Cost-Push vs. Demand-Pull Inflation: Causes and Key Differences

Demand-pull and cost-push are two traditional ways to organize possible inflation mechanisms. In a demand-pull account, households, businesses, governments, or foreign buyers seek more goods and services than producers can supply quickly at existing prices. Firms may respond by increasing production, using capacity more intensively, and raising prices where demand is strong or additional output is costly.

In a cost-push account, a producer’s expenses rise or an essential input becomes harder to obtain. If the firm cannot fully absorb the cost through margins, improve productivity, or substitute another input, it may pass some of the increase through to customers. Production may also slow when the constrained input limits how much can be made. The Federal Reserve Bank of San Francisco’s introductory explanation uses demand growing faster than productive capacity for the first mechanism, and more expensive inputs such as wages or raw materials for the second. It also cautions that actual price-level changes can have more complicated causes. {source:frbsfDoctorEconInflationFactors2002}

The categories are not labels that can be read directly from one price index. “Demand-pull” is a causal story about spending relative to supply; “cost-push” is a causal story about production costs or constrained supply. An increase in a price is an observation. Explaining why it rose requires a model and evidence about the forces affecting buyers and sellers.

The terms also do not mean that every demand increase is inflationary or every cost increase raises the overall inflation rate. A business can meet stronger demand with unused capacity, productivity improvements, or inventories. A firm can absorb a temporary cost increase in its profit margin. The macroeconomic result depends on scale, duration, market structure, substitution, and what else changes at the same time.

A simple aggregate-demand and aggregate-supply diagram gives the distinction an intuitive form. Suppose, hypothetically, that a broad increase in planned spending moves demand outward while productive capacity is slow to adjust. Businesses may sell more and raise prices. In this stylized case, both prices and real output tend to rise in the short run. This is the textbook demand-pull pattern, not a claim that all real demand shocks produce the same outcome.

Now imagine a separate hypothetical disruption that makes an important production input scarce. Firms can supply fewer units at each price, so the supply relationship shifts inward. The combination can be higher prices and lower production than otherwise. This is the textbook adverse-supply or cost-push pattern. The example describes a possible mechanism; it does not show that a particular observed episode had that cause.

This first-pass model helps explain why the two stories can have different implications for activity. Stronger demand may support output and employment even as it adds price pressure. A negative supply shock can increase prices while weakening output, creating a more difficult tradeoff. Frederic Mishkin’s Federal Reserve speech describes the broad directional contrast between demand shocks and adverse supply shocks, while also discussing the role of expectations and policy. It is a dated individual speech rather than current policy guidance. {source:fedMishkinInflationActivity2008}

A diagram leaves many details out. It compresses diverse households, industries, workers, importers, and markets into two curves. It may treat prices as if they adjust together, though some contracts reset slowly and some markets reprice quickly. It does not by itself identify whether an observed change began with demand, supply, market power, policy, or a combination. Use it to form questions about mechanisms, not as a test that converts a single chart into a causal verdict.

Prices and quantities together can provide more information than prices alone. In a simplified market, an outward demand shift tends to raise both the quantity sold and the price. A negative supply shift tends to raise price while reducing quantity. If several product categories show unexpected price and quantity changes moving in the same direction, that pattern can be consistent with demand pressure; opposite-direction movements can be consistent with supply pressure.

Consistency is not proof. A measured quantity sold is affected by inventories, delivery timing, product quality, changes in who buys, and supply-chain bottlenecks. A price index can combine items whose prices move in opposite directions. If supply and demand shift at once, the observed direction of quantity can depend on which shift is larger. A supply interruption might reduce availability just as emergency spending raises demand; both stories can be true, while the final price and quantity reveal only their net effects.

