Break in UHRSWORK 2000-2001

USA
1

Thank you very much Matthew, I understand now!
I would have another question: I am analysing how (unintervalled) weekly hours worked varies over time, using Census data until 2000 and ACS afterwards. I noticed that in 2001, the year in which I start using ACS, there is an upward blip in the trend. Is this “normal”/is there anything I should do to make the two data sources more comparable?
I can attach a figure if it helps.

Thanks a lot!

2

There are a few differences in the question that could cause a break between Census 2000 and ACS 2001. You can see the exact text for each year at this page. I noticed two things:

  1. Census refers to “last year” and ACS refers to “last 12 months” - doesn’t seem like it would cause a big difference

  2. Questions are worded and bolded differently. This may cause some difference.

There are also differences in the mode of enumeration and sampling that might cause some discrepancies. How big is the blip you are seeing?

3

Thank you Matthew. The blip is not small, for example the average weekly hours of men move from 33 in 2000 to 35 in 2001. The jump is smaller for women (as you can see in the immage uploaded)

4

I’d suggest a couple more things to see what is causing this:

  1. check 2000 Census vs 2000 ACS
  2. check the series from CPS over this time period.

2001 was also a recession year so there could be some impact from that.

5

Thanks. I thought about using ACS for 2000 but the number of observations is much smaller than the usual ACS so I am a bit concerned about that, do you think it can be a problem?

6

The smaller number of observations will reduce the precision of your estimates and might make a difference depending on what you’re calculating. But for the purposes of checking this one series I wouldn’t worry about it too much. You can get a confidence interval for your estimate using any standard statistical package. See this page for more information on this and sample code for Stata and SAS.