WHY REGIME TYPE AFFECTS THE LINK BETWEEN GENDER AND CORRUPTION. First draft. Mattias Agerberg, Aksel Sundström & Lena Wängnerud 1

WHY REGIME TYPE AFFECTS THE LINK BETWEEN GENDER AND CORRUPTION First draft Mattias Agerberg, Aksel Sundström & Lena Wängnerud 1 The Quality of Government Institute Department of Political Science, University of Gothenburg Box 711, SE 405 30 Gothenburg, SWEDEN Abstract: While research have confirmed that there is an association between higher shares of women in parliaments and lower levels of corruption, recent studies suggests that this relationship may be conditioned upon factors such as whether a country is democratic or not. In this paper, we propose another way of understanding this relationship. Drawing upon literature that characterizes endemic corruption as a collective action problem, we hypothesize that while the share of elected women should have little or no impact on the levels of corruption in contexts of corrupt equilibrias, in a setting where good governance is the state's operative norm, increasing women's political representation could have potential to curb corruption. We distinguish between three regime types: Neo-Patrimonial states; Competitive Particularism states; and Good Governance states. Analyzing a global cross-sectional dataset we find support for that this distinction is useful. Our findings indicate that regime types affect the relationship between gender and corruption. Keywords: Corruption; Gender; Regime types; Good governance Paper to be presented at IPSA 23 rd World Congress of Political Science July 19-24, Montreal, Quebec, Canada 1 E-mail: m.agerberg@gmail.com, aksel.sundstrom@pol.gu.se, lena.wangnerud@pol.gu.se 1

Introduction After more than two decades of research on corruption and other aspects of government quality, it is clear that the top echelon of society is of fundamental importance, not only because of their direct influence on politics, administration and economy, but also since their behavior indirectly signals how things are done. Dysfunctional elites undermine economic growth but also various dimensions of human wellbeing, such as health, access to clean water and education, and subjective features such as human subjective wellbeing and happiness (cf. Rothstein and Holmberg, 2012 for an overview). This paper departs from two facts; first, that corruption represents, in most cases, a stable equilibrium and second, that even highly stable equilibria are subject to disruption. Thus, it makes sense to explore factors that affect levels of corruption. Genderbalance in government institutions is one such factor attracting more and more attention in research. The pioneering studies in the field demonstrated a correlation between the proportion of women in government and levels of corruption a high proportion of women was linked to a low level of corruption and, vice versa, a low proportion of women was linked to a high level of corruption and discussed plausible explanations for this correlation (Dollar et al., 2001; Swamy et al., 2001). More recent studies however has moved beyond this step and presented fine-tuned understandings on the importance of context; the effect of gender is found in democracies but not in autocracies, and suggested that the core mechanism at work is accountability. The line of reasoning is the following: Women are less willing than men to violate social norms (risk-aversion) and in societies where government officials are at risk of being caught (accountability) this affects the behavior of women and make them less prone to engage in corruption (Esarey & Chirillo, 2013; Esarey & Schwindt-Bayer 2014). Recent studies, as the ones referred to above, are important since they help to answer the fundamental questions on how and under what conditions the gender-balance of 2

elites has an effect on corruption. Gender is indeed a fluid category and therefore one need to ask what makes the gender-factor tick in some settings but not in other. Our main critique however is that most studies in the field of gender and corruption depart from a principalagent-perspective where the decision to engage, or not to engage, in corrupt interactions implicitly is seen as an individual-level decision. We believe that much could be gained by viewing corruption as a collective action dilemma. What a collective-action-perspective contributes with is a focus on joint forces that need to be in place in order to spark changes. More specifically, we argue that Esarey and Chirillo (2013) and Esarey and Schwindt-Bayer (2014) elaborate a good understanding on pull-factors, what prevents women government officials from engaging in corruption, but in order for this strand of research to flourish we also need a good understanding of push-factors how an influx of women in government institutions can change how things are done in a more fundamental way. One could argue that push and pull factors are two sides of the same coin. We build the analyses in this paper on empirical findings in our own ongoing research: We have found a correlation between the proportion of women in locally elected councils and regional levels of corruption (Sundström & Wängnerud, 2014; Grimes & Wängnerud, 2014). Such variation is difficult to explain with the assumptions of the above-mentioned studies. Thus, we need a theory that helps to understand variation within countries and believe that the understanding we propose in this article is better equipped at explaining such patterns. The paper will proceed as follows: First we present major findings in previous research on the link between gender and corruption. Thereafter we develop our understanding of corruption as a collective action dilemma and discuss the implications of such a perspective. We present results from an empirical analysis where we distinguish between three regime types: Neo-Patrimonial states (NeP); Competitive Particularism states (CoP); and Good Governance states (GoG). Analyzing a global cross-sectional dataset we find 3

