Social networks and technology adoption: Evidence from church mergers in the U.S. Midwest

2.1 Church and congregational mergers

In the 1950s and 1960s, national Lutheran church bodies underwent significant institutional consolidation. At an April 1960 meeting in Minneapolis, Minnesota, three of the largest national Lutheran church bodies—the American Lutheran Church (ALC), the United Evangelical Lutheran Church (UELC), and the Evangelical Lutheran Church (ELC)—voted to merge and form The American Lutheran Church (TALC). This merger officially took effect on January 1, 1961.

A similar merger between the United Lutheran Church in America, the Finnish Evangelical Lutheran Church of America, the American Evangelical Lutheran Church, and the Augustana Evangelical Lutheran Church created the Lutheran Church in America (LCA) in 1962. In 1963, the Lutheran Free Church (LFC), composed largely of congregations that originally opted out of the 1960 TALC merger on theological grounds, decided to join TALC as well, extending the scope of this major Lutheran branch (Wolf, 1966).12

Figure 3 depicts the major mergers between Lutheran church bodies in the United States since the 1950s. For historical context, we focus primarily on TALC for two reasons. First, congregations of TALC were geographically clustered in the Upper Midwest whereas congregations of the LCA were more dispersed throughout the country. Second, we have access to yearbooks from TALC detailing a handful of congregational-level statistics throughout the 1960s.

National-level mergers, arranged by the constituent churches' theological and institutional leadership, had far-reaching impacts. The TALC merger was reported in local newspapers across the Upper Midwest (Dugan, 1960; Press, 1960), and national mergers forced local congregations to adopt new constitutions, bringing them into alignment with the newly formed national church (Nelson, 1975). Prior to the national mergers, many towns had congregations from multiple church branches. As a result of the national mergers, these congregations suddenly found themselves in the same denomination. This frequently led to mergers between local congregations that were previously impossible (Trinity Lutheran Church, 2012; United Lutheran Church Laurel, 2013). These mergers brought previously socially disparate groups of people into contact with one another.

Each of the merging national-level church bodies (and their associated congregations) were linked to a different ethnic group: The ALC had German roots, the ELC had a Norwegian background, and the UELC was historically Danish (St. John Evangelical Lutheran Congregation, 2014). Especially in the early parts of the 20th century, this often meant that congregations across the street from one another were holding services in different languages. Some congregations were even conducting multiple services, each in a different language (Bethel Lutheran Northfield Church, 2014; Murray County, 2014).13 Cross-branch mergers between local congregations were large shocks to churchgoers' social networks, because the congregants were not likely to have interacted nearly as frequently prior to the merger.

In addition to the local mergers that were precipitated by national church changes, a number of congregational mergers resulted from other plausibly random events. Several congregations initiated mergers after natural disasters destroyed congregation buildings (Bethlehem Lutheran Church, 2014; St. Mark's Lutheran Church, 2014). Other congregations merged due to difficulties hiring full-time clergy. Pastors, trained in centralized seminary programs, were a scarce and expensive resource, occasionally serving multiple congregations at once. Pastors in these roles frequently pressured their congregations to consolidate resources and merge into a single entity (Grace Lutheran Church, 2014; J. Thoreson, personal communication, November 25, 2013).14

In our data, we do not have definitive information about the proximate cause of each congregational merger. Therefore, we are unable to distinguish mergers resulting from national-level branch mergers from mergers following a church fire, for example. However, as we have argued above, the historical record suggests all these congregational mergers occurred for reasons that were orthogonal to agricultural fertilizer use. In Table 1 and Figure 5, described in further detail below, we present evidence that counties with merging congregations are statistically similar to counties without merging congregations along a variety of dimensions both in the pre-merger cross-section and in premerger trends.

Even conditional on mergers being plausibly exogenous, whether these mergers led to increased information diffusion and technology adoption is an empirical question. If the new congregations encouraged cross-group discussion, social integration, and trust, this likely led to technology transfer. However, it is also possible that mergers did not facilitate social integration—for example if members who previously belonged to separate congregations attended separate services postmerger or if this increased antagonism between groups. As a result, we use data to empirically test the extent to which mergers drove technology adoption.

3.1 Church data

We acquired a dataset on mergers between Lutheran congregations from the archives of the ELCA. Church archivists compiled a database of mergers between Lutheran congregations, the earliest of which occurred in 1810.15 These data are derived from two main sources: annual national church yearbooks, which in turn were compiled from reports congregations made to their governing bodies; and The American Lutheran Church (1960), a record of all the details surrounding the TALC merger. This dataset includes all mergers between congregations of the UELC, ALC, and ELC, which merged to form TALC in 1960, including mergers between congregations that no longer exist. For each congregation involved in a merger, the dataset records the state, county, local post office, location,16 synod,17 congregation name, founding date, merger date, details on which other congregations were involved in the merger, and additional historical notes. Ideally, these data would also include information on congregational demographic composition. Unfortunately, such variables were not recorded, and we are unable to explore how merger effects might differ by congregants' average age or ethnic background.

