<\/span><\/h2>\nI develop an empirical strategy to identify the causal impact of Pentecostal expansion on electoral outcomes in Brazil. I focus on a unique historical driver: the activities of the Summer Institute of Linguistics (SIL), an evangelical missionary organisation founded in the US that specialises in Bible translation. Most members belonged to the conservative wing of US evangelicalism and aimed to promote their values where they worked (Hvalkof and Aaby 1981). While Latin America has been SIL\u2019s main field of operation, it has also worked in Asia and Africa.<\/p>\n
Around 1960, SIL began translating the Bible into indigenous languages across Brazil. The process of Bible translation is highly involved and typically takes about ten years. Although missionaries lived in central towns and were not permitted to establish churches or schools, they maintained continuous contact with indigenous populations. Through these interactions and the availability of the Bible in local languages, SIL missionaries spread their beliefs and conservative views. Even for bilingual indigenous individuals, contact with SIL members meant exposure to proselytising efforts.<\/p>\n
The translation of the Bible into a specific indigenous language is used as a proxy to measure SIL\u2019s activities. As shown in Figure 2, between 1980 and 2010, SIL activities expanded into new linguistic regions through a staggered rollout across Brazil. This timing and geography variation allows for an empirical analysis of SIL\u2019s influence on religious affiliation.<\/p>\n
Figure 2<\/strong> SIL activities rollout<\/p>\n<\/span>Note<\/em>: Each polygon represents the geographic region of a distinct language spoken in Brazil, according to the 14th edition of Ethnologue (2000). Data on the timing of Bible translations were obtained from the Joshua Project, an evangelical organisation based in the US. Red polygons indicate languages with a Bible translation, while blue polygons represent those without one.<\/span><\/h6>\n<\/span>Measuring the impact of SIL activities<\/span><\/h2>\nFocusing on municipalities where indigenous languages are spoken, I examine religious affiliation and voting outcomes before and after the first Bible translation into the local languages. Using a difference-in-differences approach, I compare outcomes across time and municipalities to isolate SIL\u2019s effects and test for pre-existing trends.<\/p>\n
Figure 3 presents event-study estimates for four outcomes: (a) the share of Pentecostal affiliations, (b) the share of other (non-Pentecostal) evangelical affiliations, (c) the vote share of far-right candidates in presidential elections, and (d) the vote share of evangelical candidates in federal elections. Across all panels, there is no evidence of pre-trends, supporting the assumption that, conditional on controls, the timing of SIL activities is effectively random.<\/p>\n
Pentecostal affiliation rises notably following SIL\u2019s first translation (Panel a), while other evangelical affiliation remains flat (Panel b). Political outcomes also increase, but more gradually: both far-right vote share and evangelical candidate vote share rise following SIL engagement (Panels c and d).<\/p>\n
Figure 3 <\/strong>Pre-trend analysis<\/p>\n<\/span>Note<\/em>: These graphs display estimated coefficients from event-study regressions measuring the impact of SIL Bible translations on religious affiliation and voting outcomes. The x-axis shows years relative to the first Bible translation in each municipality.\u00a0 Confidence intervals are based on robust standard errors clustered at the language level.<\/span><\/h6>\nNext, I construct a municipality-level measure of SIL exposure between 1980 and 2010. This measure accounts for multiple indigenous languages within a municipality and 1980 population size, varying over time only with the timing of translations. I use this measure as my main independent variable in models that control for fixed municipal differences, overall time trends, and interactions between time trends and baseline characteristics such as income, urbanisation, school attendance, and ethnic composition.<\/p>\n
Results show that SIL exposure led to higher Pentecostal affiliation. This rise mainly reflected people switching from other religious groups. A breakdown by ethnic group indicates that the effects occurred mainly among indigenous and mixed-race populations.<\/p>\n
Despite finding no evidence that SIL targeted municipalities where Pentecostalism was already growing, some potential threats to the identification strategy remain. For example, SIL might have targeted municipalities based on factors that were predictive of future evangelical growth and visible to missionaries but unobserved by researchers.<\/p>\n
To address this, I construct an expected exposure measure based on an exogenous translation cost shifter. As translating the Bible into a language requires significant investment, new translations were more likely in languages linguistically similar to those that already had a translation. Accordingly, I build this measure by replacing the timing of translations into Brazilian indigenous languages with the timing of translations into similar languages spoken outside Brazil and its neighbours. In this setup, time variation mainly reflects Bible translations in North America, Asia, and Africa. Results using expected exposure closely mirror those with actual SIL exposure, reinforcing the causal interpretation.<\/p>\n
<\/span>From conversion to votes<\/span><\/h2>\nLeveraging the Pentecostal variation driven by SIL exposure, I assess the electoral effects of Pentecostal expansion. I focus on two main outcomes: the vote share of far-right candidates in presidential elections and the vote share of candidates affiliated with evangelical churches in federal elections.<\/p>\n
A 10 percentage point increase in Pentecostal affiliation, about the average change between 1990 and 2010, is associated with a 13 percentage point rise in evangelical candidates\u2019 vote share in federal elections and a 2 percentage point rise for far-right presidential candidates.<\/p>\n
These estimates rely on the assumption that SIL exposure affects political outcomes only through its impact on religious affiliation. Consistent with this, I find no significant effects of SIL exposure on other key municipal characteristics, such as literacy, school attendance, urbanisation, or agricultural employment. Moreover, event-study estimates show a more gradual increase in evangelical and far-right voting compared to the sharper rise in Pentecostal affiliation.<\/p>\n
I also analyse the extent to which SIL-driven Pentecostal growth contributed to Bolsonaro\u2019s 2018 success. While surveys indicate nearly 70% of evangelicals supported him, this correlation could reflect underlying characteristics. I employ a cross-sectional analysis linking SIL-driven Pentecostal growth to Bolsonaro\u2019s vote share in 2018. This analysis suggests that he received about 9.8 percentage points more votes in municipalities that experienced a one standard deviation larger increase in Pentecostal affiliation.<\/p>\n
<\/span>Spillovers beyond indigenous regions<\/span><\/h2>\nAlthough SIL\u2019s work focused on indigenous-language communities, their influence may have spread more widely. Using a market-access approach that accounts for geographic proximity, I find evidence of Pentecostal increases in nearby municipalities without indigenous-language communities. This likely reflects the role of migrants, commuters, and social networks in transmitting religious practices. Furthermore, in municipalities where no indigenous languages are spoken, increases in Pentecostal affiliation also appear linked to greater support for evangelical and far-right candidates, though the magnitudes are smaller than in regions with direct SIL exposure.<\/p>\n
<\/span>Why do Pentecostal communities wield such political influence?<\/span><\/h2>\nOne key mechanism is the organisational capacity of Pentecostal churches. These churches often hold frequent services where pastors discuss politics, endorse candidates, and host blessing ceremonies. The findings indicate that candidates endorsed by major Pentecostal churches are significantly more successful in capturing votes. In contrast, candidates affiliated with smaller Pentecostal churches do not experience the same electoral advantage. These results suggest that identifying with any Pentecostal church does not automatically guarantee electoral support; rather, the organisational strength is a key element. This advantage might be especially relevant since Brazil\u2019s 2015 campaign finance reforms imposed strict spending limits (Avis et al. 2022), making grassroots mobilisation critical for electoral success.<\/p>\n