Gender quotas, candidate background and the election of women: A paradox of gender quotas in open-list proportional representation systems

Electoral Studies 36 (2014) 65e80

Contents lists available at ScienceDirect

Electoral Studies journal homepage: www.elsevier.com/locate/electstud

Gender quotas, candidate background and the election of women: A paradox of gender quotas in open-list proportional representation systems
recki*, Paula Kukołowicz Maciej A. Go ERC “Public Goods through Private Eyes” Project, Institute of Sociology, University of Warsaw, Obozna 7/41, 00-332 Warsaw, Poland

a r t i c l e i n f o

a b s t r a c t

Article history: Received 4 February 2013 Received in revised form 25 March 2014 Accepted 17 June 2014 Available online 26 June 2014

We study the effects of mandatory (legislated) gender quotas in Poland, a country utilising an open-list proportional representation electoral system. We use a unique data set comprising multiple characteristics of all candidates running in two consecutive elections to the lower chamber of the Polish parliament (the Sejm). The first of them (held in 2007) was the last pre-quota election and the second (held in 2011) the first post-quota one. We show that quotas have an inherently paradoxical nature: they cause a substantial increase in the number of female candidates but the increase is accompanied by a sharp decline in women's electoral performance. This regularity holds even if we account for multiple indicators of candidate background, including previous political experience. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Women and elections Gender quotas Open-list proportional representation systems Poland

Save for a few exceptions, women's descriptive (quantitative) underrepresentation in the democratically elected parliaments around the world is an indisputable fact. On average, only approximately 22 per cent of the world's elected representatives are women (Inter-Parliamentary Union, 2014). Poor descriptive representation may contribute to poor representation of women's substantive political interests (Chen, 2010; Holli and Wass, 2010). Thus, propositions of how to reduce gender disproportions in national legislatures have attracted much interest on part of both scholarly community and policy-makers. Gender quotas are one such mechanism. They are commonly considered to be a ‘fast track’ solution (Schwindt-Bayer, 2009: 5), a mechanism which, if correctly applied, can yield a reasonably quick increase in the numbers of women holding elected office. Since 1991, the year Argentina

introduced the world's first national quota law, a number of countries have thus decided to rely on this solution.1 An increasing number of studies have attempted to assess the effectiveness of legislated gender quotas in various political contexts. A type of a political setting where, we argue, quotas have not been given due attention is countries using variants of an open-list proportional representation (PR) electoral system. While there exist various descriptive and aggregate-level accounts of the impact of quotas on women's electoral success in such systems, especially in Brazil (Miguel, 2008), Indonesia (Hoodfar and Tajali, 2011) and Peru (Schmidt, 2003a, 2003b), there are virtually no systematic micro-level (i.e. candidate-level) studies on this topic. This gap calls for a new analysis as open-list PR constitutes an arguably critical case for the assessment of the effectiveness of quotas. It is so for at least three reasons. First, unlike closed-list PR or single-member district (SMD) systems, open-list PR does

* Corresponding author.
recki), paula. E-mail addresses: [email protected] (M.A. Go [email protected] (P. Kukołowicz).

1 For details, see the Quota Project website: http://www.quotaproject. org/.

1. Introduction

http://dx.doi.org/10.1016/j.electstud.2014.06.009 0261-3794/© 2014 Elsevier Ltd. All rights reserved.

66


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

not put major constraints on voters' ability to choose between male and female candidates, nor does it limit the effectiveness of their electoral choices. The scale of an existing gender bias is thus fully visible (Htun and Jones, 2002; Jones and Navia, 1999). Second, open-list PR stimulates high levels of intra-party competition and therefore candidates' background - especially political experience and the associated name recognition e constitutes an electoral advantage (Carey and Shugart, 1995). Given women's political underrepresentation inherited from the past, such an advantage is most often enjoyed by male candidates. Therefore, it might partly suppress the positive effects of quotas on female candidates' electoral success (Schwindt-Bayer, 2009: 16). Most importantly, however, being a woman might also prove advantageous to a candidate competing under open-list PR. As open-list PR creates strong incentives to ‘cultivate a personal vote’ (Carey and Shugart, 1995), minority candidates of various sorts e including women e might find it easier to distinguish themselves from the crowd and target specific groups of voters (Shugart, 1994; Valdini, 2012). In a political system dominated by men, voters might also be inclined to favour female candidates in order to promote equality (Rule, 1994; Kittilson, 2006; Valdini, 2012). An installation of gender quotas and the associated increase in the number of female candidates certainly limits women's ability to attract votes based on gender, especially as the ‘women-friendly’ niches in the electorate are certainly finite. We thus argue throughout this paper that under open-list PR quota laws have an inherently paradoxical nature: while they are introduced to enhance women's political representation, they actually tend to magnify the exactingness of an electoral contest for an ‘average’ female candidate. In this paper, we utilise a unique micro-level data set from Poland to simultaneously test the above-mentioned

characteristics of open list PR systems with gender quota laws. To this end, we focus on the 2007 and the 2011 election to the lower chamber of the parliament of Poland (the Sejm). Quotas were introduced before the latter election and thus we study candidates' electoral fortunes before and after the introduction of these regulations. Our point of departure is an aggregate-level observation that during the 2011 post-quota election a very large gap arose between the proportion of women among the candidates and the corresponding proportion for the elected MPs (see Fig. 1). Clearly, the introduction of quotas has resulted in a sharp increase in the number of female candidates but the increase in the number of female MPs has been rather small (note that before 2011 the proportion of women among elected MPs had been rising steadily even in the absence of quotas). While such a discrepancy between the proportion of women among candidates and the analogous proportion among the elected MPs is an entirely new phenomenon in the electoral history of post-communist Poland, it mirrors the regularities observed in other open-list PR countries that have introduced gender quotas. In this paper, we offer what we believe is a deeper look at this apparent paradox. We study the impact gender and candidate background exert on both the way candidates are ranked on party lists and on voter choice patterns. Our study shows that both the 2007 and the 2011 election saw a gender gap with respect to candidates' background, broadly conceived. Our evidence also suggests that this background was an important predictor of candidates' chances of being ranked high on their party lists and of subsequently attracting large numbers of electorate votes. More importantly, we also observe a paradoxical situation, an ‘average’ female candidate having had a much harder time contesting the 2011 (post-quota) election than the 2007 (pre-quota) one. This is suggestively reinforced by additional analyses showing

Fig. 1. Proportion of women among candidates and among members of the Polish Sejm (1991e2011).


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

that within a single electoral contest, especially during the 2007 election, the electoral performance of an individual female candidate is negatively correlated with the proportion of women featuring on the candidate's party (district-level) list. The paper proceeds as follows. The next section elaborates on the conceptual and empirical aspects of gender quotas under open-list PR. The third section describes the Polish case. It points to the 2007 and 2011 elections to the Sejm as an ideal ‘ground’ for a systematic test of the impact of gender and candidate background on electoral results before and after the quota introduction. The fourth section outlines the analyses to be performed and describes the data to be used. The fifth section presents initial descriptive statistics. The sixth section describes the results with respect to party ranking strategies. The seventh section presents the analyses with respect to voter choice patterns. The last section concludes with a discussion of the contributions our results make to the understanding of the effectiveness of gender quotas.

2. Gender quotas in open-list proportional representation systems: concepts and experiences Open-list PR systems offer voters an opportunity to support one or more candidates featuring on lists prepared by parties. The total number of votes cast for a district-level party list decides the number of seats a party obtains from that district. Seats won by a party are then allocated to its candidates solely on the basis of the numbers of votes cast for every candidate individually2 (see Carey and Shugart, 1995, for a description of variants of open-list PR). Systems of this sort constitute an arguably critical case for the analysis of the effects of gender quotas. The two main alternatives, closed-list PR and single-member district (SMD) systems, offer rather limited opportunities to understand how quotas really operate. In the former type of system, the effects of quotas on women's political representation are of purely mechanical nature and depend solely on whether quotas are accompanied by placement mandates, that is, a requirement that women are put in winnable positions on party lists (Baldez, 2004; Gray, 2003; Htun and Jones, 2002; Jones, 1996, 2004; Schmidt, 2008). Since voters are not allowed to disturb party ballots, the potential gender bias cannot be expressed by means of voter choice. In SMD systems, on the other hand, every party runs only one candidate in a district, thus potentially making voters' preferences with respect to a candidate gender interfere with partisanship. Neither of these constraints is present under open-list PR. Here, the effect of quotas on the election of women has a clear non-mechanical component, resulting from voters' ability to disturb party ballots. At the same time, unlike in SMD systems, a choice between male and female candidates does not interfere with party choice and thus the scale of an existing gender bias is fully visible (Htun and Jones, 2002; Jones and Navia, 1999).

