Female presence in local government and cost efficiency: the case of Italian municipalities

ABSTRACT This paper investigates how female presence in political institutions influences local governments’ cost efficiency, implementing a one-stage stochastic frontier model, based on a panel of Italian municipalities, with municipal fixed effects. The empirical results provide significant evidence for the presence of a gender effect in politics: higher efficiency is observed in municipalities with a female mayor and a greater presence of women in their councils. Further, the gender effect is confirmed taking into account the regional perspective, that is, the historic North–South Italian dualism, and performing several robustness checks. The empirical results support the usefulness of policies aimed at favouring female empowerment and at promoting gender equality in politics.


INTRODUCTION
The gender gap that prevents women from accessing leadership positions is a persistent problem worldwide.Women are underrepresented in several contexts and, in particular, in politics, where female participation is limited at both central and local levels (Braga & Scervini, 2017;Crotti et al., 2020).
Several motivations can explain female underrepresentation in politics.A widely shared reason is that the entry of women into politics is hindered by social norms that usually associate leadership positions with men (Eagly & Karau, 2002).Such norms may lead political parties to reduce support for female candidacy, and voters to prefer male candidates (Baltrunaite et al., 2019).Another rationale is the lack of interest of women in competing for political seats, due to time constraints associated with family and care responsibilities, usually children, but also disabled siblings or parents (Tilly & Scott, 1989).Furthermore, differences in personality traits, well documented by the existing literature, could represent additional barriers, implying female underrepresentation in different sectors (Costa et al., 2001;Fossen, 2012).Higher risk aversion and anxiety, combined with lower self-confidence and willingness to face competitive situations, compared with their male counterparts could explain the gender gap in different competitive environments, such as the political arena (Croson & Gneezy, 2009;Niederle & Vesterlund, 2007).
During the last decades there has been a widespread adoption of public regulations aimed at reaching a more gender-balance presence in political institutions.Gender quotas in public elections represent one of the main measures introduced by public authorities in several countries with the aim to favour female political empowerment. 1  Different factors support the promotion of gender equality, not exclusively limited to equity, ethics or justice, but also linked to welfare considerations (Stevens, 2007).Women are usually more concerned with the well-being of the people, show higher cooperation and teamworking skills, and are less likely to engage in corruption (Chattopadhyay & Duflo, 2004).Therefore, a higher female political participation may affect the policies implemented and the allocation of resources across different programmes in favour of health and welfare expenditures (Funk & Gathmann, 2015).
The literature has extensively explored the gender effect at the firm level investigating whether female presence in leadership positions could affect companies' performance (e.g., Adams & Ferreira, 2009;Krishnan & Park, 2005;Luckerath-Rovers, 2013;Post & Byron, 2015).In contrast, gender diversity in the public context has attracted the interest of researchers only in recent years, when a growing number of studies has tried to address the effect of female political empowerment in several settings.However, the empirical results are rather heterogeneous.On the one side, a shared consensus emerged about the impact of female presence on policymaking process, as highlighted by Hessami and da Fonseca (2020) in a survey of the empirical literature.Female politicians prefer issues traditionally linked to their concerns, such as healthcare, social services, family and childcare (Chattopadhyay & Duflo, 2004;Clots-Figueras, 2012;Funk & Gathmann, 2015;Holman, 2014).On the other, weak results have been reached by the empirical studies analysing the effect of women's representation in a variety of policy outcome measures, such as public good provision or employment (Duflo & Topalova, 2004;Ferreira & Gyourko, 2014).Further, government stability represents another issue investigated by the gender literature, highlighting a more likely early termination of the legislature in the presence of a female mayor (Acconcia & Ronza, 2021;Gagliarducci & Paserman, 2012).
In view of the well-documented effect of female public empowerment on policymaking process, we analyse if the gender of local policymakers could play a significant role in affecting local governments' efficiency, an issue of high interest in the field of the public sector, still at the centre of the political and academic debate.The decentralization policies largely adopted during the last decades, devolving to local governments decision-making powers concerning expenditures and revenues, combined with the growing resources constrains following the 2008 economic and financial crisis, made the efficient management of available resources one of the main challenges for local governments.Such entities play a key role in providing public goods and services to citizens, ensure the main national infrastructures, hold critical responsibilities for healthcare and education, and create the environment where companies gain competitive advantages and have the faculty to stimulate and upgrade domestic demand (Porter, 2011).
