What explains election-driven family conflicts?

Abstract

As affective polarisation rises worldwide, so do concerns about how political disagreements affect social relationships, including families – a core building block of societies. The impact of political division on families has been a topic of interest for media outlets, but scholarship remains scarce. To contribute to this emerging debate, we ask: What factors increase the risk of election-driven family conflicts? Employing data from a national survey in Brazil, a highly polarised context, we find that the prevalence of election-driven family conflicts is low, affecting roughly only 3.5% of respondents. As our theory-development exercise reveals, even if not widely spread, election-driven family conflicts disproportionately affect one group: Lesbian, gay, and bisexual respondents. Our study adds to a growing scholarship on the detrimental impacts of polarised elections on the LGBTQI+ community – and reinforces the need for further research to investigate the heterogeneous effects of emerging political dynamics on marginalised groups.

Introduction

‘Affective polarisation’ – individuals’ positive attitudes towards their party/candidate and negative attitudes towards the ‘opposite’ side – is on the rise across many democracies (Iyengar et al., 2019). A growing scholarship seeks to understand the causes and consequences of affective polarisation around the globe (Guedes-Neto, 2023; Waldrop, 2021). Emerging research points to detrimental repercussions across several areas: From aggregate-level threats to democracy (Finkel et al., 2020), to elite level processes (Leonard et al., 2021), and individuals’ preferences (Hobolt et al., 2024) and day-to-day lives (Iyengar et al., 2019).

As a phenomenon characterised by its tendency to brew antagonistic out-group attitudes and behaviours (Finkel et al., 2020), the rise of affective polarisation also raises concerns about how political disagreements affect social relationships (Torcal and Thomson, 2023). This includes concerns that affective polarisation may negatively impact a core building block of societies: The family unit.

Indeed, news articles emphasising the detrimental impacts of affective polarisation on family relationships have been abound. With the headline ‘Political Divide Splits Relationships – and Thanksgiving, Too’ (Tavernise and Seelye, 2016), The New York Times captured a concern many other news outlets shared (e.g. Pinsker, 2021). The potential negative consequences of political division on family units also raised concerns across the pond. As one headline from The Guardian read: ‘Family rifts over Brexit: I can barely look at my parents’ (Cosslett, 2016).

Although media speculation has been widespread, scholarly evidence of election-driven family conflicts has remained scarce. While emerging scholarship suggests that opposing electoral preferences may lead to family conflicts (Chen and Rohla, 2018; Davies, 2022), we do not know the conditions under which this may be the case. To contribute to this emerging debate, we ask: What individual-level factors increase the risk of election-driven family conflicts?

To the best of our knowledge, this is the first study that seeks to answer this question. Venturing into largely uncharted territory requires us to engage in a theory-development exercise.

As a large body of work shows, families tend to be politically cohesive (Baker et al., 2020). Because individuals often seek partners who think like themselves, electoral congruence is commonly high among spouses (Coffé and Need, 2010; Osborn and Morehouse Mendez, 2011; Strøm, 2014). Through political socialisation within the household, partisanship (Beck and Jennings, 1991; De Vries et al., 2009; Macfarlane, 2022) and negative partisanship (Boonen, 2019) are also transmitted across generations. Due to self-selection and political socialisation, family units have thus often been characterised as political ‘echo chambers’ (Iyengar et al., 2018). The political cohesiveness of family units should make them the least likely scenario for electoral disagreement brewing conflict. In other words, contrasting with alarming headlines, this scholarship suggests that election-driven family conflicts should be rare.

Even if that is the case, like other political phenomena, election-driven family conflicts are unlikely to affect everyone equally. For example, as the literature on affective polarisation indicates, those with strong negative attitudes towards political opponents are potentially more likely to cut ties with family members over politics. The same is true for those for whom politics is salient and elections are, thus, higher stakes. Finally, it is also possible that social marginalisation heightened by electoral dynamics motivates individuals to stop talking to family members who they view as supporting a side that threatens their beliefs or existence.

To explore the factors underpinning election-driven family conflicts, we employ data from an original national survey from Brazil (N = 2,501) conducted in the aftermath of the 2022 presidential elections. Striking levels of affective polarisation in Brazil (see Appendix A) warrant it a plausible setting for the emergence of election-driven family falldowns.

