Real Hourly Pay Fell 0.2% Over the Year. Real Weekly Pay Rose 0.1%. Both Numbers Are in BLS Table A-1.

by Tomas Ferreira · August 17, 2026

Markets Reporter, US Market Current

The Bureau of Labor Statistics published its Real Earnings release for July 2026 on Wednesday, August 12, at 8:30 a.m. Eastern, under release number USDL-26-1379. Table A-1 of that release, titled "Current and real (constant 1982-1984 dollars) earnings for all employees on private nonfarm payrolls, seasonally adjusted," carries two over-the-year figures that point in opposite directions. Real average hourly earnings fell 0.2 percent from July 2025 to July 2026. Real average weekly earnings rose 0.1 percent over the same twelve months.

Both readings are in the same table, cover the same twelve months and rest on the same deflator. What separates them is a single line in between: average weekly hours. Table A-1 shows the workweek for all employees on private nonfarm payrolls at 34.3 hours in July 2026 against 34.2 hours in July 2025, and its own over-the-year line for average weekly hours reads 0.3 percent. BLS draws the connection itself in the release summary, which says the change in real average hourly earnings “combined with a 0.3-percent increase in the average workweek resulted in a 0.1-percent increase in real average weekly earnings over this period.” On the published, rounded figures the same arithmetic closes by inspection — minus 0.2 plus 0.3 equals plus 0.1 — and that last step, the addition, is ours.

In constant 1982-1984 dollars — the units Table A-1 uses for its real series — real average hourly earnings were $11.30 in July 2026 against $11.32 in July 2025. Real average weekly earnings were $387.72 against $387.15. By our subtraction, that is two cents less per hour and 57 cents more per week than a year earlier. In current dollars, the same table puts average hourly earnings at $37.62 versus $36.47, an over-the-year change it reports as 3.2 percent, and average weekly earnings at $1,290.37 versus $1,247.27, an over-the-year change it reports as 3.5 percent.

Before any of those directional statements can be used, it is worth being explicit about which column they stand on. This release publishes seasonally adjusted data only. Its table of contents lists exactly two tables — A-1 for all employees and A-2 for production and nonsupervisory employees — and the word "seasonally adjusted" appears in the printed title of each. There is no unadjusted companion table inside the Real Earnings release. Every figure quoted above is therefore a seasonally adjusted figure, and there is no raw column in this document to disagree with it.

That distinction matters most for the price index sitting inside the earnings table. The CPI-U row of Table A-1 reads 332.813 for July 2026 on the 1982-84 base, and the table's over-the-year line for that row reads 3.3 percent. The Consumer Price Index release that BLS issued the same morning, under release number USDL-26-1378, puts the all-items index at 333.918 on the same base and the over-the-year change at 3.4 percent, an increase it reports “before seasonal adjustment.”

Neither number is wrong, and neither release explains the gap. The Real Earnings technical note names the CPI-U as the deflator for the all-employee series and the CPI-W as the deflator for the production-worker series, but it does not say whether either index is used in seasonally adjusted form, and the CPI release publishes no seasonally adjusted index level to set beside 332.813; we looked in both releases and in the technical note. What can be checked is the monthly line. Table A-1 puts the over-the-month change in its CPI-U row at 0.1 percent, which is the figure the CPI release reports on a seasonally adjusted basis, while that release says the index was unchanged prior to seasonal adjustment over the same month. On that evidence, and on a table whose printed title reads seasonally adjusted, our reading is that the deflator inside Table A-1 is the seasonally adjusted CPI-U. That is an inference from the two documents, not a statement BLS makes. The other half of the comparison is firmer: the CPI release itself reports its 3.4 percent as a 12-month increase before seasonal adjustment, and the Real Earnings technical note says seasonally adjusted data are used for the release's over-the-year percent changes in current and constant earnings. Subtracting that 3.4 percent headline rate from the 3.2 percent nominal earnings growth in Table A-1 therefore mixes two different bases.