A wide empirical literature has long explored the role of different external factors in affecting local governments' efficiency without a shared consensus.Indeed, as highlighted by Narbon-Perpina and De Witte (2018) in an extensive and comprehensive survey of the existing literature, different types of socio-economic, geographical and political characteristics of the territory present ambiguous effects on local governments' efficiency. 2 Further, the role of female political empowerment on local governments' efficiency has rather been neglected in such literature.The studies by Boetti et al. (2012) and Agasisti et al. (2016) represent two of the very few investigating such issues.Analysing Italian municipalities, respectively located in the province of Turin and in the region of Lombardia, both papers fail in finding a significant effect for the presence of a female mayor on local governments' efficiency.
The main purpose of the present paper is to investigate whether the efficiency level of municipalities is affected by female empowerment in public governance.Considered that women are usually more concerned with the wellbeing of people and less engaged in corruption, we expect to find that local governments' efficiency is positively affected by female political empowerment (Chattopadhyay & Duflo, 2004).When developing our empirical model, we follow the approach of Afonso et al. (2005), who define government efficiency as the relationship between the outcomes and the resources employed.Further, we proxy female political empowerment using two different measures: mayors' gender and the share of women in the municipal council.
The empirical analysis is based on a sample of Italian municipalities observed during the period 2010-16.Italian municipalities provide an interesting case study since they have experienced a significant empowerment of women in public governance in the last decade, although, according to Crotti et al. (2020), Italian municipalities are still far from achieving full gender parity of political empowerment (in 2018, Italy closed 26.7% of its gap).Further, the Italian case is worthy of attention because it shows a considerable heterogeneity between the different areas of the country in terms of local development, in the private sector, in the public expenditures and revenues received by municipalities, and considerable municipal disparities exist also in terms of land size, density and age structure of the population (Fratesi & Percoco, 2014;Greco et al., 2018;Patrizii & Resce, 2015).
This paper provides an original contribution to the existing literature in several aspects.First, it gives new clear evidence that female political empowerment significantly affects municipalities' efficiency, to the best of our knowledge an issue not well documented so far by empirical studies.We show that having a female mayor and the greater presence of women in the municipal council favour cost-efficiency performance.Second, the empirical analysis is based on a large sample of Italian municipalities belonging to different regions, allowing us to investigate the potential role of regional disparities characterizing the Italian territory, that is, the historic North-South Italian dualism (Fratesi & Percoco, 2014).When performing the empirical analysis taking into account regional disparities, we confirm that a higher female presence in the council and a female mayor have a positive effect on cost efficiency in both areas, but with a weak significance in the Centre-South.Finally, the gender effect in politics is also confirmed when several robustness checks are performed.Overall, the results provide new evidence for the political process, supporting the usefulness of public regulations aimed at favouring female presence and at promoting gender equality in political institutions.
The remainder of the paper proceeds as follows.Section 2 summarizes the main findings of previous literature investigating the effects of female political empowerment.Section 3 presents the empirical model.Section 4 describes the data.Section 5 present the empirical findings.Section 6 concludes.

LITERATURE: THE EFFECTS OF FEMALE POLITICAL EMPOWERMENT
During the last decades, the gender issue has caught the interest of policymakers worldwide and, at the same time, of the empirical literature.Worldwide, policymakers started to introduce policies aimed at reaching a more gender-balance presence in political institutions.Consequently, several studies investigated the consequences of female political empowerment, mainly at the local level, in several settings: how female preferences affect the policies to implement, the effect of politicians' gender on government stability, and the impact of female presence in policy institutions on outcome measures.
A shared consensus emerged about women' capacity to affect policy decisions according to their preferences.Taking advantage of political reservations for women introduced in India since the 1990s, Chattopadhyay and Duflo (2004) find that politicians' gender affects policy decisions: female leaders prefer public investments more closely linked to their concerns, such as drinking water.In line with this finding, Clots-Figueras (2012) provides convincing evidence of gender effects in India in another policy setting: education.Investigating whether politicians' gender matters for education achievements, the author finds that increasing female political representation in an urban area of a district in India increases the probability of primary education achievement.A possible explanation for this result is based on the consideration that female politicians may care about the needs of women and, consequently, will invest more in education.For the developed context, Holman (2014) investigates the effect of female participation in politics on the provision of social welfare services, studying a random sample of US cities.The empirical findings show that cities headed by women have a higher probability of participating in a social welfare programme and dedicating greater funding to these programmes.However, it emerges that women in the councils do not influence the propensity or the amount of social welfare funding in the absence of a female mayor.Focusing on female representation in Swedish local councils, Svaleryd (2009) finds that a larger share of women in local councils increases spending on childcare and education relative to elderly care.Funk and Gathmann (2015) identify gender gaps in policy preferences, analysing the different allocation of government resources of female and male voters in Switzerland, a country where, due to extensive direct democratic rights, citizens regularly decide on a board range of issues.The authors find that female voters strongly differ from their male counterparts: women are more favourable to allocate public funds to environmental protection, healthy lifestyle, equal rights, and less supportive of nuclear energy and military spending.Analysing a sample of Spanish municipalities in 2008, Cabaleiro and Buch (2020) find that increasing the female presence in municipal councils produces significant effects on several budgetary policy measures, ranging from total current expenditure to tax revenues, including current spending in different functions.Further, analysing Italian municipalities, Ordine et al. (2022) show that an increase in the number of elected women reduces the overall amount of public expenditure, without affecting public spending concerning services related to females' needs, such as kindergarten, primary education and social care.
