ACS Commute Flow Data measurement of time

I'm working with the commute flow data set from 2012-2016, specifically for census tracts. I have not been able to find any information regarding what unit of time the estimates are measured over. I have looked through the documentation file for the data (here) as well as through the technical documentation, methodology, and the FactFinder (here), but have not been able to find it. Is the estimate yearly? Daily? For all five years in the data set?


Thank you!

  • Hello Ford -

    Which commute measure(s) are you using, specifically?
    ACS is in the field continuously, so data from a 5-year file reflect the entire period BUT specific questions are asked differently. Commute mode and place-of-work location, for example, refer to where the person worked LAST WEEK.

    For more info, try the 2016 questionnaire
  • In reply to Beth Jarosz:

    Hi Beth! Thank you for the response and linking to the questionnaire, that provided some good context.

    I am looking at the commute flow data from one census tract to another, specifically the total amount of people and the number of people who drive alone to work in a census tract. I was looking at one tract in Texas that had ~70 million people commuting into it from all other census tracts, and that caused my interest in what the amount of time the measurements cover is.

    It sounds like I should assume the ~70 million number and others computed the same way cover the entire five year period since the ACS is in the field continuously over that period. Is that reasonable?
  • In reply to Ford:

    Hi Ford,

    I might suggest you look at the Census's Longitudinal Employer-Household Dynamics (LEHD) data, which has the information you're looking for. Unlike ACS, it's based on administrative records, not survey data, and it's available at the block, rather than block group or tract, level (which is both amazing and a logistical challenge on account of how much data it includes). Furthermore, the data represents a single year, rather than a range, which could make it more useful in some circumstances.
  • In reply to Bernie:

    ~70 million commuters into one tract? That seems... implausible. The place of work question is place of work the prior week, so the distribution should roughly match employment flows in an average week across that 5-year sample period--NOT the whole 5 years.

    I know TX likes to be biggest in all things, but I don't know of any tract that has inflow/outflow of 70,000,000 workers. I wonder if there might be a weighting or coding issue with the dataset you're working with?

    Also, since what you're looking for is flows, I second the recommendation to use LEHD data!