Legislative Quotas and the Gender Gap in Campaign Finance: The Case of the 2014 and 2018 Legislative Elections in Colombia

Abstract

Earlier empirical research on party list proportional representation systems shows that women spend less on campaigns than men, particularly when quotas are applied. An analysis of the candidate campaign expenses for the 2014 and 2018 Colombian Lower Chamber elections provides a novel test of this gender gap and its underlying causes. The research design leverages Colombia’s unique context of electoral institutions, with interdistrict variation in terms of quota rules, and the availability of detailed information on campaign spending and funding. The regression models show that the gender gap in campaign spending is limited to districts with quota rules and disappears among incumbents and candidates listed first on the ballot. As for funding, women candidates are most disadvantaged with regard to personal funds and corporate donations but attract as many individual donations as men do.

Although a large number of political systems have introduced legislative gender quotas to boost women’s representational levels, these countries seldom reach full legislative parity (e.g., Dahlerup 2006; Krook 2010; Franceschet et al. 2012). This is primarily due to the quota rules themselves, which lack effective sanctions or which have thresholds falling short of parity. Yet there is a growing awareness among political scientists that women’s representation is also hampered because women candidates lack the financial resources to get elected. Hence the emphasis on gendered political financing in the recent literature, involving measures either to create financial incentives for parties to nominate women or to subsidize women candidates directly (e.g., Muriaas et al. 2020; Ohman 2018; and specifically on Latin America, Ferreira Rubio 2009). However, while few scholars doubt that political finance is an important impediment for women, there was, until recently, relatively little empirical research on PR systems mapping the gender gap in campaign spending quantitatively, and investigating the underlying causes.

Earlier research on gender and campaign spending mainly focused on the United States, and did not find a significant gender gap. But more recent studies on PR election systems have confirmed that women spend significantly less than men. Several recent studies (e.g., Smulders et al. 2019; Buckley and Mariani 2021) have shown that the introduction of quotas coincides with an increase in the spending gap, at least in the short run. The fact that the gender gap in spending increases when quotas are in place is obviously not a problem of quotas as such. It is instead related to a gendered opportunity structure favoring men (Piscopo and Kenny 2020).

On the one hand, selectorates may put women candidates in non-winnable positions on the lists, simply to comply with the quota rules. If so, it can be considered a rational decision for women candidates not to throw money at a campaign that is lost in advance. A second possible explanation is that women candidates have less access to funding sources. Particularly in the Global South, women’s weaker economic position, due to structural labor market forces and gendered traditional family values, may limit their access to personal resources. A belief that women are not politically capable and therefore not likely to obtain a seat may also inhibit corporate donors from financing women’s campaigns and also may discourage party elites from funding their campaigns. On the other hand, it has also been shown that women politicians are more apt at collecting small amounts from a broad network of individual donors (Feo et al. 2021). Investigating how the gender gap is related to funding patterns requires an in-depth analysis of the ways campaigns are funded.

This article aims to shed more light on the link between gender and campaign finance by analyzing candidate campaign expenses for the Colombian Lower Chamber elections. Colombia can hardly be considered a pioneer with regard to women’s representation in politics. Yet it allows for a strong empirical test of the gender gap in campaign spending and its causes, for two reasons. First, legislative gender quotas are applied only in the larger districts (M > 4), which leads to between-district variation as regards quota rules. This study leverages this variation to analyze the impact of quotas on the spending gap without making a cross-temporal comparison, in which candidates running for election in different decades are necessarily pooled and compared, as in earlier research. Second, the available campaign finance data in Colombia make it possible to investigate the funding patterns of men and women candidates in detail. We collected data on individual funding sources, such as personal funds, private and corporate donations, and party transfers.

These data provide strong new evidence that quotas coincide with a larger spending gap. The regression models show that the gender gap in campaign spending is limited to districts with quota rules. The gender gap also disappears among incumbents and first-placed candidates. As concerns funding, women candidates are most disadvantaged with regard to personal funds and corporate donations, and also obtain less transfers from the party. On the other hand, they attract as many individual donations as men.

Gender and Campaign Finance: Arguments and Hypotheses

A first set of studies on the relationship between gender and campaign finance focused on majoritarian systems, particularly the United States. The consistent finding was that gender does not have a significant effect on candidates’ fundraising skills and capacities, or on their campaign spending levels (Werner 1997; Jenkins 2007; Hogan 2007; Adams and Schreiber 2011). Nevertheless, survey research among candidates indicates that women are more concerned than men about their ability to raise funds (Sanbonmatsu et al. 2009). This concern translates into a reliance on a wider variety of financial sources and setting up more extensive fundraising operations (Jenkins 2007). This may be one of the reasons why, according to a recent study, women candidates have higher spending levels compared to men (Fulton and Dhima 2021).

Research on campaign finance and gender in proportional systems is more recent and points in another direction: women candidates consistently spend less than their male competitors. This result has been reported for elections in Switzerland (Engeli and Lutz 2014), Chile (Gamboa and Morales 2021), Belgium (Wauters et al. 2010; Smulders et al. 2019), Ireland (Buckley and Mariani 2021), and, with respect to funding, Brazil (Speck and Mancuso 2014, 46; Sacchet 2018). The latter four cases also indicate that the gender gap increases or—in the case of Ireland—comes into being after the introduction of quotas, even though it is always possible that other changes between elections have also affected this gap.