support for that this distinction is useful. The paper ends with a section where we lay the ground for further theorizing on push factors pressure for change in different contexts and the role that gender can play in such processes. In the concluding part we suggest directions for future research. Literature review History is full of examples of powerful female leaders (Indira Gandhi is a case in point) who have emerged from political dynasties with seemingly little aversion to corruption a concept most often defined as the abuse of public power for private gain. There is certainly plenty of anecdotal evidence of women being elected in highly corrupt contexts. Systematic research, however, tells an interesting story. In their seminal study, Are Women Really the Fairer Sex? Corruption and Women in Government, Dollar et al. (2001) demonstrate that higher rates of female participation in government are associated with lower national levels of corruption. They measure corruption using data from the International Country Risk Guide and include a range of variables in their analysis to control for various underlying institutional characteristics that could be responsible for a spurious correlation. Since that study, the association between women and lower levels of corruption has been reproduced in repeated studies (see Treisman, 2007). Today it is relatively uncontroversial to state that there is an association between gender and corruption the question of causality, however, is the subject of heated debate. The observation that countries with a higher number of elected women generally have lower levels of corruption has produced an assumption among a number of policy-makers that increasing the proportion of women in government will provide a quick fix for corruption (e.g. World Bank, 2001). In contrast, however, Sung launched a rival explanation: the fairer system thesis. This argument contends that liberal democratic institutions and spirit increase female 4

participation in government and restrain systematic corruption, but the latter two factors are not causally related (2003:708, 2012). In addition Goetz (2007) argues that it is opportunities for corruption that differ for women and men (see also Bjarnegård, 2013; Branisa and Ziegler, 2010; Pande and Ford, 2012; Vijayalakshmi, 2008). Esarey and Chirillo (2013) take this research one step further when they point to the fact that a higher share of elected women is associated with less corruption in democratic countries but not in authoritarian ones. Moreover, they find that tolerance of corruption among the general public also follow this trend. They suggest that this pattern is explained by a mechanism where women are more averse to the risks of violating political norms, and because gender discrimination makes violating institutional norms a riskier proposition for women than men (p.382). Accordingly, where corruption is stigmatized, women will be less tolerant of corruption and less likely to engage in it compared to men. But if corrupt behaviors are an ordinary part of governance supported by political institutions, there will be no corruption gender gap (p. 382 384). Esarey and Schwindt-Bayer (2014) propose a further nuance in this relationship, exploring the role of accountability as a condition for the mechanism of risk aversion to take place: Where the probability of being held accountable for corruption is high, we claim that women should be more resistant to corruption than men because they are more risk averse and corruption is more likely to be detected and/or be punished. However, where accountability is low (i.e., corruption is unlikely to be caught and/or is unlikely to be harshly punished), we argue that women s engagement in corruption will be no different than men s (p. 7). We believe that the recent trends in this literature point to the important fact that researchers need to understand how the gender-corruption nexus operates in different settings. Our interpretation is that Esarey and Chirillo (2013) and Esarey and Schwindt-Bayer (2014) comprehend the mechanism linking gender and corruption to women s self-interest to survive 5

in societies that discriminates along gender-dimensions. Following these assumptions, the pressure on women for norm following should apply in the same way throughout a country and thus it is hard to make sense of empirical findings showing substantial subnational variation between regions and/or between different types of government institutions (e.g. Sundström & Wängnerud, 2014). Corruption as a collective action dilemma There is a major strand of research that sees corruption as an individual-level departure from rules and policies, and therefore highlights effects of systems of monitoring and sanctioning. The idea is that principals, in the form of higher level officials or audit offices, fail to detect and punish instances of rule-breaking among agents (Rose-Ackerman, 1999). Research on policy change identifies a number of mechanisms such as decision costs, transaction costs, information costs and cognitive costs (Jones et al., 2003), all of which contribute to path dependency in policies and institutions, and can help to explain why the rules and monitoring systems needed to combat corruption may be slow to emerge. In any context in which corruption is moderately to highly prevalent, however, transforming the modus operandi from corruption and collusive transactions to publicregarding and rule-bound behavior also involves overcoming a large-scale collective action dilemma, as all who benefit from corruption have strong material incentives to continue acting corruptly if they believe that others also intend to opt for self-interested rent-seeking strategies (Grimes & Wängnerud, 2014). Unlike other collective action dilemmas, defection in the form of engaging in corruption also entails violation of the law. Engaging in corruption even in the instances in which it may be profitable for all parties involved in the transaction, requires a degree of confidence among those involved that transgressions will not be revealed. Corruption 6

