Explaining Backlash: Social Hierarchy and Men’s Rejection of Women’s Rights Reforms

Abstract

Governments promote gender-sensitive policies, yet little is known about why reform campaigns evoke backlash. Drawing on social position theory, we test whether marginalized (women’s organizations) or intrusive (Western donors) messengers cause resistance across public rights (quotas) and private rights (land reform). Using a framing experiment implemented among 1,704 Malawians, we find that females’ attitudes are unaffected by campaigns, while backlash occurs among patrilineal and matrilineal males. Backlash among men is more common for sensitive private rights (land reform) than public rights (quotas) and Western donors than women’s organizations, suggesting complex effects generally more consistent with the intrusiveness hypothesis.

Governments promote gender-sensitive policies, yet little is known about whether and why campaigns lead to backlash among groups whose power is challenged by women’s empowerment. Public agencies and development practitioners craft campaigns aimed at increasing public acceptance of gender-sensitive reforms (Cloward 2014, 2016; Duflo 2012; True 2003, 377). For example, in Morocco, a 2003 legal reform raised the minimum age of marriage for girls to 18 years and for the first time required women to sign their marriage contract. To support these legal changes and counter social resistance, women’s groups organized a bus campaign. Joined by trusted leaders including Islamist women’s groups, they visited rural areas to educate women, girls, and their communities about their rights under the reformed Personal Status Code (Salimé 2011).

Yet emerging research highlights troubling evidence that messaging campaigns may actually increase public resistance, leading social groups who are already resistant to change to double down in their opposition (Bush and Jamal 2015; Hughes, Krook, and Paxton 2015; Molden and Higgins 2005). If this happens, the governments and development organizations behind the proposed changes may do more harm than good. For example, in Malawi, Muriaas et al. (2019) found that endorsements designed to reduce public acceptance of child marriage triggered backfire effects among men, prompting them to become more hostile toward banning child marriage when they heard about a campaign to end the practice.

This article addresses whether and why endorsements of reforms that impact resources and power distributed unevenly along gender and lineage hierarchies cause backlash. Specifically, it explores how domestic women’s organizations and Western donors—two actors that differ in terms of their resources and therefore ability to enforce reforms—influence the acceptance or rejection of different types of women’s rights reforms. It does so among men and women from patrilineal and matrilineal groups characterized by different gender relations.

Drawn from social hierarchy theory, which expects opposition from members of high-status groups when low-status groups challenge existing status hierarchies (Ridgeway 2014; Rudman et al. 2012), we test two competing hypotheses about messenger impact: (i) The “intrusiveness” hypothesis (Blalock 1967), which anticipates backlash among members of high-status groups who are threatened by someone with power (e.g., Western donors); and (ii) the “marginalized” hypothesis, which expects backlash among members of high-status groups if members of a disadvantaged group (e.g., women’s organizations) try to elevate their position (Bobo 1999; Ridgeway 2014). As messengers, women’s organizations are the more marginalized and Western donors the more intrusive and powerful with resources to press for potentially threatening reforms.

Our research is novel in several ways. Unlike previous endorsement studies, we examine two types of gender rights—public and private rights—linked conceptually to power (public rights) and sensitive resources (private rights). Crucially, some types of gender reform are likely to be met with more resistance than others, depending on whether they challenge formal power relations (such as in the case of public or non-doctrinal rights) or resource distribution (in the case of private or doctrinal rights), with the latter often most prone to social resistance in Middle Eastern and African societies (Htun and Laurel Weldon 2018; Sadiqi 2008). Further, we do so in the context of two lineage groups with differing degrees of rights afforded to women. Specifically, Malawi’s society is made up of patrilineal and matrilineal groups, each with different gender-based distributions of resources and power in areas such as land rights, marriage rituals, custody of children, and residence. Matrilineal and patrilineal social institutions shape the underlying social norms and create a climate affecting the likelihood of a backlash against reform efforts.

To explain why backlash effects occur, we argue that it is crucial to consider social hierarchies—namely the power differential between the messenger’s group and the recipient’s group—paying attention to whether the messenger is intrusive (i.e., Western donors) or marginalized (i.e., domestic women’s organizations; Ridgeway 2014). We theorize and test for complex, intersectional impacts of endorsements across several groups of recipients (Crenshaw 1991; Hancock 2007) and expect greater backlash for sensitive private rights (e.g., land reform) than public rights (e.g., quotas) and among male and patrilineal citizens than female and matrilineal respondents.

To test these mechanisms, we embedded a survey experiment in the Local Governance Performance Index (LGPI), an original survey conducted in Malawi in 2016 among 1,704 respondents, in which we randomly assigned respondents to six groups to receive a control or endorsement of land reform or quotas from Malawian women’s organizations or Western donors. We find that females’ attitudes are unaffected by campaigns, but both messengers depress support among patrilineal and matrilineal males. Backlash among men is more common for sensitive private rights (i.e., land reform) than public rights (i.e., quotas) and Western donors than women’s organizations, suggesting complex effects generally more consistent with the intrusiveness hypothesis.

By demonstrating the importance of studying identity to understand contemporary social and political development, our study complements and extends existing literature on identity politics, gender, and international development. Our finding contributes to the comparative literature on the politics of women’s rights and social customs by demonstrating that private and public gender rights are quite different from one another in terms of the potential threats that they pose to the status quo (Bush and Jamal 2015; Htun and Laurel Weldon 2018; Muriaas et al. 2019; Sadiqi 2008). It also speaks to work that shows that men’s views are shaped by endorsements and other context cues much more than women’s (Barnes and Córdova 2016).

Our article proceeds as follows. First, we provide an overview of the Malawi context, focusing specifically on lineage systems and the issues of electoral gender quotas and land reform. Second, we outline our theory and hypotheses. Third, we describe our data and methods. Fourth, we present and discuss the results. Finally, we discuss the conclusions and implications of the study on identity politics and development.

