Hostname: page-component-848d4c4894-jbqgn Total loading time: 0 Render date: 2024-06-25T20:00:36.197Z Has data issue: false hasContentIssue false

Insecurity and Support for Female Leadership in Conflict States: Evidence from Afghanistan

Published online by Cambridge University Press:  15 December 2023

Jasmine Bhatia
Affiliation:
Department of Politics, Birkbeck, University of London, London, UK
Steve L. Monroe*
Affiliation:
Department of Political Science, Faculty of Arts and Sciences, National University of Singapore, Singapore
*
Corresponding author: Steve L. Monroe; Email: smonroe@yale-nus.edu.sg
Rights & Permissions [Opens in a new window]

Abstract

While women's political inclusion is central to international conflict resolution efforts, public attitudes in conflict states towards women's political inclusion remain understudied. We expect insecurity to depress support for female political leadership in conflicts where women's political inclusion is violently contested. Citizens wanting security through force prefer male leaders because of stereotypes privileging men's military prowess. However, citizens wanting security through reconciliation also favour men for fear that female leadership would provoke more violence. We assess these expectations with experimental and observational data from the former Islamic Republic of Afghanistan. In the survey experiment, priming respondents to think about insecurity decreases support for female leadership, but only among women. In observational data, insecurity correlates with more polarized attitudes towards women's political representation in some regions and greater support for female leaders in others. Insecurity's impact on public support for female leadership in conflict states may be highly heterogeneous.

Type
Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
Copyright © The Author(s), 2023. Published by Cambridge University Press

Ensuring women's political representation is a pillar of the United Nations (UN) conflict resolution and peacekeeping efforts (United States Institute of Peace 2021). However, how citizens in conflict statesFootnote 1 view women's political inclusion remains understudied. Extensive research finds that feelings of insecurity – prevalent in conflict states – weaken popular support for female political leadership. This work blames gendered stereotypes privileging male leadership in security affairs (Dolan Reference Dolan2014b; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2011; Lawless Reference Lawless2004). Most of this scholarship, however, stems from secure states or post-conflict ones. For citizens living in conflict states where peace has yet to be attained, insecurity may deepen desires for leadership characteristics associated with female stereotypes: inclusion, reconciliation, and clean governance (Anderlini Reference Anderlini2007; Karim Reference Karim2019; Shair-Rosenfield and Wood Reference Shair-Rosenfield and Wood2017). Does insecurity weaken or bolster public support for female leadership in conflict states?

We hypothesize that insecurity places female leaders in a double bind in the court of public opinion in conflicts where women's political inclusion is violently contested. For citizens favouring security through force, insecurity heightens public demands for male leaders because of gendered stereotypes of men's military prowess. At the same time, insecurity causes citizens seeking security through reconciliation to fear that female political leadership will provoke a violent backlash from militant groups opposed to women's political inclusion. Insecurity deprives female leaders of support from citizens who want security through force and citizens who want security through reconciliation.

Insecurity in these types of conflicts may also weaken support for female leaders from their most likely proponents: women. Citizens generally prefer leaders of the same gender (Schwarz and Coppock Reference Schwarz and Coppock2022). However, women often bear disproportionate costs in conflicts (Chishti Reference Chishti2020; Kandiyoti Reference Kandiyoti2007a; Kandiyoti Reference Kandiyoti2007b; United States Institute of Peace 2021) and have stronger preferences for peace than men (Maoz Reference Maoz2009; Maoz Reference Maoz2011; Wood and Ramirez Reference Wood and Ramirez2018). Where women's political inclusion is an axis of contention between warring parties, women may forfeit their greater affinity for female political leadership in favour of a more male-dominated political system they believe can more successfully deliver security. This trade-off between female representation and security is less salient for men, who (to begin with) are likely to have less affinity towards female leaders. Insecurity will, therefore, be more deleterious to women's support for female leadership than men's in conflicts where women's political inclusion is violently contested.

We assess these expectations with experimental and observational survey data from the former Islamic Republic of Afghanistan (IRoA), which fell to Taliban insurgents in August 2021. In spite of persistent gender inequalities in Afghan society, Afghan women experienced significant gains in political representation during the Republic's twenty-year rule. The IRoA came to power in 2001 after the United States and its allies invaded Afghanistan and displaced the Taliban. Less than two decades after the Taliban barred Afghan women from the political arena (Rashid Reference Rashid2010, Appendix 1), over six thousand Afghan women were judges, defence attorneys, and police officers, and eighteen were ministers or deputy ministers in the last years of the IRoA (Ahmadi Reference Ahmadi2019). Propped by international support and gender quotas (Bush Reference Bush2011), Afghanistan's parliament in 2019 had a higher percentage of female representatives than the US Congress (Barr Reference Barr2020).

Afghan women's gains in political representation coincided with mounting insecurity. Insecurity came in many forms: clashes between Taliban insurgents and NATO-supported government forces, thuggery from pro-IRoA militias, and extortion and kidnappings by criminal groups (Gopal Reference Gopal2021). The IRoA, marked by both a deepening in female political representation and insecurity, offers a unique lens to explore how insecurity shapes public attitudes toward female political leadership in states where women's political inclusion is violently contested.

Estimating the impact of insecurity on support for female leadership is difficult. There are observable and unobservable differences between civilians in conflict states who experience insecurity and those who do not. These differences likely influence attitudes toward female leadership.

Measuring support for female leadership is challenging. Social desirability biases mask true feelings towards female leadership (Streb et al. Reference Streb2008). That women's rights were an ideological divide between the IRoA and the Taliban undoubtedly amplifies these biases. The direction of these biases is unclear. Respondents could feel pressured to satisfy enumerators hired by Western donors and exaggerate their support for female leadership. Conversely, they may fear that disclosing strong support for female leadership could possibly out them to Taliban sympathizers. Feelings of insecurity also exacerbate social desirability biases (Singh and Tir Reference Singh and Tir2021).

We confront these challenges with an original survey experiment conducted in three Afghan provinces between 2016 and 2017. Our survey experiment first primes some respondents with information about Afghanistan's worsening security situation. We then determine preferences for female political leadership by asking all respondents to complete a conjoint exercise where they chose and ranked hypothetical leadership profiles with varying characteristics, including gender. The experiment's prime helps causally identify whether information about insecurity influences preferences for female political leadership. The ensuing conjoint exercise reveals respondents' preferences towards female leadership while attenuating social desirability biases (Horiuchi, Markovich, and Yamamoto Reference Horiuchi, Markovich and Yamamoto2022).

We expected the insecurity prime to weaken all respondents' support for female leadership. Instead, the insecurity prime only dampened women's preferences for female leaders. Absent of the prime, female respondents were indifferent towards the gender of a hypothetical political leader. However, when exposed to information about Afghanistan's deteriorating security, female respondents mimicked male respondents' consistent preferences for male over female leaders. They also became more supportive of local governing institutions. This greater support may reflect women's beliefs that these more proximate, typically conservative, male-dominated institutions can provide greater security through force or reconciliation with the Taliban.

Observational data, however, challenges our experimental findings. Using the Asia Foundation's nationally representative Survey of the Afghan People, we find that feelings of insecurity and exposure to violence correlate with more polarized attitudes towards women's political representation among men and women. The null results from our survey experiment may reflect a balancing out of these opposing reactions to the insecurity prime among treated respondents. That women's leadership preferences are also more polarized, amid feelings of and exposure to insecurity, contrasts our experimental finding of insecurity depressing women's support for female leadership. Furthermore, unlike in the rest of Afghanistan, insecurity positively correlates with support for female leadership among men and women in the provinces where we ran our survey experiment. These divergences across gender, region, and research method suggest that the impact of insecurity on preferences for female leadership in conflict states is highly heterogeneous.

Our analysis advances scholarship on insecurity and support for female leadership in three ways. First, we depart from existing scholarship by presenting mixed and varied evidence of insecurity undermining public support for female leadership. The survey experiment illustrates that women's support for female leadership is more vulnerable to insecurity than men's. Our observational analysis – though unable to generate causal claims – reveals that insecurity correlates with greater polarization among men and women's attitudes toward female leadership in some places but not others. We believe these diverging findings stem from the particular context of our study.

This points to our second contribution. We broaden the scholarship of insecurity and support for female leadership by investigating these dynamics in an ongoing conflict where women's political inclusion is an ideological cleavage between warring factions. Afghanistan is not unique in this regard. We expect our findings to apply mostly to conflict states such as Chad, Iraq, Libya, Mali, Niger, Nigeria, Syria, and Yemen.

The third contribution is methodological. By pairing a survey prime with a conjoint exercise, our survey experiment mitigates social desirability biases to offer a clear measure of the impact of information about insecurity on support for female leadership. Our analysis of observational data, however, underscores a limitation of conjoint analysis. In calculating averages, conjoint estimates mask diverging within group preferences and intensity of preferences (Abramson, Koçak, and Magazinnik Reference Abramson, Koçak and Magazinnik2022).

