Introduction
Recurring financial crises in American local governments bring attention to the priority issues in financial resource allocation. The choices of local governments to spend their money among different policy areas are influenced by various factors. The importance and proportion of local welfare spending has been increasing over the past few decades (e.g. Witco and Newmark Reference Witco and Newmark2009). Redistributive services, such as welfare, health and public safety, are traditionally provided by counties as political subdivisions of state governments (Schneider and Park Reference Schneider and Park1989; Berman Reference Berman1993; Choi et al. Reference Choi, Bae, Kwon and Feiock2010; Farmer Reference Farmer2011). Most of these redistributive services are mandated and take up a large share of county spending (Schneider and Park Reference Schneider and Park1989; Park Reference Park1996). Among the spending categories of California counties, public protection and public assistance take up more than 25 per cent of total county expenditures.
This paper examines factors that affect a county’s preferences on welfare spending in California, focusing on the gender representation of county boards of supervisors and the state legislature. Social welfare is viewed as a gendered policy domain, as it is particularly related to the interests of minorities including women (Riccucci and Meyers Reference Riccucci and Meyers2004; Wilkins and Keiser Reference Wilkins and Keiser2006). Despite a relatively large proportion of women at the local level, the role of female local officials and its impact on the politics of financial choices is largely underexplored (Garber and Turner Reference Garber and Turner1995; MacManus Reference MacManus1996; Fox and Schuhmann Reference Fox and Schuhmann2000; Alozie and McNamara Reference Alozie and McNamara2010). We seek to explore this and also attempt to test whether women officeholders moderate the impact of the local economy and institutions on welfare spending, because social welfare is not only related to gender, but is also politically vulnerable during an economic recession.
California counties constitute an excellent testing ground for the impact of various factors including gender representation on welfare spending. First, local governments in California have been through notable institutional changes under continuing fiscal constraints and the growing service demands of an ever-increasing population in the state. Notably, California recently suffered from a bubble-burst caused by the stock market decline in early 2000 (Musso et al. Reference Musso, Graddy and Grizard2006). In response, Governor Schwarzenegger proposed several solutions, including significant expenditure cuts for schools, health and human services, local assistance and state employee compensation (Spilberg and Alexander Reference Spilberg and Alexander2003).Footnote 1
Second, based on its 1879 constitution, California has historically given wide autonomy to local governments in structural, programmatic and fiscal areas (Sokolow and Detwiler Reference Sokolow and Detwiler2001, 59). In 1911, “California was the first state to amend its constitution to permit home rule for counties” (Menzel Reference Menzel1996, 5). Local governments with home rule charter usually have discretionary power in choosing the forms of government, functions and services they provide and in dealing with managerial and fiscal issues. Although Proposition 13 in 1978 has limited local autonomy, several counteractive institutional measures have been taken since then.
Third, there exists a considerable difference in fiscal, political and administrative capacities among counties (Berman Reference Berman1993). In particular, California counties are large and diverse enough to examine redistributive efforts within counties in relation to institutional, political and economic factors. For example, Alpine County has the smallest county population at little more than 1,000 persons, while Los Angeles County has a population of nearly ten million, which is more than some states in the United States have in total, including Georgia, Indiana, Washington and Massachusetts. In addition, there are significant variations in welfare spending across counties. In terms of the size of the total welfare budget in 2010, Los Angeles County spends more than 5,000 times what Alpine County does.
One of the prominent issues in county research is how leadership affects policy choices at the county level (Menzel Reference Menzel1996; Svara Reference Svara1996; Benton Reference Benton2005). A substantial amount of evidence has been collected in the public management scholarship on the effect of the quality of leaders/CEOs/managers on organisational outcomes or results (e.g. Meier and O’Toole Reference Meier and O’Toole2002). MacManus (Reference MacManus1996) offers a comprehensive description of governing boards, including the form, name, size, partisanship and term of office. Avellandeda (Reference Avellandeda2009) reveals that mayoral qualities, such as educational background and job-related expertise, have been associated with local public finance in 40 Colombian municipalities over the five years between 2000 and 2004. This research also shows that the positive influence of mayoral qualities on property tax collection and social spending per capita is moderated by external constraints (Avellandeda Reference Avellandeda2009).
Among various leadership qualities, we focus on gender representation in county and state leadership positions as an important factor in the politics of redistribution at the local level.Footnote 2 To address this link in the context of local government expenditures, we first use models to capture any differences that the gender variable may have caused in spending decisions on welfare. Subsequently, we take into account the economic and institutional factors that might affect the degree of influence of the gender variable. By including interaction terms in the model, this paper explores the mediating effect of the county leadership role in times of economic hardship and in counties with home rule. We test conditional hypotheses that the effect of female supervisors on county boards depends on the economic situation, measured by the unemployment rate at the county level every year, and the institutional situation, measured by the charter status of counties as a time-invariant factor.