Malawi: Gender Reforms in a Society with Two Lineage Systems

With its two lineage systems and history of public and private gender reforms advocated by women’s organizations, Malawi provides an ideal context in which to study how social hierarchy shapes the presence and extent of backlash. In matrilineal systems, the position of women is understood as less subordinate to that of men than in patrilineal systems. The two lineage systems are distinguished by land and property inheritance, succession patterns among traditional leaders, and marriage customs—the core features of lineage systems with repercussions for how resources, power, and status are distributed in society. In patrilineal societies, inheritance passes through the male line—that is, from father to son. The patrilineal lineage system in Malawi is thus established by tracing descent exclusively through the male bloodline (Kishindo 2010). In typical patrilineal traditions, a man often pays a bride price or lobola to a prospective wife’s parents to establish his right to take the wife and any subsequent offspring to his village (Bhaumik, Dimova, and Gang 2013). In matrilineal societies, family property is technically the wife’s property, and inheritance practices follow the mother’s line (Peters 2010). As opposed to patrilineal groups, family residence is uxorilocal, whereby at marriage, the husband relocates to live with his wife and her relations, and offspring take on their mother’s ethnic identity (White 2007). Ancestral property is transmitted from mother to daughter, while wealth management is more or less jointly shared by both sexes (Brulé and Gaikwad 2020). Men might still occupy positions of authority in the household and the community, as well as in the state and traditional political institutions, but only in matrilineal societies is traditional authority (i.e., the position of chief) also bestowed upon women.

Most of Malawi’s population lives in areas with dominant matrilineal lineage systems (e.g., the Chewa, Yao, Lomwe), although a minority in these areas follow patrilineal customs (e.g., the Tumbuka, North Ngoni, Sena). Matrilineal men are overall more acquainted with sharing resources and power and this may affect the extent to which they consider access to land and elected offices as their “rightful prerogatives” (Bobo 1999). Yet one could still argue that women in Malawi, regardless of their lineage customs, belong to a subordinate social group, given that the country has a high level of gender inequality and is ranked 174th out of 187 countries on the Gender Inequality Index (UNDP 2019). How men from patrilineal and matrilineal groups will respond to campaigns to improve women’s power and resources must be empirically verified (Crenshaw 1991). Matrilineal men already have a relatively less privileged status and may resist further erosion of their rights; or, matrilineal men may hold more gender-equal views and respond with little backlash to potential reforms. Patrilineal men may see any threat to their power as unwelcome and react with hostility toward equal rights.

Promoting Gender Status Reforms

Many gender-sensitive reforms in both the private as well as public arenas have been proposed or implemented in Malawi with the support of women’s organizations and Western donors. For example, working with Malawi’s National Women’s Lobby and Rights Group, multilateral and bilateral Western donors fund a gender commission within the Malawi Law Commission which ensures that laws promote and do not hinder women’s advancement (Chiweza, Wang, and Maganga 2016). Examples include proposed electoral gender quotas, which have been debated but not adopted, and gender-sensitive land reforms, which recently passed into law.

Yet while crucial to combating inequalities, reforms challenge the rights, resources, and status that men as a group feel “entitled to enjoy” (Bobo 1999, 450). This is particularly true of private rights, linked to resources. Private or doctrinal rights, such as those related to marriage, divorce, and inheritance, challenge status and power to a greater degree than public or non-doctrinal rights linked to power, such as women’s access to education, politics, and the labor force (Htun and Laurel Weldon 2018; Sadiqi 2008). But public rights such as electoral gender quotas are also contested. Equal representation of women in politics—a public right that alters the formal distribution of power—is a recurring theme in Malawi’s public debate, as men win approximately 80 percent of the seats in the National Assembly. Even if experimental research finds that Malawians have a slight preference for female candidates, women often experience “sexist slurs from male competitors” in real elections (Clayton et al. 2020).

Further, research demonstrates that women in Malawian politics are disadvantaged by patronage politics and cultural norms (Kayuni 2016; Tiessen 2008). As long as women remain marginalized in political parties (Kayuni and Muriaas 2014), women’s activists will most likely continue their call for more effective measures to increase the number of women in politics. Indeed, a provision on quotas for women in politics was removed from the Gender Equality Bill before its tabling in parliament in 2013. Yet, efforts to implement an electoral gender quota have so far failed due to elite resistance (Wang et al. 2020). In an interview conducted by the author, an officer in the Ministry of Gender explained that male MPs strongly opposed the provision on quotas in politics in the bill, arguing that it was discriminatory. “Male MPs were so loud on this [quotas in the gender equality bill]. They argued that everyone should compete on equal footing.”¹ This indicates that at least male representatives react negatively towards reforms that challenge the “gender status hierarchy” in elected politics (Ridgeway 2014).

Land reform that is done in a gender-sensitive way is a private rights issue in Malawi affecting the distribution of resources in society (Chiweza, Wang, and Maganga 2016). Because land rights (i.e., property rights) are generally allocated by customary law, land acquisition is closely related to family norms and cultural practices linked to lineage customs (Pashane Zuka 2019). The formalization of land rights at times leads to a dissonance between traditional modes of intergenerational transmission of land in matrilineal societies, which pass land on through the maternal line, and the formal landholding system modeled on patrilineal English legislation (Berge et al. 2014). Because questions related to women’s matrilineal rights to land have typically been left out of discussions on tenure security and equitable access to land (Kaarhus 2010), women may lose their land if the underlying reform process does not respect matrilineal customs. In 2016, for instance, the parliament passed four land-related bills that the opposition claimed would disadvantage poor landowners, including most women (Nyondo 2016). Domestic women’s organizations, including the leading NGO Gender Coordination Network, expressed concern that, “the laws that are there governing land are not as gender-sensitive as one would want them to be, and … they [the Customary Land Bill and the Land Bill] may not improve things in favor of women as we want them to.”² With the stakes so high for women, who face discrimination in elections and maintaining their land rights, we need to know more about how reformers can sensitize the public to gender equality in land reform without creating backlash. Given customs in their communities, matrilineal men may be more threatened than patrilineal men by reforms that ensure women’s continued access to land. Yet how patrilineal and matrilineal men will respond to campaigns that seek to safeguard women’s access to land ownership has not been explored.