therefore becomes a system in its own right with norms of reciprocity (regarding favors, benefits, and turning a blind eye), a socialization component, and is a system heavily dependent on interpersonal relationships (e.g. Graham, 1990; Karklins, 2005; Persson et al., 2012). Even highly stable equilibria are subject to disruption, however, given sufficient pressure for change. In order to fully understand the dynamics at work we need to theorize on potentially decisive factors in such transformations. We believe that theorizing carefully on effects of regime-type is a first important step towards well-grounded theoretical models. Analyzing effects of regime-types There is no straightforward answer to the question on how to distinguish between states since categorizations take place against the backdrop of a specific research question. Our research question concern gender-balance in government institutions and how this can affect levels of corruption. The distinction democracies/autocracies may be a first hand choice in some studies, but an alternative is to focus more directly on corruption as a stable equilibrium and what this means for the role of women in processes towards reduced levels of corruption. North and colleagues (2009) argue that one can divide the states of the world in two categories: Limited access orders and Open access orders. The vast majority of states of the world is, and has been, Limited access orders. These states maintain state monopoly of violence by excluding competitors from power and the political manipulation of economy. Open access orders has developed in a small number of countries over the last 500 years, and unlike Limited access orders, these states are characterized by political and economic competition (North et al., 2009). Building on a similar reasoning Mungiu-Pippidi (2011) argues that the category Limited access orders can be divided into two subcategories: Neo-Patrimonialism (NeP) and 7

Competitive particularism (CoP). NeP States have often an individual ruler who treats the state more or less as their property, and a shift in power in these states are often violent and bloody. The distribution of resources and power in these states is relatively predictable and consistent to the particulate basis in client elitist network (Mungiu-Pippidi, 2011:31-33). NeP States lacks almost entirely the legitimacy of population and the starting point here is a corrupt equilibrium. States like Zimbabwe and Chad fall into this category. CoP-states constitute the majority of the world's states and have replaced the violent seizure of power by public choice fairly free, but usually not fair. The allocation of power and resources are often unpredictable and characterized much of who happens to have the political power at a certain moment. CoP-states have generally low legitimacy, and when people do not expect public treat to be equal, their first priority is rather to belong to the privileged group than challenge the political rules of the game (Mungiu-Pippidi, 2011:32-33). In regard to corruption, the equilibrium position is again the starting point (for example, in states such as Russia), but the category also includes states that are on its way towards a more positive trend with more universalistic and less corrupt institutions (like for example in Poland or Lithuania). In contrast to NeP States and CoP-states we find states which have achieved a certain level of good governance (GoG). Mungiu-Pippidi (2011:18-21) argues that these states can be divided into four categories: Historical achievers (such as Denmark and the UK), early achievers (e.g. Japan), contemporary achievers (such as Spain and Chile) and mostly free or not free achievers (e.g. Singapore). A common feature of these states is that state power generally operates on a universalistic basis and that they have managed to fight the worst forms of corruption. These states are not suffering from a corrupt equilibrium even though they still may face significant levels of corruption. 8

