What difference does seasonality make in the unemployment rate over time?
If you’ve spent time graphing in FRED, you’ve run across the term “seasonally adjusted.” Seasonal adjustment uses mathematical algorithms to adjust data for predictable seasonal variations using the changes between new observations and those from a year prior. When analyzing long-term trends in data such as unemployment, which tends to increase in January and June, seasonal adjustment helps economists see the cyclical trends underlying month-to-month patterns caused by school schedules, holidays, harvests, and a plethora of other factors. For more information on seasonal adjustment, see this explanation from the Federal Reserve Bank of Dallas.
The difference between the seasonally adjusted unemployment rate and the original is presented in the top graph, where subtracting the raw data from the adjusted data shows the impact of seasonality on the data over time. Overall, seasonal adjustment made a bigger difference in the unemployment rate in the 1950s than it has in the past several decades because of the recent trend toward year-round employment. The distribution of workers across various industries has shifted significantly over the past 60 years, and different industries are subject to varying degrees of seasonal variation.
The graph below shows the overall declines of employment in two U.S. industries: Agriculture, which hires more workers during harvest seasons, employed 4.4% of workers in 1970 but only 1.2% in 2012. Likewise, manufacturing employment has declined by 16.1% over the same span of time and also employs workers based on seasonal factors and fluctuating production needs. By contrast, the Bureau of Labor Statistics reports that, between 1939 and 2015, employment in the private education and health services sector, known as the least cyclical sector in the economy, grew from 4.6% to 15.6% of all nonfarm jobs. Other sectors that have grown in employment share include retail trade and financial activities, which employ most workers year-round and thus do not contribute to seasonal variation as much as the industries that have declined.
How these graphs were created: For the first graph, search for “monthly unemployment rate U.S.” and select the seasonally adjusted series. Click “Add to Graph.” Under “Edit line 1,” add another series by searching for “monthly unemployment rate U.S.” again and selecting the not seasonally adjusted data. Click “Add.” In the formula tab, enter a-b and click “Apply.” For the second graph, search for “percent unemployment in agriculture” and select the relevant series. Click “Add Line” in the “Edit Graph” page and search for “percent unemployment in manufacturing” and select the relevant series.
Suggested by Maria Hyrc.
View on FRED, series used in this post: