Churches as organizational resources

Advertised Political Church Activities in Detroit, 1950-1951

The Detroit Free Press reserved a page or two of every Saturday edition to church news. These pages contained church and religion-related articles, church advertisements of their services and announcements of their special speakers and events. A systematic review of these pages from 1950 reveals some of the political activities in which Detroit denominations were engaged. The most prominent and overtly political content to these announcements was anti-communist. Five different churches announced special speeches or series of speeches that stressed the dangers of communism. Four of the five were held at mainline Protestant churches, and one at a Catholic church. The October 7, 1950 church page announced that “church goers will be given the opportunity to sign the Crusade for Freedom Declaration this weekend.” According to the article, all churches and synagogues had been provided with the declaration, which enlisted churchgoers’ support in the fight against communism. And church members were urged to sign before leaving their places of worship. Represented on the Michigan state committee of this crusade were a Rabbi, a representative from the Catholic archdiocese a representative from the Lutheran church, and Methodist and Episcopal Bishops.

Other politically conservative activities include a speech by former Michigan Governor Wilber Brucker (Republican) at a mainline Protestant church, and an annual “patriotic dinner” at another mainline church. Several mainline Protestant churches announced events planned by their businessmen’s clubs and the Episcopal Diocese (mainline Protestant) announced a $1,000,000 gift from Mrs. Ford.¹³

There were fewer announcements of overtly progressive political events. Of those that appeared in the 1950 church pages, include a “civil rights send off” with participation by a black Protestant church, a Rabbi from a reform synagogue, and a speaker from the Catholic Archdiocese. Temple Israel (reform) presented a book review on John Gunther’s Roosevelt in Retrospect, and Central Methodist Church announced a talk by an annual speaker, Norman Thomas (who had run for president on the Socialist Party ticket) (XXXX 2002). In general, this evidence reveals church political activities that are consistent with our argument about the political relevance of churches in neighborhoods.

Hypotheses Concerning the Effects of the Five Socio-Religious Groups on Voting

We have argued that the black community in Detroit was almost exclusively working class, that the black Protestant Churches constituted the largest organizational presence in the Black community, that Black congregants were very open to political leadership by their pastors, who by 1950 supported progressive causes, and that black Protestants had strong communal ties. Additionally, due to segregation, black churches tended to be located in congregants’ neighborhoods. Pastors controlled a significant resource in the black community (meeting space and access to communicants), and had the ability to make it available to the community for political purposes. Together, this provided a strong case for considerable influence for both high concentrations of black Protestants as well as the physical presence of black churches in neighborhoods on left-wing voting there. The combination of high concentrations and the presence of a church in the neighborhood should yield the highest effects. We are less confident that the black church would support the rights of political minorities (as manifested in opposition to Proposal 3), because this was posed as mainly an immigrant issue.

The neighborhood integration of Catholics, their medium level of communal ties, and their working-class economic position in Detroit’s economy should have served to enhance Democratic Party voting. Catholics high levels of church attendance detracted from Democratic voting. The inability of individual priests to take stands on particular candidates likely moderated opposition to the Democratic Party.¹⁴ The presence of churches in the neighborhood gave Catholics a place to meet and enhance their political community, even if it conflicted with the more conservative politics of the clergy. So overall, we expect both the concentration of Catholics and the presence of Catholic churches in neighborhoods to have enhanced Democratic Party voting.

We expect Catholic concentrations, due to their strong support of their ethnic communities, to have been associated with an increase in support for the Progressive Party and against Proposal 3. Even though the Catholic Church had a strong anti-communist stance, we have the same expectations for the presence of Catholic churches, since we argue that the Catholics used their churches as resources.

Mainline Protestants were strongly Republican, in part, due to their class standing. Although Protestant theology tended to be more liberal, most Protestants disapproved of their pastors’ taking a stance on political candidates. Many did not live near their churches; they had medium levels of attendance, and moderate levels of communal ties. Although it is unlikely that the church served as a unifying force in the neighborhood around its theology, it assembled like-minded voters and provided them with potential meeting spaces. Therefore, we expect heavy concentrations of mainline Protestants and the presence of Protestant churches to detract from both Democratic Party and Progressive Party voting, and to have enhanced support for Proposal 3 (a position against immigrant rights).

