POLITICS AND THE GLASS CLIFF: EVIDENCE THAT WOMEN ARE PREFERENTIALLY SELECTED TO CONTEST HARD-TO-WIN SEATS
Michelle K. Ryan and S. Alexander Haslam, School of Psychology, University of Exeter; Clara Kulich, School of Business, University of Exeter.
The research reported in this article was funded by a grant from the Economic and Social Research Council (RES-062-23-0135) and an RCUK Academic Fellowship awarded to the first author. We are grateful to Andrew Hindmoor and Cate Atkins for their help with data collection.
Address correspondence and reprint requests to: Michelle Ryan, School of Psychology, The University of Exeter, Exeter, EX4 4QG, UK. E-mail: M.Ryan@exeter.ac.uk
Recent archival and experimental research has revealed that women are more likely than men to be appointed to leadership positions when an organization is in crisis. As a result, women often confront a “glass cliff” in which their position as leader is precarious. Our first archival study examined the 2005 UK general election and found that, in the Conservative party, women contested harder to win seats than did men. Our second study experimentally investigated the selection of a candidate by 80 undergraduates in a British political science class to contest a by-election in a seat that was either safe (held by own party with a large margin) or risky (held by an opposition party with a large margin). Results indicated that a male candidate was more likely than a woman to be selected to contest a safe seat, but there was a strong preference for a female rather than a male appointment when the seat was described as hard to win. Implications for women's participation in politics are discussed.
Despite evidence that women are beginning to break through the “glass ceiling” that has hindered their full participation in economic and political life, they have yet to achieve true gender equality. Women continue to be underrepresented in the most powerful positions of society. This observation is particularly true in the corporate world, where women are clearly a minority among those in power. For example, within the European Union, women make up, on average, just over 10% of the top executives in the top 50 publicly quoted companies (European Commission, 2005). Similarly, in the United States, women compose less than 16% of corporate officers and fewer than 15% of members of boards of directors within Fortune 500 companies (Catalyst, 2009).
In the political sphere the picture is quite similar, with women making up only 19% of UK Members of Parliament (UK Parliament, 2009), 17% of the U.S. Congress (Center for American Women and Politics, 2009), 21% of the European parliament, 18% of politicians in Asia, and only 10% of politicians in the Arab States (Inter-Parliamentary Union, 2009). Furthermore, research suggests that the candidate-selection process favors male candidates over female candidates (Equal Opportunities Commission [EOC], 2002), and that voters generally have a preference for male candidates (e.g., Falk & Kenski, 2006; Fox & Smith, 1998; cf. Norris & Lovenduski, 1995).
Many investigations into the gender imbalance in politics focus on the traits and abilities of male and female politicians. Some commentaries explore the differences between men and women as an argument against excluding women from politics, arguing for the distinctive contribution that women can (and do) make in the political sphere (Sawer, 2000). The notion that women bring something different to the political table is echoed by women politicians themselves. Britain's first (and only) female Prime Minister, Margaret Thatcher, suggested that male and female politicians do indeed behave differently: “In politics, if you want something said, ask a man; if you want something done, ask a woman” (Davies, 2002). Also pointing to the potential for women to make a positive impact in the political arena, British Labour Member for Parliament (MP) and former Cabinet Minister, Clare Short, has claimed that, “[a]s more women come into the Commons, the culture will change, the agenda of politics will broaden, and the institution itself will be transformed” (quoted in McDougal, 1998, p. 137).
Examining differences between men and women can help us explain women's underrepresentation and provide arguments for equality. However, such a strategy runs the risk of being appropriated as an argument for exactly the opposite case—the exclusion of women from political life. Indeed, Fukuyama (1998) considered gender disparities that are said to be rooted in genetic and evolutionary differences and argued that, due to “essential” differences between men and women, the face of politics would change for the worse as a result of women's increased political participation.
