Corruption, Gender and Accountability
A quantitative assessment of Gender and Attitudes towards Corruption in Swedish Municipalities
Viktoria
HallmansMaster Thesis, 30 credits Supervisor: Sven Oskarsson
Department of Government Word count:15 440
Uppsala University, Autumn 2020
Abstract
Previous research has shown that there is a negative relationship between female elected officials and corruption in democracies. Especially in high accountability settings where corruption is stigmatized, it makes female elected officials more risk averse as a result of higher probability of being held accountable. Through a quantitative approach, this thesis examines the relationship of gender and corruption on the individual level in Swedish municipalities. To investigate if there are any differences in attitudes towards corruption among female and male local politicians, and what might cause that difference, a survey answered by local politicians from all the 290 municipalities in Sweden was used. The results showed that there is a difference in attitudes towards corruption, female local politicians seem to be more negative than their male counterparts. Female local politicians also seem to become even more negative in their attitudes towards corruption if the accountability is higher. The explanation could be that female local politicians are more risk averse than male local politicians in Swedish local councils, as the same attitude could not be found for them. However, these findings cannot exclude that other explanations can be important for the relationship between gender and attitudes towards corruption. Nevertheless, this thesis gives a starting point for further research of the connection between the individual level of attitudes towards corruption, gender and accountability in Swedish municipalities.
Keywords: Gender, attitudes towards corruption, accountability and risk aversion, local politicians, Sweden.
Table of contents
1. Introduction 4
2. Theoretical framework 7
2.1 Gender and Corruption 7
2.2 Mechanism of Accountability and Risk Aversion 10
3. The contextual background of Sweden 15
4. Method 18
4.1 Case selection 18
4.2 Data 20
4.3 Operationalization and variables of interest 22
4.4 Descriptive data of mean 26
5. Results 27
5.1 Attitudes towards Corruption 27
5.2 Accountability and Risk Aversion 33
6. Discussion 38
7. Conclusion 42
References 44
Appendix 47
1. Introduction
“it should not come as a surprise that if women were to hold a greater number of positions of power in a political system, corruption would become less common”
(Rothstein, 2018: 42-43).
Corruption and bad governance are documented to harm democratic development as well as political leadership in both new and advanced democracies (Erlingsson, Bergh & Sjölin, 2008;
Bergh et al, 2013; Jha & Sarangi, 2018). It can even in the least corrupt setting undermine the democratic political system and hinder the quality of government (Rothstein & Teorell, 2008;
Linde & Erlingsson, 2013). Further, a generally agreed-upon solution to diminish corruption is increasing female representation. Higher rates of female elected officials are associated with lower levels of corruption (Barnes & Beaulieu, 2014; Jha & Sarangi, 2018). Research has shown that women in governments where corruption is not stigmatised will not lead to a negative relationship. However, in democracies where corruption is stigmatized, both by social and political norms, female representatives tend to reduce corruption. Therefore, cultural factors might influence representation and corruption (Swamy et al, 2001). Moreover, if citizens perceive female officials as less corrupt, they also tend to be so (Barnes & Beaulieu, 2019;
Swamy et al, 2001; Dollar et al, 2001, Esarey & Chirillo, 2013; Sundström & Wängnerud, 2016; Esarey & Schwindt-Bayer, 2018; Stockemer & Sundström, 2019). Esarey and Chirillo (2013) conclude that; “If voters, judges, and administrators are quicker to condemn and punish women’s corruption (compared to men’s), then women will probably be cleaner—but only in political systems that work to fight corruption” (366). For women in these countries, any detection of involvement in corrupt activities will lead to more severe punishment and removal from office (Esarey & Chirillo, 2013; Eggers, Vivyan & Wagner, 2017; Schwindt-Bayer &
Esarey, 2018). With this background of how female representation might reduce corruption, it is interesting to look at the variations within Sweden, as it is an advanced homogenous democracy to see if this pattern occurs (Sundström & Wängnerud, 2016).
Sweden is considered to be one of the least corrupt countries by the Corruption Perception Index (CPI) (Erlingsson et al, 2014; Transparency International, 2019). But a leading way is not a guarantee that corruption does not exist. A problem with the CPI is that it is mostly designed to investigate the central level of administration; it does not include the local levels of
administration and how it looks further down at the chain. Which is where we can find the most cases of corruption in Sweden (Berg et al, 2013). Nonetheless, Sweden is not only outstanding with its low levels of corruption, it is also outstanding in an international comparison of female representation in politics. The female representation in the Swedish parliament is about 40 percent women compared to the international average around 22 percent (Wängnerud &
Högmark, 2014). Sweden is, concludingly, one of the least corrupt countries as well as one of the countries with the most female representation in the national parliament (Sundström &
Wängnerud, 2016). This is interesting for my case because, in general, Swedish local politicians should have the same background and any variation would not be dependent on the cultural differences between states (country-specific attributes). Notably, if there is a far-reaching decentralization where local governments have high levels of autonomy, as in Sweden, the combination can lead to variation within countries' subnational institutions. Hence, the quality of government level differences within a country can exceed those between countries (Linde &
Erlingsson, 2013; Erlingsson & Lundåsen, 2019).
In this research field, there is a need to “clarifying how social and cultural attitudes combine with institutional variation to influence how much men and women in government engage in corruption” (Esarey & Chirillo, 2013: 387). As to my knowledge, no previous research has investigated gender and the individual level of local politician’s attitudes towards corruption in Sweden. Hence, are there any differences in attitudes towards corruption due to the gender of local politicians in Swedish municipalities? Especially if previous research has shown that female elected officials are less approving of activities that are considered to be corrupt and partial. In countries with greater equality between men and women, there should be less of a difference, as women have more participation. However, research has shown that women tend to be more risk-averse even when they are equal to men (Esarey & Chirillo, 2013; Frank et al, 2011; Rothstein, 2018). So even if female local politicians are on an individual level more or less corrupt than their male counterparts, are there any circumstances where they are even more negative? Therefore, these questions lead to the interest of investigating the relationship between gender and corruption on the local level in an already established democracy, like Sweden, as it is not yet fully explored. The research questions this study will answer is the following:
Do female and male local politicians have different attitudes towards corruption in Swedish
Purpose/ Aim
This research aims to investigate the relationship between gender and attitudes towards corruption within the Swedish municipalities on an individual level. More precisely, to make a quantitative ordinary least square regression analysis of the attitudes towards corruption and partiality from female and male local politicians in the 290 Swedish municipalities. This will be measured with answers from about 8500 Swedish local politicians elected 2010 to 2014, in the period 2012-2013 through the KOLFU 2012 survey (Karlsson & Gilljam, 2018). The KOLFU 2012 survey is selected as it is the largest and most recent survey on local politicians in Sweden that currently exists and contain questions regarding corruption and impartiality.
