The prudent entrepreneurs: women and public sector innovation

ABSTRACT Despite the large literature on gender differences in politics, there have been relatively few empirical studies testing the effects of gender in public administration. This paper examines how male and female public managers show attitudinal differences toward innovation in the public sector. We hypothesize that male and female managers differ in three aspects. Firstly, female managers are more result-oriented than rule-following, and more oriented toward societal interests. Secondly, female public managers are more open to new ideas and creativity, and more willing to challenge the status quo. Yet, thirdly, female leaders are less eager to take risks when would-be innovations may put their organizations in peril. Thus, we argue that female managers are more prudent and entrepreneurial than their male counterparts. We test these hypotheses using a data set of 5,909 senior public managers from 20 European countries. The results of multilevel model analysis find statistically significant gender differences in attitudes toward innovation. Despite the small size of gender impacts, our findings challenge prevailing stereotypes on women’s entrepreneurial attitudes.


Introduction
Innovation in the public sector allows for 'revolutionary improvements' (Maranto & Wolf, 2013, p. 238) in normal times and solving 'wicked problems' (De Vries et al., 2016, p. 164) in times of crisis. Yet innovating in the public sector is more difficult than in the private sector (Laegreid et al., 2011). This is mainly because of various differences between the two sectors in the external environment and internal constraints for management in the public sector, such as a lack of mechanisms to use monetary incentives and the strength of media scrutiny (Rainey, 2009). Nevertheless, many public administrations experience constant innovation (Demircioglu & Audretsch, 2017, 2018Walker, 2014). A growing literature tries to understand why some public bureaucracies innovate more than others (for a recent review, see De Vries et al. (2016) and Demircioglu, 2019). This paper examines to what extent male and female public managers differ in terms of their attitudes toward public sector innovation.
In particular, this paper focuses on three attitudes related to innovation: openness to new ideas and creativity, willingness to take actions that might upset the status quo, and risk acceptance. These attitudes have been found to be key for launching innovation (Vigoda-Gadot, 2009) and determine the extent to which there is an innovative-oriented culture (Laegreid et al., 2011) in a given public administration.
This paper aims to contribute to this literature by exploring the effects of gender on innovative-oriented culture using a multilevel analysis based, first, on the responses of 5,909 public managers from 20 countries from the COCOPS (Coordinating for Cohesion in the Public Sector of the Future) Executive Survey on Public Sector Reform in Europe ( Van de Walle et al., 2016); and, second, on the characteristics of national public administrations, constructed with the opinions of over 1,200 experts in the Quality of Government (QoG) Expert Survey (Dahlström et al., 2015).
After controlling for country-, organizational-and individual-level factors, the results of our analysis show there exist statistically significant gender differences in innovation-related attitudes. Although the magnitude of the gender differences is not large, many of them run counter to conventional views on attitudinal gender differences in public sector innovation. First, our study finds that female senior managers are more likely to be result-oriented and more willing to serve societal interests than their male counterparts. Second, female senior managers are likely to be slightly more open to new ideas and creative solutions and show more willingness to change the current situation. However, and thirdly, female managers show slightly more risk-averse attitudes than male managers.

Theory
Past research on bureaucracy and organizational management suggests two different perspectives regarding the effects of personal and demographic factors, such as gender, on public managers' beliefs and attitudes (Christensen & Laegreid, 2009;Egeberg, 2012Egeberg & Stigen, 2018). In the demographic perspective argument, scholars emphasizes the importance of personal factors such as socio-economic background for managers' beliefs and attitudes (Egeberg & Stigen, 2018Pfeffer, 1983). Alternatively, in the structural perspective, the effects of demographic and personal factors, including gender, are considered as less important, because they are strongly constrained by structural and organizational factors through a socialization process (Christensen et al., 2012;Christensen & Laegreid, 2009;Egeberg & Stigen, 2018). The purpose of this study is to test the gender effects in civil servants' attitudes toward innovation, which is in line with the demographic perspective. We argue that, despite organizational constraints, male and female public managers show different attitudes toward innovation. Findings from our multilevel model provide empirical support for our argument. However, we find that such gender differences are subtle, contrary to our expectation. We suspect that such diminished effects may be due to institutional and organizational constraints on the effects of gender on public managers' attitudes toward innovation.
