Table 8.1 Multivariate regression results (sub-district data)
A provincial meso perspective from Maluku
To triangulate the findings from the qualitative interviews and counterfactual comparisons, this section complements the local data collected in Cotabato and Ambon with a provincial meso level analysis. It takes a closer look at a unique provincial dataset to assess the association between church-based activities and reconciliation indicators. The discussion that follows will be organized in two parts. The first part describes the compilation and assembly of our subnational data. The second part presents and interprets the results of the multivariate regression estimates.
As outlined in the research design, the phenomenon to be explained, the dependent variable, is the degree of non-violence. For assessing the absence of violence, we drew on Indonesian1 household survey data (PODES), which include village-level perceptions of various aspects of communal life and development. The PODES data were carefully compiled for all subdistricts in the province of Maluku.2 Indicators on local conflict and violence, which are highly relevant for our study, feature prominently in two particular survey years, 2004 and 2013. We use these two points in time and create a “declining death toll” indicator by subtracting reported conflict-related homicides in 2013 from those in 2004. The resulting values (which turn positive once homicide numbers decline over time) constitute a proxy for the overarching objective of peaceful cohabitation in local communities.
The unit of analysis for all variables in the regression estimates is the sub-district level (kecamatan). The province of Maluku consists of 112 sub-districts with an average population of approximately 15,000 people. The “district level” was not chosen because it would have resulted in a very limited number of observation points (11 districts); and a focus on the “village level” was unfeasible because identifying church-based project efforts in over one thousand villages would have been virtually impossible, not least because local churches in Maluku have insufficient data on the spatial boundaries and time frames of their project activities, but also because project documentation has been partially destroyed in conflict-induced fires.
The independent variable in this analysis is given by the presence of church-based activities (CBAs) that foster peaceful cohabitation. There are no standardized or easily accessible sources that provide a good approximation of these activities across subnational units. In the absence of such, we constructed perception-based measures by drawing on in-depth interviews and expert cross-checks. We consulted four of Ambon’s most knowledgeable church actors (Father Costan, Father Böhm, Sister Brigitta Renyaan, and Rev Jacky Manuputty), who were chiefly involved in Catholic and Protestant reconciliation projects and could provide valuable information on the “where” and “what” of church-based activities. Accordingly, the indicator of church-based activities is constructed by combining these two aspects. The first aspect captures which “church community” is active in which subdistricts. Based on the distinction of (a) Catholic, (b) Protestant, and (c) joint Catholic/Protestant efforts, we generated three dummy variables for each of these categories. The second aspect pertains to the general scope of church-based efforts. Here, we assigned scale values in ascending order reflecting the comprehensiveness of observable church activities, ranging from “no activity” (value 0), through “activities fostering the physical reconstruction of housing and infrastructure” (value 1), to “activities promoting interfaith dialogue, trauma healing, or reconciliation” (value 2). By multiplying the outlined scope variables with the respective church community variables, we generated three indicators of church-based activities for the quantitative analyses, namely: (1) “Catholic CBAs,” (2) “Protestant CBAs,” and (3) “Joint CBAs.” The integral values of these independent variables range between 0 and 2, respectively.
In addition, we added three control variables. These include the potential oversight of local security forces, approximated by the distance to the next district police station (in kilometers). We also controlled for relative economic status: by calculating the share of families in each sub-district that are considered “prosperous” or “pre-prosperous” it was possible to construct a proxy for household prosperity. Moreover, we controlled for the size of the respective sub-district communities by adding a population variable to the equation.
The quantitative analyses shed further light on the research proposition regarding the association between church-based activities and the attainment of non-violence. The results of the ordinary least square (OLS) estimates are summarized in Table 8.1 below.
Two results stand out in particular. First, OLS estimates indicate that coefficients of the independent variable “Catholic CBAs” are positive and significant at the 0.01 level (without controls) or 0.05 level (with controls). Second, regression results also show that “Joint CBAs” (i.e., instances of combined Catholic and Protestant project efforts), are positively associated with reduced degrees of local violence (coefficients are also significant at the 0.01 or 0.05 levels). The estimates for “Protestant CBA” efforts, in contrast, do not indicate a clear relationship. OLS coefficients are not significant: t statistics amount to 1.11 (without controls) and 0.25 (with controls), respectively, and hence the data does not allow us to specify an association with measures of reduced violence.
Table 8.1 Multivariate regression results (sub-district data)
Overall, these results provide further support to the hypothesized link between church-based activities and non-violent cohabitation in the case of Maluku. Due to endogeneity restrictions and the limited scope of our data, we do not claim that this OLS analysis, of and by itself, establishes a causal relationship. It does, however, highlight a significant positive association between Catholic and joint Catholic/Protestant project efforts, and a reduced incidence of conflict-related violence. This finding, combined with the findings from the counterfactual case comparisons (in chapter 7), lend further credibility to the claim that church-based activities have made a notable difference to local community life in Ambon and its surrounding sub-districts.
1 In the Philippines, the limited availability of subnational data did not allow for regression analyses to be conducted; unfortunately, the Philippine National Statistics Bureau does not collect/provide key indicators below the provincial level (which made it impossible to assemble a local conflict dataset).
2 The Indonesian Central Bureau of Statistics (BPS) collects village-level survey data (PODES) on an annual basis.