This enables for an understanding of female and male local politicians’ attitudes towards corruption and what might influence any relationship based on previous research. Interaction models will also be made to examine the accountability and risk aversion mechanism that might be an explanation for even more negative attitude towards corruption. The main focus will be to answer the hypotheses; 1) Female local politicians are more negative in their attitude towards corruption than their male counterparts in Swedish municipalities and 2) Female local politicians in Swedish municipalities who believe the accountability to be higher will be more risk-averse and have a more negative attitude towards corruption. Moreover, the data used in this research is great for the individual perspective of local politicians as it is a survey made in all of the Swedish local municipalities. This is applicable as previous research has demonstrated that there are reasons to believe that a female politician has a more negative attitude towards corruption than their male counterpart. Yet this has not been investigated in Swedish municipalities. The summary of this research is that it finds support for the two hypotheses.
Female local politicians in Sweden seems to be more negative in their attitude towards corruption than their male counterparts. When accountability and risk aversion is accounted for it makes female local politicians even more negative in their attitude towards corruption.
Limitations
Other important factors that this research would have wanted to include in its scope, but unfortunately not, are other possible explanations for understanding the relationship between gender and corruption. Due to limitations of time and data available it will be left for further research to investigate. Whether or not female elected officials are the fairer sex this thesis will not be able to tell. The motivation for any differences, biological attributes or social construction can only be resonated about, but not fully determined, as this thesis will not make any claims of causality. It should be noted that attitudes might influence differences in the
substantial corruption between gender, but the effects it has on actual level of corruption is not examine in this thesis.
Outline
The next chapter introduce previous research and theory on gender and corruption, and possible mechanisms as an explanation to these differences. A contextual background of Sweden concerning corruption and gender will be presented in chapter 3. The methodology, including case selection, data and operationalisation used for the regression analyses will be presented and discussed in Chapter 4. In chapter 5, the results will be presented with tables and figures in. This will be discussed in chapter 6 and reconnected to theory. The study will, in chapter, 7, end by concluding remarks and suggestions for possible further research.
2. Theoretical framework
In this section, the previous research and theory about female representation and its impact on corruption will be presented. It will begin in the first section to give an overview of combined findings on why females are considered to be less corrupt from aggregated levels to an individual focus. The second section will go more in depth of a possible mechanism that can explain the negative relationship between gender and corruption.
2.1 Gender and Corruption
In general, corruption is less profound in democracies, in comparison to non-democracies.
However, to some extent, corruption is present in all societies. Therefore, different settings can affect whether women’s representation will lead to lower levels of corruption in government (Esarey & Schwindt-Bayer, 2018; Esarey & Chirillo, 2013). There is no “quick fix” by providing a greater proportion of women in government to resolve the problems with corruption (Sundström & Wängnerud, 2016). Other variables than gender play a part when women’s representation makes a difference (Esarey & Schwindt-Bayer, 2018). Neither is corruption gender neutral in democracies (Stockemer & Sundström, 2019) even though it has been stated that “when a country performs well in gender equality, it performs well in quality of government and vice versa” (Alexander, 2018: 22). Swamy et al (2001) conclude that; “(a)in hypothetical situations, women are less likely to condone corruption, (b) women managers are less involved in bribery, and (c) countries which have greater representation of women in government or in
market work have lower levels of corruption” (26). Further, women in parliament are associated with lower levels of corruption (Frank et al, 2011) and are less likely to engage in both bribery and bribe-taking (Swamy et al, 2001) and seem to be more trustworthy (Dollar et al, 2001).
Concerning these gender differences, Swamy et al (2001) found, when measuring data from the World Value Survey in the 1980s and 90s across countries, that a man’s likelihood of answering
“never justified” of accepting bribery” were less than a woman. This is followed by Hernandez and McGee’s (2012) findings in a research of four European countries, Germany, Italy, France and Great Britain with sample population, that showed that women were more against bribery than men. Gender was a vital differential but also age, religious, cultural, historical, psychological and sociological factors were statistically significant for attitudes towards bribery. Although, the impact of female representation on corruption varies on the context.
Esarey and Chirillo (2013) found in their measured data from the World Value Survey that women were more disapproving of bribe-taking than men but only when there are democratic institutions. Moreover, in a regional survey of 85,000 Europeans, Bauhr, Charron and Wängnerud (2019) found evidence that female political representatives reduced both petty and grand corruption. In local councils with increased female representation, bribery experience decreases for both men and women but decreases the most among women. This follows by another research of 80.000 respondents in 21 EU countries that showed that women have lower greed corruption than men. Women were more involved in need corruption as a way to access basic public services. The only variable in the research that could explain the difference between greed and need corruption was gender (Bauhr & Charron, 2020).
The individual differences
At a more individual level, it has been reasoned that women have been brought up to be more honest and risk averse than men and that they feel that their probability to be caught is bigger (Eggers, Vivyan & Wagner, 2017; Esarey & Schwindt-Bayer, 2017). Or that they have been brought up to have higher levels of self-control (Swamy et al, 2001). Furthermore, Swamy et al (2001) resonate that it could be a “stylized fact”. Women are often the main caretaker of children and might therefore feel an urge to teach them appropriate values. Another explanation could be that women feel like they must follow the rules as they are there to protect them (as they are less physically strong and more vulnerable in society). If so, it might be the level of trust in society that affects the attitude of a politician.
Swamy et al (2001) made a survey of the Nordic countries, Denmark, Finland, Sweden and Norway, which all are at the top of women’s representation in parliament in the world. Even with more female representation it showed that there was a difference in attitudes towards corruption and accepting bribes between genders. Moreover, in a large research project conducted in Italy, Sweden, UK and US women were in every country under every condition more tax compliant. They also found the largest difference between the genders in Sweden, which is highest ranked compared to others by the UN Gender Inequality Index. Women in these countries were more prone to pay their taxes even when the risk acceptance was kept identical to men (Rothstein, 2018). Swedish women are also more positive towards ethics and codes of conduct within companies and more likely to report suspicious behaviour (Tillsammans mot korruption, 2020). Rothstein (2018) therefore argue that honesty is not automatically transformed with the improvement of equality. Women have more prosperity to tell the truth even when they are more equal to men. Importantly, not all agree that the differences are due to “biological justifications” and refer to the social construction of gender roles. As Swamy et al (2001) so firmly put, “there is a plethora of theories regarding the sources of gender differentials in crime, with potential applications to corruption” (53). Grimes and Wängnerud (2018) and Frank et al (2011) argue that corruption is not due to a lack of moral on the individual level or that women are not intrinsically more honest or averse to corruption than men. Rather, it is a result of behavioural norms, rational action and incentives that play a part in these differences between men and women. In some contexts, it is not unusual that sanctions are made if these behavioural norms are not fulfilled. Frank et al (2011) found in their laboratory experiment that women react strongly to detection. They explain it as women are less willing to engage with corrupt criminals and tend to let the corrupt transactions fail and are less likely to strike any successful corruption deals. Higher involvement of women can therefore reduce corruption, but it is not as simple as to address women as the fairer sex.