Counties under fiscal stress
The service role of county governments has been significantly expanded in recent years (Benton Reference Benton2002a; Streib et al. Reference Streib, Svara, Waugh, Klase, Menzel, Salant, Byers and Cigler2007). Since the state-local realignment proposals in 1991 and 1994–1995, counties have had more flexibility and fiscal incentives to determine budget decisions based on local needs.Footnote 3 With little difference from municipal governments, county governments practically function as a general-purpose government (Benton Reference Benton2002a). In particular, California counties become a primary general government to the residents in regions that are not incorporated to cities. Despite the public support and recognition of counties as “full-service governments” (Menzel Reference Menzel1996; Benton Reference Benton2002a, 196), however, scholars have devoted relatively little attention to county governments.Footnote 4
Counties are required to play several roles in the state-local relationship. On the one hand, they operate as the administrative arm of the state government within federalism, and on the other, they serve as the direct providers of services that meet the needs of their residents (Ostrom and Ostrom Reference Ostrom and Ostrom1988; Berman Reference Berman1993; Berman and Salant Reference Berman and Salant1996; Benton Reference Benton2002a; Chapman Reference Chapman2003; Choi et al. Reference Choi, Bae, Kwon and Feiock2010). As an important coordinating unit in the state-local relationship, counties with large jurisdictions take on complicated issues, such as environment, transportation and other urban services (Benton Reference Benton2002b, 475).
Counties have a considerable amount of flexibility in applying mandates (Morgan and Kickham Reference Morgan and Kickham1999; Sokolow and Detwiler Reference Sokolow and Detwiler2001; Streib et al. Reference Streib, Svara, Waugh, Klase, Menzel, Salant, Byers and Cigler2007). In particular, urban counties have more options in choosing the specific range of services and programmes they provide (Morgan and Kickham Reference Morgan and Kickham1999). Oakerson and Parks (Reference Oakerson and Parks1989, 290) admit that counties are a legal and political subdivision of state government, but the conception of “creature of the state” is “both empirically incorrect and normatively misleading”. Under fiscal federalism, there is a division of economic responsibilities between the federal, state and local governments (Stiglitz Reference Stiglitz2000, 728).
Although the division of responsibilities between state and local government is complicated, providing redistributive services has been one of the most important responsibilities of counties.
For example, California’s child welfare service programmes are provided by the Department of Social Services in the state government and by county agencies, including county welfare departments and county social service departments.Footnote 5 Current state law requires all counties to establish specialised organisational entities within county welfare departments that have sole responsibility for the operation of the child welfare services programmes (California Department of Finance 1997, 1). The state has discretion to determine the services to be provided and the eligibility criteria for receipt of services; however, counties have greater discretion in the structuring of programmes and the allocation of resources under the state’s block-grant funding (California Department of Finance 1997).Footnote 6 The State Department does not have enough information and the counties have different criteria for determining cases in the programme. In a study of the Temporary Assistance for Needy Families programme, Riccucci and Meyers (2004) also posits that programme goals vary across counties even though policy goals are centralised at the state level. Thus, the success of local welfare services depends on whether the state obtains commitments from the county agencies.
Accordingly, counties, as a local service provider and the first contact with clients, have fairly high levels of discretion in implementing the service delivery systems in which the funds are spent (Pagano Reference Pagano1997; Dougherty et al. Reference Dougherty, Klase and Song2003; Krane et al. Reference Krane, Ebdon and Bartle2004). In other words, the local budget decisions on welfare will pretty much depend on the preferences of county officials in welfare service agencies. In particular, county officials on the boards exercise the executive, legislative and quasi-judicial authority of the county. In partnership with county staff, county supervisors are deeply involved in the delivery of services and programmes to the county region. They set priorities on budgets, supervise county officials and spend money on programmes aimed to meet the needs of county residents (California State Association of Counties 2009).
With decreasing revenue from other levels of government and increasing demand for urban services, counties are under fiscal pressure to balance their budgets (Pagano Reference Pagano1997; Dougherty et al. Reference Dougherty, Klase and Song2003). Local officials in counties must therefore choose which services they will provide and find more efficient and effective ways to provide these services (Cigler Reference Cigler1994; Benton Reference Benton2002a, 204).Footnote 7 County governments have the flexibility to create or distribute surplus funds by changing expenditure categories (Pagano Reference Pagano1997; Morgan and Kickham Reference Morgan and Kickham1999; Dougherty et al. Reference Dougherty, Klase and Song2003) and to resist unfunded state mandates (Berman and Salant Reference Berman and Salant1996). Several studies suggest that state governments under fiscal pressure allow more discretion, transfer state and federal responsibilities, and impose unfunded mandates on local governments (Berman Reference Berman1993; Krane et al. Reference Krane, Ebdon and Bartle2004; Sosin et al. Reference Sosin, Smith, Hilton and Jordan2010). Since October 2011, the California state government has expanded mandates, including those related to Child Welfare Services, Adult Protective Services, mental health, and drug and alcohol programmes to counties (Danielson and Mejia Reference Danielson and Mejia2011).
Hypotheses
Gender and welfare politics
This study aims to test whether female officeholders at the state and local levels make a difference in local welfare expenditures. Women are playing an increasingly important role in local politics (Fox and Schuhmann Reference Fox and Schuhmann2000; Alozie and McNamara Reference Alozie and McNamara2010). Compared to the studies of women in state legislatures and Congress, however, few works exist that attempt to analyse the impact of women in local office (MacManus and Bullock Reference MacManus and Bullock1995).