Insecurity and Support for Female Political Leadership

Most scholarship finds that feelings of insecurity undermine public support for female political leadership. Survey respondents generally see male politicians as better suited to manage defence and security issues than female politicians (Dolan Reference Dolan2014a; Dolan Reference Dolan2014b; Schwartz and Blair Reference Schwartz and Blair2020). In the wake of 9/11, US citizens viewed men as more competent in national security issues than women (Lawless Reference Lawless2004). Holman, Merolla, and Zechmeister (Reference Holman, Merolla and Zechmeister2011) apply observational and experimental survey data to demonstrate that terrorist threats depress respondents' evaluations of female candidates, while male candidates with similar profiles faced no such penalties. In the UK, Holman, Merolla, and Zechmeister (Reference Holman, Merolla and Zechmeister2022) reveal that the 2017 Manchester Arena terrorist attack did not boost Prime Minister Teresa May's popularity: rally-around-the-flag effects primarily benefit men. Context nuances these findings. Belonging to a hawkish party, however, shelters female candidates from this insecurity penalty (Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2011; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Ono and Burden Reference Ono and Burden2019). Nevertheless, public opinion generally favours male over female leaders in security affairs.

There is also evidence that feelings of insecurity bolster support for more conservative ideologies. Psychologists have found that terrorist attacks in the US and Europe (Echebarria-Echabe and Fernández-Guede Reference Echebarria-Echabe and Fernández-Guede2006; Schüller Reference Schüller2015) and feelings of threat in a lab experiment (Nail et al. Reference Nail2009) increase preferences for conservative policies. This shift towards conservatism was greatest among liberals (Landau et al. Reference Landau2004; Van de Vyver et al. Reference Van de Vyver2016). Feelings of insecurity in settings where conservatism favours male political representation could generate a greater shift in support for male political leadership among constituents who are most open to female leadership.

These patterns extend outside North America and Western Europe (Carlin, Carreras, and Love Reference Carlin, Carreras and Love2020). Kim and Kang (Reference Kim and Kang2022) use survey data from over eighty countries to show that respondents in states with higher external security threats are more likely to believe that men make better political leaders than women. In Bosnia, Hadzic, and Tavits (Reference Hadzic and Tavits2021) uncover that while women are more likely to run in districts that experienced higher levels of violence during the civil war, they are less likely to get elected. They also find that priming survey respondents to think about the violence their ethnic group experienced during the civil war decreases female respondents' but increases male respondents' willingness to engage in politics (Hadzic and Tavits Reference Hadzic and Tavits2019).

Much of this work blames gender-trait stereotypes, defined as ‘pervasive, durable, shared beliefs held about groups on the basis of certain (often ascriptive) characteristics’ (Schwartz and Blair Reference Schwartz and Blair2020, 877), for these outcomes. Stereotypes make people believe that men are more qualified than women in handling security and defence affairs (Dolan Reference Dolan2014a; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Huddy and Terkildsen Reference Huddy and Terkildsen1993; Lawless Reference Lawless2004). These beliefs weaken public support for female leadership in times of insecurity. However, most of this evidence comes from relatively secure states where external actors are the cause of insecurity or post-conflict states, particularly Bosnia (Butler, Tavits, and Hadzic Reference Butler, Tavits and Hadzic2023; Hadzic and Tavits Reference Hadzic and Tavits2019; Hadzic and Tavits Reference Hadzic and Tavits2021), where the war ended more than two decades ago.

Whether insecurity reduces support for female leadership in conflict states is less clear. Insecurity in conflict states is more than a memory, a once-in-a-lifetime terrorist attack, or a nightmarish hypothetical. It touches all facets of governance, not just foreign affairs or national security. Citizens in states with ongoing conflicts may value different leadership characteristics than citizens in relatively secure states, even those with a recent history of conflict. Insecurity in secure states may prompt public demands for leaders who can deliver order and justice. Insecurity in conflict states, however, may elicit public demands for leaders who can deliver peace and reconciliation. The characteristics desired for a leader to restore security in a post-conflict state might diverge from those desired for a leader to obtain security in a conflict state.

These characteristics could reflect prevailing stereotypes about women. Often referred to as benevolent sexism, these stereotypes cast women as more compassionate, moderate, and peace-seeking than men (Huddy and Terkildsen Reference Huddy and Terkildsen1993; Shapiro and Mahajan Reference Shapiro and Mahajan1986). Karim (Reference Karim2019, 800) finds that increasing female representation in post-conflict states' security institutions improves public confidence in the security sector. This is because prevailing stereotypes of women being less abusive and more inclusive improve public perceptions of restraint and inclusiveness in security institutions with female representation. Women are often political outsiders, too (Tripp Reference Tripp2015, Chapter 8). Citizens may believe that female leaders' outsider status makes them uniquely suited to heal violent divisions (Anderlini Reference Anderlini2007, 127). Shair-Rosenfield and Wood (Reference Shair-Rosenfield and Wood2017) claim that female representatives in post-conflict states boost public perceptions of good governance and elite credibility.

Does insecurity in conflict states heighten citizens' demands for female political leadership because of peace-seeking stereotypes about women? Answering this question requires determining what citizens in conflict states want from their leaders. Citizens preferring security through force – hawks – may want male political leaders because of stereotypes privileging men's security prowess. Citizens wanting security through dialogue and reconciliation – doves – could prefer female political leaders because of stereotypes privileging women's peacemaking.

In conflicts where women's political inclusion is violently contested, however, we hypothesize that both the desire for security through force and security through reconciliation decreases support for female political leadership. Hawks in these conflicts likely subscribe to the seemingly global stereotype of men's greater military abilities (Kim and Kang Reference Kim and Kang2022). While some hawks may not believe these stereotypes, men generally monopolize ‘law and order’ security institutions like the military and the police in conflict states (Karim Reference Karim2019). Women's exclusion from these male-dominated institutions may signal a weaker capacity to impose peace through force due to a lack of professional experience.

Congruently, doves in these conflicts may worry that female political leadership would provoke violence from militant groups opposed to women's political inclusion. Therefore, they may accept forfeiting female political representation if it reduces violence. Insecurity thus pushes both hawks and doves to favour male over female political leadership in conflicts with militant groups opposed to women's political inclusion. This generates our first hypothesis:

Hypothesis 1 (H1): Insecurity decreases support for female political leadership in conflict states where women's political inclusion is violently contested.

Insecurity may have different effects on men's and women's preferences for female political leadership. A significant body of research finds that women are generally more supportive of female political leadership than men (Kao and Benstead Reference Kao and Benstead2021; Schwarz and Coppock Reference Schwarz and Coppock2022). At the same time, women have higher preferences for peace and are more open to compromise than men (Maoz Reference Maoz2009; Wood and Ramirez Reference Wood and Ramirez2018). Feelings of insecurity can cause women to prioritize security over female representation in conflicts where women's inclusion is violently contested – women are more fearful of the consequences of insecurity than men. This trade-off is less apparent for men, who care relatively less about security and have lower baseline preferences for female leadership. This produces our second hypothesis:Footnote 2

Hypothesis 2 (H2): Insecurity has a more negative effect on women's support for female political leadership than men in conflict states where women's political inclusion is violently contested.

The Afghan Case

The IRoA made substantial, albeit uneven, progress in advancing women's political representation during its two decades of rule. At the national level, the IRoA introduced quotas for female political participation, resulting in women eventually holding approximately twenty-seven per cent of seats in parliament (Bahesh Reference Bahesh2021). At the local level, some rural development programmes mandated gender equality in programme governance (Beath, Christia, and Enikolopov Reference Beath, Christia and Enikolopov2013). Women also made gains in the public sector. By 2019, women represented twenty-one per cent of the Afghan civil service, including sixteen per cent of senior management positions (Haque Reference Haque2020).

Women's growing political representation coincided with mounting security challenges. After relative peace in the early 2000s, a growing Taliban insurgency derailed security and toppled the IRoA in 2021. The Watson Institute of International and Public Affairs estimates that nearly 50,000 Afghan civilians were killed due to conflict over the past two decades (Watson Institute for International and Public Affairs 2021). More died from indirect causes.

How did insecurity shape public attitudes towards female leadership in the IRoA? Some argue that insecurity amplified Afghans' reliance on existing tribal and village structures (Weigand Reference Weigand2022). These structures, typically situated in areas far from the reach of formal state institutions, often reinforced traditional gender hierarchies (Chishti Reference Chishti2020). Others observe that Afghanistan's deteriorating security and the IRoA's foreign allies’ embrace of women's political inclusion tainted gender rights as a foreign import (Bahri Reference Bahri2014; Kandiyoti Reference Kandiyoti2007a). This pushed Afghans to support more conservative policies on gender issues. By contrast, Beath, Christia, and Enikolopov (Reference Beath, Christia and Enikolopov2013) find little relation between rural villages' exposure to violence and their inhabitants' attitudes towards women's political inclusion.

