Extreme non-viable candidates and quota maneuvering in Brazilian legislative elections

This article explores the causes and consequences of extreme non-viable candidacies, also known as “laranja” (orange) candidacies in the Brazilian political lore. We first define and delineate what makes a candidate a laranja, engaging the comparative literature on sacrificial lambs and using district-level electoral results to operationalize the concept. We then advance a typology of laranjas with four ideal types that vary along dimensions of legality and intentionality. Next, we apply descriptive statistics and a hierarchical logistic regression model to explore the individual, party, and district-level characteristics of extreme non-viable candidates and assess whether and how laranjas are distinct from non-laranjas. Finally, we illustrate the gendered character of laranjas, documenting how the candidate gender quota law in Brazil has been associated with a proliferation of candidatas laranjas (women extreme non-viable candidates).


Introduction 4
Every four years, a staggering number of candidates-5,876 in 2014-run for a seat in Brazil's lower house (Chamber of Deputies), launching a highly competitive and expensive contest to occupy one of the 513 seats representing 27 statewide multimember districts. In contrast to most proportional representation (PR) elections, Brazil's open-list variant of PR results in candidate-centered, personalistic campaigns in which co-partisans compete against one another (Ames, 1995;Nicolau, 2006;Samuels, 2001 Mancuso, 2014). From 1994 to 2014, the number of candidacies for the Chamber of Deputies nearly doubled.
Amidst this competitive electoral landscape is a concurrent increase in the share of extreme non-viable candidacies-candidates receiving zero or exceedingly few votes. In past local, state, and federal elections, the majority of those candidates receiving zero votes-dubbed laranja candidates in the media and in the popular vernacular-were women (Delgado, 2017;Holanda, Monnerat, and Vassallo, 2018;Queiroga, 2018). Electoral officials have treated instances of women candidates receiving zero votes as suggestive evidence of party attempts to nominally comply with the gender quota law and thus avoid punishment by the electoral courts while preserving the status quo (Oms, 2018;Tribunal Regional Eleitoral-São Paulo, 2017).
This article contributes to the literature on electoral politics in and beyond Brazil in two ways. First, we conceptualize and operationalize extreme non-viable candidacies, bridging the academic literature on sacrificial lambs with popular accounts of "laranja candidacies". Descriptively assessing patterns across Brazil, we use district-level electoral results to measure the frequency of and rise in laranja candidacies in recent elections. The descriptive data offer rich insights into the prevalence and patterns of laranjas. Yet, in recognizing that there are important differences underlying the motivations for laranja candidacies, we then further our conceptualization with a typology of laranjas. We discuss four ideal types of laranjas, differentiating along the dimensions of legality and intentionality.
Second, we engage the comparative literature on gender quotas and gendered institutional change to contend that parties have used laranja candidacies to circumvent the country's gender quota. We apply candidate, party, and district-level data to construct a hierarchical logit regression model evaluating the multilevel conditions under which candidates are likely to run as laranjas. We find empirical support for the hypothesis that women are more likely to be laranja candidates, particularly in the wake of quota reforms emboldening enforcement. We conclude by discussing prospects for future research.

Brazilian electoral context: candidates and the gender quota law
In the wake of redemocratization, the Brazilian political system has experienced significant expansion. From 1986 to 2014, the number of voters more than doubled, and the effective number of parties at the electoral level (ENEP) grew from 3.54 to 14.60. The country's party system has become substantially more fragmented, with a record 30 parties winning a seat in the 2018 elections for the Chamber of Deputies. Meanwhile, the average number of candidates per seat across the country increased from 7 to 11 candidates, a 57% net increase since 1986. Brazil's fragmented party system and hypercompetitive electoral arena, fueled by its high-magnitude open-list proportional representation (OLPR) electoral system, yields intraparty competition (Ames, 2002;Carey and Shugart, 1995;Nicolau, 2006;Samuels, 2008) and extremely expensive candidatecentered campaigns (Lemos, Marcelino and Pederiva, 2010;Samuels, 2001).