Summary
To sum up this section, research have found that there is negative relationship between women and corruption in democracies (Esarey & Schwindt-Bayer, 2018; Esarey & Chirillo, 2013).
Women are less prone to bribes or to accept bribery and are not as welling to engage in corrupt activities for their own private gain (Hernandez & McGee, 2012; Swamy et al 2001). A European study on the local level showed that increased female representation decreased the overall bribery experiences (Bauhr, Charron & Wängnerud, 2019). On the individual level there
be due to how the behaviour norms in our society construct gender roles (Frank et al, 2011;
Grimes & Wängnerud, 2018). However, it seems that women and men act differently concerning corruption regardless of the different explanations and that women have been demonstrated on both a national and a local level to be less positive to corruption. At least when it comes to bribes (Swamy et al, 2001; Esarey & Schwindt-Bayer, 2018; Esarey & Chirillo, 2013). Based on previous research it should be expected that female elected officials even on the local level in a democratic country like Sweden will have a more negative attitude towards corruption than their male counterparts. Therefore, the first hypotheses estimate the following:
H1: Female local politicians are more negative in their attitude towards corruption than their male counterparts in Swedish municipalities.
Different focus has been brought forth on the relationship between gender and corruption and what is causing the effect. The next section will review the main focus of the possible mechanism of accountability and women’s risk-aversion as an explanation of women’s more negative attitudes towards corruption.
2.2 Mechanism of Accountability and Risk Aversion
So how can we interpret gender differences in corrupt behaviour? To begin with previous research has found a negative relationship between corruption and women’s representation in high accountability democracies (Esarey & Schwindt-Bayer, 2018; Esarey & Chirillo, 2013).
The explanation for the negative relationship between women’s representation and corruption in high accountability countries is the mechanism of “risk aversion”. It is an institutional incentive that creates a relationship between the participation of female representatives and lower corruption as the culture concerning corruption has been stigmatised (Esarey & Chirillo, 2013; Eggers, Vivyan & Wagner, 2017; Esarey & Schwindt-Bayer, 2018). When the norm concerning corruption is stigmatised, it creates a higher risk of involvement for women but not for men. Thus, in countries with low accountability, female representatives will not have any substantial effect on corruption as it is not stigmatised in the same extent (Esarey & Schwindt- Bayer, 2018). However, other research has shown that there might be other explanations to the negative relationship between gender and corruption. One could be that women lack possibility of accessing corrupt networks or the “know how” to engage in corrupt practices (Bjarnegård, 2010; Swamy et al, 2001). This is especially the case in democracies when information of the existence of corruption is needed for any engagement (Esarey & Chirillo, 2013). It is labelled
as “the exclusion mechanism” and makes female representatives mobilise against grand corruption as they are not engaged or are not allowed into the corrupt networks (Bauhr, Charron
& Wängnerud, 2019; Goetz, 2007). Sundström and Wängnerud (2016) argue from the European context that it is more useful to see it as female politicians are lacking a rooted position in the social network of male party gatekeepers. They become excluded from any “sensitive exchange” and are not considered to be a part of the “recruitment process”. Another explanation is according to Klasnja and Tucker (2013) that women are more likely than their male counterparts to improve public service delivery. Which means that improvements could make any need to engage in corruption to decrease. It is labelled as the “women’s interest mechanism”
and argued as the reason female elected representatives can mobilise against petty corruption.
It is according to them the most common in Sweden. However, this thesis will primarily focus on the first mechanism presented of accountability and risk aversion.
Democratic system and accountability
As mentioned, the institutional features in democratic governments might play a big role in the culture around corruption (Hernandez & McGee, 2012). In democratic settings with high accountability, female politicians tend to be less corrupt as an effect of believing that it is more likely that they will be held accountable (Esarey & Chirillo, 2013; Eggers, Vivyan & Wagner, 2017). Furthermore, in a parliamentary system, on the highest level of government, it is riskier to engage in corruption than in the presidential system. In a parliamentary system voters can hold the elected official to account all the time and easier replace them. Therefore, there is a connection between lower levels of corruption and women’s representation in parliamentary systems as it would mean more risk which women are not willing to take (Esarey & Schwindt- Bayer, 2018). Another interesting contribution is that in a party-centric systems elected officials have a lower risk to engage in corruption, as it is more difficult to identify which politicians who have engaged in corruption (Esarey and Schwindt-Bayer, 2018). Hence, a proportional representation system with open list for candidates might have a greater effect on gender differences as it makes it easier to cast a personal vote (Eggers, Vivyan & Wagner, 2017). To win an open and competitive election enables different candidates who holds systematically different attitudes towards corruption (Klasnja & Tucker, 2013).
Voters response
Klasnja and Tucker (2013) found in their comparative research between a low corruption country, Sweden, and a high corruption country, Moldova, that Swedish voters respond to
corruption on a societal level. Voters react strongly to allegations of corruption and any behaviour deviating from the optimal performance or a bad record of corruption will be punished immediately. This strong reaction will halt the corrupt politician from entering politics as well as constrain the politicians in the political arena to involve in corrupt behaviour. This is likely to keep dishonest and incompetent politicians from running for office (ibid). Yet, Eggers Vivyan and Wagner (2017) could not in their research of accountability of politicians find that the perception of men and women’s similar behaviour was different. However, they could not rule out that any misconduct of female politicians was more visual by voters as they also found female voters to respond harder to corrupt behaviour from female politicians as well as their good behaviour than to male politicians. Female politicians also tend to be held accountable for any improvement or mistakes by female citizens (Bauhr, Charron & Wängnerud, 2019).
Especially in local councils where gender equality increases the feeling that female elected officials decrease petty corruption in health and education sector (Bauhr, Charron &
Wängnerud, 2019). Research have found women to be less tolerant of corruption in contexts where the spending on health and human services are high. In this setting it could be that women have more to lose when the system is corrupt as they depend more on the services provided.
Women exercise accountability through the polls by choosing another party that is clean from corruption scandals and allegations. In settings of low public spending women are not as unaccepting of corruption as it would not affect them as much anyway. Welfare-states reproduces a virtuous cycle where women are contributing to democratic accountability of corrupt politicians (Alexander et al, 2020).