Theoretical Framework: Challenging Status Hierarchies

Hogg, Abrams, and Brewer (2017) argue that women’s rights messages may be understood as group-directed criticism toward men and more conservative or patrilineal groups in society, sharpening identity-based polarization and leading to backfire effects. Developed by Tajfel to understand racial prejudice, social identity theory posits that individuals place themselves and others into social groups and hold more positive views of their ingroup than other outgroups (Abrams and Hogg 1990; Turner 1982). Groups that are in conflict or competition will be particularly likely to develop strong biases about the outgroup and come to regard their social ingroup more highly. Experimental research demonstrates the rapid emergence of outgroup bias even in low-stakes situations such as games played by teams in the lab with few organic, real-world bases for social affiliation (Tajfel et al. 1971).

In view of this, studies in social position and status condition theory warn of backlash among high-status groups whose power, status, and resources are threatened. High-status individuals have more to lose when their power and resources are challenged than lower-status individuals. As a result, members of high-status groups are psychologically primed to view campaigns as personal criticism and react defensively to the reforms (Hornsey, Trembath, and Gunthorpe 2004, 500). One way of understanding this mechanism is that dominant group members react because they believe that “subordinate group members are encroaching on their rightful prerogatives” (Bobo 1999, 450). Hence, even if we cannot assume that all women are supportive of a women’s rights agenda (Krook 2015), social position theory leads us to expect reform messaging to have fewer negative effects on women than men because women do not belong to a group whose privileges are threatened by gender reforms (Ridgeway 2014).

Numerous studies offer evidence of such backlash to messages, in the form of a backfire effect (Ahluwalia 2000; Blalock 1967; Bush and Jamal 2015; Muriaas et al. 2019; Ridgeway, Johnson, and Diekema 1994). The attitudes of those whose interests are not threatened by the reform remained unchanged, while social groups who are threatened by the reforms may defend status hierarchies when they think that the proposed reforms “go too far” (Ridgeway 2014, 7). Status-based social differences, such as gender and lineage, are woven into the resources and power of social actors, impacting the likelihood of backfire effects.

Reforms targeting gender inequality are particularly likely to cause a reaction because the elevation of the status of the subordinate group—that is, women—implies a contrast to the opposite sex—men—whose status as dominant would be challenged (Ridgeway and Correll 2004). Nevertheless, endorsement effects may only occur when recipients believe reforms will affect their power and resources. Whether a message causes a backfire effect among those who feel their prerogatives are threatened or criticized is likely to depend on whether the messenger is perceived as “intrusive” (Blalock 1967) and capable of implementing the reform. Those who do not think that the reform will impact their status will remain unaffected by the endorsement, regardless of the intrusiveness and capability of the endorser.

The question, therefore, is what kind of messenger is likely to trigger resistance, among whom, when, and why. The intrusiveness hypothesis posits that as the power of the messenger increases, so too does the perceived likelihood of the reforms coming to fruition, thus increasing the likelihood of a backlash among members of the threatened group (Blalock 1967). Status beliefs tend to bias people’s willingness to pay attention to and evaluate the content of a message. As argued by Ridgeway (2014), the same argument might sound more convincing coming from the advantaged rather than the disadvantaged, hence there will be backfire effects if the messenger is advantaged. In contrast, the “marginalized” hypothesis expects a backlash effect among members of high-status groups if members of disadvantaged groups try to elevate their position (Bobo 1999; Ridgeway 2014). The potential for disadvantaged groups to cause a hostile reaction can be pronounced, for instance when such groups receive increased public exposure for their cause and convey a message that challenges the status hierarchy—especially from high-status reformers (Ridgeway, Johnson, and Diekema 1994).

Hypotheses: Messenger, Recipient, and Gender-Sensitive Issues

We develop an experimental research design that allows us to test the outlined expectations. Based on previous research that finds an endorsement backlash among men (Muriaas et al. 2019), and considering social status theories, we expect that endorsements of women’s rights reforms decrease support among men but do not affect support among women. However, we expect that the pattern of resistance to women’s rights reforms is likely to be more complex than this and suggests some additional conditions that are likely to interact with this general pattern. When confronted with an endorsement, the social hierarchies that men belong to may affect whether they experience the messenger as a threat or not. Recipients and messengers are embedded within a social structure in which groups enjoy unequal status and power that is intersectional (that is, mutually constituted by social hierarchies). Given this, the precise nature of the impact is likely to be complex and intersectional, meaning that it must be empirically verified (Crenshaw 1991; Hancock 2007). We, therefore, build a theoretical framework that allows us to test the “intrusiveness” and “marginalized” hypotheses while considering the social status of the recipient and the sensitivity of the reform.

Messenger Status

The “intrusiveness” and “marginalized” hypotheses offer different expectations about the impact of Western donors and women’s organizations on attitudes on the potential for backlash. The “intrusiveness” hypothesis (Blalock 1967) expects a backlash effect among members of high-status groups who are threatened by someone with power, such that endorsements from Western donors will decrease support more than those from domestic women’s organizations. The “marginalized” hypothesis expects a backlash effect among members of high-status groups if members of a disadvantaged group seek to elevate their position (Bobo 1999; Ridgeway 2014).

Existing literature and data point to the “intrusiveness” hypothesis being more plausible. Afrobarometer data show a higher level of distrust of external donors than domestic NGOs. According to the Afrobarometer, Western organizations are seen more negatively than Malawian civic organizations, even though there is some overlap in their priorities, which include promoting gender equality. The Afrobarometer data do not include questions on women’s organizations specifically. However, 7 percent of Malawians see their local civic organizations and NGOs as too influential, while 15 percent see Western donors and organizations as having too much influence (figure 1).