Why would women candidates spend less than men, particularly when quota rules are in place? We argue that this gender gap in spending is related to the political opportunities offered to women candidates. First of all, the lower spending levels of women candidates are probably due to the fact that the often male-dominated party elites allow women to run only from non-winnable positions, with the sole purpose of complying with quota rules. If women candidates have no realistic prospect of winning a seat and sense that they are being used only as “sacrificial lambs,” it is logical that they will engage less in fundraising and run less expensive campaigns (Wylie 2020, 127). Such a decision to spend less may also be affected by an alleged tendency of women politicians to be more risk-averse than men (Murray 2021). In the same vein, women are also more critical of their own electability and tend to hold back as candidates until they consider the odds of winning a seat realistic (Fulton et al. 2006).

Parallel to the influx of new women candidates is the deselection of medium- and low-quality male candidates. This was shown to be the case in Sweden, where the introduction of a quota system with a zipper rule in local elections led to an increase of the competence of male candidates as a result of the removal of mediocre males, while the competence of women candidates remained constant (Besley et al. 2017). The strong male candidates retained by party elites will most probably also be those who showed considerable fundraising and spending capacities. As a result, the average spending levels will increase for men, which will increase the spending gap.

Research on the Belgian case (Smulders et al. 2019) supports the hypothesis that the gender spending gap is related to electability. The increase of this gap after the introduction of a parity rule was limited to candidates in unrealistic list positions (i.e., positions on the party-ranked ballot list that do not provide a realistic chance of getting elected to Parliament in the Belgian flexible-list PR electoral system). Initially, the gender gap also increased for the “realistic” candidates, but this was only temporary. Contrary to the unrealistic candidates, women realistic candidates were rapidly able to catch up financially with their male counterparts.

Apparently, if women candidates spend less, it is because they are mostly put in non-winnable positions and do not want to throw money at a seemingly “mission impossible” campaign. The important role of electability is also corroborated by the finding that the spending gap is affected by the incumbency status, taking into account that incumbents normally obtain a realistic position on the list. In Brazil (Wylie 2020, 130–33), Ireland (Buckley and Mariani 2021, 8), and Italy (Feo et al. 2021, 10–11), gendered finance inequities disappear among incumbents. In Chile, the gender gap is even reversed among incumbents, in the sense that women MPs, on average, spend more than men MPs (Gamboa and Morales 2021, 264–65).

Following this line of reasoning and building on the explanations and findings from prior research, we expect that male candidates will outspend women candidates (H1) and that the spending gap will be significantly larger in districts with quotas (H2). Furthermore, assuming that the lower expenditures of women are related to the opportunity structure and the chances of winning a seat, we also hypothesize that the gender gap will disappear among incumbents and candidates in top positions (H3).

This reasoning assumes that it is a rational decision of women candidates to spend less, given the poor chances of winning a seat. But there may be another explanation. Perhaps it is simply not an option for women to invest in their campaign. Particularly in the Global South, women candidates are often in an economically weaker position compared to men. Apart from time constraints, there are economic inequities, as women often have lower wages and fewer personal means at their disposal (e.g., Piscopo et al. 2021, 14–15). Even if they come from relatively wealthy families or have extended family networks, gendered social norms may make it difficult for women candidates to draw on these family resources, as they are expected to invest in the children and the household instead of “taking bread off the table” by funding a political campaign (Wylie 2020, 127–30). Recent research on Chile (Piscopo et al. 2021) confirms that women candidates draw fewer resources from their private funds to finance their campaign, compared to men. In line with these studies, we hypothesize that women candidates will invest fewer personal funds than men (H4).

If women have less access to personal funds, they will, arguably, be more dependent on donors for their campaign funding. But they also appear to be disadvantaged in that respect. The mostly male donors appear to be reluctant to fund the campaign of women candidates. They do not really take women candidates seriously; they doubt that women have enough ambition and stamina for a political career or stand a chance to be elected. Another prejudice is that elected women will not have sufficient time to fully devote themselves to the job. It also appears that male donors do not feel at ease when they have to negotiate campaign funding with women candidates (Casas-Zamora and Falguera 2017, 27–31). If such gendered prejudices affect the behavior of donors, they may set in motion a downward spiral. As women receive less funding, they will have to tone down their campaign, thus giving the impression of being less motivated and thereby confirming donors’ prejudices, leading to further underfunding. The expectation that women will not be able to attract sufficient private funding may even jeopardize their chances of getting selected (e.g., Josefsson 2020 on Uruguay; Wang and Muriaas 2019 on Zambia). Such a gender gap in private funding was found in Chile (Piscopo et al. 2021).

At the same time, it is important in this respect to distinguish between funding from individual donors and funding from corporate donors. It is especially with regard to corporate funding that women appear to be disadvantaged. Women have less access to large, male-dominated political and financial networks (Casas-Zamora and Falguera 2017, 27–31; Muñoz-Pogossian and Finn 2017, 181; Hinojosa and Vázquez Correa 2018, 46–47). More particularly, women are not as well connected to the mostly male-dominated business community, which is generally the main funder of campaigns (Hillman 2018, 330). For instance, in Brazil, where the elections are dominated by male candidates waging expensive and entrepreneurial campaigns (Wylie 2020), the largest gap in campaign revenue is found in corporate donations (Sacchet 2018).