2 Defining open-list PR this way, we clearly distinguish it from ‘flexible’ €uninger et al., 2012). list systems (see e.g. Bra

67

Crucial as the above-mentioned aspects might be, they do not exhaust the whole list of the nuances of open-list PR that should be taken into account by students of gender quota effects. We argue here that open-list PR systems may create radically divergent contexts for the performance of women candidates, depending on whether or not a gender quota law is in place. Unlike with closed-list PR, under open-list PR the relatively high intra-party competition results in the candidates facing strong incentives to develop personal reputations (Carey and Shugart, 1995). This is highly important given the large numbers of candidates contesting an average parliamentary seat. These numbers can easily exceed 10 or even 15. Indonesia (Ufen, 2010: 283) and Poland (see later on in this paper) are very telling examples here. Candidates' background should thus strongly affect the probability of an electoral success. Previous political experience and the associated name recognition are certainly an advantage. But, in some cases, the fact of being a woman might help as well. For not only does open-list PR stimulate personal vote seeking, but it also facilitates the diversification of the strategies by means of which candidates build their personal reputations. Like with any other identity politics argument, women can use to their advantage the fact of their being of a different gender than most of the candidates. As Valdini (2012: 742) argues, gender can operate in the same way as regional or religious background, facilitating access to some pre-formed groups of voters, in this case the existing women's groups. In other words, not only does being a woman distinguishes a candidate from the crowd, but it also potentially helps her create a personal bond with a specific niche in the electorate. In preferential voting systems, including open-list PR, it is this possession of distinguishing politically relevant traits that might give an edge to a candidate (Shugart, 1994). In this type of electoral system, a minority status may thus be converted into an electoral advantage. It could be a priceless asset for women participating in an intense inter-party and intra-party electoral competition. Yet another argument posits that under open-list PR (some) voters, being entirely unconstrained in their choice between male and female candidates, will choose to act in such a way as to promote equality and diversity. They will thus tend to favour female candidates (Kittilson, 2006). Under open-list PR, legislated gender quotas may, paradoxically, reduce women's (potential) advantage resulting from their minority status, especially if the quota size is large. After a quota is installed, a larger number of female candidates appear on party lists but, by definition, their minority status is much more questionable than it was before. A strategy of building a personal reputation based on gender will now be, on average, less efficient. Briefly speaking, in this type of electoral system quotas may create a context in which being a female candidate is no longer rare or unique. At the same time, the existing specific groups to which women have traditionally turned for electoral support (Valdini, 2012: 742) are undoubtedly finite and thus the competition in those niches should become more intense. To make our reasoning more explicit, let us consider the following (slightly stylised) example. Suppose that all members of a given electorate display a strong preference for a

68


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

candidate gender and that 30 per cent of those voters prefer a woman while 70 per cent prefer a man. In a situation where women constitute a small proportion of candidates e e.g. about 20 per cent e an ‘average’ female candidate will enjoy an electoral advantage over an ‘average’ male candidate. For the number of women running is small relative to the number of voters who are ready to support female candidates. Now suppose that the proportion of female candidates increases to 35 or 40 per cent e e.g. as a result of a quota law e while the distribution of the electorate's preferences with respect to a candidate gender remains the same as before or changes only incrementally. The situation is now radically different as a relatively large number of female candidates will compete for the support of the relatively small ‘women-friendly’ part of the electorate. One might be tempted to underestimate the relevance of such a reasoning, especially in the light of the fact that the available evidence from countries such as the United States (Sanbonmatsu, 2002: 23) or Poland (Kukołowicz, 2013: 230) indicates that the distributions of those electorates' ‘baseline’ preferences as to a candidate gender are not skewed in favour of either men or women. However, it should be remembered that the concept of a ‘baseline’ gender preference refers to a hypothetical situation of a choice between two ‘equally qualified’ candidates who differ solely with respect to gender. Such a concept might be somewhat misleading as gender tends to serve as a cue for a candidate's political competence and issue positions. While women are seen as somewhat more competent with regard to certain areas, broadly related to ‘protecting social security’, the perceived male candidates' supremacy in other domains, including ‘handling foreign affairs’, ‘dealing with the crime problem’, ‘big business’ and ‘economy’, is much more overwhelming (Sanbonmatsu, 2002: 23; Huddy and Terkildsen, 1993). Moreover, there is evidence to suggest that the stereotypically ‘masculine’ traits tend to be perceived as desirable for candidates seeking a nationallevel office, be it a legislative or an executive one (Huddy and Terkildsen, 1993). Stereotypes of this sort certainly limit female candidates' potential to attract votes which forces them to search for support in the abovementioned (limited) niches in the electorate. In addition, as Sanbonmatsu (2002: 23) argues, survey-based evidence on the distribution of electorates' ‘baseline’ preferences with respect to a candidate gender is most likely prone to ‘social desirability bias’ (Humm and Humm, 1944). In this particular case, it is expected that respondents will perceive as socially desirable an expression of a preference for a female candidate or simply a neutral answer. The scale of the ‘true’ prejudice against women might thus be underestimated (Sanbonmatsu, 2002: 23). All in all, we believe there are strong reasons to argue that the changes in the proportions of women among candidates (such as those initiated by a quota law introduction), especially under the conditions of fierce intra-party competition stimulated by open-list PR, might affect the electoral prospects of an ‘average’ female candidate. A resulting paradox of gender quotas e whereby this sort of regulations increase

the number of female candidates but decrease those candidates' success rate e appears to be a logical consequence of open-list PR. Existing evidence with respect to the implementation of gender quotas in countries using open-list PR e especially the evidence from Brazil, Indonesia and Peru e is very indicative. It points to the central role of personal vote seeking in such systems. Brazil is perhaps the world's most quoted example of an ‘atomised’ polity where candidates tend to compete on their own and where intra-party discipline is extraordinarily hard to maintain (Mainwaring,1991; Ames, 1995). Gender quotas, introduced in the 1990s, thus necessarily interacted with the above-mentioned tendencies. The most visible sign of it was the fact that, following the introduction of a (district-level) gender quota of 30 per cent, parties put forward many female candidates with renowned names. Most often, these were daughters or wives of local political leaders and their success rate was very high (Miguel, 2008: 205). However, parties used the nuances of the quota law to put forward lists with proportions of female candidates significantly below the district-level quota (Miguel, 2008: 202). In general, the probability of an ‘average’ female candidate winning a seat in the Brazilian Congress sharply declined after quotas were installed even though it was rising in the pre-quota period (Miguel, 2008: 204). In Indonesia, the first post-quota election, held in 2009, was contested by a large number of female celebrities, broadly conceived, and the probability of success was again very high for such candidates (Hoodfar and Tajali, 2011: 136). However, most parties applied the mandatory gender quota of 30 per cent at the national level only, thus disregarding the exact (district-level) meaning of those regulations. It is therefore likely that female candidates tended to occupy lists in districts a given party deemed unwinnable (Hoodfar and Tajali, 2011: 136). By contrast, the institutional and geographic aspects of the Peruvian polity (Schmidt, 2003a, 2003b) e including the fact that the country constitutes a single electoral district and the fact that the urban population of Lima constitutes a large portion of the entire electorate e facilitated the election of a considerable number of professional women and a resulting rapid increase in the number of female legislators. However, same as in Brazil, female candidates' success rate dropped following the quota introduction. Overall, open-list PR systems constitute a particularly interesting environment in which to test the effectiveness of gender quotas. Two directions seem to invite research efforts. First, the relative impact of gender and candidate background on candidates' electoral fortunes under openlist PR with quota regulations in place should be examined more systematically than it had been done before. So far, however, the only quantitative micro-level (candidatelevel) multivariate analysis comparing the effects of gender and candidate background on women's electoral success under quota law has been conducted with respect to the SMD tier of the Mexican system (Langston and Aparicio, 2011, 2012). That analysis indicates that candidate background, in particular political experience, is a better predictor of electoral success than is candidate gender. An extension of such an analysis to open-list PR seems to be a most worthwhile research effort and it should most


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

preferably distinguish between parties' actions e for instance list ranking strategies e and voter choice patterns (Matland, 1998 : 81) Second, this sort of systematic analysis should compare the relevant regularities before the quota introduction with the corresponding regularities observed when the quota law is already in place. This should facilitate a profound examination of our hypothesised paradox of quotas under open-list PR. Here again, the suggestive evidence from Brazil (Miguel, 2008) and Peru (Schmidt, 2003a), unsupported by analyses of multiple candidate characteristics, can only be considered circumstantial. In this paper, we analyse the Polish case in our effort to address all the above-mentioned concerns. 3. Quotas in open-list proportional representations systems: Poland Poland introduced its gender quota law in January of 2011. The adoption of quotas was a process initiated and pursued by the Congress of Women, a new broadly-based women's rights association. Although the campaign for the regulations ‘became the field of an ideological battle’ (Sledzinska-Simon and Bodnar, 2013: 163), the wider public remained largely ignorant of the new law: only 22 per cent of Poles surveyed one month before the first post-quota election correctly identified the legal provisions in question (Centrum Badania Opinii Spolecznej, 2011). The Polish quota law requires that the district-level representation of each gender on parties' electoral lists be not lower than 35 per cent.3 The first legislative election following the introduction of these new regulations was conducted on the 9th of October, 2011. In our view, that election and the last pre-quota election, held in 2007, constitute an unparallelled setting in which to test the effectiveness of gender quotas on female candidates' electoral success. It is so for at least six reasons. First, unlike in Brazil and Indonesia, Polish regulations had guaranteed perfect enforcement of quotas starting from the 2011 election. The State Electoral Commission is unqualifiedly obliged to refuse registration of a party's district-level list if the list does not abide by the quota. In the light of the fact that in many other cases it was inefficient enforcement mechanisms that was blamed for the delaying of the positive impact of quotas on the descriptive representation of women (Schwindt-Bayer, 2009: 16), the instantaneous enforcement of quotas in Poland is a clear advantage of our analysed case. In our study, we can leave aside the enforcement issue and focus our attention on the assessment of the effectiveness of quotas in fostering women's legislative representation. Second, the Polish quota law is entirely devoid of placement mandates, that is, it does not make any provisions with respect to the ranking of candidates on party lists.4 In an electoral system where parties can freely rank

3 An exception from this rule is lists with three candidates which must include at least one candidate of each gender (see Poland's entry at the Quota Project website). 4 Following the 2011 election, a new quota law proposal, including placement mandates, was submitted to the Sejm. At the time of this paper going to print, it has not, however, been decided whether placement mandates will be introduced.