Summing up this discussion we would like to point out core features in these regime-types. What we are looking for is the likelihood for a window of opportunity to emerge. Note that corruption, in most cases, represent a stable equilibrium and the starting point for any theory is that it is very hard to achieve fundamental changes. Thus one need to look for joint forces in different settings. If we start with GoG states, it is important to recognize that they generally have lower levels of corruption than other states. However, even more important to note is that they are characterized by elite competition and universalistic values. One can expect that once the issue of corruption has been set in motion, changes can be comparatively fast and farreaching. The electoral arena can be expected to play a central role since electoral politics has to do with values and the electoral arena is a sphere where it is necessary for a political party, but also for individual candidates, to stand out if one is to attract attention and votes (cf. Manin, 2007). Individual female candidates may, for various reasons, use gender attributes such as being a clean outsider to increase their chances of being elected, but the influx of women into political parties may also start a reflective process that goes beyond that. Most research show that noticeable increases in the number of women elected is the result of conscious acts on behalf of party leadership (Kittilson 2006; Wängnerud 2009). The influx of women may contribute to puncturing otherwise quite stable norms and initiate a process of change in several spheres such as agenda setting (Phillips 1995) but also norms of reciprocity regarding favors and benefits. If we then move to a discussion on NeP and CoP-states it is important to note the role of predictability and violence. A high risk of violence is a strong pull factor. In fact, it is hard to imagine how the gender-balance in government institutions could have an impact in cases of collusive corruption in combination with a violent political climate which is the case in NeP-states. CoP-states are characterized by less violence but also by higher levels of 9

unpredictability. Our expectation is that this would lead to large variation within this group of states sometimes the gender-balance may have an effect and sometimes not. If we would be able to follow developments over time we would also expect large fluctuations within individual countries in the CoP-category of states. Data and design Our starting point is to divide the world countries in NeP-States, CoP-states and GoG-States. The degree of good governance can generally be said to be very low in NeP States, low or medium in CoP-states and comparatively high in GoG States. It is of course difficult to be categorical in a rough subdivision thereof; some countries are definitely on the borderline between two categories. We believe that this division serves the purpose to show that regimetypes affect the gender-corruption nexus. Using the distinction in Mungiu-Pippidi (2011) we divided the world s countries on theses characteristics in to three clusters. The availability of corruption measures limit our focus to the years from 1996 and onwards. A pooled cross-sectional dataset gives us a sample of over 2700 country-years, dispersed across these three categories. In the analysis we present summary statistics for the different type of states. As indicated, each of the three clusters shows widely differing features, a fact that we argue support this distinction of states. The analysis first present bivariate statistics between our two variables of main interests the representation of women and levels of corruption and then proceed to report the multivariate models, where we control for factors that may explain the identified association. Dependent variable Our main dependent variable consists of the World Bank's data on the perceived level of corruption from a large cross-section of countries, covering the years 1996 to 2010. The 10

World Bank defines corruption as the extent to which public power is exercised for private gain, including both petty and grand forms of corruption as well as capture of the state by elites and private interests (Kaufmann et al., 2010: 4), and uses the indicator Control of Corruption (WBCC) to operationalize this definition. This is one of several perception-based measures of corruption that combines data from multiple sources and uses surveys and riskassessments of country experts to compile a general index of corruption levels in different countries. These indicators are thus by and large subjective measures, and not hard measures of actual corrupt activities in a country. 2 Independent variable Our key independent variable is a measure of the percentage of women in parliament. Data on the share of elected women is collected by Inter-Parliamentary Union (Inter-Parliamentary Union, 2013). In countries with both a lower and an upper house we use the share of women in the lower house. This variable is lagged by one year and then matched with the corruption data for a given country. Each observation thus consists of a country with corruption data from year t, paired with the percentage of women in parliament from the year t - 1. Control variables To check for potential spurious correlations we include a number of control variables that have shown to affect corruption levels in previous studies. Economic factors are generally considered important in explaining corruption (see Treisman, 2007). Therefore we include measures of log per capita GDP and a measure of trade openness (trade as % of GDP). 3 The 2 That being said, measuring corruption objectively is much more problematic since measures like corruption conviction rates might be a better test of a country's legal system than the actual level of corruption. Perception based indicators like WBCC are therefore currently the best and most commonly used way to measure corruption that are widely available for a large number of countries (see Ko & Samajdar 2010). 3 These variables are based on figures from the United nations Development Program (UNDP) and the World Bank and taken from the Quality of Government dataset (Teorell et al. 2013). 11

