Inflation Is Cooling in Speed but Widening in Reach

Inflation Is Cooling in Speed but Widening in Reach

Two of the most useful gauges of US inflation are now moving in opposite directions. The share of personal consumption expenditure categories rising more than 3% from a year earlier has climbed from roughly 38% in early 2025 to about 50%. Yet the Dallas Federal Reserve's twelve-month trimmed-mean PCE inflation rate has fallen from around 3% to 2.3%. One measure says excessive inflation is spreading across more of the consumption basket; the other says the underlying magnitude of price increases is moderating.

This is not a statistical curiosity. It is a compact description of the policy problem. Inflation can improve in intensity while deteriorating in breadth. A small set of extreme price increases can fade, pulling a robust mean lower, even as many ordinary categories drift from benign readings into the 3%-to-4% range. The economy then has less acute inflation but more generalized inflation. For the Federal Reserve, that is progress without comfort: the tail of the distribution is becoming less dangerous, while its center remains too far above a 2% objective.

The correct conclusion is deliberately narrower than either bullish or bearish rhetoric. The data do not establish a renewed inflation acceleration, because the trimmed mean is still declining. They also do not validate a clean return to price stability, because half the basket is rising faster than 3%. The evidence instead describes a broad, low-amplitude inflation regime whose persistence matters more than its drama. That regime can end benignly if wage growth, housing costs and service-sector margins continue to normalize. It can also become sticky if widespread moderate price increases reinforce one another through contracts, expectations and income growth.

Understanding which outcome is more likely requires treating inflation as a distribution, not a single number.

What the two gauges actually measure

Inflation breadth asks how many categories are experiencing unusually rapid price increases. Here the threshold is 3% year over year, one percentage point above the Federal Reserve's long-run target. If the basket contains N components and I_i is the annual inflation rate of component i, a simple unweighted breadth statistic can be written as B = (1/N) times the sum of indicators 1(I_i > 3%). A weighted version assigns each category its expenditure share. Either form deliberately discards the size of each overshoot. A category at 3.1% and one at 13% both count as above threshold.

That apparent crudeness is the point. Breadth measures participation. It asks whether inflation is concentrated in a handful of volatile components or dispersed through the economy. A broad rise is harder to dismiss as one-off noise because it can reflect common forces: labor costs, rent resets, insurance repricing, financing expenses, regulatory charges or a generalized willingness of firms to pass through costs.

The Dallas Fed trimmed-mean PCE performs a different operation. Each month it ranks detailed price changes, removes a portion of the most extreme increases and decreases, and averages the middle of the distribution using expenditure weights. The twelve-month rate compounds those underlying monthly readings. Trimming limits the influence of volatile outliers without permanently labeling particular categories as core or non-core. Food or energy can survive the trim in a quiet month; an extreme service category can be excluded in a turbulent one.

If ordinary price changes become smaller while fewer spectacular outliers dominate the upper tail, the trimmed mean can decline. Yet if many categories settle just above 3%, breadth can rise at the same time. The measures answer different questions: breadth asks how common above-comfort inflation is; the trimmed mean asks how fast the representative middle of the distribution is moving.

Neither dominates the other. A policymaker who watches only breadth can mistake a cluster of small threshold crossings for a major acceleration. A policymaker who watches only the trimmed mean can underestimate the institutional persistence of moderate inflation spread across many sectors. Together they reveal the shape of the distribution.

A numerical example of the divergence

Imagine a ten-category economy. In the first period, three categories rise 8%, 6% and 5%, two rise 3.2%, and the remaining five rise 1%. Seven observations sit in the central region after the most extreme tails are removed. Despite several very high readings, only half the categories exceed 3%, and the middle can already be moderating if the largest increases are treated as outliers.

Now suppose the next period records inflation rates of 4.0%, 3.8%, 3.6%, 3.5%, 3.4%, 3.2%, 2.8%, 2.2%, 1.8% and 1.5%. The severe 8% and 6% shocks are gone. The average and trimmed mean fall substantially, but six of ten categories now exceed 3%. Inflation has become less intense and more diffuse.