Figure 1.

Perceptions of international organizations and Malawian civic organizations’ influence.

Note:Afrobarometer (2008). “How much influence do international donors/NGOs have?” “How much influence do Malawian civic organizations/NGOs have?”

Existing research suggests the importance of local sources of authority in ensuring effective attitude change and this leads us to expect smaller backfire effects from the endorsements of women’s organizations than those from Western donors. In her work on international norm messaging and gender issues in Kenya, Cloward (2015) argues that when transnational activism of international norms collides with local norms, local practices on female genital mutilation are most likely to be abandoned when there are non-circumcising groups in proximity and local elites are among the first movers.

Recipient Status

As noted, men have the most to lose when it comes to potential reforms. It is instructive that in the Afrobarometer, men are more likely than women to believe international and domestic non-governmental organizations wield too much influence (figure 2). Eleven percent of women and 15 percent of men state that these organizations have too much influence, a significant gender difference (P < 0.05). This suggests that men may be more critical and distrustful of the donor agendas than women even on topics that do not include gender issues.

Figure 2.

Perceptions of international organizations, by respondent gender.

Note:Afrobarometer (2008). “How much influence do international donors/NGOs have?”

In the context of Malawi, recipients living in the context of patrilineal lineage norms may react differently to both the type of messenger and the message than those with matrilineal norms. Recent research shows that more progressive norms about the role of women in society exist in areas of Malawi where women live in matrilineal societies and have greater access to land and other assets than those in patrilineal societies (Robinson and Gottlieb 2021). We thus expect that respondents’ lineage tradition—at the intersection of their gender—will shape the effects of endorsements. Accordingly, we expect a backfire effect to occur more strongly among patrilineal men than matrilineal men irrespective of whether the endorsement is from a Western donor or a women’s organization, but we recognize that the effects may be complex and intersectional (Crenshaw 1991). The mechanisms explaining this are, however, different. An endorsement from Western donors appears like a threat to patriarchal structures as they are known for their gender equality agenda. The backlash when members of the domestic women’s movement speak up is due to not being used to—or feeling uneasy—when women “complain” about their subordinate position.

Issue Type

Finally, we expect that backfire effects among men are likely to be more pronounced when they relate to kinship practices and traditions, including inheritance and property rights, than for issues concerning women’s political participation because of the crucial relationship between kinship, reproduction, and group interests (Charrad 2001). Private rights threaten entrenched interests to a greater extent than public rights because they fundamentally reorder power relations between men and women. Thus, we expect backfire effects will be larger for land rights than women’s access to political leadership, independent of the messenger.

Hypotheses:

General:

H1: Endorsements will decrease support among men but not affect support among women.

Intrusiveness—considering lineage:

H2a: Endorsements from Western donors will decrease support more than those from domestic women’s organizations, but more so among patrilineal men than matrilineal men.

Intrusiveness—considering gender reform:

H2b: Endorsements from Western donors will decrease support more than those from domestic women’s organizations, but more so for land reform than quotas.

Marginalized—considering lineage:

H3a: Endorsements from domestic women’s organizations will decrease support more than those from Western donors, but more so among patrilineal men than matrilineal men.

Marginalized—considering gender reform:

H3b: Endorsements from domestic women’s organizations will decrease support more than those from Western donors, but more so for land reform than quotas.

Data and Methods

To test these hypotheses, we use a survey experiment embedded in the LGPI, an original household survey conducted in 2016 with a team from the Program on Governance and Development (GLD). The LGPI used probability proportional to size (PPS) and random within-household sampling in seventeen traditional authority (TA) areas in rural Malawi and an additional five local government wards in urban centers to select a nationally representative sample of Malawians. We added post-stratification weights to correct for differential response rates across population subgroups, especially an underrepresentation of men, many of whom were working in neighboring villages, towns, or outside the country. The survey was administered by a team of trained Malawian interviewers managed by Dr Boniface Dulani and conducted face-to-face in three languages. The survey interview lasted an hour and had a response rate of 94.5 percent (AAPOR Response Rate 1).³ While 7,500 Malawians participated in the survey, 1,704 randomly selected respondents received the experimental prompts, including eight who refused to answer the outcome variable.

Design and Outcome Measures

We assigned respondents to one of six endorsement conditions using a randomized block design so there were roughly equal numbers of respondents in each of the six groups (see table 1). At random, respondents were told either about a quota reform or a land reform. The statement about the reform was by an endorsement from Malawian women’s organizations (“Malawian WOMEN’S ORGANIZATIONS are supporting a new law to …”), Western donors (“A group of WESTERN DONORS is supporting a new law …”), or no endorsement (i.e., the control). They were then asked to indicate their support for the reform (“Would you be not at all likely [=1], somewhat unlikely [=2], somewhat likely [=3], or very likely [=4] to support this policy?”). The dependent variable—support for the quota or land reform—was measured on a four-point scale.

Table 1

Experimental design and distribution of 1,696 respondents across conditions

Quotas	Condition 1. WDs: (N = 282)
	A group of WESTERN DONORS is supporting a new law to increase the number of women elected to parliament. They say that reserving a set number of seats for women will result in a higher number of women being elected.
	Condition 2. WOs: (N = 272)
	Malawian WOMEN’S ORGANIZATIONS are supporting a new law to increase the number of women elected to parliament. They say that reserving a set number of seats for women will result in a higher number of women being elected.
	Condition 3. Control: (N = 293)
	There are discussions of a new law to increase the number of women elected to parliament. Reserving a set number of seats for women will result in a higher number of women being elected.
Land Rights	Condition 4. WDs: (N = 276)
	A group of WESTERN DONORS is supporting a new law to ensure that all Malawian women have and can exercise equal rights to own and control land. They say that changing the law will make women more economically independent.
	Condition 5. WOs: (N = 298)
	Malawian WOMEN’S ORGANIZATIONS are supporting a new law to ensure all Malawian women have and can exercise equal rights to own and control land. They say that changing the law will make women more economically independent.
	Condition 6. Control: (N = 275)
	There are discussions of a new law to ensure all Malawian women have and can exercise equal rights to own and control land. Changing the law will make women more economically independent.