An alternative to a group difference approach would involve examination of the context surrounding women's political participation. One way in which this approach can be achieved is by examining the types of political opportunities women are offered, the positions they achieve when they take on political roles, and the barriers they face both in attaining and in filling those roles. This article will examine these issues from the perspective of research that suggests that women are more likely than men to occupy leadership positions that have an increased risk of failure (Ryan & Haslam, 2005, 2007). Extending the metaphor of the glass ceiling, this phenomenon is labeled the “glass cliff” to evoke the precariousness of such positions due to the dangers of falling from the heights of leadership.
The Glass Cliff
In order to understand women's underrepresentation as political leaders, we can look to research into gender equality in organizational leadership. Previous research in the organizational realm suggests that women leaders may occupy positions that are very different from those of their male counterparts (Ryan & Haslam, 2005, 2007). For example, in an examination of the top 100 companies on the London Stock Exchange (the FTSE 100), Ryan and Haslam (2005) analyzed patterns of share price performance before and after the appointment of a male or female board member. This archival research revealed that, in a time of financial downturn in the stock market, companies that appointed a woman to their board had experienced consistently poorer performance in the 5 months preceding the appointment than a matched sample of companies that had appointed men to their boards. We argued that these women confronted a glass cliff because they were more likely than men to be appointed to leadership positions in problematic organizational circumstances. In these cases, women came to occupy leadership roles that had an increased risk of failure and could thus be seen as more precarious than those of their male counterparts. Moreover, these positions are likely to attract more attention than those of men, to expose women to greater stress than men (Ryan, Haslam, Hersby, Kulich, & Atkins, 2007), and to make women vulnerable to higher risk of blame for negative outcomes for which they are not responsible (Ryan & Haslam, 2007; but cf. Kulich, Ryan, & Haslam, 2007, where women leaders were held less responsible for organizational outcomes than were men).
Support for the notion of glass cliffs comes from quantitative research that suggests that women's leadership opportunities are very different from those of men. In particular, studies in which organizational performance is experimentally manipulated demonstrate the preferential selection of women for leadership positions in times of crisis. A consistent finding in this research is that women are selected for leadership positions ahead of equally qualified men when (and only when) there is a high risk of organizational (and leader) failure (Ashby, Ryan, & Haslam, 2007; Bruckmüller & Branscombe, in press; Haslam & Ryan, 2008).
Within the political realm there is already some evidence that women are selected to run for seats that have different characteristics from those for which their male counterparts are candidates. In the United States, former Democratic National Committee Chair, John Bailey, once stated that “the only time to run a women is when things look so bad that your only chance is to do something dramatic” (quoted in Burrell, 1993, p. 123). Indeed, historical research suggests that, in the United States between 1916 and 1978, women were more likely than men to run for Congress in “hopeless” districts, that is, districts where their party had received less than 40% of the vote in the previous election (Gertzog & Simard, 1981).
In the United Kingdom, political commentary has acknowledged that, whereas male candidates and MPs are disproportionately represented in safe (i.e., easy to win) seats, women are more likely to be concentrated in marginal seats (Lovenduski, 2005; Short, 1996). Indeed, in recognition of the fact that men and women may not be competing on level playing fields, the UK Labour party introduced the affirmative action strategies of “twinning,” where constituencies with the same likelihood of being won by a political party are paired and then a male candidate is placed in one constituency and a female candidate in the other (Stokes, 2005).
Anecdotally, one also can interpret the careers of a number of high-profile female politicians in terms of glass cliffs. In the United Kingdom, Baroness Margaret Thatcher twice ran as a Conservative candidate for a strong, safe Labour seat, losing at both attempts, before winning and going on to serve as Prime Minister. Kim Campbell, the first female Canadian Prime Minister, took over the leadership position of the Conservative party in troubled times and then duly lost the next election only months later. In Australian politics, it is notable that there have only ever been three female State Premiers—Joan Kirner in Victoria, Carmen Lawrence in Western Australia, and Anna Bligh in Queensland. All three were appointed mid-term (replacing male counterparts), Kirner and Lawrence after their party had been exposed to humiliating scandals. As a result, both Kirner and Lawrence faced the prospect of unwinnable elections, which they duly went on to lose, while only Bligh went on to win the position in her own right.