The concept of representativeness or diversity has received more attention since traditional representative institutions like legislatures have been challenged in dealing with complex issues in communities.Footnote 8 The literature on representative bureaucracy offers a theoretical framework for this research. It assumes that the presence or increased presence of minorities in bureaucracy will have a substantive effect on policy choices or outputs. Minority bureaucrats can produce substantive benefits for their social groups directly through their own administrative behaviour and indirectly by influencing others’ behaviours (Lim Reference Lim2006).Footnote 9
The evidence on gender difference in policy preference is quite strong. Women have a more distinctive focus on “women’s issues” as policy priorities, including policies that impact women, children, family, welfare and health, while focusing less on other issues, such as economic development (Burns and Schumaker Reference Burns and Schumaker1987; Schumaker and Burns Reference Schumaker and Burns1988; Saint-Germain Reference Saint-Germain1989; Thomas Reference Thomas1991; Carroll Reference Carroll1994, Reference Carroll2003; Schlesinger and Heldman Reference Schlesinger and Heldman2001; Dolan Reference Dolan2002; Keiser et al. Reference Keiser, Wilkins, Meier and Holland2002; Stivers Reference Stivers2002; Swers Reference Swers2002; Wilkins Reference Wilkins2006; Wilkins and Keiser Reference Wilkins and Keiser2006; Alozie and McNamara Reference Alozie and McNamara2010). Several scholars argue that the gender gap is distinctive enough that it cannot be erased by organisational socialisation (Dolan Reference Dolan2002) and will make a substantial contribution rather than a symbolic one (Carroll Reference Carroll2003; Lim Reference Lim2006; Park Reference Park2013).
An interesting question here is whether the difference in attitudes and priorities between men and women makes a real difference in policy outcomes (Dolan Reference Dolan2002; Andrews et al. Reference Andrews, Boyne, Meier, O’Tool and Walker2005; Wilkins and Keiser Reference Wilkins and Keiser2006; Bradbury and Kellough Reference Bradbury and Kellough2007; Jacobson et al. Reference Jacobson, Palus and Bowling2010; Park Reference Park2013). Women and minorities in high-level positions are more likely to push for social welfare programmes (Keiser et al. Reference Keiser, Wilkins, Meier and Holland2002; Riccucci and Meyers Reference Riccucci and Meyers2004; Wilkins Reference Wilkins2006; Bradbury and Kellough Reference Bradbury and Kellough2007), which might link demographic or passive representation to substantive or active representation in specific areas (Dolan Reference Dolan2002; Keiser et al. Reference Keiser, Wilkins, Meier and Holland2002; Riccucci and Meyers Reference Riccucci and Meyers2004; Meier and Nicholson-Crotty Reference Meier and Nicholson-Crotty2006; Wilkins and Keiser Reference Wilkins and Keiser2006).Footnote 10
Studies have found that women leaders and officeholders bring up different policy issues and solutions by exercising a different leadership style from that of their male counterparts (Fox and Schuhmann Reference Fox and Schuhmann2000; Carroll Reference Carroll2003; Mandell and Pherwani Reference Mandell and Pherwani2003). For example, Selden (Reference Selden1997, 69) explained the variations in housing loan approval decisions by county supervisors with the concept of the “minority representative role”. She argues that the mediating effect of the minority representative role is more important than minority status itself. Furthermore, empirical research has identified the group-benefitting behaviour of women in legislative bodies (Saint-Germain Reference Saint-Germain1989; Thomas Reference Thomas1991; Carroll Reference Carroll1994; Bratton and Haynie Reference Bratton and Haynie1999; Saidel and Loscocco Reference Saidel and Loscocco2005). Women legislators are more likely to initiate bills on issues associated with women’s concerns. The legislative agendas of female state legislators display different policy priorities and pursue distinctive legislative politics as compared to male legislators.
Given the gendered characteristics of the welfare policy area, policy consequences are expected from an increased number of females represented in local governments and state legislature. We include gender representation in the state legislature because of the dynamics of California’s budget process and the state influence on local budget. We expect that there will be a positive relationship between the percentage of women officeholders in county boards and the state legislature and the portion of the county budget spent on welfare. Further, we examine whether or not the influence of gender representation on welfare spending has a linear relationship.
(H1a): County welfare spending will be positively associated with an increase in the number of women supervisors on county boards.
(H1b): County welfare spending will be positively associated with an increase in the percentage of women in state legislature leadership positions.
According to literature on representative bureaucracy, the presence of minorities, i.e. passive representation, does not guarantee policy choices for minorities (Riccuci and Meyers Reference Riccucci and Meyers2004; Lim Reference Lim2006; Bradbury and Kellough Reference Bradbury and Kellough2007); rather, active representation is needed. Active representation occurs when some specific conditions are met, such as critical mass of minority groups (Kanter Reference Kanter1977), policy discretion (Keiser et al. Reference Keiser, Wilkins, Meier and Holland2002) and institutional context and policy areas (Wilkins and Keiser 2006). This line of reasoning leads us to a detailed hypothesis that expects a moderating effect on local welfare decisions when county leadership includes women supervisors. We identify the economic and institutional conditions that expand the degree of discretion of local governments. We expect that an increase in the number of women officeholders has a substantive effect on local welfare choices in counties under economic pressures and/or in counties with home rule. Assuming that women officeholders may not always have the same attitude in dealing with fiscal challenges, we expect that there may be an interactive effect for the leadership and economy variables. Women officeholders may play the role of minority representative when the economy worsens, since state governments may allow more local discretion (Berman Reference Berman1993; Krane et al. Reference Krane, Ebdon and Bartle2004; Sosin et al. Reference Sosin, Smith, Hilton and Jordan2010), and when policy choices benefitting women and minorities are at stake. Considering the gendered nature of social welfare, we expect that women officeholders will be more protective of welfare expenditures when they are threatened by economic and fiscal crises. As the degree of administrative discretion available to county boards is a key issue linking passive and active representation in this study, we also expect that institutional conditions increasing local discretion will positively moderate the effect of women supervisors on local welfare spending.