A provision of Brazil's Electoral Law that allows parties to advance excess candidacies, or candidacies in a district that exceed the number of seats contested in proportional elections (4.737/1965, 6.990/1982, 7.454/1985, 9.100/1995, and 9.504/1997) (Lamounier and Amorim Neto, 2005;Wylie and Santos, 2016), contributes to the crowded electoral landscape. The current iteration of the Electoral Law (13.165/2015) allows each party or coalition to advance total candidacies equal to 150% of available seats in proportional elections. For elections to the Chamber of Deputies and state Legislative Assemblies, that allowance is increased for smaller states (less than 12 seats in the Chamber of Deputies), where each party or coalition is permitted total candidacies equal to 200% of the available seats (13.165/2015).
Before a 2009 mini-reform (12.034/2009), the gender quota target-initially at least 20% candidates "of each sex" and extended to 30% in 1997-applied to candidacies permitted rather than advanced. With the generous allowance of candidacies, parties need not use all of their permitted candidacies to be competitive, instead "reserving" slots for women without actually advancing women candidates (Araújo, 1999(Araújo, , 2003(Araújo, , 2010Wylie, 2018;Wylie and Santos, 2016;Miguel, 2008). Indeed, as late as 2010, 44.2% of the 607 state parties contesting the Chamber of Deputies elections did not advance a single woman candidate, with just over one-quarter of state parties meeting the quota target. Brazil's gender quota law is thus widely considered a failure (Araújo, 2010(Araújo, , 2017Miguel, 2008;Sacchet, 2012;Wylie and Santos, 2016;Wylie, 2018). Research has shown that women can succeed in the electoral arena in Brazil provided they have the political capital to do so (Araújo, 2010;Araújo and Borges, 2013;Bolognesi, Perissinotto, and Codato, 2016;Codato, Costa and Massimo, 2014;Miguel, 2003;Miguel, Marques, and Machado, 2015;Pinheiro, 2007, Sacchet andSpeck, 2012). Nevertheless, most parties lack the will and capacity to recruit and support viable women candidates (Wylie, 2018).
Instead, parties persist in creating men-only lists, with women continuing to be underrepresented as candidates and elected deputies, and overrepresented as laranja candidates, as documented below.
We contend that the rise in laranja candidates is a consequence of Brazil's gendered political institutions and a systemic practice by party leaders used to nominally comply with the gender quota requirement while maintaining the status quo. The electoral context in Brazil, including the implementation of and reforms to the gender quota law, have incentivized the use of extreme non-viable candidates. As we document below, the majority of laranjas are women, which we argue constitutes an observable implication of party resistance to the gender quota.

Sacrificial lambs and laranjas: Brazil in context
Within Brazil's formidably competitive electoral environment exists the intriguing phenomenon of laranja candidacies, an umbrella term encompassing several variants of extreme non-viable candidates, including sacrificial lambs. The term sacrificial lamb has been used widely in the literature to describe candidates in precarious or nearly impossible scenarios. Thomas and Bodet (2013, p. 154) define sacrificial lambs as candidates who run as "mere standard bearers in riding where the party does not expect to win". The term is especially used when discussing majoritarian elections, where parties may nominate a candidate to contest a race it does not expect to win due to a strong incumbent and/or limited party support (Canon, 1993;Kaplan, Park and Ridout, 2006;Roscoe et al., 2006;Thomas and Bodet, 2013;Wilcox, 1987). The term has also been widely used in the gender and politics literature to better understand (the absence of) party support for women politicians, with mixed results (Davidson-Schmich, 2010;Dolan, 2006;Hennings and Urbatsch, 2014;Ondercin and Welch, 2009;Studlar and Matland, 1996;Thomas and Bodet, 2013;Welch et al., 1985).
Understanding the what and why of sacrificial lambs is important for the study of candidate strategy, party politics, and electoral behavior. One major problem in conceptualizing and measuring a sacrificial lamb relates to the necessity of combining both a priori and post-hoc tests to operationalize the concept. Nevertheless, we argue that by conceptualizing and measuring sacrificial lambs (and laranja candidacies), even if this conceptual exercise includes post-election empirics, we can help scholars better understand other phenomena related to the practice, including the role of parties in "creating" sacrificial lambs.