The Present Research
Given existing gender inequities in the political sphere and preliminary evidence for a political glass cliff, the present research sought to establish empirically whether women are differentially selected to contest seats that are difficult to win. Study 1 documents the phenomenon by way of archival research into the 2005 UK general election. Yet, although this study may provide evidence of the existence of a glass cliff in the political arena, it still raises questions about whether the phenomenon is due to women's political (in)abilities, such as being less experienced or less qualified or whether it is more directly due to gender per se. These questions are addressed in experimental research, which allows these (and other) variables to be controlled. More specifically, Study 2 uses experimental methods to examine relative preference for male and female candidates in the selection phase of a fictitious by-election where the winnability of the seat is manipulated.
Within the United Kingdom, women are clearly underrepresented in the political sphere. Women make up 21% of the current Labour cabinet (Labour party, 2008), 19% of the lower house (UK Parliament, 2009), 20% of the upper house (Centre for Advancement of Women in Politics, 2008), 35% of the Scottish parliament (Scottish Parliament, 2009), and 17% of the Northern Ireland Assembly (Northern Ireland Assembly, 2009), with equality being achieved only in the National Assembly for Wales (where the Labour party used affirmative action measures; Equality and Human Rights Commission, 2008).
In the 2005 general election, 720 women ran as candidates (20.3% of all candidates), providing an opportunity to explore the electorates in which women were chosen to run as a function of party and seat winnability. In this election, within the two main parties, 27% of the Labour candidates were women, compared to 19% of Conservative candidates. Of the 646 constituencies, almost half of them (300) had all men as the three main challengers (Fawcett Society, 2005). As noted above, of the two main parties, only the Labour party took affirmative action in relation to gender, using all-women shortlists to ensure that women were selected to contest winnable seats.
Study 1 looked at this election more closely in order to examine the electoral success of women and the nature of the seats they were selected to contest. Our study rests on observations by Gertzog and Simard (1981) that women are more likely than men to run in unwinnable (or hopeless) seats. Our research extends this work by looking at winnability as a continuous (rather than discrete) variable and by examining the impact of seat winnability on votes received. In light of Labour's affirmative action policies, we expected that the normative practice of placing women in precarious glass cliff positions would be disrupted and thus we did not expect to find gender differences in electoral success or seat winnability for the Labour party. However, in line with the evidence of glass cliffs in other domains (Ryan & Haslam, 2005, 2007), we hypothesized that, in the Conservative party, women would be selected to contest seats that were significantly less winnable than those in which their male counterparts ran (Hypothesis 1). Moreover, we predicted that such selection biases would have an impact on women's electoral success, such that they would receive a smaller percentage of the vote than men (Hypothesis 2).
Design and Procedure
Using data provided by the UK Electoral Commission (Electoral Commission, 2006), a data set was compiled that contained detailed information about the 2005 UK general election. Only data pertaining to the two main parties (Labour, the governing party that retained power after the election, and Conservative) were included in the analysis. In all, data from 1,258 candidates were analyzed, including 289 women and 969 men. The data set contained the following information: (a) constituency, (b) candidate name, (c) candidate gender, (d) party affiliation, (e) percentage of votes won, (f) percentage of votes won by party in the previous election (in 2001), and (g) the relative winnability of the seat for each candidate, calculated as the percentage margin by which their party had won (or lost) the previous election.
In order to investigate the impact of candidate gender and political party on seat winnability, a 2 (gender of candidate) × 2 (party: Conservative or Labour) between-groups analysis of covariance was conducted. Unsurprisingly, the analyses revealed a significant main effect for party, F(1, 1,161) = 258.34, p < .001, η2= .18, such that the Labour party (the governing party) candidates ran for seats that were significantly more winnable than did the Conservative party. The analyses also revealed a significant main effect for gender, F(1, 1,161) = 10.67, p = .001, η2= .009, indicating that, overall, women ran for less winnable seats. On average, women contested seats that were held by the opposition by a margin of 5.1% of the votes. In contrast, men contested seats that were held by the opposition by 2.6% of the vote.