Conditional hypotheses have been constructed, suggesting that the effect of women supervisors on county boards may depend on economic and institutional factors. More specifically, an increase in welfare expenditures is positively associated with an increase in the number of women on county boards when the county economy goes bad and/or counties have home rule. As we are interested in the effect of gender representation in county governments, we specify the following hypotheses:
(H2a): During an economic downturn, county welfare spending will be positively associated with an increase in the number of women supervisors on county boards.
(H2b): In county governments with home rule, county welfare spending will be positively associated with an increase in the number of women supervisors on county boards.
Three categories of control variables are discussed below: institutional factors, such as the introduction of Proposition 1A and chartered or general-law counties; political factors, such as the support for the democratic presidential nominee of each county and the strength of interest groups; and socio-economic factors, such as intergovernmental revenue, unemployment rates, the percentage of females in the population, the percentage of the population over 65 and under 14, per capita personal income and median income.
Institutional factors
County home rule
Several studies have presented evidence that the government structure associated with a home rule charter affects the spending of county governments (Duncombe Reference Duncombe1977; DeSantis and Renner Reference DeSantis and Renner1994; Benton Reference Benton2002a, 2002Reference Bentonb; Choi et al. Reference Choi, Bae, Kwon and Feiock2010). Efforts to modernise city and county governments are closely related to home rule, which involves greater autonomy in structural, functional and fiscal areas (Benton Reference Benton2002b, 473). Often, cities and counties adopt charters because of structural reasons rather than functional ones (Sokolow and Detwiler Reference Sokolow and Detwiler2001, 61). However, the effect of home rule seems more limited in counties than cities, because counties are more dependent on state control compared to cities (Sokolow and Detwiler Reference Sokolow and Detwiler2001; Benton Reference Benton2002b). Yet, we assume that charter status affects counties in the same way it does cities (Benton Reference Benton2002b, 474). Chartered counties are expected to respond to the needs of citizens more effectively (Benton Reference Benton2002b; Chapman Reference Chapman2003; Choi et al. Reference Choi, Bae, Kwon and Feiock2010).Footnote 11
Although we expect that an institutional condition that increases local discretion will be associated with positive moderating effects on local welfare spending, the effect of county home rule on welfare spending may be complex. On one hand, counties with home rule have an incentive to focus more on services that match local needs. Average taxpayers in counties may choose to expand redistributive policies for social responsibility and long-term positive policy outcomes (Moane and Wallerstein Reference Moane and Wallerstein2001; Farmer Reference Farmer2011).Footnote 12 In this case, charter status may be positively associated with welfare expenditures. However, counties are also tempted to focus on generating revenue with economic development rather than social welfare (Peterson Reference Peterson1981; Benton Reference Benton2002b; Choi et al. Reference Choi, Bae, Kwon and Feiock2010; Farmer Reference Farmer2011). According to Peterson (Reference Peterson1981), localities have distinct preferences for developmental policy areas rather than redistributive programmes, because they are afraid of losing residents and sources of revenue (Bahl et al. 2002; Stein Reference Stein2003; Farmer Reference Farmer2011).
Proposition 1A in 2004
The approval of Proposition 1A in 2004 is one of the important institutional changes to affect local finance after Proposition 13. Enacted in 1978, Proposition 13 forbade local officials from setting their own property tax rates in order to protect homeowners. The direct effect of Proposition 13 was a sharp drop in property taxes, whereas the long-term effect of this measure was more complex than expected.Footnote 13 Proposition 13 fundamentally undermined the fiscal and political independence of local governments in California (Sokolow and Detwiler Reference Sokolow and Detwiler2001, 63; Chapman Reference Chapman2003; Danielson and Mejia Reference Danielson and Mejia2011; Misczynski and Mejia Reference Misczynski and Mejia2011). Correlatively, tensions between the state and local governments have been growing since the proposition (Chapman Reference Chapman2003; Misczynski and Mejia Reference Misczynski and Mejia2011).
As a response of local authorities, there have been several attempts to modify the state-local relationship and to limit the state’s fiscal interference (Danielson and Mejia Reference Danielson and Mejia2011; Misczynski and Mejia Reference Misczynski and Mejia2011). In November 2004, Proposition 1A was approved as a way to protect local government budgets from being transferred to the state government and to prohibit unfunded state mandates. As a consequence, local governments have more flexibility in managing local programmes, as this proposition limits state control over local property tax and local sales tax.
Political factors
Political ideology: Democratic votes in presidential elections
As choices for redistributive services are closely related to state and local politics, it is essential to include political variables in the study of local finance. We expect that the political/ideological attitudes of county residents influence local spending decisions on welfare. The influence of the ideological environment of the local community on bureaucratic output has been elaborated in the Kaufman’s (Reference Kaufman1959) classic study. The political preferences of the public are reflected in the choices of the people who have power, according to pluralist theory (Stein Reference Stein2003). Most elected county officials are willing to respond to local needs that reflect the political preference of county residents (Park Reference Park1996). As such, it is necessary to control for political/ideological preference when examining the factors that result in different spending levels in different policy areas (Park Reference Park1996; Fox and Schuhmann Reference Fox and Schuhmann2000).