A research of gender stereotypes in American elections showed that it is not always gender that influences voters, rather it is the political party (Dolan, 2014). Further, Barnes and Beaulieu (2014) argue that the higher ethic stereotype addressed to women can also help them depending on the preferences of the voter. If the voter prioritizes ethics before leadership, social welfare, education, health care etc, before foreign affairs, economy, security etc, women have an advantage. Moreover, male voters and voters who do not share the same partisan connection are less likely to suspect female candidates being engaged in any fraud activities (Barnes &
Beaulieu, 2014). Voters perception of women is that they are more risk-averse which reduces their suspicion of engaging in corrupt behaviour (Barnes & Beaulieu, 2019). If this is true, we cannot be sure. Regardless of whether women are more risk-averse or not, people’s perceptions are that they are (Ibid.; Esarey & Chirillo, 2013; Esarey & Schwindt-Bayer, 2018).
Gender stereotypes
Female politicians can allay citizens' concerns about corruption and be seen as a restorer of trust for governments if a scandal happens (Barnes & Beaulieu; 2019). It could be due to that women in positions of power are asking for monitoring of the state (Agerberg et al, 2018). Or the fact that women are characterized as more warm, compassionate, communal and genuinely concerned of welfare policies compared to men. Although, men are described as more competent and assertive in politics. On the one hand, gender stereotypes characterise women as more honest, ethical and trustworthy (Barnes & Beaulieu, 2014; 2019; Eggers, Vivyan &
Wagner, 2017; Ionescu, 2014). On the other hand, findings of political stereotypes demonstrate that politicians are considered to be more dishonest. An interesting finding is that there is mixed evidence when it concerns female politicians (ibid.). Thus, in general when there is a high risk of institutional sanctions people are less sceptical of politicians (Barnes & Beaulieu; 2019). So, do women benefit from the gender stereotype of being considered more honest? Even though greater representation of females has reduced gender stereotypes as a disadvantage for female candidates (Dolan, 2014) research have found that it is not always the case. Women will still be punished if they do not conform to stereotypes (Schwindt-Bayer and Esarey, 2018; Benstead
& Lust: 2018; Grimes & Wängnerud, 2018). Female politicians have a hard time to convey stereotypes, either they break the one of a leader (the male attribute of being in a power position) or the one of a woman (warm, caring, friendly and helpful). This can hurt elections for female politicians as well as their longevity in politics as they are portrayed with stereotypes that have nothing to do with them (Van der Pas & Aaldering, 2020; Morinière, Zarić & Docquir, 2017).
Moreover, the media can increase the cost of corrupt behaviour (Swamy et al, 2001; Easarey &
Schwindt-Bayer, 2018). It reproduces gender stereotype biases and normative expectations of women as well as portraying them according to a male perspective (Bromander, 2012;
Morinière, Zarić & Docquir, 2017; Van der Pas & Aaldering, 2020). Research in Sweden has found that female politicians not only are underrepresented in the daily news media but also as a group more audited by the media than male politicians. When women are not adherent of their stereotype of being more honest and compassionate the media will punish them. Especially scandals involving women are more covered by the media than scandals involving male politicians (Bromander, 2012; Säll, 2019). The systematically more targeted position has led to a gender difference in accountability and removal from office. Female politicians are forced to resign more often than their male politicians due to a political scandal in Sweden (Bromander, 2012).
Women’s attitudes towards corruption and risk aversion
Importantly, female politicians are extremely responsive to the political norms in their society.
Consequently, “the more that the society disapproves of corruption, the more women disproportionately express disapproval of corruption” (Esarey & Chirillo, 2013: 384). The explanation is according to Esarey and Schwindt-Bayer (2018) women’s stronger risk-averse behaviour of being detected in corrupt activities. The relationship between lower corruption and female participation is a by-product of differential treatment of women (both by voters and political elites). Due to gender discrimination, women have more incentives to adopt political norms concerning corruption in democracies (Esarey & Chirillo, 2013). Further, females are judged harder or even inferior to equally qualified men just because they are females. As a result, they always have to work harder and perform better than men (Säll, 2019). If women engage in corruption and are expected not to, they will face strong hostility when they break out from their subscribed gender roles (Benstead & Lust, 2018). The stereotypes reinforce gender inequality and affect women’s chances of being elected but also once they step into office (Benstead & Lust, 2018). As a result, women become more risk-averse to stay in office (Esarey & Chirillo, 2013; Esarey & Schwindt-Bayer, 2018; Barnes & Beaulieu, 2019). When punishment is harder for women, women will profess greater aversion compared to men. It is cultural discourages that makes them suffer more severe consequences than their male counterparts (Esarey & Chirillo, 2013). Risk aversion seems according to Schwindt-Bayer and Esarey (2018) be the strongest explanation of the differences between men and women. The authors argue that there is significant evidence that supports this in high accountability countries. Although Bauhr, Charron and Wängnerud (2019) argue that norm-compliance and risk aversion only provide partial explanations to the differences in corruption between men and women. Surprisingly, elite women might differ from the pattern of women to be more risk- averse than men (Bauhr, Charron & Wängnerud, 2019). In settings where women have more access to informal and formal political networks their lower propensity of engage in corruption might be reduced (Klasnja & Tucker, 2013). Women that repetitively engage in bribery exchange, adapt a behaviour more similar to men (Frank et al, 2011). The significant differences in gender stereotypes would decrease if female leadership becomes normalized. Then we can no longer expect that the female elite will have an improving perception of corruption (Barnes
& Beaulieu, 2014).
Summary
In a high accountability democracy female elected officials become more risk averse in order to stay in office as they must stay cleaner - otherwise they are likely to be replaced. This is even more likely when there is a parliamentary system as voters can easier replace their candidates (Esarey & Schwindt-Bayer, 2018). If the public preferences stigmatise corruption it creates pressure on female elected officials, who reinforce a less acceptable attitude to corrupt behaviour (Esarey & Chirillo, 2013). Gender discrimination reproduces a stereotype of female politicians as more trustworthy and honest, any deviation from that behaviour becomes punishable (Esarey & Schwindt-Bayer, 2018; Schwindt-Bayer et al, 2018; Benstead & Lust:
2018; Grimes & Wängnerud, 2018). Moreover, as voters are likely to react when corruption do occur, the accountability and risk aversion mechanism makes women more restrained to engage in any corrupt activity (Klasnja & Tucker, 2013; Esarey & Chirillo, 2013; Esarey & Schwindt- Bayer, 2018). Conclusively, we should a expect a negative relationship between female representation in government and corruption in high electoral accountability country. The explanation is that women are more risk averse when corruption is stigmatised, both compared to men and other women who are not held accountable at the same extent (Esarey & Chirillo, 2013; Eggers, Vivyan & Wagner, 2017; Esarey & Schwindt-Bayer, 2018). So, the more accountability the female elected official experience should create a more negative attitude towards corruption. Even though the previous research has been on national level, the mechanism of accountability and risk aversion should be expected on local level of politics between gender and attitudes towards corruption. Especially in Sweden, as it is a low corruption country with corruption stigmatised on a societal level (Klasnja & Tucker, 2013) and where the local level of politics is meaningful for the welfare (Bauhr, Charron & Wängnerud, 2019). The second hypotheses estimate the following based on accountability and risk aversion:
H2: Female local politicians in Swedish municipalities who believe the accountability to be higher will be more risk-averse and have a more negative attitude towards corruption.