However, these two main effects were qualified by a two-way interaction, F(1, 1,161) = 20.46, p < .001, η2= .017. In line with Hypothesis 1, tests of simple effects to decompose this interaction revealed that gender differences in the winnability of seats occurred only for Conservative party candidates. Here, women ran for seats that were significantly less winnable in the sense that they were held by the opposition party by a much larger margin (M = 26.2%, SD = 21.9) than those contested by men (M = 12.4%, SD = 23.4), F(1, 1,161) = 27.42, p < .001, η2= .023. Indeed, based on these figures, the seats in which Conservative women ran can be seen as extremely safe Labour seats. No gender difference was apparent for the Labour party, F(1, 1,161) = .88, p = .35.
In order to investigate electoral success (measured as percentage of votes won) as a function of candidates' gender and political party, a 2 (gender of candidate) × 2 (party: Conservative or Labour) between-groups analysis of variance (ANOVA) was conducted. As expected, given that Labour won the election, the analyses revealed a significant main effect for party, F(1, 1,254) = 68.93, p < .001, η2= .052, such that, on average, Labour candidates won a higher percentage of votes (M = 37.83, SD = 15.17) than did the Conservative candidates (M = 31.40, SD = 14.11). The analyses also revealed a significant main effect for gender, F(1, 1, 254) = 12.34, p < .001, η2= .010, such that, on average, women won a smaller percentage of votes (M = 32.67, SD = 14.42) than did men (M = 35.18, SD = 15.12). These main effects, however, were qualified by a significant two-way interaction, F(1, 1,254) = 7.97, p = .005, η2= .006. In line with Hypothesis 2, simple effects demonstrated that gender differences in electoral success occurred only for the Conservative party, where female candidates achieved a significantly smaller proportion of the vote (M = 26.38, SD = 12.76) than did men (M = 32.61, SD = 14.17), F(1, 1,254) = 18.17, p < .001, η2= .014. A gender difference was not apparent for the Labour party, F(1, 1,254) = .27, p = .61.
The above analyses demonstrate that, within the UK Conservative party, women won a significantly smaller proportion of the vote than men. In order to examine whether women's lack of success could be explained by the winnability of the seats they contested, a stepwise regression analysis was conducted. In a first step, regression demonstrated that, within the Conservative party, there was a significant negative relationship between gender and electoral success (β=−.22, p < .001) such that female candidates enjoyed significantly less success than male candidates. However, when electoral success was simultaneously regressed on gender and seat winnability, the previously significant relationship between candidate gender and electoral success became nonsignificant (β=−.003, p = .87), while at the same time there was a significant relationship between winnability and electoral success (β= .92, p < .001). Taken together, these results indicate that although Conservative women performed less well than their male counterparts, receiving, on average, smaller percentages of the vote in their electorates, this lack of success disappeared entirely when seat winnability was taken into account.
The results from this study provide some archival support for the existence of the glass cliff in the 2005 UK general election. Consistent with our first hypothesis, within the Conservative party, women were selected to contest seats that were significantly less winnable than those of their male counterparts. Indeed, the seats in which Conservative women ran were clearly Labour strongholds, needing, on average, an almost impossible swing of over 26% of the vote to win. Moreover, supporting our second hypothesis, female candidates actually performed less well than did male candidates, securing a significantly smaller proportion of the vote in their elections.
Conservative women's lack of electoral success could potentially lead commentators to conclude—as they have in the organizational domain (Judge, 2003)—that women underperform in the political arena and are poorly equipped for meeting the challenges of pursuing electoral office. However, countering this argument, it is clear from the results of the regression analyses that there were no gender differences in performance when seat winnability was taken into account. Such a finding suggests that women's lack of electoral success is not due to their inability to convince the voters, but rather, consistent with the glass cliff, to the winnability of the seats in which they ran.