This distinction can also be expressed through moments of the cross-sectional distribution. Let mu be the weighted mean inflation rate, sigma its dispersion and q the share above a chosen threshold. A decline in mu does not determine the direction of q because q depends on the full distribution relative to the threshold. If sigma falls sharply and the center of the distribution remains a little above 3%, mass can move out of both tails and pile up just above the threshold. The result is lower average inflation, lower dispersion and higher breadth.

Threshold effects amplify the visual tension. Moving from 2.9% to 3.1% changes the breadth count even though the economic change is small. Moving from 9% to 5% has no effect on breadth even though it creates substantial disinflation. This is why the recent move from about 38% to 50% should be interpreted as a warning light, not a sufficient statistic. Its significance depends on the distance above 3%, the weights of the categories crossing the line, and whether the movement persists.

The journey from concentrated shock to generalized residue

The chart's longer history helps. Before the pandemic, the share of categories above 3% generally occupied a low range, while trimmed-mean inflation hovered around 2%. The 2021-23 inflation episode pushed both measures sharply higher. Breadth rose toward three quarters of the basket, while the trimmed mean approached 5%. That was a genuinely systemic inflation shock: large changes and widespread participation arrived together.

The first phase of disinflation reversed both. Supply chains normalized, commodity pressures eased, goods demand rotated toward services, and the most extreme price changes retreated. Breadth fell and the trimmed mean declined. This was the easy part of disinflation in the sense that removal of exceptional disturbances delivered rapid gains.

The current phase is different. Trimmed-mean inflation has continued toward 2.3%, but breadth stopped improving and turned higher. That pattern is consistent with a transition from shock-driven inflation to residual inflation. The dramatic components have normalized, yet many prices still reset at rates inconsistent with a stable 2% aggregate trend. Insurance premiums, rents, medical services, personal services, administered fees and labor-intensive categories need not be exploding to keep the distribution broad.

The distinction resembles the difference between putting out a fire and cooling the building. Eliminating the flames produces visible progress. Heat stored in walls, machinery and air dissipates slowly. In inflation, that stored heat lives in multiyear contracts, staggered wage bargaining, rent renewals, regulated pricing schedules, cost recovery and backward-looking expectations. A trimmed mean near 2.3% says much of the heat has left. Breadth near 50% says it has not left every room.

Why 2.3% is encouraging but not identical to target

A trimmed-mean rate of 2.3% is close enough to 2% to represent major improvement, especially compared with the peak near 5%. But close is not identical, and robust measures do not map mechanically into the headline index the Fed targets. The central bank's objective concerns total PCE inflation over time, not a promise that every core or trimmed statistic will equal exactly 2% every month.

There are at least three reasons to resist false precision. First, real-time inflation measures are noisy and revised. Second, relative prices must move: a productive economy can have some categories rising faster than 3% while others fall. Third, the relationship between a trimmed mean and future headline PCE varies across regimes. A 2.3% trimmed mean with declining wages and narrow breadth would be more reassuring than the same number with rising wages and widening breadth.

Still, the distance matters cumulatively. If the price level grows at 2.3% rather than 2% for five years, the difference compounds to roughly 1.5% in the level. That is not an inflation crisis, but persistent misses can affect credibility when households have already absorbed a large prior price-level shock. People experience prices in levels, while policymakers usually communicate rates of change. After a burst of inflation, even normalizing inflation leaves the cost-of-living level permanently higher unless outright deflation occurs. A broad set of continuing 3%-plus increases therefore feels less benign to households than a macro chart may suggest.

The right policy standard is not mathematical purity. It is confidence that inflation will average 2% sustainably without unnecessary damage to employment. The 2.3% reading raises that confidence; the breadth reading limits it.

Breadth as a signal of persistence

Inflation persistence is not simply high autocorrelation in an index. It emerges from adjustment mechanisms. Firms change prices at different times because information, menu costs, contracts and customer relationships differ. In Calvo-style sticky-price models, only a fraction of firms can reset prices in any period. After a common cost shock, the adjustment is staggered. Even when the original shock fades, firms that have not yet reset may catch up, spreading moderate inflation over time.

Breadth can capture that staggered propagation. If successively more categories cross 3% while the average shock diminishes, the data may show late adjusters rather than a new impulse. This is less alarming than synchronized reacceleration, but it is still relevant because delayed adjustments lengthen the return to target.