3. The contextual background of Sweden
This section will give a short introduction to the background of Sweden and its political system and what previous research have found about corruption and female representation.
The Swedish multilevel system is described to have the shape of an “hourglass”, the local and state levels are strong, and the regional level is weak and quite anonymous (Erlingsson &
Lundåsen, 2019). The largest group of elected representatives in Sweden can be found at the local level as local elected officials. For the Swedish elected representatives, the party affiliation is strong (Karlsson & Gilljam, 2014a) and politics on both local and regional level is, like the national level, highly politicized. Thus, the local levels and the national levels are much alike and can have similarities. However, a big difference between local politicians and national politicians in the parliament is that the former are mostly leisure politicians, and the latter are full-time politicians with high wages. The parliamentarians also have authority to regulate the municipalities and regions scope as they answer to the national level. Needless to say, the national level has more authority than the local levels (Karlsson & Gilljam, 2014b).
As mentioned, Sweden is one of the countries with the highest number of female representatives in politics (Karlsson & Gilljam, 2014a). Since 1994, when “every other lady’s” was introduced in the Swedish election system, the number of women engaged in politics and being elected have been steady at 40 percent both national and local elections (Wängnerud & Högmark, 2014). The parliament and the government in Sweden have a more descriptive female representation than the other levels (Säll, 2019; Karlsson & Gilljam, 2014a). In the municipality assemblies the representation of women differs between municipalities; during the mandate period 2010-2014 it was as low as 29 percent in one municipality and it was as high as 58 percent in another (Wängnerud & Högmark, 2014). Overall, men are still dominating the municipality assemblies and the local level of equal representation is not as high as the national one (Säll, 2019). In the local elections, the highest percentage of female elected officials can be found in the Social Democrats Party and the lowest in the Swedish Democrats Party (Karlsson
& Gilljam, 2014a). The position of Chairman in Municipal Council in Swedish local politics is the most male-dominated. As a result, male dominance exists in higher positions in local councils (Säll, 2019).
The local governments in Sweden have the responsibilities to provide many of the important welfare inquiries that their citizens need in their everyday life. By providing these tasks, education, water and sewage, social services etc, the municipalities levy taxes (Erlingsson &
Lundåsen, 2019; Karlsson & Gilljam, 2014b). Swedish municipalities are financed mainly by the income of tax and the GDP for municipalities is higher than the central government, 20 percent compared to 7 percent. Hence, Sweden ranks high in decentralization indices of local
government autonomy and capacity (Bergh et al, 2020). The local politicians are closer to public services and more active in the output phase of politics. They are also closer to their voters and have more mobility in participatory activities (Karlsson & Gilljam, 2014b). Local governments are in direct contact with citizens (Erlingsson & Lundåsen, 2019; Karlsson &
Gilljam, 2014a; 2014c) and can thereby increase or decrease social trust (Erlingsson &
Lundåsen, 2019). This area of responsibility is a risk area of partial impact (Bergh et al, 2013).
Besides this relationship to the voters makes the local executive organs throughout Sweden different from each other (Karlsson & Gilljam, 2014b).
According to a report from the Swedish National Council of Crime Prevention, the Swedish municipalities are in the risk zone of corruption. Both through direct bribes and subtle bribes, the first is a monetary exchange for an advantageous decision, and the other is based on a more long-term relationship between those involved. Especially friendships and close ties, so-called
“friendship corruption”, makes it challenging for public officials as it quickly can move from a grey area to an illegal one. Especially during dinners, events and similar where oral agreements are made. This is more common in smaller municipalities as the officials are living closer to the citizens and subconsciously benefiting the local businesses (Hols Salén & Korsell, 2013;
Gustafsson & Sørensson, 2012). Rural citizens perceive less corruption compared to urban citizens. Thus, the bigger cities are more affected by corruption, so it seems that demographic effect and income matters of corruption. However, there is no evidence that the three biggest cities Stockholm, Gothenburg and Malmö differ from each other (Hols Salén & Korsell, 2013;
Gustafsson & Sørensson, 2012). A possible explanation is that larger councils lead to more corruption and bribery when there are larger administrative ties (Hols Salén & Korsell, 2013;
Berg et al, 2017). The incentives for monitoring the elected representatives could be weaker in larger councils than in smaller councils (Ibid). Another explanation for the variation in size is that local politicians from larger municipalities have more knowledge of what is classified as corrupt activities in comparison to the smaller municipalities (Gustafsson & Sørensson, 2012).
The media is also less active in investigating local levels (Erlingsson, Bergh & Sjölin, 2008).
The current development in Sweden is that the local media have become weaker and decreased in numbers (Gustafsson & Sørensson, 2012). This is argued by Transparency International Sweden (2012) as the main determent of differences between the Swedish municipalities.
In Sweden, no coherent legislation of corruption exists, it is criminalised through separate
partial behaviour can also change with public opinion (Sjölin, 2014). Additionally, The Swedish Agency for Public Management estimates that it is the absence of ethics, codes of conduct and review of the internal culture in the municipal sector that has led to an increased risk of corruption (Hols Salén & Korsell, 2013). Further, in a report from The Swedish National Council for Crime prevention, they conclude that the most corrupt elected representatives can be found in a longer appointment and middle-aged groups. The report also concluded that middle-age women and men are the most active in all different sectors. Thus, even if middle- age women are the most active in corruption cases they are not involved in as many cases as middle-aged men. Middle-aged civil servant men who hold a power position are in the majority of corruption cases received by the Swedish Prosecution Authority. According to them power affects more than gender in the cases of corruption (Hols Salén & Korsell, 2013). Lastly, the Swedish Government has declared that the municipal boards have the initial responsibility for internal control. However, the revision has a responsibility to make sure that the internal control from the boards is to satisfaction. The revision is an independent control organ which is for the citizen an important democratic instrument for accountability as it offers review and audit of the municipal activities (Gustafsson & Sørensson, 2012).
4. Method
This section will present the methodology used to interpret the results. It will start with a discussion of case selection and its relevance for the selected data in this research together with the selected method. It will be followed by a presentation of the discussion of the data used and operationalisation of the variables and how they are measured.