69

their candidates in every district but voters are subsequently allowed to disturb party ballot (Kunovich, 2012), it is possible to study unrestrictedly both parties' and voters' behaviour affecting candidates' electoral fortunes. Third, in Poland the percentage of female members of the Sejm remains relatively low, slightly above 20 per cent in recent years (Karp and Banducci, 2008: 108; Dubrow, 2010), and the mandatory gender quotas proved to be only moderately successful in enhancing women's legislative representation. As Fig. 1 above indicates, the proportions of women among candidates to the Sejm have been steadily increasing since the first post-communist-era parliamentary election, held in 1991. The proportion of women among those winning seats has been increasing as well, reflecting by and large the share of female candidates. The last pre-quota election, held in 2007, saw 23 per cent of women among the candidates and 20.4 per cent of women among the elected MPs. Quota regulations brought a major change to this pattern in that they had a strong effect on the proportion of women running: 43.5 per cent of the candidates running in the 2011 election were female. At the same time, the proportion of women among the candidates winning seats rose to the mere 23.9 per cent, which indicates a rather moderate success of the new quota law (note once again that before 2011 we saw a gradual increase in the proportion of women among elected MPs despite no quotas being installed). As we already mentioned above, for the first time in the post-communist era we observe an enormous, and thus far unexplained, gap between the proportion of women among the candidates and the corresponding proportion among the elected MPs. Fourth, access to the Sejm is still relatively open to challengers. The proportion of re-elected members is relatively low, having accounted for approximately 60 per cent and approximately 66 per cent respectively in 2007 and 2011. This suggests that the impact of incumbency e accounting for the largest part of variation in candidates' electoral success in established democracies (SchwindtBayer, 2005; McElroy and Marsh, 2010) e may be somewhat weaker in Poland. This might have made room for the effects of other electorally relevant background. Fifth, an important characteristic of the Polish case is a large within-election-year variation in the proportions of women featuring on parties' electoral lists, especially during the 2007 (pre-quota) election. If only the parties whose candidates entered the Sejm are considered, that election saw lists with the proportions of women ranging from as low as 3.8 to as high as 50 per cent. The variation for the 2011 election was way smaller, even if still considerable (see appendix figure A1 for histograms of the proportion of women on a party list in both election years). No wonder, the quota law set a clear lower bound for the proportion of women on those lists. We argue here that the withinelection-year variation in the proportion of women on electoral lists creates an opportunity for an additional test of our hypothesised paradox of gender quotas. For what we essentially propose is that an ‘average’ female candidate's electoral success decreases as the proportion of women contesting elected office increases. On the one hand, such an increase in the proportion of women running as candidates does not necessarily require an installation of quotas.

70


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

On the other, one might criticise our proposition and argue that it is negative views on the introduction of the quota law itself that causes many voters to withdraw their support for female candidates.5 From such an alternative viewpoint, the quota law is hypothesised to operate as a ‘turnoff’, diminishing women's chances for electoral success. This could complicate highly the interpretation of our proposition that women's success rate should dwindle following the introduction of quotas. However, if we are able to demonstrate an analogous within-election-year effect of the varying proportions of women on electoral lists e especially in 2007 when mandatory quotas were not even discussed in public domain e then our findings related directly to quotas will be suggestively reinforced. Finally, we believe some further characteristics of the recent political and electoral landscape in Poland contribute to the emergence of favourable conditions for our research effort. Most importantly, the hypothesised paradox of gender quotas assumes that, all else being equal, female candidates' success rate decreases following the quota introduction. As we see above (Fig. 1), such a regularity would contradict the, slow but persistent, trend toward an increase in women's representation, observed since early the 1990s. We thus seem to be on the safe side here as what we propose is intuitively unexpected based on the long-term trends and, as such, its roots should be searched for solely in the quota introduction. Moreover, with respect to electoral support for the main parties the 2011 election was marked by continuity rather than by change, albeit an entirely new party entered the Sejm (see next section for details). Claims that the political landscape has stabilised are thus certainly justified, especially as the government coalition survived the election e a scenario that happened for the first time in the entire postcommunist transition period. This makes the results of our study more credible. Perhaps the only potentially worrying aspect is a significant turnout decline: from 53.8 per cent in 2007 to 48.9 per cent in 2011. In the United States, for instance, low turnout has been shown to be associated with a decrease in the Democratic, that is, presumably socially liberal, vote shares (see for example Gomez et al., 2007). However, as studies on European countries show virtually no effect of turnout decline on left-of-centre vote (Bernhagen and Marsh, 2007; van der

5 The question of whether or not the quota law enjoyed support on part of the wider public is a potentially controversial issue. An opinion poll conducted in September of 2010 (Centrum Badania Opinii Spolecznej, 2010) e that is, at the time when the parliament was proceeding on the quota law e yielded rather confusing results. While around a half of the respondents expressed their support for the idea of greater participation of women in political life, just around 10 per cent pointed to quotas as the mechanism by means of which this goal should be achieved. On the other hand, 56 per cent of those same respondents indicated that they ‘strongly supported’ or ‘rather supported’ the particular quota law proposal proceeded on in the parliament. We believe this suggests a pervasive lack of awareness and confusion as to the exact nature of the quota law, a thesis supported by the mentioned fact that one month before the 2011 election only 22 per cent of the respondents surveyed knew exactly the content of the legal provisions in question (Centrum Badania Opinii Spolecznej, 2011).

Eijk and van Egmond, 2007), we believe it will not contaminate the results of our study.

4. The data and the concept of analysis In this paper, we utilise a unique data set, containing information about all candidates having competed for seats in the Sejm in the 2007 and 2011 elections.6 The 460 Sejm seats were contested by 6187 and 7035 candidates in 2007 and 2011 respectively; this gave 13.5 candidates per seat in the former election and 15.3 candidates per seat in the latter one. The large number of candidates contesting Polish elections is obviously advantageous: the results we present later in this paper are based on a very respectable number of observations. On the other hand, the analysis poses certain difficulties. Most importantly, there has thus far been little research on candidate selection processes in Poland and the bulk of the available evidence focuses on the decade of the 1990s. This evidence echoes findings from established democracies (see for example Norris and Lovenduski, 1993; Wessels, 1997) by pointing to the emergence of career politicians as a key to understanding candidates' ranking on party lists (Kunovich, 2003) and their overall electoral success (Shabad and Slomczynski, 2002). We build on these findings and include in our analysis multiple indicators of candidates' political experience. Incumbent status, so frequently emphasised by the existing literature on gender and voter choice, is one such variable. We have also established which candidates ran (unsuccessfully) in the Sejm election directly preceding one of the two elections analysed. Furthermore, our data set contains two dummy variables relating to candidates' experience in local politics. We distinguish between experience as a local, sub-regional or regional councillor and executive experience (that is, as a commune leader, town mayor or regional marshal).7 We also include three dummy variables that relate to the very rare statuses as a party leader, senator or minister. Also, the literature on candidate selection in Poland (Siemienska, 2000: 44) mentions economic and social capital as the assets Polish parties value in potential candidates. We thus include dummies for business owners and for what we call ‘publicly visible professions’. The latter category comprises doctors, lawyers, journalists, civil servants and teachers (including university lecturers). We claim that the representatives of those professions have access to broader social networks. In most cases, those professionals are also believed to adhere to a certain ethics of public service. Both these traits might constitute valuable (social) capital.

6 Raw information we used while compiling our data set, including electoral results and some basic candidate characteristics, was made available to us, upon our request, by the State Electoral Commission (http://pkw.gov.pl/). We supplemented it with additional analyses of careers of the candidates. 7 According to Polish regulations, a local official automatically loses her position in local government if elected to the Sejm. Simultaneously holding a seat in the Sejm and a seat in the European Parliament is forbidden as well. Nonetheless, while no Polish members of the European Parliament contest elections to the Sejm, large numbers of local officials do so.