On the other hand, women candidates appear to be less at a disadvantage with regard to donations from individuals. In Canada, women citizens are more inclined than men to donate to campaigns in small amounts, and they also favor women politicians (Tolley et al. 2020). That may be one of the reasons why women candidates in general rely on smaller and more differentiated sources than men, who are instead supported by a more limited number of powerful donors (Feo et al. 2021, 3). The latter authors have shown that, while male candidates in Italy draw by far most of their funding from corporations, they consistently report fewer individual donations than women candidates do. The candidates themselves attribute this to a stronger local embeddedness of women candidates (Feo et al. 2021, 9). These insights from previous research lead to the hypotheses that women candidates in Colombia will receive fewer corporate donations than men (H5), but more individual donations (H6).

The gendered disadvantage of women with regard to personal and corporate funding might be compensated by the political parties. Indeed, one of the possible gender-targeted funding schemes involves the requirement that parties use a part of their subsidies to support the campaigns of woman candidates (Ohman 2018, 12–13). But such a scheme is seldom applied, and the normal practice is that parties are left free in the allocation of funds to individual campaigns. There are various reasons to expect that party elites will financially favor male candidates. Women have less access to party networks and are therefore less involved in the decisionmaking on campaign funding. Also, the male-dominated party elites will arguably be prone to the same gendered stereotypes mentioned above (e.g., Ferreira Rubio 2015, 140).

Most studies confirm that political parties generally do not compensate gendered funding inequalities (e.g., Ohman 2018; Piscopo et al. 2021 on Chile; Murray 2021, 9–10 on the UK), or they even reinforce them, such as in Brazil (Wylie and Dos Santos 2016, 436–37). An exception is Ireland, where, since a quota was introduced in 2016, parties have given twice as much financial support to women candidates as to men (Buckley and Mariani 2021, 8). Even so, the most plausible expectation on the basis of the literature is that women candidates will receive less funding from the party than will men (H7).

The Case of Colombia

The Congress of Colombia consists of a Senate (Senado) and a Lower House or Chamber (Cámara). The Senate has 100 members elected in a single, nationwide district. The Chamber has 161 members elected in 33 districts, which correspond to the 32 departments plus the capital district of Bogotá. ¹ The average district magnitude (M) is 4.9, but the variation is substantial. There are 12 districts with only 2 seats. At the other extreme are the 3 districts containing the main urban centers: Bogotá (18 seats), Antioquia (17 seats), and Valle (13 seats). The district magnitude in the 18 remaining districts ranges from 3 to 7. The seats are allocated to the lists on the basis of the D’Hondt method, with a threshold of half of the Hare quota (or 30 percent in the districts with 2 seats) (Taylor 2009, 90–92).

A unique characteristic of the Colombian system is that since the reform of 2003, parties can choose between an open-list and a closed-list format (Duque Daza 2017, 241–45; Achury et al. 2017; Taylor and Shugart 2018). The percentage of closed lists has gradually declined, from 22.1 percent in 2006 and 24.7 percent in 2010 to 15.8 percent in 2014 and only 12.8 percent in 2018 (authors’ calculations). There are significantly more closed lists in the smallest districts (Achury et al. 2017). In the 2014 and 2018 elections 18.2 percent of the lists in the small districts (M < 5) were closed and 81.2 percent were open, compared to 10.2 percent closed and 89.9 percent open in the larger ones. In these small districts, closed lists are often only a reelection vehicle of the first-placed popular candidate (Pachón and Shugart 2010, 657).

As regards women’s representation, Colombia has a comparatively poor track record. It was one of the last Latin American countries to grant voting rights to women, in 1954. Until the 1970s, only 4 to 5 percent of the members of the Lower Chamber were women. In the 1990s this percentage was only twice as high, reaching 13 percent in 2010, before the introduction of a quota rule (Casas-Zamora and Falguera 2017, 23–24).

The quota legislation in Colombia dates back to 2000, when it was decided that women should make up at least 30 percent of all positions at the highest political and administrative levels. But the Constitutional Court abrogated the quota rules with regard to elected mandates, with the argument that this was a violation of the parties’ freedom of association. (Puyana 2017, 187–212; Pachón and Lacouture 2018, 228–44). It was only in 2011 that a quota rule of 30 percent was imposed for candidates in local, regional, and national elections. As a result, in both the 2014 and 2018 legislative elections, at least 30 percent of the candidates on both the open and closed lists had to be women.

Lists that do not comply with the quota rule are rejected. But this quota rule applies only to electoral districts with at least 5 seats; that is,14 of the 33 districts. In the other 19 districts it is still possible to nominate only male candidates. In these small districts without quotas, 49 (or 30.4 percent) of the 161 seats are elected. The quotas in the larger districts do not involve a placement mandate. In other words, it is possible for parties to put all the women candidates in low list positions (Tula 2016, 135–64). Such a placement mandate would be crucial for closed lists, but less so for the vast majority of open lists, where intralist competition between candidates is high and positions on the list are less important. Because of these shortcomings, the impact of quotas on the representation of women in Parliament has been modest. The percentage of women representatives in the Chamber increased from 12.6 percent in 2010 to 19.9 percent in 2014 and 18.1 percent in 2018 (Vargas and Batlle 2019).