The studies of gender differences in the public sector have traditionally focused on the political level. However, the recent studies on gender differences go beyond the political sphere. Several studies have explored gender differences in public bureaucracy, showing that gender affects behavior and attitudes (Grissom et al., 2012;Hamidullah et al., 2015;Nielsen, 2015), as well as organizational performance and behavior (Opstrup & Villadsen, 2015;Suzuki & Avellaneda, 2018). However, studies on gender differences in public bureaucracy are still few compared to the studies on gender effects in private organizations and politics (Hamidullah et al., 2015). Furthermore, most studies focus on the theory of representative bureaucracy (e.g., Meier & Nicholson-Crotty, 2006). In particular, little is known about how the gender of public managers is linked to their attitudes toward innovation. 1 We think that the scarcity of studies on gender and attitudes toward innovation may be due to the strong assumption of Weberian bureaucracy (Weber, 2009), i.e., that 'the Weberian public manager is a genderless bureaucrat' (Opstrup & Villadsen, 2015, p. 292). Compared to the political/electoral arena, where scholars have identified significant influences of gender on the attitudes and behavior of elected officials, as well as policy outcomes (Volden et al., 2013;Wängnerud, 2009), gender effects in bureaucracy are considered suppressed (Stensöta, 2018Stensöta et al., 2015). In the political/electoral arena, individual political candidates need to 'stand out' to attract voters' attention, sometimes by utilizing their group-specific or personal characteristics such as gender. However, such personal and individual characteristics, acquired through social-origin or demographic factors, tend to be suppressed in a bureaucracy (Stensöta et al., 2015). This is because adherence to a norm of impartiality is considered one of the core principles of Weberian bureaucracy (Gualmini, 2008;Rothstein & Teorell, 2008).
In fact, such an assumption aligns with one of the perspectives regarding the effects of demographic and personal factors on public managers' attitudes and values: structural perspective. According to this view, civil servants and decision makers behave 'in a manner appropriate to his or her position' or other structural factors such as type of organization and organizational tasks, thereby relying on the concept of bounded rationality (Christensen & Laegreid, 2009, p. 953). In this view, the values and attitudes associated with civil servants' personal backgrounds are considered constrained (Christensen et al., 2012;Christensen et al., 2019;Christensen & Laegreid, 2009;Egeberg & Stigen, 2018). On the other hand, the demographic perspective argument (Pfeffer, 1983) focuses on the importance of 'pre-packed' demographic and personal factors in explaining civil servants' attitudes and preferences, as demonstrated by empirical studies of representative bureaucracy (Egeberg & Stigen, 2018, p. 4). In this view, 'who you are and where you come from' matters more than structural and organizational factors (Christensen & Laegreid, 2009, p. 954).
In this paper, we argue that gender does matter for public managers' attitudes toward innovation. A large amount of social psychology literature shows gender differences in terms of specializations, linguistic capabilities, predisposition toward stress, and general behavior (Hamidullah et al., 2015). According to Gilligan (1993), men and women perceive and construe social reality differently, leading to differences in values and behaviors (Hamidullah et al., 2015). Furthermore, social role theory explains that people's internalized societal cues and norms are gendered and different, reflecting 'the sexual division of labor and gender hierarchy of the society (Eagly et al., 2000, p. 124 as cited in Hamidullah et al. (2015). In fact a growing number of studies on representative bureaucracy identify that gender in public bureaucracy, along with other elements of the composition of bureaucracy, such as race and ethnicity, matters to the efficiency, legitimacy, and performance of bureaucracy (Hong, 2017;Riccucci & Van Ryzin, 2017;Suzuki & Avellaneda, 2018). Suzuki and Avellaneda (2018) suggest that decision-making in administrations is impacted by the differing life experiences and preferences of men and women. This paper examines three attitudes of public managers pertaining to innovation. First, we ask whether female managers have a more goal-oriented motivation (measured by the motivation to achieve results and doing something socially useful) than men. Secondly, we explore whether female managers possess more entrepreneurial attitudes (measured by openness to new ideas and creative solutions and willingness to challenge the status quo). Thirdly, we test whether female and male managers exhibit different preferences towards risk.