It is widely acknowledged by scholars that Democrats are more favourable to redistributive policies than Republicans, while Republicans are more likely to increase spending on developmental policies (Fry and Winters Reference Fry and Winters1970; Keiser Reference Keiser1997; Mead Reference Mead1999; Alt and Lowery Reference Alt and Lowery2000; Nicholson-Crotty et al. Reference Nicholson-Crotty, Theobald and Wood2006; Witco and Newmark Reference Witco and Newmark2009; Choi et al. Reference Choi, Bae, Kwon and Feiock2010; Soss et al. Reference Soss, Fording and Schram2011). Thus, we hypothesize that the more votes cast in a county for a presidential nominee from the Democratic Party, the more the county will spend on welfare. The political ideology of county residents is measured by counting the Democratic vote share in general presidential elections every four years.
Strength of interest groups: number of non-profit organisations
We consider interest groups as one of the political actors in local politics related to welfare. Interest groups are known to have a negative effect on the welfare of lower income populations, because these populations are not represented by interest groups (Reference SchattschneiderSchattschneider Reference Schattschneider1960; Fry and Winters Reference Fry and Winters1970). Yet, under the competition for funding, interest groups with professional expertise may influence the behaviours of public officials in various ways (Fry and Winters Reference Fry and Winters1970; Chapman Reference Chapman2003). We expect that interest groups having common interests in human services may contribute to increasing, or at least maintaining, the level of welfare spending in local governments.
Socio-economic factors
Intergovernmental revenue: state and federal aid
Revenue from other levels of government accounts for a substantial share of county revenue (Duncombe Reference Duncombe1977; Park Reference Park1996). Most of the social welfare programmes in California, such as California Work Opportunity and Responsibility to Kids and In-Home Supportive Services, are funded by the federal and state levels of government. Several important programmes, such as welfare, education and health, are supported by a mix of federal, state and local funding (Misczynski and Mejia Reference Misczynski and Mejia2011). State and local government provide different services with different funding sources, although there is no clear division in programmes and funding among the levels of government. Considering the stimulative effect of funding (Courant et al. Reference Courant, Gramlich and Rubinfeld1979; Benton Reference Benton1992), we expect more local spending on the areas aided from the higher levels of government. Therefore, it is necessary to control intergovernmental transfers to identify factors affecting county spending on redistributive services (Park Reference Park1996; Witco and Newmark Reference Witco and Newmark2009, 219).
Economic hardship: unemployment rate
The relationship between economy and local public finance is well established. Early political scientists seem to agree that socio-economic variables account for local expenditures better than political variables do (Fabricant Reference Fabricant1952; Dye Reference Dye1966; Sharkansky and Hofferbert Reference Sharkansky and Hofferbert1969). Since California counties have suffered from chronic economic recessions, the potential effect of this economic variable will be considered in the redistributive expenditures in California counties.Footnote 14 However, the effect of economy on local welfare services can be equivocal (Watkins-Hayes Reference Watkins-Hayes2011). In times of economic hardship, local governments experience both a decrease in the state budget and a counties’ own revenue sources (Berman 1993; Berman and Salant Reference Berman and Salant1996; California State Association of Counties 2009) and an increase in local needs for welfare at the same time.
Demographics: the ratio of female population, population over 65, population under 14, per capita personal income and median income
In general, demographic characteristics reflect local demands for public services. The composition of the population can be a strong determinant of local choices on spending (Dye Reference Dye1966; Witco and Newmark Reference Witco and Newmark2009). Counties may increase or decrease spending on certain programmes according to demands for the provision of the redistributive services (Benton Reference Benton2002a). California counties vary greatly in size, economy, geography and demographics, which leads them to have different demands on welfare. In the end, we need to consider the characteristics of the communities in which the politics of redistribution occur (Fry and Winters Reference Fry and Winters1970; Fox and Schuhmann Reference Fox and Schuhmann2000).
County choices for redistributive services may depend on community characteristics, such as population size and growth (Schneider and Park Reference Schneider and Park1989; Park Reference Park1996; Choi et al. Reference Choi, Bae, Kwon and Feiock2010; Farmer Reference Farmer2011), problem intensity (Fosset and Thompson Reference Fosset and Thompson2005) and racial diversity (Fosset and Thompson Reference Fosset and Thompson2005; Choi et al. Reference Choi, Bae, Kwon and Feiock2010; Farmer Reference Farmer2011). We expect that there is a greater need for redistributive services in counties with a higher percentage of minority groups that may benefit from welfare programmes. Women, children and seniors are also included as target groups for services in California counties, as these counties already have various social service programmes assisting these groups. We also consider the wealth and income of county residents. Counties with wealthy residents may spend more on growth-oriented, developmental services than redistributive, social services (Park Reference Park1996; Bahl et al. Reference Bahl, Martinez-Vazquez and Wallace2002). The wealth and income of county residents are captured by the per capita personal income and median income.
Data and Methods
Our hypotheses are tested with panel data across 58 counties in California over ten years (2001–2010). Data are drawn from various sources, such as the Counties Annual Report, the Bureau of Labor Statistics, RAND California and the California Elections Data Archive (CEDA). Data and measures are shown in Table 1.
Table 1 Data and measures
Note: All financial data are adjusted for inflation based on statewide Consumer Price Indexes for urban consumers using 2011 dollars. The variables with (L) are lagged one year.
