A diffusion index is a number that measures how widespread a change is across a group of components, like industries, companies, or economic sectors. Instead of telling you how much something grew or shrank in total, it tells you how many parts of the whole are moving in the same direction. This distinction matters more than it sounds: an economy where nearly every industry is growing looks very different from one where a single booming sector is masking declines everywhere else, even if the total growth number is identical.
Breadth vs. Magnitude
Traditional economic statistics report magnitude. They tell you total employment rose by 200,000 jobs, or GDP grew 2.3%. A diffusion index answers a different question: how many of the individual pieces contributed to that change? An increase in total employment caused by growth in just a few industries can be the same size as one driven by growth across dozens of industries, but the two situations carry very different implications for the health of an economy.
Think of it like a classroom vote. Knowing that a proposal passed 28 to 2 tells you something fundamentally different from knowing it passed 16 to 14, even though the outcome is the same. A diffusion index captures that breadth of agreement among the components.
How the Calculation Works
The most common version is straightforward. Each component in the index (an industry, a survey respondent, a sector) gets sorted into one of three buckets: improving, unchanged, or declining. The index equals the percentage reporting improvement, plus half the percentage reporting no change. Those reporting decline are effectively left out.
Here’s a concrete example from the ISM Purchasing Managers’ Index: if 20% of manufacturers say business got better, 70% say it stayed the same, and 10% say it got worse, the diffusion index is 20 + (0.50 × 70) = 55. The result falls on a scale from 0 to 100.
Some surveys use a simpler version where each component is assigned 100 (if it increased), 50 (unchanged), or 0 (decreased), and the results are averaged. The math differs slightly but the logic is identical. Other indices, like some Federal Reserve regional surveys, center their scale at zero instead of 50, producing values that range from roughly negative 100 to positive 100. The interpretation is the same: the midpoint means an even split between expansion and contraction.
What the 50 Threshold Means
On the standard 0-to-100 scale, 50 is the dividing line. A reading of exactly 50 means the number of components expanding equals the number contracting. Above 50, growth is more widespread than decline. Below 50, more components are shrinking than growing. A reading of 100 would mean every single component improved, while 0 would mean universal decline.
The further the index moves from 50 in either direction, the more lopsided the picture. A reading of 65 doesn’t mean the economy grew by 65% of anything. It means expansion is broad-based, with a clear majority of sectors moving upward. Employment diffusion values below 50 tend to coincide with recessions, because widespread job losses across industries are a hallmark of economic downturns.
The ISM Purchasing Managers’ Index
The most widely followed diffusion index in the U.S. is the ISM Manufacturing PMI, released on the first business day of each month. It’s a composite of five equally weighted diffusion indexes covering new orders, production, employment, supplier deliveries, and inventories, each getting 20% of the total. Purchasing managers at hundreds of manufacturing firms report whether each activity improved, stayed the same, or worsened. The results are combined using the standard diffusion formula.
Because the PMI comes out quickly and surveys people with direct knowledge of business conditions, it’s treated as one of the earliest signals of where the economy is heading. A sustained move below 50 often precedes broader economic weakness, while readings consistently above 50 suggest manufacturing is expanding.
Federal Reserve Regional Surveys
Several regional Federal Reserve banks publish their own diffusion indexes based on monthly surveys of businesses in their districts. The Philadelphia Fed runs a Manufacturing Business Outlook Survey and a Nonmanufacturing Business Outlook Survey covering the Third Federal Reserve District. The Richmond Fed and other regional banks produce similar indexes for their territories. These surveys ask about general business activity, new orders, shipments, employment, and prices, then report each as a diffusion index.
Because these regional reports come out before the national ISM data, traders and economists watch them for early clues about the direction of the broader economy. A sharp drop in the Philly Fed index, for instance, often sparks discussion about whether a national slowdown is forming.
Diffusion Indexes as Recession Signals
One of the most valuable properties of a diffusion index is that it tends to turn down before the economy officially enters recession. Research by Geoffrey Moore in the 1950s found that historical diffusion indexes led business cycle peaks by about eight months. More recent work has pushed that estimate further: researchers at Columbia University found that diffusion indexes they developed would have predicted the onset of the 2007-2009 Great Recession as early as October 2006, roughly 14 months before the recession officially began in December 2007. On average, their approach provided warnings about one year ahead of time.
This leading behavior makes intuitive sense. Before a recession shows up in total GDP or employment numbers, the weakness starts appearing across more and more individual industries. The aggregate statistics might still look positive because a few strong sectors are masking the spreading decline. A diffusion index catches that erosion of breadth before the magnitude statistics reflect it.
Weighted vs. Unweighted Versions
The standard diffusion index treats every component equally. A tiny industry counts the same as a massive one. Researchers at the American Statistical Association tested whether weighting by industry size or by the amplitude of fluctuations would improve the index. The results were instructive: size weighting made little difference, because the average size of industries in the expanding and contracting groups stayed fairly constant even as the number of industries in each group fluctuated. Amplitude weighting, which gave more influence to industries with bigger swings, tended to exaggerate month-to-month movements without adding meaningful information.
This is part of why the unweighted version remains standard. The whole point of a diffusion index is to measure breadth, and weighting by size or intensity starts blurring the line between a breadth measure and a magnitude measure.
Limitations to Keep in Mind
A diffusion index deliberately ignores how large the changes are. If 80% of industries added jobs but each added only one position, while 20% of industries cut thousands of jobs each, the diffusion index would still read a healthy 80, even though total employment might have fallen. This is a feature, not a bug, since the index is designed to measure breadth. But it means you should always read a diffusion index alongside traditional measures of magnitude.
Diffusion indexes also work best for variables where breadth genuinely drives the aggregate trend. For variables where a single large player can dominate the total, like oil production in a country with one major oil field, the index adds less insight. The strength of the tool is in capturing the collective direction of many independent parts, which is why it’s most useful in diverse economies with many sectors moving somewhat independently of each other.