Survey Experiment on Insecurity and Preferences for Female Leadership

We examined our hypotheses with an original survey of over two thousand Afghan respondents. We surveyed 2,485 households between 2016 and 2017 in three northern provinces: Balkh, Kunduz, and Sar-e-Pul. All three provinces are ethnically and socio-economically diverse. While the northern region of Afghanistan has historically been a locus of opposition to the Taliban, Taliban insurgents were present in all three provinces when the survey was administered. This survey was part of a broader project on Afghans' attitudes towards leadership in relation to insecurity, ethnicity, and corruption.Footnote 3

We investigate how insecurity influences preferences for female leadership by pairing a priming experiment with a conjoint exercise. The survey randomly primed over 600 respondents with information about worsening security in the Afghan conflict. Another 600 respondents randomly received a neutral ‘control’ prime. Finally, approximately 1,200 respondents were randomly primed with information about different types of corruption. Following the priming experiment, the survey asked all respondents to complete a conjoint exercise where they had to rank and choose between profiles of hypothetical leaders.Footnote 4

We compare the insecurity-primed respondents' preferences towards female leadership with respondents who randomly received a neutral ‘control’ prime. Because we are primarily interested in how insecurity shapes attitudes towards female leadership, our main analysis excludes respondents who received primes about corruption. We nevertheless include respondents who received the corruption primes in numerous robustness checks.

Insecurity Prime and Conjoint Exercise

Enumerators primed treated respondents with information about insecurity by reading the following text, which was based on news reports at the time:

The past year is believed to have been the most difficult year for Afghanistan in terms of insecurity since 2001. Last year, the level of civilian casualties rose to unprecedented levels. Officials recently implored ‘those inflicting this pain on the people of Afghanistan to take concrete actions to protect civilians and to put a stop to the killing and maiming of civilians’.

There is no pure control in this survey. We opted for a ‘neutral’ informational control to make the treatment and control conditions as similar as possible. Enumerators read the following ‘neutral’ text to respondents in the control group:

Afghanistan has a population of about 33 million people, making it the 41st largest country in the world. There are 34 provinces and nearly 400 districts in Afghanistan. Afghanistan shares a border with six different countries in Central and South Asia, including a 76km border with China. There are four major rivers in Afghanistan: the Amu Darya, the Hari River, the Kabul River, and the Helmand River.

Enumerators then gave treated respondents a manipulation check to verify whether they understood the vignette's information. Eighty-seven per cent of respondents answered the insecurity prime's manipulation check correctly.Footnote 5 Table 1 in Section 2.1 of the Supplementary Information (SI) depicts few demographic differences between respondents in the control and treatment groups regarding education, employment status, and ethnicity, except for treated respondents who have stronger preferences for peace and security. Note that we asked respondents' security preferences after the insecurity or neutral prime. This confirms that the insecurity prime was successfully randomized and made treated respondents think more about insecurity when assessing leadership profiles than respondents in the control group.

The prime experiment is a hard test for our treatment effect. All survey respondents were likely to think of insecurity when evaluating hypothetical leaders' profiles. Indeed, when asked to rank from 1 to 6 the importance of leaders' ability to bring peace and security, the respondents averaged 5.72. Respondents were also likely to have already forged their attitudes about the suitability of women in leadership positions. Afghanistan's gender quota system in parliament and international efforts to include Afghan women in government prompted public controversy and debate over the role of women in politics (Broadbent Reference Broadbent2010).

Support for female political leadership is the analysis' dependent variable. We measure and compare respondents' preferences for female leadership with a conjoint experiment of leadership profiles. Conjoint experiments do not ask respondents to state their preferences directly; they require respondents to choose between profiles with numerous attributes, including non-sensitive ones. This helps mitigate social desirability biases (Horiuchi, Markovich, and Yamamoto Reference Horiuchi, Markovich and Yamamoto2022).

After receiving the treatment or control prime, respondents were read the following text:

Now, I am going to show you a few pairs of profiles of potential leaders and ask you to choose between the two, the one that you think would be the best advocate for you. Given a choice between these two profiles, which person would you prefer as a leader?

We define political leadership broadly in terms of a leader's ability to advocate for their constituents. We prefer this broader definition over referencing precise leadership positions like mayor, parliamentarian, or president to prevent respondents from thinking of actual candidates for these positions when choosing between leadership profiles in the conjoint experiment. Enumerators then read the leaders' profiles with the following attributes and values (Table 1):

Table 1. Leadership profile attributes

The survey randomized leadership attributes' values and the order in which they were read. There were three constraints in the randomization of leadership attribute values. No hypothetical female leader had a military professional background. This is because there were so few women in the Afghan National Army (ANA) (Jones Reference Jones2018). Leaders born in Kandahar – a Pashtun-dominant province – were Pashtun.Footnote 6 Hypothetical leaders younger than thirty would have been formally educated to at least a secondary school level.Footnote 7 Though party affiliation has an important effect on public perceptions of women's competency in security affairs (Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Ono and Burden Reference Ono and Burden2019), we did not include a political party attribute; many Afghan candidates ran for positions independently, and most major parties are associated with particular individuals and/or ethnic factions.

Finally, we assess whether the insecurity prime had diverging effects on hawks and doves. We identify respondents with more ‘dovish’ conciliatory views towards the Taliban based on their response to the question, ‘To what extent do you agree that international forces should remain present in Afghanistan for the foreseeable future?’, with 1 indicating ‘Strongly Agree’ and 5 indicating ‘Strongly Disagree’. The median response was 4, ‘Disagree’, revealing significant variation in responsesFootnote 8 and mitigating concerns of social desirability bias. We categorize all respondents who disagreed with the statement as dovesFootnote 9 and those who agreed as hawks.Footnote 10 We asked this question after the respondents received the control or insecurity prime. These responses may suffer from a post-treatment bias, though there is no statistically significant mean difference between the treatment and control group regarding their support for international forces remaining in Afghanistan (SI Table 3).

We estimate the effect of the insecurity prime on support for female political leaders. We measure leadership preferences through a forced choice conjoint experimental design (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2014), where respondents had to choose between a pair of leadership profiles with randomized attributes three times. Respondents also had to rank each profile from 1 to 5. This produced 7,800 profile observations.

We investigate whether respondents' preferences for female leaders vary across treatment groups and genders. Most conjoint analysis examines attributes' Average Marginal Component Effect (AMCE). However, AMCEs estimated across subgroups are sensitive to reference or baseline category specification (Leeper, Hobolt, and Tilley Reference Leeper, Hobolt and Tilley2020). We follow Leeper, Hobolt, and Tilley (Reference Leeper, Hobolt and Tilley2020) and use the cregg package to calculate and plot conditional marginal means (MM), and conduct omnibus F-tests to determine whether respondents' preferences for female leaders differ across subgroups.Footnote 11 We cluster standard errors at the respondent level to account for the repeated number of observations per respondent.

In a forced choice design, an attribute with an MM value of 1 indicates that respondents would choose a profile with that attribute, with a probability of 1. MMs in a forced choice design average 0.5 by definition, with values greater (lower) than 0.5 representing a positive (negative) bias towards an attribute (p.210).

We hypothesize:

H1: Respondents who receive the insecurity prime (treatment group) will have lower preferences for female political leadership than respondents who receive the neutral text (control group).

H2: The negative effect of the insecurity prime on respondents' preferences for female political leadership will be greater among female respondents than male respondents.

Results

Insecurity and Preferences for Female Leadership Across Treatment Groups

There is no statistically significant difference in preferences for female political leadership between respondents in the insecurity (treatment) and neutral text (control) groups. An F-test comparing a model of respondents' likelihood to choose a profile with a treatment variable interacting with the female attribute and a reduced model without the treatment variable interaction term is not statistically significant (p = 0.705). Nor is the F-test statistically significant for respondents' ratings of profiles (p = 0.28) (SI Section 2.4). This null finding is not from a lack of power (power ratio: 0.872, SI Section 2.13).

Figure 1 illustrates a persistent gender gap in preferences for male over female leaders in the treatment and control groups. It plots the MMs of the gender attribute for the choice outcome for both groups. Each dot and error bar represents the MM and its 95% confidence interval for the gender attribute. Confidence intervals crossing an MM value of 0.5 (the vertical line) indicate that the MM estimate of a gender bias is not statistically significant at the five per cent level. MM estimates and their confidence intervals for profiles with male leaders are greater than 0.5 in both groups.

Figure 1. Insecurity and preferences for male and female leadership across control and treatment groups (H1): estimated marginal means (MM) and 95% confidence intervals.

Profiles with a male leader had a roughly fifty-three per cent probability of being chosen (plus or minus one percentage point) in the treatment and control groups (SI Section 2.4). The insecurity prime had no statistically significant effect on either hawks or doves’ female leadership preferences (F-test p: 0.99; SI Section 2.5).

Insecurity and Preferences for Female Leadership Across Gender and Treatment Groups

In support of H2, the insecurity prime had a much greater negative effect on female respondents' preferences for female political leadership than male respondents. The insecurity prime had no effect on men's preferences for female leadership (F-test p: 0.485; Figure 2). Men in the treatment and control groups were consistently more likely to choose male over female leaders.Footnote 12

Figure 2. Insecurity and preferences for male and female leadership among male respondents (H2): estimated marginal means (MM) and 95% confidence intervals.