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

Another variable we include in our models is a dummy for celebrities, including distinguished athletes, artists and media personalities. Due to name recognition, celebrities might be put forward as candidates with the hope to attract many electorate votes (West and Orman, 2002). In addition, we control for candidate age and its square as we expect that middle-aged candidates will, on average, do better at the polls than will their young and elderly counterparts. Although we believe our data set contains all the feasibly obtainable information, we do not claim it is exhaustive. For instance, party service experience cannot be adequately described for such a large number of candidates as in our study. This is because a deeper inspection of such experience is not feasible given frequent transformations that Polish party system had undergone (Markowski, 2006, 2008; McMenamin and Gwiazda, 2011). Detailed histories of regional and local party organisations are thus not readily available for an analysis. We have experienced the same difficulty with respect to candidates' participation in various voluntary associations as such activities are nowhere reported. We nonetheless believe that, despite these limitations, our data set contains enough information to reveal interesting regularities.8 Our research effort encompasses three steps, undertaken with respect to both the 2007 and the 2011 election. First, we present descriptive statistics regarding gender disparities in candidates' background. Then, we proceed to an analysis of the impact of gender and background on a candidate list ranking. We define a candidate ranking as ‘viable’ if the candidate is ranked among the top N candidates where, for parties winning seats in a given district, N equals twice the number of seats a party wins in that district. For parties not winning seats in a district we classify as ‘viable’ the top two positions on a list.9 We estimate two models for ranking. In the first one, we just include gender, its interactions with party dummies, age, and age squared. In this first step, we are thus able to study the effect of gender on a candidate list ranking separately for every party but uncontrolled for candidate background. In order to simplify the analyses, we include dummy variables for selected parties only. In 2007, we include dummies for all the parties that have reached the electoral threshold of 5 per cent and thus won seats in the Sejm: the Civic Platform, the Law and Justice, the Democratic Left Alliance/Left and

8 A factor potentially affecting candidates' success in various electoral systems, but not included in our analyses, is campaign spending (see e.g., Benoit and Marsh, 2008). The reason for not taking spending into account in our study is the fact that individual candidate spending is not allowed in Poland. Campaigns are funded from central party budgets and only the aggregate figures for parties are reported. As we include party dummies in our analysis, all party-level variation, including inter-party variation in campaign spending, is automatically controlled for. As a result, in the absence of figures regarding individual candidate spending, the available (party-level) spending data are of no use for our purposes. 9 We argue that, in order to study gender disparities with respect to list ranking, the ‘viable’ positions must be defined in such a way as to guarantee that gender parity is theoretically achievable. Hence, there should be an even number of ‘viable’ positions on every list, two such positions being a minimum. We are aware that, like any other classification, this one too introduces some arbitrariness. On the other hand, our efforts to validate the analyses presented later in this paper, referring to other classification schemes, have yielded essentially the same results.

71

Democrats and the Polish People's Party. In addition, we include a dummy for the Women's Party. That party was supported by just 0.3 per cent of voters. Nonetheless, as only three out of 109 candidates running on its behalf were men, we worry that such a disproportion would distort the results if a dummy for the party was not included in our model. In 2011, the Women's Party did not contest the Sejm election. All the four parties that entered the 2007 Sejm won seats also in 2011. Apart from the dummy variables for those parties, we also include dummies for the Palikot's Movement10 and Poland Comes First, that is, splinters from respectively the Civic Platform and the Law and Justice. The former party did enter the Sejm while the latter did not. In the second step, we add candidate background variables to the just described basic model. Proceeding this way, we aim to establish what proportion of the gender effect on a candidate ranking can be ascribed to genderrelated disproportions with respect to background variables rather than to a direct effect of gender. Finally, we take quite similar steps to analyse the impact of gender and background on the number of votes cast for a candidate. More precisely, we first estimate a model including gender as well as party dummies and interactions between the former and the latter, controlling for age and age squared. In order to test the above-mentioned within-election-year effect of the proportions of women on electoral lists on female candidates' electoral performance, we also include in the model the proportion of women on an electoral list and its interaction with candidate gender. Then, in our intermediate model, we add a dummy distinguishing between ‘viable’ and ‘unviable’ positions on a list. Finally, our full model includes also the indicators of candidate background. Same as in the case of our analyses of candidates' list ranking, we are then able to establish if the effect of gender on the number of votes cast for a candidate ‘vanishes’ after ‘viability’ and background are added to the model.

5. Candidate background and women's electoral success: descriptive statistics Table 1 presents basic descriptive statistics on candidates who contested the 2007 and 2011 Sejm elections.11 Most importantly, as the proportions of those successfully elected among men and women indicate, female candidates' success rate declined sharply after the quota introduction while the analogous figure for men rose. This echoes findings from Brazil (Miguel, 2008) and Peru (Schmidt, 2003a). The discrepancy between male and female candidates' probabilities of being placed on ‘viable’ list positions did increase substantially as well. Clearly then, an ‘average’ female candidate found it more difficult to compete under the quota law than before its

10 The party was named after its founder and leader, Janusz Palikot, a former Civic Platform MP. In 2013, the party's name was changed to Your Movement. 11 Our inspection of analogous regularities within parties did not reveal any major deviations from the overall picture. For the sake of simplicity, we thus refrain from analysing within-party distributions here.


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

72

Table 1 Candidate background by election year and gender. In%

Elected Candidates at ‘viable’ positions Incumbents Ministers Senators Party leaders Councillors Local executives Candidates in preceding election Visible professions Business owners Celebrities

2007

2011

Among men (N ¼ 4759)

Among women (N ¼ 1428)

Among men (N ¼ 3972)

Among women (N ¼ 3063)

7.69 20.42 6.81 0.46 0.19 0.17 21.96 2.54 32.19 17.57 8.01 0.15

6.58 17.51 5.88 0.35 0.14 0.07 14.22 0.77 24.44 24.72 3.29 0.21

8.81 22.33 8.26 0.38 0.28 0.13 17.85 1.89 22.63 16.21 8.53 0.38

3.59 12.18 2.87 0.23 0.07 0.00 10.64 0.82 9.79 23.34 3.95 0.26

introduction. Furthermore, female candidates in both elections were less politically experienced than were their male counterparts. In 2011, an ‘average’ female candidate's level of political experience dropped even further as the percentage of female incumbents and past candidates decreased comparing to the 2007 election. This should have been expected, given the sudden (legally enforced) increase in the number of female candidates. Multivariate analyses should reveal if it is the decline in an ‘average’ female candidate's experience level, or our hypothesised paradox of quotas, that stands behind women's low success rate in 2011. Finally, while male candidates are relatively frequently recruited from among business owners, women tend to be recruited from among the representatives of visible professions. Despite the sudden huge increase in the overall number of female candidates in 2011, the proportion of those in visible professions hardly dropped. The recruitment of professional women thus seems to be one of the Polish parties' response to quotas. Again, multivariate analyses should uncover the potential link between this strategy and female candidates' electoral success. 6. Candidate ranking: multivariate analyses To model the probability of a candidate being ranked at a ‘viable’ position on her party list, we apply conditional (fixed-effects) logistic regression. This estimator is most appropriate for our purposes as it utilises only withinparty-list variation. As this is a fixed-effects model, only variables that differ within party lists can enter the equation in the form of a constitutive term. The variables constant within lists, such as party dummies, enter the equation only in interaction with other variables. In our models, gender is such a variable as we are interested in, among others, how the effect of gender on list ranking varies by party. In Table 2, we present marginal effects of gender on the probability of a candidate being placed at a ‘viable’ position (see also appendix table A1 for regression coefficients and standard errors). As figures in Table 2 indicate, in the 2007 election the effects of being a woman on the probability of a candidate being ranked at a ‘viable’ position on a party list are pervasively negative e regardless of whether or not candidate background was taken into account. The sole exception is the largest party e

the Civic Platform. At the same time, the effects of gender on a candidate ranking are all statistically insignificant in both the basic and the full model. Surprisingly, women running on behalf of the Law and Justice and the Democratic Left Alliance/Left and Democrats tend to lose somewhat after the indicators of candidate background are included in the model. This means female candidates from those parties who did not have relevant background found it slightly more difficult to get ranked at a ‘viable’ position than did their male counterparts. These within-electionyear differences are, however, of negligible size. A major shift from this pattern takes place in 2011. For almost all the parties for which a comparison with the 2007 election is possible, the effects of being a woman on the probability of a candidate being ranked as ‘viable’ move in the negative direction. Moreover, in the basic model those effects are substantial, ranging from
25.9 per cent to
11.7 per cent, and statistically significant. The same is true of the candidates from the two new parties e the Palikot's Movement and Poland Comes First. It can thus be said that in 2011 women were universally placed at low list positions, a radical change in comparison to the more nuanced 2007 picture. After including candidate background variables, the effects of gender on a candidate ranking unsurprisingly tend toward zero. This confirms the hypothesis that the lack of relevant background is a major reason behind women's relatively low list rankings. This concerns the 2011 election in particular, as an ‘average’ female candidate's level of political experience dropped significantly in comparison to the pre-quota period. The effects of relevant background on candidates' ranking on party lists are presented in Table 3. As the magnitude of the effects does not differ substantially between male and female candidates, we analyse the general picture only, that is, the average effects for the entire population of candidates. Relevant background boosts nearly universally candidates' chances of being placed at a viable position albeit the strength and statistical significance of these effects vary considerably depending on the type of background and, less substantially, between the two election years. The privilege of being ranked high is obviously often enjoyed by incumbents, but also by councillors and those with the rare statuses as senators, party leaders or ministers. Business owners are ranked high only in 2011,


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

73

Table 2 The effects of gender on the predicted probability of a candidate being ranked at a ‘viable’ position on a party list. 2007

Civic Platform Law and Justice Democratic Left Alliance/Left and Democrats Polish People's Party Women's Party

2011

Basic model

Full model

Basic model

Full model

3.8% [
4.1%; 11.7%]
2.2% [
10.2%; 5.8%]
9.1% [
20.7%; 2.4%]
11.5% [
24.9%; 2.0%]
22.0% [
64.2%; 20.1%]