Another gender issue investigated by the empirical literature concerns the effect of politicians' gender on local government stability. 3Analysing a sample of Italian municipalities, Gagliarducci and Paserman (2012) find that in the presence of a female mayor the likelihood of an early termination of the legislature is greater.Further, the probability that a female mayor ended her term is lowest in the presence of an entirely male council, and in southern regions where unfavourable attitudes towards female empowerment still prevail.These findings highlight more difficulties for female mayors in fostering cooperation among men and keeping the coalition together, because of the resistance and the reluctance of male councillors to be headed by a woman.Similar conclusions have been reached in a recent study by Acconcia and Ronza (2021) who document a negative effect of a female mayor on Italian municipalities' stability.However, the authors document a positive effect for the presence of female councillors on local governments' stability.
If some consensus has been reached about the impact of female public empowerment on policymaking process and government stability, the evidence on the effect of politicians' gender on policy outcomes is rather scant and usually conflicting.An example is provided by Duflo and Topalova (2004), who compare the performance of council leaders by gender in Indian villages and how the villagers perceive them.The authors find that despite female leaders providing more public goods, their quality does not differ among villages and, further, female villagers (i.e., living in villages headed by women) are less satisfied with the public goods received.Also the studies focused on the developed world fail in providing evidence that female political empowerment positively affects outcome measures.Studying US mayoral election between 1950 and 2005 through a regression discontinuity design, Ferreira and Gyourko (2014) do not find a significant impact of mayors' gender on a variety of policy outcome measures, such as a local government's size, the composition of its spending, employment or local crime rates.Analysing a rich dataset of Brazilian municipalities, Brollo and Troiano (2016) obtain different findings.The authors show that to have a female mayor implies better health outcomes related to prenatal care delivery.In addition, it emerges that in the presence of a female mayor, federal discretionary transfers are higher, while administrative irregularities decrease.
The role of gender considerations on local governments' efficiency has been analysed by few studies that fail to find a significant effect for the presence of a female mayor.In a study based on Finnish city mayors' characteristics, Loikkanen and Susiluoto (2005) find that gender is a variable not significant in explaining differences in cost efficiency.Similar conclusions have been reached by Boetti et al. (2012) and Agasisti et al. (2016) for Italian municipalities.Boetti et al. (2012) find no beneficial effect of gender of the mayor on the level of efficiency for a sample of Italian municipalities belonging to the province of Turin.Analysing a sample of large Italian municipalities located in the Lombardia region (i.e., having more than 5000 inhabitants), Agasisti et al. (2016) find that mayors' gender is not significant in affecting local governments' efficiency.The main limits of both these studies consist of the very scant number of Italian municipalities considered in both the empirical analyses, 262 and 331, respectively, and the focus on a single province/region, implying that the regional perspective has been lacking, and in particular the regional disparities characterizing the Italian territory.
The increasing adoption of public policies aimed at promoting gender equality in political institution combined with the scant and conflicting results reached by the empirical studies on the consequences of female public empowerment suggest the usefulness of further investigations and explanations.

METHODS
To investigate the effect of female empowerment on municipalities' efficiency in the provision of local services, we estimate a cost function through a stochastic frontier analysis (Kumbhakar & Lovell, 2000).Into the parametric stochastic frontier approach we include exogenous variables that may exert an influence on municipalities' performance, controlling for potential heterogeneity.