level of democracy plays an important role as a determinant of the level of corruption (Keefer, 2007), and several authors argue that it affects the relationship between gender and corruption (Esarey & Chirillo, 2013; Esarey & Schwindt-Bayer, 2014; Sung, 2003; Sung, 2012). We measure the level of democracy with the combined Freedom House / Polity scores. 4 Following Swamy et al. (2001) we also control for the percent of population who are Muslim and the percent of population who are Catholic in a country 5. These variables should be considered to be proxies for cultural factors that may affect women's participation in politics and/or corruption. Swamy et al. (2001) notes that for example catholic countries like Italy and Spain have higher levels of corruption and lower rates of women's participation compared to protestant Scandinavian nations. In the full multivariate models we include these variables simultaneously in order to analyze the impact from women s representation on corruption levels. The equation for each of the three models is the following: National levels of corruption = β 0j + β 1 *(Share of women in parliament)+ β 2 *(Levels of democracy) + β 3 *(Log GDP per capita) + β 4 *(Openness to trade) + β 5 *(Catholic share of population) + β 6 *(Muslim share of population) + e Results We first corroborate the reasoning of dividing the world s states in to three clusters by analyzing the broad features of this divide. As can be seen from table 1 it is evident that the countries within these clusters have a context that in terms of economic development and corruption levels differ substantially from each other. 4 This data is based on Hadenius & Teorell (2005) and taken from the Quality of Government dataset (Teorell et al. 2013). 5 This data is based on La Porta et al. (1999) and taken from the Quality of Government dataset (Teorell et al. 2013). 12

Table 1. Summary statistics average values for the states in the three categories Variable NeP states CoP states GoG states Control of corruption -0.84 (0.52) -0.41 (0.46) 1.41 (0.58) Women in parliament 13.62 (10.23) 12.87 (8.81) 16.95 (11.21) Level of democracy 2.21 (1.22) 7.39 (1.87) 7.72 (3.02) GDP per capita 5634.50 (7411.71) 6308.13 (6547.94) 19231.43 (15708.79) Trade (% of GDP) 87.50 (39.07) 85.23 (38.38) 94.15 (58.15) Catholic (% of population) 13.28 (22.09) 37.03 (37.01) 38.51 (37.67) Muslim (% of population) 47.79 (42.22) 18.92 (30.96) 14.16 (29.77) Number of states 455 1112 1255 Comments: The standard deviation is reported in the parenthesis Pertaining to the relationship between women in parliament and national levels of corruption we first analyze the bivariate associations in each of the three clusters. First, as can be seen from Figures 1 to 3, it should be noted that there is a large variation in both variables in all three samples. Second, it is evident that the relationship between the variables is quite different as we move from the NeP-states to the GoG-states. While the regression line is almost flat in the first sample of countries indicating a non-existing correlation among NePstates the slope of the line is positive among the GoG-states. Accordingly, there seem to exist a relationship between the share of women in parliament and levels of corruption in the latter but not in the former. In Figure 4 all clusters of nations are presented in the same graph, visualizing how the regression slopes differ between the three different samples. 13

Figure 1. The association between women in parliament and corruption in NeP-states Figure 2. The association between women in parliament and corruption in CoP-states Figure 3. The association between women in parliament and corruption in GoG-states 14

Figure 4. The association between women in parliament and corruption in all samples When proceeding to analyze this relationship in a multivariate framework we include the control variables in a full model. In doing so, a similar pattern emerges. As visible in table 2, the NeP-states still do not show any association between our main independent variable and 15

the dependent variable the coefficient measuring this effect is not significant. As a contrast, within the GoG-states there is still an association to be found also when controlling for a range of plausible explanatory factors. This relationship is positive and statistically significant. In more detail, it suggest that a one-point increase in women s representation in parliament would generate a higher value in the measure of combating corruption of 0.022 units. It can also be noted that we find an effect from the share of elected women on corruption levels in the middle category. Within the CoP-states, the effect is positive, indicating an effect of 0.005 on the corruption indicator from a one-point increase in the share of elected women. The results from the multivariate models support the argument developed in this article, that while the share of elected women should have little or no impact on the levels of corruption in contexts of corrupt equilibrias, in a setting where good governance is the state's operative norm, increasing women's political representation could have potential to curb corruption. We believe that the argument developed in this article has stronger explanatory power than the reasoning outlined by Esaray and Chrillo (2013), focusing on the role played by democracy. The findings outlined above support this suggestion and they are also robust for an alternative specification of the main independent variable. When measuring the variable of women in parliament without the one-year lag, the results remain basically identical. These figures are presented in table 3 in the appendix. 16