4.1 Case selection
The further motivation of Sweden as case selection, Dahlström and Sundell (2013) have made three conclusions based on previous research why Sweden is an excellent selection when looking at corruption on a local level: 1) Corruption exists in Sweden even though in comparison it is low levels of corruption. 2) The local level in Sweden seems to be more corrupt than the national level. This is not usually the case as the national and supranational levels have been considered to be more corrupt. Besides, in the European context, it is a citizen concern located on the national level (Bergh et al, 2013). Even in comparison with the other Nordic states, the local levels in Sweden are more corrupt (Dahlström & Sundell, 2013). 3) The in-
country variation is large (Dahlström & Sundell, 2013). Countries with the model as the Nordic ones the local level is favourable as the local governments are strong (Erlingsson & Lundåsen, 2019) and highly decentralized in international comparison (Karlsson & Gilljam, 2014b).
Additionally, the local level has a great variation of female representation and it might affect the results (Wängnerud & Högmark, 2014). Notably, this research will not be able to compare the different levels of female local politicians in different municipalities.
Further cultural factors have been pointed out to affect representation and corruption (Swamy et al, 2001; Hernandez & McGee, 2012). For that reason, it is relevant to construct a study in Sweden where the social terms should be homogenous (Sundström & Wängnerud, 2016). By having a homogeneous country as an object of analysis it controls for the political culture and socioeconomic development in the different municipalities. Sweden has small within-country variation regarding the engagement of civil society, economic equality, literacy, education and the egalitarian character (Erlingsson, Bergh & Sjölin, 2008).
Sweden is also optimal as a case selection by being a high accountability democratic country (Esarey & Chirillo, 2013; Klasnja & Tucker, 2013). Further some context has been lifted as strongly associated with lower-level perceived corruption when female representation increase:
“(1) where corruption is not an institutional norm, (2) where freedom of the press is respected, (3) in parliamentary rather than presidential systems and (4) under personalistic rather than party-centered electoral rules” (Esarey & Schwindt-Bayer, 2018: 661). The three first fits perfectly in Sweden as a case, thus one could argue that the fourth one is not a perfect match.
The institutional system of Sweden allows individuals to hold elected officials accountable at the polls and the party list system could make it possible to select candidates if one is dissatisfied with a candidate higher on the list. However, as all Swedish political candidates are elected and part of an open list which is decided by the parties, the party itself can decide the ranking and eligibility of the candidates (Säll, 2019).
This thesis will use a quantitative method with ordinary least square (OLS) regressions on a survey made on local and regional politicians in Sweden (KOLFU 2012) to answer the research question. This will be done in two parts, the first relates to the first hypothesis and will investigate how attitudes towards corruption and gender correlate to answer the first hypothesis.
The second part will investigate the accountability and risk aversion mechanism to answer the
of variation within the group of women. This is because there is an interaction between the independent variable, gender is “interactively involved” with another independent variable (Lewis-Beck & Lewis-Beck, 2016).
As the research is based on a survey, it is appropriate to use a quantitative method. But also, as it has an aim to generalize the results. The quantitative approach reduces the risk of subjectivity.
Although, it still requires a thorough insight to the contextual background. The statistical significance will be one estimate for verifying or falsifying the hypotheses along with theoretical reflections. Quantitative methods are also useful when establishing causality, thus this research will not have the possibility to establish any direction of causality, but it will give a more descriptive analysis of Swedish local politician’s attitudes towards corruption. This research aims to move from the aggregated levels of female representation towards the individual level of attitudes. It will give an overview of local politicians’ attitudes and if there is any relationship between gender and attitudes towards corruption. That will say between the independent variable of gender and the dependent variable of attitude towards corruption.
4.2 Data
The material used in this thesis to measure gender and attitudes towards corruption is the Municipalities and Regions Survey made 2012 and 2013, called KOLFU 2012, in Sweden (Karlsson & Gilljam, 2018). This survey includes responses from about 78 percent of the 13 361 active politicians at the time in the Swedish municipalities and regions (Dahlström &
Sundell, 2013). This can be seen as satisfactory because of the decrease of respondents in the developed world in the last decades (Bergh et al, 2020). The absolute number of respondent regional and local politicians is 10 491 (Karlsson & Gilljam, 2018) but the number of observations in this research will be around 8200-500 respondent local politicians. This is due to what they have answered in the survey and what is useful for this research. In comparison to other surveys made in Sweden, this one is particularly large and includes all different levels of local politicians, as chairman, vice-chairman and members of the local councils. From each municipality there are between 13 and 68 respondents, depending on the size of the municipality (Dahlström & Sundell, 2013). It also measures both corruption and impartiality, which is great for this research as it will capture the individual perception of them. Important to mention is that the answer from respondents in this survey are measured through both perception and experience elements. By using the KOLFU 2012 survey it gives an insight into the experience
and individual perception of politicians from 290 Swedish local councils during the election period 2010-2014. This is unique access to the local politician’s attitudes and what may or may not affect their view on the individual level and what they consider to be acceptable behaviour concerning impartiality.
The KOLFU 2012 has its limitations as it will not be possible to evaluate any effect of size of the municipality and if there are any differences between urban and rural areas. This was done to protect the anonymity of the respondents, as it would be easy otherwise to track each answer to a certain politician. It holds sensitive information as it includes direct questions of corruption and impartiality, among other information a politician would want to keep private. Besides, it is already difficult to measure corruption and impartiality due to the social norm within modern societies that corruption is illegal, and impartiality is important (Dahlström & Sundell, 2013).
Another way to measure corruption could have been through media and official crime statistics.
The benefit with KOLFU 2012 survey is that Dahlström and Sundell (2013) have validated the questions by correlating them to legal bribery charges, newspaper articles and a similar survey with less than 1000 respondents (Bergh et al, 2020). To measure convictions or committed crimes is not as useful according to Dahlström and Sundell (2014), as we become limited to the corruption that has been discovered.
Measuring corruption through perceptions and experiences have its limitations. Perceptions might be biased and not correspond to reality. Further, previous results from survey data research have shown that the individual experience of being asked to give bribes from Nordic citizens is extremely limited. And only a few cases have been brought to court (Erlingsson et al, 2014). Therefore, experience-based measures are risky as it would only capture visible corruption. This is solved by questioning the politicians themselves and by evaluating a question based on what they consider to be legit or not it can move past the prolonged problem of how to measure. Nor does it matter how they perceive or have experienced corruption currently or before, as it is their view of it that becomes important (Teorell, 2014). What they say and what they do are two different things, one might say that one is against any partial and corrupt behaviour but still act in a certain way. Additionally, it highlights nontrivial within- country differences regarding violations of the norm of impartiality (Erlingsson & Lundåsen, 2019). Nonetheless, a problem that prolongs is how to ensure that the respondents will give truthful answers as there is a risk that the respondent will answer what is desirable and what is expected of social norms. A possible solution is to ensure that the answers are confidential,
which is the case with the KOLFU 2012 survey. The conclusion is that the survey is almost “a real-world phenomenon”, and it needs “further attention in empirical studies'' (Dahlström &
Sundell, 2013: 22). Thus, it is not perfect, nor can it be. By using the KOLFU 2012 survey it is limited to the questions asked in the survey.