The classical stochastic frontier model for panel data can be written as: where, for each unit i at time t, f (•) defines a cost (frontier) relationship between the total cost (y) and the corresponding vector of outputs and prices (x), depending on the corresponding parameter vector b, which represents the technology to be estimated.Therefore, f (•) defines the minimum feasible cost for a given level of outputs and prices (the input-oriented analysis).The error term has two components: vis a symmetric two-sided error representing random disturbance due to measurement errors (i.e., the classical noise), while u is a one-sided error term which represents technical inefficiency (u .0). v is assumed to follow a two-sided normal distribution (i.e.,v iid N (0, s 2 v )), while u follows a half-normal distribution on the non-negative part of the real number line (i.e., u iid N + (0, s 2 u )).The presence of inefficiency determinants is handled by relaxing the constant variance assumption on the inefficiency term.s 2 u is specified as a function of some explanatory variables potentially associated with technical inefficiency.Therefore, we assume that the inefficiency term depends on a set of exogenous covariates via the so called 'scale' transformation (Wang & Schmidt, 2002) of some underlying process h it : where z it is a vector of contextual variables, gis a vector of unknown coefficients, and h it is an i.i.d.random variable that does not depend onz it , non-negative (h it ≥ 0), with mean one (E(h) = 1) and finite variance (Kumbhakar & Lovell, 2000).A single element of g is positive (negative) when the corresponding element of z has a negative (positive) effect on technical efficiency.This formulation has the computational advantage to automatically constrain s u to be positive and to directly nest the homoscedastic case.Further, any change in z it implies a variation in the scale, but not in the shape of the distribution of u it (Alvarez et al., 2006).We use this specification following Wang and Ho (2010) after adopting the within transformation which in the case of panel data permit one to distinguish unobserved individual heterogeneity from inefficiency and to take into account potential endogeneity.
To allow production units to have rather heterogeneous technologies, we specify f (•) by considering the translog cost function, which represents a class of flexible functional forms for the cost functions since it provides a second-order approximation to any arbitrarily twice differentiable function. 4More specifically, the cost function can be written as: with b jk = b kj and input elasticities given by: The inefficiency score of each unit is estimated by the conditional distribution of the u component with respect to compound error 1 = v + u (Jondrow et al., 1982).Denoting l = s u /s v and s 2 = s 2 u + s 2 v , the estimates of individual efficiency score are: where f(•) and F(•) are the standard normal density and cumulative distribution functions, respectively, and: Finally, individual cost efficiency can be obtained from CE = exp { − û}.municipalities' expenditure and output information for all sample years. 7 For each municipality we consider the expenditure and the output variables with regard to six different services: public education, local police, viability and territory, social services, waste, and general administration.These represent the fundamental functions of ordinary statute municipalities used to classify municipalities' standard needs, as defined by legislation (LD 216/2010).
In our model specification (equation 1) we define the total costs as the sum of municipalities' current expenditures overall for the six services.On the contrary, as regard to the output side, for each service we follow the methodology developed by the Solution for the Economic System (SOSE) (2018), which aggregates elementary subservices into the corresponding macro-service adopting a combination of two methods widely implemented by previous studies: the benefit of the doubt (BoD) and the pricebased method (Greco et al., 2019).The BoD has been extensively implemented in the evaluation of public services because it allows one to include the heterogeneity in the preferences of the decision-maker (Barra et al., 2022;D'Inverno et al., 2022); this is extremely relevant in the municipalities because the 'mix' in the allocation of different goods and services is the result of political choices of systems whose leadership is elective (Patrizii & Resce, 2015).The price-based method has been also widely implemented because it allows to include the costs into the aggregations (Card, 1999;Murphy & Topel, 2006).It has been shown that this framework is a reliable solution when the efficiency of public services needs to be computed, mainly because it avoids that the relative importance of the different macro-services depends on the number of the corresponding subservices, also avoiding the curse of dimensionality (Barra et al., 2022;D'Inverno et al., 2022;Narbon-Perpina & De Witte, 2018).
Different methods have been adopted in order to obtain the output measures for the six identified municipalities' services: public education, local police, viability and territory, social services, waste, and general administration.
The output measure for public education was computed starting from the consideration of the different services provided by municipalities in support of this service.Therefore, the output measure is obtained as the weighted sum of transported students, canteen users and students engaged in extracurricular activities managed out of school hours.The weights are obtained by the marginal costs of the function used to determine the municipalities' standard needs (i.e., higher weights are assigned to more costly services).
For local police, the output indicator is the weighted sum of seven different micro-outputs related to the services: fines for traffic offences, fines for commercial offences, administrative seizures, removal of vehicles, investigations for other local governments, work for judicial authority, and public land occupation permits. 8This variable reflects the volume of activity realized by police in municipal territory.
A two-step procedure was implemented in order to obtain the output for viability and territory.In the first step, two macro-outputs, one for territory and the other for viability, were obtained by a weighted sum 9 of several micro-outputs reflecting activities connected to each service. 10In the second step, the two macro-outputs were aggregated using the correlation between macro-outputs and current expenditures as weights.
Also, the computation of social services output follows a two-step procedure.In the first step, users of social sector services were grouped into six homogeneous groups, defined target: family and minors, disabled, people with addictions or mental health, old people, immigrants and nomads, and poor adults and homeless.Then, for each target a macro-output was computed by adding the elementary provisions of services, composed of interventions and services, economic funding and organizations.In the second step, the six macro-outputs were aggregated using the correlation between macro-outputs and current expenditures as weights.
