Causes of inflation

Inflation is a single outcome, the general price level rising, but it can be produced by four distinct engines, and the type matters enormously because it dictates the right response. A central bank should lean hard against some kinds and look through others; an analyst who cannot tell them apart will misjudge both the policy reaction and the effect on companies. Think of these as four causes that can run singly or, dangerously, together.

What is demand-pull inflation?

Engine 1. Demand-pull inflation: "too much money chasing too few goods." This is inflation from the demand side: aggregate demand outruns the economy's capacity to produce, so buyers bid up prices for a limited supply. The mechanism: when spending (consumption + investment + government + net exports) exceeds what the economy can produce at full employment (a positive output gap), the excess demand spills into prices rather than volumes, because the factories and workers to make more simply are not there. Demand-pull is the "good problem" kind of inflation in the sense that it comes with a hot economy, low unemployment, and rising wages; it is the kind the central bank's interest-rate tool is genuinely designed to cool (raise rates → dampen demand → close the output gap). Example: the post-COVID US inflation of 2021–22 had a large demand-pull component. Trillions of dollars of fiscal transfers (stimulus cheques) and pent-up savings hit an economy whose supply had been constrained by the pandemic, and demand for goods in particular surged well past what could be supplied, pulling prices up. India example: classic demand-pull shows up in India around festival and wedding seasons and in credit-fuelled booms, when demand for gold, autos, and housing outruns near-term supply; more structurally, a period of rapid credit growth and government capex feeding into an economy near capacity produces demand-pull pressure the RBI leans against.

What is cost-push inflation?

Engine 2. Cost-push inflation: the supply side gets more expensive. This is inflation from the cost side: the price of a key input jumps, and producers pass the higher cost through into output prices, so the supply curve shifts, not the demand curve. The signature of cost-push is the nasty one: prices rise even as the economy weakens, because the shock makes producing anything more expensive while also acting like a tax on consumers (they must spend more on the pricier input, leaving less for everything else). The archetypal cost-push shock is energy: the 1973 OPEC oil embargo roughly quadrupled the oil price and the 1979 Iranian revolution doubled it again, and because oil is an input to almost everything, the general price level was pushed up while output fell. That is the recipe for stagflation. US example: both 1970s oil shocks; and in 2022 the Russia–Ukraine war spiking European gas and global food/energy prices was a cost-push overlay on the existing demand-pull inflation. India example: India imports roughly 85% of its crude oil, so a global oil spike is a direct cost-push shock to Indian inflation and the current account simultaneously (M7.06); domestically, India's recurring food-supply shocks (a failed monsoon spiking cereal and vegetable prices, or an onion or pulses shortage) are textbook cost-push, and because food is ~46% of India's CPI basket (see CPI vs WPI in India) these supply shocks dominate Indian headline inflation in a way they never dominate US inflation. The policy dilemma with cost-push is acute: raising rates does nothing to grow more oil or onions; it can only crush demand further into an already-weakening economy. This is why central banks usually try to look through a one-off cost-push shock — unless it threatens to un-anchor expectations (Engine 4).

How does money growth cause inflation?

Engine 3. Monetary inflation: too much money. This is the classical, long-run engine, associated above all with Milton Friedman's dictum: "Inflation is always and everywhere a monetary phenomenon in the sense that it is and can be produced only by a more rapid increase in the quantity of money than in output." The intuition runs through the quantity theory of money, M × V = P × Y (money supply × velocity = price level × real output): if the quantity of money M grows much faster than real output Y, and velocity V is stable, the price level P must rise to absorb the extra money. Money is worth what it is scarce enough to be worth; flood the economy with it and each unit buys less. The crucial calibration from M7.01 and M7.02: this is a claim about broad money and spending, not about central-bank reserves, which is exactly why a decade of QE (which created reserves, not spendable broad money) did not produce the inflation that naive monetarists predicted, while the 2020–21 combination of QE plus fiscal transfers that put money directly into spending accounts did. The monetary engine is the dominant one over long horizons and at the extremes: every hyperinflation in history is a monetary phenomenon in the purest sense, a government printing money to pay its bills until the currency is worthless. Over shorter horizons, the strict money-to-prices link is loose and unreliable (velocity is not stable, and modern money is endogenous), which is why almost no central bank targets a money aggregate anymore. But as the explanation of sustained, large inflation, the monetary engine is inescapable: you cannot have a persistent, general inflation without an accommodating expansion of money and credit.

How do expectations cause inflation?