4.3 Operationalization and variables of interest
To measure how Swedish local politicians' attitudes are concerning corruption and what those activities might be demonstrated in real life, and operationalization of its concept must be made.
Corruption is the opposite to the Quality of Government, that will say “impartiality in the exercise of public power” (Rothstein & Teorell, 2008: 166; Linde & Erlingsson, 2013: 587).
Further impartiality rules out patronage, clientelism and discrimination, which all is classified as corrupt behaviour. The output side of corruption can minimize the legitimacy and trust of a democratic institution. As legitimacy is characterized by impartiality and procedural fairness when exercising civil servants perform their tasks. Corruption is not impartial and is based on a political or personal relationship (Rothstein & Teorell, 2008; Erlingsson & Lundåsen, 2019).
Further, it can be described as “abuse of public office for private gain” (Bergh et al, 2020: 9, Dahlström & Sundell, 2013: 7; Alexander, 2018; Rose-Ackerman, 1978). The power holder uses the public authority for one's selfish interest and gains. Low levels of QoG can increase negative experiences from government institutions and undermine social trust as the public good (Alexander, 2018). Unfair treatment, corruption and partisan behaviour can also harm the sense of belonging to a community (Erlingsson and Lundåsen, 2019).
Dependent variable
In this research, the dependent variable will be any tendency to corrupt behaviour. It will be measured through the local politicians’ attitudes concerning partial behaviour. It will be measured by the following questions from the KOLFU 2012 survey:
- A public servant requests a fee to perform a service that is actually part of his / her duties
- An entrepreneur offers a gift or service to a public servant in connection with a procurement
- A teacher in a municipal school provides extra tuition to a close relative's child during regular working hours
- A public employee offers a relative a job even though the person lacks the right qualifications
- A teacher in a free school provides extra tuition to a close relative's child during regular working hours
The respondents could answer these questions on a scale from 1 to 7, where 1 was never acceptable and 7 was always acceptable (Karlsson & Gilljam, 2018). This variable has been coded to a dummy variable so that every step from never acceptable is considered to be
“acceptance of corrupt attitudes” in all five scenarios. Surely, there is a difference between answering a 2 and a 7 on the scale. However, this is done as the social norm in Sweden is that corruption and partial behaviour is not acceptable and that, as always with corruption perceptions, the politicians might answer what they know to be acceptable. Which would be the answer “never justified”. Nonetheless, all five questions are a measurement of impartial behaviour and if one of the scenarios is being considered to be more acceptable it could demonstrate that the view of impartial behaviour is not so strict by that individual. By relocating all of them at least once, answering something else than “never justified” will give a more overall view of how the local politicians view impartial behaviour.
Independent variable
The main independent variable to measure attitudes is the answers to the question of impartiality is gender. As it only has two values, male or female, it is used as a dummy variable.
Control variables
The most common to engage in corrupt activities are middle-aged men and women (Hols Salén
& Korsell, 2013) which makes relevant to control for the duration of appointment. However, as there is no such variable, but it is possible to control for age. Age is measured through five categories: 18-24, 25-34, 35-49, 50-64 and 65 – 80. Being middle age could also mean one has been appointed for a long while and probably holds a power position. But also, as females become integrated into the internal networks, it could imply that longer time in office could increase corruption (Esarey & Chirillo, 2013). This could be further investigated by controlling for education. Education is measured in four categories: elementary school, upper secondary school, university degree and PhD. To control for party affiliation is also important as it in both international and national research have indicated that parties located to the right spectrum have
is accepted corrupt behaviour within each party. Party affiliation consist of eleven categories but only the eight biggest will be measured in this thesis: The Left Party (V), the Social Democrats (S), the Green Party (MP), the Centre Party (C), the Christian Democrats (KD), the Liberal Party (L), the Moderate Party (M) and the Swedish Democrats (SD).
In the KOLFU 2012 survey the three questions, position in the local council, majority/opposition party in local council and the perseverance of municipal revision, were found that could have implications on high accountability and risk aversion. These have been analysed in interaction models as the aim is to understand how they interact with gender. As it is hard in this particular research to distinguish between what is being accountability and what is being risk-aversion they are measured together. Risk aversion can be understood as a result of high accountability. Therefore, it is relevant to check if the politician is a part of the majority rule or the opposition. This could have an impact on how audit one might be by others in the local council and the voters in the municipality. It is measured as a dummy variable, either the politician belongs to the majority or the opposition. Research has found that women on an elite level are not any different from their male colleagues it is relevant to control the position in the local council (Frank et al, 2011; Bauhr, Charron & Wängnerud, 2019). This variable has three values, the local politician is a chairman of the local council, vice chairman/second vice- chairman or member of the council. Being in a power position enables for more audit of the female politician. It should be a relevant measurement of women being more risk averse as they are more subjected to social norms and less willing to accept bribes if they can be detected and punished (Esarey & Chirillo, 2013). The variable of position in the local assemblies will be coded as a dummy variable. It will be separated between chairman and vice chairman/second vice-chairman in the local council and only being a member of local council. As every municipality has a revision it is also relevant to control for when the revision identifies deficiencies and alleged deficiencies have led to repercussions by the revision. A functioning audit of the municipality’s activities can be considered to be effective measure against corruption and have a strengthening effect of the routines of internal control (Hanberger, 2014).
By having more revision control in a municipality, it could also mean that they have a more thorough internal control of the activities within that council. The role of the revision is to supervise and make sure that the municipal assembly's internal control is to satisfaction (Gustafsson & Sørensson, 2012). Female local politicians should become more risk averse and have a more negative attitudes towards corruption when the municipality have a more active revision. The revision variable is measured as an interval between the value 1, the revision
remarks have to a small extent led to repercussions, and 5, the revision remarks have to a large extent led to repercussions. Notable, when used in the interaction model the values 1 to 3 and 4 to 5 have been separated and made into a dummy variable.