Though this difference is statistically insignificant, Figure 2 demonstrates that men who received the insecurity prime had a slightly higher likelihood of choosing a female profile than men in the control group. The insecurity prime may have ameliorated some men's attitudes towards female leadership more than others.

Figure 3 in the SI examines the insecurity prime's effect among male hawks and doves' choice for female leadership. Though differences in these MM estimates are not statistically significant (F-test p: 0.73), primed respondents with hawkish preferences are the only male subgroup who exhibit no bias in female leadership preferences. Unlike doves, they may not care if female leadership provokes a violent backlash from patriarchal militant groups. Hawkish men's greater support for female leadership may explain the positive though statistically insignificant change in primed men's preferences for female leadership.

Figure 3. Insecurity and preferences for male and female leadership among female respondents (H2): estimated marginal means (MM) and 95% confidence intervals.

The insecurity prime had a much greater impact on female respondents. Women who were read the neutral text were just as likely to choose a profile with a male or female leader (Figure 3). However, women who received the insecurity prime were more likely to choose a male leader (52.7 per cent mean probability, plus or minus 2.6 percentage points; Table 9 in SI Section 2.6). These differences in gender preferences among women in the treatment and control groups are statistically significant at the five per cent level (F-test, p < 0.05). The insecurity prime forged a gender gap in female respondents' female leadership preferences (Figure 3).

This finding is robust to respondents' ranking of potential leaders (F-test,p < 0.05). These patterns hold when adding respondents who received primes about corruption and the neutral text as the control group (SI Table 10). However, subgroup differences among women for the choice variable are no longer statistically significant at the ten per cent level (p = 0.16). Female respondents mirror men's preferences for male leaders when they receive the insecurity prime. Women's support for female leadership is more vulnerable to information about insecurity than men's.

Unlike with men, the insecurity prime pushed female respondents with hawkish views to become more supportive of male leadership. Though differences in MM estimates across groups are not statistically significant (p = 0.14), female hawks in the treatment group were the only subgroup to prefer male over female leadership. When primed to think about insecurity, Afghan women with more belligerent attitudes towards the Taliban may be more likely to believe that men can better provide security. More dovish women did not favour male leadership profiles when primed. They perhaps doubted that male leaders would seek peace with the Taliban. The insecurity prime's diverging effect on hawkish men (SI Figure 3) and women's (SI Figure 4) support for female leadership proposes that policy preferences and gender may interact to produce diverging responses to insecurity.

Some may wonder whether the insecurity prime had less of an effect on men's preferences because men are more likely to be already informed about Afghanistan's deteriorating security. Men in our sample were, on average, more educated and more likely to be employed than women (SI Table 2). These socio-economic differences may expose men to more information about national politics, dulling the informational effect of the insecurity prime. A higher percentage of male respondents also resided in Kunduz, a province that the Taliban briefly occupied during our survey collection.

Our results for H2 are robust to excluding respondents who have a university education (SI Section 2.9), who are from Kunduz (SI Section 2.10), and who are unemployed (SI Section 2.11). Greater access to information about insecurity and possible exposure to insecurity cannot explain why the insecurity prime had no effect on men's leadership preferences but weakened women's support for female leadership.

Conversely, perhaps men were less attentive to the insecurity prime than women. A higher percentage of treated male respondents failed the manipulation check than treated women (sixteen versus ten per cent) – and this difference is statistically significant (p < 0.05) (SI Table 6). Our enumerators corrected respondents when they failed the manipulation check. And, though men were more likely to fail the manipulation check, treated male respondents did value leaders who provide peace and security more than men in the control group (SI Table 5). The prime did impact male respondents' valuations of leaders' ability to provide peace and security.

Others may suspect that it is not insecurity but simply priming respondents with information about poor governance that drives these results. That our findings persist when expanding the control group to include respondents who received primes about corruption mitigates this concern (SI Section 2.8). We further test whether information about poor governance undermines women's support for female leadership by re-running our analysis and categorizing respondents who received the corruption prime as the treatment group and respondents who received the neutral text as the control group (SI Section 2.12). Unlike the insecurity prime, the corruption prime has no statistically significant effect on women's preferences for female leaders or men's leadership preferences (SI Figure 5).

Finally, some may worry that our findings are underpowered, given the treatment's small effect size and the numerous subgroups under analysis. Using Schuessler and Freitag's (Reference Schuessler and Freitag2020) cjpowr R package, we estimate that our analysis of the insecurity prime's effect on men's leadership preferences (H2) has a power ratio of 0.677. However, expanding the control group to include respondents who received the corruption primes augments the power ratio past the conventional level of 0.8. That the insecurity prime's effect remained statistically insignificant with the broader control group lessens our concern that the null finding among male respondents for H2 stems from low statistical power (SI Section 2.13).

Mechanisms

We hypothesize that desires for security in conflicts where women's inclusion is violently contested push both hawks and doves to favour male over female leaders. Hawks prefer men due to stereotypes privileging men's military expertise. Doves also prefer men so as not to provoke a backlash from militants opposed to female political leadership.

These mechanisms are fully compatible with two others in the literature. The first, drawn from psychology, argues that feelings of insecurity trigger stronger conservative beliefs (Echebarria-Echabe and Fernández-Guede Reference Echebarria-Echabe and Fernández-Guede2006; Landau et al. Reference Landau2004; Nail et al. Reference Nail2009), which would favour male leaders in the Afghan context. The second, stemming from scholarship on Afghanistan (Weigand Reference Weigand2022), maintains that insecurity amplifies Afghans' trust in local governing institutions, which tend to reinforce traditional gender hierarchies relative to national institutions. Neither of these mechanisms predicts whether insecurity privileges hawkish or dovish leadership. All of these mechanisms, however, expect insecurity to benefit male over female leaders in the court of public opinion.

To adjudicate between these mechanisms, we examined respondents' answers to a series of questions posed after the priming experiment about the attributes they value most in a leader.Footnote 13 These include questions about a leader's piety, military experience, and willingness to punish criminals. We also probed respondents' support for international military forces in Afghanistan, as well as their confidence in their provincial and central government. We then subsetted our analysis of these answers across genders to ascertain why female respondents' gender leadership preferences were more sensitive to the insecurity prime than men's. Respondents' answers to these questions are prone to social desirability biases. Nevertheless, these biases should not disproportionately affect the treatment or control group because the insecurity prime was randomized.

Leadership Attributes

After the prime, enumerators asked respondents to rate, on a scale of 1 to 6, the importance of a series of leadership attributes; 1 indicated the ‘least important’ and 6 the ‘the most important’ attribute a leader could have. To identify demands for leaders who can bring security through force, we examine respondents' preferences for leaders with combat experience, whether fighting for the Afghan National Army (ANA) or as a Mujahideen.Footnote 14 We also examine preferences for leaders who always ‘punish those who break the law’. This attribute's emphasis on punishment invokes the use of force, which is frequently associated with the Taliban's approach to justice (Swenson Reference Swenson2017).

To gauge respondents' demands for leaders who can bring security through reconciliation with the Taliban, we analyze respondents' support for leaders with strong religious values. Though an imperfect proxy, respondents with greater preferences for leaders with strong religious values are plausibly more supportive of reconciliation or compromise with the Taliban. Furthermore, if primed respondents are more supportive of religious leaders, this would align with work from Europe and the US that demonstrate feelings of insecurity deepen support for conservative ideologies (Echebarria-Echabe and Fernández-Guede Reference Echebarria-Echabe and Fernández-Guede2006; Nail et al. Reference Nail2009; Schüller Reference Schüller2015).

Governing institutions

We pair our analysis of respondents' preferred leadership attributes with their attitudes towards governing institutions. Enumerators asked respondents to list on a scale from 1 to 5, with 1 indicating ‘no confidence at all’ and 5 ‘a lot of confidence’, their confidence in the National Unity Government (NUG) and their respective provincial governments. Provincial governments were generally much more enmeshed in local tribal networks and militant groups than the NUG, which was seated in Kabul (Mukhopadhyay Reference Mukhopadhyay2009b). Evidence of the insecurity prime propping greater support for provincial government would corroborate existing accounts of Afghans turning to more local, tribal, and typically male-dominated governing structures amid mounting insecurity.

Finally, we asked respondents to rate on a scale of 1 to 5 the extent to which they agree that international forces should remain in Afghanistan for the foreseeable future, with 1 representing ‘Strongly Disagree’ and 5 ‘Strongly Agree’. We interpret greater support for international forces to signify less support for reconciliation with the Taliban.Footnote 15

Table 3 in SI Section 2.2 is a balance table of treated and control group respondents' preferences towards leadership attributes and governing institutions. The first row reiterates that primed respondents value leaders who can provide peace and security more than the respondents in the control group (p < 0.05), though average support is high in both groups. We find no statistically significant mean differences in treated and control group respondents' attitudes towards leadership attributes or governing institutions, except for trust in provincial governments. On average, primed respondents exhibit higher confidence in their provincial government than the control group (p < 0.05). This could attest to respondents' belief that their local government can obtain better security, whether through peace or force. Due to the heightened influence of traditional institutions at subnational levels, this local bias may skew support for male over female leaders. However, greater support for local government did not translate to stronger preferences for male leaders in the pooled analysis.