Waste disposed, data directly provided by Italian Institute for Environmental Protection and Research (ISPRA) were used as a proxy for the output of municipalities' waste management service.
Despite population not representing a proper output measure, however, it is typically used as a proxy to quantify the intangible production of general administration for efficiency analysis in Italian municipalities (e.g., Boetti et al., 2012;D'Inverno et al., 2018).Therefore, we proxy municipalities' general administration output using population data directly provided by ISTAT.
Finally, we include in the cost function two price variables: the local cost for renting offices/apartments and the revenues declared by resident population for fiscal reasons.Such variables are used as a proxy for the cost of the two classical production factors: capital and labour.According to the World Bank (2011) framework, the capital is multidimensional because it includes the produced, the natural, the human and the institutional components.However, among these, the only asset price available at the municipal level in Italy is the average rent per m 2 for offices/apartments.Therefore, we use this measure as a proxy of the price index of productive factors different from labour, and assume that the prices for other inputs are constant across the whole sample, as in Arcelus et al. (2015). 11Furthermore, both the local cost for renting and the revenues declared can be considered exogenous enough to be included in the cost function for the public sector given that municipalities' public spending cannot affect their levels in the short run.According to the Tiebout model, an increase in the quality of local public services driven by an increased public expenditure could result in higher property values and income because of location decisions (Oates, 1969), but such a mechanism requires a medium-to-long-run time to be observed (Gyourko & Tracy, 1989).Therefore, it is unlikely that municipalities' expenditure in a year could affect the local house prices and income in the same year.Further, in the short run, the municipality is a price-taker mainly because the dimension 20 Valeria Bucci et al.

REGIONAL STUDIES
of its budget is marginal compared with the total economy. 12Table 1 reports descriptive statistics of costs, outputs and prices used in our model specification in 2010, 2013, 2015 and 2016.When analysing costs, it emerged that the average value of current expenditures per capita for general administration decreased from €347 to €274.A slightly decreasing trend is observed for education (from €79 to €69) and social services (from €75 to €66).In contrast, the average expenditure for waste disposed increased from €128 to €138.Quite stationary are the expenditures for both local policing and urbanization.
On the output side, a decreasing trend characterized most of the variables.Waste disposed per capita decreases from 471 to 450 tons, public education output decreases from 1.1 to 0.9, viability and territory output decreases from 0.7 to 0.3, and local police output decreases from 3.8 to 2.1.Quite stationary is the output of social services, while the population value increases, reaching the value of 6,266,000 from 5,956,000.Overall, as can be seen from the standard deviations, the confidence intervals are all overlapping, meaning that the trends in the average output measures are not significant.
Regarding the prices used in our cost function, the average local cost for renting decreased from €4600 to €4200/m 2 per month, and average revenues per capita increased from €28,200 to €30,600.
Table 2 summarizes information about the contextual variables that can potentially influence municipalities' efficiency.Such information concerns demographic factors and institutional frameworks.The two main variables of our empirical model are the proxies for the effect of female empowerment: the presence of a female mayor and the share of women in the municipal council.The data show an increase during the period 2010-16 for both the share of Italian municipalities headed by a female mayor and the average presence of women in the councils, from 11.4% to 14.7% and from 19.8% to 27.1%, respectively.Further, Figure 1 confirms the female political empowerment process occurred between 2010 and 2016 in the Italian regions: municipal councils having no female members fade, while municipalities with a significant female political presence intensified.

RESULTS
This section presents the findings concerning the association between female empowerment in public governance and the efficiency of Italian municipalities in the period 2010-16.To compare our results with the previous literature, we start our empirical analysis by modelling municipalities' efficiency as a function of female presence in politics and adding additional controls over two steps.We estimate three nested specifications of the model (1): . Model 1: municipalities' inefficiency is considered exclusively a function of female presence in politics.
. Model 2: additional institutional characteristics are included in the cost function. .Model 3: institutional factors, socio-economic and demographic characteristic that can affect municipalities' inefficiency are included in the empirical model.
All model specifications include municipal fixed effects in the cost function to account for individual (i.e., municipal) heterogeneity.