When thinking about sacrificial lambs in the Brazilian context, it is important to acknowledge the cultural and institutional peculiarities of the country. Such differences influence both the definition of a sacrificial lamb and the institutional context in which politicians operate. The term sacrificial lamb is not used by Brazilians to describe candidates who have a near impossible chance of winning. Instead, the term laranja (literally meaning "orange") is widely used in the popular vernacular to describe extreme non-viable candidates, including those who may not even mount a campaign (Perissé, 2010). Recent media coverage of laranjas has focused on candidates receiving zero votes (Bertho, 2018;Delgado, 2017;Holanda, Monnerat and Vassallo, 2018;Queiroga, 2018).
Much like the sacrificial lamb concept, the laranja candidacies concept remains undertheorized and conceptually muddied. This article examines the term laranja and the extent and purpose of laranjas within Brazil's electoral context. We aim to contribute to a broader discussion of sacrificial lamb candidates in comparative perspective, priming future research to explain and explore in greater detail the similarities and differences between and among extreme non-viable candidates both within and across cases.
Portuguese dictionaries define laranja, beyond its more common definition of orange, as a person who is naïve and meek, and as a person used as an intermediary in fraudulent and other suspicious businesses. While we were unable to locate additional etymological information regarding the use of the word from scholarly sources, popular media reinforce these definitions of the term. The use of laranja to identify a straw-man, or person used as an intermediary in fraudulent business is the most common use of the term (Bielschowsky, 2009;Moreno, 2009;Perissé, 2010). Today, the term is used in a variety of contexts, the three most common being laranja as an individual involved in fraudulent activities, empresa laranja (front business) or dummy companies that exist only in name and are fronts for money laundering and corruption schemes (D'Ambrosio, 2005;Rolli and Fernandes, 2007), and laranja candidacies.
Following the term's common usage within Brazil, we conceptualize laranja candidacies as extreme non-viable candidates. The popular term most likely originated in connection with the use of the word laranja to reference fronts for fraudulent activities in Brazil and to describe at least three types of candidates: candidates who register but do not run a campaign-also called phantom candidates in the media (Caram, 2018;Douglas and Iglesias, 2018;Fábio, 2016)-those who actively run with no chances of winning (sacrificial lambs), and the "front-person", placeholder, or attack dog for another influential politician. One of the more recent infamous cases of a laranja as a stand-in or front-person for an influential politician happened in 2010, when Weslian Roriz, wife of former Federal District governor Joaquim Roriz, ran for that same office. Barred from running for reelection one week before the campaign started because of the Lei Ficha Limpa (Clean Record Law, a law that made candidates who had been impeached, who resigned to avoid impeachment, or who were convicted of corruption ineligible of running for office), Federal District Governor Joaquim Roriz and his party nominated his wife in his place (Doin et al., 2012;Maltchik, Brígido and Weber, 2010).

Laranjas and quota maneuvering
The literatures on gender quotas and gendered institutional change speak to the prevalence of quota maneuvering and backlash by resistant parties (Gatto, 2016;Krook, 2009Krook, , 2015Mackay, 2014;Waylen, 2014Waylen, , 2017. In general, reforms that stand to redistribute power will provoke resistance, and once established, will face subversion attempts (Gatto, 2016;Helmke andLevitsky 2004, 2006;Mackay, 2014;Mahoney and Thelen, 2009;Waylen, 2014Waylen, , 2017. Gender equity reforms such as gender quotas are especially "vulnerable to regress": new institutions are nested within existing (gendered) institutions, and mechanisms for attenuating the potential displacement of men abound (Mackay, 2014). We argue that laranjas constitute one such mechanism.
Since the establishment of the gender quota in 1995, but especially since the 2009 mini-reform, parties have been required to field at least 30 percent women candidates in legislative elections (Araújo, 2010;Rangel, 2009). Parties that lack a continuous project to promote women's participation often scramble at the final hour to find sufficient women candidates; laranjas can help solve the party's immediate needs for women candidates without diverting party resources (Wylie and Santos, 2016). In this last-ditch effort, some party officials have apparently registered unwitting candidates without their consent. In Minas Gerais, party officials allegedly used women's Facebook profile photos to register their candidacies; the regional electoral official (Procuradoria Regional Eleitoral) has reportedly investigated the case as identity fraud, punishable by up to five years in prison and a fine (Affonso, 2014; Barba, 2014). More commonly, laranjas constitute violations of the spirit rather than the letter of the electoral law. A candidate for municipal council in 2016 that won zero votes confessed her laranja status, "I only gave my name to fill the quota" (quoted in Bertho, 2018).
