A pattern hidden in a century of data
By now you can read the unemployment rate as a headline that hides many human stories, and you have met its different flavours — frictional, structural, and the cyclical kind that swells in a slump. This guide adds the relationship that, more than any other, has tempted and tormented policymakers: the apparent deal between unemployment and inflation. Lower one, the story went, and you must pay with more of the other.
In 1958 a New Zealand–born economist, A. W. Phillips, plotted nearly a century of British data — from 1861 onward — with unemployment on one axis and the rate at which wages rose on the other. The dots traced a clean downward-sloping curve. When unemployment was low, wages climbed fast; when unemployment was high, wages barely moved. Later economists swapped wage growth for price inflation, and the [[phillips-curve|Phillips curve]] was born: a tidy, negative relationship between joblessness and inflation that seemed to hold across decades.
Why a short-run trade-off makes sense
Before judging the curve, it helps to see why such a trade-off is plausible at all in the short run. Picture an economy where demand suddenly surges. Firms scramble to produce more, so they hire harder; the pool of jobless shrinks and unemployment falls. But with workers now scarce, employers must offer fatter pay packets to lure and keep them — and those rising labour costs get passed into prices. Strong demand thus shows up as *both* lower unemployment *and* faster inflation at the same time. This is the demand-driven engine behind the curve.
Run the same logic in reverse and you get the other end of the curve. When demand is weak, firms shed workers, unemployment climbs, and with so many people chasing few jobs nobody can push for big raises — so wage and price growth cools. The same forces that ease inflation are the ones that throw people out of work. To a 1960s policymaker this looked less like a warning and more like a *menu*: pick the inflation–unemployment combination you like, then use monetary policy or fiscal levers to steer demand and land on that spot.
The 1970s: when the menu vanished
Then came the decade that wrecked the menu. Through the 1970s many rich economies suffered *both* high unemployment *and* high inflation at the same time — a combination the simple curve said should be nearly impossible. Economists coined a word for it: [[stagflation|stagflation]], a stagnant economy plus inflation. The neat downward-sloping line, when you plotted the new years, scattered into a cloud. The map no longer matched the road.
Two things had happened. First, brutal [[supply-shock|supply shocks]] — above all the oil-price jumps of 1973 and 1979 — pushed up costs and prices *while* choking output and jobs. A supply shock moves inflation and unemployment in the *same* direction, the opposite of the demand story, so it shifts the whole curve rather than sliding along it. Second, and more deeply, economists Milton Friedman and Edmund Phelps had warned even before the crisis that the trade-off would not survive being exploited. Their reasoning is the heart of the matter.
Expectations: the trap inside the trade-off
The crack in the curve runs through one word: expectations. Workers and firms do not care about inflation in the abstract; they care about *real* outcomes — what a wage actually buys. A pay rise feels generous only if it beats the price rises you expect. So the bargaining that drives wages and prices is built on [[inflation-expectations|inflation expectations]], and a curve drawn at one expected inflation rate quietly assumes those expectations stay put — a ceteris paribus hiding in plain sight.
Now watch the trap spring. A government decides to 'buy' lower unemployment by letting demand and inflation run a little hotter — say it pushes inflation to 5%. At first it works: prices rise faster than wages people negotiated when they expected, say, 2%, so real wages dip, hiring looks cheap, and unemployment falls. But people are not fools forever. Once they expect 5%, they demand 5% raises just to stand still. Real wages climb back, the hiring advantage evaporates, and unemployment drifts back to where it started — only now with 5% inflation baked in. The job gain was temporary; the inflation is permanent.
Short run (expectations lag): inflation 2% -> 5%, unemployment 6% -> 4% (gain!) People re-learn, expect 5%: wages catch up, unemployment 4% -> 6% (back) Long run: inflation stuck 5%, unemployment 6% (no gain)
Why the long-run curve is vertical
String those episodes together and a different shape emerges. In the short run there really is a downward-sloping curve — but there is a *different* one for every level of expected inflation, and each act of surprising people shifts you onto a higher one. The economy keeps returning to the same unemployment rate, just at ever-higher inflation. Connect those resting points and you get a vertical line: the [[long-run-phillips-curve|long-run Phillips curve]]. In the long run, there is no trade-off to buy — only the inflation rate is yours to choose.
The unemployment rate that line sits at has names you have met: the natural rate of unemployment, or in its inflation-anchoring form the NAIRU — the lowest unemployment compatible with steady, non-accelerating inflation. It is set by the plumbing of the labour market (how people search, how skills match jobs, how unions and regulations work), not by how much money is sloshing around. Demand can pull unemployment below it only at the cost of *rising* inflation, and push it above only with *falling* inflation. Trying to hold it permanently below the natural rate does not buy lasting jobs — it buys an inflation spiral.
The cautionary tale
Step back and the Phillips curve becomes a lesson larger than itself. It is a textbook case of an empirical relationship that held for a hundred years, was treated as a dependable law, and then dissolved the moment policymakers tried to lean on it. The deepest reason is that economies are made of people who learn. A pattern in *human* behaviour can rewrite itself once people understand the pattern and start to game it — a relationship that exists only because nobody is exploiting it can vanish the instant somebody does.
This is why economists are so insistent that correlation is not causation, and why they distrust any model that quietly assumes people will keep behaving as they did before. The curve has not been abandoned — modern versions add expectations and supply shocks and still guide monetary policy every day — but no one again mistakes a fitted line for a free lunch. Carry that humility up the ladder: the prettier a historical relationship looks, the more carefully you should ask *why* it held, and whether that reason will survive you acting on it.