Additionally, the results of Study 1 show that female Labour party candidates did not face the same glass cliff as did Conservative women. Labour women were placed in seats that were as winnable as those of their male counterparts, and they enjoyed the same level of electoral success. This finding points to the success of affirmative action programs in countering the particular problems that glass cliffs pose for women who aspire to leadership positions. Moreover, the findings suggest that having been given the opportunity to vote for female candidates in winnable seats, the Labour voting public did not differentiate between male and female candidates. Such a pattern of outcomes across parties argues that the pre-selection of political candidates plays some role in ensuring equal gender representation in public life.
Although the results of Study 1 provide some evidence of the glass cliff in a real political context, archival research of this form has significant limitations. In particular, although the external validity of the research is high, it does not help understand why women occupy less winnable seats than men (Haslam & McGarty, 2003). For example, it could be argued that gender differences in seat allocation are a reflection of multiple realities: that female candidates (a) are less experienced or less qualified than their male counterparts (Palmer & Simon, 2006; Shvedova, 2005; cf. EOC, 2002), (b) have different strengths or abilities from those of men (Childs, 2004), or (c) themselves choose to contest seats that are less winnable.
One way of addressing this issue is to conduct experimental research that manipulates candidate gender while controlling for other variables (such as experience or age) that may covary with gender in the world at large. In short, this method allows us to ascertain whether women are selected to contest less winnable seats precisely because they are women. Drawing on insights gained from Study 1, and congruent with experimental research conducted in the organizational realm (Haslam & Ryan, 2008), Study 2 utilized a by-election (i.e., a special election) scenario in which political science students selected a candidate to contest a hypothetical seat that was either “safe” (i.e., easily winnable) or “risky”‘ (i.e., hard to win). Adapting the method devised by Haslam and Ryan (2008, Studies 1 and 2), two target candidates were described—a male and a female candidate whose experience and qualifications were matched on key dimensions. In line with previous demonstrations of the glass cliff, we predicted that the female candidate would be more likely to be chosen to contest the risky seat than the safe one. In contrast, we expected that the male candidate would be more likely to be chosen to contest the safe seat rather than the risky one.
Participants and Design
Participants were 80 final year undergraduate political science students from a British university who participated in the study during a class exercise. Ages ranged from 18 to 37 years with median age of 19 years; 24 (30%) were women and 56 were men. Participants received a description about a by-election with information about two unnamed political parties—“Party X” and “Party Y.” The information made clear that the contested seat was either held, or not held, by Party X. Participants then evaluated three potential Party X candidates (including the two focal candidates, a male and female candidate) to contest the seat in the by-election. The study thus had a 2 (winnability: high, low) × 2 (gender of candidate) × 2 (gender of participant) design, with repeated measures on the second factor.
Materials and Procedure
Participants were randomly assigned to one of two versions of a six-page questionnaire. The first page began with a newspaper article describing a fictitious by-election to replace a retiring MP (equivalent to a member of the U.S. Congress) in the electoral seat of Caldwell. In order to manipulate the winnability of the seat, the article included seat information from the previous election. In the high-winnability condition (“Caldwell By-Election: Party X Strong”), participants read that in the previous election Party X “won the seat convincingly” and that “all recent polls point to the Party X trend continuing.” In addition, a graph was included showing that Party X won the seat with a total vote of 55.01% in the last election. In the low-winnability condition (“Caldwell By-Election: Party Y Strong”), different participants saw parallel text and a chart that here favored Party Y. In both versions of the questionnaire, the article stated that candidates would be announced shortly. As a manipulation check, participants were asked to indicate who they thought would win the election on a 7-point scale from 1 (Party X will win) through 4 (it will be even) to 7 (Party Y will win).