Input-output links create another mechanism. A rise in insurance, rent, freight, software subscriptions or professional services enters the cost base of many firms. Each pass-through may be modest, but the network distributes it widely. The result can be a high-breadth, low-amplitude pattern. Research on inflation microdata has repeatedly shown that sectoral shocks can propagate through production networks and that the extensive margin—the fraction of items changing price—contains information beyond the aggregate rate.

Yet breadth is not destiny. If nominal demand slows, firms may absorb cost increases in margins rather than pass them through. If productivity improves, unit labor costs can decline even with healthy wage gains. If inflation expectations remain anchored, workers and firms are less likely to extrapolate recent changes. Persistence depends on feedback, not just participation.

The key diagnostic is therefore duration. A one-month breadth increase may be threshold noise. A year-long rise across economically diverse, heavily weighted components is stronger evidence that inflation's center has not fully normalized.

The labor-cost channel

Services dominate consumer spending and are labor intensive. For many service providers, wages are the largest controllable cost. The relevant relationship is not wage growth alone but unit labor cost growth: nominal compensation growth minus productivity growth. If compensation rises 4% and productivity rises 2%, unit labor costs increase about 2%, compatible with target inflation after allowing for margins and measurement. If compensation rises 4% while productivity is flat, businesses face a stronger incentive to raise prices.

This channel can explain why breadth rises while the trimmed mean falls. Wage growth may cool from extreme levels but remain high enough to support 3%-plus price increases in a wide range of ordinary services. No category needs a double-digit surge. Restaurants, repair businesses, health providers, recreation services and personal-care firms can all reprice moderately as labor contracts and payrolls reset.

The Phillips curve provides a useful but incomplete frame. Inflation tends to respond to economic slack, expected inflation and supply disturbances, but the relationship is nonlinear and time varying. A labor market moving from overheated to balanced can deliver substantial disinflation without a recession. Once slack is near neutral, however, the final decline may be slower because wage and price setting reflect both current conditions and inherited nominal contracts.

For investors, labor data must be read jointly. Payroll growth, vacancies, quits, hours, participation, unemployment, wage trackers and productivity each illuminate different parts of the mechanism. A modest rise in unemployment accompanied by stable productivity and slower wage growth would make broad 3%-plus inflation less persistent. Renewed wage acceleration without productivity support would make the breadth signal more consequential.

Housing, insurance and slow-resetting prices

Housing inflation is famous for its lag. Market rents for new leases can slow long before official shelter measures do because statistical indices capture the full stock of occupied units and update gradually. The same principle applies to insurance policies, annual tuition, medical contracts, local fees and subscriptions. Prices that reset infrequently create delayed waves.

This makes the composition of breadth essential. If the categories moving above 3% are mainly lagging components whose forward indicators are cooling, breadth can worsen temporarily even as future inflation improves. In that case, the trimmed mean may be closer to the true direction of travel. If instead the crossings occur in flexible prices, new-lease rents, discretionary services and categories with frequent repricing, they may signal fresh demand pressure.

Insurance illustrates the ambiguity. Premiums can rise because replacement costs, disaster losses, litigation expenses and capital requirements increased in prior years. Those increases are painful and broad but may be a lagged recovery of costs rather than evidence of current overheating. They still affect measured inflation and household budgets. Monetary policy, however, cannot manufacture cars, homes or catastrophe reinsurance. Aggressive tightening aimed at such prices could suppress demand elsewhere without quickly reducing the source of the increase.

The policy task is to distinguish lagged relative-price adjustment from self-sustaining aggregate inflation. Relative-price changes should eventually be offset by slower increases or declines elsewhere if nominal demand is constrained. Aggregate inflation persists when too many categories can pass through costs without offset because total nominal spending keeps expanding faster than real capacity.

Demand, margins and the nominal-spending test

The most useful bridge between micro breadth and macro inflation is nominal spending. In broad terms, nominal GDP growth equals real output growth plus economy-wide inflation, with composition and measurement complications. If nominal demand repeatedly grows faster than the economy's productive capacity, firms can raise prices without sacrificing much volume. Breadth then reflects generalized demand support.