Our findings are twofold. First, we find that election-driven conflict within family units is less pronounced in Brazil than otherwise anticipated by the media, affecting roughly only 3.5% of respondents. In large part, this is a product of the maintenance of political ‘echo chambers’ (Iyengar et al., 2018) within families: Even amidst high levels of society-wide affective polarisation, divergences in electoral preferences remain low within families.

Second, we reveal that a critical factor explaining the higher risk of family conflicts is voters’ sexual orientation. Specifically, we find that when electoral preferences diverge within family units, lesbian, gay, and bisexual (LGB) voters are more than 1.5 times more likely to report severing ties with at least one relative and represent roughly one-third of the cases of family conflicts reported within our sample. We interpret this result as the product of LGB voters’ higher levels of political engagement and a heightened sense of social marginalisation: During his first term in office, Jair Bolsonaro, who was attempting reelection in 2022, consistently undermined the rights and protection of the LGBTQI+ community (Feitosa, 2021; Martins de Oliveira, 2022). The 2022 electoral dispute was not equally costly to all voters.

This result is aligned with studies that show that sexual minorities are more electorally engaged (Daoust et al., 2024; Çakır et al., 2023), and face increased levels of stress in electoral disputes with viable far-right candidates (Gonzalez et al., 2020; Swank, 2020) – and reveals yet another dimension of the heterogeneous impacts of elections on marginalised groups.

Data

We employ data from a survey conducted online via Qualtrics between 22 November and 5 December 2022, shortly after the second round of Brazil’s presidential elections on 30 October 2022. The sample includes 2,501 respondents and mirrors the Brazilian adult population with respect to gender, age, socioeconomic class, and regional distribution. ¹ Appendix B details the questions employed, and Appendix C (Table 1) provides descriptive statistics for all variables.

An initial survey question asked participants to identify living family members, including spouses, parents, children, and grandparents. ² Respondents without living family members were disqualified from further participation. ³ Appendix C (Table 2) displays the share of respondents who declared having each type of family member.

We aim to investigate which factors increase the likelihood of election-driven family conflict in polarised elections. To capture our dependent variable, we employ answers to the question: ‘Because of the 2022 elections, did you stop talking to any of your family members?’ By asking whether respondents severed ties with any of their family members, this question aimed to reduce variations in respondents’ interpretation of ‘conflict’ and identify lasting tensions rather than temporary animosities. ⁴ The variable family conflicts counts the number of family members with whom respondents severed ties; it ranges from 0 to 4 and has a mean of 0.04, indicating few respondents reported such conflicts.

Scholarship has yet to examine whether affective polarisation has repercussions for family relations and, if so, for whom. We advance and test possible factors that increase the likelihood of election-driven family conflict.

First, an underlying assumption of election-prompted conflicts is that family tensions emerge due to electoral disagreements in contexts where affective polarisation is widespread. We explore whether family tensions arise from electoral disagreements, with the variable electoral divergences, which counts the number of family members supporting the respondent’s opposing candidate. ⁵ This variable ranges from 0 to 7 and has a mean of 0.60, indicative of the low levels of within-family electoral divergences in our sample.

Second, and aligned with the literature on affective polarisation (Druckman et al., 2021; Iyengar et al., 2019; Wagner, 2021), we propose that greater antagonism towards a political side increases the likelihood of severing ties with family members who support that side. In Brazil, this dynamic revolves around the left-wing Workers’ Party (PT) and its leader, Luiz Inácio Lula da Silva (Samuels and Zucco, 2018). The popularity and subsequent 2018 election of Jair Bolsonaro – a personalist candidate with weak partisan ties – contributed to further polarisation and provided a source of political identification for the right (Rennó, 2020). To capture this, we employ the variables antagonism Lula supporters (ALS) and antagonism Bolsonaro supporters (ABS), which are measured with 11-point Likert scales and have a mean of 5.78.

Third, we expect that political conflict is more likely to arise among those for whom politics is important (Guedes-Neto, 2023). We capture this with the variable political salience, which measures how often respondents discuss politics. The variable ranges from 0 (‘never’) to 3 (‘every day’) and has a mean of 1.44.

We do not expect electoral disputes to pose equal stakes to all individuals. Specifically, we anticipate stakes to be higher for groups for which the election heightened social marginalisation. Prior research indicates that polarising political events can reshape social identities (Guedes-Neto, 2023) and lead individuals to distance themselves from perceived threatening opponents (Mason, 2016; Renström et al., 2023).