In our main analysis, we analyze the mergers that took place between 1959 and 1964.18 A total of 56 mergers occurred during this time in Minnesota, North Dakota, South Dakota, and Wisconsin. In our main study period, between 1959 and 1964, 47 counties experienced one merger, three counties experienced two mergers, and one county experienced three mergers. We consider counties that experienced at least one merger between 1959 and 1964 to be treated, and counties that did not experience a merger during this time to be untreated. Figure 4 displays the spatial distribution of treated and untreated counties. The counties in blue experienced at least one merger between 1959 and 1964, the counties in white did not, and the counties in gray included a major urban area and were excluded from the sample.

We restrict our analysis to Minnesota, North Dakota, South Dakota, and Wisconsin for two main reasons. First, as Figure 2 shows, these are the states where Lutheranism is most prevalent. We believe Lutheran congregational mergers are likely to be more economically relevant in areas where Lutheranism is a prevalent or predominant religion. Second, these four states are fairly geographically and agronomically similar, meaning that farmers are likely to have adopted fertilizer for use in similar cropping systems. There is a steep drop-off in the prevalence of Lutheranism between our four chosen states and the next-most-Lutheran states: Nebraska and Iowa. Additionally, agricultural practices and crop choices in the 1960s in Nebraska and Iowa differed in meaningful ways from Minnesota, Wisconsin, and the Dakotas.

3.2 Agriculture data

We combine our data on merging congregations with data from the United States Department of Agriculture (USDA)'s Census of Agriculture. In the 1950s and 1960s, the census was designed to have full coverage of every farm in the U.S. (U.S. Census Bureau, 1967). Censuses were taken every 5 years, and data gathering took place in the fall. Enumerators visited every dwelling and administered the census to any household engaged in agriculture. After collection, the census underwent a multistage quality control process. The final dataset is available at the county level. Wherever possible, we use a digitized version of the dataset maintained by the University of Michigan's Inter-university Consortium for Political and Social Research (Haines et al., 2015). Some variables were unavailable in the digitized data; we hand coded these from PDFs made available from the USDA's own archive.19

We use the data from the Census of Agriculture from 1945, 1950, 1954, 1959, and 1964. Our main analysis uses the 1959 and 1964 censuses due to data limitations with earlier and later data. In particular, the 1950 and 1945 censuses did not include county-level information on fertilizer use. We are unable to use later waves of data, because after the 1964 wave, data were only collected for farms selling over $2500 worth of goods per year, and there is no way to reconcile the two sampling frames. We use the 1954 census to test for differential trends in fertilizer use among counties with and without congregational mergers. Though the 1950 and 1945 censuses did not include information on fertilizer, we can and do use data from these datasets to analyze premerger trends in other variables of interest. Having performed these tests for parallel trends using the earlier data, we perform our main analyzes using the 1959 and 1964 censuses. We combine the Census of Agriculture data from these years with our congregational data to create a balanced panel of 262 counties.

The Census of Agriculture contains data on our main outcomes of interest: the number of farms using fertilizer, acres fertilized, tons of chemical fertilizer used, corn acres fertilized, and tons of dry and liquid fertilizer used on corn. It also contains information on the use of agricultural lime, a complement to chemical fertilizer. In addition, the census includes data on other agricultural practices, such as strip cropping and irrigation; other types of land use, such as orchards; and capital-intensive farm durables, including vehicles.

3.3 Other data

In addition to the data described above, we also use data on historical weather including average annual precipitation, average annual temperature, heating degree days, and cooling degree days, aggregated to the climatic division level, all sourced from the National Oceanic and Atmospheric Administration (2007). We also use county-year-level data on open two-digit highways sourced from Baum-Snow (2007). Finally, we obtain basic county-level demographic data in 1950, 1960, and 1970 from the U.S. Census Bureau.

3.4 Summary statistics

Table 1 presents summary statistics from the 1954 Census of Agriculture and several population censuses for counties that did and did not experience congregational mergers between 1959 and 1964. Although there are some apparent differences between treated and untreated counties (treated counties tend to be more rural and agriculturally focused), few of these differences are statistically significant. The major exceptions are in harvested acreage and miles of highway: treatment counties harvested approximately 57,500 more acres and had 0.02 more miles of open two-digit highway than control counties in 1954. These differences are statistically significant at the 5% level.

