The Winter of Our Content
How this winter’s unseasonably warm weather jump-started job creation.
Photograph by Justin Sullivan/Getty Images.
Everyone likes a mild winter, but perhaps no one likes it more than President Obama. A recent Macro Musings newsletter from the respected forecasting firm Macroeconomic Advisers suggested that warm weather in December, January, and February added 72,000 extra jobs to the U.S. economy. This report helps explain one of Recovery Winter’s minor puzzles: Why did the economy add an unusually high level of jobs relative to very modest growth in GDP? But it also leaves us with a question: If unseasonable weather boosted winter job growth, would a return to normal this spring undo those gains? Macroeconomic Advisors thinks it will, but there’s reason to believe they’re wrong.
It’s no secret that the economy is seasonal, in part for weather reasons. Employment at ski resorts is obviously highest in the winter. Many businesses in beach and vacation communities keep longer hours during the summer. That means part-time employees start working full-time, and new workers get hired to fill the needed shifts. These sorts of issues, along with the predictable Christmas retail cycle, are at the core of the “seasonal adjustments” that the Bureau of Labor Statistics applies to jobs data. According to the BLS, for example, more than 2 million fewer people were working in January 2012 than in December 2011 which was reported in the press as a hiring surge of 284,000 new jobs.
The discrepancy between the numbers discussed in the press and the actual reality is the result of the seasonal adjustment. Nearly 600,000 jobs lost in “retail trade” from December to January, for example, were transformed into 26,000 new jobs via the magic of statistical adjustment.
This sounds a little fishy, but the system actually works quite well. These fluctuations—stores hire people for the Christmas rush and then lay them off in January—are fairly predictable and the BLS is good at doing its job. But it does leave the system open to possible errors induced by the failure of predictable things to happen. For example, between December and January we saw an unadjusted decline of 280,000 people working in the construction industry. With the seasonal adjustment factor in place, that turned into 21,000 new jobs. What this really means is that construction employment declined more slowly in January than the BLS’s statistical model taught it to expect. One major reason why the BLS was expecting a construction decline is that frigid weather tends to slow or halt construction projects. If you get an unusually warm January, you get an unusually small decline in construction employment and a better headline number out of the seasonally adjusted series.
What’s more, as a recent BLS publication emphasized (PDF) it’s not unusual for people to miss a day or two of work because of bad weather, and these incidents are concentrated in the wintertime. Citing data from 1977 to 2010, BLS says that in the average January almost 2 percent of workers miss a day on the job, and about 1.5 percent do so in February. From April through November, fewer than 0.5 percent of workers miss a day for weather. But this past winter we not only saw unusually warm temperatures, but also unusually low levels of weather-related work disruptions.
Macroeconomic Advisers built a model using temperature deviation and deviation from normal levels of work disruption as their data points for bad weather. By correlating historical data about bad weather with historical data about employment growth, they are able to create a weather adjustment to the standard seasonal adjustment model. They conclude that mild weather boosted employment by 29,000 in December, by 35,000 in January, and by 1,000 in February, adding up to a substantial total job boost. They also find 7,000 weather-related bonus jobs already in place by November, adding up to 72,000 in total. I wouldn’t take the precise figures too literally since the model hasn’t been rigorously tested on out-of-sample data over the years, but it illustrates the scale of the effect.
Weather also helps explain why the winter saw the reverse of a jobless recovery, a time period in which employment growth was faster than you’d predict based on the GDP numbers.
The reason is that one man’s utility bills are another man’s income stream. When mild weather leads to more pleasant days and less use of your home heating system, that’s nice for you, but terrible for the utilities you buy heat from. But precisely because unusually warm weather is neither predictable or likely to repeat, utilities don’t shrink their workforce in response.
What about the future? Using a similar historical modeling exercise, Macroeconomic Advisers concludes that the warm weather employment boost will fade out if conditions return to normal. But here I think they may have gone awry. The problem, as with many analyses of the present-day economy that are based on historical data, is that we’re in a unique situation. Normally the Federal Reserve attempts to set interest rates at a level where desired savings matches desired investment and willing workers can find jobs. Under that kind of policy regime, temporary shifts in weather can only have a temporary impact. But today that balancing interest rate would be negative and has been for some time, so the economy’s been in an extended funk. Instead of operating in correct balance, we are slowly groping toward recovery as population growth and capital depreciation lead to a gradually increasing desire to invest. Under these circumstances, a run of good luck isn’t necessarily balanced out by a return to normal. Arguably it represents a permanent bonus that simply pushes us back to normal conditions more quickly.
The bad news is that the same calculus applies to the risk of bad weather going forward. In other words, if you’re hoping for a return to full employment, pray we don’t have any hurricanes this summer.