Inflation: measurement, mechanisms, and expectations

Inflation is the rate of change of the price level of a basket of goods and services over time. Three official measures dominate, each constructed differently.

• Consumer Price Index (CPI). A weighted average of prices of a fixed basket representative of consumer spending. Weights are updated periodically (typically every 1–2 years) based on consumer expenditure surveys. The fixed-basket approach is computationally simple but lags behind real-time substitution.
• Personal Consumption Expenditures (PCE) price index. Used by the US Federal Reserve as its preferred inflation measure. Weights are updated more frequently and reflect actual spending patterns from national accounts. PCE typically runs 0.3–0.5 percentage points below CPI over long horizons because of methodological differences.
• GDP deflator. The ratio of nominal GDP to real GDP, capturing price changes across the entire economy (not just consumer goods). Includes investment, government spending, and net exports. The deflator can diverge from CPI when those components have different price dynamics than consumer goods.

For any monetary or fiscal policy claim, which inflation measure is being discussed matters. A central bank targeting 2% PCE inflation is implicitly targeting a different number than 2% CPI inflation. The first cursus reading rule: ask which measure.

Most inflation measures publish two versions:

• Headline inflation — the full basket including food and energy.
• Core inflation — the same basket with food and energy removed (and sometimes other volatile items).

Food and energy prices are excluded from core because they are dominated by supply-side shocks (weather, geopolitical disruptions to oil markets, harvest variation) that move prices in ways monetary policy cannot quickly offset. Excluding them produces a measure that responds more cleanly to underlying demand and wage dynamics.

The trade-off is interpretive:

• Headline matters for households. Real purchasing power tracks the full basket, including the things people buy frequently (gasoline, groceries). Wage demands respond to headline.
• Core matters for policy. Central banks set policy with longer lags than commodity price shocks, so they typically guide expectations using core trends. Persistent gaps between headline and core can themselves transmit through expectations.

The lesson is to read both. A sustained headline-only inflation often reverses when supply normalizes; a sustained core inflation suggests something deeper that policy is more likely to need to address.

Demand-pull inflation arises when aggregate demand for goods and services exceeds the economy's productive capacity at current prices. Formally:

$\text{aggregate demand} > \text{potential output at current prices.}$

The excess demand bids up prices. Mechanisms:

• Consumers buy more than producers can supply at current capacity, so prices clear the market upward.
• Firms facing strong order books raise prices because they can.
• Labor markets tighten and wages rise, which firms pass through to prices.

A demand-pull episode typically shows broad price increases across many sectors simultaneously, accompanied by low unemployment and rising wages.

The canonical policy response is to reduce aggregate demand: raise interest rates, tighten fiscal policy, or both. The transmission channel from policy to demand is the subject of the central-bank-toolkit lesson, but the basic logic is straightforward — higher rates make borrowing more expensive, slow investment and consumption, and bring demand back in line with supply.

Demand-pull is the inflation pattern central banks are best equipped to address, because their primary tools (interest rates) work directly on demand.

Cost-push inflation arises when production costs rise independently of demand: an oil price shock, a wage shock, a supply disruption that raises input costs across many sectors. Firms raise output prices to maintain margins; prices rise even though demand may be unchanged or weak.

A cost-push episode typically shows:

• Prices rising concentratedly in the affected sector first, then spreading as firms downstream pass through costs.
• Margins compressed before prices fully adjust.
• Output and employment often falling — stagflation in the extreme case.

The canonical policy response is harder. Tightening monetary policy reduces demand and can suppress the inflation, but at the cost of further reducing output and employment in an economy already supply-constrained. Accommodating the shock (not tightening) lets inflation persist longer but reduces output losses.

The stagflation episodes of the 1970s — concentrated oil shocks combined with accommodative monetary policy — were the canonical cost-push example whose policy lessons shaped central-bank doctrine for the following decades. The pandemic-era inflation of 2021–2023 had substantial cost-push components (supply chains, energy) layered with demand factors (large fiscal transfers, reopening surges), which is why it was harder to diagnose in real time.

Distinguishing demand-pull from cost-push in real data is hard. Most actual episodes have both components, and the diagnosis influences which policy response is appropriate.

