Posted by:
Henry Bemis
(
)
Date: March 28, 2024 12:11PM
Very interesting. Thanks for posting.
I was unable to find any actual published study by Devin Pope supporting the linked article. Specifically, I was curious (and skeptical) as to the reliability of the methodology used. (Sometimes, the enthusiasm of the social sciences for a given idea or technology outrun its credible application.)
In the present case, I did find a similar study by Pope that identifies his general smart phone tracking methods:
http://devingpope.com/assets/files/Racial_Disparities_in_Voting_Wait_Times.pdfNote the following very basic description of the process, which presumably is similar to the study identified in your link:
"The three primary datasets that we use in our paper include: (1) SafeGraph cell phone location records, (2) Polling locations, and (3) Census demographics. We use anonymized location data for more than 10 million U.S. smartphones provided by SafeGraph, a rm that aggregates location data across a number of smartphone applications (Chen and Rohla 2018)."
"Pings are recorded anytime an application on a phone requests information about the phones location. Depending on the application, this can happen when the application (e.g. a navigation or weather app) is being used, or can occur at regular intervals when the application is in the background. The modal time between pings in our sample for a given device is 5 minutes."
Note also the admitted limitations:
"However, it also has its limitations. One key limitation is that the sample is an imperfect representation of the overall U.S. population. Given the nature of the data, our sample can by construction only be representative of the population of approximately 77% of U.S. adults who owned a smartphone in 2016. Additionally, we will be restricting our sample to phones that receive regular pings and thus will disproportionately remove individuals who turn off their phones for extended periods of time, do not allow location-tracking services on their phone, or live in areas with poor cell phone coverage."
Applying these limitations to the current study, it would appear that the accuracy of the cell phone tracking would importantly depend upon (1) Whether the church attenders had smart phones; (2) Whether they took their phones with them to Church; (3) Whether their smart phones were actually on; (4) Whether their smart phones were set to geographical tracking; and (5) Whether the location of the Church was sufficiently urban to allow such effective tracking.
Intuitively, it seems that these limitations were seriously undermine any study of Church attendance, which may be why the study was apparently never published. Although the vast majority of the population now have smart phones, one can assume that the anti-hacking, and anti-tracking sophistication of smart phones has also increased, perhaps making statistical tracking more difficult as well.
Moreover, one is probably more likely to leave their smart phone home, or turn it off in a Church setting (as compared to voting). Also, since Church attendance involves a substantial number of older persons not smart phone literate (even if they have one), it just seems like a reach that any meaningful data could be collected by this method in this context.
In any event, I am not sure that this methodology is any better than the old-fashioned survey method, as weak as that method is.