PMI explained: how purchasing managers’ indexes work

A purchasing managers' index (PMI) is a monthly survey-based gauge of business activity. Purchasing and supply executives at a panel of companies are asked a simple question about each part of their business: compared with last month, is it higher, the same or lower? The answers are turned into a number on a 0-to-100 scale where 50 separates expansion from contraction.

PMIs are among the first economic data published for any month, arriving within days of its end, long before official figures on output or orders. They are produced for most major economies, cover manufacturing and services separately, and are watched by investors as an early read on the business cycle.

The catch is in how they are built. A PMI measures how widespread an improvement is, not how large. This guide explains the diffusion-index arithmetic, what goes into a headline manufacturing PMI, how the two main US PMIs differ, and how to read them against hard data. ISM and S&P Global PMI data are proprietary, so this page explains their construction without reproducing their readings.

The diffusion index

Every PMI is built from diffusion indexes, one per question. Each respondent answers only "higher", "the same" or "lower"; nobody reports a number.

Reading the result:

• 50: the share reporting improvement equals the share reporting decline. If all 300 firms said "same", the index would also be 50.
• Above 50: more firms improving than worsening. 100 would mean every firm reported "higher".
• Below 50: more firms worsening. 0 would mean every firm reported "lower".
• The change in the index says whether the balance shifted. A move from 55 to 52 means expansion became less widespread, not that activity fell.

Breadth, not size

A diffusion index counts firms, not dollars. That is its strength (it is simple to answer and to aggregate) and its main weakness.

Real surveys rarely look that lopsided, and across many firms breadth and size tend to move together. But the example is why a reading of 59 does not mean orders grew 9%, and why a PMI just above or below 50 tells you little about the size of the change.

From sub-indexes to a headline PMI

A manufacturing survey asks the higher/same/lower question about several parts of the business and publishes a diffusion index for each. The headline PMI is a weighted average of five of them:

The supplier-deliveries index is inverted: slower deliveries push the PMI up. The logic is that suppliers usually fall behind when demand is strong and their order books are full. The inversion can mislead when deliveries slow for reasons unrelated to demand, such as port closures, strikes or shortages: the PMI then rises even though conditions are worse.

That is why analysts read the components, not just the headline. New orders are often watched as the most forward-looking piece: orders received now become production later. Many surveys also publish indexes not in the headline, such as prices paid, backlogs, export orders and imports. The prices index is followed as an early signal of cost pressure on its way to producer and consumer prices; see Producer Price Index (PPI): final demand, pipeline prices and CPI.

ISM and S&P Global: the two US PMIs

Two organizations publish widely followed PMIs for the United States.

The Institute for Supply Management (ISM), a US association of supply-management professionals, publishes the monthly Report On Business, with a Manufacturing PMI and a separate Services PMI. Its manufacturing PMI is a composite of the five diffusion indexes above with equal weights, 20% each. ISM's reports usually come out early in the following month, the manufacturing report typically on the first business day.

S&P Global (formerly IHS Markit) publishes PMIs for the US and dozens of other countries using a common method, which makes them the standard source for international comparisons. Its US manufacturing PMI uses the same five concepts but different weights: a 2018 IHS Markit research note on the two indexes gives its output index a weight of 0.25 in the composite, against 0.20 for ISM's production index. S&P Global also publishes an early "flash" estimate based on most of the month's responses before its final release.

The two indexes usually tell a similar story, but not always. The same 2018 note documented a stretch beginning in early 2017 when the ISM index ran persistently above the IHS Markit index. When two surveys of the same sector disagree, the gap itself is a reason for caution.

How markets read a PMI

Because PMIs arrive first and are rarely revised much, markets use them as an early check on the month's hard data. A few common readings:

• The 50 line and the trend. A move from below 50 to above it, or the reverse, gets attention, but so does a run of declines that is still above 50.
• New orders minus inventories. When orders run ahead of stocks, firms may need to raise production; analysts often watch the gap.
• Prices paid, for inflation pressure.
• Services versus manufacturing. Services make up most of the US economy, so a weak manufacturing PMI with a firm services PMI usually points to a sector slowdown, not an economy-wide one.