The Phillips curve is the historically observed inverse relationship between unemployment and wage (or price) inflation. In its simplest form:

$\pi = -\alpha(u - u^*) + \text{error}$

where $\pi$ is inflation, $u$ is the unemployment rate, $u^*$ is the natural rate of unemployment, and $\alpha > 0$ is the slope. When $u < u^*$ (labor market tighter than equilibrium), inflation rises; when $u > u^*$, inflation falls.

The simple Phillips curve appeared to fit data well in the 1950s and 1960s, predicted a stable trade-off between inflation and unemployment, and was used as a policy menu (accept higher inflation, get lower unemployment).

The 1970s broke the simple model. Both inflation and unemployment rose together (stagflation), inconsistent with the stable trade-off.

Friedman and Phelps had predicted this failure: the simple Phillips curve omitted inflation expectations. The corrected version is:

$\pi = \pi^e - \alpha(u - u^*) + \text{error}$

where $\pi^e$ is the expected inflation rate. If the public expects 5% inflation, employers and workers negotiate wage increases that reflect that expectation; firms set prices accordingly. The unemployment-inflation trade-off shifts to whatever expectation prevails. Over the long run, no permanent trade-off exists — only short-run trade-offs when expectations have not yet adjusted.

The expectations-augmented Phillips curve has a strong implication: inflation outcomes depend on what the public expects future inflation to be.

• If expectations are anchored at a low target (e.g., 2%), supply shocks raise headline inflation temporarily but do not feed through to wages and prices because workers and firms expect inflation to return to target. The shock is one-time.
• If expectations are unanchored, a supply shock that raises inflation gets built into wage demands ('I need a 7% raise because inflation is 7%'), which firms pass through to prices, which raises expectations further. The system can enter a self-reinforcing loop.

This is why central banks devote so much attention to communications about future policy. The first-order goal is not the current inflation rate per se; it is keeping the expectation $\pi^e$ at the target, because $\pi^e$ then constrains what current inflation can drift to.

Measurement of expectations is itself imperfect:

• Survey-based — household and professional forecaster surveys. Slow to update, but broad coverage.
• Market-based — break-evens between nominal and inflation-protected bonds. Real-time, but affected by inflation risk premiums and liquidity.
• Implied from behavior — wage bargaining patterns, contract clauses, retail price-setting frequency.

A central bank watching whether expectations are 'still anchored' is watching all three with a heavy weight on whether they have diverged from the target — and acting before they do.

Why does the response of inflation to monetary policy operate with such long lags? The answer is sticky prices.

Firms do not adjust prices continuously. Each price change has costs: menu costs (literally repricing menus, labels, catalogs), customer reactions to frequent changes, internal decision processes. Empirically, firms adjust prices every several months to several years depending on the sector. Restaurant menus update less often than gasoline prices; clothing prices adjust seasonally; rent contracts adjust annually.

A standard formalization is the Calvo model: in each period, a random fraction $1 - \theta$ of firms can adjust their prices, while the rest hold prices fixed. The average duration between price changes is $1/(1-\theta)$ periods. For monthly data with $\theta = 0.85$, the average duration is about 7 months.

The consequence for inflation dynamics:

• A shock to demand or costs propagates gradually as more firms get their chance to reprice.
• The full pass-through of a shock to the price level takes several months to years.
• Monetary policy effects on inflation typically have peak impact 12–24 months after the policy change.

This lag is the structural reason central banks must act on forecasts, not on current data. By the time current inflation is observed, the policy decision needed for the next 12–24 months has already been made. The cycle-reading lesson returns to this point.

Three pieces of structural framework carry forward.

• Inflation is a number plus a story. A given inflation print can come from demand-pull, cost-push, expectations drift, supply chain effects, or some combination. The policy response depends on which combination is acting. The next lesson on central-bank tools is implicitly conditional on this diagnosis.
• Expectations are an input, not an output. A central bank's primary asset is the credibility that lets it keep $\pi^e$ anchored. Loss of that credibility shifts the entire Phillips-curve relationship and makes every subsequent policy choice more costly. Central-bank independence, communications strategy, and policy consistency are all in service of this credibility.
• Lags and dynamics matter. The Calvo-style sticky-price model is the structural reason policy must be forward-looking. Real-time data is necessary but not sufficient; what matters is the projection 12–24 months out.

With inflation framed, the next lesson moves to the central bank's actual instruments — the rate-setting, balance-sheet, and reserve-management tools by which it acts on the demand and expectation channels described here.