The Difference Between Inflation Falling and Inflation Being Solved
A falling inflation rate is an arithmetic fact; a solved inflation problem is a change in the price-setting process — and multiples that price the first as the second carry an uncompensated tail.
Sophisticated investors long ago learned not to confuse disinflation with deflation. The error that survives at the sophisticated level is subtler: treating a falling inflation rate as evidence of a solved inflation problem. These are different claims about different objects. The rate is a twelve-month arithmetic result that can fall for reasons that have nothing to do with the underlying process being repaired. The problem is a property of how the economy sets prices — and it can remain fully intact underneath several consecutive prints of improving arithmetic. Markets trade the rate because the rate is a number; portfolios are exposed to the process.
The working distinction we use is earned disinflation versus mechanical disinflation. Mechanical disinflation is what you get when the arithmetic moves: a supply shock reverses, a commodity spike falls out of the twelve-month comparison, goods prices normalise after an inventory cycle. Nothing about the economy's price-setting behaviour has changed — the shock simply stopped. Earned disinflation is what you get when the process changes: price increases narrow from broad to concentrated, the wage-bargaining ratchet releases because workers stop expecting the recent past to repeat, and long-horizon expectations sit where they sat before the episode. Only earned disinflation changes what a central bank can safely do, which means only earned disinflation changes what an equity multiple can safely assume.
Three ways inflation falls — only one is a solution
It helps to be concrete about the channels, because a single disinflation typically blends all three. The first channel is shock reversal. Globally traded inputs — energy, freight, agricultural commodities, durable goods moving through congested supply chains — spike and mean-revert on their own physics. Supply responds, substitution happens, inventories rebuild. Inflation driven by this channel falls without anyone earning anything; it was a level shift in relative prices that the annual comparison eventually digests. The second channel is base arithmetic: the year-over-year rate falls simply because last year's extreme months roll out of the denominator. This channel is fully forecastable months in advance and carries no information at all about the present — we treat it at length in a sibling piece on base effects, but for this framework the point is that it is disinflation produced by the calendar, not by the economy.
The third channel is the only one that constitutes a solution: the process cooling. Demand softens enough, or credibility holds firmly enough, that the behavioural machinery which propagates inflation — annual wage rounds pricing in realised inflation, firms passing costs through because everyone else is, households pulling purchases forward because waiting costs money — loses its grip. You can observe this channel, but not in the headline rate. It shows up in breadth: the share of the basket inflating above target shrinks, trimmed and median measures converge down toward the target rather than hovering above it, and price increases re-concentrate in idiosyncratic categories rather than spanning the basket. Breadth is the fingerprint of process change, because a demand-and-expectations problem is by nature a broad problem. A falling rate with persistent breadth is channels one and two doing the work while channel three waits.
Why the market conflates the two: three layers of why
The illustration: the declared victories of the 1970s
The cleanest historical warning is the stop-go pattern of the 1970s, stated as established history. After the first oil shock's inflation surge subsided mid-decade, the rate fell substantially from its peak. By the arithmetic, the problem was receding; policy eased into the improvement, and the political system moved on. But the process had never been repaired — wage indexation was widespread, longer-horizon expectations had ratcheted up with each episode, and pricing behaviour remained broad. When the next supply shock arrived at the end of the decade, it landed on an economy whose price-setting machinery was still primed, and the second inflation wave crested higher than the first. Breaking the process, when it finally came, required a policy stance severe enough to cause a deep recession — a price that would not have been necessary had the first disinflation been treated as mechanical rather than earned.
The 2021–2023 episode offers the counter-illustration in its early phase: much of the initial disinflation was visibly mechanical — energy reversal and goods normalisation — while services breadth and wage growth cooled far more slowly. The reason that episode is remembered differently from the 1970s is precisely the process variables: long-horizon expectations held, and the wage ratchet, in an economy with little formal indexation, turned slowly rather than spinning. The general lesson survives any particular decade: the rate falling tells you the shocks are fading; only the process tells you whether the economy has stopped being flammable.
The strongest case against the distinction
The serious objection is that "solved" is unfalsifiable in real time — that the distinction collapses into hindsight. If inflation was caused by shocks, the objection runs, then the reversal of the shocks is the solution, and demanding additional evidence of process change is demanding proof of the repair of something that was never broken. On this view the famous "last mile" of disinflation is not a hard grind at all; it is simply the tail of mean reversion arriving on schedule, and investors who waited for process confirmation sacrificed returns to a distinction without a difference.
We accept the logical structure and note that it concedes the framework's point. If inflation was purely a shock phenomenon, then breadth should never have widened, wage growth should never have accelerated beyond productivity, and expectations should never have drifted — and where those things are true, the framework itself reads the disinflation as requiring no further proof. The distinction only bites when the process measures did move, and that is observable, not hindsight. The honest version of the debate is therefore an empirical question asked in real time: did the shock reach the wage base and the expectations structure, or not? Where it did not, mechanical disinflation is genuinely sufficient and the re-rating is safe to own early. Where it did, the reversal of the original shock removes the spark while leaving the flammability — and the 1970s stand as the canonical record of what that asymmetry costs. The framework does not demand pessimism; it demands that the multiple you pay match the evidence tier you actually have.
What this changes for a portfolio
The portfolio consequence is concentrated in one place: the price of duration. A solved inflation problem licenses a lower policy path, a compressed inflation risk premium, and the restoration of the bond-equity hedge that makes balanced portfolios work — all of which support higher multiples on long-duration equity. A merely falling rate licenses none of that durably, because it leaves the re-acceleration tail alive: the scenario in which the process reasserts itself, policy reverses, and the assets priced for the solved world reprice for the unsolved one. The asymmetry matters more than the base case. In the falling-but-unsolved phase, long-duration equity offers solved-world upside that is already substantially priced, against unsolved-world downside that is not — a distribution a portfolio owner should recognise as selling insurance without collecting the premium.
Sterling's Embedded Intelligence reads every disinflation through this two-ledger lens: one ledger for the rate, which sets the market's mood, and one for the process, which sets the risk. Sterling's operating rule is blunt — mechanical disinflation changes the forecast; only earned disinflation changes the regime. Position sizing on duration should follow the second ledger, and the width of the gap between the two ledgers is itself a measure of how much optimism is embedded in the price.
How to apply this framework
- Score every disinflation by channel before extending duration. Attribute the fall across shock reversal, base arithmetic, and process cooling. The first two justify revising forecasts; only the third justifies revising the regime assumption embedded in multiples.
- Treat breadth as the verdict, not the footnote. Trimmed and median inflation measures and the share of the basket inflating above target are the fingerprint of process change. A falling headline with stubborn breadth is arithmetic outrunning reality.
- Check whether the shock ever reached the wage base. If wage growth net of productivity never left its anchor, mechanical disinflation is genuinely sufficient and waiting for more proof has a real cost. If it did, shock reversal removes the spark but not the flammability.
- Distinguish easing-because-solved from easing-because-weak. Policy loosening on process evidence supports multiples; policy loosening into unrepaired breadth is the historical setup for the second wave — and for the sharpest duration drawdowns.