Table 3 summarizes the findings of the empirical estimations of the three model specifications. 13When considering municipalities' inefficiency exclusively as a function of the variables proxying for female presence in political institutions (model 1) we provide evidence for a significant gender effect on municipalities' efficiency.A significant negative association emerges between the gender of the mayor and inefficiency: to have a female mayor favours the municipalities' cost efficiency performance.This result is in contrast with the previous literature, particularly with the findings of Boetti et al. (2012) and Agasisti et al. (2016), who failed to provide convincing evidence of the beneficial effect of mayors' gender on municipalities' efficiency, whose main limit is the very scant dataset used to implement the empirical analysis.Therefore, we show that implementing an accurate empirical analysis based on a large sample ensuring heterogeneity in the data could help in providing convincing evidence of a gender effect on municipalities' efficiency performance.Furthermore, the positive effect of female empowerment on efficiency is confirmed if we analyse the effect of women's presence in municipal councils: increasing the share of women in councils significantly reduces municipalities' inefficiency.These results confirm and extend the findings reached by previous studies investigating the effect of female presence on public expenditure (Cabaleiro & Buch, 2020;Ordine et al., 2022), education outcomes (Clots-Figueras, 2012) and social services (Holman, 2014).Following Gagliarducci and Paserman (2012), we include in the estimation model the interaction term between a female mayor and the share of women on the council.The interaction term results are not statistically significant, showing that a higher female presence in municipalities' councils does not affect the probability of having a female mayor.
The positive effect of female political empowerment is confirmed after the inclusion of additional institutional characteristics able to explain municipalities' efficiency (model 2 in Table 3).Further, the empirical findings provide evidence of a significant negative effect of the average age of the council members on municipalities' inefficiency.The decrease of inefficiency with the increase of the age of administrators implies that older policymakers are more efficient.This evidence is in line with the findings of Alesina et al. (2019), showing that younger politicians, in order to ensure their re-election, are more likely to adopt strategic choices increasing public spending.Furthermore, the share of employees with higher education is negatively associated with inefficiency.This result is in line with the previous literature showing that human capital, such as a Female presence in local government and cost efficiency: the case of Italian municipalities 21 REGIONAL STUDIES highly educated workforce, represents a strategic asset that contributes to the outcomes of local government authorities (Carmeli, 2004).Further, from a wider perspective, this result emphasizes the key role of education in economic development, both at the micro-and macro-levels (Barro, 2001).Finally, we include in the empirical model an additional set of variables concerning institutional factors and socio-economic and demographic characteristics able to affect municipalities' efficiency, following previous studies (Boetti et al., 2012;Geys et al., 2010).The estimation results are provided in model 3 in Table 3.We control for the demand for the education services provided at the municipal level including the share of the young population (3-14 years).The positive association between municipalities' inefficiency and the young population confirms that a higher number of students increases the likelihood of requiring more specific and costly services.In

REGIONAL STUDIES
addition, we control for population density, a variable able to classify rural/urban areas and the potential advantages of agglomeration economies.In this regard, we observe a negative association between population density and cost inefficiency, confirming the presence of agglomeration economies: a higher density could reduce costs for providing several local services, such as transportation.Further, we control also for land size, but such variable results are not statistically significant.To take into account the effect of environmental issues on municipalities' efficiency, we control for the share of waste sorted collection.We find that the greater the share of waste sorted collected, therefore the better the environmental conditions, the lower the inefficiency level of municipalities.Finally, to account for the local economic development, we include in the empirical model the number of companies and the employment rate.It has been highlighted that the effect of economic development on municipal cost efficiency is not trivial (Geys et al., 2010): higher development implies lower spending on social benefits (a 'cost effect') as well as a greater demand for costly public services (a 'preference effect').In line with Geys et al. (2010), the positive impact of both the number of companies and the employment rate on municipalities' inefficiency shows that the preference effect outweighs the cost effect in local governments.
With the aim of providing brief considerations about municipalities' cost-efficiency performance, Figure 2 reports the density of efficiency scores estimated by adopting the full model specification (model 3).We obtain a narrow distribution of efficiency scores of close to 0.9 for the model highlighting the high cost efficiency of Italian municipalities.
Further, in order to take into account the regional perspective, considering the historic dualism of the Italian territory, we estimated the full model (model 3) using two restricted subsamples, limited to North and Centre-South regions, respectively. 14Overall, the empirical findings show that the considerable heterogeneity between the different areas of Italy does not affect the positive effect of female political empowerment on municipalities' efficiency.The negative effect of the presence of a female mayor on municipalities' inefficiency preserves its significance and magnitude in both the territorial restrictions.However, we find that the territorial geography affects the effect of female council representation: to increase the share of women in the municipal councils has a stronger and higher significant effect on the inefficiency level of northern municipalities, while the effect for municipalities located in Centre-South is lower and weakly significant.Note: Values are mean (standard deviation).Female Mayor is a dummy variable assuming a value of 1 for municipalities headed by a woman; share of Women in the Council is the ratio of female to total council members; average Age of the Council is the average age of the members of municipal council; share of highly educated employees (Education) is the share of municipal employees having at least a secondary school diploma; Population 3-14 Years is the share in terms of the total population; Population Density is people per km 2 of land area; Land Size is the km 2 of land area; Waste Sorted Collected is the share of total waste; companies (Local Units) is the per capita value of all local companies; and Employment Rate is the regional value.