Participants then were informed that Party X was choosing a candidate to run in the Caldwell by-election. They were given brief descriptions of three potential candidates, modeled on the paradigm developed by Haslam and Ryan (2008). From the candidate descriptions, it was readily apparent that two of the three candidates stood out as being extremely qualified for the position: Candidate 1 (a male candidate) and Candidate 3 (a female candidate). Both candidates were described as long-standing party members, living in the Caldwell district, currently holding local council seats, and sitting on a number of high-profile committees. Although the descriptions were matched on these key dimensions, they varied slightly in their details (e.g., the university that the candidate attended and the Council that they represented), and so the descriptions were counterbalanced on these details across gender. A third (male) candidate (Candidate 2) was included to enhance mundane realism and capture features of the typical candidate shortlists—in which women are a numerical minority. However, this candidate was obviously much weaker than Candidates 1 and 3 because he had recently moved to the area and had only very limited political experience.
After reading about the candidates, participants were asked to evaluate each in turn by indicating their level of agreement with a series of statements on 7-point scales from 1 (do not agree at all) to 7 (agree completely). Three items were designed to measure perceptions of the candidate's suitability for the position (“The candidate's past experience is relevant to the position,”“The candidate would be suitable for the position,” and “The candidate will bring the right skills to the job”; for the male candidate, α= .81, and for the female candidate, α= .87). Two items measured the candidate's leadership ability (“The candidate has clear leadership credentials” and “The candidate has the skills and experience to lead other people”; for the male candidate, r = .59, p < .001, and for the female candidate, r = .69, p < .001). Each variable was computed by averaging over its items.
Participants were then asked to rank the three candidates, from 1 to 3 (where 1 was the preferred representative). Finally participants were asked to provide some basic demographic information. After completing the questionnaire, they were debriefed in full.
Analysis of the manipulation check item suggested that participants clearly understood the manipulation. Participants believed that Party X was more likely to win the election in the highly winnable condition (M = 2.77, SD = 1.22) than in the hard-to-win condition (M = 5.26, SD = 1.52), t(71) = 7.70, p < .001.
Preliminary analyses of variance on both ranking and evaluation measures with all three factors revealed, as expected, that Candidate 2 was indeed seen by participants as significantly less suitable for the position. Accordingly, subsequent analyses focused solely on evaluations of Candidates 1 and 3.
Winnability and Ranking of Candidates
In order to investigate the impact of seat winnability and participant gender on the ranking of the focal candidates, a 2 (winnability: high, low) × 2 (gender of candidate) × 2 (gender of participant) ANOVA was conducted with repeated measures on the candidates' gender. The results revealed no significant main effects (pwin= 84, pcgen= .74, ppgen= .47), but they did identify the predicted two-way interaction between gender of the evaluated candidate and the winnability of the seat as significant, F(1, 76) = 7.08, p = .009, η2= .085. Tests of simple effects to decompose this interaction revealed that in line with predictions, the winnability of the seat had a significant impact on the ranking of the female candidate—she was ranked higher when the seat was hard to win (M = 1.51, SD = .75) than when it was winnable (M = 1.92, SD = .62), F(1, 76) = 4.55, p = .04, η2= .056. Furthermore, for the male candidate, the winnability of the seat also had a significant impact on his rankings, such that he was ranked higher when the seat was winnable (M = 1.41, SD = .64) than when it was hard to win (M = 1.83, SD = .63), F(1, 76) = 6.46, p = .01, η2= .078. There was no effect for participant gender, either on its own or in interaction with the other variables (ppgen= .47, ppgen × cgen= .29, ppgen × cgen × win= .41).
In addition, a chi-square analysis was performed to examine the ranking of the focal candidates as a function of seat winnability. Participants favored the candidates overall almost equally, with 47.9% ranking the female candidate first and 52.1% of participants ranking the male candidate first. However, again in line with predictions, chi-square analysis revealed that winnability had a significant impact on the ranking of the candidates, χ2(1) = 13.31, p < .001. As can be seen in Table 1, when the seat was unwinnable, 68.4% of participants preferred the female candidate to the male candidate, but when the seat was winnable, preference for the female candidate dropped to 25.7%.
