Wages, Unit Labour Costs and the Phillips Curve

Unit labour costs: the bridge between the two reports

The labour report and the inflation report are two halves of one mechanism, and unit labour costs are the bridge. Compensation per hour measures what an hour of labour costs; output per hour measures how much output that hour produces; dividing one by the other gives the labour cost embedded in a single unit of output. The growth-rate version is the one quoted: unit labour cost growth is compensation growth less productivity growth, so a 4.6% pay increase with 1.4% productivity growth is a 3.2% increase in the labour cost of a unit of output. Nothing about the pay rise is inherently inflationary — productivity is what decides how much of it is absorbed. From there the chain to consumer prices runs through the markup. If a firm holds its percentage markup constant, its output price rises with unit labour costs, and the pass-through is close to complete in a service business where labour is most of the cost. If the firm absorbs the increase, its margin compresses and the price rise is smaller. That is why the same wage print means different things in different years: when margins are fat, an increase can be absorbed for a while, and when they are thin it is passed on immediately. The observable version is the labour share of income — compensation as a share of output — which tells you whether workers are gaining relative to capital or whether the wage rise is simply keeping pace with prices. The chain has a second leg, and it is the one the credibility lesson was about. Nominal wage growth is what workers ask for; real wage growth is what they actually get. If prices rise 5% and wages rise 4.6%, real pay has fallen, and the next wage negotiation opens with that gap as its starting point. The self-reinforcing version — wages chasing prices while prices chase wages — is the wage-price spiral, and it requires expectations to stop being anchored: workers must believe the price increases will continue for their wage demands to anticipate them. That is why the same wage number can be evidence of a spiral or evidence of a catch-up, and why the anchor is what distinguishes the two cases rather than the level of the number. From a pay rise to a price — Compensation per hour: +4.6%: the wage number everyone quotes · Output per hour: +1.4%: what productivity absorbs · Unit labour cost: +3.2%: the cost increase that actually has to go somewhere ← · Markup expanding 2.7%: prices 5.9% — the firm passes on the cost and takes more · Wage growth consistent with 2% inflation: 3.4% at this productivity and a stable markup The last row is the number a policy maker actually wants: with 1.4% productivity growth and no change in margins, 3.4% wage growth is consistent with 2% price inflation. Wage growth above that is either a margin squeeze, a price increase, or both — which is what makes it the cleanest single statistic about inflation pressure that exists.

The Phillips curve, flattened rather than dead

The classic relationship says unemployment and inflation trade off: a tighter labour market produces faster wage growth, which becomes faster price growth. The relationship held strongly enough in the nineteen-sixties and seventies to be treated as a menu, and the treatment failed for the reason this curriculum has already built twice — if policy makers try to hold unemployment below its sustainable level, expectations adjust, the trade-off disappears, and the result is accelerating inflation with no permanent employment gain. What replaced it is a relationship that is still there but much flatter. The modern curve has three terms: expected inflation, the gap between unemployment and its sustainable level, and a supply shock. Two of the three are usually dominant. Expectations explain most of the level of inflation — which is why a credible central bank can bring inflation down with less unemployment than the historical ratio implied — and supply shocks explain most of the volatility, which is why inflation can spike with a labour market that is not tight at all. The unemployment term survives as the direction of pressure: a very tight market pushes inflation up and a very slack one pulls it down, but the same unemployment rate is consistent with a wide range of inflation outcomes, and no committee can read a rate off the curve and set policy from it. Three pieces of evidence are worth holding together. The very flat curve of the twenty-tens, when unemployment fell to levels once thought inflationary and inflation stayed below target, showed how weak the slope had become. The inflation surge of 2021–2022 arrived with a labour market that was tight, but the increase was led by energy and goods rather than by wages, and unit labour costs lagged the price move — which is evidence for the supply-shock term rather than against the labour-market term. And the disinflation that followed cost far less output than the historical relationship implied, which is exactly what an anchored expectation buys: the wage-price leg never engaged, so the shock drained out through the supply channel instead of through a recession. Read a wage number as a direction of pressure and a test of the anchor rather than as a dial to turn. • The curve flattened rather than died: the slope is small and the level is set by expectations. • Expectations explain most of the level; supply shocks explain most of the volatility. • The unemployment term survives as a direction of pressure, not as a usable trade-off. • Read a wage print against productivity and the anchor, or it says almost nothing. • A disinflation that costs little output is evidence that the wage-price leg never engaged. The most common error in reading wages is comparing a nominal wage number with the target. Wage growth of 4% with 2% productivity growth implies about 2% price inflation with a stable margin — but wage growth of 4% with 0.5% productivity implies 3.5%, and the two look identical in a headline.