Female presence in local government and cost efficiency: the case of Italian municipalities 23 REGIONAL STUDIES

Robustness check
In order to take into account further differences between Italian municipalities potentially able to impact the gender effect on cost efficiency, we perform several robustness checks.
The first robustness check takes into account the dimension of municipalities, exploiting the institutional differences related to municipalities' size established by Law n. 56/2014.In order to increase female presence in municipal offices, the law introduced a gender quota, establishing that in municipalities with more than 3000 residents neither gender could be represented by less than 40%.Further, the same law increased from two to three the number of possible consecutive mandates for the mayor of municipalities with fewer than 3000 residents.Therefore, we investigate if the impact of female political empowerment changes depending on municipalities' size, restricting the estimation of model 3 to smaller municipalities, setting the threshold to 3000 inhabitants.The estimation results (column R1 in Table 4) show that the main results are not affected by the institutional differences, and the absence of a gender quota.The positive effect of a higher share of women in the municipal councils preserves its high significance, such as its negative impact on inefficiency.The presence of a female major implies a reduction of municipalities' inefficiency, even if it loses its significance. 15 The second robustness check is based on the consideration that the financial conditions of municipalities could differ significantly in the presence of a financial distress, a special procedure of insolvency introduced by Law n. 66 in 1989 to respond to the growing financial problems of local authorities (Antulov-Fantulin et al., 2021).Results in column R2 of Table 4 show that the effect of female empowerment on municipal inefficiency remains negative and significant when municipalities experimenting a financial distress in the period considered are excluded from the sample.
Third, we investigate if a change of the mayor and of political administrations in the time span considered could affect the gender effect on efficiency.Therefore, we limited the estimation of the main model to the subsample of municipalities experimenting at least one change of the mayor and of political administrations in the period considered (column R3 of Table 4).The empirical findings show that the effect of both a female mayor and the share of women in the council on municipal inefficiency is still negative and significant, implying the persistent female effect when there is political change in the administration.
Finally, the fourth robustness check is performed with the aim to verify if our empirical results could depend on the model specification adopted.Considered that our estimates are based on a translog function, we perform an additional model estimation using the less flexible Cobb-Douglas specification for the cost function.The results (column R4 of Table 4) confirm our main evidence, showing that the significant and negative effect of female empowerment on municipal inefficiency is robust to the cost function implemented.

CONCLUSIONS
This paper evaluates the effect of female political empowerment on local governments' cost efficiency by analysing Italian municipalities in 2010, 2013, 2015 and 2016.
The significant empowerment of women in public governance occurred in Italy during the last decade combined with the considerable heterogeneity of the different areas of the territory make Italian municipalities an interesting case study.
The empirical findings provide an original contribution to the existing literature, showing that female political empowerment is significant in affecting municipalities' cost efficiency, an issue not well documented so far in the empirical literature.We show that to have a female mayor and a higher share of women in the municipal council favours municipalities' cost-efficiency performance.The positive effect of a female presence in local governments is confirmed when additional contextual factors able to explain municipalities' efficiency are included in the estimation model.
The empirical analysis exploits a large sample of Italian municipalities belonging to different regions allowing one to take into account the regional perspective, and in particular the historic North-South dualism of territory.The empirical results show that a higher female presence in the council has a stronger effect on cost efficiency for municipalities located in northern areas, while the positive effect for the presence of a female mayor is similar when the territorial analysis is performed.Female presence in local government and cost efficiency: the case of Italian municipalities 25

REGIONAL STUDIES
Finally, in order to verify if further differences between Italian municipalities could impact the gender effect on cost efficiency we perform several robustness checks.The empirical findings are confirmed when we take into account the presence of institutional differences and the absence of a gender quota, linked to municipalities' size.No different results are due to the exclusion from the sample of municipalities having different financial conditions, documented by the experience of a financial distress.Further, the gender effect in politics is also confirmed when taking into consideration any change of the mayor and of political administration that has occurred during the sample period, and when we adopt of a different model specification.
The policy implications of our findings are quite interesting in the light of the consideration that for the most of the countries, in both the developed and developing worlds, to reduce gender disparities represents an urgent need, included in the environmental, social and governance strategy.We provide significant support for the usefulness of public regulations aimed at favouring a female presence and at promoting gender equality.Such findings are in line with the European Commission's (2020) strategy, highlighting the need for a higher female presence in leadership positions, in particular in political institutions.During the last decades, the Italian government has been very active in gender strategies, ranging from the introduction of gender quotas to the promotion of gender mainstreaming and gender balance, measures aimed at favouring gender equality overall in social, economic and political contexts and at introducing tools able to monitor the impact of public action.The large adoption of gender strategies raises interesting issues and provides useful instruments for future researches, in particular allowing one to verify if the strong association between female public employment and municipality efficiency is robust to an updated analysis, taking into account also the effect of the COVID pandemic crisis.Note: For definitions of the variables, see Table 2.The models were estimated by taking into account: municipalities with a number of inhabitants less then Valeria Bucci et al.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the author(s).