Tables 4 and 5 in SI Section 2.2 help explain why. Table 4 presents mean differences in female respondents’ leadership and institutional preferences in treated and control groups. Unlike the pooled sample, the insecurity prime does not increase women's preferences for leaders who can bring peace and security. On a scale from 1 to 6, both groups' average support is roughly 5.86 (p = 0.721), greater than the pooled sample-treated group's mean support of 5.775. However, because women's responses to this question are already close to the maximum, ceiling effects may limit the insecurity prime's ability to increase female respondents' preferences for greater peace and security. Both women in the prime and control groups strongly desire leaders who can bring peace and security. There is also no evidence of primed women valuing hawkish leadership attributes more than women in the control group. Nor are primed women more supportive of religious leaders.

The only statistically significant difference, which is at the ten per cent level, is treated women's stronger confidence in provincial government (3.492 v. 3.321, out of 5; p = 0.093). We interpret primed women's greater support for the provincial government to reflect a stronger belief that local governing institutions can better provide peace and security, whether through conflict or accommodation with the Taliban. The insecurity prime did not substantially change women's already high valuation of leaders who can bring peace and security, but it did increase their support for an institution which can provide peace and security: local government.

We suspect that greater support for provincial government underpins primed women's stronger preferences for male leaders. Indeed, among non-treated female respondents – women in the control group and those who received the corruption primes – women who report having high confidence in their provincial government are also more likely to favour male over female leaders in the survey's conjoint exercise (F-test p < 0.01). Women who express low confidence in their provincial government are more likely to choose a female profile over a male one (Figure 1 in SI Section 2.2).

Our analysis does not disclose whether insecurity causes female respondents to prefer leaders who bring security through force (hawks) or reconciliation (doves). However, we find that female respondents value peace and security more than men (SI Section 2.2.1). Female respondents also become more supportive of local governing institutions when primed to think about insecurity. We infer that women may believe local governing institutions are better equipped to provide security than the national government. Because these institutions tend to be more male-dominated, this may correspond with greater support for male leadership among women who received the insecurity prime.Footnote 16

Treated men, by contrast, do not display greater support for the provincial government than men in the control group (SI Table 5; p = 0.34). Unlike women, we find no statistically significant difference in support for female leadership in the conjoint exercise between untreated men who express high and low levels of confidence in their provincial government (p = 0.25).

Though men have lower mean preferences for leaders who can provide peace and security than women, the insecurity prime did increase men's support for this leadership attribute (p < 0.05) – the only statistically significant difference in the level of support between men in the treatment and control groups. Insecurity-primed male respondents support leaders who provide peace and security more than the men in the control group; this increase does not change men's already low support for female leadership.

In sum, when weighing mechanisms from existing literature that might help explain our experimental findings, we find more convincing evidence that treated women's higher propensity to select male leaders is driven by a heightened affinity for more proximate governance institutions in the face of insecurity rather than an attraction to more conservative values. Indeed, it is notable that women in the treated group were less likely to desire a leader with strong religious values, although the difference in support between the two groups is not statistically significant.

Stepping back, some may wonder whether our control group does not represent a real control because all respondents in a conflict state care about peace and security when evaluating potential political leaders. The six percentage point gender gap in preferences for male over female leaders in the treatment and control group may derive from Afghans' prevailing security concerns.

Though important, the magnitude of this gender gap is not extraordinary (Schwarz and Coppock Reference Schwarz and Coppock2022, 662). It is comparable to gender gaps in leadership preferences found in conjoint experiments in Jordan (Kao and Benstead Reference Kao and Benstead2021) and Tunisia (Blackman and Jackson Reference Blackman and Jackson2021). Similar gaps in non-conflict states suggest that insecurity is neither the sole nor the main driver of respondents' preferences for male over female political leaders.

Finally, some may suspect that the null finding from our pooled sample (H1) derives from the prime being too weak. Our respondents strongly valued peace and security. And, while treated respondents did value a leader's ability to bring peace and security more than respondents in the control group, that difference was slight: a 0.1 average difference on a scale from 1 to 6 (SI Table 1). Perhaps our sample's high valuation of peace and security mitigates the insecurity prime's effect on the treatment group.

A stronger insecurity prime might have impacted men's leadership preferences, but it is hard to imagine what a stronger but still ethical insecurity prime would entail. Furthermore, concerns over a too-weak prime cannot explain why the prime was strong enough to lower female respondents' female leadership preferences. Observational data, however, points to an alternative explanation: our null findings could reflect an averaging of insecurity's diverging effects on preferences for female leadership. We discuss this evidence below.

Evidence from the Asia Foundation Survey

The Asia Foundation's Survey of the Afghan People is the longest-running barometer of Afghan public opinion. Beginning in 2004, it surveyed nearly 130,000 Afghans in all thirty-four provinces across more than fifteen waves (Burbidge et al. Reference Burbidge2016, 257). The survey solicits Afghans' past exposure to violence and insecurity and their attitudes towards women in politics.

As in the survey experiment, observational data reveals inconclusive evidence of feelings of insecurity decreasing support for female leadership (H1). Unlike the survey experiment, however, the survey exposes polarized preferences for female leadership in relation to past exposure to violence and feelings of insecurity. Insecure respondents are either more supportive or less supportive of female leadership than secure respondents. Our null finding in the conjoint experiment's pooled analysis (H1) may reflect a balancing out of the treatment group's pro-female and pro-male leadership preferences. These patterns persist when subsetting the survey across genders. Insecurity correlates with more polarized gender leadership preferences among male and female respondents. This contradicts our second experimental finding – insecurity lowers women's support for female leadership (H2). Finally, there is a subnational variation in the relationship between insecurity and preferences for female leadership. Insecurity is positively associated with men and women's support for female leadership in the three provinces where we conducted the survey experiment. It correlates with polarized preferences in the rest of the country.

These results should be interpreted with caution. Afghans who report experiencing violence differ from those who have not across a host of observable and unobservable characteristics (SI Section 3.1). These findings do not take into account social desirability bias. The survey asked respondents directly about their attitudes towards female political leadership. Gendered norms may make women more likely to admit feeling or experiencing insecurity than men.

Nevertheless, when paired with the conjoint experiment, these findings point to possible limitations of conjoint analysis. By averaging the intensity and direction of respondents' preferences (Abramson, Koçak, and Magazinnik Reference Abramson, Koçak and Magazinnik2022), conjoint estimates may obscure diverging preferences within groups. The disjuncture between our experimental and observational results from the same three provinces flags some of the challenges of reconciling experimental and observational analyses. We suspect differences in the wording of survey questions, social desirability biases, and historical context explain some of the disconnects between our observational and experimental analysis, as well as between the three provinces surveyed in our experiment and the rest of Afghanistan. Combined, our experimental and observational analyses convey that the impact of insecurity on female leadership preferences in conflict states may vary both across and within genders and provinces.

Measuring Insecurity and Support for Female Leadership

The survey probes respondents' current feelings of, and past exposure to, insecurity. This differs from respondents' awareness of insecurity in Afghanistan, a closer proxy to the survey experiment's insecurity prime. We suspect that both the informational and experiential components of insecurity can impact preferences for female leadership.

We used two questions from the survey to measure respondents' feelings of insecurity. The first asks, ‘Have you or anyone in your family been a victim of violence or of some criminal act in your home or community in the past year?’Footnote 17 Almost nineteen per cent of respondents (n = 24,201) said ‘yes’. Over eighty per cent of respondents (n = 104,594) said ‘no’, and the remainder (n = 1,039) said ‘don't know’ or refused to answer. We first measure insecurity (Insecurity) as a dummy variable equal to 1 if a respondent confirmed that they or someone in their family had been a victim of violence in the past year.

Our second measure of insecurity focuses on respondents' reported levels of fear. The survey asks, ‘How often do you fear for your own personal safety or security or for that of your family these days? Often, sometimes, rarely, or never?’Footnote 18 Almost sixty-two per cent (n = 79,925) of respondents answered ‘Sometimes’ or ‘Often’. Fear is a dummy variable equal to 1 if a respondent feels fear ‘Sometimes’ or ‘Often’.

We assess the respondents' support for female political leadership based on their responses to, ‘Do you think that political leadership positions should be mostly for men, mostly for women, or do you think that both men and women should have equal representation in the political leadership?’Footnote 19 Forty-three per cent of respondents (n = 56,024) answered ‘mostly for men’.Footnote 20 An almost equal percentage (Forty per cent, n = 51,645) believed in equal gender representation in leadership. Eleven per cent support mostly women in leadership (n = 14,616), almost five per cent responded ‘anyone based on merit’ (n = 6,021), and the remaining six per cent answered that they did not know or refused to answer.

