Opinion

Never Average Your Leading Indicators Into One Number

The worst thing you can do with a panel of leading indicators is average it into a single mushy verdict. The agreements and the disagreements are the information. Read the series individually, and today's panel is telling a specific story about how aligned, or divided, the cycle really is.

Opinion | By Emmett Cole, Macro & the Cycle. The argument here is the columnist's own; every figure links to the live series behind it, and opinion is not measurement.

The instinct when reading the manufacturing cycle is to gather the leading indicators, mash them into one composite, and announce a verdict. It feels rigorous. It is the opposite. Collapsing a panel to a single number throws away the most valuable thing on the board, which is the pattern of agreement and disagreement among the series. Of the 5 leaders I weight, new orders, truck tonnage, weekly hours, the regional Fed composite, and capacity utilization, 4 are rising, 0 are falling, and 1 is roughly flat. That distribution is the signal. The average of it is noise wearing a decimal point.

What today's distribution actually says

Read voice by voice, the panel is currently in unusual near-agreement, mostly pointing up with none in outright retreat. New orders sit at $657B (Jun 2026, up about 7.4% from a year ago); truck tonnage at 114.30 index (2015=100) (up about 0.8% from a year ago); average weekly hours at 41.7 hours/week (up about 1.5% from a year ago); the regional Fed composite at 19.43 index (Jul 1, 2026); and capacity utilization at 75.56% of capacity. When a panel lines up this tightly in one direction, that near-consensus is itself a signal, and a more reliable one than any single series, because it is hard to dismiss four aligned indicators as noise. A lopsided panel is the closest thing the cycle offers to a clear read. It is also, notably, not what you would infer from a doom-heavy headline or a single soft print.

The levels disagree even when the directions agree

Here is where averaging really costs you, and where the long lens earns its keep. Direction is only half of each series; the other half is level, and the levels are not aligned at all. Measured against its full 34-year archive, new orders sit at the top of its 34-year range, near the strong end of the range. Over the same window, capacity utilization sits mid-range over the 54-year archive, near the soft end. A panel where demand is booking at a multi-year high while the plants filling it run with multi-year slack is not a contradiction to average away, it is the single most interesting fact on the board: orders strong, throughput loose. A blended composite would net those two into a bland "fine" and delete the one tension worth thinking about.

The contradiction an average would have deleted

Now run those two series across the whole archive instead of the current print, and you get the single most interesting fact in American manufacturing right now. New orders are up 169% since 1992. Capacity utilization over the same years is down 13%. Demand went up. The share of capacity in use did not. There is only one arithmetic explanation: capacity grew at least as fast as the orders did. American manufacturers spent six years building the ability to make more things, and they succeeded, and the reward for succeeding is that none of them can raise a price.

A composite index would have blended those two into a serene middling number and told you the sector is fine. It IS fine, in the sense that the work is there. It is also structurally unable to price, which is a completely different problem with a completely different response, and averaging is what hides the difference. This is why I will not collapse a panel. The disagreement between orders and utilization is not noise to be smoothed. It is the finding.

The real payoff comes when they split

Panels do not stay aligned forever, and the method earns its keep when they diverge. When new orders and freight part ways, it means something specific: orders booking while freight softens can signal demand written but not yet shipped, an early-expansion pattern, or shipments running down a backlog while new demand fades, an early-contraction one. Same two series, opposite readings, and the tie-breaker is whichever moves next. An averaged composite would blend that contradiction into a gray number and hide exactly the question worth asking. Reading the series individually keeps the question visible, which is the whole point.

The average of a panel tells you where the middle is. The spread tells you what is about to happen. Only one of those is worth acting on.

The arbiter when the vote is close

When the leaders split and refuse to resolve, I lean hardest on the inventories-to-sales ratio, currently 1.47 ratio (May 2026), down about 6.4% from a year ago, because it sits downstream of the argument and shows which side is winning. If goods are piling up faster than they sell, the optimistic reading of strong orders is on borrowed time. If the ratio is behaving, soft sentiment is more likely noise than signal. Right now the referee is not raising a flag: on its full 34-year archive the ratio is the lowest since January 2021, and it has come DOWN over the past four years rather than building, which is the benign direction. That tilts the tie toward the orders. It does not settle every debate, but it is the closest thing the panel has to a referee. Read the series individually, respect a lopsided vote, and let the inventory ratio break the ties. That is a cycle read you can defend, and it is the opposite of averaging.

See the leading series individually, not blended, on the live data and signals pages. Read the full panel

Published 2026-08-06.