Our first hypothesis is that female managers will show more motivation to achieve results and do something socially useful than male managers. In particular, we argue that female managers, when faced with the dilemma of either achieving substantive results or following the rules, will be more likely to choose the latter than their male counterparts. This hypothesis goes against a widespread view of women in managerial positions. Leadership has largely been seen as culturally masculine, both in objective termsbecause most managers are menbut also in subjective termsbecause a 'leader should look male and pale' (DeHart-Davis et al., 2006). Women have generally been considered more rule abiding than men (Gordon, 1970) and evidence, both from surveys of city employees in the U.S. (Portillo & DeHart-Davis, 2009) as well as in experimental settings (Morrison, 2006), seems to provide some support for this view.
However, the fact that women are less likely to break the rules than men (Morrison) does not imply that women are more likely to strictly follow the rules. More importantly, the results of social psychology studies relying on social role theory suggest several reasons why one should expect women to be more oriented towards achieving substantive policy results and, particularly, welfare-enhancing outcomes for their communities. First, women generally score higher in 'communal' attributes and have a higher concern for the welfare of other people due to differences in patterns of social behavior acquired by men and women throughout life (Eagly et al., 2000). For instance, women foster more spending on in-kind family benefits (Ennser-Jedenastik, 2017). Second, women display a leadership style that is more interpersonally oriented than men's. Consequently, women tend to be more focused on improving others' morale and welfare (Eagly & Johannesen-Schmidt, 2001). Similarly, results of meta-analysis show that female managers score higher than male managers on measures of task-oriented leadership (Eagly et al., 1992). In contrast, male managers may be more inclined to take a conservative approach to policy, prioritizing the implementation of the existing rules over the achievement of results. Third, there is mounting evidence indicating than women care more for the well-being of society than men. Women are overrepresented in positions that involve 'emotional labor' (Guy & Newman, 2004), and female public managers show more compassion than their male counterparts (DeHart-Davis et al., 2006). Consequently, our first hypothesis is: Hypothesis 1: Female managers will show more motivation towards achieving results and doing something socially useful than their male counterparts.
Our second hypothesis is concerned with receptiveness to new ideas and creative solutions, and change orientation. Again, the stereotype here is that, ceteris paribus, men are more open to new ideas than women. To start with, women tend to rate themselves lower than men on innovative attitudes, as studies in the private sector have traditionally indicated (DiTomaso & Farris, 1992). Similarly, the pioneering study on gender differences in the public sector by Burns (1978) found differences in terms of professional ambition among 590 municipal administrators. Fox and Schuhmann (1999) also noticed that female city managers were less likely to see themselves as policy entrepreneurs than men but rather saw themselves more as managers and facilitators. However, numerous studies have revealed that women exhibit many attitudes that seem clearly related to pro-innovative behavior. For instance, female city managers are more likely to incorporate citizen input, to be concerned with community involvement in their decisions, and to emphasize more communication than male counterparts (Fox & Schuhmann, 1999).
Our hypothesis is that female managers are more open to creativity and new ideas than their male counterparts. Firstly, women in leadership positions tend to favor empowerment of those surrounding them (Van Engen & Willemsen, 2004). Secondly, female managers adopt a more democratic (or participative) style of leadership than their male counterparts, who tend to adopt a more autocratic style (Eagly et al., 1992), and this inherently implies more openness to new ideas. Thirdly, female leaders show more emotional intelligence (Mandell & Pherwani, 2003) and better ability to operate within informal networks (Bass, 1981).
We expect female managers to be more open to new ideas, but also more willing to challenge the status quo. Organizational scholars note that there are two types of leadership: transactional and transformational (Burns, 1978). While transactional leaders aim to clarify their subordinates' responsibilities, transformational leaders aim to innovate and challenge the status quo through inspiration and encouragement. Several studies indicate that women tend to adopt a more transformational leadership style than men, e.g., (Bass & Avolio, 1993). Additionally, numerous studies have dismantled the conventionally held view that men are more committed to policy change than women. For instance, evidence from state health and human service agencies indicates that female leaders in the public sector score higher on 'attraction to policy making' than men (DeHart-Davis et al., 2006). Furthermore, women score similarly to men on 'commitment to public service,' another stereotypically masculine characteristic. From a different cultural context, a study of the Seoul Metropolitan Government found that female employees were more satisfied with their jobsdespite enjoying fewer advantages in terms of pay, autonomy, or promotionsthan men (Kim, 2005). Indeed, gender was the only significant predictor of job satisfaction among the demographic variables. Consequently, our second hypothesis is the following: Hypothesis 2: Female public managers show more openness to new ideas and creative solutions to policy problems, and are more change-oriented than male public managers.
