It was a holiday-shortened trading week across North America, with Canada Day on July 1 and the U.S. observing the Independence Day holiday on Friday, July 3.
During the final couple of weeks of June, selling pressure accelerated across precious metals and miners. This was primarily the result of quarter-end window dressing; at the end of every quarter, portfolio managers often sell holdings that performed poorly during the previous three months and look to add stocks that outperformed.
At the end of Q2 2026, this meant selling gold and gold miners while buying semiconductor stocks and AI-related plays such as CAT and TSLA.
Thursday’s U.S. nonfarm payrolls report delivered a big downside surprise, including downward revisions totaling 74,000 jobs for April and May:
There are mounting signs that the US labor market may not be as strong as many would like us to believe. In recent months, the household employment survey has strongly diverged to the downside from the establishment survey:
The key difference between the two employment surveys is the household survey counts people, while the establishment survey counts jobs. In the household survey, a person with two jobs is counted once as employed. In the establishment survey, that same person can be counted twice if both jobs are on nonfarm payrolls.
The household survey is where the unemployment rate comes from. It includes broader categories of workers, such as the unincorporated self-employed, agricultural workers, unpaid family workers, private household workers, and people on unpaid leave. That makes it better for understanding the labor force, demographics, unemployment, participation, and employment-population ratio.
The establishment survey is where the headline nonfarm payrolls number comes from. It is generally considered more reliable for month-to-month job growth because it has a much larger sample and a smaller margin of error. BLS says a monthly change of about 122,000 jobs is statistically significant in the establishment survey, versus about 650,000 in the household survey.
The surveys can diverge because they have different coverage. The establishment survey excludes the self-employed, most agricultural workers, unpaid family workers, private household workers, and people on unpaid leave, while the household survey includes many of these groups. The establishment survey also counts payroll jobs during the employer pay period that includes the 12th of the month, while the household survey generally uses the calendar week that includes the 12th.
The downward revisions to April and May could be confirming the downtrend in the household employment data.