Engine 4. Expectations-driven (built-in) inflation: the self-fulfilling prophecy. This is the most insidious engine because it needs no fresh shock: it runs on belief. If workers, firms, and lenders come to expect 8% inflation next year, they act on it today: workers demand 8%+ wage rises to protect their real pay, firms pre-emptively raise prices 8% to protect their margins, lenders demand 8% higher nominal rates. Those actions, in aggregate, produce the 8% inflation everyone expected. Expectations become self-fulfilling, and once embedded they are extraordinarily hard to dislodge because everyone is rationally protecting themselves against everyone else. The vicious form is the wage-price spiral: prices rise → workers demand higher wages to keep up → higher wages raise firms' costs → firms raise prices further → workers demand more → round and round, each lap ratcheting the price level higher with no external shock needed. This is why central bankers are obsessed, to the point of what looks like overkill, with keeping inflation expectations "anchored" near target, because the entire modern framework of inflation targeting (M7.02) is, at bottom, a machine for managing this one variable. A central bank with credibility can suffer a large cost-push shock and have inflation fall back on its own, because everyone believes it will return to target and so does not build the shock into wages and prices; a central bank without credibility suffers the same shock and watches it metastasize into a spiral, because no one believes it will be contained. Expectations are the transmission belt that turns a temporary shock into a permanent inflation, which is precisely what happened in the 1970s (see stagflation) and precisely what Volcker had to break.

The analyst's synthesis: real inflations are usually a blend, and the blend determines the policy reaction and the market consequence. The 2021–23 inflation was demand-pull (fiscal stimulus) plus cost-push (supply chains, then energy and food) plus an accommodative monetary backdrop plus, dangerously, the beginnings of an expectations un-anchoring. That is why central banks eventually reacted so hard: they could look through the cost-push piece, but not the risk that expectations would embed it. When you read an inflation print, the professional question is not just "how high?" but "which engine, and is it the kind the central bank will fight or the kind it will look through?", because that judgment is what tells you where the policy rate, and therefore the gravity, is heading.

How do demand and supply shocks differ?

The four engines are causes. The diagram that holds them in one frame is the aggregate demand and aggregate supply model, and it is worth building explicitly, because it is the only device that tells you from the direction of a shock whether output and prices move together or in opposite directions. Three curves, on axes of the price level (vertical) against real output (horizontal).

Aggregate demand slopes down. At a lower price level real balances buy more, real interest rates and the real exchange rate adjust, and total planned spending on domestic output is higher. The curve is consumption, investment, government purchases and net exports summed at each price level, which is the same spending identity M7.01 built.

Short-run aggregate supply slopes up. With wages and many prices fixed by contract, a higher price level raises the margin on each unit sold, so firms produce more. That upward slope exists only because something is sticky, and it disappears when the stickiness does.

Long-run aggregate supply is vertical at potential output. Over the long run, output is set by the labour force, the capital stock and productivity, and not by the price level at all. The vertical line is monetary neutrality drawn as geometry.

Now the discipline EC1.01 enforces for a single market, applied to the economy. A movement along a curve is caused only by a change in the price level. Anything else that changes output or the price level at every point (fiscal policy, monetary policy, confidence, an input-price shock, a change in potential) shifts a curve. Confuse the two and you will predict the wrong sign about half the time, which is the same failure rate EC1.01 records for the single-market version.

The two shocks the model exists to separate:

Size them and the point stops being a diagram. Take an economy whose nominal spending grows 9% while potential output grows 3%. Real output cannot exceed potential for long, so in the long run the price level must absorb the rest: 1.09/1.03 − 1 = 5.83% of inflation. That is a demand shock, it is engine 1 running into a vertical LRAS, and the central bank's answer is simple even when it is painful: slow nominal demand.

Now hold nominal spending growth at 5% and knock 2% off long-run supply, which is what a durable energy or supply-chain shock does. The price level rises 1.05/0.98 − 1 = 7.14% and output falls 2%. That is stagflation, it is engine 2, and it is why the stagflation page calls it the hardest pathology: tightening to kill the inflation deepens the output loss, and easing to protect output validates the price rise. There is no lever that does both.

Two habits fall out of this and are worth carrying to the practice set (P19 and P20 drill them). First, before asking how large a shock is, ask which curve it moved, because that decides the sign of the output response and the sign is what a portfolio is positioned on. Second, an expectations shift (engine 4) moves SRAS by changing what wage and price setters build into contracts, which is why an anchored expectation is worth more to a central bank than any single good inflation print.