Diffusion Index

A diffusion index is a summary statistic that measures how broadly a change is spread across an economy or survey sample. Rather than averaging responses or weighting by size, a diffusion index counts the fraction of respondents reporting expansion versus contraction. When diffusion is high, the entire economy (or most sectors) is moving in the same direction; when diffusion is low, the movement is narrow or mixed. This breadth metric is a powerful leading and concurrent signal of cycle strength.

The logic of counting rather than averaging

Imagine two scenarios. In the first, 80% of firms report rising sales and 20% report falling sales. In the second, 51% report rising sales and 49% report falling sales. Both have positive average growth, but they tell very different stories. The first is a broad, healthy expansion; the second is an anaemic one where half the economy is contracting.

A simple average of sales growth obscures this distinction. A diffusion index does not. By counting the fraction in each category, it reveals breadth—whether the economic movement is concentrated in a few large firms or spread across the entire sample. This matters enormously for forecasting. A broad expansion is more sustainable and less vulnerable to shocks than a narrow one led by a handful of sectors.

The diffusion index essentially asks: “What percentage of the economy is expanding?” A reading above 50 means more firms or regions are growing than shrinking; a reading below 50 means contraction is dominant. The distance from 50 signals intensity: a reading of 75 is a roaring expansion; a reading of 25 is deep contraction.

The ISM manufacturing diffusion index

The most famous diffusion index in the U.S. is published by the Institute for Supply Management (ISM). The ISM surveys purchasing managers at about 400 manufacturers monthly, asking them about new orders, production, employment, deliveries, and inventories. For each category, the manager reports whether it is “up,” “same,” or “down” compared to the prior month.

The ISM computes a diffusion index for each category:

Diffusion = (% reporting “up”) + 0.5 × (% reporting “same”)

This formula gives equal weight to the binary choice of expansion versus contraction, but rewards consensus (if everyone says “same,” diffusion stays near 50 and signals stability). The result ranges from 0 to 100.

An ISM diffusion index above 50 indicates manufacturing expansion; below 50 indicates contraction. Values above 60 suggest strong momentum; values below 40 suggest recession-level weakness. Because the ISM surveys are released early in the month (usually the first business day after month-end), they provide a fast read on industrial momentum and are watched closely by Federal Reserve officials, traders, and forecasters.

Why diffusion matters for cycle confirmation

When the economy is turning from expansion to contraction, diffusion indices typically begin rolling over months before headline GDP turns negative. Here is the mechanism: in the early stages of a slowdown, only certain sectors weaken—perhaps construction falters while services hold steady, or technology cuts spending while financials do not. Diffusion starts to decline because the fraction of firms reporting weakness is rising.

But as the slowdown spreads, more sectors join the contraction. Diffusion falls further. By the time diffusion has dropped to the 30s or 40s, it is nearly certain that gross domestic product will turn negative in the next one to two quarters. Conversely, when diffusion has risen above 50 from deep contraction (say, from 20s to 50s), it signals that recovery is broadening and will likely persist.

The NBER Business Cycle Dating Committee does not formally incorporate diffusion indices into its official methodology, but staffers and members track them carefully. A recession call is far more credible if multiple diffusion indices are signalling broad contraction than if weakness is confined to one or two sectors.

Regional diffusion indices and sectoral disaggregation

Beyond the ISM, many regional Federal Reserve banks publish manufacturing diffusion indices. The Philadelphia Fed’s Business Outlook Survey, the Richmond Fed’s Manufacturing Index, and others ask similar questions to firms in their districts. These regional indices allow forecasters to see whether weakness (or strength) is national or geographically concentrated.

Sectoral diffusion indices also exist. You can compute a diffusion index of how many industries within the S&P 500 are in bull markets versus bear markets; or how many sectors of the housing market (detached homes, condos, apartments) are in expansion. The logic is the same: breadth signals confidence in the move.

The relationship to other cycle measures

Diffusion indices work alongside leading indicators, coincident indicators, and lagging indicators. A broad diffusion index of new orders is a leading signal (firms are placing orders based on expected demand). A diffusion index of current production is a coincident signal (factories are running or sitting idle). A diffusion index of employment (percentage of firms hiring versus laying off) is slightly lagging, because hiring and firing decisions follow output changes by a few weeks.

The ISM manufacturing diffusion index for new orders, in particular, is highly correlated with future gross domestic product growth two to three quarters ahead. When new-order diffusion falls from 70 to 45, recession forecasters sit up and take notice.

Interpretation pitfalls

One common mistake is treating a diffusion index as a level-change indicator. A reading of 45 does not mean the economy will shrink by 45%; it means 45% of respondents report expansion, implying net contraction. The magnitude of contraction depends on the depth of the declines among those shrinking and the pace of the growth among those expanding, neither of which is captured by a simple diffusion count.

Another pitfall is assuming that all firms and sectors are equally important. The ISM surveys purchasing managers, who are more sensitive to inventory and order cycles than other stakeholders might be. A purchasing manager may report “orders down” because inventory destocking, even if the firm’s underlying business is fine. Context matters; diffusion is one input, not an oracle.

Also, in globalized economies, diffusion indices based on domestic surveys may miss shocks or strength coming from abroad. A U.S. manufacturing diffusion index might be weak while European order books are surging, or vice versa.

Diffusion in non-manufacturing sectors

The ISM also publishes a non-manufacturing (services) diffusion index. Services now account for roughly 80% of U.S. gross domestic product, so the breadth of service-sector expansion and contraction is increasingly important. A services diffusion index that remains elevated even as manufacturing diffusion rolls over can signal that recession risk is limited; conversely, simultaneous weakness in both manufacturing and services diffusion is a potent recession signal.

Private firms and research boutiques also compute ad-hoc diffusion indices. The Conference Board publishes a composite economic index with diffusion components. Stock market breadth (percentage of stocks in uptrends) is a variant used by equity traders. The logic is always the same: measure the fraction of the sample in expansion, and interpret the signal in the context of recent history and other cyclical measures.

The forward-looking nature of surveys

What makes diffusion indices useful is that they capture expectations embedded in survey responses. When a purchasing manager reports “new orders up,” they are not just reporting past facts; they are implicitly asserting confidence that demand will remain strong (or they would not bother ordering). This forward-looking bias gives diffusion indices a slight lead over backward-looking coincident indicators.

This is why ISM diffusion (a survey-based metric) often turns months before official employment data (backward-looking administrative records). The manager is optimistic and hiring now; the payroll numbers will reflect that hiring weeks or months later.

See also