I’m doing some work looking at commute mode over the course of the day, and I see a big jump in Other mode usage between 12:00 AM and 1:00 AM—in the 2022 ACS, 18% (!) of (weighted) respondents reported this mode at that time. I don’t think it’s just an issue of a small number of individuals being weighted up heavily—there are just over 1000 individuals who report arriving at this time by “other” mode. These individuals don’t seem to have anything particular in common among the industries they work in, their ages, their incomes, or the states they live in. Does anyone have any idea what this “other” mode might be? I feel like it must be something obvious that I’m just missing.
This isn’t a data error, at least not one introduced by IPUMS—I confirmed that this same trend is present in the original Census bureau data for the 2022 PUMS. There’s a similar trend in the 2021 ACS, but not in 2019 and prior.
I looked into what you are describing and I agree that it looks odd. I created a data extract that includes ACS data from 2015-2022 and both the IPUMS variables TRANWORK, TRANTIME, DEPARTS, and ARRIVES and their corresponding original source variables. I also included the data quality flags for DEPARTS (QDEPARTS), TRANTIME (QTRANTIM), and TRANWORK (QTRANWOR). These flags identify when a variable’s value has been allocated.
First, I see that these unusually high rates of “other” responses to TRANWORK for those leaving for work at 12:15 am (DEPARTS=15) compared to other departure times are present in the source (original) data from the Census Bureau. Second, I see that about 50 percent of respondents with both DEPARTS=15 and TRANWORK=70 (other) had one or both of these values allocated in the 2022 ACS. This is higher than the allocation rate of about 30 percent for respondents with DEPARTS!=15 and TRANWORK!=70 in the 2022 ACS. However, among those with neither DEPARTS nor TRANWORK allocated, there is still a similarly high rate of “other” responses to TRANWORK for those with DEPARTS=15.
It is possible that the Census Bureau may have performed other edits to either TRANWORK or DEPARTS that affected one or both of these variables. It is also possible that these values of TRANWORK, DEPARTS, or both, are errors introduced during the data transcription, editing, or recoding processes. I have not found any Census Bureau documentation referencing intentional edits of either variable that would explain their unexpected association. I am sorry that I cannot provide a definitive explanation for what you are observing. You could certainly reach out to the Census Bureau to ask if they are aware of any known issues or edits with either or both of these variables. We would be curious to know if you uncover any more information.