This paper uses expenditure data to measure the decisions or choices of county governments with regard to welfare services. The expenditure data is useful for capturing the trends of county choices in funding redistributive services (Benton Reference Benton2002a). The dependent variable is the amount of county expenditures on public assistance, including welfare, social services, general relief and other public assistance programmes. Care of court wards and veteran services are included in this category according to the official classification. We tested the variable total welfare spending of the model specification for stationarity with the Levin-Lin-Chu test (time trend included) and found evidence against the null hypothesis of a unit root. We therefore concluded that total welfare spending is stationary. The changes in the total welfare spending of counties with a representative shape over ten years are presented in Figure 1. The trends show no consistent shape after inflation adjustment – they are either increasing or decreasing, showing smooth changes or changes with fluctuations.
Figure 1 Changes of total welfare spending of nine counties with a representative shape by year.
Our major independent variable is the gender representation in state and local government. For the causal inference based on the temporal precedence condition, all the independent variables except a charter dummy were lagged one year. Two measures of gender representation at different levels of government are (1) the number of women on the board of supervisors of each county and (2) the percentage of women officeholders in state legislature leadership positions. Two conditional hypotheses are tested using multiplicative interaction models (Brambor et al. Reference Brambor, Clark and Golder2006). The interaction variables were developed to further explore the county leadership role in chartered counties and in times of fiscal stress and its impact on policy choices related to welfare.
In counting the number of female supervisors-elect, we check the name of the successful candidates in CEDA and identify the supervisor’s gender through a name gender guesser (www.genderguesser.com) and through search websites if the gender is not clear. California counties each have five supervisors with four-year staggered terms on a non-partisan ballot (www.guidetogov.org).Footnote 15 Two supervisors are elected in one general election and three supervisors in the next. The four-year staggered term of county supervisors posed a challenge in counting the number of women supervisors-elect in each county. To sort out the effect of incumbents, we used the number of successful female candidates rather than the proportion of women in county boards.
A state legislature is a legislative branch of the government that holds the principal law-making powers of the state (www.legislature.ca.gov). Every state has its own legislative body. As a measure of gender representation at the state level, we have used the percentage of women in leadership positions in the California state legislature from the Center for American Women and Politics. It not only reflects the distribution and penetration of women in higher leadership positions, but also shows more dynamic changes than the percentage of women in the state legislature, which has increased gradually and has now reached over 20 per cent (for more about the measures of representation, refer Greene et al. Reference Greene, Selden and Brewer2000). The leadership positions include Senate President and President Pro Tempore, Majority and Minority Leaders, Republican and Democratic Caucus Chairs, Secretary of the Senate, Speakers of the State Assembly including Pro Tempore, Majority Floor Leader, Assistant Majority Whip and Democratic Caucus Chairs, among others.
Intergovernmental revenue is drawn from the RAND California County Finance Statistics and Counties Annual Report. Federal and state aid for social assistance programmes and administration is included. The political/ideological attitude of each county’s residents is measured by the proportion of Democratic votes in the presidential general elections in 2000, 2004 and 2008 in the California State Archives (www.sos.ca.gov).
As a measure for the strength of interest groups, we include the number of registered 501(c)(3) non-profit organisations offering human services in each county.Footnote 16 Major categories include organisations offering child daycare, in-home assistance, family care, family services for adolescent parents, single-parent services, family-violence shelters, pregnancy centres and emergency assistance, as well as specific organisations, such as senior centres, developmental disability centres, women’s centres, ethnic and immigrant centres, homeless centres, blind and visually impaired centres, deaf and hearing-impaired centres and LGBT centres. Demographic variables are from the Bureau of Labor Statistics and United States Census Bureau.
Results and Discussion
Tables 2 and 3 provide descriptive statistics and pairwise correlations among the variables.
Table 2 Summary statistics
Table 3 Pairwise correlation among continuous variables
Note: Coefficients with asterisks are statistically significant at Bonferroni-adjusted 0.05 levels.
To examine the robustness of the results shown in Tables 4 and 5, we have re-estimated the analysis using sensitivity tests. In addition, a number of diagnostics, including the variance inflation factor (VIF) test for multicollinearity, confirm that the assessments in Tables 4 and 5 legitimately judge the influence of these factors.Footnote 17 In the context of this study, however, the city and county of San Francisco has been excluded as an outlier from the analysis, most importantly with regard to its unique status as the only consolidated city-county in California. The number of members on the San Francisco Board of Supervisors is 11, while all other counties have five members on their boards. Given the available data, the exclusion of San Francisco does not affect the results associated with gender representation in the present study.
Table 4 The linear fixed effect models
Note:
† Charter variable is included in the model as a constitutional term of the interaction models but omitted because of collinearity. The results are not affected by the inclusion.
L1, variables that lagged one year; LN, variables with natural logarithm.
***p<0.01, **p<0.05, *p<0.1, SE noted in parentheses are adjusted for 57 clusters in id (county).
Table 5 The non-linear fixed effect models
Note:
† Charter variable is included in the model as a constitutive term of the interaction models but omitted because of collinearity. The results are not affected by the inclusion.
L1, variables that lagged one year; LN, variables with natural logarithm.
***p<0.01, **p<0.05, *p<0.1, SE noted in parentheses are adjusted for 57 clusters in id (county).
Panel data can be estimated using several different models according to the assumption of the characteristics of residuals. Based on theoretical reasons and statistical tests, such as the F-test and Hausman test (Hsiao Reference Hsiao2007), the fixed effect model turned out to be an appropriate method to estimate the relative impacts of the independent variables in the equation. A choice of either fixed or random effects estimation can be justified in the theoretical and technical dimensions. However, econometricians now argue that the Hausman test does not help in deciding between fixed and random effects. Instead, more weight is given to the size of the data set, the extent of variability within units and the level of correlation between and within the covariates and units (Clark and Linzer Reference Clark and Linzer2012). The modelling of random effects usually follows strong assumptions, such as a normal distribution and independence of residuals of explanatory variables (without omitted variables), while fixed effects estimators do not rely on as strong assumptions as the random effects model and therefore are not likely to fail. In addition, fixed effects estimation is desired for the purpose of investigating “the effects that are in the sample”, while the alternative is prescribed for that of identifying “the population characteristics” based on a supposedly random sample (Hsiao Reference Hsiao2007). Therefore, the fixed effects approach is preferable for this work, based not only on the Hausman test, but also on the theoretical considerations.
