Full Size Image
Available for download here (Its a 50mb PDF, it will take a minute to load!)
Highlights
The power of this one was in the data processing, so the visuals could use some work. This is the legend up close. Nondenominational is the catchall/”other” of the dataset; its where I parked things I couldn’t identify.
The downtown core. White Catholics to the north and Hispanic Catholics in west Oak Cliff, Baptists in South Dallas, east Oak Cliff, and PG, and an intense Pentecostal cluster in West Dallas.
Look at all those evangelicals in the burbs! This area is at the core of DFW fascist organizing. Orange is confirmed evangelical, the light green is non-denominational. What’s interesting too about this area is that its not an all white vibe- lots of East and South Asian immigrants with high incomes. Evangelicalism as a cultural glue for a wealthy suburban multiculturalism? White racial radicalism as a release valve for expansion of the bounds of whiteness? Its been, ahem, done before.
Mid cities. Interesting blend of Catholic, Baptist, and nondenominational. No real guess on the lack of Catholics in Pantego and dumbly named Dalworthington Gardens.
About
For a fleeting window during COVID, companies that sell cellphone location data were so desperate enough for exposure that they offered detailed data products to researchers for free. All sorts of papers were coming out, some of them dumb as hell, modelling lock down impacts on mobility and the virus spread. These companies wanted their data used to drive traffic to their product, so they were offering academics free access. Through the clever use of a ‘.edu’ email I’ve maintained since undergrad, I managed to convince the kind folks over at one of these companies that I was an affiliated researcher studying mobility changes from lock down rules. I was not. But that didn’t stop them from giving me 3 years of extremely detailed aggregated cellphone data. It was broken down into POIs, and then counts of activity from Census Blocks to each POI for whoever was unlucky enough to have their phone caught up in the surveillance dragnet.
I looked at a few things out of curiosity. I reconstructed commute patterns for businesses on Cedar Springs to build a gayborhood commute shed. I investigated the distribution patterns of jails and detention centers. But what I spent the most time on was an examination of religious affiliation. Dallas, as many know, is a hotbed of theocratic policy innovation. So I was thinking, what is the relationship between church affiliation, politics and class consciousness? How can it be documented or studied? What does a geographic method look like for this? I got as far as “here’s where people live who attend churches by denomination.” Good enough.
Around this time I was also constrained to an 8gb RAM, 160gb SSD, 4 year old laptop that my workplace had lying around, so my compute power was extremely limited, and I had to get creative with how I processed the data. I don’t remember the precise processing parameters, but it went something like this: download csv, chunk csv into sqlite db, query query query, pull into Python for analysis. I wasn’t as savvy with DBs then as I am now so I’m sure there are more elegant solutions.
To get these local population estimates for denomination concentrations, I wrote two algorithms. The first found all the churches from the company’s database of places. The next algorithm, which linked churches to a denomination, was imprecise, iterative, and introduced way more error. Fortunately many church names follow patterns specific to a denomination, and some even say the denomination in the name. AME churches are typically Something Something AME Church, St. Whatever Episcopal Church, and so on. I did reach out to a separate company that seems to maintain a database of churches by denominations, hoping to use that to cross reference. They were, perhaps unsurprisingly, not as enthused about my ‘.edu’ scheme and wouldn’t give me the data for free. And they call us stingy! Regardless, the algorithm I came up with was right about 70% of the time on the small sample I reviewed, so I said “I’m not getting paid for this, good enough.” I also ignored non-Christian religions, on an emotional level because I just don’t have personal history with proselytizers from non-Christian religions, on a rational level because they don’t seem to have formalized their congregations into political power bases in North Texas.
The map itself is not amazing quality, and was constrained by time and technology. I also never got to develop the methodology for actually exploring the denominations links to power. My theory is that Baptist and Catholic denominational affiliation informs local family structures, home prices, and other neighborhood conditions. I’ve never seen a map of religious affiliation by denomination at a small scale, much less at the scale of Census blocks, so it felt worthwhile on its own.
Camille Kaplan 