Purchasing managers’ indices arrive before almost any hard economic data, which is why they move markets. They are also diffusion indices, which means they measure something narrower than most coverage implies — and the difference matters most precisely when the economy is turning.
What a diffusion index measures
The survey asks purchasing managers whether a given activity — new orders, production, employment, deliveries — is higher, lower or unchanged than last month. The index is calculated as the percentage reporting an increase plus half the percentage reporting no change.
A reading of 50 means the number reporting improvement equals the number reporting deterioration. Above 50 means more firms are improving than declining. Below 50 means the reverse.
Here is the critical property: the index does not measure magnitude. A firm reporting orders up 1 percent and a firm reporting orders up 40 percent both count as one “higher” response. The index measures breadth of change, not size of change.
This produces a specific failure. An economy in which a few large sectors are contracting sharply while many small ones improve marginally will produce a reading above 50 while output falls. The index says the majority is improving. It is right, and it is misleading.
What it is genuinely good at
Timeliness is the main virtue, and it is a substantial one. The survey is published within days of the month ending, weeks ahead of hard data on production or trade, and it is essentially never revised — the responses are what they were.
It is also good at identifying turning points, which is where breadth is exactly the right thing to measure. A move from 52 to 48 means the balance of firms has flipped from expanding to contracting. That is genuine information about direction, and it typically arrives earlier than the same signal in output data.
Where it is weak is in quantifying. A reading of 55 does not correspond to a particular growth rate, and attempts to map the index onto GDP growth work poorly outside the range where they were calibrated.
The sub-indices carry more than the headline
The headline is a composite. The components beneath it are more informative and are frequently ignored.
- New orders. The most forward-looking component. Orders precede production, so new orders leads the headline. The ratio of new orders to inventories is a standard early indicator — rising orders against falling stock implies production must increase.
- Supplier delivery times. The most misread component. Slower deliveries normally indicate strong demand straining capacity, and the index treats them as expansionary. But deliveries also slow during supply disruptions, when demand is weak. In that situation the component pushes the headline up while conditions deteriorate.
- Prices paid. A timely read on input cost pressure, available well before producer price data.
- Employment. Leads official payroll data, though with meaningful noise.
Manufacturing gets attention out of proportion
Manufacturing PMI is the one that moves markets, and manufacturing is a modest share of output and employment in most developed economies. Services are the substantial majority.
The disproportion has a defensible basis: manufacturing is more cyclical, so it turns earlier and more sharply, which makes it a useful leading indicator despite its size. Inventory cycles amplify swings in a way services do not experience.
But the two can diverge for extended periods, and when they do, the manufacturing reading is not describing the economy most people work in. A manufacturing index below 50 alongside services comfortably above it is a common configuration and does not indicate recession.
Soft data is not hard data
The deeper caution is that these are surveys of sentiment about activity, not measurements of activity. Respondents answer from impression, and impressions are influenced by news coverage, recent experience and mood.
The gap between survey data and hard data has at times been wide and persistent. When it opens, the resolution is not consistently in one direction — sometimes surveys correctly anticipated a turn hard data had not yet captured, and sometimes surveys reflected a pessimism that never materialised in output.
The honest position is that a divergence is a signal to look harder, not a forecast.
How to read one
- Look at the direction and the distance from 50, not the level alone. A fall from 58 to 53 is a substantial deterioration in breadth even though the reading remains expansionary.
- Read new orders before the headline.
- Check whether delivery times are flattering the composite. If the headline rose because deliveries slowed, establish why they slowed.
- Weight services by their share of the economy.
- Do not convert it into a growth forecast. It was not built to support that.
Used as an early read on direction, it is among the most useful releases available. Used as a measurement of how much the economy grew, it is the wrong instrument — a distinction that applies to most of the monthly data calendar.