The Colombian campaign finance system is generally described as mixed, in the sense that the campaigns are, in principle, funded by both public and private means (Roll and Cruz 2010, 21–52). As concerns public funding, both political parties and candidates are entitled to state subsidies, calculated on the basis of the number of obtained votes. Up to 80 percent of the available public funding can be deposited to the lists and candidates in the form of advance payments at the start of the campaign (Espinosa 2015, 21–22; Restrepo 2011, 185–224).

In practice, however, the role of public funding in the campaign is marginal. In 2014 only 4.6 percent of the average candidate’s expenses was financed with these advance payments, a percentage that dropped to a mere 1 percent in 2018 (authors’ calculation). One of the main problems is that candidates have to comply with several legal requirements before they can receive an advance payment (Transparencia por Colombia 2014, 32–37, 42–45). Therefore, it is often only at the very end of the campaign that a candidate receives the advance payment, if he or she receives it at all. This incentivizes candidates primarily to search private means to cover campaign expenses and not to bother with the advance payment. The eventual state subsidy is paid to the national parties, sometimes years after the election. To what extent these subsidies are transferred to the parties at the district level and finally to the individual candidates is not known. ²

As a result, campaigns are overwhelmingly privately funded. This is also because the legal restrictions on private campaign funding for legislative elections are limited. Not only natural persons but also corporations can donate to a campaign. The total amount of donations to a party or candidate may not exceed the legal spending limit. A single person or corporation can contribute only up to 10 percent of this spending limit. This restriction does not apply to contributions from relatives. These are legally equivalent to funding on the basis of the candidate’s personal means, which is not limited (Espinosa 2015, 21; Restrepo 2011, 219).

There is also a spending limit for parties and candidates. The maximum amount that parties may spend during the three months preceding the election is determined for each election and district by the Consejo Nacional Electoral (CNE) (Espinosa 2015, 24; Restrepo 2011, 220). In practice, however, the caps are too high to have a real impact on spending. In 2014, an average candidate spent only 9.3 percent of what he or she was allowed to, and in 2018 a mere 8.3 percent (authors’ calculations).

Lists and candidates have to comply with detailed disclosure requirements. The financial data about expenses and funding of the candidates’ campaigns are first consolidated at the list level and subsequently sent to the party within one month after the election. Within two months after the election, the party has to send all these financial reports to the CNE. Since the 2014 election, this information is publicly available on the website Cuentas Claras (Espinosa 2015, 24; Transparencia por Colombia 2014, 33–37; Lodoño 2018, 147–64).

It is a public secret in Colombia that the real campaign costs are much higher than those officially reported. Some reports even estimate that they are ten times as high, but this remains highly speculative (Casas-Zamora and Falguera 2017, 30). At the same time, the publicly available information is amazingly detailed. While the official monitoring body, the CNE, is generally considered lax (Roll and Cruz 2010, 38–41; Espinosa 2015, 34–35), the Cuentas Claras data are increasingly scrutinized by well-staffed and internationally sponsored NGOs, such as Transparencia por Colombia and Misión de Observación Electoral. As a result, it is hazardous for a candidate to declare minimal expenses while waging a very expensive campaign. Experts and candidates agree that there is a correlation between the actual size of the campaign and the reported expenses. ³ Moreover, these expenses correlate in a meaningful and predictable way with a number of background variables. It is therefore safe to assume that the declared expenses, while undoubtedly underreported, are a valid indicator of the real campaign costs.

The Colombian political finance regime also includes two modest gendered political finance measures that have no direct impact on campaign spending. Five percent of the yearly state subsidies to the parties is distributed according to the number of women in elected institutions. Also, at least 15 percent of this subsidy has to be set aside for financing think tanks, political education, and activities promoting the participation of young people, women, and ethnic minorities in the political process (Restrepo 2011, 216–17; Estrada Ruiz 2018, 105–6). A detailed analysis of the parties’ financial accounts has shown that in the period 2016–2018, a mere 3 percent of the subsidies was used for enhancing the participation of women. But on a closer look, the larger part of this 3 percent consisted of general operational expenses unrelated to gender politics (Transparencia por Colombia and ONU Mujeres 2019, 28–42).

Data and Methods

We collected campaign finance data on the 1,435 and 1,642 candidates in, respectively, the 2014 and 2018 Lower Chamber elections. These data were obtained from the website of the Registraduría del Estado Civil, which also hosts the Cuentas Claras application. ⁴

The empirical section focuses on two types of dependent variables: Candidate expenditures and Candidate funding sources. The expenditures of candidates can be expressed in absolute or in relative terms. A straightforward relative measure is the cost per registered voter, which controls for the large differences in district magnitude. The result is a measure that varies strongly between districts, with a much higher cost per voter in the smaller districts.

