I’ve had a suspicion, but I haven’t tested it...

My suspicion is: Some share of people misunderstand ACS question #32 and decide that “rode to work” does not include the driver.


I asked Census Bureau staff last winter… They don’t know why.





From: Sarah K Heimel (CENSUS/DSSD FED) [mailto:Sarah.K.Heimel@census.gov]
Sent: Tuesday, January 06, 2015 5:33 PM
To: Graham, Todd
Subject: Re: ACS questionnaire wording



Hi Todd,


I hope you had an enjoyable holiday.  I spoke with the Census subject matter expert on commuting statistics about your question from last month.  He said that they are aware of the same anomaly that you and the transportation planners have seen, which has apparently been appearing in the data for a while now, at least since 2000.  He thus ruled out my hypothesis of the emerging car-sharing option. 


In 2006, Census modified the allocation algorithm slightly to jointly assess these two variables when in need of allocation.  Still, there seems to be a lot of people who legitimately chose both no vehicles available and drove alone.  We are not currently aware of cognitive test results showing an issue with the wording on Question 32.  This is on the list of things to research, to hopefully gain insight into who these people are and what the confusion might be.  For now though, I unfortunately cannot give you an answer. 







Sarah K. Heimel

Mathematical Statistician

Decennial Statistical Studies Division

U.S. Census Bureau



From: Graham, Todd <todd.graham@metc.state.mn.us>
Sent: Thursday, December 18, 2014 11:21 AM
To: Sarah K Heimel (CENSUS/DSSD FED)
Subject: ACS questionnaire wording


By the way Sarah… is there a team in ACSO that evaluates question wording and quality of response accuracy (or: risk of respondent error)?

If so can you forward this on?


I was asked earlier this month about commute mode cross-tabbed with “number of vehicles.”  Transportation planners are asking: how are there so many workers driving to work alone when they live in zero-vehicle households?


I speculated: there could be some respondent error accruing to one of the JTW questions?


The commute mode questions on the questionnaire are:

·         #31. How did this person usually get to work LAST WEEK?

·         #32. How many people, including this person, usually rode to work in the car, truck, or van LAST WEEK?


If a respondent perceives a difference between being a passenger vs a driver – and thus “rode to work” vs “drove to work” – then he/she could underreport the number of vehicle occupants by -1. 


What does ACSO think about respondent error on question #32.

Are there high-occupancy carpools that are being misreported as single-occupancy?


Let me know if there’s any answer – thanks for your attention to this.



Todd Graham

Principal Forecaster  |  Metropolitan Council   |  Regional Policy and Research


P. 651.602.1322  |  F. 651.602.1674

390 North Robert Street  |  St. Paul, MN 55101  |  metrocouncil.org/data

ConWUsFaceBook_32x32Twitter_32x32Youtube_32x32 EmailSU