Evangelical Protestants had lower class standing than mainline Protestants, and there seems to be no coherent sense of how evangelical Protestant theology translated into political positions in Detroit during the 1950s. Additionally, the numbers of evangelical Protestants and churches were relatively small. Hence, we expect concentrations of evangelical Protestants to be associated with more Democratic voting than concentrations of mainline Protestants, but less Democratic than the other socio-religious groups. Since there is no strong political tendency among the communicants, we expect the presence of the churches to have a weaker negative (compared to mainline Protestant) or no impact on neighborhood left-wing voting.

Jewish politics were strongly liberal, with significant radical strains. Their communal ties were strong and they tended to live in areas surrounding the synagogues. Still, Jews in Detroit during the 1950s did not attend synagogue often, and the majority of congregations were orthodox. Overall, we expect concentrations of Jews and the placement of synagogues in neighborhoods to have enhanced Democratic Party and Progressive Party voting and to have repressed the vote for Proposal 3.

A critical reader might ask how we disentangle the overlap between race (especially in the case of blacks and black Protestants), ethnicity, immigration status, language and the occupational-class factors we use as controls. This we are able to do only partially, as data on denomination and church presence do not allow us to determine if a congregant of a specific religion went to the church of the same faith in his or her neighborhood. Nor is it reasonable to conclude that support of left-wing causes from one of the few Jewish neighborhoods is strictly a result of the population of Jews independently of the population of new immigrants. In many cases, these were the same people. These overlapping identities, however, should not eliminate the influence of the most important neighborhood institutions, controlling for the presence of religious congregations.

We expect that the influence of the various religious groups will be commensurate with concentrations of its members and placement of churches. In addition, we expect that when members of a given religious group “dominate” a neighborhood where a church is located, the effect will be exaggerated. Below we first test to see if religiously dominant neighborhoods disproportionately voted in the predicted directions.

Data and Methods

Using geographic information systems, we link together three types of mapped data on Detroit: individual-level indicators on religious denominations and voting, church addresses by denomination, and aggregate information on neighborhoods. Conceptually, these data form three levels – people, nested within churches, nested within neighborhoods.¹⁵ Some respondents are not within churches (“no denomination”) but are still within neighborhoods and some neighborhoods have, and others don’t have a church or synagogue. At the individual level, we have the respondent’s denomination, but not his or her membership in specific congregations, so we collapse the “church” level into the neighborhood. Hierarchical linear models are used to specify the relationship between the probability of an individual’s vote for the Democratic Party and the neighborhood context. Robust regression¹⁶ models of aggregate data are used to examine the relationship between, on the one hand, denominational populations and religious institutions, and on the other, actual election results for the presidential campaign and a politically charged ballot measure.

Data on the dependent variables for the aggregate analysis come from the City of Detroit Election Commission (1952) and Michigan Department of State (1950). We matched voting precincts from the 1952 elections and church addresses to census tracts using maps from the Detroit Public Library Map Room (City Election Commission 1952).¹⁷

The Directory of Churches (Detroit Council of Churches 1951) lists churches by specific denomination, and their addresses. It indicates either “Negro” or the name of the dominant ethnic group. Most are without these ethnic/racial indicators, and we interpret these as majority group, white churches.¹⁸

The Detroit Area Study (DAS) provides self-report data on individual voting in the 1952 election. All DAS from the 1950s contain information about respondents’ religious affiliation and basic demographics and household characteristics. We combine DAS from 1953 to 1958 to estimate the religious makeup of census tracts, something not available from other sources.¹⁹ DAS, however, were not intended to be spatially representative. The first three years under consideration have fewer primary sampling units (tracts), an average of 81 or 21.9% of the 369 total for 1950. The latter three years used a different sampling scheme and cover an average of 243.33 tracts or 65.9% of the total. Because we are interested in the intersection between spatial location and political behavior, respondents who reported having moved into a neighborhood after 1952 were deleted from the sample. Overall, once we delete tracts outside the city limits and merge two pairs of tracts due to missing data, DAS provides data on 299 out of 367 geographic units in Detroit.²⁰