An alternative way of standardizing the expenses across districts is to calculate the share each candidate spends out of all the expenses incurred by the candidates in the district (Benoit and Marsh 2008; Speck and Mancuso 2014, 39–40). As this share is highly dependent on the number of candidates, we divided it by the expected share if expenses were distributed equally across all candidates in the district. Take, for instance, a district with only two candidates, the first of whom spends 70 percent and the second 30 percent. The standardized expense of the first candidate at the district level is then 70/50 = 1.4, and of the second candidate 30/50 = 0.6. In other words, this second candidate spends only 60 percent of what we would expect if expenses were distributed equally across all the candidates of the district. We can also control for differences between lists; for example, between marginal and strong lists, by standardizing at the list level. This means that we calculate the extent that a candidate spends relatively more or less than the other candidates on the same list. In addition to the per capita expenses and the absolute expenses, we used both standardized measures (at the district and at the list level) in the analysis, so as to gain confidence in the robustness of the empirical results.

Regarding the analysis for the funding gap, the Cuentas Claras data also allow us to make a distinction between various income categories of the candidates. We distinguished between the candidates’ personal funds and their next of kin, individual donations, corporate donations, transfers from the parties, and other sources. The party transfers comprise both donations from the parties’ own means and advance payments of state subsidies. ⁵ In Colombia, the total expenses almost always equal the total income of the candidates. The average candidate spends 99.7 percent of their total reported income. As a consequence, there is no need to distinguish between total income and total expenses in the analysis.

The crucial independent variables are Gender and Quota rules. For the latter variable, we worked with a dummy variable that distinguished the districts with and those without quotas. At the same time, we controlled for district size by including the logged total number of registered voters in the models. ⁶ It can be assumed that candidates will have higher expenses in more competitive elections. While the competition between candidates is normally more intense in larger districts (Carey and Shugart 1995), the competitiveness between lists also depends on the number of parties effectively competing for seats and the electoral margin with which the seats are obtained. We therefore also controlled for the effective number of electoral parties in the district (ENEP) and the safety of the last allocated seat in the district (distance). ⁷

To capture the effect of quotas on spending, it is also important to control for ballot format. As mentioned earlier, the smallest, quotaless districts have more closed lists. The lack of intralist electoral and financial competition on closed lists may also lead to different spending patterns across gender. This was taken into account by including both the ballot format and its interaction with gender in the models.

Nonfinancial variables concerning the candidates include Incumbency status and List position. A candidate was considered incumbent if he or she was a member of the Lower Chamber or the Senate during the preceding legislature. The list position of candidates is less important in Colombia than in other list PR systems. This is partly due to the layout of the ballot, which contains only a row of numbers for each open list. A voter has to memorize the number of his or her preferred candidate, and the candidate has to draw attention to it (Pachón et al. 2017). Nevertheless, the first position is especially important. This is all the more so in the small districts, where lists are sometimes only a reelection vehicle of the first-placed popular candidate.For these reasons, we also distinguished the first-placed candidates from other candidates on the list.

In 2014 and 2018, on average 32.8 percent of the candidates on the 570 Lower Chamber lists were women, while 22.,3 percent of all lists included only male candidates. Gender equality on lists improved slightly between these two elections, as the average percentage of women candidates increased from 30.1 percent to 35.1 percent. In the districts where quota rules did not apply, on average 24.7 percent of the candidates were women. This percentage increased to 40.9 percent in the districts with quotas.

The representation of women on the lists did not vary substantially between parties. For almost all major parties, which presented lists in most districts, the average percentage of women on the list hovered around 30 percent. Left-wing parties were not more gender-inclusive than right-wing ones. While left-wing parties take more women-friendly positions, this is not necessarily translated into practice, as was already evident from earlier research in Colombia (Wills Obregón and Cardozo Garcia 2010).

In line with this discrepancy, the only party with a substantially higher percentage of women on the lists was a conservative one: the Evangelical party MIRA (Movimiento Independiente de Renovación Absoluta). While this party is small, polling only 2.9 percent in 2014 and 4 percent in 2018, it is well organized, and it presented lists in 30 districts in both elections (Velasco et al. 2019). The average percentage of women on these lists was 52.2 percent. Even in the small quotaless districts, MIRA reached an average of 54.4 percent. Hence, the MIRA politicians were heavily overrepresented among the relatively few women candidates in these districts (23.5 percent out of 204 women). As a result, 5 of the 7 MIRA MPs elected in 2014 and 2018 were women.

It is paradoxical that women politicians play such a prominent role in a party that heralds traditional family values and thereby seems to embody the gendered stereotypes that impede women’s participation in politics. Yet while this focus on family values by Pentecostal parties in Colombia involves stances against abortion and LGBT rights, and sometimes against the “gender ideology” in general (Beltrán and Creely 2018), it has also led these parties to promote laws in favor of protecting and empowering women; for instance, against drug and alcohol abuse by men. In fact, the 2000 quota law was an initiative of Viviane Morales Hoyos, a senator with a strong Pentecostal background. Some observers also attribute this woman-friendly face of Pentecostal parties to the fact that the Evangelical churchgoers in Colombia are predominantly women (Beltrán and Quiroga 2017, 206–7).