Given Bolsonaro’s history of misogynistic, racist, and anti-LGBTQI+ rhetoric and policies, it is likely that women, non-white, and LGBTQI+ voters perceive his supporters as a threat to their rights (Alfonso, 2020; Feitosa, 2021; Martins de Oliveira, 2022; Pereira and Aguilar, 2021). Conversely, given the PT’s closer association with feminist movements and LGBTQI+, evangelical voters may view Lula as a threat to their world views (Araújo, 2022). We expect individuals identifying with these social groups to be more likely to distance themselves from family members who support a side they perceive as threatening. To explore this, we employ the variables: woman (1 = women, 0 = men); non-white (1 = non-white individuals, 0 = white); non-heterosexual (1 = LGB, 0 = heterosexual); and evangelical (1 = evangelical respondents, 0 = otherwise).

Finally, we include several controls. Family size indicates the number of living family members reported (range: 1 to 7, mean 3.20). We also control for respondent characteristics: age (18 to 83, mean 40.94), education (1 to 8, from ‘Incomplete primary education’ to ‘Complete postgraduate education’), and income (1 to 7, from ‘Up to R$1,212’ to ‘More than R$18,181’).

Results

We aim to examine the conditions under which elections prompt family conflict. This first requires exploring the scale of the issue. Figure 1 (panel A) shows the prevalence of such conflicts during the 2022 Brazilian presidential elections. As it indicates, only 3.5% (N = 88) of respondents reported severing ties with a family member over election-related issues, indicating that family fallouts may be less common than often portrayed. ⁶ This is consistent with studies from other contexts, which indicate election-driven family animosities may not be as widely disseminated as alarming headlines suggest. For example, as a UK opinion poll reveals, only about 6% of respondents reported severing ties with family members due to Brexit disagreements (Woodcock, 2019). As Chen and Rohla (2018) reiterate, although in 2016, the year of Trump’s election, families with divergent electoral preferences had shorter Thanksgiving dinners, they were still breaking bread during the holiday.

Figure 1.

The prevalence of election-prompted conflicts (panel A) and electoral disagreements within families (panel B). Note: The unit of analysis is the individual respondent (N = 2,501). Panel A displays the share of respondents who declared having experienced conflict with at least one member of their family. Panel B displays the share of respondents who declared having family members who voted for the opposing candidate. We assign a value of 0 to respondents (a) whose relatives voted for the same candidate as themselves, (b) whose relatives did not vote, and (c) who did not know how their relatives voted.

The rarity of election-driven conflict aligns with findings on the family unit as the bedrock of political socialisation. Despite widespread affective polarisation among voters in the 2022 presidential elections, family units largely maintained cohesive electoral preferences. As shown in Figure 1(panel B), in our sample, 66.69% of respondents reported fully cohesive families, where all members supported the same candidate or had unknown disagreements. Among those with disagreements, 53.54% had only one family member supporting the opposing candidate. These figures reflect strong congruence in electoral choices within Brazilian families, with few cases of relationship severing due to the elections.

In this paper, we investigate the factors that explain election-driven family conflicts. To account for the low incidence and high dispersion of family conflicts (our dependent variable), we employ negative binomial regressions, a generalisation of Poisson regression, which loosens the restrictive assumption that the variance and the mean are equivalents. This is appropriate since, in our data, the variance is greater than the mean. This empirical strategy allows us to identify factors that increase or decrease the risk of a given event occurring in samples where the outcome of interest is skewed. To facilitate interpretation, we report odds ratios: values above 1 indicate factors that increase the odds of family conflicts; values below 1 indicate reduced odds of the event taking place.

As shown in Figure 2, Model 1, political salience, gender, ethnicity, and religious affiliation do not appear to be significantly associated with the risk of election-driven family conflict. Four variables produce statistically significant coefficients. First, as expected, respondents whose families had a higher number of electoral divergences (i.e. relatives who supported the ‘other side’ of the electoral dispute) have higher odds of severing ties with at least one family member. More specifically, for each additional family member who voted differently from the respondent, the risk of conflict increases by 1.56 (i.e. 56%). ⁷ In contrast with our expectations and the scholarship on affective polarisation, we also find that respondents’ antagonistic attitudes towards Lula and Bolsonaro supporters decrease the risks of family conflict. Specifically, a one-unit increase in the scales of ALS and ABS is associated with a 13% and 7% lower risk of election-driven conflicts, respectively. Finally, we also find that respondents’ sexual orientation is highly associated with the risk of family conflict. Specifically, LGB respondents have a chance of experiencing family conflict that is 2.66 times (i.e. 166%) higher than their heterosexual counterparts.