Study 2: Percentage of Participants Who Ranked Candidate First as a Function of Candidate Gender and Seat Winnability
Evaluation of Candidates
In order to explore participants' perceptions of the target candidates, evaluations of their suitability for the position and their perceived leadership ability were subjected to a 2 (winnability: low, high) × 2 (evaluated candidate: male, female) × 2 (participant gender: male, female) ANOVA with repeated measures on the second variable. The results from each of these analyses revealed that none of the independent variables, either in isolation or in conjunction with one another, had an impact on either evaluation of the target candidates (ps ranged between .15 and .87).
The results of Study 2 provide experimental evidence for the glass cliff phenomenon in the political arena. In line with hypotheses and other findings in the organizational realm (Haslam & Ryan, 2008), a female candidate was more likely to be chosen to run for a seat that was hard to win than for one that was winnable, whereas a male candidate was more likely to be chosen to run for a seat that was winnable than for one that was more risky.
Importantly, this preference for the female candidate in difficult times cannot be said to be due to her being seen as more or less qualified for the position. Reflecting the matched descriptions of the male and female candidates, there were no gender differences in judgments of suitability or leadership ability, either overall or as a function of winnability. It is also important to note that, consistent with previous research (Ashby et al., 2007; Haslam & Ryan, 2008), participant gender did not play a role in candidate rankings. Study 2 therefore builds on Study 1 by suggesting that the political glass cliff is, at least to some extent, due to the candidate selection process. Women are more likely to be selected for difficult to win seats, even when past experiences and qualifications are controlled.
GENERAL DISCUSSION AND CONCLUSIONS
The two studies presented here provide evidence that glass cliff positions are not restricted to organizational settings; rather, they also extend to the political arena. Study 1 demonstrates the existence of a glass cliff within the Conservative party in the 2005 UK general election, such that Conservative women contested seats that were significantly less winnable than those in which Conservative men ran. Such a trend is problematic because the over-representation of women in difficult (if not impossible) to win seats potentially contributes to the notion that women are not “cut out” for politics. However, it is important to note that when the gender disparity in seat winnability was taken into account, there were no longer gender differences in voter support, suggesting that female candidates have the capacity to perform just as well as their male counterparts if given equal opportunities. This conclusion was echoed in the findings from the Labour party. Here, with affirmative action procedures in place to ensure women's access to winnable seats, female candidates performed no differently from men in this election.
The political glass cliff was also demonstrated under experimental conditions in Study 2. Here, the relationship between seat winnability and gender was examined directly under conditions where potential alternative explanations (such as gender differences in candidates' biographical details, their qualifications, and their past experience) were controlled for and party politics taken out of the equation. Under these hypothetical conditions, a woman emerged as the preferred candidate for a seat that was unwinnable, but when the seat was winnable, a male candidate was preferred. This finding is important because it establishes a causal relationship between seat unwinnability and the preferential selection of women. The results also question speculation that the preponderance of women in unwinnable seats can be explained solely by the fact that women themselves prefer, and actively choose, to run in seats that are less winnable—an explanation that female leaders sometimes endorse in other organizational contexts (e.g., because “they relish the challenge”; Ryan, Haslam, & Postmes, 2007). Moreover, because we controlled for candidate past experience across gender, our results also suggest that the preference for women in unwinnable seats is not due to objective gender differences in qualifications and experience. Interestingly, in contrast to previous organizational research (Haslam & Ryan, 2008), our results also cannot be attributed to perceived differences in leadership ability or suitability for the position. It may be that other (gendered) abilities, such as charisma or communication might be seen as more important for political candidates. Future research should explore this possibility. Finally, whereas the use of political science students may not be an exact proxy for the real party politics involved in selecting political candidates, the fact that the glass cliff was demonstrated using both archival and experimental methodologies strengthens our confidence that the political glass cliff is a genuine phenomenon that is due, at least in part, to the candidate selection procedure.