With regard to our third hypothesis (i.e., preferences towards risk), there are competing views toward gender differences. The relevant literature has reported that when it comes to financial decision-making in the private sector, a strong connection exists between gender and individual preferences and behavior (e.g., Filippin & Crosetto, 2016), with women demonstrating more risk aversion than men.
Such a view is also spread through popular books, e.g., John Gray's Men are From Mars, Women are From Venus (2003) or Simon Baron-Cohen's The Essential Difference (2004). Contrary to this, other scholars underline that existing systematic reviews of the evidence indicate that 'men and women tend to be much more similar in their responses to risk than the popular Marsversus-Venus understanding would imply' (Nelson, 2015, p. 580).
Nevertheless, there are reasons to expect that in the context of the public sector, female managers are, ceteris paribus, less prone to take risks than their male counterparts. Women are more concerned with the negative externalities of risky decisions for their organization. Certain risks could imperil the organizations they lead, and female managers may be more sensitive to this outcome than their male counterparts. We take this insight from the literature on finance, where two apparently contradictory results have been found. On the one side, in experimental settings, female subjects do not make significantly less risky financial choices than male subjects (Schubert et al., 1999). On the other, there is growing evidence indicating that female leaders are more prudent than their male counterparts. Women are more risk averse when it comes to decisions with potentially damaging financial consequences for the private sector firms they lead (e.g., Sunden & Surette, 1998). Generally speaking, female executives take less risky financial decisions (Huang & Kisgen, 2013).
We hypothesize that the same gender differences in managerial prudency may be found in the public sector. Female public managers, even if they dare to challenge the status quo more than men, will be less prone to taking risky decisions that could hamper the operations of the organization they lead. Our hypothesis resembles Suzuki and Avellaneda's (2018), who, using data from Japanese municipalities, show that female representatives in local councils are associated with risk-averse behavior in financial decision making. Consequently, we argue that: Hypothesis 3: Female public managers are less risk taking than male public managers.

Data Collection
This study aims to contribute to the growing scholarly interest in comparative studies of bureaucracy and civil servants (e.g., Dahlström et al., 2012;Meier et al., 2017;Suzuki & Hur, 2020). We use the COCOPS Executive Survey on Public Sector Reform in Europe (Hammerschmid, 2015), which contains the survey answers of 9,333 senior public sector executives from 21 European countries (see the appendix for further information on the data set). All of the individual-level variables come from this COCOPS survey, while the country-level variables are from the QoG Expert Survey (Dahlström et al., 2015), the QoG Standard Dataset (Teorell et al., 2017), Hofstede et al. (2010), UNdata (United Nations Statistics Division, 2017, and the Global Innovation Index (GII) database (Cornell University et al., 2014).

Dependent Variables
This study utilizes five dependent variables from the COCOPS survey that aim to collect public managers' pro-innovation attitudes: (1) motivation to achieve results, (2) willingness to serve societal interests, (3) openness to new ideas and creative solutions, (4) willingness to challenge the status quo, and (5) risk taking. All five variables come from alignment with the following questions. The first dependent variable is collected from the statement, 'Public services often need to balance different priorities. Where would you place your own position?' Respondents are asked to rank their management style from 1 (following rules) to 7 (achieving results). Higher values represent being more result oriented. The second dependent variable is derived from the statement, 'How important do you personally think it is in a job to have … 'Doing something that is useful to society'?' Respondents are asked to select their responses from 1 (not important at all) to 7 (very important). The third dependent variable is collected from the statement, 'Being creative and thinking up new ideas are important to me'; the fourth from the statement, 'I avoid doing anything that might upset the status quo'; and the fifth dependent variable from the statement, 'I like to take risks.' In the latter three questions respondents are asked to select their responses from 1 (strongly disagree) to 7 (strongly agree). These dependent variables are thus ordinal variables. We reverse scale the fourth dependent variable for interpretation. Therefore, higher values mean more willingness to challenge the status quo, while lower values mean more willingness to maintain the current condition.

Independent Variable
The main interest in our study is the gender of senior managers. The COCOPS Survey includes a question asking about gender. In our analysis we set male as the baseline, 1, and female as 2.