Three wage measures, three different answers

This lesson has treated wage pressure as something to compute, and in practice it is something to read — with three headline measures that disagree often enough that knowing why is the difference between seeing a signal and seeing an artefact. They are not competing estimates of one number; they are answers to three different questions. The first is **average hourly earnings**, published monthly alongside the employment report. It is timely and simple, and it is a mean, which makes it vulnerable to composition. If a month’s job losses are concentrated in low-wage sectors, the average rises without anyone receiving a raise, because the mix underneath it changed. The report’s own revisions and the birth-death modelling of firms add further noise. As a monthly indicator of the pace it is useful; as a measure of whether a particular worker’s pay is rising, it can be actively misleading. The second is the **employment cost index**, published quarterly and built from a fixed set of jobs, which is what makes it the cleanest available measure of wage inflation: because the basket does not change, composition cannot flatter it. It is also the measure policy-makers treat as the reference when they argue about whether pay is consistent with the inflation target. The price of that quality is timing — a quarterly figure with a lag cannot lead anything, and it is often published after the story it describes has already moved. The third is the household-survey tracker that follows the same individuals from one year to the next, and it answers the question the other two cannot: what happened to the pay of people who stayed in their jobs. That distinction matters, because aggregate wage growth is the sum of raises for incumbents and the effect of people changing jobs, and the two behave differently across the cycle. Quits and job-switching drive the second term, which is why the quits rate from the job-openings survey belongs beside these numbers rather than in a separate conversation: a worker who changes employer is usually the channel through which pay accelerates. The reason they diverge is mechanical, and a single example shows it. In a month when employment gains are concentrated in high-paying sectors, average hourly earnings can rise quickly while the fixed-basket index shows much less, and the tracker shows something in between, because the composition effect is inflating the first and absent from the others. None of the three is wrong. They are measuring different populations with different controls, and the disagreement itself is informative — a wide gap between the mean and the fixed-basket measure is a statement about which jobs were added, not about pay. That suggests how to use them together rather than choosing a favourite. The fixed-basket index carries the trend; average hourly earnings carry the timing; the tracker carries the pressure on people who did not move; the quits rate carries the part of the story that comes from changing jobs. And none of them translates into consumer prices on its own — only the cost of labour per unit of output does, which is the comparison the earlier reads built. A wage print is evidence about one term of that equation, and it becomes an inflation argument only when productivity is in the same sentence. The practical rule is the same one this subject applies everywhere: never treat one release as a turn. Read the four-quarter trend in the fixed-basket measure, the three-month annualised rate in the monthly one, and check whether the tracker and the quits rate agree. When all four point the same way, the signal is real. When two of them do, the question is which composition effect is doing the talking. • Average hourly earnings are timely and distorted by which jobs were added. • The employment cost index holds the basket fixed, which makes it the cleanest trend measure and the slowest. • The same-worker tracker isolates raises for people who stayed, and the quits rate captures the switching channel. • Only unit labour costs translate a wage move into a price move; a print alone is one term of the equation. A useful cross-check is to compare the trend in the fixed-basket measure with the trend in unit labour costs. When they move together, pay is feeding through; when they separate, productivity or margins are absorbing the difference, which is exactly the mechanism the Phillips-curve section described.

Sources

• The wage-price process and unit labour costsStandard Phillips-curve literature; BLS productivity and costs release
• The flattening of the Phillips curveThe post-2009 literature on the missing disinflation and the missing inflation; Federal Reserve Board research
• Expectations as the transmission belt from wages to pricesMR7 on expectations and credibility in this curriculum
• Wage growth consistent with a 2% inflation objectiveFederal Reserve commentary on wage growth and productivity; standard labour-share arithmetic