If nominal spending moderates while breadth rises, margins become the adjustment valve. Some companies will accept lower profitability, seek efficiencies or lose volume. The inflation process should eventually cool. Corporate earnings calls, small-business pricing plans and sector margins can therefore help determine whether widespread price increases are being validated by demand.

This is where the markups literature matters. During a sudden shortage, firms with market power may expand margins because customers tolerate price increases. As competition and supply normalize, those margins can compress, contributing to disinflation. But margins are heterogeneous. Economy-wide claims that profits either caused inflation or are irrelevant miss the mechanism. In some sectors, margin normalization offsets wage costs; in others, constrained capacity permits continued pass-through.

The divergence between falling trimmed inflation and rising breadth may mean the high-markup outliers are fading while ordinary firms continue modest pass-through. That is a healthier distribution than the 2022 peak, but it is not full normalization. Confirmation would come from slower nominal consumption, softer pricing intentions and a declining frequency of above-threshold increases without a collapse in real activity.

Expectations, credibility and the difference between levels and rates

Inflation expectations influence wage bargaining, contract indexation, inventory decisions and required returns. Modern monetary theory does not require every household to forecast PCE precisely. It requires price setters and wage setters to believe that persistent deviations will eventually be corrected.

Broad inflation can challenge that belief because it is visible. Consumers may not notice a trimmed mean falling from 3% to 2.3%, but they notice repeated increases across insurance, meals, rent and services. Frequency shapes perception. If many prices rise, households may infer that inflation remains normal business practice even when the average increase is smaller.

Anchored long-term expectations are therefore necessary but not sufficient evidence of victory. Survey medians can hide dispersion, and market-based measures contain risk and liquidity premia. Short-term expectations react to salient prices, especially food and energy, while longer-term expectations reflect confidence in policy. The ideal configuration is declining short-term expectations, stable long-term expectations and reduced disagreement.

The Fed's credibility allows it to tolerate small, temporary misses. Credibility is a stock that can be used, but not carelessly depleted. Overreacting to breadth could cause needless employment losses; ignoring it for too long could allow 3%-plus pricing norms to become embedded. A reaction function that explains the distinction between intensity and diffusion can preserve credibility better than declaring victory or panic.

What would constitute genuine reacceleration

The current chart alone does not show reacceleration. For that conclusion, multiple pieces of evidence should turn together. The trimmed mean would need to flatten and rise on a three- or six-month annualized basis; median inflation and other robust measures should confirm; breadth should increase across several thresholds, not only 3%; and the distribution should shift right rather than merely compress around the cutoff.

Sectoral confirmation matters. Renewed pressure in market rents, wages, discretionary services and goods would be more concerning than isolated rises in administered or lagged prices. Nominal consumption and credit growth would need to indicate demand capable of validating higher prices. Inflation expectations and corporate pricing plans would ideally corroborate the move.

Base effects also require care. A year-over-year rate can rise because a low monthly observation drops out of the comparison window, even if current monthly inflation is unchanged. Analysts should compare one-, three-, six- and twelve-month annualized rates, adjusted and unadjusted data, diffusion at several horizons, and revision histories. No single transformation deserves monopoly status.

A genuine second wave would therefore look like rising location, rising breadth and possibly rising dispersion: the center shifts higher, more categories participate, and upside tails thicken. Today's combination is different: breadth has risen, but the robust center continues to fall. That is uncomfortable persistence, not yet renewed acceleration.

What would constitute durable normalization

Durable normalization would not require every category to remain below 3%. Relative prices always change, and an inflation target applies to an aggregate. It would require the share above 3% to trend down over several quarters, the trimmed mean to settle near a target-consistent range, and shorter-horizon measures to avoid repeated upside pulses.

The composition should improve as well. Shelter measures should converge toward cooler market-rent signals. Wage growth should align with productivity and 2% inflation. Service-sector pricing intentions should normalize. Inflation expectations should remain anchored, and the price-change distribution should regain a balanced shape rather than depending on a few falling goods prices to offset widespread service increases.