It is important to acknowledge that while some non-viable candidates run unwittingly as laranjas, others do so intentionally. The practice also entails varying degrees of legality; although all cases certainly violate the spirit of the electoral law, some of the more egregious instances of laranja candidacies involve electoral fraud. Those varying levels of candidate complicity in being a laranja candidate and approaches to deploying laranjas motivate our typology of laranjas (see Table 1). Next, we elucidate four ideal types of extreme non-viable candidates that vary along the dimensions of legality and intentionality: laranjas on leave, non-consensual laranjas, naïve laranjas, and strategic laranjas. First, we discuss two subtypes of laranjas that exist outside the letter of the law.  (Maciel, 2017).
In contrast to laranjas on leave, the second type of illegal laranja lacks candidate intentionality. Non-consensual laranjas are typically unwitting candidates that have their candidacies registered by party officials without their consent. As discussed above, some women's names have allegedly been used without their consent by parties in need of women candidates to bring their candidacy list into compliance with the quota. A Regional Electoral Prosecutor in Minas Gerais explained, "The fraud is evident; some women don't even know about their candidacy. We have seen absurd cases of women registered with photos taken from Facebook". Parties commit such fraud "out of desperation" due to their insufficient number of women candidates (Affonso, 2014). Non-consensual laranjas can help meet the party's immediate needs for women candidates with minimal cost to the party (Wylie and Santos, 2016). Unaware of their candidacy, non-consensual laranjas do not mount a campaign.
Accounts of non-consensual laranjas have proliferated especially since 2010 (Scalzer, 2016), and MPs across the country have paid special attention to the candidacy of women receiving little to no votes. Both types of illegal laranjas described above are clear targets for investigation after an election. The 2016 Manual of the Regional Electoral Prosecutor states that after each election, prosecutors must "pay attention to the campaign contributions reports for the inspection of 'laranja' candidacies of public servants and women" (Ministério Público Federal, 2016). A reporter for The Intercept and Revista AzMina contacted a woman candidate to ask about her candidacy and discovered not only that the woman was unaware of her candidacy, but also that she had explicitly rejected her party's invitation to run (Bertho, 2018).
Moving beyond laranjas whose candidacies are illegal, we discuss two types of legal (but perhaps ethically questionable) extreme non-viable candidacies. We call the nonintentional but legal candidate a naïve laranja. Naïve laranjas are typically women party members who are asked by party officials to put their name on the list and actually run a campaign in earnest, only to receive little to no party support (sacrificial lambs). Parties often use naïve laranjas to gesture toward the quota target without constituting an electoral threat to the established (usually men) candidates on the list. That strategic advancement of laranjas corresponds with the deployment of sacrificial lambs common in other political systems, including majoritarian systems (Thomas and Bodet, 2013).
The last type of laranja in our typology is called the strategic laranja. These are candidates whose extreme non-viable candidacies represent a deliberate strategy on the part of the candidates. Popular accounts of laranjas suggest that they are at times complicit in the extreme non-viability of their campaign, driven by a desire to do what the party asks of them (Bertho, 2018). Among parties that have neglected to promote women's participation, last-minute appeals to rank-and-file women members to put their names on the ballot to "help the party" meet the quota target are common. In an interview with a newspaper, O Povo, a municipal party president openly stated, "female candidates willingly lend us their names so the party can complete its ticket" (quoted in Barba, 2014). As explained by the Regional Electoral Prosecutor in Santa Catarina, "She knows, and herself signs the documents, but the end goal is not to actually run a campaign or win votes. The end goal is to comply formally with the gender quota for proportional tickets" (Scalzer, 2016).
Strategic laranjas typically register their candidacy but run a bare bones campaign without actively campaigning. By showing loyalty and/or increasing their own electoral viability in subsequent contests, strategic laranjas earn the promise of party support for their own genuine candidacy in subsequent elections, representing a strategic decision on the part of both the party and the candidate. They can help the party by assisting its (often last-minute efforts) to meet the gender quota target while receiving a small number of votes that pool to the party/coalition list and/or actively campaigning for the party's priority candidate(s). While strategic laranjas often do not even campaign on their own behalf, in contrast to non-consensual laranjas, they do consent to their candidacy.
