Policy makers look to the employment numbers to gain insight regarding economic trends: are more people seeking work? Are more businesses hiring? Tracking these trends can better inform economic policy making. However, in a recent WSJ Op-Ed, Daniel Quinn Mills, a professor of business administration at Harvard took issue with how the BLS data are reported.
Employment data is often impacted by seasonally-recurring phenomena. For example, the 295K “jobs added” figure for the month of February is for “nonfarm employment,” because as one might expect, agricultural hiring spikes in the summertime and falls in the winter. In other words, tracking long term trends requires accounting for predictable fluctuations – a process known as seasonal adjustment.
Enter Mills, who thinks the practice of seasonal adjustment is misleading and outdated:
Seasonal-adjustment factors were developed seven decades ago for an economy that predominantly produced goods. Construction shut down in winter and automobile manufacturing closed in January. So employment fell and the monthly estimates were adjusted upward for an annual rate. Today’s economy is a service economy, with far less weather-related employment variation. For example, when seasonal adjustment began, total American employment fell almost 10% due to bad weather in the winter months. Now it is less than 2%.
January’s seasonal adjustment implies that reduced levels of employment and output will automatically return to higher levels when spring arrives. There will be nothing automatic about the return of the almost three million jobs lost. In fact, the largest declines in employment were in retail trade and professional service, neither of which is weather related.
It is useful for policy-making purposes to adjust monthly data to an expected annual rate. But the current method needs to be updated and based on something other than the weather. Because the economy has changed so much over the decades, we don’t understand what reality, if any, current seasonal adjustment represents. Seasonal adjustment is probably a statistical reflection of underlying changes in the economics of various industries, and it is therefore distorting exactly what investors and the Fed want to know about.
[…] seasonal adjustment isn’t entirely, or even primarily, about the weather. It’s about accounting for recurring patterns, whatever they may be. Tax preparation firms hire lots of people every spring and then lay them off after April 15. Landscaping firms employ far more people in the summer than in the winter. Automakers shut down their factories each summer to change over to the new model year.
Know what else happens every year? Christmas. Far from being immune to seasonal patterns, as Mills apparently believes, the retail sector is among the economy’s most predictable industries. Every November and December, even in the depths of the recession, retailers add hundreds of thousands of workers to handle the holiday rush. And every January, they lay them off.
Knowing that retailers cut jobs in January, in other words, tells you precisely nothing about the state of the economy. What you need to know is whether that cut was bigger or smaller than you’d expect — exactly the question that seasonal adjustment attempts to answer.
To a reader who comprehends the objective of reporting data on long-term employment trends to better inform policy making, this makes perfect sense. A huge spike in employment in December, followed by a huge drop in employment in January, is more indicative of holiday shopping (which occurs every year) than a long-term trend.
Moreover, the BLS is transparent about how their data is reported. By titling their second table on the opening page of their jobs report “Nonfarm payroll employment over-the-month change, seasonally adjusted, February 2013 – February 2015,” they are acknowledging the limitations of the data. Additionally, Mills wants the BLS to report non-seasonally adjusted data – which they do in the link above (Table A-2 for household data). Mills seems to have a bone to pick with the BLS, but by going after them, he showcases his gross misunderstanding of their very data in the process.