We estimate support for female political leadership in three ways. In the following analyses, Model 1 employs a Pro-Women Score. This is a continuous measure where the response ‘Mostly for Men’ equals − 1, ‘Equal Representation’ and ‘Anyone Based on Merit’ (hereafter referred to as ‘Equal Representation’) equals 0, and ‘Mostly for Women’ is 1. Model 2 applies an ordinal logit model with Support Ordered as the outcome variable. In Support Ordered, the lowest order category is ‘Mostly for Men’. ‘Equal Representation’ is the middle-order category, and ‘Mostly for Women’ is the highest-order category. Ordinal logit models assume that the independent variable (Insecurity) has the same effect on the odds of ascending to a higher-order category across the three categories. Models 3 and 4 discard this assumption with a multinomial logit model. We use ‘Equal Representation’ as the baseline comparison category. We assess the relationship between feelings of insecurity and the log odds of responding that leadership positions should be reserved ‘Mostly for Men’ (Model 3) and ‘Mostly for Women’ (Model 4), relative to ‘Equal Representation’.

Hypotheses and Model Specifications

We hypothesize that feelings of, and exposure to, insecurity lower support for female political leadership (H1). The equation below summarizes our model. Insecurity is the main independent variable. All models control for respondents' Gender, Age, Education,Footnote 21 and marital status (Marriage).Footnote 22 Our main models include province and wave fixed effects. Standard errors are clustered at the wave level. We use an ordinary least squares (OLS) model for Model 1, an ordinal logit model for Model 2, and a multinomial logit model for Models 3 and 4. We expect the β coefficient to be negative in Models 1, 2 and 4 and positive in Model 3.

For respondent i, we model:

(1)$$Y_i = \beta \; Insecurity_i + Controls_i + \epsilon _i$$

In our second hypothesis, we expect women who experience insecurity to be less supportive of female political leadership (H2). We test this hypothesis by re-running our analysis on female respondents. For the sample of female-only respondents, we hypothesize the β coefficient to be negative in Models 1, 2, and 4 and positive in Model 3.

Results

Figure 4 previews our main findings. It plots the proportion of respondents who answered ‘Mostly for Men’, ‘Equal Representation’, and ‘Mostly for Women’ between respondents who experienced violence over the past year (Insecure) and those who did not (Secure). A higher rate of insecure respondents answered ‘Mostly for Men’ and ‘Mostly for Women’ than secure respondents. Insecurity may polarize preferences for female leadership.Footnote 23

Figure 4. Gender leadership preferences between secure and insecure respondents.

Regressions corroborate these patterns (SI Section 3.3). Models 1 and 2 in SI Table 21 present a negative but statistically insignificant relationship between exposure to violence (Insecurity) and support for female political leadership. However, Models 3 and 4 demonstrate that insecurity is positively associated with a respondent's likelihood of answering that political leadership positions should be reserved ‘Mostly for Men’ relative to ‘Equal Representation’ (Model 3) and ‘Mostly for Women’ relative to ‘Equal Representation’ (Model 4). These findings persist when using the Fear measure of insecurity (SI Table 22) and removing wave fixed effects (SI Table 23).

These findings remain when subsetting our analysis to female and male respondents (SI Section 3.4.1; 3.4.2). In contradiction of H2, women's exposure to insecurity correlates with diverging levels of support for female leadership (SI Table 27). These diverging responses are robust to different measures of insecurity (SI Table 28) and model specifications (SI Tables 29, 30). Men mirror these patterns as well (SI Tables 33 to 35).

Unlike in the survey experiment, insecurity is negatively associated with both hawks and doves' support for female leadership (SI Tables 25, 26), irrespective of the respondents' gender (SI Tables 31, 32, 37, 38). Furthermore, insecurity is positively associated with support for female leadership among survey respondents from the three provinces where we deployed the survey experiment (SI Tables 24, 30, 36). This contrasts with the rest of Afghanistan, where insecurity correlates with polarized preferences for female leadership (SI Tables 39, 40, and 41). Subnational variation in feelings of insecurity may explain these differences. Respondents from the three provinces in the survey experiment disclosed lower feelings of insecurity than the others (SI Table 15). Perhaps heightened levels of collective insecurity polarize attitudes towards female leadership.

Observational analysis also presents mixed evidence of insecurity amplifying women's support for local governing institutions. Insecure respondents have less support for their provincial government (SI Table 42). However, among female respondents, insecurity is positively associated with women's preferences for more male leadership in provincial government (SI Table 43).

These findings must be interpreted with caution. Less popular provincial governments may abet greater insecurity. A host of confounding variables likely explain both a respondent's exposure to insecurity and their attitudes towards local government and female leadership. These findings do not take social desirability into account. Likewise, respondents with more polarized attitudes towards female leadership may be more likely to be exposed to violence from competing factions. We place greater weight on the experimental findings for these reasons. Nonetheless, by uncovering polarized preferences for female leadership among insecure men and women, our observational analysis highlights a possible blind spot in the conjoint analysis. Diverging responses to insecurity may underpin the pooled sample's null finding in the survey experiment (H1). That insecurity is positively associated with female leadership support among some women and men, and this support varies subnationally invites future research and deeper theorizing on the conditions that orient opposing reactions to insecurity in conflict states.

Conclusion

Combining an original survey experiment with observational data, we find mixed evidence that insecurity decreases public support for female political leadership in Afghanistan. Our survey experiment proposes that women's support for female leadership is more vulnerable to insecurity than men's. We suspect that insecurity pushes women to look to local, male-dominated governing institutions to provide security. Observational data, however, conveys that insecurity may polarize public preferences for female leadership in some places and increase support for female leadership in others. We place more empirical weight on the survey experiment's findings because the experimental design tackles the social desirability biases and endogeneity concerns that undermine observational studies. Nevertheless, observational analysis cautions that our experimental design may miss diverging reactions to insecurity among women and men (Abramson, Koçak, and Magazinnik Reference Abramson, Koçak and Magazinnik2022). Combined, our experimental and observational findings stress that the impact of insecurity on support for female leadership in conflicts where women's political inclusion is violently contested may be highly heterogeneous.

These findings offer important theoretical and empirical contributions to the conflict and gender literature. Our mixed results depart from most scholarship, which demonstrates a consistently negative effect of insecurity on public support for female leadership. We locate the point of this departure from existing literature to the context of our study: a conflict state with militants opposed to women's political inclusion. In this environment, we theorized insecurity would dampen support for female leaders, especially among women.

Our mixed findings, however, beckon further theory-building and testing. We did not expect insecurity to ever be positively associated with support for female leadership in Afghanistan. Future work can interrogate the conditions in which men and women in conflict states view insecurity and female leadership as trade-offs or complements. While we believe our findings to be most applicable to conflicts where women's rights are an axis of contention, we urge scholars to examine these processes in other conflict and non-conflict settings, including those in which men are more open to female political leadership.

Methodologically, our two-stage survey experiment – combining a prime and conjoint exercise – mitigates endogeneity concerns and social desirability biases to clarify how insecurity influences public preferences for female leadership in a conflict state. This methodology holds promise for further empirical work on sensitive topics in such environments. Our findings point to potential limitations of conjoint designs; they may obscure important variations within different subgroups. Assessing the external validity of future experimental analysis through observational and/or other types of data will be crucial in mitigating these limitations and teasing out mechanisms.

A lack of informational equivalence across experimental conditions may also confound our results (Dafoe, Zhang, and Caughey Reference Dafoe, Zhang and Caughey2018). Respondents may use gender as a proxy for more specific competencies and policies not captured in the experimental design. To account for these limitations, future studies could provide more specific information about a potential candidate's platform, particularly their security policies and personal attributes.Footnote 24

We expect these findings to most apply to conflicts where women's political inclusion is violently contested. They may have less relevance in contexts where men's baseline support for female leadership is significantly higher than in Afghanistan. A floor effect would not constrain insecurity's impact on men's support for female leadership in these environments. We expect less divergence between men and women's support for female leadership in the wake of insecurity in these settings.

The policy implications of our findings merit further inquiry. Our research proposes that insecurity generally (though not always) penalizes, if not polarizes, public support for female leadership in conflicts where women's political inclusion is violently contested. Neither outcome is conducive to robust public backing of women's political representation. While institutional remedies such as quotas may expand women's political representation, they are unlikely to strengthen public support for female political representation without accompanying improvements in security.

Finally, our findings are even more salient in the wake of the Taliban takeover of Afghanistan in August 2021. Since assuming control of the country, the new government has taken steps to reverse many of the institutional gains experienced by female leaders in the post-2001 era. Shortly after seizing power, a Taliban spokesperson dismissed the prospect of female ministers in government and asserted that women protesting for greater representation are unrepresentative of Afghan women (Sabin Reference Sabin2021).

Contrary to Taliban claims, Afghan women's attitudes towards women's political leadership are invariably linked to the broader security context. Our research suggests that insecurity depresses, if not polarizes, women's support for female leadership. A secure environment is vital to upholding popular support for women's political inclusion. In spite of efforts to promote women's rights over its twenty-year incumbency, the IRoA and its international supporters ultimately failed to provide such an environment. Afghan women and girls now bear the costs of this failure.