Figure 2.

Factors associated with a higher risk of election-driven family conflicts. Note: *p < 0.1, **p < 0.05, ***p < 0.01. The unit of analysis is the individual respondent (N=2,501). The figure displays odds-ratio estimates from negative binomial regression models with robust standard errors. Models also include controls for family size, age, education, and income. Confidence intervals are at the 95% level. Appendix E reports tables of full estimates.

We expect electoral divergences to be at the root of family conflicts, so we produce models that interact the variable electoral divergences with other variables that achieve statistical significance in Model 1, namely, ALS, ABS, and non-heterosexual. These interactions better capture our theoretical expectations as they allow us to estimate how each of these factors influences the odds of conflict when there are electoral divergences within one’s family. As estimates from Models 2 and 3 show, when family members vote for different candidates, antagonistic attitudes towards Lula and Bolsonaro supporters do not seem to affect the risk of family conflict.

Conversely, as Model 4 indicates, when embedded into family units where there are divergences of electoral preferences, LGB respondents continue to be at a greater risk of severing ties with their relatives. Precisely, non-heterosexual respondents who face electoral divergence in their families are exposed to a risk of experiencing conflict that is 1.53 (53%) higher than heterosexuals. Employing estimates from Model 1 to plot the marginal effects of electoral divergences for heterosexual and non-heterosexual respondents offers similar results (Appendix F), confirming that LGB respondents face a higher risk of election-driven conflicts when they encounter electoral divergences within their families. Analyses presented in Appendix G, which reports standardised beta coefficients, also indicate that, after ED, non-heterosexual is the strongest predictor of electorally driven family conflicts. Our results are substantively equivalent when employing OLS regressions (Appendix H) and conducting analyses with an alternative operationalisation of non-heterosexual (Appendix I). ⁸

We interpret these results as evidence that sexual orientation drives most of the conflict in our sample. In fact, LGB respondents account for more than 30% of the cases of election-driven family conflicts. There are various possible mechanisms underpinning this result.

First, sexual minorities may be more inclined than other groups to develop political identities that differ from those of their parents (Turnbull-Dugarte, 2024). Second, LGBTQ+ voters tend to be more politically engaged than other demographics (Daoust et al., 2024; Çakır et al., 2023), which may lead them to view elections as having higher stakes. Finally, it is possible that sexual minorities experience increased stress compared to other groups when facing the prospect of a far-right candidate winning an election (Gonzalez et al., 2020; Swank, 2020). Although this is not the main focus of our paper, our exploratory analyses suggest that, in the Brazilian context, LGBTQ+ voters’ heightened political engagement and strong opposition to Bolsonaro (Appendix J) underpin their greater propensity to face election-driven family conflicts.

Conclusion

From the US to Europe and Latin America, recent elections have divided societies, raising concerns about the future of social relations (Torcal and Thomson, 2023). This paper offers the first assessment of election-driven conflict within families – a core building block of societies.

Employing novel data from a survey carried out in the aftermath of the highly polarised 2022 Brazilian elections, we find that the prevalence of election-driven family conflicts in Brazil is low. This finding aligns with UK polls indicating the low impact of Brexit on family animosities (Woodcock, 2019) and Iyengar et al. (2018), which conveys US families tend to share political views despite broader societal polarisation. This suggests that anecdotal evidence may be skewing ‘conventional wisdom’ in overestimating the capacity of elections to break up families. However, our study is limited to conflicts among close family members, and that persist in the aftermath of elections; further research is needed to explore conflicts among distant relatives or friends, and that take place before and during election campaigns.

As our study also shows, even if not widely spread, election-driven family conflicts disproportionately affect one group: LGB respondents. Non-heterosexual respondents represent one-third of the cases of election-driven family conflicts. These findings contribute to a growing scholarship on the harmful effects of polarised elections on the LGBTQI+ community (e.g. Gonzalez et al., 2020; Swank, 2020) and underscore the necessity for further investigation into the varied impacts of emerging political dynamics on marginalised groups.

Supplemental Material

Supplemental Material - What explains election-driven family conflicts?

Supplemental Material for What explains election-driven family conflicts? by Malu A. C. Gatto, João Gabriel Leal and Débora Thomé in Research & Politics