The same split runs through the production-worker table. Table A-2 deflates by the CPI-W rather than the CPI-U, and its CPI-W row reads 325.850 for July 2026 with an over-the-year change of 3.3 percent. The CPI release puts the CPI-W all-items index at 327.104 with a 12-month change of 3.4 percent, and prints no seasonally adjusted index level for that series either. Two BLS releases embargoed until the same minute — both until 8:30 a.m. Eastern on August 12 — carry two different July index levels for the same price series, and neither is in error.

For production and nonsupervisory employees, the earnings data take the same shape as for all employees, with a smaller hourly decline and a slightly larger weekly gain. Table A-2 shows real average hourly earnings for that group at $9.94 in July 2026, an over-the-year change of minus 0.1 percent, and real average weekly earnings at $336.08, an over-the-year change of plus 0.2 percent. Average weekly hours for the group stand at 33.8 against 33.7 a year earlier, an over-the-year change the table reports as 0.3 percent. Again the hours line is what turns a negative hourly reading into a positive weekly one.

The July hourly figure is negative, but it is not the weakest of the year, and it is worth recomputing that before anyone calls it one. May 2026 printed an over-the-year change of minus 0.8 percent in real average hourly earnings, and it printed that figure in both the July 14 release and the August 12 release. April 2026 printed minus 0.3 percent in the July 14 release; the August table prints only four month columns — July 2025, May, June and July 2026 — so April is available only from the earlier release. Across the four months April through July as those months were published — minus 0.3, minus 0.8, 0.0 and minus 0.2 percent — July's decline is the smallest of the three negative readings, and June is the only one of the four that is not below zero.

The one month in that stretch that read positive no longer does. In the Real Earnings release for June, published July 14, 2026, Table A-1 showed real average hourly earnings up 0.1 percent over the year in June. In the August 12 release, the same June figure reads 0.0 percent. The revision runs through the nominal series as well: June average hourly earnings were first published at $37.64 and now read $37.60, and June real average weekly earnings were revised from $388.21 to $387.79.

A second revision moved in the other direction, and it landed on the workweek. In the July 14 release, average weekly hours for production and nonsupervisory employees in June read 33.7, and Table A-2 reported the over-the-month change in that line as minus 0.3 percent. In the August 12 release, June reads 33.8 hours and the over-the-month change reads 0.0 percent. June real average weekly earnings for the group were revised up from $335.10 to $335.88 — revised higher for production workers in the same release that revised the all-employee June weekly figure lower.

The month-to-month movement in July itself was small on every line. Table A-1 reports real average hourly earnings down 0.1 percent from June to July, with nominal average hourly earnings up 0.1 percent and the CPI-U up 0.1 percent over the month. Real average weekly earnings were unchanged over the month, and the over-the-month change in average weekly hours was 0.0 percent. On the production side, real average hourly earnings were unchanged from June to July and real average weekly earnings rose 0.1 percent.

July looks flat mainly because June did not. In the revised August 12 data, real average hourly earnings rose 0.7 percent from May to June — the largest month-to-month move in that line among the four months Table A-1 prints — because the table's CPI-U row fell 0.4 percent that month while nominal hourly earnings rose 0.3 percent. That June jump was itself trimmed in revision, from the 0.8 percent originally published on July 14. July gave back a tenth of it.

What the release does not contain is as relevant as what it does. Real Earnings is two tables, a summary and a technical note; it carries no industry breakdown, no regional detail and no adjustment for shifts in the composition of employment, which can move an economy-wide average hourly earnings figure without any individual worker's pay changing. The release makes no attribution for the movement in its own series beyond the arithmetic of earnings and prices, and neither do we.

The next installment, covering August 2026, is scheduled for Friday, September 11, 2026, at 8:30 a.m. Eastern, according to the July release. Until then the operative statement from this data is narrow and worth stating in full: on a seasonally adjusted basis, the average private-sector worker's real hourly pay in July was slightly below where it stood a year earlier, the real weekly paycheck was slightly above, and the difference was a workweek one-tenth of an hour longer than it was in July 2025.