Where the Problem with Historical Data about U.S. Religion Really Lies

One of my side projects (eventually to turn into a main project) is figuring out what can be done with historical data about religious groups in the United States. This ground is in some ways well trodden. The field has a very fine atlas in the form of Gaustad, Barlow, and Dishno’s New Historical Atlas of Religion in America, as well as an experimental Digital Atlas of American Religion for the twentieth century. Then too, the field has more or less decided that this ground is not worth treading anyway. There are a number of sophisticated critiques of the whole enterprise of dealing with religious statistics and mapping. If I can sum these up in a broad statement, the point is that numbers don’t tell us anything that the field actually wants to know. As Laurie Maffly-Kipp puts it in a well-argued review essay, “our dazzling new technologies and spatial theories” might only have “brought us back to much more circumscribed definitions of religious experience.”1 I recognize the weight of these arguments, and a full justification for dealing with religious statistics will eventually have to take them into account.

But not yet. I want to argue that historians of American religion have barely begun to take advantage of the quantitative data available to them. While we have to keep the theoretical arguments I alluded to in mind at all times, the pressing issue at the moment is one of basic research. Until we make a fuller attempt at using these quantitative records, we can’t really know whether we will find anything useful from them.

Here is the argument. Mapping and quantitative analysis of historical statistics about U.S. religion have been sorely limited by the kinds of data that have typically been used, namely county-level aggregates of Federal census data, and by the way that mapping has focused on general comparisons rather than the specifics of the data.

First, the kinds of data. Leafing through the New Historical Atlas it is apparent that, with the exception of a few colonial maps, most of the maps in that work follow the pattern of mapping the number of churches per county for three years (1850, 1890, 1950). If you look the data used in the encyclopedia, you’ll find that it all comes from federal censuses, which began asking a few questions about religion in 1850, and which took more detailed censuses of religious bodies in 1906, 1916, 1926, and 1936. The exact same data is used in the Digital Atlas for all years before WWII.2 Furthermore, both the New Historical Atlas and the Digital Atlas use county-level aggregates of the data. (By the way, the same county-level aggregates are available at the NHGIS. Once one has mastered the basics of making maps—it’s not hard—one could reproduce most of the maps in the Gaustad atlas.)

This data is extremely limited. The federal censuses that asked about religion occurred at most once a decade and didn’t start six or seven decades into U.S. history. Because of concerns about church and state, the questions about religion asked by the census were limited questions such as the number of churches, the value of church property and amount of seating, and in only a few years the number of adherents. Because the data is aggregated by county or (worse yet) the state, we are seeing data which has a very low geographic resolution. Furthermore, no real attempt has been made to get beyond the surface of the data to do diachronic calculations. For instance, neither atlas tries to connect individual counties to see how they changed over time. Nor has this data been connected to other kinds of data that are available. Knowing the number of churches per county just isn’t that interesting a fact, so it is unsurprising that the field remains unconvinced that this kind of work is useful.

While the federal censuses with county-level data are the easiest to access, historians aren’t obliged to limit themselves to just those sources. In fact, we are in the midst of a renewal of mapping religion in cities in several different projects which have gone far beyond the census data. I’ll offer a fuller account of what else is available eventually, but for now let’s look at just two tables from two different years’ worth of Methodist annual conference minutes. The first table is from the Minutes of the Annual Conferences of the Methodist Episcopal Church for 1787.

Figure 1: Table from Methodist minutes for 1787 [PNG]

The second table is from the same minutes for 1858.

Figure 2: Table from Methodist minutes for 1787 [PNG]

Looking at these tables we see how these denominational records escape some of the constraints of county-level census data. Let’s make a partial list of what is useful here. First the data in these tables is far more geographically detailed, focusing on circuits and stations. Not only is it more detailed, but the geographic divisions are linked to the institutional structures of the denomination rather than to arbitrary administrative boundaries. Second, with some effort the names of places can be linked from year to year, so that we can get an estimate not just of how the denomination changed as a whole but what was happening over time in particular places. Third, the data is annual. This is hugely important, since not only does it give us ten times more decade per data, but it also lets us evaluate the quality of the data by looking for anomalies. Fourth, the data starts much earlier than the federal data and goes on for longer. Fifth, we have far more interesting variables available to us. The number of adherents is a more useful thing to know than the number of churches. By the 1850s we can start to see things like the number of adherents per minister or per church, the ratio between infant and adult baptism, and so on. And most important even in the earliest Methodist data we have separate figures for white and black Methodists.

Not all of the data is interesting. No one is waiting for a field-changing article on the number of volumes in Sunday school libraries, to be sure. And there is much that we might want to know about, such as gender, which these tables don’t tell us about. But what is readily available is far more interesting than what is in the census. As anyone who has worked with data knows, when you have only one variable in a dataset, you can’t do much more than the obvious with it. But when you have multiple variables you can combine them in many interesting ways.

That leads to my second point about why the current mapping and quantitative analysis. They tend to suffer from the “one map to rule them all” disease. In the Digital Atlas there is one interface to all the data, and in the New Historical Atlas almost all the maps follow exactly the same template of churches per county. There’s a lot that could be said about why this is a poor design for the interface, not least because it puts the burden on the user to do the discovery of what is interesting. My main point is that putting everything in a single map forces you to use a lowest-common-denominator approach to the data. The reason both atlases have essentially one map is that they want to make comparisons across time, space, and denominations. The strength of the Gaustad atlas is that you can look at any of the maps and compare the group being depicted across time, or to other religious groups.3 But this also means that only data which is common to all denominations, times, and places ends up being used. To try to make this concrete, that approach might mean that one would make a map of churches and forgo making a map of the ratio between probationers and members because Methodists kept track of that data while other denominations did not. Instead of making all the data fit the template, it is better to make maps that fit the peculiarity of the data. In other words, as humanities scholars we delight in the telling detail, the peculiarity of a particular narrative or text. But when it comes to mapping religion, we’ve only done the most general work and forsaken the particulars.

To sum up: before we can resolve questions about whether mapping and quantitative analysis is useful, we can do a lot better with the sources and methods that are available to us. There are many challenges to doing that work, not least in finding data is available, digitizing it, figuring out whether the data is reliable, and discovering whether anything actually interesting be said from it. What will we find? Well, all that will have to be the subject of the future posts, except to observe that the scholars who are working on mapping religion in cities are doing extraordinary work asking interesting historical questions well-informed by theoretical concerns. My point for now is to try to turn high-level questions about whether mapping fits the aims of the field into low-level research questions about what sources are out there and what can be done with them.

  1. Laurie F. Maffly-Kipp, “If It’s South Dakota You Must Be Episcopalian: Lies, Truth-Telling, and the Mapping of U.S. Religion,” Church History 71, no. 1 (2002): 132–42. Cf. Laurie F. Maffly-Kipp, “Putting Religion on the Map,” Journal of American History 94, no. 2 (2007): 522–29, doi:10.2307/25094964.

  2. Post-WWII is really a separate question when it comes to gathering data: I’m content to leave that period to the sociologists and late-twentieth century historians.

  3. Well to a certain extent. The Gaustad atlas occasionally makes some truly bizarre changes to the scales of the maps so it is difficult to actually make comparisons.