The levels of significance used in this paper are 10 per cent, 5 per cent and 1 per cent. Although the conventional significance level is 5 per cent, the limit for statistical significance was set at 10 per cent (p=0.10) in this paper. Although reducing the alpha level is one way to protect against Type I error, it also increases the chance of a Type II error. The total number of observations is 507, as we lost ten by excluding one panel and the city and county of San Francisco and another 58 when creating interaction or lagged variables. The remaining five missing values are missing from the original data set. Robust SE were clustered by county. The relative impact of the independent variables between the models with and without an interaction term creates little change in sign, strength or statistical significance.
The fixed effects models with and without interaction terms assuming linear and non-linear relationships are tested in this study. From the models accounting for a linear relationship, there are some noteworthy aspects to the findings in Tables 4 and 5. First, gender representation in state legislature leadership positions had a statistically significant positive effect on the amount of total welfare spending in California counties. This result seems particularly noteworthy in light of the fact that there is a positive relationship between women and welfare politics. The variable turns out to be consistently significant across all the models in the study. However, the variable of the number of women on boards of supervisors in each county was not significant in the linear fixed effect models.
Also noteworthy is the fact that the interaction term for the number of women supervisors and county unemployment rate reached the significance level with a positive sign at the 90 per cent confidence interval (with t-value of 1.76). Although the inclusion of interaction terms requires that the coefficients be interpreted carefully, this suggests that women supervisors are willing to spend more on welfare when a county faces financial stress. Since the unemployment rate itself was not significant, this also indicates that macro-economic variables moderated by the county leadership role affect county welfare expenditures.
We can gain a better understanding by being sensitive to the changes in the marginal effect of the variable. Figure 2 presents the changes in the marginal effect of the number of women on county boards on total welfare expenditures. The solid line indicates the marginal effect of the number of women on county boards on unemployment rates, while the dashed line shows the 90 per cent confidence interval around the solid line. Any particular point on the solid line is
$${{\partial {\rm Total}\,{\rm welfare}} \over {\partial {\rm Women}\,{\rm supervisors}}}\,=\,\beta_1\,{\plus}\,\beta_3\,{\rm Unemployment}\,{\rm rate}$$
(Bramber et al. Reference Brambor, Clark and Golder2006, 75). When the county unemployment rate is over 1.4 per cent, the percentage of women supervisors on county boards are likely to be positively associated with welfare spending.
Figure 2 The marginal effect of gender representation in county level on total welfare spending. Note: The dashed line shows the 90 per cent confidence interval around the solid line.
The dashed lines marking the upper and lower bounds of the 90 per cent confidence interval around the estimate are both above the zero line at 10.3 per cent. This implies that the marginal effect of the number of women supervisors becomes positively significant when the county unemployment rate exceeds 10.3 per cent. Therefore, when the economy sufficiently worsens, the number of county board positions held by women has a statistically significant positive effect on welfare spending, and the positive effect increases as the number increases. Other than the importance of the county leadership role in expanding welfare services, this finding suggests that gender representation on county boards matters in times of economic hardship. Contrary to the findings from other studies, which show that institutional factors including charter status have an impact on changes in county expenditures (Choi et al. Reference Choi, Bae, Kwon and Feiock2010; Farmer Reference Farmer2011), however, both the interaction term and the dummy variable for charter status were not statistically significant.Footnote 18
Second, another variable revealed to have a statistically significant influence on total welfare spending is the political preference of county residents. The importance of political factors in the politics of redistribution cannot be overstated in the study of expenditures and policy outcomes (Fry and Winters Reference Fry and Winters1970, 508; Fosset and Thompson Reference Fosset and Thompson2005). The effect of a political variable was not overshadowed by socio-economic variables, a finding that is contrary to some previous studies (see, e.g. Sharkansky and Hofferbert Reference Sharkansky and Hofferbert1969, 510). Interestingly, the residents’ political/ideological attitude measured by the percentage of Democratic votes in presidential elections had a strong positive effect on welfare spending in each county. This relationship clearly reveals how county residents’ political preferences affect choices on county welfare and confirms much of what we expected, namely that the Democratic Party stimulates welfare spending.
Yet, the variable representing the strength of interest groups was not significantly related to local welfare expenditures once the other variables were controlled, although the sign and the correlation coefficient were consistent with expectations derived from our theory. Some possible explanations include the fact that it is quite demanding to measure interest group strength and its impact on spending choices. Although they are assumed to be interest groups, non-profit organisations classified as 501(c)(3)s might not function as expected. Organisations in this category are permitted to lobby to an extent, but they are more committed to charitable contributions than political purposes (Boris and Steuerle Reference Boris and Steuerle2006, 70). Another explanation may have to do with government-non-profit relations. Most of the social welfare programmes after the Reagan administration were delivered directly to consumers and not through non-profits or producers of services (Grønbjerg and Salamon Reference Grønbjerg and Salamon2003, 454).
