Unusual proportion of part-time workers due to holidays in Sep. 2015?

I want to produce a monthly full-time vs. part-time rate series with monthly CPS base samples. I noticed that the full-time employment rate appears surprisingly low and part-time employment rate appears surprisingly high in Sep. 2015. A closer examination reveals that an unusual number of workers reported to be on holiday; see table below:

Sep. 2015

In comparison, the normal range for number of workers reporting to be holiday is anywhere between 0 and 2000. As examples, in Aug. & Oct. 2015, 22 and 1,822 workers reported to be on holiday, respectively.


  1. Could this be a coding error in the data cleaning process?
  2. If not coding error, are you aware of events that could explain this pattern?


I suspect this is the result of Labor Day falling on September 7 in 2015. Labor Day is always on the first Monday in September; September 7 is the latest possible date this holiday can occur. When the holiday occurs on September 7 it overlaps with the CPS reference week, which is the week that includes the 12th. This is not the case when Labor Day occurs on September 1-6. There is a similar spike in 2009, the last year prior to 2015 that Labor Day also occurred on September 7 (see below). The variable WKSTAT seems robust to this anomaly if it is appropriate for your analysis.

This makes sense.

Really appreciate your response!

1 Like