Note: A total of 1,704 people were included in the experiment. In all, 1,696 answered the outcome measure.

The treatments were effectively randomized across sampling units and response rates were constant across frames.⁴ Because the treatment was significantly related to gender and education level, the models control for respondent gender, interviewer gender (Benstead 2014), lineage, gender attitudes, rural residence, education, income, and age, and also include post-stratification weights.

We conducted qualitative interviews between January 2014 and January 2015 and focus groups in December 2015 and used this information to develop the question wording of the experiment.⁵ We selected gender quotas and land reforms because of their appropriateness as examples of private and public rights and the fact that they were salient issues in Malawi at the time of the survey. Following this field research, we carefully crafted the control and treatment statements, utilizing a neutral control that takes the same amount of time to read as the treatment prompts. The control statement refers simply to the discussions of a new law—the closest approximation to a control statement feasible.

Measurement of the Independent and Control Variables

Recipient social group and status. To measure recipient social group and status, we divided respondents into four groups according to their gender and lineage system using the question: “In your family, is lobola (bride price/dowry) paid when people get married?” measured as “Yes” (=0/patrilineal) and “No” (=1/matrilineal). The group sizes are: patrilineal male (n = 323), matrilineal male (n = 276), patrilineal female (n = 525), and matrilineal female (n = 553) (table 2).

Table 2

Distribution of respondents across conditions

		Experimental condition
		Quotas			Land reform
		1	2	3	4	5	6	Total
Recipient identity (intersectional identities)		Condition 1 (WDs)	Condition 2 (WOs)	Condition 3 (Control)	Condition 4 (WDs)	Condition 5 (WOs)	Condition 6 (Control)
Male respondents	Patrilineal	52	46	47	64	59	55	323
	Matrilineal	48	39	40	53	49	47	276
Female respondents	Patrilineal	84	89	99	76	96	81	525
	Matrilineal	92	96	103	79	93	90	553
Total		276	270	289	272	297	273	1,677

Respondent intersectional identity × endorsement interaction indicators. The primary independent variables in the analyses are eleven interaction indicators for each combination of respondent identity by endorsement conditions (Quotas: WD, WO, or control; Land reform: WD, WO, or control). Patrilineal males serve as the reference category. This coding scheme rather than the traditional full factorial of the two variables (i.e., main effects for the condition and recipient identity group with their interactions) simplifies testing of our hypotheses. However, the two coding schemes are mathematically equivalent (table 3).⁶

Table 3

Description of unweighted dependent measures

	Mean	SD	N	Patrilineal man	Matrilineal man	Patrilineal woman	Matrilineal woman
All experimental conditions	3.57	0.85	1,665	3.52	3.55	3.51	3.48
				Pearson’s χ2 (9) = 63.3094/Pr = 0.000***
Comparison across the six conditions
Condition 1 (WDs): Quotas	3.54	0.88	275	3.59	3.48	3.54	3.57
				Pearson’s χ2 (9) = 17.4605/Pr = 0.042*
Condition 2 (WOs): Quotas	3.49	0.90	265	3.33	3.47	3.51	3.57
				Pearson’s χ2 (9) = 12.2919/Pr = 0.197
Condition 3 (Control): Quotas	3.55	0.90	288	3.38	3.72	3.47	3.65
				Pearson’s χ2 (9) = 24.8283/Pr = 0.003**
Condition 4 (WDs): Land Reform	3.52	0.82	271	3.54	3.43	3.31	3.80
				Pearson’s χ2 (9) = 18.6169/Pr = 0.029*
Condition 5 (WOs): Land Reform	3.61	0.83	294	3.48	3.43	3.64	3.76
				Pearson’s χ2 (9) = 30.0041/Pr = 0.000***
Condition 6 (Control): Land Reform	3.70	0.71	270	3.73	3.80	3.56	3.75
				Pearson’s χ2 (9) = 12.1897/Pr = 0.203

***

P < 0.001,

**

P < 0.01,

*

P < 0.05.

The treatments were effectively randomized across sampling units and response rates were constant across frames. Inclusion of additional demographic control variables for heterogeneous treatment effects, unless theoretically justified, is unnecessary in a randomized treatment assignment and can bias estimation of the average treatment effects (Mutz 2011). However, because the treatment was significantly related to gender and education level—and because prior research shows that respondent gender, interviewer gender (Benstead 2014), lineage, gender attitudes, rural residence, education, income, and age are related to attitudes toward gender equality in different societies (Alexander and Welzel 2011; Muriaas et al. 2019)—we control for these factors and include post-stratification weights the regressions.

Analytic Strategy

We use ordered logistic regression to examine how endorsements by the international and domestic messengers shape attitudes toward the public and private rights in the sample as a whole and within the four identity groups. Because the dependent variable is measured on a Likert-type scale, we use ordered logistic regression rather than means comparisons (Long and Freese 2014). This allows us to see whether endorsements are effective and whether their efficacy depends on the recipient–messenger interaction and the sensitivity of the reform.

For clarity of presentation, the average effects of the endorsement in the sample as a whole are summarized in figure 3 and table 4, followed by a similar presentation of the heterogenous effects in figures 4 and 5 and tables 5 and 6.⁷ The figures summarize the predicted probabilities of each response and thus illustrate the substantive size of the effects of each endorsement relative to the control. The tables show the significance level when comparing the effects of each endorsement to the other (i.e., WD vs. WO) and to the control (i.e., WD vs. Control and WO vs. Control) for each reform type (i.e., Quotas vs. Land Reform).