Strong PMIs tend to lift cyclical stocks such as industrials and materials, and push Treasury yields and commodity prices such as copper up; weak ones do the reverse. The size of the reaction depends on the surprise relative to forecasts, as discussed in How markets react to economic data: surprises, regimes, revisions, and on which risk, slowdown or inflation, is worrying investors at the time. Sector patterns over the cycle are covered in How sectors behave across the business cycle: cyclicals vs. defensives.

Checking a PMI against hard data

The closest hard-data counterpart for manufacturing is the Federal Reserve's industrial production index, which measures actual output of factories, mines and utilities each month. The PMI's production and output sub-indexes are designed to track it, its new orders index to track manufacturers' orders, and its employment index to track manufacturing jobs.

Whatever a PMI says about a month, data like these are what eventually settle the question. The production index and the capacity utilization rate that comes with it are explained in Industrial production and capacity utilization: the Fed’s G.17.

Pitfalls and limitations

• Breadth, not size. A diffusion index above 50 can coexist with falling activity, and a reading of 52 is not "2% growth".
• Inverted supplier deliveries. Supply disruptions slow deliveries and raise the PMI, making a supply shock look like strong demand.
• Survey noise. Each panel is a limited sample of firms; month-to-month moves of a point or two are often noise.
• Sector weight. Manufacturing is a minority of US output. A manufacturing PMI below 50 does not by itself mean the economy is shrinking.
• Thresholds differ. The level of a PMI consistent with overall economic growth is not exactly 50, and publishers' own estimates of it change over time.
• Proprietary data. ISM and S&P Global own their indexes and license them; historical series are not freely redistributable.
• Different publishers, different answers. ISM and S&P Global weight components differently and survey different panels, so their readings for the same month can differ.

PMIs compared with other business indicators

Surveys of households are the subject of Consumer sentiment and confidence: how the surveys work; the idea of soft data leading hard data is covered in Leading, lagging and coincident indicators: what turns first.

How Finabase does it

Finabase does not publish PMI readings or an economic calendar; the PMI data belong to ISM and S&P Global. The diagram on this page is ours and uses illustrative numbers.

The app does show the markets that PMIs tend to move. The Industrials and Materials sector pages list those sectors' S&P 500 stocks with their sector indexes, and the Sector page compares all 11 sectors' returns, a quick way to see whether cyclical sectors outperformed on a PMI day. Commodity futures, including copper and crude oil, are on the Futures page, and Treasury yields are on the Treasury page. The Discover page's Economy topic gathers web articles that report each month's PMIs.

Frequently asked questions

What does a PMI above 50 mean?

It means more surveyed firms reported an improvement than a decline compared with the previous month. It signals that expansion is widespread, but not how fast activity grew.

How is the PMI calculated?

Each sub-index is the percentage of firms reporting "higher" plus half the percentage reporting "the same". The ISM manufacturing PMI averages five sub-indexes, new orders, production, employment, supplier deliveries and inventories, with equal weights.

What is the difference between the ISM PMI and the S&P Global PMI?

They survey different panels and weight the components differently: ISM gives each of its five sub-indexes 20%, while S&P Global gives output, for example, 25%. S&P Global also publishes an early flash estimate and comparable PMIs for many other countries.

Why do slower supplier deliveries raise the PMI?

Because suppliers usually slow down when demand is strong, the survey counts slower deliveries as a positive sign. That convention can mislead during supply disruptions, when deliveries slow even though demand is not stronger.

Is the PMI a leading indicator?

It is an early indicator: it arrives before the hard data for the same month, and its new orders component tends to lead production. It is still a survey, so its signals need confirming by measured output and orders.

Summary

A purchasing managers' index turns simple higher/same/lower answers from company purchasing executives into diffusion indexes, where 50 marks an even split between firms improving and worsening. A headline manufacturing PMI combines five of them: ISM weights new orders, production, employment, supplier deliveries (inverted) and inventories equally, while S&P Global's version weights them differently. PMIs are early, rarely revised and widely followed, but they measure the breadth of change rather than its size, so a reading of 59 is not 9% growth and a supply shock can masquerade as strength. Read the components, compare publishers, and check the signal against hard data such as industrial production.