Adam Hale Shapiro’s 2022 San Francisco Fed analysis illustrates one empirical approach. It examines more than 100 goods and services categories in the U.S. personal consumption expenditures (PCE) price index, using rolling 10-year regressions to estimate unexpected price and quantity movements. The study classifies same-direction movements as demand-related, opposite-direction movements as supply-related, and movements statistically indistinguishable from their predicted values as ambiguous. Its data window ends in April 2022, so this is a model-based, method-specific classification of a historical U.S. episode, not a direct reading of every price change or a current estimate. The sample and period matter; do not repeat episode-specific shares as timeless or universal. {source:frbsfSupplyDemandInflation2022}

Causal attribution therefore needs more than a sign pattern. Researchers may compare multiple industries, use data on input costs and spending, study timing, examine supply disruptions, and test alternative models. Different methods can disagree because they define shocks differently, use different data, or ask different counterfactual questions. A careful report distinguishes what was observed, how the analysis classified it, and what remains uncertain.

Inputs include labor, energy, materials, freight, rent, financing, and intermediate goods. When one becomes more expensive, a producer’s total costs may rise, but the effect on per-unit costs depends on how much of the input is used for each unit of output and whether productivity changes. Firms can respond by accepting a lower margin, raising prices, substituting inputs, redesigning products, changing suppliers, or reducing production. The response is likely to vary by industry and over time.

This is why wage growth by itself does not prove cost-push inflation. A wage is a price paid for labor, but labor cost per unit of output also depends on how much output workers produce. If compensation per hour rises while output per hour improves, unit labor cost may rise more slowly than wages, remain steady, or fall. The reverse can happen when productivity slows. Unit costs are still only one part of total cost, and an observed increase does not prove firms passed it through to consumer prices. For the labor-cost calculation, see What Are Unit Labor Costs? Formula, Productivity, and Inflation.

Energy or transport disruptions illustrate the difference between an input-price change and broad, persistent inflation. A higher energy price can affect the energy component of household bills directly and raise costs for firms that use fuel or electricity. How far that increase spreads depends on energy intensity, substitution options, contracts, margins, and whether firms and customers expect the change to last. The price level can remain higher after the inflation rate’s direct contribution fades, if the input price stops increasing.

A cost increase can also meet strong demand. Producers that face higher expenses may find it easier to pass them on when customers are still spending and competing supply is limited. Under weak demand, firms may instead absorb more of the cost, reduce output, or lose sales. These examples show why it is often more accurate to ask which forces contributed and through which channels than to choose one exclusive label.

The same event can affect both sides of the market. A disruption may reduce available supply and also shift spending toward substitute products. Fiscal support can raise demand while a production bottleneck limits the response. Exchange-rate changes can make imported inputs costlier while domestic demand is changing. Inventory rebuilding can create a burst of orders, even as shipping capacity remains constrained. In each case, it is possible for the demand and supply channels to reinforce or offset one another.

The timing of measurement complicates the picture. Spot prices may adjust before contracts, wages, or consumer prices. Firms can hold inventories for a while before changing prices; customers can postpone purchases; suppliers can redirect goods. Monthly price and quantity series may capture different phases of one adjustment. A price rise followed by a quantity decline does not necessarily mean that the initial shock was supply-only, and simultaneous price and output growth does not establish a pure demand story.

Broader inflation measures also aggregate very different markets. Some categories may show demand-driven patterns, some supply constraints, and some little movement. A headline rate can rise because many prices are increasing moderately or because a few heavily weighted items rise sharply. The composition matters for explaining the movement, even when the published aggregate is correctly calculated.

It is also useful to separate the price level from its rate of change. A one-time input shock can lift the price level and temporarily raise measured inflation while that adjustment enters the year-over-year comparison. If the shock does not continue or spread, inflation can later slow even though the affected prices do not return to their prior levels. Persistent inflation generally requires price increases to continue across time or spread through additional categories; the original shock need not be the only cause. The correct conclusion depends on the index, period, and data vintage.

An initial shock changes relative prices: one good or service becomes more expensive compared with others. Expectations can influence whether that relative-price change remains localized or affects broader wage and pricing decisions. If workers expect inflation to stay high, they may seek compensation that protects purchasing power. Firms considering future labor, materials, and financing costs may change prices sooner or by more. Those choices can add persistence, especially if businesses and workers repeatedly adjust to past inflation.

This process is not automatic. A worker may have limited bargaining power or a wage contract that updates infrequently. A business may be unable to pass costs along if customers can switch easily. Expectations are not directly observed; surveys ask people or firms what they expect, while market-based measures embed risk premia and other factors. Different measures cover different populations and time horizons, so no single reading settles whether expectations are stable or how strongly they influence behavior.