Control Variables
We control for individual-level factors as well as country-level factors that may affect individual attitudes towards the aforementioned dependent variables, drawing on the prevailing literature on the importance of structural, Weberian and other demographic factors. Individual-level factors include organizational type, organizational size, respondent's current position, age, public sector experience, educational level, job satisfaction, organizational goal clarity, job autonomy, degree of political interference, organizational social capital, and respondent's organizational commitment. The latter four variables are included in our robustness check models. Country-level variables include professionalism of bureaucracy, bureaucratic closedness, women representation in the public sector, gender inequality, country-level innovation, and three national cultural factors from Hofsted's dimension of cultural values, namely power distance, individualism-collectivism, and uncertainty avoidance. The first three variables are included in our main models (Table 4), and the remaining variables are included in our robustness check models (Table A2 in the appendix). Detailed operationalization and data source information are described in the appendix. Table 1 presents descriptive statistics of all variables in the analysis. We conducted collinearity diagnostics using variance inflation factors (VIF) based on our main models. Mean values of VIF are less than 1.65 in all main models. The highest individual VIF score for individual variables is 3.28 (public sector experience). These results suggest that the models do not have serious multicollinearity issues.

Methods
Our data set has a hierarchical structure, with public sector managers (level 1) nested in country-level factors (level 2), and thus multilevel analysis seems to be an appropriate method (Jones, 2008). We assume that intercepts of individual-level variables can vary across countries due to the country-level factors and therefore a random intercept model is applied. Since the dependent variables in this study are ordinal variables that have a natural ordering measured on scales from 1 to 7, we employ multilevel ordered logit models. The first set of models (models 1-5 in Table 3) includes only individual-level independent and control variables. The second set of models (models 1.1-5.1 in Table 4) adds country-level variables. A small number of country-level samples might create a problem with multilevel analysis (Stegmueller, 2013). However, we limit the number of country-level variables in a single model to three. Moreover, we do not have cross-level interactions. Therefore, we believe that our results do not suffer from a bias arising from country factors. Please see the appendix for our robustness check strategies and results. Table 2 shows mean comparisons of the dependent variables by gender.

Analysis and Results
There are gender-based differences in the second, fourth and fifth dependent variables. The mean values of motivation to do something useful for society are higher among female managers than their male counterparts (p-value < 0.000). Female public managers also score higher in their willingness to challenge the status quo than male managers, even if this result is not statistically so robust (p = 0.075). However, female managers' scores on willingness to take risks are 0.17 points lower than their male counterparts (p < 0.001). Female managers show more motivation to achieve results (p = 0.116) and slightly more open attitudes towards new ideas than male managers. Nevertheless, these difference are not statistically significant (p = 0.343).
Having shown a bivariate relationship between gender and the dependent variables, we move on to the results of the multilevel regression models.   Table 3 reports the results of the multilevel ordered logistic models with only individual-level variables. Gender has a statistically significant association with all of the dependent variables. Being a female manager is positively associated with motivation to achieve results (p < 0.001), motivation to serve societal interests (p < 0.001), openness to new ideas and creative solutions (p < 0.01), and willingness to challenge the status quo (p < 0.01). On the other hand, gender is negatively associated with willingness to take risks (p < 0.05). Some of the individual-level controls show statistical significance. Age (p < 0.001), doctoral degree (p < 0.01), working at an agency or subordinate governmental body at state government (p < 0.05), and organizational size (p < 0.05) are positively associated with motivation towards achieving results. On the other hand, working at an agency or subordinate governmental body at central government (p < 0.05), having a second hierarchical level position (p < 0.05), and working years in the public sector (p < 0.01 and p < 0.05) are negatively correlated to the motivation to achieve results. Age (p < 0.05), private sector experience (p < 0.05), education (p < 0.05 and p < 0.001), job satisfaction (p < 0.001), and goal clarity (p < 0.001) are positively associated with motivation to serve societal interests. Age (p < 0.01), private sector experience (p < 0.001), doctoral-level education (p < 0.01), and job satisfaction (p < 0.001) are positively correlated to openness to new ideas and creative solutions. On the other hand, working at an agency or subordinate governmental body at central government (p < 0.01) and a ministry at state or regional government (p < 0.05) are negatively associated