Table 2. Predicting corruption levels, OLS regression Three types of countries (DV: Corruption Perception Index) Predictors NeP States CoP States GoG States Share women in parliament.003.005***.022*** (1 year lag) (.002) (.001) (.002) Level of democracy.001.079*** -.065*** (.019) (.008) (.009) Log GDP/capita.103***.184***.493*** (.023) (.012) (.030) Trade (% of GDP) -.001*.001*** -.000 (.001) (.000) (.000) Muslim share of population.000.001 -.012*** (.001) (.000) (.001) Catholic share of population -.006*** -.002*** -.002*** (.001) (.000) (.000) Constant -1.513*** 2.636*** 3.230*** (.208) (.093) (.251) N 387 1003 506 Adj R-squared.14.42.65 Notes: The coding of the variables are described in the methods section and taken from the Quality of Government dataset (Teorell et al. 2013). * = p<.05 ** = p<.01 *** = p<.001 Discussion The question remains; what sets the process of reducing corruption in motion in the first place? Our answer is that it does not have to be female actors. The upshot of our study is that there need to be several elements in place in order for the gender-balance among government officials to play a role in processes towards reduced levels of corruption. 17

A crucial factor for GoG-states is probably the combination of elite competition and the fact that corruption is comparatively high on the policy agenda. The entry of a new set of actors, who bear with them a somewhat different background, may help to start a process of self-reflection in organizations such as political parties. The recruitment of women is, also in established democracies, a signal of a break with old habits that usually receive a lot of attention, not the least from the media. Political parties want to stay in power and it would be particularly risky for parties that have started to recruit women on a large scale to turn a blindeye to acts of corruption. Thus we comprehend the mechanism linking gender to corruption to political parties self-interest to survive in societies that build on elite competition and universalistic values. Conclusion We started off this paper by referring to Indira Gandhi and anecdotal evidence of powerful female leaders that have emerged from dynasties with seemingly little aversion to corruption. We could end this paper with counter examples such as Dilma Rousseff, president of Brazil, who has sacked several ministers over allegations of corruption and has repeatedly stressed that she will not tolerate wrongdoing. We could also exemplify with countries such as Rwanda where the bloody conflicts in the 1990 s paved the ground for a gender quota system and strategies to combat corruption. Currently Rwanda has the highest number of women elected in the world, 63.8% (Inter-Parliamentary Union, 2014) and the country is doing comparatively well on rankings produced by Transparency International (TI). In 2013, Rwanda, was ranked as 49 out of 177 countries on the list of least to most corrupt countries in the world (TI, 2013). We believe that what is needed in further research is more systematic studies on the micro-foundations at work when the gender-balance in government institutions has an 18

effect on levels of corruption. We are not asking for idiosyncratic stories unusual factors at work but core features that need to be in place in order to spark changes. The important conclusion from our study is that such core features should cluster differently depending on regime-type. Future research also needs to produce more concrete information on how changes take place. Elites are important because of their direct influence on politics, administration and economy. Political elites are of especial importance since they produce policies and legislation. However, most under-researched, when it comes to the effects of elites, is probably signal effects. For example, do gender equality at the top signal impartiality equal treatment to the population? These issues are all worthy to be further discussed by the scholarly community. 19

References Adsera, A., Boix, C., Payne, M. (2003) Are you being served? Political accountability and quality of government. The Journal of Law, Economics & Organization, 19 (2), 445-90. Alatas, V., Cameron, C., Chaudhuri, A., Erual, N. and Gangadharan, L. (2009) Gender, culture and corruption: insights from an experimental analysis, Southern Economic Journal, 75 (3), 663-80. Alhassan-Alolo, N. (2007) Gender and corruption: testing the new consensus. Public Administration and Development, 2 (3), 227-37. Bjarnegård, E. (2013) Gender, Informal Institutions and Political Recruitment: Examining Male Dominance in Parliamentary Representation. New York: Palgrave McMillan. Branisa, B. and M. Ziegler (2010) Re-examining the Link Between Gender and Corruption: The Role of Social Institutions, Discussion Paper No. 24. Göttingen: Courant Research Centre. Dollar, D., R. Fisman, and R. Gatti (2001) Are Women Really the Fairer Sex? Corruption and Women in Government, Journal of Economic Behaviour & Organization 46, 423-9. Esarey, J. and G. Chirillo (2013) Fairer Sex or Purity Myth? Corruption, Gender, and Institutional Context, Gender and Politics 9 (4): 390-413. Esarey, J. and L. Schwindt-Bayer (2014) Gender, Corruption, and Accountability: Why Women are (Sometimes) More Resistant to Corruption Than Men, work in progress. Goetz, A. M. (2007) Political Cleaners: Women as the New Anti-Corruption Force?, Development & Change 38 (1): 87-105. Hadenius, A. and J. Teorell. (2005) Assessing Alternative Indices of Democracy, Concepts & Methods Working Papers 6, IPSA, August 2005. Inter-Parliamentary Union. (2014). Women in Parliaments: World and Regional Averages. http://www.ipu.org/wmn-e/world.htm 20