That said, the estimates of the proportion of each denomination in a census tract are based on small samples. The pooled DAS provide an average of 8.75 cases per tract.²¹

Where data on denominations are missing, the median value is substituted. Spatial averages (spatial lags) are used to control for the clustering of denominational populations and improve estimates in areas with missing data. The spatial average is computed as the average value of all contiguous tracts on a given variable. For example, if a tract shared a boarder with three other tracts, containing 1, 0 and 2 Catholic churches respectively, then the average adjacent Catholic churches would be 1. This permits the effect of churches to be larger than a census tract, an arbitrary unit with respect to social segregation. This is a reason to give more weight to the spatial averages, as they represent conglomerations of tracts with more cases and, hence, more stable estimates.²²

Our estimates of tract-level religious populations are thus derived differently from those of the U.S. Census Bureau data on religious affiliation (Zelinsky 1960), enumerations of congregants carried out by the National Council of Churches of Christ (National Council of Churches of Christ 1956), and more recent studies by the Glenmary Research Center (Bradley et al. 1992; Newman and Halvorson 2000). Those county-level counts come from the churches’ own enumerations, numbers that differ depending on how the religion counts membership. The summaries from six years of Detroit Area Study surveys are means of self-reported affiliation from a random sample and, as a result, describe smaller geographic units. No other study to our knowledge has used a random sample to assess the spatial distribution of denominations. Data on household income, occupational populations (craftsmen, operatives and laborers) and the population (over 21 and in the civilian labor force) come from the 1950 Census (United States Department of Commerce 1952).

There are several prominent schemas for categorizing denominations into larger groups within the social science study of religion (Smith 1990; Steensland et al. 2000; Manza and Brooks 1997; Lenski 1963; Greeley 1972). While use the scheme of Steensland et al., we do not assign denomination partly on the basis church attendance.²³ We consider the following categories of religious denominations: Catholics are Roman Catholics; Jews are Orthodox;²⁴ Protestants are divided into black Protestants, and white Protestants. The latter are divided into mainline and evangelical here as per Steensland et al. (2000:314-16);²⁵ the small numbers of Black Catholics are grouped with all other Catholics.

Spatial Distribution of Religious Groups in Detroit

Figures 1-5 show the religious populations and their churches throughout the city of Detroit. Darker shading shows a higher quintile of the distribution; a plus sign indicates the presence of one or more church. Members of the different religious denominations are highly segregated, but their religious institutions are much less so. Note, for example, white mainline Protestants (Figure 3): members are concentrated in the north west and north east sections of the city; the mainline churches are more spread out throughout the city. Due to their smaller numbers, evangelical Protestants (Figure 4) have smaller pockets of concentration that appear to be somewhat randomly dispersed throughout the city, and their churches are also spread out. Black Protestants (Figure 5) are highly concentrated, as are Blacks in general, and their churches tend to be located in their neighborhoods. Catholics (Figure 1) have high concentrations in the north central and south west regions of the city. Their churches tend to be located near the city center, which reflects their long history of presence in Detroit and their tendency to establish churches that represent territories, rather than members. Jews (Figure 2) are segregated in the west-center north-center regions, consisting of a small number of tracts. Their synagogues are located in their areas of concentration.

Figure 6 identifies neighborhoods with a dominant denominational group (where the population of congregants is in the top quintile of that denomination and the denomination had at least one church in the tract). The capital letter indicates which group is dominant.²⁶ Black Protestants dominate in the central city; mainline Protestants on the east and west sides of Detroit. Because evangelical Protestants dominate in only 5 tracts scattered throughout the city they are not depicted on the map. Catholics dominate the areas to the east of Dearborn and in the city’s northeastern corner; Jews dominate in the areas to the north and west of Highland Park (a separate incorporated area that is part of the blank polygon in the center of the map).