with openness. Organizational size (p < 0.001), doctoral degree (p < 0.05), job satisfaction (p < 0.05), and organizational goal clarity (p < 0.01) are positively related to respondents' willingness to challenge the status quo. Working at an agency or subordinate governmental body at central government (p < 0.01), being in a second hierarchical (p < 0.01) or third hierarchical (p < 0.001) position, and having more than 20 years of public sector experience (p < 0.01) are negatively associated with motivation to challenge the status quo. Finally, organizational size (p < 0.01) and private sector experience (p < 0.05 and p < 0.001) are positively correlated with risk taking. Working at an agency or subordinate governmental body at state government (p < 0.05) and being in a third hierarchical level position (p < 0.01) are negatively correlated to risk taking. Estimation results of models with country-level variables are reported in Table 4. Even after controlling for country factors, the gender impact results do not change. Being a female manager is positively associated with motivation to achieve results and serve societal interests, openness to new ideas, and willingness to change the current condition, and negatively correlated with risk taking. Most of the significant individual-level controls also hold statistical significance. As for the country-level factors, a professional bureaucracy (as opposed to a politicized one) is positively associated with motivation to achieve results (p < 0.001) and willingness to challenge the status quo (p < 0.01). Contrary to our expectation, the percentage of women in the public sector is negatively associated with motivation to achieve results (p < 0.001).
Since multilevel ordered logit coefficients are difficult to directly interpret, we calculated the predicted probabilities and marginal effects of gender for each of the five values of the dependent variables. Tables 5-9 report predicted probabilities of all responses for five dependent variables. Table 5 reports predicted probabilities of seven outcomes for motivation to achieve results. Holding other factors at the mean, almost half of female senior managers say that they are willing to achieve results rather than follow rules (responses '5,' '6,' and '7'), compared to 44% of male senior managers. On the other hand, 28.4% of female managers report that they prioritize following rules (responses '1,' '2,' and '3'), while 32.8% of male managers do so. All of these results are statistically significant. Results suggest that compared to their male counterparts, female senior managers are more likely to put priority on achieving results rather than following rules. Table 6 reports predicted probability of motivation to serve societal interests. When all other things are equal, 93.2% of female senior managers have high motivation to serve societal interests (responses '5,' '6,' and '7'), compared to 90.6% of male managers. Thus, results suggest that a higher percentage of female managers are more likely to report that they want to do something good for society than their male counterparts. Table 7 shows predicted probability of openness to new ideas and creative solutions by gender: 81.4% of senior female managers report they agree with the importance of openness to new ideas (responses '5,' '6,' and '7') compared to 79.9% of male senior managers, holding other factors at the mean. These results are also statistically significant. Table 8 reports that 74.7% of female senior managers say that they have greater willingness to challenge the status quo (responses '5,' '6,' and '7') compared to 72.1% of male senior managers, holding other factors constant. Results suggest that more female managers exhibit willingness to change the current situation than their male counterparts. Finally, Table 9 shows Predicted probability is calculated based on model 1.1 in Table 4.
predicted probability of risk taking. While 52.9% of male senior managers show agreement with risk taking (responses '5,' '6,' and '7'), 50.5% of female managers say that they are willing to take risks, holding other factors at the mean.    Tables 10-14 report average marginal effects of being a female manager on each outcome of the dependent variables. According to Table 10, and consistent with the earlier result, female managers are 5.2 percentage points more likely than male counterparts to think that achieving results is important. Table 11 suggests that female managers are 2.5 percentage points more likely to say that serving societal interests is important compared with male managers. Table 12 shows that the result of outcome '6' is not statistically significant (p = 0.858). However, female managers are 2.4 percentage points more likely than male managers to strongly agree with openness to new ideas and creative solutions (outcome '7'). Table 13 shows that female managers

Conclusions
If research on gender differences in the public sector only explores the direct effect of gender, it may fall in to 'a myriad of stereotypes and generalizations'   Table 4. (Jacobson et al., 2010, p. 478). So, many studies have noted the importance of contextual factors and interactions (Sowa & Coleman Selden, 2003). This paper contributes to this literature by expanding the analysis of gender with a large, interorganizational, cross-country data set. The inclusion of these additional variables in a large-N sample allows for a better, more nuanced, understanding of the role of gender in public administrations and particularly for public innovation.