We contend that the proliferation of laranjas, across subtypes, is in part a byproduct of party resistance to the gender quota. Next, we offer empirical evidence of the prevalence of and temporal and inter-party variation in laranjas in elections for the Brazilian Chamber of Deputies.

Operationalizing laranjas
The data we use were collected from the Brazilian national electoral authority, the Supreme Electoral Court (TSE), and are available online through the institution's repository. The dataset contains detailed individual-level information about the candidates' profile including their name, sex, racial identity (as of 2014), age, occupation, education level, party label, number of votes, and the electoral outcome (elected, not elected, or elected by the list electoral quotient). We also compiled reports on candidate spending on their campaigns and their share of campaign revenue from different sources (personal, individuals, parties, and corporations). We supplement those data with original and secondary data on party and district-level characteristics. As Table 2 shows, the pool of Brazilian candidates is stable. Women run for office at far lower rates than men do, although the gender quota provision has led to increases in the number of women candidates over the period analyzed. We also see that highly educated individuals dominate the candidate stock despite the growing participation of aspirants with just a high school education. As average age suggests, the system favors candidates with established professional careers and past political experience (Codato, Costa and Massimo, 2014 Based on that threshold, we show in Table 3 the gendered prevalence of estimated laranjas over time.  We also conducted two tests using analysis of variance (Anova) to assess intergroup differences 8 . We evaluate whether men and women candidates differ in their average vote share and in their rates of laranjas. Overall, we find that: 1) a male candidate  Source: Elaborated by the authors and based on TSE data (2016). 8 We thank an anonymous reviewer for this suggestion.  Source: Elaborated by the authors and based on TSE data (2016). Notes: Two-tailed tests of the differences (between women and men laranjas, and between laranjas and non-laranjas) are statistically significant at the following levels: * p < .05, ** p < .01, *** p < .001. Data on campaign contributions from 1998-2014.
The gendered differences displayed here among laranjas reflect discrepancies common to all candidacies, with women candidates being significantly less likely to be married and college-educated than men candidates (although the education gap disappears among elected deputies). One interesting gendered distinction that is unique to laranjas is in the average party magnitude-the party seat share, or number of candidates elected per party and per district (Matland, 1993) 9 , which is significantly higher for women laranjas than for men laranjas 10 . While the magnitude of the difference is small, it offers tentative evidence that such extreme non-viable candidacies are not solely the domain of the country's smaller, less competitive parties, but rather, represent an element of parties' electoral strategy. In Table 6, we see that indeed, Brazil's larger parties are no stranger to the phenomenon of laranjas 11 . Table 6 shows, by party (major parties) and by ideology (all parties), for 1998-2014, the percent of all candidates that classify as laranjas, the ratio of women laranjas to men laranjas (a number above 1 means there are more women laranjas than men laranjas) and then for 2014, the percent of women candidates that classify as laranjas, and the percent women elected. We see from Table 6 that non-leftist parties have a greater proportion of laranjas than do leftist parties 12 . Among all parties, with a single exception (PCO), a substantially greater proportion of parties' women candidates are 9 Unlike district magnitude (each district's number of seats contested in a race), party magnitude captures party fragmentation. A district can have a large magnitude, with most parties still only electing 1-2 deputies, particularly in the OLPR context (Matland, 1993). 10 Although not displayed here, for non-laranjas and for elected deputies, the party magnitude of women candidates' parties is significantly lower than that of men candidates. We evaluate the relationship further in the multivariate model below. 11 Table A3 in the Appendix includes the percentage of total candidates and women candidates that classify as laranjas, as well as the ratio of women laranjas to men laranjas, for all parties. 12 The left/non-left differences are statistically significant at the p<0.001 level.
laranjas relative to their laranjas overall. Next, we examine whether these descriptive patterns of laranjas prevalence hold in multivariate analyses.