Crucially, this should happen without a collapse in employment or output. The best disinflation is supply-assisted: productivity, labor-force participation, housing supply, logistics and investment expand capacity while nominal demand slows gradually. Such an outcome lowers inflation and preserves real income. Demand destruction can also reduce breadth, but at a much higher social and portfolio cost.

The current data leave that soft-landing route open. A 2.3% trimmed mean suggests the process is advanced. The breadth reversal says the last mile remains unfinished.

Implications for Federal Reserve policy

The divergence argues for patience with a conditional bias, not mechanical easing or renewed tightening. Monetary policy acts with long and variable lags. If the Fed responds only to the breadth increase, it risks tightening after the original shock has already faded. If it responds only to the trimmed mean, it may underestimate the probability that moderate inflation becomes entrenched.

The sensible reaction function places weight on both the level and momentum of robust inflation, labor-market balance, expectations and financial conditions. With the trimmed mean falling, the burden of proof for a hike is high. With half the basket above 3%, the burden of proof for rapid easing is also high. Policy can remain restrictive enough to prevent renewed demand pressure while allowing accumulated disinflation to work through slow-resetting categories.

Risk management is asymmetric but not one-sided. Inflation above target imposes cumulative purchasing-power costs and threatens credibility. Excessive restraint imposes unemployment, investment losses and financial stress. The probability-weighted policy choice depends not only on the modal forecast but on the cost of tails. Broad inflation raises the persistence tail; falling trimmed inflation lowers the acute acceleration tail.

Communication matters. Policymakers should state that falling underlying inflation is welcome, widening breadth is a reason for vigilance, and neither measure alone determines the next decision. This avoids the credibility damage of categorical claims that later data may reverse.

Rates and inflation-market implications

For government bonds, the signal is more supportive of duration than the breadth line alone implies, because the magnitude of underlying inflation is falling. But it does not justify treating the 2% destination as guaranteed. Term premia should still compensate investors for uncertainty about persistence, fiscal supply and the covariance between inflation and growth.

The Fisher decomposition offers a basic map: a nominal yield is approximately the expected real rate plus expected inflation plus risk premia. A falling trimmed mean can reduce expected inflation and the inflation risk premium. Rising breadth can slow that decline by increasing uncertainty about how long above-target inflation lasts. The result may be lower front-end policy expectations without an equally large decline in long-maturity yields.

Inflation breakevens and swaps require similar nuance. If markets price a renewed inflation wave solely because breadth reached 50%, inflation protection may become expensive relative to the evidence. If they focus only on the 2.3% trimmed mean and price away nearly all upside risk, protection may be attractive. The trade is about the distribution of outcomes, not a point forecast.

Curve shape can express the distinction. Near-term rates respond to the expected policy path; longer maturities incorporate secular inflation uncertainty, fiscal risk and term premium. A patient Fed facing sticky breadth can produce a relatively range-bound front end and volatile long end. Investors should separate a disinflation view from a duration view because fiscal issuance and real-rate uncertainty can overwhelm modest inflation improvement.

Equity, credit and sector implications

Equities care about inflation through discount rates, nominal revenue, margins and policy risk. A decline in underlying inflation is normally helpful because it reduces the probability of aggressive tightening and lowers uncertainty around the discount rate. Rising breadth complicates the benefit. It suggests that pricing power remains distributed across the economy, which may protect nominal revenues but also keeps labor and input costs alive.

The sector effect is not uniform. Businesses with recurring revenue, low labor intensity and strong productivity may preserve margins as inflation cools. Labor-intensive companies with limited pricing power can be squeezed if wages and insurance costs remain above 3% while final-demand growth slows. Banks may benefit from higher nominal rates but face credit deterioration if policy stays restrictive. Homebuilders and real-estate businesses are sensitive less to one inflation print than to the duration of high financing costs and the evolution of housing supply.

Quality matters more in this regime. When inflation was surging, nominal revenue growth could conceal weak unit economics. When inflation intensity falls but breadth remains elevated, companies must translate modest pricing into real volume and productivity. Free cash flow, balance-sheet flexibility, working-capital discipline and the ability to automate become more valuable than raw top-line growth.