The results of our study show that ceteris paribus, female public managers exhibit more pro-innovation attitudes than their male counterparts, but with some caveats. First of all, the results are consistently statistically significant throughout the models, even after controlling for a large number of demographic and organizational variables, and yet the effects are not substantially very relevant. For instance, holding other factors constant, being a female leader increases result orientation by 5.2 percentage points, serving societal interests by 2.5 percentage points, openness to new ideas and creativity by 2.4 percentage points, and challenging the current condition by 2.6 percentage points. On the other hand, being a female manager decreases risk-taking attitudes by 2.2 percentage points. Nevertheless, despite relatively small size of gender effects, the results are relevant because two of the three main results go against the conventional view of gender differences which shows female managers are less motivated to achieve results (and more motivated to follow rules), less open to new ideas and creativity, and less willing to challenge the status quo.  Table 4.
We think that such small effects of gender may be due to strong constraints on the effects of demographic factors in bureaucracy, as the structural perspective explains (Christensen & Laegreid, 2009). In particular, this study focuses on senior public managers, not middle-or street level-bureaucrats. As the literature on street-level bureaucrats suggests, frontline public officials tend to have more discretion, which, in turn, leads to lowered 'suppressive power of the bureaucracy' among female officials (Stensöta, 2018, p. 132). Therefore, gender effects are likely to emerge more at lower levels of bureaucracy (Stensöta, 2018) than at the senior level. In order to empirically test the relationship between the effects of gender and position, we would thus need a data set containing public officials from a wider range of positions, not only senior positions.
1wAdditionally, gender may matter only in specific contexts. To rule out this possibility, we have tested various interactions with gender variables by adding interaction terms between gender and individual-and country-level variables. Such country-level variables include gender representation in the public sector, professional bureaucracy, and closed/open bureaucracy. Individuallevel factors include the respondent's position, organizational size, private sector experience, job autonomy, organizational commitment, and educational background. However, we did not find any meaningful interaction effects. Thus, in light of the data analyzed here, we can conclude that gender exerts a direct and significant effect on different dependent variables linked to public sector innovation, and that this effect is not mediated by other variables. Predicted probability is calculated based on model 5.1 in Table 4.
Regarding the implications of this paper, our results seem to complement Bratton and Ray (2002), which finds that gender representation has the greatest effect on policy outcomes (i.e., the coverage of child-care services) during periods of policy innovation. The reason why women have an especially strong impact when governments engage in innovation may be due, as this paper indicates (and against the conventional view of women as being less entrepreneurial than men), to the fact that female leaders actually generate a more innovative-oriented culture than their male counterparts.
Our study is not without limitations. First, this study has the same constraint as other previous studies using survey data from public officials. It relies on selfreported variables, which potentially cause a social desirability bias (Jakobsen & Jensen, 2015;Meier & O'Toole, 2012). The COCOPS data set implemented a cleaning procedure dropping all respondents who answered less than 25% of all survey items so as to reduce respondent bias. However, we cannot exclude the possibility that respondents' self-assessment has an upward bias. Another shortcoming is a lack of longitudinal data. Without such data we are not able to assess socialization effects on gender-based attitudes and behavior (Quintelier, 2013). In male-dominant cultures, female senior managers may face pressure to behave like their male counterparts, which may suppress the expected effects of gender on attitudes. We controlled for the country level of female representation and cultural values, as well as age and working years in the public sector, in order to mitigate this problem. However, without longitudinal data we are not able to measure socialization effects on gender differences in behavior. At present, as Van de Walle et al. (2016) have indicated, data collection on comparative bureaucratic behavior on a large scale is still in its early stages, and therefore opportunities to undertake the suggested tasks will arise as such data emerges.
Supporting data and materials for this article comes from the COCOPS dataset, which has two versions: full version and reduced version. The reduced version is publicly available from Gesis webpage here https://dbk. gesis.org/dbksearch/SDesc2.asp?DB=E&no=6599. Note 1. Even in the general innovation literature, the link between innovation and gender has been seldom explored (Fogelberg Eriksson, 2014;Gry et al., 2013, Lindberg and Schiffbänker 2013, Pecis 2016. Individual innovators rarely receive any focus in innovation studies (Nählinder 2010).

Disclosure statement
No potential conflict of interest was reported by the author(s).