Figure 7 displays the distribution of Democratic Party voting, Progressive Party voting, and voting on Michigan Proposal 3. Comparing this with Figure 6 gives a visual image of the spatial correlation between the religious and political characters of the neighborhoods.

These comparisons suggest a strong relationship between denominationally dominant neighborhoods and neighborhood voting: black Protestant dominant neighborhoods constitute the vast majority of the strongly Democratic neighborhoods: black Protestant dominant tracts make up over 70% of all strongly Democratic neighborhoods. Jewish dominant neighborhoods give (comparatively) strong support to the Progressive Party (nine of ten predominantly Jewish neighborhoods are in the top 10 percentile for the Progressive Party vote and for opposition to Proposal 3). Mainline neighborhoods are located in areas with low Democratic and Progressive Party voting. The task of the next section is to determine if these visual impressions of a connection between religious populations and political outcomes hold up at both the individual and aggregate levels and net of control variables.

Religious and Political Affiliation in Detroit, 1950s

Next we turn to hierarchical linear models of individuals nested within census

Table 1 about here

tracts. Table 1 reports the means and standard deviations. A comparison of self-reported voting behaviors by religious group shows that religious denomination is associated with voting behavior in the 1952 presidential election. As expected, black Protestants (without any control variables) have the highest proportion (87%) that voted for the Democrat Party candidate in 1952. Also, a majority of Catholics (64%) voted for the Democratic Party. Approximately 58% of white evangelical Protestants, 54% of Jews, and 40% of the white mainline Protestants reported voting for the Democratic Party in this election.

Table 2 presents hierarchical logistic regression models of Democratic Party voting in 1952 using the combined DAS data²⁷ from the 1950s, and controlling for family income and class. In these logistic regression models, the dependent variable is coded 1 if an individual reports having voted for Stevenson, the Democratic Party candidate. These models are represented by the following set of multilevel equations:

Equation 1 is the level 1 or individual equation. The term P(Y_ij=1) is the probability that the ith person within the jth census tract voted for the Democratic Party candidate. Separate models are estimated within each tract where the DAS has data for 1953. Q is the number of predictors. The X_qij term represents the individual level variables for denomination, working-class occupation and family income. Individual level errors are represented by r_ij. In equation 2, the intercept, _0j, is modeled as the level 2 equation,

w
here ₀₀ represents the grand mean of the dependent variable across all tracts when the value of each of the level 1 and level 2 variables is zero. W_sj is the vector of tract-level variables, and u_0j is the level 2 error term. The u_0j term represents the residual effect of denomination after controlling for the presence of a church in the tract and the spatial mean of churches in the surrounding tracts. The combined model is shown in equation 3.

Model 1 shows that mainline Protestants were less likely to vote for the Democratic Party candidate and black Protestants were more likely to vote Democratic. Model 2 reveals that these two relationships hold after the introduction of controls for class and income.²⁸

Table 2 about here

Next, we turn to the question of how the neighborhood religious environment impacted individual-level voting behavior within neighborhoods. ²⁹ We again use the combined DAS data for the 1950s and church location.³⁰ Model 3 shows that, controlling for religious preference, individuals who lived in neighborhoods with a Catholic or a black Protestant church were more likely to vote Democratic. Those who lived in a neighborhood with one or more mainline Protestant church were less likely to vote Democratic.

The adjacency model (Model 4) finds a significant positive effect for black Protestant churches in the surrounding area on Democratic voting; this means that individuals in tracts with a higher average number of black Protestant churches in the surrounding area were more likely to vote Democratic. Model 5 shows that the spatial effect is crucial to the impact of the black church. Having one or more such churches in a tract becomes insignificant, but being in a tract with a high spatial average in the surrounding community makes resident voters more likely to vote for the Democratic Party. This corresponds to the spatial segregation of blacks in general. Controlling for all of the other significant variables, being black Protestant, having a working class occupation and a lower income, living in a neighborhood where there is a Catholic church, and living in a neighborhood surrounded by black Protestant churches enhanced voters’ propensity to vote for the Democratic Party in the 1952 presidential election. Being mainline Protestant and living in an area where there was a mainline church both decreased the likelihood of Democratic Party voting in this election. In general, regardless of respondents’ religious convictions, the presence of Catholic, mainline Protestant, and black Protestant churches mattered for their political behavior.