Figure 3.

Predicted probability of supporting quotas and equal land rights, by endorsement (average treatment effects).

Figure 4.

Predicted probability of supporting quotas and equal land rights, by endorsement (man, patrilineal).

Figure 5.

Predicted probability of supporting quotas and equal land rights, by endorsement (man, matrilineal).

Table 4

Significance levels for pairwise comparisons, average treatment effects

		Not likely at all	Somewhat unlikely	Somewhat likely	Very likely
Quotas	WO vs. WD (2 vs. 1)	0.398	494	0.550	0.457
	Control vs. WD (3 vs. 1)	0.529	0.512	0.527	0.498
	Control vs. WO (3 vs. 1)	0.168	0.263	0.289	0.214
Land rights	WO vs. WD (5 vs. 4)	0.743	0.199	0.068†	0.212
	Control vs. WD (6 vs. 4)	0.001***	0.002**	10.001***	0***
	Control vs. WO (6 vs. 5)	0.03*	0.042*	0.037*	0.017*

***

P < 0.001,

**

P < 0.01,

*

P < 0.05,

†

P < 0.10.

Table 5

Significance levels for pairwise comparisons, average treatment effects (man, patrilineal)

		Not at all likely	Somewhat unlikely	Somewhat likely	Very likely
Quotas	WO vs. WD (2 vs. 1)	0.702	0.718	0.714	0.705
	Control vs. WD (3 vs. 1)	0.314	0.291	0.266	0.274
	Control vs. WO (3 vs. 2)	0.175	0.217	0.155	0.166
Land rights	WO vs. WD (5 vs. 4)	0.382	0.673	0.996	0.628
	Control vs. WD (6 vs. 4)	0.183	0.157	0.096†	0.133
	Control vs. WO (6 vs. 5)	0.006**	0.002**	0.001***	0.001***

***

P < 0.001,

**

P < 0.01,

*

P < 0.05,

†

P < 0.10.

Table 6

Significance levels for pairwise comparisons, average treatment effects (man, matrilineal)

		Not at all likely	Somewhat unlikely	Somewhat likely	Very likely
Quotas	WO vs. WD (2 vs. 1)	0.577	0.562	0.587	0.570
	Control vs. WD (3 vs. 1)	0.084†	0.055†	0.041*	0.045*
	Control vs. WO (3 vs. 2)	0.120	0.186	0.194	0.158
Land rights	WO vs. WD (5 vs. 4)	0.677	0.368	0.202	0.374
	Control vs. WD (6 vs. 4)	0.001***	0.002**	0.002**	0.001***
	Control vs. WO (6 vs. 5)	0.004**	0.002**	0***	0***

***

P < 0.001,

**

P < 0.01,

*

P < 0.05,

†

P < 0.10.

Results and Discussion

We first examine average treatment effects, allowing us to test H2b and H3b, and then run conditional treatment effects across four respondent groups that intersect respondent gender and lineage custom—patrilineal male, matrilineal male, patrilineal female, and matrilineal female—allowing us to evaluate H1–H3.

Average Treatment Effects

First, we examine average treatment effects to test the intrusiveness hypothesis (considering gender reform (H2b)) and the marginalized hypothesis (considering gender reform (H3b)) in the sample as a whole. Because we find no significant difference between the impact of women’s organizations and Western donors in the sample as a whole, the evidence does not support either hypothesis. However, we find support for a stronger backlash for the sensitive private right than the public right, as anticipated (H2b and H3b). For land reform, the probability of being very likely to support reform when there is no endorsement (i.e., the control) is 84.9 percent, but it diminishes to 76.4 percent with a women’s organization endorsement (P < 0.05 when compared to the control) and 71.4 percent with a Western donor message (P < 0.001 when compared to the control).

Heterogeneous Treatment Effects

Second, by exploring heterogeneous treatment effects, we are able to test all of the hypotheses. Across the models, backfire effects elicited by the endorsement are present only among men, in support of H1. Among female respondents, the endorsements have no significant effect. Following social position theories, this suggests that men belong to a group that perceives its privileges as threatened by the proposed reforms, while women do not.

Messenger Status

It is also important to consider whether one of the messengers is less likely to cause a backlash than the other. The core of the intrusiveness versus the marginalized messenger comparison lies in the presence and relative size of the effects across the two messengers—the Western donors and the women’s organizations. The intrusiveness hypothesis—considering lineage (H2b)—expects that the endorsements from Western donors will decrease support more than those from domestic women’s organizations, and more among patrilineal men than matrilineal men, while the marginalized hypothesis—considering lineage (H3b)—expects endorsements from domestic women’s organizations will decrease support more than those from Western donors, but more so among patrilineal men than matrilineal men.

The evidence suggests little difference in the size of the backlash across the two messenger types. On the whole, women’s organizations elicited strong backfire effects for land reform among patrilineal men and both messengers produced backfire effects for at least one type of reform among matrilineal men. This suggests slightly stronger support for the intrusiveness hypothesis (H2a and H2b) than the marginalized hypothesis (H3a and H3b), although both receive some support.

Recipient Status

Further, the effects are slightly different across men from the two lineage groups, but they only partially support the hypothesis that backfire effects will be larger among patrilineal than matrilineal groups (H2a and H3a). Patrilineal men are significantly less likely by 27.1 percentage points to be “very likely” to support land reform when it is endorsed by women’s organizations than in the control condition for the very likely category (P < 0.001) (table 5 and figure 4). This is larger than the comparable effect among matrilineal men, who are significantly less likely by 19.1 percentage points to be “very likely” to support land reform when it is endorsed by women’s organizations than in the control condition (P < 0.001)—a difference of about 8 percentage points. But there are other large backfire effects among matrilineal men—specifically that they are significantly less likely by 25.6 percentage points to be “very likely” to support land reform when it is endorsed by Western donors than in the control condition (P < 0.001).

