Why Services Inflation Matters More Than Headline CPI
Headline inflation tells you what happened to household cash flow; services inflation tells you what happens to the discount rate — and equity investors are paid for the second one.
Most investors have absorbed the rule that you should look at core inflation rather than headline, and stopped there. That is the right instinct applied at the wrong level of resolution. The distinction that actually matters for a portfolio is not core versus headline — it is that headline and services inflation are answers to two different questions, and the common error is not watching the wrong number but attributing the wrong consequence to the right one.
Here is the framing we use. Headline CPI is a cash-flow variable. It measures what the household actually paid, energy and food included, and therefore what happened to real disposable income, to volumes, to trade-down behaviour, to the mix inside a consumer company's revenue line. Services inflation is a discount-rate variable. It is the part of the price level that carries information about persistence, and persistence is what determines the policy path, which determines the rate at which every distant cash flow in an equity market is discounted. Confuse the two and you will read a fall in headline inflation as an easing signal when it is nothing of the kind, or read a services print as a consumer-demand signal when it is mostly a wage and lease-rollover artefact.
Two numbers, two transmission channels
The reason these separate cleanly is that they enter the equity valuation identity at different points. A move in energy or food prices is close to a tax or a rebate on the household: it changes the level of nominal spending available for everything else within weeks, and the earnings consequence shows up in volumes, category mix, and gross margin for consumer-facing businesses. It is a numerator event, and it is largely mean-reverting — the commodity complex is globally traded, supply responds, and base effects mechanically flush the shock out of the year-over-year rate twelve months later whether or not the underlying inflation problem has been resolved.
Services inflation does not behave that way, and the reason is structural rather than cyclical. The services basket is dominated by two things: labour cost and shelter. Neither is traded across borders, neither is priced daily, and neither can be arbitraged away by a tanker changing course. When a central bank asks whether inflation is embedded, it is asking about this component, because this is the component whose price-setting process has memory. That makes services inflation the input into the reaction function, and the reaction function — not the CPI print itself — is what equity duration is priced off.
Why services prices are sticky: three layers of why
It is not enough to know that services inflation is stickier. The portfolio consequence depends on why, because each layer of the mechanism implies a different lag structure and a different set of things to watch.
That last step is the whole framework in compressed form. A rate-sensitive growth compounder does not care what a household paid for petrol this quarter except insofar as it changes the terminal rate; a mass-market retailer cares enormously, and cares about the policy path only through its financing costs. Sterling's Embedded Intelligence approach to these releases is to route each component to the channel it actually acts through, rather than treating the top-line print as a single market-wide signal. Sterling's discipline here is unglamorous: decompose first, then decide which portfolio exposure the decomposition speaks to.
The illustration: goods disinflation without services relief
The 2021–2023 inflation episode is the cleanest available teaching case, and it is worth stating carefully as history rather than as a live signal. The initial impulse was heavily goods-led — supply chains, shipping capacity, durable goods demand pulled forward, then a commodity and energy shock layered on top. Because those components are globally traded and inventory-cycle driven, they also unwound first: goods prices decelerated sharply, and in several categories fell outright as supply normalised and retailers cleared stock.
Headline inflation therefore fell quickly, and a large part of that fall was mechanical — base effects rolling the prior year's energy spike out of the twelve-month comparison. Yet services inflation stayed elevated well after goods had turned, because the wage rounds and annual price schedules were still repricing off the inflation that had already happened. The lesson generalises: a falling headline rate can coexist with an unresolved inflation problem, and the composition of the disinflation tells you which one you are looking at. Disinflation you get from energy base effects is disinflation you cannot own — it does not change the terminal rate. Disinflation that shows up in wage-linked services is the kind that does.
The mirror-image trap is equally durable: headline inflation can be pushed up by an energy shock while services decelerate. That is a real-income hit to the consumer with no implication for policy persistence — arguably the opposite, since it destroys demand. Reading it as an inflation resurgence, and de-rating duration accordingly, is the same error in reverse.
The strongest case against this framework
The serious objection is not that headline is irrelevant — it is that headline is where expectations are formed. Households and wage bargainers do not observe a services ex-shelter index; they observe grocery bills and fuel prices. A sustained headline shock can therefore leak into the wage round and become tomorrow's services inflation. On that view, watching services is watching the output of a process whose input is headline, and the input has the earlier information.
We accept the mechanism and would reframe rather than dismiss it. Headline matters as a leading indicator of services persistence, not as a policy variable in its own right — and it only matters in that role when the move is large and sustained enough to enter bargaining behaviour, which is a much higher bar than a single volatile print. That reframing also gives you the test: watch whether headline shocks are showing up in wage growth and in longer-horizon inflation expectations. If they are not, the second-round channel is closed and headline volatility is noise for discount-rate purposes.
The second objection is more technical and more damaging if ignored: services inflation is partly a measurement artefact. Shelter is a lagging construction by design, and several components are imputed or administered rather than transacted. A desk that anchors on the aggregate services number is deliberately staring into a rearview mirror and may conclude inflation is embedded when the printed index is merely catching up. This is why the aggregate is not the working measure. The working measure separates components by measurement lag and price-setting mechanism: market-transacted services with short repricing cycles carry the most current information; shelter carries the most stale; insurance, medical and financial-service imputations are frequently driven by methodology rather than demand.
What this changes about how you read a release
The practical output is that a CPI release stops being one number with one market consequence and becomes a decomposition with two. First, ask what happened to household real income: that is your read on consumer volumes, mix and trade-down over the coming quarters. Second, ask what happened to wage-linked, market-priced services with short repricing cycles: that is your read on how long policy stays restrictive, and therefore on the discount rate applied to long-duration equity.
Two prints can look identical at the top line and imply opposite positioning. A benign headline driven by energy with firm services is a warning for duration and a modest positive for the consumer. A firm headline driven by energy with cooling services is a hit to the consumer and, if anything, a relief for duration. The framework's value is that it stops you trading the second scenario as if it were the first — which is the single most common way a well-informed generalist gets an inflation release backwards.
How to apply this framework
- Decompose the disinflation before you own it. Ask what share of any fall in headline came from energy, food and base effects versus wage-linked services. Composition tells you whether the fall changes the policy path or merely the arithmetic of a twelve-month comparison.
- Rank services components by measurement lag, not by weight. Market-transacted services that reprice frequently carry current information; shelter and imputed components carry stale information by construction. A hot aggregate driven by lease rollovers is a different signal from a hot aggregate driven by discretionary services pricing.
- Watch the wage-to-services link, because that is the persistence channel. If unit labour costs and negotiated wage growth are decelerating, services inflation has a path down even while the printed index remains elevated. If they are not, expect the policy path to stay restrictive regardless of where headline sits.
- Route each signal to the exposure it acts on. Headline surprises belong in the consumer-volume and mix analysis; services surprises belong in the discount-rate and duration analysis. Applying one to the other is the error this framework exists to prevent.