Other variables that could have an influence on the relationship between gander and attitudes towards corruption1 is how the local politicians perceive the media’s monitoring of the local assembly2. It might tell us something if the media's monitoring affects the female politician’s attitudes towards corruption. The Swedish Agency for Public Management report shows that there are different perceptions of corruption in different municipal boards. The ones more focused on social welfare and education are found to have almost no cases whilst the boards focusing on IT, construction, business and tourism are believed to have more frequently occurring cases of corruption (Gustafsson & Sørensson, 2012). Therefore, it could be relevant to control for which board (s) a respondent is assigned to by being elected in a municipal assembly. This will be made into a dummy variable where boards that are focusing on social welfare and education are separated from other boards. Further, it could be relevant to control for the place where the individual and parents mainly grow up might affect the relationship as it could indicate that one is not a part of the homogenous networks that consists of politics.3 Research has shown that ethnic diversity is good for ensuring high quality of government (Gustafsson & Sørensson, 2012). But also, the main occupation in addition to the position as an elected official in municipalities (the most local politicians are appointed part-time or as leisure politicians). Another occupation could open up for other possibilities of building networks.4
Concludingly, a regressions analysis prefers to measure variables in a numeric precision, a categorical variable like gender is easily made to a dummy variable. Variables such as education, age and party affiliation cannot be measured in an interval, they are categorical variables in this survey and would therefore be measured as it. By coding from 0 to 1 there is no range for a negative value (Lewis-Beck & Lewis-Beck, 2016). Even other variables which do have an interval scale are coded as dummy variables. By doing so the variation might not
1 These control variables where later found to not have any substantial meaning for the result. However, regression models can be found in table 6, 7 and 8 in the appendix.
2 Interval variable with 10 values, can be found in table 6 in appendix
3 Categorical variable with 7 values can be found in table 7 in appendix.
resemble the real world as much as it would have otherwise. However, it by doing so it clarifies and simplifies the analysis of the results and to be illustrated in graphs.
4.4 Descriptive data of mean
Data description of the mean in table 1 demonstrate that there are already some differences between gender and attitude towards corruption. It also allows us to see some differences within the other variables. The description of mean gives further implications for what this research will investigate in the next section, results.
Table 1: Mean of variables
All Male Female Dummy
Attitude Corruption
.2671603 .3035214 .215446 yes
Female .4302603 yes
Chairman .0787326 .0835335 .0722981 yes
Majority .543856 .5313673 .5609106 yes
Revision .5776029 .5580261 .6053282 yes
Age 3.700734 3.750502 3.634833 no*
Education 2.434945 2.3555 2.540553 no*
Party affiliation*
V .0553808 .0532129 .0582761 yes
S .3687923 .335174 .413029 yes
MP .0557621 .0493641 .0642588 yes
C .1050424 .104585 .1056947 yes
KD .0456582 .0485274 .041879 yes
L .0746354 .0746319 .0746732 yes
M .2252407 .2443106 .2000886 yes
SD .0315509 .0455154 .0130733 yes
N 8500- 10500 4000-6000 3500- 4600
*The Left Party (V), the Social Democrats (S), the Green Party (MP), the Centre Party (C), the Christian Democrats (KD), the Liberal Party (L), the Moderate Party (M) and the Swedish Democrats (SD).
Education and age are not coded as a dummy in table 1 but will be measured categorical in regressions.
Notable, there are more observations in the group of men as a result of the Swedish local politics being dominated more by men than women. This is recurrent through all tables and figures.
5. Results
This section will present the result in two different sections to answer the research question and hypothesis. The first part will cover attitudes towards corruption among female and male local politicians and the second part will investigate the accountability and risk aversion mechanism.
5.1 Attitudes towards Corruption
To test the first hypotheses, this section will contain two regression models. The first bivariate regression will illustrate the relationship between gender and attitude towards corruption. The second, a multiple regression, will illustrate that relationship when controlled for education, age and party affiliation. Graphs will be presented to illustrate the relationship.
In the first regression in Table 2, we can see the differences in attitudes between female and male local politicians. The bivariate regression demonstrates that male politicians’ attitudes towards corruption are on average 0,304. Male local politicians are read as the constant as it is a dummy variable in which male is 0 and female are 1. Female politicians are on an average - 0,0873 less acceptable than the value of the constant in their attitude towards corruption than their male counterparts. This finding can therefore implicate that men and female local politicians have different attitudes towards corruption and what is considered to be an acceptable behaviour. In sum, the former being less negative and the latter more negative in their attitude towards corruption. Below, in figure 1, this relationship is presented in a graph.
Table 2: Female representation and attitudes towards corruption
Model 1 Model 2
Female -0.0873*** -0.0887***
(0.00973) (0.00976) Education
Elementary 0
(.)
Secondary 0.0204
(0.0162)
University 0.0265
(0.0162)
PhD 0.0343
(0.0372) Age
18-24 0
(.)
25-34 -0.0762**
(0.0380)
35-49 -0.219***
(0.0348)
50-64 -0.282***
(0.0344)
65-80 -0.283***
(0.0354) Party
V 0
(.)
S 0.0166
(0.0215)
MP 0.137***
(0.0284)
C 0.118***
(0.0246)
KD 0.0807***
(0.0300)
L 0.0864***
(0.0264)
M 0.0905***
(0.0224)
SD 0.0992***
(0.0341)
cons 0.304*** 0.467***
(0.00624) (0.0420)
N 8477 8477
Standard errors in parentheses. Statistical significance * p < 0.1, **p < 0.05, *** p < 0.01.
This first regression and figure 1 has not yet been controlled for by the other variables which are visible in regression model 25. When controlled for education, age and party affiliation the relationship between gender and corruption slightly increased, almost none from -0.0873 to - 0.0887. This could imply that there is no negative or positive bias by including education, age and party. Or that the control variables are not so meaningful for the result and gender is the most explanatory variable. For party affiliation almost all of the values have statistical significance of more positive attitudes towards corruption. This is also a pattern for the women but as they are -0.0887 negative towards corruption it reduces the increase from party affiliation in most cases. However, not in all, belonging to the Moderate Party (M) or the Swedish Democrats (SD), does not make female local politicians less negative. Although, they are less negative than the male local politicians belonging to the same party group. Party affiliation, therefore, seems to have an effect on the result. Inevitably, despite that difference, female local politicians are in general less positive towards their male counterparts. Moreover, education is not statistically significant in any different education category and could imply that it is not of the greatest importance for attitudes towards corruption. Age seems to have a negative relationship with corrupt attitudes. The higher age of the respondent, the more negative attitude towards corruption.
Before moving on, some reflections of the accuracy of the regressions in this research. Both the first and second regression have a rather low adjusted R-square value. Thus, the R-square value is only a prediction of differences and with the hypothesis made we can already expect a difference based on the average mean. But also, attitudes towards corruption is a complex matter to measure and is in this research roughly drawn, which could mean that other explanation can have influence. The R-square values on individual level will be lower as there are more possible measurement errors than on an aggregated level. Nonetheless, as this thesis is investigating the individual perception of attitudes towards corruption, the standard errors are better to evaluate the accuracy of the regressions and are therefore only included when displaying regressions. The standard errors values are low which could indicate that the results from the data are valid and have an explanatory value. In summary, it seems that there exist reasons to believe that the first hypothesis of a difference between female and male politicians’
attitudes towards corruption is verified. To get a better illustration of how education, age and
5 Some variables such as media, division of board, background and main occupation are, as mentioned in the method section, not presented in the result. This done as it did not seem to be of any substantial meaning for the
