A one-two-three series of natural disasters took its toll on the job market in September, providing the first negative BLS report in seven years. The US economy lost 33,000 jobs last month, even as the unemployment rate declined slightly — and, oddly, the workforce expanded. However, revisions to the two previous months paint a less rosy picture even outside of the hurricanes of September:

The unemployment rate declined to 4.2 percent in September, and total nonfarm payroll employment changed little (-33,000), the U.S. Bureau of Labor Statistics reported today. A sharp employment decline in food services and drinking places and below-trend growth in some other industries likely reflected the impact of Hurricanes Irma and Harvey.

The unemployment rate decreased by 0.2 percentage point to 4.2 percent in September, and the number of unemployed persons declined by 331,000 to 6.8 million. Both measures were down over the year.

The BLS provides a big caveat to these numbers:

Hurricane Irma made landfall in Florida on September 10—during the reference period for both the establishment and household surveys—causing severe damage in Florida and other parts of the Southeast. Hurricane Harvey made landfall in Texas on August 25—prior to the September reference periods—resulting in severe damage in Texas and other areas of the Gulf Coast.

Our analysis suggests that the net effect of these hurricanes was to reduce the estimate of total nonfarm payroll employment for September. There was no discernible effect on the national unemployment rate. No changes were made to either the establishment or household survey estimation procedures for the September figures. For both surveys, collection rates generally were within normal ranges, both nationally and in the affected states. In the establishment survey, employees who are not paid for the pay period that includes the 12th of the month are not counted as employed. In the household survey, persons with a job are counted as employed even if they miss work for the entire survey reference week (the week including the 12th of the month), regardless of whether or not they are paid. For both surveys, national estimates do not include Puerto Rico or the U.S. Virgin Islands.

In other words, the methodology for the surveys was not impacted by the storms, so BLS has confidence in the data. However, the storms did have an impact on jobs. Whether that lasts remains to be seen, but the revisions for July and August don’t signal much optimism. Combined, the two revisions cut previous job estimates by 38,000. The new three-month average is now 91,000, well below the maintenance rate. That’s because of the underwater results in September, but the two-month aggregate for July and August is only 307,000 — roughly 153K per month, barely keeping the pace.

With that said, the report does have some curious aspects, especially the Household survey and the drop in the unemployment rate. The civilian labor force grew by 575,000, and the number of Americans not in the labor force fell by 368,000. The Household survey also shows an additional 906,000 people employed from August to September, which … makes little sense at all. That inflated employment figure is the reason that the U-3 rate fell from 4.4% to 4.2% and U-6 fell from 8.6% to 8.3% while the economy lost jobs in the Establishment survey.

The Household survey is conducted separately from the Establishment survey. The former polls homes and asks respondents about their employment situations, while the latter gets payroll data from employers to estimate job gains and losses overall and in specific industries. The two surveys define unemployment differently, which produces a bit of an apples-oranges issue when it comes to calculating unemployment rates, but there’s not been this much of a disconnect in recent memory.

The AP’s Christopher Rugaber sees signs for optimism, but he relies on the odd results from the Household survey for his conclusions:

Looking past the hurricanes’ impact, the job market and economy generally look healthy. Job growth is likely to rebound in the coming months as businesses in the area reopen and construction companies ramp up repair and renovation work. …

The unemployment rate fell because it is calculated with a separate survey of households. That survey counted people as employed even if they were temporarily off work because of the storms.

As a result, the drop in the jobless rate is a sign the job market improved outside hurricane-hit areas. The proportion of adults with jobs rose to 60.4 percent, the highest since January 2009.

Average hourly wages rose 2.9 percent from a year earlier, a solid gain. But the government said that figure was artificially inflated by the loss of so many lower-paid workers in hurricane-hit areas.

It seems a little risky to rely on data that shows an increase of almost a million jobs in a month. As surveys go, it seems a little … outlier-ish. YMMV.

The best advice here is to wait for the revisions next month, and to see whether the Household survey snaps back to its previous trends. I’d expect modest job gains and a hike in unemployment for the October report. Everyone expected September to be an outlier, with the first hurricanes to hit the US mainland in twelve years, and at least that expectation got met.