While 35.1 percent of all candidates are women, this percentage drops to 24.7 percent among the first-placed candidates. Women are also underrepresented among the 6.4 percent of incumbent MPs on the lists. Of all these incumbents, only 12.2 percent are women. This percentage increased from 7.6 percent in 2014 to 17.6 percent in 2018, which is a consequence of the abovementioned increase in women MPs in 2014 due to the quota legislation.

Results

The Spending Gap

To begin with, a descriptive analysis presented in table 1 provides preliminary support for H1 about the gender gap in spending. For brevity’s sake, we will limit this preliminary analysis to the absolute expenses. The average male candidate spends $94,354 thousand (about US$24,000), which is almost twice as much as the average female candidate ($53,256 thousand, about US$13,500). ⁸ The gender gap can be measured by dividing the men’s expenses by the women’s expenses, resulting in a ratio of 1.77 overall. This gap has widened somewhat, from 1.72 in 2014 to 1.81 in 2018. ⁹ A minority of 11.6 percent of the candidates declare to have incurred no campaign expenses, but this percentage hardly differs between men (11.4 percent) and women (12 percent). ¹⁰

Table 1. Average Individual Campaign Expenses According to Gender, Quotas, Incumbency Status, and List Position (in thousands of Colombian pesos)

As table 1 shows, the gender gap is considerably smaller in the districts without quotas (ratio 1.20) than in those with quotas (ratio 2.01), which is in line with H2. Incumbents spend almost five times as much as nonincumbents. We already know that women are underrepresented among incumbents, but this explains the gender gap only partially. Among nonincumbents, the spending gap remains substantial, with a ratio of 1.52. On the other hand, among incumbents, the spending gap shrinks to a ratio of 1.16. The distinction between first-placed and other candidates is clearly less important. First-placed candidates spend almost two-and-a-half times as much as the others. But the gender gap hardly differs between the first-placed (ratio 1.56) and other candidates (ratio 1.66).

With regard to the analysis of expenditures, we ran four separate regression models, using as dependent variables the absolute expenses, the per capita expenses, the standardized spending share at the district level, and the standardized spending share at the list level. For all models, the dependent variables are semicontinuous, as 11.6 percent of the candidates declared no expenses, and thus the dependent variables have a value of zero. Therefore, we did not run simple linear regression models but instead followed the alternative approach of estimating generalized linear models with a Tweedie distribution (Parveen et al. 2016), which is a suitable approach to analyzing a zero-inflated continuous outcome. Apart from the variables already mentioned, we included fixed effects for the year of the election and the party.

Table 2 reports the results of eight generalized linear models with a Tweedie distribution. For each of the four versions of the dependent variable, we first ran a model without interaction terms. The results consistently show that while controlling for the presence of quotas, incumbency status, ballot format, size of the district (in terms of registered voters), effective number of electoral parties, and the safety of the last allocated seat in the district, women candidates spend significantly less on their election campaigns than their male counterparts. In terms of substantive interpretation, the average marginal effect of gender (women = 1) on campaign expenses in absolute terms is $–26,470,562 (roughly US$–6,800). In the model with per capita spending, men candidates spend, on average, $53.49 (roughly US$0.014) more per registered voter than women candidates. These results confirm H1, that women candidates generally produce lower campaign expenses than male candidates. Moreover, as could be expected, incumbents and first-placed candidates spend significantly more on their campaigns.

Table 2. Campaign Spending Levels of Candidates for the Colombian Lower Chamber Elections, 2014 and 2018

***p < 0.001; **p < 0.01; *p < 0.05.

Note: Table reports coefficients of generalized linear models with Tweedie distribution, standard errors in parentheses.

In order to test H2 and H3, we included the interaction terms between gender, on the one hand, and incumbency status, first list position, and the presence of gender quotas on the other hand. For each of the four versions of the dependent variable, we find that the gender gap in campaign spending exists only among candidates in districts with quotas. There are no significant spending differences between men and women candidates in districts without quotas. This is confirmed when separate models are run for the districts without and those with quotas: the gender effect is significant only in the latter models. These results corroborate H2 and earlier empirical work on the link between gender quotas and campaign spending.

Of course, the quota and nonquota districts differ in many other respects, such as the electoral system. The smallest nonquota district, with only two seats, has a quasimajoritarian system. In the largest quota district, on the other hand, with up to 18 seats, the seats are allocated in a highly proportional way. The absence of a gender gap in the nonquota district may be due to the different competitive dynamics induced by the electoral system rather than to the quota regulation per se. Even though we have controlled for district-level competitiveness by including the ENEP and distance variables in the models, this probably does not fully capture the variation in competitive dynamics between the small and large constituencies. It is therefore useful to provide a stricter empirical test of the gender gap by limiting the analysis to the districts with four seats (the largest districts without quotas) and those with five seats (the smallest districts with quotas).

We assume that the variation in competitive dynamics due to this one-seat difference will be minimal. The four districts with four seats (Cauca, Cesar, Huila, and Risaralda) had 279 candidates in 2014 and 2018; the five districts with five seats (Caldas, Córdoba, Magdalena, Nariño, and Norte de Santander) had 407 candidates. In the four quotaless districts, male candidates spent $96,589 thousand on average (about US$24,500), compared to $78,417 thousand (about US$19,900) for women candidates, yielding a ratio of 1.23. But in the five districts with quotas, this difference increases to $117,647 thousand (about US$29,800) for men compared to $46,371 thousand (about US$11,800) for women, yielding a ratio of 2.54. An additional set of generalized linear models (see table A in the appendix) limited to these nine electoral districts confirm the existence of a spending gap in districts with quotas and its absence in districts without quotas.