Credit investors should watch interest coverage. A 2.3% trimmed mean may encourage hopes for lower policy rates, but sticky breadth can keep nominal borrowing costs elevated longer. Firms refinancing fixed-rate debt into a higher-rate environment face a delayed cost reset analogous to the delayed price resets in the consumption basket. Spreads may remain calm while all-in yields pressure weaker borrowers. This is a reason to prefer issuers with manageable maturity walls and genuine pricing power over firms whose apparent resilience depends on refinancing being cheap.

Portfolio construction under a broad, low-amplitude regime

The central portfolio lesson is to distinguish confidence in disinflation from confidence in a specific asset response. One can believe that inflation will drift lower and still avoid a concentrated long-duration position if term premia, fiscal supply or valuation are adverse. One can worry about breadth and still own equities if productivity and earnings compensate for modestly higher inflation.

A robust portfolio can combine several exposures. High-quality nominal bonds hedge a growth-led disinflation outcome. Inflation-linked bonds or options hedge a renewed right-tail shock. Equities with pricing power and productivity exposure participate in a benign nominal-growth path. Cash and short-duration instruments preserve flexibility if the policy path remains uncertain. Commodity exposure may hedge supply shocks, though it carries negative roll, volatility and no guarantee of matching household inflation.

Position sizing should reflect estimation error. The breadth statistic depends on component definitions, weights and a chosen threshold. The trimmed mean depends on trimming conventions and seasonal adjustment. Treating either as a precise trading trigger invites false confidence. Scenario diversification is more defensible than betting the portfolio on whether one line is “right.”

The covariance structure also changes by scenario. In benign disinflation, stocks and bonds can rise together as discount-rate pressure falls and earnings hold. In growth-led disinflation, bonds may rally while cyclicals and credit weaken. In an inflation reacceleration, both nominal bonds and long-duration equities can fall. Inflation protection is valuable not merely because expected inflation may be high, but because it can perform when traditional stock-bond diversification breaks down.

Four scenarios for the next stage

The first scenario is benign convergence. Breadth turns down with a lag, the trimmed mean approaches 2%, productivity absorbs wage growth, and housing inflation continues to cool. The Fed can ease gradually because both sides of its mandate are balanced. Duration benefits, credit remains stable and equities receive both valuation and earnings support. This is the soft-landing path.

The second is sticky convergence. The trimmed mean remains around 2.3%-2.6%, while breadth stays near half the basket for several quarters. No new shock arrives, but contracts and service prices adjust slowly. The Fed maintains restrictive policy longer than markets prefer, though it does not need to hike materially. Front-end volatility persists, rate-sensitive sectors struggle, and security selection dominates broad beta.

The third is renewed acceleration. Breadth rises through several thresholds, the trimmed mean turns upward, wage growth reaccelerates relative to productivity, and nominal demand remains too strong. Inflation expectations may also drift higher. The Fed must tighten or delay easing, real yields and term premia rise, and both duration and expensive growth equities are vulnerable. Commodity and inflation-linked exposures may outperform.

The fourth is growth-led disinflation. Breadth falls quickly because demand and employment weaken, not because supply improves. The trimmed mean moves below target-consistent rates. The Fed eases faster, but credit losses and earnings downgrades offset the benefit for risk assets. Government bonds outperform, while lower-quality credit and cyclicals lag.

Today's divergence assigns meaningful probability to the first two scenarios and does not by itself validate the third. The fourth remains a policy risk if officials mistake residual breadth for a new inflation shock and over-tighten. This scenario map is more useful than a binary claim that inflation is either defeated or returning.

A dashboard that respects the distribution

A disciplined monitoring framework should begin with cross-sectional measures. Track the share of components above 2%, 3%, 4% and 5%; weighted and unweighted diffusion; median and trimmed-mean inflation; dispersion; skewness; and the persistence of threshold crossings. A rise only at 3% with declining 4% and 5% shares is less threatening than a synchronized rise across all thresholds.

Next examine time horizons. Compare one-month noise with three- and six-month annualized rates and twelve-month trends. Diffusion based on monthly changes can turn before year-over-year breadth because annual comparisons contain old observations. If shorter-horizon breadth is falling while twelve-month breadth rises, the divergence may resolve benignly as base effects roll through.