Explaining laranjas
Although there exist other strategic motivations for advancing laranjas, we build on the literatures discussed above to contend that parties' use of extreme non-viable candidacies serves to comply formally with quota provisions while preserving the status quo. We evaluate our argument by conducting a multivariate analysis that explores the candidate, party, and district characteristics associated with laranja status. The main contribution of the multivariate analysis is to evaluate whether the gendered patterns in laranjas suggested by the descriptive analyses hold after controlling for conventional indicators of candidate quality and district competitiveness. Those controls help to mitigate the limitations imposed by an ex-post measure of laranja status. We account for differences in educational attainment 13 , occupational background 14 , and incumbency 15 , factors widely associated with candidate quality (Bohn, 2007;Carson, Engstrom and Roberts, 2007;Samuels, 2001). We also take into consideration variation across state elections, controlling for the competitiveness of the election through effective number of candidates 16 . To assess our argument that laranjas represent an element of quota maneuvering, we hypothesize that woman will exercise a positive effect on laranja status even once such controls are introduced. Further, we hypothesize that heightened gendered effects will manifest in the years of quota introduction (1998) and reform (2010,2014). 13 Educational attainment is coded ordinally, with 1 representing less than high school education, 2 a high school degree, and 3 a college degree (or higher). 14 Occupation is coded dichotomously, with candidates from a feeder occupation (administrator, lawyer, doctor, politician, public servant/bureaucrat, educator, businessperson) coded as 1 and all others 0. 15 Incumbency is coded dichotomously, with candidates seeking immediate reelection coded as 1 and all others as 0. 16 The effective number of candidates speaks to district competitiveness while accounting for the fragmentation of the vote; it is equal to 1 divided by the sum of the squared vote share of all candidates running per state election (Taagepera and Shugart, 1989). This variable is highly correlated with district magnitude (ρ=0.88), which ranges from 8 to 70. Since the former offers more information, we include it in the model instead of district magnitude.
H1: Woman will exercise a positive effect on the probability of laranja status.
H2: Woman*1998, woman*2010, and woman*2014 will each exercise a positive effect on the probability of laranja status.
We hypothesize that leftist parties will be less likely to advance women laranja candidacies than non-left parties. Historically, leftist parties have been more open to women and the positive discrimination measures that would facilitate their involvement in formal politics (Kittilson, 2006). Indeed, Brazil's leftist parties have significantly higher proportions of women in their national and state-level leadership structures than do nonleft parties (Wylie, 2018;Wylie and Santos, 2016). Among the 28 parties that won seats to the Chamber of Deputies in 2014, the average proportion of women in parties' national executive committees that year was 17.6%; for leftist parties it was 29.2%, more than twice that of non-leftist parties (13.8%). When parties afford women space at the decisionmaking table, the former are more likely to incorporate a gendered frame of reference into party decisions. Women leaders will be in a position to (but will not necessarily) lobby other party leaders on behalf of women and work to recruit and support viable women candidates rather than laranjas (Kittilson, 2006;Wylie, 2018). We alternatively employ a scaled and dichotomous coding of ideology, using the latter when results are similar to facilitate interpretation 17 .
H3: Left and left*woman will each exercise a negative effect on the probability of laranja status.
Were the laranjas phenomenon solely an implication of the plethora of candidates and non-competitive parties involved in Brazilian elections, we would expect its use to be confined to (or at least more common among) smaller parties. This is due to the seeming irrationality of running laranja candidates in Brazil's OLPR elections, where candidate votes pool to the party and the Electoral Law allows excess candidacies. We therefore might expect electorally competitive parties to be less likely to advance laranjas. Nevertheless, we have argued that most of Brazil's major parties remain resistant to earnest implementation of the gender quota and thus use laranjas to formally comply with the quota provisions without disrupting the status quo. Accordingly, we hypothesize that the negative effect of party competitiveness on laranjas will only hold for men candidates. We operationalize party competitiveness with party magnitude, or the number of seats won by a party in a state election (Matland, 1993), and interact the variable with woman to assess that expectation 18 . 17 Zucco (2009, 2011) and Power and Rodrigues-Silveira (2019) draw on surveys of parliamentarians to calculate scaled estimates of party ideology; we use those as well as Mainwaring's (1999) dichotomous classification of party ideology (left/non-left), updated by referencing party websites and statutes. The parties coded as left are as follows: for 1994-2014, PCdoB, PCB, PCO, PDT, PPS, PSB, PSOL, PSTU, PT, PV; the PSDB was also classified as left in 1994. 18 Unlike district magnitude (each district's number of seats contested in a race), party magnitude captures party fragmentation. A district can have a large magnitude, with most parties still only electing 1-2 deputies, particularly in the OLPR context (Matland, 1993).