Spatial Models of Neighborhood Voting

We now turn to linear models of aggregate vote totals. Instead of looking at individual-level self-report of political preference, we consider the election returns to precincts throughout the city, grouped to census tracts. While it would be preferable to consider multi-level models for all political outcomes, we do not have individual-level survey data on voting for the Progressive Party or Proposition 3. These models do have the advantage of considering actual election returns rather than aggregates of self-reports. We use the pooled DAS data to estimate the proportion of each denomination. Table 3 displays the means and standard deviations for these variables.

Table 3 about here

Table 4 presents robust least-squares regression models of the tract-level vote totals for the Democratic Party controlling for the population of eligible voters. Model 6 (Table 4) finds that net of income and occupation-class controls, the more Jews and black Protestants in the neighborhood, the higher the Democratic vote totals.³¹ The number of mainline Protestants is negatively related to Democratic Party votes. Model 7 shows that the presence of churches of each denomination impacts the neighborhood Democratic vote totals in the expected direction. Catholic, black Protestant churches and Jewish religious institutions increase Democratic Party vote totals; mainline and evangelical Protestant churches decrease it. Model 8 again considers the spatial effect of religion: the more mainline Protestant churches in the surrounding area, the less the Democratic vote total and the more the black Protestant churches in the surrounding area, the greater the Democratic vote total. These findings imply an important role for religious institutions within neighborhoods.

Table 4 about here

The full, aggregate-level model of Democratic Party vote totals (Model 9, Table 5) shows the impact of the black Protestant religion. Concentrations of black Protestants, the presence of a black church in the neighborhood, and the number of black churches in the general area, all independently enhanced Democratic voting in Detroit neighborhoods.³² Model 9 predicts 91% of the variance in 1952 presidential votes for the Democratic candidate (Stevenson) in Detroit neighborhoods. Concentrations of Jews, and black Protestants significantly enhanced Democratic Party voting; concentrations of mainline Protestants in neighborhoods decreased it.³³ The presence of a black Protestant church in the tract and more in the surrounding area enhanced the Democratic vote and the presence of mainline and evangelical Protestant detracted from it.

Table 5 about here

To broaden our analysis, the next two columns in Table 5 consider left voting for the Progressive Party (Model 10) and against Michigan Proposal 3 (Model 11). Model 10 demonstrates that the effect of concentrations of different denominations in neighborhoods is also important for more radical (social movement-like) voting. We use poisson regression for the model on Progressive Party voting because our dependent variable is a count of a small number of values with a strong left skew. The majority of cases are zero. A vote for the Progressive Party in 1952 was a courageous act in Detroit. The House Un-American Activities Committee had been paying visits to the large cities, and had just made a stop in Detroit. The hearings made front-page headlines in Detroit’s newspapers and the committee alleged that communists dominated the Progressive Party.³⁴ This made voting for the Progressive Party a relatively extreme act. While the correlation between concentrations of Jews and Progressive Party votes is strong and positive, when entered into the poisson regression, it is not significant net of the controls. Yet the presence of orthodox synagogues and the spatial average of synagogues both positively contributed to the Progressive Party vote. Spatial averages of mainline Protestant churches have a negative impact on Progressive Party vote totals. Numbers of craftsmen detracted from the Progressive Party vote while numbers of operatives contributed to it.

Model 11 considers votes against Proposition 3, i.e. opposition to a conservative and xenophobic proposal feared by many foreigners and opposed by leftists. We find that Jewish concentrations and the presence of synagogues in neighborhoods both independently increased the numbers who opposed further restrictions on political radicals. Likewise, the presence of a Catholic church increased the level of opposition to Proposition 3. Positive coefficients here show support for the political left. The presence of black churches and mainline Protestant churches in surrounding census tracts increased the number voting for further restrictions (the political right). Voters from the least skilled portion of the working class also tended to favor more restrictions.