(Originally published here.)
Weather has been awful across much of the U.S. this winter. “Thundersleet” traumatized the Northeast yesterday, brutal cold has gripped the Midwest and California is withering in a drought. Maybe it’s a coincidence, but employment growth, car sales, home sales, capital goods orders, industrial production, and retail sales have all been weaker than economists’ forecasts. If the weather is to blame, it could skew our ability to tell how the economy is doing for the next few years.
The reason has to with something called seasonal adjustment. Spending, hiring and investment aren’t constant across the days of the week and the time of year. For example, Americans shop more in November and December than they do in other months. Employment tends to be more volatile in the summer as students take and then leave temporary jobs.
In order to filter out this noise, the analysts in the government and elsewhere use algorithms that compare new information against historical norms for that same time of year. This helps government agencies, such as the U.S. Bureau of Labor Statistics, smooth out the bumps and gives us a clearer picture of what’s actually going on. The general methodology assumes that seasonal patterns don’t change too much from year to year.
Most of the time, this works pretty well, although there is good reason to thinkthat the algorithms got messed up by the Great Recession, which caused the greatest damage in the winter of 2008-2009. According to Johns Hopkins economist Jonathan Wright, the timing of the economic slowdown confused the computer into thinking that U.S. winters had suddenly become a lot worse. As a result, job gains in subsequent winter months were consistently overstated, while job gains in the summer months looked weaker than they really were.
This effect has faded over time, but the algorithm remains vulnerable to extreme events. If the recent bad weather did hurt the economy, then next winter’s statistics might end up looking a bit stronger than they really are. Of course, the effect should be much smaller than what Wright observed because of the scale and scope of the recession spawned by the financial crisis.
And if the U.S. starts getting thundersleet every year, at least we won’t have to worry about the impact of this year’s storms on the algorithm.
(Matthew C. Klein is a writer for Bloomberg View. Follow him on Twitter at @M_C_Klein.)
To contact the writer of this article: Matthew C. Klein at email@example.com.
To contact the editor responsible for this article: James Greiff at firstname.lastname@example.org.