Thus, we find large backfire effects among patrilineal males for endorsements of sensitive land reforms emanating from women’s organizations. While this endorsement by women’s organizations of land reform produces smaller backfire effects among matrilineal men, matrilineal men also elicited backfire effects in response to other endorsements, including Western donor endorsements of land reform (relative to the control) and Western donor endorsements of quotas (relative to the control). This effect for land reform is substantial (25.6 percentage points for the “very likely” category), suggesting that matrilineality alone does not automatically translate into acceptance of gender equality when presented by messengers.

Issue Type

The larger backfire effects in the sample as a whole for private rights compared to public rights also replicate across the analysis of heterogeneous treatment effects among men in support of H2b and H3b. Patrilineal men are 27.1 percent less likely to be “very likely” to support land reform when it is endorsed by women’s organizations than in the control condition for the same category (P < 0.001) (table 5 and figure 4). On the same measure, matrilineal men are 19.1 percent less likely to be “very likely” to support land reform when it is endorsed by women’s organizations than in the control condition (P < 0.001) and significantly less likely by 25.6 percentage points to be “very likely” to support land reform when it is endorsed by Western donors than in the control condition (P < 0.001) (table 6 and figure 5). In patrilineal societies, access to land is what privileges men over women and reform is likely to have broad impacts on existing gender hierarchies for all men, regardless of their background. A land reform message may be perceived as criticism of all patrilineal men. In contrast, the need for gender quota reform may be perceived as threatening only for some men, especially those from elite backgrounds.

Complex, Intersectional Effects

Yet, the effects are more complex than our hypothesis anticipates. The backfire effect tends to be larger for land reform among men, but matrilineal men also react negatively when presented with messages advocating for quotas in partial support of H2a. Accordingly, we find partial support for the intrusiveness hypothesis (H2a), although among an unexpected group, in the form of backlash among matrilineal men when they encounter an endorsement from a Western donor. Other complex and intersectional effects emerge as well. In partial support of H2b, we also find backlash for an unexpected issue—that is, against quotas rather than land reform among matrilineal men. Yet in partial support of the marginalized hypothesis (H3a), we find large backfire effects among patrilineal males for endorsements of sensitive land reforms emanating from women’s organizations.

While we can only speculate, it is possible that matrilineal men, whose political institutions are already more open to women, find Western donors more intrusive. Matrilineal men may feel that their group already upholds gender-equal practices and that additional rights are excessive. Thus, while we find strong support for H1 and generally stronger support for the intrusiveness hypothesis, the exact nature of the effects depends on the complex intersection of the messenger power, the male recipient’s lineage customs, and the type of reform. In general, however, backlash is more likely for Western donors, patrilineal males, and land reform than women’s organizations, matrilineal males, and quotas, although effects are possible in all of these scenarios.

Summary and Short Discussion of Results

Our results show strong potential for rights campaigns to produce a significant backlash, especially—although not exclusively—for sensitive private rights and among men. We find support for both the intrusiveness as well as the marginalization hypotheses, which means that for men, different actors—whether foreign or domestic—advocating a reform appear to prime fears or contempt. Fears (intrusiveness) occur when respondents believe that such reforms will become a reality while contempt (marginalization) occurs when someone from a subordinate group in the social hierarchy tries to speak up for themselves (Blalock 1967; Ridgeway 2014).

Our study offers a strong indication that men—across both lineage customs—perceive their status to be threatened by gender reforms and are resistant to attempts by messengers to alter their resources and power. The clearest finding is that men respond to endorsements with resistance, while women are unaffected (H1). We find support for the marginalization hypotheses as women’s organizations elicited strong backfire effects for land reform among patrilineal men (H3a and H3b). Yet we also find support for the intrusiveness hypotheses, as both women’s organizations and Western donor endorsements produced backfire effects for at least one type of reform among matrilineal men (H2a and H2b). Reforms are more likely to invoke opposition (Htun and Laurel Weldon 2018) in the arena of private rights that impact resources than public rights that impact power.

We do, however, find that the effects are more complex than our hypothesis anticipate. Most importantly, we did not expect a backlash among matrilineal men when they encounter an endorsement of gender quota reform from a Western donor. In a study on political participation and lineage systems, Robinson and Gottlieb (2021) found that matriliny is robustly associated with closing the gender gap in political participation. Yet matrilineal men exhibit backlash consistent with the intrusiveness effect when Western donors endorse further erosion of men’s access to political power as a result of electoral gender quotas.

Conclusions and Implications

There is increasing concern among scholars worldwide about the threat of backlash against gender equality reforms, despite numerous examples of progress. Yet in policy discussions and the academic literature, sub-Saharan Africa is portrayed as a region in which progress is still the dominant trend. Our study contributes to this global debate by providing evidence of the potential for backlash against gender reforms and proposing and testing a novel theoretical framework for the conditions under which it is more likely to occur.

Our findings offer a clear message that neither private nor public rights are exempt from the possibility of backlash, even though the threat is greater for sensitive private rights that relate to the distribution of resources than public rights. But it is difficult to predict the circumstances in which complex backlash effects will occur. Endorsements of the private right—that is, land reform—produced a backfire effect among patrilineal and matrilineal men, while quotas did so only for matrilineal men. This indicates that subgroups of men are affected differently depending on what type of law reform being proposed.

These findings have clear implications for the scholarly literature on gender mainstreaming and backlash. Women cannot hold traditional leadership positions in patrilineal communities and they also hold fewer state offices in patrilineal than matrilineal areas (Robinson and Gottlieb 2021). Our results suggest that patrilineal men may be willing to accept a marginal increase in the number of women who hold elected office because they do not see this as a threat to their power, given that it does not change their access to resources (i.e., land) or other prerogatives. While we can only speculate, it may be in matrilineal communities that men already feel that their power and resources are threatened, leading them to worry that having more women in political office will only further undermine their position vis-à-vis women.