4.1% [
1.7%; 9.9%]
4.5% [
12.8%; 3.9%]
12.3% [
27.6%; 3.0%]
1.7% [
13.5%; 10.1%]
19.2% [
70.7%; 32.3%]


11.7% [
16.0%;
18.9% [
23.1%;
14.6% [
23.2%;
25.9% [
34.2%;

3.0% [
1.7%; 7.7%]
5.5% [
11.1%; 0.0%]
1.1% [
11.5%; 9.3%]
13.7% [
27.0%;
0.5%]

Palikot's Movement Poland Comes First
4.9% [
11.7%; 2.0%]

Other Parties


1.9% [
9.0%; 5.2%]


7.3%]
14.5%]
6.0%]
17.7%]


15.0% [
22.3%;
7.7%]
22.8% [
31.6%;
14.0%]
20.0% [
24.9%;
15.1%]


14.5% [
22.9%;
6.1%]
20.0% [
31.7%;
8.4%]
19.3% [
25.2%;
13.3%]

Note: Main entries are changes in predicted probabilities of a candidate being ranked at a ‘viable’ position on a party list when the variable ‘Woman’ changes from 0 to 1, calculated for respective parties. Positive values of the effects mean women's advantage over men while negative values mean a vice versa situation. The numbers in square brackets are 95 per cent confidence intervals of the main predictions. The effect of candidate gender is presented for the basic model and for the full model in which candidate background is taken into account. Fixed effect is assumed zero.

perhaps a result that should be explored more profoundly within parties (which goes beyond the scope of this paper), while the effect for visible professions is moderately positive for both elections. Celebrities also tend to be ranked high but the effect is insignificant in 2011. Overall, the effects of relevant background e especially previous political experience - on the probability of a candidate being placed at a ‘viable’ list position are relatively strong. Most importantly, the effects of experience are stronger relative to the effects of gender. Overall, our analysis of parties' ranking strategies leads to some noteworthy conclusions. Most importantly, the crucial role of political experience in determining a candidate ranking is evident in both the 2007 and the 2011 election. Incumbents, ministers or even local councillors are way more likely to occupy viable list positions than are those lacking such background. When background is controlled for, the effects of gender on candidate ranking tend to be less substantial than a basic model, focussing solely on the effect of gender, would suggest. Last but not

least the introduction of quotas seems to have worsened an ‘average’ female candidate's position on her party's districtlevel list during the 2011 election. 7. Voter choice patterns: multivariate analyses Our analyses of ranking strategies employed by parties suggest that a major change has taken place following the quota introduction. The evidence presented in the previous section indicates that after the quota adoption an ‘average’ female candidate was less likely to be placed at a viable list position than she was in the pre-quota period. However, this evidence is hardly sufficient to demonstrate our hypothesised paradox of quotas under open-list PR. In such systems, ranking can influence candidates' electoral success only indirectly. It is thus necessary to show how voter choice is affected by the introduction of party-list gender quotas. To this end, we estimate a model in which the number of votes cast for a candidate is regressed on candidate gender and relevant background. As the response variable is a count, we

Table 3 The effects of relevant background on the predicted probability of a candidate being ranked at a ‘viable’ position on a party list. 2007

Incumbent Councillor Candidate in preceding election Minister Party leader Senator Local executive Visible profession Business owner Celebrity

2011

Main effects

95% CIs

Main effects

95% CIs

30.1% 25.5% 20.7% 44.9% 44.8% 52.7% 27.3% 5.6% 1.8% 26.9%

[21.7%; 38.6%] [21.5%; 29.5%] [16.1%; 25.3%] [38.0%; 51.8%] [38.0%; 51.7%] [46.1%; 59.3%] [18.6%; 36.0%] [1.0%; 10.3%] [
4.8%; 8.5%] [7.7%; 46.2%]

35.6% 20.9% 18.7% 51.5% 46.5% 53.7% 11.3% 5.0% 8.9% 24.9%

[28.2%; 43.1%] [15.9%; 25.9%] [14.8%; 22.7%] [45.4%; 57.7%] [42.9%; 50.1%] [45.9%; 61.5%] [
1.7%; 24.4%] [1.6%; 8.4%] [3.2%; 14.6%] [
3.3%; 53.1%]

Note: Main entries are changes in predicted probabilities of a candidate being ranked at a ‘viable’ position on a party list when a given predictor changes from 0 to 1. The numbers in brackets are 95 per cent confidence intervals. Fixed effect is assumed zero.


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

74

Table 4 The effects of gender on the number of votes cast for a candidate. 2007

Civic Platform Law and Justice Democratic Left Alliance Polish People's Party Women's Party

2011

Basic model

Intermediate model

Full model

Basic model

Intermediate model

Full model


583 (
7.9%) [
2617; 1451] 81 (1.4%) [
1600; 1762]
105 (
4.4%) [
801; 591]
423 (
27.1%) [
725;
120] 219 (52.9%) [105; 333]


124 (
1.7%) [
1850; 1601] 718 (12.7%) [
272; 1708] 573 (24.0%) [
52; 1197]
86 (
5.5%) [
330; 157] 251 (60.5%) [77; 424]

1252 (17.0%) [
50; 2554] 1333 (23.6%) [
374; 3040] 390 (16.4%) [
94; 875] 89 (5.7%) [
228; 406] 274 (66.2%) [100; 448]


2473 (
40.2%) [
4332;
613]
2795 (
59.6%) [
4295;
1295]
644 (
49.6%) [
1006;
281]
938 (
71.6%) [
1322;
554]


1879 (
30.5%) [
3186;
572]
1764 (
37.6%) [
2827;
701]
336 (
25.9%) [
522;
150]
597 (
45.6%) [
891;
303]


225 (
3.7%) [
2057; 1606]
1250 (
26.7%) [
2482;
18]
244 (
18.8%) [
595; 106]
866 (
66.1%) [
1868; 135]


594 (
35.5%) [
1049;
140]
120 (
29.2%) [
228;
12]
103 (
58.3%) [
163;
42]

183 (10.9%) [12; 354] 62 (15.2%) [7; 117]
24 (
13.5%) [
60; 12]

75 (4.5%) [
152; 301] 61 (14.9%) [4; 118]
28 (
16.2%) [
56;
0.5]

Palikot's Movement Poland Comes First Other Parties

17 (6.4%) [
26; 60]

56 (21.0%) [16; 96]

72 (27.1%) [40; 104]

Note: Main entries are changes in the predicted number of votes cast for a candidate when a variable ‘Woman’ changes from 0 to 1. Positive values of the effects mean women's advantage over men while negative values mean a vice versa situation. The numbers in round brackets, expressed in percentages, are the main figures divided by the average number of votes cast for a candidate from a respective party. The figures in square brackets are 95 per cent confidence intervals of the main predictions.

model it using negative binomial regression.12 Two additional controls are included in our voter choice equations. To account for inter-district variation in voting population sizes and turnout rates, we control for the total number of votes cast in an electoral district. Also, the average number of votes cast for a candidate from a given party running in a given district certainly depends on the total number of candidates on the list of that party in that particular district. We thus control for the latter too. In Table 4, we present the effects of gender on the number of votes cast for a candidate (see appendix table A2 for regression coefficients and standard errors). The results are very suggestive and seem to confirm the paradoxical nature of gender quotas under open-list PR. Looking at the basic models e including gender, age, age squared, party dummies and the proportion of women candidates on a list (with the interactions of the last two with gender) e we notice a substantial difference between the two elections. In 2007, women obtained, on average, fewer votes than did men (apart from the candidates from the Women's Party and from the Law and Justice). Nonetheless, with the exception of the candidates from the Polish People's Party and the Women's Party, the between-gender differences were rather small e below 10 per cent of the number of votes cast for a party's ‘average’ candidate e and statistically insignificant. Thus, it can be said that despite women's relatively low levels of political experience and the fact of their list ranking being slightly lower than that of men, they did not suffer a major

12 Modelling voter choice relying on ordinary least squares (OLS) regression would certainly yield more easily interpretable results. On the other hand, OLS is a technique that does not account for the fact that, like with any other count variable, the number of votes cast for a candidate has a lower bound at zero. OLS estimates might produce negative predicted values. Unfortunately, this is the case in our study. Depending on which model is considered, OLS produces from 12.9 to 33.0 per cent of predicted values below zero. Thus, in order to avoid such an undesirable situation, we are forced to rely on negative binomial regression estimates.

disadvantage in terms of popular vote. The fact that during the pre-quota period the proportions of women among elected legislators reflected, by and large, the corresponding proportions among candidates (see above) is now more comprehensible. In 2011, the situation changes nearly drastically. An ‘average’ female candidate running on behalf of virtually any party is severely disadvantaged relative to her male counterpart. For the two largest parties e the Civic Platform and the Law and Justice e the between-gender gap approaches or even exceeds 2500 votes. For the Law and Justice, such a gap constitutes nearly 60 per cent of the mean number of votes cast for the party's candidate. These strong negative effects of being a woman on the number of votes obtained by a candidate explain why the large proportion of women among all the candidates contesting the 2011 election (43.5 per cent) did not translate into a comparable proportion of women among those elected to the Sejm. For both elections, after the basic model is extended by adding the ranking ‘viability’ dummy (see intermediate model) and then indicators of candidate background (see full model) the effects of gender on the number of votes cast for a candidate change substantially. In the 2007 full model, the effects of being a woman on the number of votes cast for a candidate are positive for all major parties, albeit statistically insignificant for virtually all of them. Looking at the effects for the two largest parties, we see that an ‘average’ female candidate running on behalf of any of these two enjoyed an advantage of well over 1000 votes over her male counterpart. This supports our prediction that in the prequota period female candidates found themselves in a very comfortable situation. By contrast, for the 2011 postquota election, most of the effects of gender remain negative even after controlling for candidate ranking and background. The exceptions here are the two new parties: Palikot's Movement and Poland Comes First. For all the four major parties that contested both elections, female candidates' situation worsened substantially in 2011 as compared to the pre-quota period. For instance, in 2007 female


