Write the curve as inflation = expected inflation + 0.5 × (natural rate − unemployment), with a natural rate of 4.5%. People expect 2% inflation and the central bank holds unemployment at 3.5%. Inflation is 2 + 0.5 × 1 = 2.5%. Next year people expect 2.5%, so the same 3.5% unemployment produces 3%, and the year after 3.5%. Keeping unemployment below the natural rate takes inflation that keeps rising, not a single step up. To get back to 2%, the central bank has to hold unemployment at 5.5% for three years: 3%, 2.5%, 2%. The numbers are invented. The logic is the result that replaced the 1960s menu.
A recent episode shows why the price of lower inflation is not fixed. In December 2022 the median Federal Reserve policymaker projected that unemployment would rise from 3.7% to 4.6% by the end of 2023 as inflation came down. Inflation did come down: consumer prices rose 6.5% in the year to December 2022 and 3.4% in the year to December 2023. Unemployment ended 2023 at 3.8%. Fed Chair Jerome Powell later said the reversal of pandemic distortions to supply and demand and of the energy shock was central to the decline, helped by inflation expectations that stayed anchored. Those forces shift the whole curve down. They do not move the economy along it.
The Phillips curve is not a menu of permanent choices, and it is not a law that low unemployment causes inflation. In 2019 U.S. unemployment averaged 3.7% while core PCE inflation, the Fed's preferred measure, stayed below 2% all year. A supply shock can also raise inflation and unemployment together, as in the 1970s, and then no point on the curve is a good one. The natural rate cannot be observed, only estimated, and the estimates move. This page is about the relationship itself. How unemployment is measured is in Unemployment, how beliefs about inflation form is in Expectations, and supply shocks are in Aggregate Supply.