Then examine economic composition. Separate goods from services, housing from non-housing services, flexible from sticky prices, market-set from administered prices, and current from lagging indicators. A common factor across unrelated categories is more important than a cluster tied to one regulated-price schedule.

Finally connect inflation to the mechanism: wages relative to productivity, nominal spending relative to real capacity, margins, rents, credit, expectations and financial conditions. The goal is not to assemble more data for its own sake. It is to determine whether price increases are being validated by demand, propagated by contracts or merely recording old shocks.

This dashboard also guards against narrative overfitting. Analysts often change preferred measures when the headline becomes inconvenient. Precommitting to a balanced set of indicators makes it harder to cherry-pick the series that confirms a portfolio position.

Research lessons behind the signal

Several strands of economic research clarify why the two lines can diverge. Sticky-price models emphasize staggered adjustment: firms do not all reprice simultaneously, so inflation can persist after the original disturbance. State-dependent pricing models add that large shocks cause firms to adjust more quickly, while smaller residual changes arrive more slowly. This can naturally produce a phase in which the magnitude falls before participation fully normalizes.

The New Keynesian Phillips curve links current inflation to expected future inflation and real marginal cost. Its practical lesson is that inflation can decline as supply constraints ease and expectations remain anchored even when the labor market is not deeply weak. But the model's aggregate form can conceal sectoral heterogeneity. Multi-sector extensions show that differences in price stickiness and input-output connections affect both the persistence of inflation and the real cost of bringing it down.

The “divine coincidence”—the idea that stabilizing inflation can also stabilize the output gap—breaks down under sector-specific supply shocks and relative-price distortions. When insurance, housing or energy need to reprice relative to other goods, forcing every category toward the same rate can require unnecessary output losses. A central bank should stabilize aggregate inflation while permitting efficient relative-price movement.

Macro-finance research adds that asset prices respond to inflation risk, not simply expected inflation. Inflation is especially damaging to nominal bonds when it covaries with weak growth or policy instability. The breadth signal can raise uncertainty about persistence even as the trimmed mean lowers the modal inflation forecast. This combination can reduce expected policy rates while leaving inflation risk premia sticky.

Empirical work using micro price data also distinguishes the intensive margin—how much prices change—from the extensive margin—how many prices change. During large shocks, both margins can move. During normalization, they need not retreat together. The present chart is essentially an intensive-versus-extensive-margin story expressed in accessible aggregate statistics.

Measurement caveats that should change confidence, not erase the signal

PCE data are built from detailed source information and are revised as more complete data arrive. Component-level year-over-year rates can be affected by seasonal adjustment, quality adjustment, substitutions and changes in expenditure weights. A breadth index can count tiny categories alongside economically important ones unless properly weighted. It can also jump when many readings cluster around an arbitrary 3% line.

The trimmed mean has choices too. The amount trimmed, the treatment of weights and the conversion of monthly changes into annual rates affect the result. Because the composition of the trimmed set changes each month, the statistic should not be interpreted as a fixed basket of “true” inflation. It is a robust estimator designed to reduce tail influence.

These caveats do not make the indicators useless. They define the appropriate confidence interval. If alternative breadth thresholds, median inflation and several robust means tell the same story, confidence rises. If only one cutoff diverges, caution is warranted. The aim is triangulation.

Revisions deserve special attention near turning points. A policy or trade built on a few basis points of apparent improvement can be fragile. Persistent directional movement across many releases is more informative than the latest decimal. The chart's broad message—much lower intensity than at the peak, but renewed diffusion since early 2025—is sufficiently large to merit analysis even though exact endpoints may change.

Why the 3% threshold has economic meaning without being sacred

Three percent is useful because it marks a meaningful overshoot relative to a 2% objective and is easy to communicate. But it is not a cliff. Inflation of 2.99% is not stable while 3.01% is dangerous. The threshold converts a continuous distribution into a readable participation rate, sacrificing information to gain intuition.

Analysts should therefore ask how far above the line categories sit. One useful extension is excess breadth: sum the expenditure-weighted positive gaps max(I_i - 3%, 0). Another is a family of diffusion curves showing the share above every threshold. These measures distinguish many slight overshoots from fewer severe ones.