Third, consistent with prior studies, county expenditures in redistributive policy areas were determined primarily by intergovernmental revenue from the federal and state governments (Park Reference Park1996; Bahl et al. 2002; Farmer Reference Farmer2011). Although our study could not ascertain the differential effect of state and federal financial support, estimates suggest that a one-dollar increase in intergovernmental revenue leads to an increase of about 20 cents in county spending on public assistance.Footnote 19 Accordingly, the decision to shrink welfare in state government will have a direct impact on county governments.Footnote 20 Because state governments tend to spend less on welfare programmes and more on economic development during economic recessions (Witco and Newmark Reference Witco and Newmark2009, 219), local governments have to solve the dilemma of providing more services with less money. In sum, the state redistributive decisions are critical in determining the level of welfare spending in counties, regardless of the local discretion in choosing the subcategories of services.
Finally, after controlling for economic and political variables, demographic characteristics had no discernible impact on redistributive choices. Based on this finding, the percentage of females in the population, population over 65, population under 14, median income and per capita personal income do not appear to be associated with significantly higher welfare expenditures. This seems contrary to the findings of previous research (Park Reference Park1996; Farmer Reference Farmer2011), but is consistent with Peterson’s (Reference Peterson1981) argument that social service expenditures are not related to indicators of local demands in the community. This could, in part, be due to the fact that the political/ideology variable reflects much of the socio-economic status of county residents. Other plausible explanations might include the overwhelming influence of the state government in the politics of redistribution.
To explore the potential quadratic effects of gender representation, the same series of regressions, including the squared term of both the number of female county supervisors and the percentage of women in the state legislature, are presented in Table 5. In the non-linear fixed effect models, the estimates of the squared term of the number of women county supervisors turned out to be statistically significant. Although the findings are not particularly robust or strong, it suggests a potentially positive non-linear effect of gender representation in the county level on welfare spending. The inclusion of the squared variables did not change the result of the regression substantially. The additional control variables follow the same pattern.
Conclusion
This paper has examined the politics of redistribution in local government. With hard data from 58 counties over a period of ten years, this study attempts to present explanations of the redistributive choices of California counties under chronic financial stress. Among the various leadership qualities, this study includes gender in county and state leadership positions as an important determinant of local welfare spending.
The findings from the panel data analysis leads us to conclude that the social welfare services in California counties are primarily determined by the state’s leadership, financial support from other governments and the political/ideological preferences of county residents. Welfare spending by county governments is likely to be influenced by various combinations of these factors. Although other demographic variables were not significant, the results presented here provide supporting evidence of the effect county residents have on redistributive choices. Counties with more supporters of Democratic presidential candidates are likely to spend more on welfare services.
The empirical evidence on the effect of gender representation in the leadership positions in state and county government follows the general pattern predicted by the model. Gender representation in the local government and state legislature had a different effect on policy choices related to welfare. The increase in the percentage of women in state legislature leadership positions had a statistically significant positive effect on local welfare spending, while the increase in the number of women on county boards had no significant effect. Interestingly, however, the effect of gender representation at the county level may have a quadratic relationship with welfare spending in the non-linear models. In addition, the effect of gender in county leadership seems contingent upon the economic situation. An increase in the number of women on county boards has a slight but significantly positive association with an increase in local welfare expenditures when the county economy goes bad. Put differently, the effect of economic hardship on social welfare varies across California counties, and this difference can be partly explained by gender representation on county boards.
We expect this work to give us insight into the critical problem of financial resource allocation by local governments. We have attempted to understand the impact of gender on local welfare spending and conclude that gender is a crucial variable in the study of welfare politics, especially when interacting with the economic environment. The findings of the study support the relationship between women and the politics of redistribution at the local level and provide evidence for the literature on representative bureaucracy, especially as related to the link between passive and substantive representation. This study also offers implications for leadership at the county level. Given the substantial level of discretion county authorities have, exploring and examining how leadership affects policy choices at the county level is essential to understanding the dynamics of state-local relationships. This study contributes to existing explanations of local government variations on providing and delivering social welfare services. The potential impact of gender representation on other policy areas vulnerable to economic and fiscal crises deserves more attention. Considering the need for longitudinal research on this issue, this research has the potential to contribute to the scholarship on diversity and representation in the public sector, as well as bring to attention implications for American local governments.
Despite its contribution, caution should be taken in generalising the findings from this study. The politics of redistribution may be different in other contexts, such as for cities with secondary responsibilities for welfare under competitive pressure and for counties in other states with their own historical backgrounds. There might be other important factors that influence the scope of local service provision, such as political culture (Benton Reference Benton2002a, 197) and electoral competition (Barrilleaux et al. Reference Barrilleaux, Holbrook and Langer2002). An important caveat of the analysis is that it does not directly tackle endogeneity issues among the variables. Gender representation in leadership positions cannot be achieved in a short period of time, while the ways in which gender representation affect welfare spending can be complex and have a long-term impact. Although we attempt to ensure temporal precedence by lagging independent variables, the empirical models of this study could not rule out the possibility that the gender representation may itself be a result of other processes. Future research should assess this relationship.
Acknowledgement
I would like to thank three anonymous reviewers and the editors for helpful comments and suggestions. Address correspondence to the author at selotus@gmail.com. An earlier version of this article was prepared for the 2012 American Political Science Association (APSA) Annual Meeting, New Orleans, 31 August–2 September 2012.