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

75

Table 5 The effects of list position and relevant background on the predicted number of votes cast for a candidate. 2007

‘Viable’ position Incumbent Councillor Candidate in preceding election Minister Party leader Senator Local executive Visible profession Business owner Celebrity

2011

Votes

% of mean

95% CIs

Votes

% Of mean

95% CIs

3052 4213 1009 397 7099 11,998 11,612 1058 292 14 3542

(117) (161) (39) (15) (272) (460) (445) (41) (11) (1) (136)

[2314; 3790] [3320; 5106] [795; 1223] [221; 573] [3164; 11,033] [
3536; 27,532] [6186; 17,037] [679; 1437] [144; 440] [
217; 245] [
294; 7377]

3369 2813 944 282 4353 17,993 2180 1071 204 287 3170

(165) (138) (46) (14) (213) (881) (107) (52) (10) (14) (155)

[2052; 4686] [1425; 4200] [506; 1382] [33; 530] [1687; 7020] [
11,689; 47,675] [729; 3632] [512; 1630] [31; 377] [61; 513] [589; 5750]

Note: Main entries are changes in the predicted number of votes cast for a candidate when a given predictor changes from 0 to 1. The numbers in round brackets, expressed in percentages, are main effects divided by the average number of votes cast for a candidate. The numbers in square brackets are 95 per cent confidence intervals of the main predictions.

candidates running on behalf of the Democratic Left Alliance were, other things being equal, preferred to that party's male candidates. During the post-quota 2011 election this electoral advantage vanishes completely. The change is perhaps even more evident in the case of the socially conservative parties e the Law and Justice and the Polish People's Party. Perhaps female candidates running on behalf of such parties encounter severe difficulties finding enough niches in the electorate where votes based on gender can be sought. Our results suggest that such a problem of exhausted niches is experienced by candidates from virtually all the major parties and that perhaps after the quota introduction a greater number of women compete for the support of limited groups of voters. It thus seems clear that the lack of relevant background, including political experience, is not the only factor behind women's low electoral success following the introduction of quotas. Quotas themselves seem to have indeed paradoxically lowered an ‘average’ female candidate's baseline probability of winning a parliamentary seat. These findings are suggestively reinforced by our analyses of the within-election-year effects of the proportion of women on an electoral list on female candidates' electoral fortunes (see Table 6). The pattern is quite clear, especially for the 2007 pre-quota election. When the proportion of women on an electoral list is held at a low value e Table 6 The proportion of women on a party list and the effect of gender on the predicted number of votes cast for a candidate. Proportion of women

Effect of gender (2007)

Effect of gender (2011)

5th percentile

937 (35.9%) [342; 1532] 533 (20.4%) [91; 974]
47 (
1.8%) [
439; 345]


46 (
2.3%) [
416; 324]
108 (
5.3%) [
437; 220]
199 (
9.8%) [
556; 158]

Median 95th percentile

Note: Main entries are changes in the predicted number of votes cast for a candidate when the variable ‘Woman’ changes from 0 to 1. Positive values of the effects mean women's advantage over men while negative values mean a vice versa situation. The numbers in round brackets, expressed in percentages, are the main figures divided by the average number of votes cast for a candidate. The figures in square brackets are 95 per cent confidence intervals of the main predictions.

in this case, at the 5th percentile (8.3 per cent of women among all candidates) e the number of votes cast for an ‘average’ female candidate exceeds the number of votes cast for her male counterpart by nearly 1000. This is a large difference, constituting about 36 per cent of the mean number of votes obtained by a candidate running in that election. The advantage enjoyed by female candidates drops by nearly a half (over 400 votes) when the proportion of women on a list is held at the median value (21.4 per cent). Finally, gender does not seem to affect strongly the electoral results if the proportion of women rises to the value at the 95th percentile (45.0 per cent).13 In 2011, the corresponding effects are substantially weaker and statistically insignificant. This may be result partly of a decrease in variation in the proportion of women among candidates imposed by the quota law. Yet, the general pattern is largely the same, an ‘average’ female candidate enjoying less popular support as the proportion of women on an electoral list increases. These results strengthen our argument concerning quotas and their potentially paradoxical effects. At the same time, they point to a more general source of the paradox of quotas. For our results suggest strongly that any sudden and substantial increase in the proportions of women among candidates, including the increase resulting from the introduction of quotas, might lead to a decline of female candidates' success rate. The effects of relevant background (see Table 5), including political experience, are interesting as well. They vary depending on the particular indicator of candidate background. In general, however, political experience tends to be very helpful in attracting popular vote. Obviously, incumbency effects and the effects of rare statuses (party leader, minister etc.) are way stronger than the effects of experience in local politics. The latter are nonetheless not negligible either. As regards non-political background, only

13 We checked whether the results concerning the 2007 election are robust to the exclusion of the Women's Party, on behalf of which a disproportionally high percentage of women were running, ranging, at the district level, from 87 per cent to 100 per cent. Analyses performed on the trimmed data set, with the Women's Party excluded, yield essentially the same results as those presented in our study.

76


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

celebrity status seems to work strongly in favour of a candidate. This suggests that name recognition is the most likely key to understanding of the effects of relevant background on candidates' electoral success under open-list PR. The effects for business owners and visible professions are rather weak. Such small effects for the latter suggest that in Poland, unlike in Peru (Schmidt, 2003b), the strategy of nominating professional women did not have a strong positive impact on women's overall electoral success. Finally, the strong impact of list ranking on the number of votes cast for a candidate deserves mention as well. The effect of list position ‘viability’ is comparable to that of incumbency. It thus appears that relevant background affects candidates' electoral fortunes in two ways. First, it has an independent direct effect on the number of votes a candidate obtains. In addition, such experience also helps a candidate obtain a ‘viable’ position on a party's list. Thus, indirectly, it affects a candidate electoral success as the privilege of being ranked high on a list results in large benefits in terms of the numbers of votes cast for a candidate. That said, a note of caution is necessary here. The effect of ranking on a candidate's electoral success might be potentially endogenous (Lutz, 2010). Parties might simply place their popular candidates at the top of the lists. At the theoretical level, this threat is realistic. Unfortunately, our efforts to instrument real ranking with an alphabetic order, as suggested by Lutz (2010), were unsuccessful. On the other hand, we do not believe the potential endogeneity distorts our results substantially. There are a few reasons to claim so. First of all, we control for a number of predictors of both ranking and candidates' popularity. Second, both journalistic accounts and academic research (Flis, 2011; McMenamin and Gwiazda, 2011) indicate that in Poland high positions on party lists are a priceless privilege candidates compete for. In other words, this evidence suggests that parties and candidates believe that ranking largely determines a candidate's electoral success. Also, recent survey research on ballot position effects in Poland shows that about 16 per cent of Polish citizens openly admit that they rely on candidate ranking when making their electoral choices (Centrum Badania Opinii Społecznej, 2011: 5e6). Given that such an answer is highly ‘socially undesirable’ (Humm and Humm, 1944), this proportion seems reasonably high and points to the important role of ranking in determining a candidate's electoral success. Finally, research on ballot position effects in other contexts suggests that even a purely random ranking affects candidates' electoral fortunes (Lutz, 2010). Taken together, all the above arguments boost our confidence that the positive effect of ranking, even if somewhat overestimated due to endogeneity, is largely a true part of the story told here. 8. Conclusion Severe descriptive underrepresentation of women in the democratically elected parliaments might undermine not only women's substantive political representation but also democratic system legitimacy, for example by depressing voter turnout (see Karp and Banducci, 2008). Gender quota laws e a ‘fast track’ mechanism designed to enhance women's political representation e are thus a solution applied more and more frequently by democratic states. In