To test H3, we also included interaction terms between gender and incumbency status and gender and list position. These are never significant, indicating that spending levels do not differ between women and men candidates within the group of incumbents and first-placed candidates. Indeed, the gender gap in campaign spending can be detected within the group of nonincumbent and lower-ranked candidates running for election in districts with quotas, which confirms H3. When we run separate models for incumbents (See table B.1 in the appendix), the gender effect and the interaction between gender and quotas are not statistically significant. Separate models for nonincumbents (See table B.2 in the appendix) yield a slightly larger parameter for the quota and gender interaction term than in the overall models. This provides additional support for the thesis that the gender gap is limited to nonincumbents in districts with quotas. In models with only the first-placed candidates (See table B.3 in the appendix), the quota and gender interaction coefficient reaches significance in some cases, but this result is not robust.

The Funding Gap

The data collected from Cuentas Claras allow us to distinguish between various funding categories. Just over half of the average candidate’s funding (54 percent) comes from private means or next of kin. Individual donors are the second most important source and provide about a third (29.3 percent) of the funding, while a mere 4.4 percent is donated by corporations. Transfers from the party account for only 8.3 percent, confirming earlier findings (e.g., Avellaneda and Escobar-Lemmon 2012, 120–21) that parties play a limited role in the funding of individual campaigns.

For each separate category, we can analyze the absolute or the standardized values, similarly to the previous section for the overall spending. The analysis does not include the small Evangelical party MIRA, as the campaign funding of MIRA candidates deviates drastically from that of other parties. No less than 91.6 percent of the income of an average MIRA candidate originates from the party. In fact, MIRA funnels a remarkable amount of campaign funds to its candidates. A candidate receives, on average, $52,955 thousand (about US$13,400) from the party, which is about seven times higher than candidates from other Colombian parties ($8,083 thousand, about US$2,000). Because this small party has so many women candidates, not treating it as an outlier would lead to the result that women receive significantly more party transfers than male candidates.

In table 3, the expenses of men and women candidates are split up according to income category. As the bottom rows show, the overall spending gap is reflected in the various funding categories: men always receive more than women. With regard to personal fortune and party transfers, the ratio hovers around the overall expense gap ratio (now without the MIRA candidates) of 1.84. However, the gap is substantially larger for corporate donations and smaller for individual donations. Apparently, women candidates lag behind in obtaining funding from corporations but compensate this to a certain extent by narrowing the gap in private donations.

Table 3. Average Individual Campaign Funding per Income Category (in Thousands of Colombian Pesos) by Gender and Quota/Nonquota District, Without MIRA

In the constituencies with quotas (upper rows), the gender gap is clearly visible for each of the four funding categories, although there are differences. The gap is largest for corporate funding (2.67) and party transfers (2.40) but shrinks to 1.66 for private donations. In the constituencies without quotas (middle rows), the gender gap is almost nonexistent for three of the four income categories. It is even reversed for private donations; the average woman candidate obtains slightly more private donations. But corporate donations constitute an exception, as, even in the quotaless constituencies, women candidates obtain substantially less funding than men. However, as this funding source covers a mere 3.7 percent of the expenses for the average candidate in the quotaless constituencies, this gender difference does not translate to an overall gender gap with regard to expenses.

As regards the multivariate analysis of campaign income, we treated each of the four income categories as separate dependent variables in an additional set of generalized linear models. For each of the income categories, we ran models similar to those for the spending variable (see table 2). However, we only report the results for income per capita in table 4 (see tables B.1, B.2, and B.3 in the appendix for the results using the other operationalizations). First of all, the models without interactions show that women candidates are disadvantaged for three of the four funding categories: they draw less money from their personal resources and receive less funding from corporate donors and from the party. This means that H4, H5 and H7 are confirmed. A comparison of the parameters indicates that the bias is strongest for corporate donations, and weakest, though still significant, for party transfers. It can also be seen that gender does not have a significant effect on individual donations. In this respect, women are not disadvantaged. This exception is in line with H6, even though the expectation that women would obtain more individual donations than men is not met. As can be seen in Tables B.1, B.2, and B.3 in the appendix, the three alternative operationalizations of funding yield highly similar results, the only exception being the models with absolute values, in which the effect on corporate donations is not confirmed.

Table 4. Campaign Funding Levels by Candidates for the Colombian Lower Chamber Elections, 2014 And 2018 (per capita spending only)

***p < 0.001; **p < 0.01; *p < 0.05.

Note: Table reports beta coefficients of a generalized linear model with Tweedie distribution, standard errors in parentheses.