Overall, Table 1 reveals that treatment and control counties are reasonably similar to one another prior to the congregational mergers that took place between 1959 and 1964. These statistics support the notion that mergers were not driven by the agricultural sector or other potentially endogenous factors. Additionally, our preferred empirical specification includes county fixed effects that will account for any time-invariant discrepancies between treated and control counties.

In addition to assessing balance in the cross section, we can also examine the extent to which outcomes in treated and untreated counties were trending differentially over time. Figure 5 presents graphical evidence on premerger trends for six variables: the number of farms using fertilizer, the number of farms using lime, irrigated acreage, acres in orchards, harvested acreage, and the number of farms with trucks over the full time series for which we observe each variable.

Across all these variables, we find counties that experienced a congregational merger between 1959 and 1964 were trending similarly prior to 1959. This suggests that we can compare counties with and without mergers before and after these mergers took place to recover causal estimates of the effects of mergers on agricultural outcomes.

As one such comparison using the raw data, Figure 6 reports kernel density plots of the change in farms using fertilizer between 1959 and 1964 for counties that did and did not experience a congregational merger. On average, counties without a merger saw a reduction of 21 farms using fertilizer, and counties with a merger saw an increase of seven farms. Although Figure 6 seems to tell a compelling story that congregational mergers increased the number of farms using fertilizer, it does not control for other factors that may affect fertilizer use. To capture these possible confounding factors, our empirical strategy employs a difference-in-differences approach and controls for county and state-by-year fixed effects.

5.1 Main results

We first estimate the effects of congregational mergers on fertilizer use on the extensive margin, using the number of farms using fertilizer as the dependent variable. We estimate five specifications, each with a different set of controls. Table 3 reports the results. Column (1) is the most parsimonious specification, including only the interaction term of interest (Year = 1964 × merger), a 1964 dummy, and a “merger county” dummy. In Column (2), we replace the “merger county” dummy with county fixed effects. Column (3) adds four weather controls: temperature, precipitation, heating degree days, and cooling degree days.23 In order to control for time-varying unobservables, we include state-by-year fixed effects in place of the weather controls in Column (4). In Column (5), we include both state-by-year fixed effects and weather controls. This is our preferred specification. Finally, in Column (6) we re-estimate the specification in Column (5) without observations from Wisconsin, which is the least Lutheran and most agronomically dissimilar of the four states.

The results in Table 3 are consistent with our main hypothesis: As expected, counties that experienced congregational mergers see higher rates of fertilizer adoption than those that did not. These effects are economically meaningful and relatively stable across specifications. Given our limited sample size, we are somewhat underpowered. Across specifications, we find positive—and economically meaningful—point estimates, ranging from 28.43 to 37.73 additional farms adopting fertilizer as a result of congregational mergers. In Columns (4) and (5), these estimates are statistically significant at the 10% level based on traditional standard errors clustered at the county level. With the inclusion of weather controls in Column (5), our preferred specification, we find that congregational mergers caused 36.09 additional farms per county to begin using fertilizer, an increase of 5% over the mean in the control group. These results are statistically significant at the 10% level, which, given our relatively small sample size, is encouraging. When we drop Wisconsin, the result attenuates slightly, to 33 additional farms (on a smaller mean of 552), and we lose statistical significance, though this is not surprising considering the smaller sample.

Given our relatively small sample size, traditional inference that relies on asymptotic theory may be inefficient relative to nonparametric approaches such as bootstrapping or randomization inference (Horowitz, 2019). Therefore, we supplement traditional inference methods with a randomization inference procedure. We randomly reassign exactly 51 counties to treatment 5000 times. For each iteration, we estimate every specification in Table 3 and store the estimated coefficient $\widehat{\beta }$. We display the results of this procedure in Figure 7. The gray histograms show coefficients from these 5000 random draws and the blue lines denote the treatment effect using the real assignment vector. In each case, the real effect lies in the far-right portion of the distribution—and in our preferred specification, lies above the 97th percentile—which suggests that our results are not an artifact of random chance. This randomization inference exercise also provides exact p-values that are valid for small samples. We report these “randomization inference p-values” in brackets in all our results tables and prefer this approach to inference in our small sample setting. Using this method, the p-value for our preferred specification in Table 3 is 0.029.

Because we observe each merged congregation's size (membership), it is natural to explore whether mergers have heterogeneous effects by congregation size: Do larger mergers have larger effects on technology adoption than smaller mergers? Unfortunately, our sample size is too small to answer this question: Only 51 counties experienced a congregational merger at all between 1959 and 1964. Every specification we estimate that includes a heterogeneous treatment effect fails to yield any statistically significant treatment effects whatsoever. Therefore, we instead focus on mergers' average treatment effects.

5.2 Robustness checks

We conduct a series of robustness checks to complement our primary analysis. These include analyzing alternative dependent variables, conducting placebo tests, and exploring alternative explanations.