this paper, we have focused on the effects of quotas in openlist PR systems, that is, systems that give voters the seemingly broadest spectrum of choices. Our in-depth analysis of the Polish case suggests several broad conclusions. Most crucially, in open-list PR quotas appear to have a paradoxical nature. The introduction of quotas and a sudden substantial increase in the numbers of female candidates might lower drastically an ‘average’ female candidate's electoral performance, at least in the initial post-quota elections. Such an inference can be made not only with respect to Poland but also with respect to Brazil (Miguel, 2008) and Peru (Schmidt, 2003a). Second, relevant background e especially political experience e increases substantially the probability of a candidate winning elected office. Finally, when quotas are in place open-list PR allows us to see adequately the gender bias held by both parties' ‘selectorates’ and the electorate. In Poland, the extent of both types of bias seems to be substantial enough to preserve the rather slow pace of the gradual increase in women's descriptive representation observed over the last two decades. A more definite assessment of the effectiveness of quotas in Poland and beyond will certainly be possible only from a long-term perspective. As it seems that women's electoral success is largely based on the same grounds as is men's success (Schwindt-Bayer, 2011), a few elections might have to pass until quotas become fully effective (Schwindt-Bayer, 2009: 16). An analysis of the next Sejm election, scheduled for 2015, should thus clarify a number, even if not all, issues here. Also, other mechanisms might operate and their intensity might vary depending on cultural factors. For instance, in Brazil quotas have gradually resulted in parties putting forward more competitive female candidates (Miguel, 2008) while in Indonesia the presence of female incumbents seem to have lowered gender bias (ShairRosenfield, 2012). More research, including betweencountry comparisons and survey-based studies on the electorate,14 is certainly needed. This study should contribute to that effort. Acknowledgements Earlier drafts of this article were presented at the Midwest Political Science Association 70th Annual Conference, Chicago, 12e15 April 2012, the European Political Science Association 2nd Annual General Conference, Berlin, 21e23 June 2012 and the European Consortium for Political Research Joint Sessions, Mainz, 11e13 March 2013. The paper was also presented at the Institute of Sociology Seminar, University of Warsaw, 21 February 2012, and at the Colloquium of the Mannheim Centre for European Social Research, University of Mannheim, 7 April 2014. We are grateful to the participants of the above events, and in particular to Thomas Br€ auninger, Mikolaj Czesnik, Thomas €ubler, Marc Debus, Meryl Kenny, Natalia Letki, Richard Da Matland, Inta Mierina, Filip Raciborski and Wojciech Rafalowski, for their helpful comments.

14 Unfortunately, the available data sets on the Polish electorate, including the Polish National Election Study (PGSW) data, include questions about party choice only, thus not touching upon candidate choice.


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

77

Appendix

Figure A1. Proportion of women on party lists: histograms

Table A1 Predictors of the probability of a candidate being ranked at a ‘viable’ position on a party list: conditional logistic regression estimates. 2007

Woman Woman X Civic Platform Woman X Law and Justice Woman X Democratic Left Alliance Woman X Polish People's Party Woman X Women's Party

2011

Basic model

Full model

Basic model

Full model


0.21 (0.16) 0.37 (0.23) 0.12 (0.23)
0.19 (0.35)
0.27 (0.32)
0.89 (1.32)


0.09 (0.16) 0.33 (0.24)
0.18 (0.31)
0.53 (0.43) 0.003 (0.335)
0.79 (1.32)


1.00** (0.14) 0.50** (0.18) 0.18 (0.17) 0.36 (0.23)
0.17 (0.25)


0.87** (0.15) 1.05** (0.18) 0.58** (0.19) 0.82** (0.25) 0.25 (0.30)

0.29 (0.24)
0.15 (0.33)

0.25 (0.25)
0.04 (0.35) 4.01** (0.31) 0.95** (0.12) 0.85** (0.09) 17.48** (0.70) 14.98** (0.97) 3.93** (0.88)

Woman X Palikot's Movement Woman X Poland Comes First Incumbent Councillor Candidate in preceding election Minister Party leader Senator

4.40** (0.29) 1.27** (0.09) 0.99** (0.11) 18.04** (0.44) 15.32** (0.79) 4.78** (1.11)

(continued on next page)

78


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

Table A1 (continued ) 2007 Full model

0.02** (0.00)
0.002** (0.000) 0.27
2175 0.02 6187 296

1.39** (0.28) 0.29* (0.12) 0.09 (0.17) 1.73* (0.93) 0.01** (0.00)
0.001** (0.000) 0.76
1535 0.31 6187 296

Local executive Visible profession Business owner Celebrity Age Age squared Average p-value for rejecting IIA assumption Log likelihood McFadden R2 Number of candidates Number of party lists

2011

Basic model

Basic model

Full model

0.03** (0.00)
0.001** (0.000) 0.45
2262 0.06 7035 338

0.52* (0.29) 0.24** (0.08) 0.43** (0.14) 1.30 (0.90) 0.02** (0.00)
0.001** (0.000) 0.67
1786 0.25 7035 338

Note: *p < 0.05; **p < 0.01 (one-tailed tests). Main entries are unstandardised regression coefficients and the numbers in parentheses are standard errors. Errors are clustered at the level of electoral district (41 clusters). The IIA assumption was tested in the manner proposed by Martin and Stevenson (2001).

Table A2 Predictors of the number of votes cast for a candidate: negative binomial regression estimates. 2007

Woman Civic Platform Law and Justice Democratic Left Alliance Polish People's Party Women's Party

2011

Basic model

Interm. Model

Full model

Basic model

Interm. Model

Full model

0.36** (0.13) 3.26** (0.10) 2.97** (0.08) 2.14** (0.08) 1.72** (0.09)
0.53* (0.31)

0.39** (0.13) 2.55** (0.08) 2.35** (0.06) 1.95** (0.07) 1.77** (0.08)
0.42 (0.30)

0.46** (0.09) 2.24** (0.06) 2.02** (0.07) 1.75** (0.06) 1.59** (0.06)
0.24 (0.15)


0.33 (0.27) 3.40** (0.15) 3.16** (0.16) 1.94** (0.17) 1.96** (0.22)


0.05 (0.22) 2.82** (0.15) 2.72** (0.17) 2.05** (0.17) 2.16** (0.21)

0.08 (0.22) 2.20** (0.14) 2.35** (0.14) 1.70** (0.14) 1.85** (0.18)

2.22** (0.14) 0.65** (0.13) 0.17 (0.16)
0.07 (0.16) 0.08 (0.15)
0.18 (0.14)

2.09** (0.12) 0.64** (0.13)
0.17 (0.13)
0.25* (0.12)
0.16 (0.12)
0.33** (0.12)

2.13** (0.09) 0.59** (0.12) 0.20* (0.10)
0.06 (0.10) 0.06 (0.10)
0.14 (0.11)

0.26* (0.15) 0.29* (0.15)

0.26** (0.09) 0.30** (0.10) 2.03** (0.08)

0.26** (0.06) 0.34** (0.09) 1.59** (0.04) 1.11** (0.09) 0.58** (0.06) 0.19** (0.06) 1.22** (0.20) 2.34** (0.62)

Palikot's Movement Poland Comes First Woman X Civic Platform Woman X Law and Justice Woman X Democratic Left Alliance Woman X Polish People's Party Woman X Women's Party


0.18 (0.15)
0.10 (0.17)
0.13 (0.16)
0.43** (0.11) 1.61** (0.36)


0.23* (0.11)
0.12 (0.10) 0.04 (0.15)
0.30** (0.10) 1.29** (0.46)


0.09 (0.07)
0.06 (0.08)
0.05 (0.09)
0.20* (0.09) 1.28** (0.24)

Woman X Palikot's Movement Woman X Poland Comes First Viable position Incumbent Councillor Candidate in preceding election Minister Party leader

1.99** (0.08)

1.28** (0.07) 1.34** (0.08) 0.59** (0.03) 0.24** (0.04) 1.50** (0.22) 1.87** (0.53)


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

79

Table A2 (continued ) 2007 Basic model

2011 Interm. Model

Full model


0.07 (0.23)
0.82* (0.40) 0.01** (0.00)
0.001** (0.000) 0.00* (0.00)
0.004 (0.004) 5.31** (0.09)
47,676 0.09 6187

2.43** (0.20) 0.61** (0.07) 0.12** (0.03) 0.01 (0.05) 0.91** (0.34)
0.19 (0.19)
0.90** (0.25) 0.002* (0.000) 0.000 (0.000) 0.00** (0.00)
0.01** (0.00) 5.38** (0.08)
46,698 0.11 6187

Senator Local executive Visible profession Business owner Celebrity Proportion of women Woman X Proportion of women Age Age squared Votes cast in a district Number of party candidates in a district Constant Log likelihood McFadden R2 Number of candidates

0.12 (0.29)
1.27** (0.40) 0.02** (0.00)
0.001** (0.000) 0.00** (0.00)
0.01** (0.00) 5.77** (0.10)
49,900 0.05 6187

Basic model


0.90 (0.62)
0.63 (0.58) 0.01** (0.00)
0.001** (0.000) 0.00** (0.00)
0.03** (0.01) 6.11** (0.33)
55,306 0.05 7035

Interm. Model

Full model


1.26** (0.52)
0.20 (0.44) 0.002* (0.001) 0.000 (0.000) 0.00 (0.00)
0.02 (0.01) 5.49** (0.29)
52,845 0.10 7035

1.08** (0.18) 0.64** (0.10) 0.10* (0.04) 0.13** (0.05) 0.95** (0.24)
1.12* (0.55)
0.51 (0.46) 0.00 (0.00) 0.000 (0.000) 0.00 (0.00)
0.02** (0.01) 5.53** (0.34)
52,217 0.11 7035

Note: *p < 0.05; **p < 0.01 (one-tailed tests). Main entries are unstandardised regression coefficients and the numbers in parentheses are standard errors. Errors are clustered at the level of electoral district (41 clusters).