Federal Reserve Governor Adriana Kugler’s 2025 speech discusses how firms’ pricing decisions can reflect expected demand, current and anticipated costs, and expectations, as well as ways workers and businesses can reinforce inflation through second-round effects. The remarks represent the speaker’s views and should not be read as a current committee forecast or as proof that a particular feedback loop is occurring. {source:fedKuglerInflationExpectations2025}

The distinction between first-round and second-round effects helps keep claims precise. A direct rise in an input’s price is a first-round effect on the affected market and related costs. A later pattern of wage, service-price, or broader price adjustments may be consistent with second-round propagation. Establishing that propagation requires evidence across prices, wages, contracts, expectations, and time; it should not be assumed from a single energy or food price movement.

A central bank generally cannot manufacture a missing input, reopen a blocked port, or make a harvest larger by changing an interest rate. Monetary policy works mainly through financial conditions and aggregate demand: policy rates can influence other borrowing rates and credit conditions, which affect spending and investment with delays. The Federal Reserve’s FAQ describes this indirect transmission and emphasizes that many factors outside monetary policy also influence inflation and employment. Its explanation is specific to the U.S. Federal Reserve framework and does not imply a guaranteed result. {source:fedInflationEmploymentTransmission2024}

That limitation does not mean supply-driven price pressure is irrelevant to monetary policy. A central bank may consider whether an initial price shock is likely to affect broader expectations and pricing behavior. If it lowers aggregate demand to reduce persistent inflation pressure, however, it cannot selectively reduce demand only in the market with excess spending. Output and employment may also weaken. If it accommodates every price increase, broader inflation could persist. The choices involve uncertainty, lags, and tradeoffs rather than a mechanical rule.

Policy tools and supply repair operate through different channels. Better logistics, expanded production capacity, worker training, competition, or substitute inputs can address particular supply constraints, though these measures take time and lie beyond the direct reach of interest-rate policy. Monetary policy can still affect whether demand presses against available capacity and whether inflation expectations remain consistent with the central bank’s objective. The exact mandate, tools, and institutional arrangements differ by jurisdiction.

The output gap is one framework for asking whether demand is high or low relative to estimated productive capacity, but potential output is not directly observed and estimates can be revised. A gap reading therefore cannot independently identify a cost-push shock or dictate a rate decision. For that separate concept, see What Is the Output Gap? Potential GDP Explained. The Phillips curve offers another lens on inflation and slack, with its own limits and shifting relationships: What Is the Phillips Curve? Inflation, Unemployment, and Its Limits.

Start with the claim’s measurement. Which price index is being discussed, and does it cover consumer prices, producer prices, or a broader set of domestic output? Is the reported change monthly, year-over-year, or annualized? Does the statement describe the price level or the inflation rate? Before explaining a movement, identify the series, period, revisions, and whether the rise is broad or concentrated.

Next, ask what evidence would distinguish the proposed mechanisms. For demand, look for evidence of spending, sales, output, capacity use, or order backlogs relative to supply. For cost pressure, look at input prices, availability, delivery times, and production quantities. Wage data need productivity and unit-cost context. Price-and-quantity patterns can help, but mixed shocks and measurement noise may leave the result ambiguous.

Then separate the observation from the model’s causal interpretation. “The category’s price increased while measured quantity fell” describes data. “A supply shock caused the increase” adds a causal inference. Name the method and its limits, and state when evidence cannot distinguish explanations. The San Francisco Fed’s empirical approach makes this uncertainty visible by preserving an ambiguous category rather than assigning every movement to supply or demand. {source:frbsfSupplyDemandInflation2022}

Finally, consider expectations, policy, and time. Did the initial relative-price change spread to wages and other prices, or did it remain concentrated? Could monetary policy influence broad demand, and what activity tradeoffs might follow? Are supply-side conditions changing independently? A strong explanation can include more than one cause and can change as evidence is revised. The two labels are helpful organizing concepts; they are not a substitute for describing the mechanism, scope, and limits of the evidence.