Supplementary Material

The supplementary material for this article can be found at https://doi.org/10.1017/S000712342300056X

Data availability statement

Replication data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/OIV11K

Acknowledgements

The authors would like to thank our three reviewers, in addition to Carolyn Barnett, Vivienne Badaan, Chao-Yo Cheng, Brandon De La Cuesta, Dino Hadzic, Kristen Kao, Nicholas Lotito, Elizabeth Nugent, Steve Oliver, Tim Robinson, Alejandro Seif, Leonard Wantchekon, Erin York, and participants at the Harvard Kennedy School Middle East Initiative (MEI)'s summer research colloquium, the Center for Experimental Social Sciences (CESS), Nuffield College, Oxford University's experimental design colloquium, Yale-NUS College's Global Affairs Working Paper Series, Yale's Middle East Politics Workshop, the American Political Science Association (APSA)'s 2021 Annual Conference and the 2022 Association for Analytic Learning about Islam and Muslim Societies (AALIMS) conference at Princeton University for their excellent feedback. Jonathan Chu, Sharan Grewal and Laszlo Horvath provided especially helpful guidance in revising the manuscript.

Author contributions

The authors are named in alphabetical order and contributed equally to the manuscript.

Financial support

This work was supported by the International Growth Centre (JB, grant number S-89329-AFG-1).

Competing interests

None.

Ethical standards

This research received Research Ethics Approval No SSH_DPIR_C1A_16_0_010 from Oxford University's Department of Politics and International Relations Departmental Research Ethics Committee (DREC).

Footnotes

1 We define conflict states following the World Bank's categorization of countries affected by ongoing violent conflict based on a threshold number of conflict-related deaths relative to the population (World Bank 2022).

2 Our second hypothesis (H2) is exploratory. Unlike H1, H2 and its underlying theory emerged inductively from our conjoint analysis. We urge scholars to test H2 deductively in future research. See Supplementary Information (SI) Section 1 for more information.

3 See SI Section 1 for more on survey logistics and implementation.

4 The sequence of survey questions following the prime was randomized to reduce order effects. Socio-economic and demographic questions were asked at the end of the survey.

5 We include participants who failed the manipulation check, as excluding them may introduce other types of bias into the results (for example, by excluding a disproportionate number of respondents with no education).

6 Kunduz province was excluded as a place of birth because it was too insecure for data collection at the time the survey was finalized, although security conditions subsequently improved and we were able to run the survey there as well.

7 This is because of the vast expansion of public school education in Afghanistan over the last twenty years.

8 Responses had a Standard Deviation of 1.5.

9 These are respondents who answered ‘Disagree’ or ‘Strongly Disagree’.

10 These are respondents who answered ‘Neither Agree or Disagree’, ‘Agree’, or ‘Strongly Agree.’

11 SI Section 2.3 and Leeper, Hobolt, and Tilley (Reference Leeper, Hobolt and Tilley2020) discuss why MMs are a more appropriate measure for conjoint subgroup analysis.

12 SI Section 2.6 has a more extended discussion on the insecurity prime's impact on men's female leadership preferences.

13 The order of these questions was randomized. They sometimes preceded or followed the conjoint exercise.

14 In Afghanistan, the title is associated with having fought the Soviet occupation during the 1980s and is claimed by both pro- and anti-IRoA fighters.

15 We also use this question to categorize hawks and doves.

16 Provincial governments and local institutions were not fully male dominated. Quotas existed for women within provincial, district, and local community development councils. In practice, these local bodies were hampered by a lack of authority stemming from poorly defined roles, and gender quotas were not always enforced (Bhatia, Jareer, and Mcintosh Reference Bhatia, Jareer and Mcintosh2018; Brown Reference Brown2021). By contrast, informal governing institutions and networks remained highly influential (Mukhopadhyay Reference Mukhopadhyay2009a).

17 Question x16 in the Survey.

18 Question x15. Note that Wave 8 the Survey included ‘Always’ as an option. That answer is measured as ‘Often’ in the merged dataset.

