Two Government Numbers Disagree About Your Raise. Only One Is About You.

In one week this August, the federal government published two numbers about American pay.

The first: over the twelve months to July 2026, real average hourly earnings fell 0.2 percent.^[1] Adjusted for prices, the average hour of American work bought slightly less than it did a year earlier.

The second: over the same twelve months, the median worker's wage grew 3.8 percent,^[2] against inflation of 3.4 percent.^[3] The typical worker came out roughly four-tenths of a point ahead.

Neither number is spun. Neither is a revision waiting to happen. They disagree because they are answers to two different questions, and almost every argument you will read about whether life is getting more or less affordable is really an argument about which question counts.

The average job versus the same person

Average hourly earnings is exactly what it sounds like: total wages paid, divided by total hours worked, across everyone employed right now. It is a snapshot of the whole pool.

That makes it sensitive to something that has nothing to do with anyone's raise — who is in the pool. If a year of hiring adds a million jobs in lower-paying work, the average falls even if not one existing worker's pay changed. If a wave of high earners retires, the average falls again. Economists call this a composition effect. It is not a flaw in the statistic; measuring the average price of labor is a legitimate thing to want to know. It is just not a measure of what happened to you.

The Atlanta Fed's Wage Growth Tracker is built the other way around. It uses Census microdata to find the same individual twelve months apart and compare that person's hourly wage to their own wage a year ago, then reports the median of those individual changes. Nobody enters or leaves the calculation in between. Composition cannot move it.

One measures the pool. The other measures the swimmer.

If your question is "how much is an hour of American labor worth these days," use the first. If your question is "did I get a real raise," use the second — and note that the second says the median worker did, barely.

The whole margin lives in one decision

Here is where the tracker earns its keep. It splits those individuals into two groups: people who stayed with the same employer, and people who changed employers.

Job stayers: 3.6 percent. Job switchers: 4.4 percent. Inflation: 3.4 percent.[2]

Run that through a $60,000 salary and the stayer gained about $115 of purchasing power over the year. The switcher gained about $580. The stayer's "raise that beat inflation" is real, and it is roughly one tank of gas.

This is not advice to quit your job. Switching carries costs that never appear in a wage statistic — a lost tenure clock, an unvested match, a manager you actually liked, and the plain risk that the new place is worse. But it does mean something specific: if you stayed put in the last year and got less than 3.4 percent, you took a pay cut. Not a metaphorical one. Your paycheck buys less. That is information about your position in the market, and it is worth having before your next review rather than after.

One caution on the 0.8-point gap: it is a step, not a slope. Switchers get a one-time bump when they move, and then they are stayers again. It does not compound at 0.8 points a year forever, and anyone modeling it that way is selling something.

Why a winning number still feels like losing

Suppose you are the median stayer. You got 3.6 percent, prices rose 3.4 percent, you are technically ahead. You will not feel ahead. There are two reasons, and neither one is you being bad at math.

The first is the basket.

Inflation of 3.4 percent is a weighted average across everything. Underneath it, gasoline rose 24.6 percent and airline fares 25.5 percent, while used cars fell 1.9 percent and core inflation — everything except food and energy — ran at just 2.5 percent.^[3]

The index weights each item by how many dollars Americans spend on it. Your memory weights it by how many times you had to pay. You buy gasoline weekly and a used car once a decade, so the 24.6 percent is burned in and the −1.9 percent is invisible. A 3.4 percent average can feel like a lie without anyone lying.

The second reason is that a raise and a price increase do not feel like they belong on the same ledger. Harvard economist Stefanie Stantcheva surveyed Americans about inflation and found that among people who had received a raise, more than twice as many attributed it primarily to their own job performance (20 percent) than to inflation (9 percent).^[4] A raise reads as a verdict on your work. Inflation reads as theft by an anonymous force. Even when the two cancel out to the penny, you bank one as earned and file the other as stolen.

That same survey found 81 percent of respondents believe prices rise faster than wages, and 51 percent believe inflation will pad their employer's profits without touching their pay.^[4] The University of Michigan's consumer survey now finds just 8 percent of Americans expect their income to outpace inflation over the coming year, down from 18 percent at the end of 2024.^[5]

Belief and measurement have come apart. Both are worth taking seriously — the measurement because it is true, the belief because it changes what people do with their money.

Do it on your own number

Two minutes with a calculator beats an hour of arguing about the economy.

Take your raise, add one, divide by 1.034, subtract one. That is your real raise. When a newer CPI figure comes out, swap it in — the arithmetic does not change.

Benchmark it against 3.6 percent, not 3.8 percent. The 3.8 percent median includes job switchers, and the only way into that group is to switch.

Four things worth carrying out of this

Compare your raise to 3.4 percent, not to zero. Zero is the wrong baseline and it is the one almost everyone uses.

Know which statistic answers your question. When a headline says wages are falling or rising, the first thing to ask is whether it followed the same people. Most do not, because following the same people is expensive.

Falling inflation is not falling prices. A 3.4 percent rate means prices are still climbing, just more slowly, and they are climbing on top of everything they already did. Nearly every "the data says you are fine" argument quietly swaps the rate for the level. Nearly every "nothing is affordable" argument does the reverse.

One month is not a trend. Real average hourly earnings at −0.2 percent is well inside the range that month-to-month noise can produce. Watch the shape over years; ignore the shape over weeks.

Figures as of August 19, 2026.

This article is educational and is not financial advice.