NOTES
1.The effectiveness of public regulations aimed at favouring female empowerment in politics is still an open debate (e.g., Baltrunaite et al., 2014Baltrunaite et al., , 2019)).2. For an additional survey of the existing literature on local governments' efficiency, see Aiello and Bonanno (2019).
3. For evidence about the effect of female representation on Cabinet duration, see Krauss and Kroeber (2021).4.Among the different parametric functional forms, the translog function is the most commonly used in the empirical applications, and it represents a generalization of the Cobb-Douglas function, which assumes homogeneity, unitary elasticity of substitution between inputs and separability.5.For a review of stochastic frontier models, see Ferrara (2020).6. Italy is divided into 15 ordinary statute regions (OSRs) and five at special statute regions (SSRs), mainly differing in financing.We focus exclusively on OSRs, limiting the potential bias related to the different organization that such regions can self-regulate independently (Braga & Scervini, 2017) 8.The weights were obtained through BoD approach.9. See note 8. 10.For territory services, eight micro-outputs have been used: environmental ordinances, authorizations conceded, reports faced, opinions of environmental compatibility, competitive bidding procedures and direct commitments, pruned plants during the year, new trees planted, and cleaning of rivers and streams.The micro-outputs used for viability are nine: projects, operations management, progress status and final status, trials, competitive bidding procedures and direct commitments, contracts, salt interventions and snowplough, concessions of public land for construction or archaeological site, and disabled parking and accesses for traffic-restricted zone allowed or renewed during the year.11.The potential spatial heterogeneity in prices is mitigated by the presence of a public company (Consip SpA) which centralizes public procurement in Italy.
12. In our sample, the total expenditure for essential services is €683 per capita on average in 2016, while the total public expenditure accounts for about €14,000 per capita, and the gross domestic product (GDP) per capita is about €26,000 in the same year (Eurostat, 2022).13.We estimated the three model specifications also using a balanced dataset and neither the signs nor the statistical significance of all the control variables change.The results are available from the authors upon request.14.North regions: Emilia-Romagna, Liguria, Lombardia, Piemonte and Veneto; Centre-South regions: Abruzzo, Basilicata, Calabria, Campania, Lazio, Marche, Molise, Puglia, Toscana and Umbria.15.We perform the robustness check also on the subsample of bigger municipalities (i.e., those having more than 3000 inhabitants), which confirms that both the presence of a female mayor and a higher share of women in the municipal councils preserve their negative and significant effect on inefficiency.The results are available from the authors upon request. 5 4. DATAOur empirical analysis is based on a dataset providing information concerning economic, socio-demographic and institutional aspects of Italian municipalities.The data cover a sample of more than 5000 Italian municipalities belonging to ordinary statute regions 6 observed over four years: 2010, 2013, 2015 and 2016.The discontinuity in the panel is due to the unavailability of data concerning Female presence in local government and cost efficiency: the case of Italian municipalities 19

Figure 2 .
Figure 2. Distribution of efficiency scores for local governments.

Table 1 .
Descriptive statistics: costs, outputs and prices.Values are mean (standard deviation).All expenditure and output variables are per capita measures; expenditure variables are expressed in thousands of euros; waste output is expressed in tons; population is expressed in thousands; the average local cost of renting is measured in monthly euros per m 2 ; and average revenues are expressed in thousands of euros.Sources: Data are obtained by integrating the information from different sources: Italian automobile association (ACI); Italian Institute for Environmental Protection and Research (ISPRA); Italian National Institute of Statistics (ISTAT); Italian Ministry of Economy and Finance (MEF); Italian Ministry of the Interiors (MI); real estate market observatory of the Italian Revenue Agency (OMI); and Solution for the Economic System (SOSE), an agency of the MEF specialized in econometric analysis. Note:
Note: For definitions of the variables, see Table2.
. 7. Municipalities' expenditure and output data are provided by the database OpenCivitas, an initiative promoted by the Ministry of Economy and Finance and SOSE.The project was born with Law 42/2009 and Legislative Decree 216/2010 with the aim of defining for Italian municipalities the standard expenditure needs, in relation to territorial features and socio-demographic characteristics of the resident population.Such values, used to distribute intergovernmental funds in an equitable and transparent way, are collected through a detailed questionnaire for municipalities in different waves: 2010, 2013, 2015 and 2016, which are our years of analysis.