Duration above threshold also matters. A category spending one month at 3.1% differs from one spending two years at 4%. A hazard-style framework could estimate the probability that a component returns below 3% conditional on how long it has remained elevated. Persistent categories deserve more policy attention than transient crossers.

The published chart should be read as the first page of that analysis, not its final page. Its power is to expose the question: why is participation worsening while the center improves? The answer requires the richer distribution.

The political economy of an unfinished disinflation

Inflation debates are difficult because aggregate improvement and household dissatisfaction can both be valid. A lower trimmed mean is genuine macroeconomic progress. Yet widespread price increases arrive on top of a permanently higher price level. Households do not receive a refund for past inflation, and their personal baskets differ from official weights.

This gap affects policy credibility. If officials emphasize only declining inflation, they may sound detached from lived costs. If they imply that the price level should broadly reverse, they create an unrealistic expectation of deflation that would often require recession. The honest message is that the rate of increase has fallen substantially, while affordability depends on real wages, productivity, housing supply and competition catching up with the price level.

Broad inflation also distributes pain unevenly. Lower-income households spend more on necessities and have less capacity to substitute or hedge. Retirees with fixed nominal income face different exposures from leveraged homeowners or workers receiving wage gains. A single PCE rate averages those experiences. Fiscal policy targeted at supply and vulnerable households may address distributional problems more efficiently than monetary policy, which works by cooling aggregate demand.

The institutional challenge is coordination without fiscal dominance. Monetary policy must preserve the nominal anchor; fiscal and structural policy should improve capacity rather than offset restraint with generalized demand. Housing reform, labor-force participation, infrastructure reliability and productivity-enhancing investment can lower the sacrifice ratio—the output cost of reducing inflation.

What the chart does and does not say

It says that the disinflation process has changed character. The violent upper tail of the inflation distribution has receded enough to bring trimmed-mean PCE to about 2.3%. At the same time, the fraction of categories exceeding 3% has risen to about half, reversing part of the earlier breadth improvement.

It says that inflation is broad and still above comfort. It says the final approach to target may be slower than a smooth headline forecast implies. It says policymakers and investors should watch diffusion, composition and duration rather than celebrating one robust average.

It does not say that inflation is accelerating in the aggregate. It does not prove expectations are unanchored, wages are in a spiral or monetary policy has failed. It does not establish that rates must rise. For those claims, the center, short-horizon momentum, wages, demand and expectations would need to turn together.

It also does not say the breadth series is irrelevant because the trimmed mean is lower. Breadth identifies a persistence risk that an average can obscure. The two measures are complementary precisely because their divergence forces a better question.

Conclusion: progress without permission to relax

The US inflation distribution is sending a nuanced message. Underlying intensity has moderated dramatically: the Dallas Fed trimmed-mean PCE rate has fallen from around 3% to 2.3%, far below its pandemic-era peak. But participation has worsened: about 50% of PCE components are now rising more than 3% year over year, up from roughly 38% in early 2025.

The most coherent interpretation is neither renewed inflation panic nor completed price stability. Extreme increases are fading while moderate above-target increases have become more common. Inflation is cooling in speed but widening in reach.

That configuration supports patient, data-dependent monetary policy. It lowers the case for reacting to an acute inflation surge, but it raises the standard of evidence required before declaring victory or easing rapidly. The next phase will be decided by whether broad moderate increases lose support from wages, nominal demand and lagged contracts, or whether they become a stable 3%-plus pricing norm.

For markets, the distinction separates a favorable decline in inflation from an unconditional duration or risk-asset signal. Bonds benefit from lower underlying inflation but remain exposed to persistence and term premia. Equities benefit from lower policy risk but face uneven margin pressure. Diversified portfolios should hedge both growth-led disinflation and a renewed inflation tail rather than treating either line as an oracle.

Most importantly, the divergence restores intellectual discipline. A price index is a distribution compressed into one number. When the center and participation disagree, the answer is not to choose the line that fits a preferred narrative. It is to understand why the distribution is changing. Today that distribution shows substantial progress, meaningful residual heat and no definitive evidence of a new inflation wave. That is enough to justify vigilance—and not enough to justify alarm.