISM) | Five indexes, equal weights | First business day of the month, 10:00 a.m. New York time |
| ISM Services PMI | Institute for Supply Management | Four indexes, equal weights | Third business day, 10:00 a.m. New York time |
| S&P Global US Manufacturing PMI | S&P Global | Five indexes, unequal weights | First business day, 9:45 a.m. New York time |
| S&P Global US Services PMI | S&P Global | One question on business activity | Third business day, 9:45 a.m. New York time |
S&P Global also publishes a flash estimate of both indexes before the month is over: September's came out on the 23rd. ISM has no preliminary reading. In UTC, the releases land at 13:45 and 14:00 until the US clocks change in November, and an hour later in winter.
How is each headline built?
In manufacturing, both surveys average the same five topics but weight them differently.
| Component | ISM | S&P Global |
|---|---|---|
| New orders | 20% | 30% |
| Production (output) | 20% | 25% |
| Employment | 20% | 20% |
| Supplier deliveries | 20% | 15% |
| Inventories (stocks of purchases) | 20% | 10% |
In services the difference is larger. ISM's Services PMI is the average of four indexes: business activity, new orders, employment and supplier deliveries. S&P Global's services headline is a single question about business activity compared with a month earlier, and S&P Global itself says it "is not comparable with the headline manufacturing PMI figure".
The panels differ too. S&P Global's services panel covers consumer services (excluding retail), transport, information, communication, finance, insurance, real estate and business services. ISM's also takes in retail, construction, utilities, agriculture, mining and public administration.
The timing differs as well. S&P Global collects its responses in the second half of the month; ISM sends its survey early in the month and says most respondents answer late in it.
What did September 2026 look like?
| September 2026 | ISM | S&P Global |
|---|---|---|
| Manufacturing headline | 54.5 (54.6 in August) | 55.9 (57.0 in the flash) |
| Services headline | 54.9 (55.4 in August) | 58.8 (56.5 in August) |
| Services business activity | 56.5 (61.7 in August) | 58.8 (56.5 in August) |
The last row is the closest to like for like, since both surveys ask about business activity. ISM's fell 5.2 points and S&P Global's rose 2.3. Both still read well above 50, so the two surveys agreed that services were growing and disagreed on whether growth was speeding up.
In manufacturing they were closer. ISM's headline barely moved, and S&P Global's final reading came in 1.1 points below its own flash.
Which one moves the dollar?
We measured the high-to-low range of EUR/USD and USD/JPY in the fifteen minutes after each release, on one-minute bars from 17 August to 16 October 2026. S&P Global publishes at 13:45 UTC and ISM at 14:00, so on the days both report, each gets its own window. The comparison is an ordinary day: the median of the 38 other weekdays in that period, Labor Day aside, measured over the same fifteen minutes.
| Release | Times measured | EUR/USD (pips) | USD/JPY (pips) | Ordinary day (EUR/USD and USD/JPY) |
|---|---|---|---|---|
| S&P Global flash | 2 | 6.0 and 17.6 | 14.0 and 30.6 | 7.8 and 13.4 |
| S&P Global final | 4 | median 8.2 | median 9.1 | 7.8 and 13.4 |
| ISM | 4 | median 9.5 | median 16.5 | 7.5 and 11.3 |
The flash of 23 September moved EUR/USD 2.3 times as much as an ordinary day, while the August flash moved it less than usual. ISM moved it a little more than usual, and the final S&P Global readings about as much as an ordinary day. The largest single USD/JPY window, 31.7 pips after the final reading of 3 September, came on a day of sharp moves in the yen, so we do not credit it to the survey.
The size of the surprise lines up with that. The September flash beat the forecasts quoted by investingLive by 2.7 points in services and 3.4 in manufacturing, while none of the four ISM releases landed more than 1.2 points from the forecasts in our roundups.
A final reading starts from a flash the market already has. In September the services final came within 0.1 point of it, and the manufacturing final came in 1.1 points lower.
For scale, payrolls moved EUR/USD between 3.6 and 5.2 times as much as an ordinary day in our comparison of the jobs numbers.
How do you read the two together?
Start with what they agree on, then look at the parts.
- Same direction, stronger signal. When both surveys rise or fall together, two different panels are telling you the same thing.
- Opposite directions, check the components. In September the services headlines split, and so did the business activity question that both surveys ask. That is a real disagreement, not a difference in formula.
- The flash is the early read. S&P Global's flash comes out about a week before ISM manufacturing, and in our sample the September flash beat forecasts by about three points. On 23 September the 10-year Treasury yield hit its highest since 2007, as our roundup of 24 September recorded.
- Read past the headline. On 1 October ISM's manufacturing index came in at 54.5, below forecasts of 54.8 to 55.0, but its prices index jumped to 77.9, and investingLive wrote that the prices number was "grabbing the market's reaction" in the first minutes.
In general, what moves price is the surprise against the forecast, not the level of the index.
What this post does not give you
It does not tell you which survey is more accurate. S&P Global has published comparisons of the two, but it is one of the two publishers, and one month of disagreement is not a pattern.
The price table rests on two flashes, four final readings and four ISM releases inside one two-month window. It shows the order of magnitude, not a rule.
Other releases in the same minute are not separated out: US job openings came with ISM manufacturing on 1 September, and construction spending on 1 October. The ordinary days are not empty either: some of them carried other US data at the same times.
The takeaway
The two PMIs answer the same question with different panels, weights and timing, so they can disagree in the same month. Watch S&P Global's flash for the first surprise, read ISM's components for the detail, and judge each against its own forecast.
Methodology from S&P Global's US Manufacturing PMI (March 2024) and US Services PMI (March 2026) news releases and from ISM's Manufacturing and Services PMI reports for September 2026; September readings from those releases as reported by investingLive, FXStreet and Action Forex; the date of the August flash from Trading Economics. Forecasts from our roundups and from investingLive; price ranges are measured on our own one-minute bars.
