It's time to weather-adjust the jobs reportFebruary 6, 2014: 2:07 PM ET
Aren't there better things we can talk about other than the weather? Let's talk about it.
FORTUNE -- On Friday, the government will report how many people were employed (and unemployed) in January. Once again, the number could be a messy one. The reason: the weather.
December's report said companies added only 74,000 positions, which was much lower than what was expected. But many economists said that was due to weather, and to ignore it. The weather in January wasn't much better, certainly colder that average. Auto dealers and real estate agents have complained that frigid temperatures slowed activity. And construction certainly slackened more than normal. Still, the week, or two weeks, depending on which of the government's two jobs surveys you are talking about, in which the Labor Department counted the number of workers, had relatively little snow.
Even so, the fact that some companies put off hiring in December because of the weather may artificially boost January's job growth. So, one way or another, it's hard to say the number won't be affected somehow. A number of Wall Street's top economists have noted how weather will or will not play into January's tally. Clearly, it's a guess, on top of all the other
guesses forecasts they make to come to their predicted number.
January will mark the third month in a row that weather has clouded the jobs report. November was warmer than usual, potentially leading to more hires than normal, which also could have contributed to December's drop.
But wait, this is all kind of silly. First of all, the significance of the weather effect may be overstated. The jobs number that everyone quotes each month comes from the survey of employers. In that survey, all you have to do is work one day of the pay period in question in order to get counted as employed. And, according to the BLS, only about 22% of the workers in its survey are paid weekly. So weather would have to stop them from getting to work for two full weeks for it to really have an effect on the number. That's like a Hurricane Katrina or Sandy event, not a snow storm or a cold front. What's more, as Mark Zandi of Moody's Analytics has pointed out, about 20,000 fewer accountants were employed in December than the month before. It's hard to blame that on the weather.
But let's ignore all of that, because everyone else seems to, and assume that weather affects the numbers. If that's the case, I have a proposal: Why doesn't the government adjust the jobs number based on the weather?
MORE: Did the Fed screw up?
The government already seasonally adjusts the jobs numbers. But those adjustments are really about seasonal industry trends and not necessarily weather. So the BLS adjusts for more retail and fewer construction jobs in November and December. And those adjustments are all put into place in advance of the actual month, before we know what the weather will bring. And they are heavily based on hiring patterns from the past few years. So get one year of especially wacky weather, which we're experiencing more and more often, and all the adjustments get thrown off.
What's more, the government is already collecting some data on how the weather affects employment, so why not use it? Here's how I would do it: Each month, the government asks individuals if they have a job but weren't able to get to work because of the weather. (Those people are counted as unemployed.) Last month, 273,000 people responded yes to that question. That was higher than usual. The five-year December average is 181,000. So the number of additional people who weren't able to get to work because of last month's unusual weather came to 92,000.
The problem is that each month the government conducts two employment surveys. The weather question comes from the survey that goes out to individuals. The widely quoted number of jobs added each month, which was 74,000 in December, comes from the payroll survey. Historically, the payroll survey number is, on average, about 7% lower than the household survey number. Factor that in, and the weather-adjusted payroll number should include an additional 87,000 workers in December. (Zandi says his models point to 50,000, so we are in the same ballpark.) Based on this line of thinking, the December jobs number should be 161,000.
But we're not done yet. You can't figure out how many jobs were added in December without knowing how many weather-adjusted jobs we had in November. And remember, the weather in November was quite lovely. And that, according to my math, means that 78,000 more people landed jobs than should have. So you have to take those jobs out. Bottom line: On a weather adjusted basis, the economy added 239,000 jobs in December.
Is that the right number for job growth in December? GDP rose 3.2% in the fourth quarter. So 239,000 feels more in line with that fact than 74,000. But who knows? What's clear is that not adjusting for the weather makes an already messy number even messier. One seasoned Wall Street economist said he and his colleagues should take November to March off. The winter jobs numbers have basically become useless, so taking a vacation might offer one solution. But then we would also have to adjust for all those out-of-work economists. And no one wants to do that.