Our study complements and extends existing literature on identity politics, international development, and gender in several ways. First, our work shows that social position theory is a useful lens through which to understand the impact of campaigns on public support and resistance. We extend this theory by demonstrating backlash in a novel context—that of Malawi—and show that intrusiveness and marginalization are not competing hypotheses but operate in tandem, leading to complex effects in societies with multiple intersecting identity groups.

Second, our study contributes to a broad literature demonstrating the importance of studying identity to understand contemporary social and political development (Hoffarth and Jost 2017; Jost et al. 2017; Prusaczyk and Hodson 2018). For instance, Robinson and Gottlieb (2021) theorize about the role lineage customs play in shaping social attitudes and find more progressive norms about the role of women in society in matrilineal than patrilineal areas of Malawi.

Third, our findings contribute to the comparative literature on the politics of women’s rights and social customs by demonstrating that gender rights are quite different from one another in terms of the potential threats that they pose to the status quo (Bush and Jamal 2015; Htun and Laurel Weldon 2018; Muriaas et al. 2019; Sadiqi 2008). Like these scholars, we offer evidence in a novel context that endorsements of the more sensitive private right, land reform, produced backfire effects among men from both lineage groups, while quotas were perceived as most threatening among matrilineal men.

Finally, our work speaks to existing studies demonstrating that men’s views are shaped by endorsements and other context cues much more than women’s (Muriaas et al. 2019), whose views are more likely to be “settled” about gender issues. As an example, Barnes and Córdova (2016) find in a cross-national study of support for gender quotas in Latin America that men’s support for quotas is more sensitive to contextual conditions and the credibility of the state, based on their perceptions of good governance. Other research on sexual harassment panels finds larger effects of mixed-gender panels on perceptions of legitimacy among men, who have less well-formulated views on the issue (Clayton, O’Brien, and Piscopo 2019). Together, this extant research along with our results suggests that the use of trusted messengers and organizations is particularly crucial when working with groups who, while not directly affected by reforms, are needed to ensure positive social change.

It is likely that our main hypotheses—intrusiveness and marginalization—are relevant for understanding opposition to gender equality reform beyond Africa. More work is needed both to test these hypotheses in a comparative context, as well as to assess whether Western donors and women’s organizations are the most valid measures of these concepts beyond Malawi and other African cases. We expect that our operationalization and findings may apply to Africa and Asia and perhaps beyond to cases in which reformers are seeking to fundamentally challenge men’s rights and where powerful external actors, such as Western donors, have leverage to make these reforms a reality. It is also possible that internal actors such as a monarch who seeks to reform women’s rights from the top might qualify as an “intrusive actor,” as in the case of Morocco (Sadiqi 2008).

Our findings illustrate the importance of comparative research on the effects of different types of campaigns on public opinion if we are to fully understand the current wave of backlash to gender equality. We need more studies that identify the mechanisms that evoke opposition to reform. Who says what, when, and about what type of law clearly matters. With these new studies, it may be possible to craft a comparative framework that explains the conditions under which efforts by “intrusive” and “marginalized” actors will be effective in shifting public opinion in ways that reduce—or at least do not increase—resistance among segments of the population whose support is most needed to achieve legal and social change.

Notes

1

Interview February 3, 2014.

2

Interview August 12, 2014.

3

Post-stratification weights were added to correct for differential response rates, especially the underrepresentation of men, many of whom were working in neighboring villages or South Africa.

4

See Supplementary material.

5

Forty semi-structured interviews on gender reforms were conducted mainly in Blantyre, Lilongwe, and Zomba with politicians, development partners, civil society actors, civil servants, and academics.

6

table 4 shows that there are significant differences in unweighted groups in the sample as a whole (P < 0.0000), and across groups for all groups for all endorsements except for quota endorsements by WOs and the control statement for land reform.

7

The model includes post-stratification weights for design effects.

Supplementary Data

Supplementary data can be found at [email protected]

Acknowledgments

We thank all the collaborators involved in the Malawi Local Governance Performance Index (LGPI) survey. This project reflects the fruitful collaboration of researchers at the Chr. Michelsen Institute, including Ragnhild Muriaas, Lise Rakner, and Vibeke Wang; the Institute of Public Opinion and Research, including Asiyati Chiweza, Boniface Dulani, Happy Kayuni, Hannah Swila, and Atusaye Zgambo; and the Program on Governance and Local Development, including Adam Harris, Kristen Kao, Ellen Lust, Maria Thorson, Jens Ewald, Petter Holmgren, and Pierre Landry, and Lindsay Benstead during implementation and Ruth Carlitz, Sebastian Nickel, Benjamin Akinyemi, Laura Lungu, and Tove Wikehult in the process of data cleaning and analysis. We gratefully recognize the hard work of colleagues at the Institute of Public Opinion and Research who lead the survey research teams. These include Ellasy Chimimba, Grace Gundula, Steve Liwera, Shonduri Manda, Alfred Mangani, Razak Mussa, Bernard Nyirenda, Charles Sisya, and Elizabeth Tizola. We are grateful for thorough comments from anonymous reviewers at Social Politics and comments on earlier drafts from Shareen Hertel, Sarah Bellows-Blakely, Alice Kang, Farida Jalalzai, Jennifer Piscopo, Henning Melber, Hasan Muhammad Baniamin, Liv Tønnessen, and Vegard Vibe.

Funding

We thank all the funders of the Malawi Local Governance Performance Index (LGPI) survey. This project was made possible with the financial support of the Hicham Alaoui Foundation, the Carnegie Corporation of New York, the World Bank, and Yale University, which funded the development of the LGPI, and the Swedish Research Council (grant number 439201438) and the Research Council of Norway (grant number 233803), which funded implementation in Malawi.