References Ames, B., 1995. Electoral rules, constituency pressures, and pork barrel: bases of voting in the Brazilian Congress. J. Polit. 57 (2), 324e343. Baldez, L., 2004. Elected bodies: the gender quota law for legislative candidates in Mexico. Legis. Stud. Q. 29, 231e258. Benoit, K., Marsh, M., 2008. The campaign value of incumbency: a new solution to the puzzle of less effective incumbent spending. Am. J. Polit. Sci. 52 (4), 874e890. Bernhagen, M., Marsh, M., 2007. The partisan effects of low turnout: analysing vote abstention as a missing data problem. Elect. Stud. 26, 548e560. €uninger, T., Brunner, M., Da €ubler, T., 2012. Personal vote-seeking in Bra flexible list systems: how electoral incentives shape Belgian MPs' bill initiation behaviour. Eur. J. Polit. Res. 51 (5), 607e645. Carey, J.M., Shugart, M.S., 1995. Incentives to cultivate a personal vote: a rank ordering of electoral formulas. Elect. Stud. 14, 417e439. Centrum Badania Opinii Spolecznej, 2010. Parytety Prawna Gwarancja Wiekszego Udzialu Kobiet w Polityce. Komunikat z Badan BS/130/ 2010. CBOS, Warsaw. Centrum Badania Opinii Spolecznej, 2011. Wiedza i Opinie Polakow o Prawie Wyborczym i Mechanizmach Kampanii. Komunikat z Badan BS/109/2011. CBOS, Warsaw. Chen, L.-J., 2010. Do gender quotas influence women's representation and policies? Eur. J. Comp. Econ. 7 (1), 13e60. Dubrow, J.K., 2010. The importance of party ideology: explaining parliamentarian support for political party gender quotas in Eastern Europe. Party Polit. 17 (5), 561e580. van der Eijk, C., van Egmond, M., 2007. Political effects of low turnout in national and European elections. Elect. Stud. 26, 561e573. Flis, J., 2011. Instytucjonalny labirynt demokracji wewnatrzpartyjnej. In: Chmaj, M., Waszak, M., Zbieranek, J. (Eds.), O Demokracji w Polskich Partiach Politycznych. ISP, Warsaw. Gomez, B.T., Hansford, T.G., Krause, G.A., 2007. The republicans should pray for rain: weather, turnout, and voting in U.S. presidential elections. J. Polit. 69, 649e663. Gray, T., 2003. Electoral gender quotas: lessons from Argentina and Chile. Bull. Lat. Am. Res. 22, 52e78. Holli, A.M., Wass, H., 2010. Gender-based voting in the parliamentary elections of 2007 on Finland. Eur. J. Polit. Res. 49, 598e630.

Hoodfar, H., Tajali, M., 2011. Electoral Politics: Making Quotas Work for Women. WLUML, London. Htun, M.N., Jones, M.P., 2002. Engendering the right to participate in decision-making: electoral quotas and women's leadership in Latin America. In: Craske, N., Molyneux, M. (Eds.), Gender and the Politics of Rights and Democracy in Latin America. Palgrave, New York. Huddy, L., Terkildsen, N., 1993. The consequences of gender stereotypes for women candidates at different levels and types of office. Polit. Res. Q. 46 (3), 503e525. Humm, D.G., Humm, K.A., 1944. Validity of Humm-Wadsworth temperament scale: with consideration of the effects of subjects' responsebias. J. Psychol., 55e64. Inter-Parliamentary Union, 2014. Women in National Parliaments. Situation as of 1 January 2014. http://www.ipu.org/wmn-e/classif.htm (accessed 11.03.14.). Jones, M.P., 1996. Increasing women's representation via gender quotas: the Argentine Ley de Cupos. Women Polit. 16, 75e96. Jones, M.P., 2004. Quota legislation and the election of women: learning from the Costa Rican experience. J. Polit. 66, 1203e1223. Jones, M.P., Navia, P., 1999. Assessing the effectiveness of gender quotas in open-list proportional representation electoral systems. Soc. Sci. Q. 80, 341e355. Karp, J.A., Banducci, S.A., 2008. When politics is not just a man's game: women's representation and political engagement. Elect. Stud. 27, 105e115. Kittilson, M.C., 2006. Challenging Parties, Challenging Parliaments: Women and Elected Office in Contemporary Western Europe. Ohio State University Press, Columbus. Kukołowicz, P., 2013. Do voters read gender? Stereotypes as voting cues in electoral settings. Pol. Sociol. Rev. 182 (2), 223e238. Kunovich, S., 2003. The representation of Polish and Czech women in national politics: predicting electoral list position. Comp. Polit. 35 (3), 273e291. Kunovich, S., 2012. Unexpected winners: the significance of an openlist system on women's representation in Poland. Polit. Gend. 8, 153e177. Langston, J., Aparicio, F.J., 2011. Gender quotas are not enough: how background experience and campaigning affect electoral outcomes. In: Paper Presented at the Midwest Political Science Association Meeting, March 31eApril 3, 2011.

80


recki, P. Kukołowicz / Electoral Studies 36 (2014) 65e80 M.A. Go

Langston, J., Aparicio, F.J., 2012. Beyond the incumbency disadvantage: why more women do not win legislative seats?. In: Paper Presented at the Midwest Political Science Association Meeting, April 12e15, 2012. Lutz, G., 2010. First come, first served: the effect of ballot position on electoral success in open ballot PR elections. Representation 46 (2), 167e181. Mainwaring, S.P., 1991. Politicians, electoral systems, and parties: Brazil in comparative perspective. Comp. Polit. 24 (1), 21e43. Markowski, R., 2006. The Polish elections of 2005: pure chaos or a restructuring of the party system? West Eur. Polit. 29 (4), 814e832. Markowski, R., 2008. The 2007 Polish parliamentary election: some structuring, still a lot of chaos. West Eur. Polit. 31 (5), 1055e1068. Martin, L.W., Stevenson, R.T., 2001. Government formation in parliamentary democracies. Am. J. Polit. Sci. 45 (1), 33e50. Matland, R.E., 1998. Enhancing women's political participation: legislative recruitment and electoral systems. In: Karam, A. (Ed.), Women in Parliament: beyond Numbers. IDEA, Stockholm. McElroy, G., Marsh, M., 2010. Candidate gender and voter choice. Analysis from a multimember preferential voting system. Polit. Res. Q. 63 (4), 822e833. McMenamin, I., Gwiazda, A., 2011. Three roads to institutionalisation: Vote-, office- and policy-seeking explanations of party switching in Poland. Eur. J. Polit. Res. 50 (6), 838e866. Miguel, L.F., 2008. Political representation and gender in Brazil: quotas for women and their impact. Bull. Lat. Am. Res. 27 (2), 197e214. Norris, P., Lovenduski, J., 1993. ‘If only more candidates came forward’: supply-side explanations of candidate selection in Britain. Br. J. Polit. Sci. 23 (3), 373e408. Rule, W., 1994. Parliaments of, by, and for the people: except for women? In: Rule, W., Zimmerman, J.F. (Eds.), Electoral Systems in Comparative Perspective: their Impact on Women and Minorities. Greenwood Press, Westport. Sanbonmatsu, K., 2002. Gender stereotypes and vote choice. Am. J. Polit. Sci. 46 (1), 20e34. Schmidt, G.D., 2003a. The implementation of gender quotas in Peru: legal reform, discourse and impacts. In: Paper Presented at the

International IDEA Workshop ‘The Implementation of Quotas: Latin American Experiences’, February 23e24, 2003. Schmidt, G.D., 2003b. Unanticipated successes: lessons from Peru's experiences in majoritarian closed list and open list PR systems. In: Paper Presented at the International IDEA Workshop ‘The Implementation of Quotas: Latin American Experiences’, February 23e24, 2003. Schmidt, G.D., 2008. The election of women in list PR systems: testing the conventional wisdom. Elect. Stud. 28, 190e203. Schwindt-Bayer, L.A., 2005. The incumbency disadvantage and women's election to legislative office. Elect. Stud. 24, 227e244. Schwindt-Bayer, L.A., 2009. Making quotas work: the effect of gender quota laws on the election of women. Legis. Stud. Q. 34, 5e28. Schwindt-Bayer, L.A., 2011. Women who win: social background, paths to power, and political ambition in Latin American legislatures. Polit. Gend. 7, 1e33. Shabad, G., Slomczynski, K., 2002. The emergence of career politicians in post-communist democracies: Poland and the Czech Republic. Legis. Stud. Q. 27 (3), 333e359. Shair-Rosenfield, S., 2012. The alternative incumbency effect: electing women legislators in Indonesia. Elect. Stud. 31, 576e587. Shugart, M.S., 1994. Minorities represented and underrepresented. In: Rule, W., Zimmerman, J.F. (Eds.), Electoral Systems in Comparative Perspective: their Impact on Women and Minorities. Greenwood Press, Westport. Siemienska, R., 2000. Nie Moga Czy Nie Potrafia? O Postawach i Uczestnictwie Politycznym Kobiet w Polsce ‘Scholar’: Warsaw. Sledzinska-Simon, A., Bodnar, A., 2013. Gender equality from beneath: electoral gender quotas in Poland. Can. J. Law Soc. 28 (2), 151e168. Ufen, A., 2010. The legislative and presidential elections in Indonesia in 2009. Elect. Stud. 29, 276e285. Valdini, M.E., 2012. A deterrent to diversity: the conditional effect of electoral rules on the nomination of women candidates. Elect. Stud. 31, 740e749. Wessels, B., 1997. Germany. In: Norris, P. (Ed.), Routes to Power: Legislative Recruitment in Advanced Democracies. Cambridge University Press, Cambridge. West, D.M., Orman, J.M., 2002. Celebrity Politics. Prentice Hall, Upper Saddle River, NJ.