Although the two studies reported here provide evidence for the processes underlying women's lack of political success by demonstrating the existence of political glass cliffs, what is less certain is what causes the glass cliffs themselves. Research in the organizational realm suggests that, despite clear stereotypes equating leadership with men (e.g., Eagly & Karau, 2002), women may be preferentially selected for precarious or risky positions because they are seen to “have what it takes” (Ryan & Haslam, 2007; Ryan, Haslam, Hersby, & Bongiorno, 2009; Ryan, Haslam, & Postmes, 2007). Here, traditionally feminine traits such as being intuitive, creative, or understanding are regarded as useful in times of crisis. Extending this analysis into the political realm, it may be the case that despite stereotypes associating politics with masculinity (e.g., Huddy & Terkildsen, 1993a; Ross, 2002; Rosenwasser & Dean, 1989), women are seen as better political representatives in times of crisis. Although we controlled for overt differences in abilities in Study 2, this is not to say that people do not expect women to lead in a different way. Indeed, research suggests that voters expect women to be more able than men to manage “compassion” issues, such as dealing with poverty and improving education and health (e.g., Huddy & Terkildsen, 1993b; Leeper, 1991; Shapiro & Mahajan, 1986), and are more likely to vote for women when social welfare is an important issue (McDermott, 1998). In this way, it may be expected gender differences (stereotyping), rather than actual gender differences, that may play a role in the appointment of women to glass cliff positions.
Of course, an alternative explanation for the political glass cliff is that it is simply a form of sexism—women are selected for difficult-to-win seats precisely because they are difficult to win. Indeed, our finding that women's selection for such seats was not accompanied by raters' views that these women were more suitable or were better leaders suggests that there is more to the political glass cliff than women being seen to have what it takes. Although gender attitudes have been shown to be unrelated to the tendency for individuals to prefer women for precarious positions (Ashby et al., 2007), certainly women themselves believe that sexism and the desire to set women up for failure plays a large role in glass cliff positions (Ryan et al., 2007).
Along these lines, research in the organizational arena (Ryan & Haslam, 2007; Ryan et al., 2007) suggests that glass cliffs are likely to be multiply determined and that overt discrimination, party strategy, lack of networks, and the limited options for women wanting to enter politics are all likely to play a role. Thus, future research should investigate the roles that sexism and stereotyping play in the political glass cliff, and the way in which these interact with other processes to contribute to the selection of women for precarious political positions. Such research should also examine whether other minority groups, which are also underrepresented in the political arena (and that may share stereotypes with women, Fiske, Cuddy, Glick, & Xu, 2002), are similarly at risk of being placed on glass cliffs. Indeed, archival research is already under way to examine the seats that ethnic minorities are selected to contest in UK general elections (Kulich, Ryan, & Haslam, 2009).
Finally, it is important to note that even when women break through the political glass ceiling, they may still confront glass cliffs. Indeed Palmer and Simon (2006) reported that in the U.S. Congress, female incumbents enjoy significantly less electoral security than their male counterparts. Here, women in relatively safe seats (having won more than 55% of the vote in the previous election) were found to be significantly less likely to have their seat remain uncontested (11.2% versus 18.8% for safe male incumbents) and more likely to face contested primaries within their own party (33.1% versus 29.7% for men). Palmer and Simon (2006) also suggest that the picture for women in marginal seats is even more precarious, with fierce opposition likely for their seats.
The existence of glass cliffs in the political arena has important implications for women considering a career in politics. Merely selecting women to run as candidates does nothing to improve the underrepresentation of women in politics if these positions are unwinnable. Indeed, given the electoral failure that almost inevitably follows, casting women in the role of “sacrificial lamb” may simply reinforce the notion that women are not suitable for political office, while at the same time discouraging potential female candidates from entering the political arena. Taken together, the results presented here suggest that simply opening the door for female candidates pays mere lip-service to equality if women are not given reasonable opportunities to be elected and then contribute to politics, government, and public decision making. At the same time, on a more optimistic note, our findings suggest that active affirmative action can pave the road to the elimination of glass cliffs. Although these strategies have proven controversial, they thus present something of a Gordian solution to what might otherwise appear to be an intractable problem.