Kaufmann, D., A. Kraay, and M. Mastruzzi (2010). The Worldwide Governance Indicators: Methodology and Analytical Issues. http://ssrn.com/paper=1682130 (Accessed August 22, 2013). Keefer, P. (2007) Clientelism, Credibility, and the Policy Choices of Young Democracies, American Journal of Political Science, 51(3): 433-448. Ko, K. and A. Samajdar (2010) Evaluation of international corruption indexes: Should we believe them or not?, The Social Science Journal, 47: 508-540. La Porta, R., López-de-Silanes, F., Shleifer, A.. and Vishny, R. (1999). The Quality of Government. Journal of Law, Economics and Organization, 15(1): 222-279. Michailova, J. and I. Melnykoskova (2009) Gender, corruption and sustainable growth in transition countries, Journal of Applied Economic Sciences, 9: 387-407. Mishra, A. (2005) Persistence of corruption: Some theoretical perspectives, World Development, 34 (2): 349-58. Mungiu-Pippidi, A. (2006) Corruption: Diagnosis and Treatment, Journal of Democracy, 17 (3), 86-99. Mungiu-Pippidi, A. (2011) Chasing Moby Dick across every sea and ocean? Contextual choices in fighting corruption, Hertie School of Goverance: Report commissioned by the Norwegian Agency for Development Cooperation, Berlin. Stockemer, D. (2011) Women s Parliamentary Representation in Africa: The Impact of Democracy and Corruption on the Number of Female Deputies in National Parliaments, Political Studies, 59 (3): 693-712. Sundström, A. and L. Wängnerud (forthcoming) Corruption as an obstacle to women s political representation: Evidence from local councils in 18 European countries, Party Politics. 21

Sung, H.-E. (2012) Women in Government, Public Corruption, and Liberal Democracy: A Panel Analysis, Crime, Law and Social Change, 58: 195-219. Sung, H.-E. (2003) Fairer Sex or Fairer System? Gender and Corruption Revisited, Social Forces, 82 (2): 703-23. Swamy, A., S. Knack, Y. Lee, and O. Azfar (2001) Gender and Corruption, Journal of Development Economics, 64 (1): 25 55. Transparency International (2014). Corruption Perceptions index, 2013. Available online, http://cpi.transparency.org/cpi2013/results/ Treisman, D. (2007) What Have We Learned About the Causes of Corruption from Ten Years of Cross-National Empirical Research?, Annual Review of Political Science 10: 211-44. Vijayalakshmi, V. (2008) Rent-Seeking and Gender in Local Governance, Journal of Development Studies, 44 (9): 1262-88. World Bank (2001) Engendering Development: Through Gender Equality in Rights, Resources and Voice, A World Bank Policy Research Report. Oxford: Oxford University Press. Wängnerud, L. (2012) Why Women Are Less Corrupt than Men, In Good Government: The Relevance of Political Science, Northhampton, MA: Edward Elgar. 22

Appendix Table 3. Predicting corruption levels, OLS regression Three types of countries (DV: Corruption Perception Index) Predictors NeP States CoP States GoG States Share women in parliament.004.004***.023*** (No time lag) (.002) (.001) (.002) Level of democracy.001.080*** -.068*** (.019) (.008) (.015) Log GDP/capita.105***.186***.492*** (.023) (.012) (.030) Trade (% of GDP) -.001*.001*** -.000 (.001) (.000) (.000) Muslim share of population.000.001 -.012*** (.001) (.000) (.001) Catholic share of population -.006*** -.002*** -.002*** (.001) (.000) (.000) Constant -1.542*** 2.648*** 3.200*** (.210) (.093) (.251) N 387 1004 506 Adj R-squared.16.42.65 Notes: The coding of the variables are described in the methods section and taken from the Quality of Government dataset (Teorell et al. 2013). * = p<.05 ** = p<.01 *** = p<.001 23