The logical final step of the analysis is to enter the interactions in the models with the funding categories. However, we are now hitting the limits of our data. Because a large number of candidates have zeros, particularly for the smaller funding categories, these complex Tweedie models with interactions either cannot be estimated with all the variables or yield unstable results. In the models presented in table 4, with the funding per voter as dependent variable, both the main effect of gender and the interaction with quotas on corporate funding and party transfers lose significance. There is a significant main effect of gender on personal resources and a significant interaction effect with quotas on private donations. But we should not make too much of these findings, because the models with the three other operationalizations yield slightly different results (see appendix tables C.1, C.2, and C.3). In order to circumvent the statistical problems with the interactions, we also ran separate analyses for the constituencies without and those with quotas (results not shown). In the latter constituencies, the results were roughly identical to the models presented in table 4. In the constituencies without quotas, by contrast, all the gender effects on funding disappear.

Conclusions

Colombia can be considered a least likely case for gender equality in politics. It has introduced quotas late and only halfheartedly. We know from comparative research that contextual factors are much more important than institutional ones for explaining gender (in)equality in politics (Schmidt 2009). The broader literature on gender and society in Colombia draws a rather bleak picture of this context, pointing to the culture of machismo and the patriarchal family structure (e.g., Pachón and Lacouture 2018, 234–35).

It would take us too far afield to elaborate on these issues. The point is that in such a context, it is probably naïve to expect miracles from institutional changes. Nevertheless, the quota rules have clearly resulted in a substantial increase of women MPs in Colombia. Yet our analysis adds to the growing evidence that campaign finance is one of the factors that slow down this effect. Quotas appear to coincide with a larger gender gap with regard to campaign spending: the newly attracted women candidates spend less than male politicians who are strong enough to be retained on the list. This gap is clearly apparent in the districts with gender quotas, but it is nonexistent in Colombian districts without quotas. We find the same result when we keep district magnitude more or less constant by comparing the districts with four seats without quotas to the districts with five seats and quotas. Research in other countries shows that this quota-related spending gap tends to disappear in the long run, but since only two elections have been held with quotas in Colombia, such a long-term effect has not yet materialized.

A first explanation for this gender gap concerns the selection of women candidates by often male-dominated party selectorates. Women politicians are not inherently inclined to invest less in their campaigns. They do so only to the extent that they are put in unrealistic positions on the list, with small chances of winning a seat. This is confirmed by our finding, in line with previous research, that the gender gap disappears among incumbents or candidates at the top of the list. In other words, women incumbents and candidates in winnable list positions spend as much as men in comparable positions.

A second possible explanation for the spending gap relates to the fundraising capacities of women compared to men. Women candidates may simply not have the resources to wage an expensive campaign and to fight their way to the electoral top. Our analysis shows that the gender gap is most visible with regard to personal funding, in the sense that women candidates dispose of significantly fewer resources drawn from their own or their family’s fortune. This may be related to women’s weaker economic position, but it is probably also caused by gendered family norms and stereotypes, which discourage women from investing their resources in political campaigns. In any case, also because these personal means are the most important funding source in Colombia, it appears that the lack of personal or family resources is an important driver of the gender gap in spending.

This disadvantage is, to a certain extent, counterbalanced by women candidates’ strong ability to collect individual donations. Indeed, for this funding source, the gender gap disappears. By contrast, the gender gap is broadest for corporate donations. These findings are perfectly in line with recent studies indicating that women politicians are particularly adept at drawing small amounts of money from a broad network of individual donors, while the powerful male-dominated corporations still favor men.

In Colombia, the large gender gap with regard to corporate donations may seem to be a minor issue, given that only few candidates report corporate donations, and these sources constitute only 4.4 percent of the candidates’ declared income. But it is common knowledge in Colombia that candidates disclose only part of their total campaign expenses and that corporate donations in particular remain below the radar (Transparencia por Colombia 2014, 10; Espinosa 2015, 34). Therefore, we suspect that the gender gap observed in the scarce reported donations is only the tip of the iceberg. There is no reason to assume that the undisclosed corporate donations are more equally distributed among men and women.

This touches on another issue regarding women and politics in Colombia, which falls beyond the scope of this research. Women candidates are often said to be electorally disadvantaged because they are less visible in the media than men. This may be related to the “corporate” gender gap found in our research. A lot of the illicit corporate campaign funding comes from commercial media concerns. These donate in kind, in the form of radio or television ads below the market price (Casas-Zamora and Falguera 2017, 21, 25–26). If this hidden funding is primarily directed to men, just like the disclosed corporate funding, the result will be that women candidates are less present in the media.

Our analysis brought additional evidence that women candidates are also disadvantaged with regard to party funding, even though this is also a minor source of their income. We already know from the literature on gendered political finance that the public funding of parties is not the most efficient way to financially empower women candidates. Even if a part of the subsidy is earmarked for promoting gender equality, it is often difficult to monitor whether the money is actually used for that purpose, as the case of Colombia also attests (Ohman 2018, 12–14; 24). Our results underscore the recommendation that public campaign subsidies should be allocated directly to women candidates, especially if quotas are in place.

In sum, the results of this research not only confirm that the introduction of quotas coincides with a gender gap in campaign spending, but also provide strong evidence that it is not the quotas themselves that lead to these inequalities. Instead, it is the behavior of parties and corporate donors and. more generally, the political opportunity structure that lead to this outcome. Therefore we want to emphasize that this article should not be read as a critique of quota rules, but rather as a plea for accompanying measures aimed at the financial empowerment of women.