19 Question x69.

20 This includes six respondents who answered ‘women should do house work’.

21 Education is a continuous variable where ‘No Formal School’ equals 0 and ‘University Education’ equals 4.

22 Marriage equals 1 if the respondent is married.

23 SI Tables 18–20 provide more descriptive statistics.

24 See Blackman and Jackson (Reference Blackman and Jackson2021) for an excellent example of this.

References

Abramson, SF, Koçak, K and Magazinnik, A (2022) What do we learn about voter preferences from conjoint experiments? American Journal of Political Science 66(4), 1008–20.CrossRefGoogle Scholar
Ahmadi, B (2019) Afghanistan Talks: No Women, No Peace. Technical report. United States Institute of Peace. Available at https://www.usip.org/publications/2019/03/afghanistan-talks-no-women-no-peace (accessed 12 October 2022).Google Scholar
Anderlini, SN (2007) Women Building Peace: What They do, Why it Matters. Boulder, CO: Lynne Rienner Pub.CrossRefGoogle Scholar
Bahesh, H (2021) Towards Equality: How Afghan Women Conquer 27 percent share in parliament after decades of war. The Asahi Shimbun (21 June 2021). Available at https://www.asahi.com/ajw/articles/14377179 (accessed 12 October 2022).Google Scholar
Bahri, J (2014) Western gender policies in Afghanistan: Failing women and provoking men. Gender, Technology and Development 18(2), 163–85.CrossRefGoogle Scholar
Barr, H (2020) A Crucial Moment for Women's Rights in Afghanistan. Technical report. Human Rights Watch. Available at https://www.hrw.org/news/2020/03/05/crucial-moment-womens-rights-afghanistan (accessed 12 October 2022).Google Scholar
Beath, A, Christia, F and Enikolopov, R (2013) Empowering women through development aid: evidence from a field experiment in Afghanistan. American Political Science Review 107(3), 540–57.CrossRefGoogle Scholar
Bhatia, J and Monroe, SL (2023) Replication Data for: Insecurity and Support for Female Leadership in Conflict States: Evidence from Afghanistan. Available at https://doi.org/10.7910/DVN/OIV11K, Harvard Dataverse, V1.CrossRefGoogle Scholar
Bhatia, J, Jareer, N and Mcintosh, R (2018) Community-driven development in Afghanistan. Asian Survey 58(6), 1042–65.CrossRefGoogle Scholar
Blackman, AD and Jackson, M (2021) Gender stereotypes, political leadership, and voting behavior in Tunisia. Political Behavior 43(3), 1037–66.CrossRefGoogle Scholar
Broadbent, E (2010) Women's Participation in Elections in Afghanistan. Technical report. Governance and Social Development Resource Centre. Available at https://gsdrc.org/publications/womens-participation-in-elections-in-afghanistan/ (accessed 12 October 2022).Google Scholar
Brown, FZ (2021) Aiding Afghan Local Governance: What Went Wrong? Washington, DC: Carnegie endowment for international peace.Google Scholar
Burbidge, HD et al. (2016) Afghanistan In 2016: A Survey of the Afghan People. Technical report. The Asia Foundation.Google Scholar
Bush, SS (2011) International politics and the spread of quotas for women in legislatures. International Organization 65(1), 103–37.CrossRefGoogle Scholar
Butler, DM, Tavits, M and Hadzic, D (2023) Gender bias in policy representation in post-conflict societies. Political Research Quarterly 76(1), 200–12.CrossRefGoogle Scholar
Carlin, RE, Carreras, M and Love, GJ (2020) Presidents’ sex and popularity: Baselines, dynamics and policy performance. British Journal of Political Science 50(4), 1359–79.CrossRefGoogle Scholar
Chishti, M (2020) The pull to the liberal public: Gender, orientalism, and peace building in Afghanistan. Signs: Journal of Women in Culture and Society 45(3), 581603.CrossRefGoogle Scholar
Dafoe, A, Zhang, B and Caughey, D (2018) Information equivalence in survey experiments. Political Analysis 26(4), 399416.CrossRefGoogle Scholar
Dolan, K (2014a) Gender stereotypes, candidate evaluations, and voting for women candidates: What really matters? Political Research Quarterly 67(1), 96107.CrossRefGoogle Scholar
Dolan, K (2014b) When Does Gender Matter?: Women Candidates and Gender Stereotypes in American Elections. New York, NY: Oxford University Press.CrossRefGoogle Scholar
Echebarria-Echabe, A and Fernández-Guede, E (2006) Effects of terrorism on attitudes and ideological orientation. European Journal of Social Psychology 36(2), 259–65.CrossRefGoogle Scholar
Gopal, A (2021) The Other Afghan Women. The New Yorker (6 September 2021). Available at https://www.newyorker.com/magazine/2021/09/13/the-other-afghan-women (accessed 12 October 2022).Google Scholar
Hadzic, D and Tavits, M (2019) The gendered effects of violence on political engagement. The Journal of Politics 81(2), 676–80.CrossRefGoogle Scholar
Hadzic, D and Tavits, M (2021) Wartime violence and post-war women's representation. British Journal of Political Science 53(3), 1024–39.CrossRefGoogle Scholar
Hainmueller, J, Hopkins, DJ and Yamamoto, T (2014) Causal inference in conjoint analysis: Understanding multidimensional choices via stated preference experiments. Political Analysis 22(1), 130.CrossRefGoogle Scholar
Haque, T (2020) Afghanistan's Development Gains. Technical report. World Bank. Available at https://openknowledge.worldbank.org/handle/10986/33209 (accessed 12 October 2022).Google Scholar
Holman, MR, Merolla, JL and Zechmeister, EJ (2011) Sex, stereotypes, and security: A study of the effects of terrorist threat on assessments of female leadership. Journal of Women, Politics & Policy 32(3), 173–92.CrossRefGoogle Scholar
Holman, MR, Merolla, JL and Zechmeister, EJ (2016) Terrorist threat, male stereotypes, and candidate evaluations. Political Research Quarterly 69(1), 134–47.CrossRefGoogle Scholar
Holman, MR, Merolla, JL and Zechmeister, EJ (2022) The curious case of Theresa May and the public that did not rally: Gendered reactions to terrorist attacks can cause slumps not bumps. American Political Science Review 116(1), 249–64.CrossRefGoogle Scholar
Horiuchi, Y, Markovich, Z and Yamamoto, T (2022) Does conjoint analysis mitigate social desirability bias? Political Analysis 30(4), 535–49.CrossRefGoogle Scholar
Huddy, L and Terkildsen, N (1993) Gender stereotypes and the perception of male and female candidates. American Journal of Political Science 37, 119–47.CrossRefGoogle Scholar
Jones, S (2018) The Many Dangers of Being an Afghan Woman in Uniform. New York Times Magazine (5 October 2018). Available at https://www.nytimes.com/2018/10/05/magazine/afghanistan-women-security-forces.html (accessed 12 October 2022).Google Scholar
Kandiyoti, D (2007a) Between the hammer and the anvil: Post-conflict reconstruction, Islam and women's rights. Third World Quarterly 28(3), 503–17.CrossRefGoogle Scholar
Kandiyoti, D (2007b) Old dilemmas or new challenges? The politics of gender and reconstruction in Afghanistan. Development and Change 38(2), 169–99.CrossRefGoogle Scholar
Kao, K and Benstead, LJ (2021) Female electability in the Arab World: The advantages of intersectionality. Comparative Politics 53(3), 427–64.CrossRefGoogle Scholar
Karim, S (2019) Restoring confidence in post-conflict security sectors: Survey evidence from Liberia on female ratio balancing reforms. British Journal of Political Science 49(3), 799821.CrossRefGoogle Scholar
Kim, NK and Kang, AJ (2022) External threat environments and individual bias against female leaders. Political Science Research and Methods 10(1), 117.CrossRefGoogle Scholar
Landau, MJ et al. (2004) Deliver us from evil: The effects of mortality salience and reminders of 9/11 on support for President George W. Bush. Personality and Social Psychology Bulletin 30(9), 1136–50.CrossRefGoogle ScholarPubMed
Lawless, JL (2004) Women, war, and winning elections: Gender stereotyping in the post-September 11th era. Political Research Quarterly 57(3), 479–90.CrossRefGoogle Scholar
Leeper, TJ, Hobolt, SB and Tilley, J (2020) Measuring subgroup preferences in conjoint experiments. Political Analysis 28(2), 207–21.CrossRefGoogle Scholar
Maoz, I (2009) The women and peace hypothesis? The effect of opponent negotiators’ gender on the evaluation of compromise solutions in the Israeli-Palestinian conflict. International Negotiation 14(3), 519–36.CrossRefGoogle Scholar
Maoz, I (2011) Women and peace hypothesis. The Encyclopedia of Peace Psychology, 1–4. https://doi.org/10.1002/9780470672532.wbepp296.Google Scholar
Mukhopadhyay, D (2009a) Disguised warlordism and combatanthood in Balkh: The persistence of informal power in the formal Afghan state. Conflict, Security & Development 9(4), 535–64.CrossRefGoogle Scholar
Mukhopadhyay, D (2009b) Warlords as Bureaucrats: The Afghan Experience. Washington, DC: CEIP: Carnegie Endowment for International Peace.Google Scholar
Nail, PR et al. (2009) Threat causes liberals to think like conservatives. Journal of Experimental Social Psychology 45(4), 901–7.CrossRefGoogle Scholar
Ono, Y and Burden, BC (2019) The contingent effects of candidate sex on voter choice. Political Behavior 41(3), 583607.CrossRefGoogle Scholar
Rashid, A (2010) Taliban: Militant Islam, oil and Fundamentalism in Central Asia. New Haven, CT: Yale University Press.Google Scholar
Sabin, L (2021) Taliban Confirm that no women will be made ministers in Afghanistan, saying they should give birth instead. The Independent (10 September 2021). Available at https://www.independent.co.uk/asia/south-asia/taliban-spokesperson-afghanistan-women-ministers-b1917937.html (accessed 12 October 2022).Google Scholar
Schuessler, J and Freitag, M (2020) Power Analysis for Conjoint Experiments. Available at https://osf.io/preprints/socarxiv/9yuhp/.CrossRefGoogle Scholar
Schüller, S (2015) The 9/11 conservative shift. Economics Letters 135, 8084.CrossRefGoogle Scholar
Schwartz, JA and Blair, CW (2020) Do women make more credible threats? Gender stereotypes, audience costs, and crisis bargaining. International Organization 74(4), 872–95.CrossRefGoogle Scholar
Schwarz, S and Coppock, A (2022) What have we learned about gender from candidate choice experiments? A meta-analysis of sixty-seven factorial survey experiments. The Journal of Politics 84(2), 655–68.CrossRefGoogle Scholar
Shair-Rosenfield, S and Wood, RM (2017) Governing well after war: How improving female representation prolongs post-conflict peace. The Journal of Politics 79(3), 9951009.CrossRefGoogle Scholar
Shapiro, RY and Mahajan, H (1986) Gender differences in policy preferences: A summary of trends from the 1960s to the 1980s. Public Opinion Quarterly 50(1), 4261.CrossRefGoogle Scholar
Singh, SP and Tir, J (2021) Threat-inducing violent events exacerbate social desirability bias in survey responses. American Journal of Political Science 67(1), 154–69.CrossRefGoogle Scholar
Streb, MJ et al. (2008) Social desirability effects and support for a female American president. Public Opinion Quarterly 72(1), 7689.CrossRefGoogle Scholar
Swenson, G (2017) Why US efforts to promote the rule of law in Afghanistan failed. International Security 42(1), 114–51.CrossRefGoogle Scholar
Tripp, AM (2015) Women and Power in Post-Conflict Africa. Cambridge, UK: Cambridge University Press.CrossRefGoogle Scholar
United States Institute of Peace (2021) What is UNSCR 1325? United States Institute of Peace. https://www.usip.org/gender_peacebuilding/about_UNSCR_1325 (accessed 12 October 2022).Google Scholar
Van de Vyver, J et al. (2016) Boosting belligerence: How the July 7, 2005, London bombings affected liberals’ moral foundations and prejudice. Psychological Science 27(2), 169–77.CrossRefGoogle ScholarPubMed
Watson Institute for International and Public Affairs (2021) U.S. Costs to Date for the War in Afghanistan, in $ Billions FY2001-FY2022. Brown University. Available at https://watson.brown.edu/costsofwar/figures/2021/human-and-budgetary-costs-date-us-war-afghanistan-2001-2022 (accessed 12 October 2022).Google Scholar
Weigand, F (2022) Waiting for Dignity: Legitimacy and Authority in Afghanistan. New York, NY: Columbia University Press.CrossRefGoogle Scholar
Wood, R and Ramirez, MD (2018) Exploring the microfoundations of the gender equality peace hypothesis. International Studies Review 20(3), 345–67.CrossRefGoogle Scholar
World Bank (2022) FY23 List of Fragile and Conflict-Affected Situations. Available at https://thedocs.worldbank.org/en/doc/69b1d088e3c48ebe2cdf451e30284f04-0090082022/original/FCSList-FY23.pdf (accessed 11 June 2022).Google Scholar
Figure 0

Table 1. Leadership profile attributes

Figure 1

Figure 1. Insecurity and preferences for male and female leadership across control and treatment groups (H1): estimated marginal means (MM) and 95% confidence intervals.

Figure 2

Figure 2. Insecurity and preferences for male and female leadership among male respondents (H2): estimated marginal means (MM) and 95% confidence intervals.

Figure 3

Figure 3. Insecurity and preferences for male and female leadership among female respondents (H2): estimated marginal means (MM) and 95% confidence intervals.

Figure 4

Figure 4. Gender leadership preferences between secure and insecure respondents.

Supplementary material: File

Bhatia and Monroe supplementary material
Download undefined(File)
File 649.8 KB
Supplementary material: Link

Bhatia and Monroe Dataset

Link