H4: Party magnitude will exercise a negative effect on the probability of laranja status. Reflective of the anticipated gendered effect, Party magnitude*woman will exercise a positive effect on laranja status. Table 7 displays odds ratios from a hierarchical logistic model of laranja status (pr (laranja=1)) in the 1994-2014 elections for the Chamber of Deputies. Statistically significant odds ratios above one indicate that the variable makes the probability of laranja status greater; those below one indicate that the variable makes laranja status less likely.
We confirm a gendered pattern in extreme non-viable candidacies; women were 3.4 times more likely than men candidates to classify as laranjas. As demonstrated in Table 7 Table 7. Looking to party characteristics, as predicted in H3, leftist parties were significantly less likely than non-leftist parties to advance laranjas. And as expected, leftist parties appear less reliant on women laranjas than non-leftist parties. Interactive effects demonstrate that while leftist women candidates were no more likely than men leftist candidates to classify as laranjas, among non-leftist candidates, women were three times more likely than men candidates to classify as laranjas. While there is party variation in the use of extreme non-viable candidates overall and among women, leftist parties are not exempt. Among men, the predicted probabilities of leftist and non-leftist candidates classifying as laranjas (holding all else constant) are 0.11 and 0.14, respectively. For women candidates, the predicted probability of classifying as a laranja is 0.26 for leftists and 0.32 for non-leftists. The findings also yield support for our expectations about party magnitude articulated in H4. While candidates running with more competitive parties were less likely to classify as laranjas, the positive significant interaction of party magnitude with woman demonstrates that the effect of party magnitude is gendered; for women, the negative effect of party competitiveness on their likelihood of laranja status was diminished and more varied than the same effect among men.
Together, the descriptive and multivariate analyses demonstrate that the phenomenon of laranjas is a salient feature of Brazilian elections with a remarkably gendered character. As expected by the literature on gender quotas and gendered institutional change, the use of laranjas has intensified in the wake of reforms to the gender quota that incentivize its enforcement. We also find that the prevalence and gender dynamics of laranjas varies by party, with leftist parties being less likely to advance laranjas, and non-left women candidates significantly more likely than non-left men candidates to classify as laranjas. Future research should further examine variation across parties and contexts to shed light on why some parties have proven to be better able to adapt than others to legal and societal changes demanding greater inclusion of women; research from Funk, Hinojosa and Piscopo (2017) suggests that the exogenous "decision environment" may be more salient than party ideology. Evidence of the relative electoral costs and benefits of such adaptation could help to convince intransigent party elites of the electoral utility of cultivating viable women candidacies in lieu of widespread offerings of sacrificial lambs.

Conclusions
The advancement of laranja candidacies is evident and follows a gendered pattern.
As this article has shown, laranjas constitute an important element of party strategy, exploited to comply formally with legal provisions such as the gender quota and the Clean Record Law without disturbing the status quo, in a typical instance of layered institutional change (Mahoney and Thelen, 2009;Waylen, 2014). This article lends additional weight to studies by Araújo (1999Araújo ( , 2009Araújo ( , 2010, Gatto (2016), Wylie (2018), and Wylie and Santos (2016), who point to the designed dilution of Brazil's quota by men politicians keen to limit its potential for change. In the wake of attempts to close the loopholes that had undermined quota effectiveness, the frequency of laranjas increased, spiking precisely at those moments of quota reform (2010,2014).
In line with the sacrificial lamb metaphor, naïve laranjas are not privy to their party's strategic decision to treat their candidacy as one in name only. Yet strategic laranjas willingly lend their names to the party without campaigning, typically with an eye toward garnering support in subsequent elections or landing an administrative position. Future studies should apply the typology and operationalization introduced here to consider those variations motivating laranja candidacies, which could help electoral officials as they work to curtail the prevalence of laranjas. We encourage the broader literature on sacrificial lambs to also consider variation in intent; the sacrificial lambs terminology conjures a naïve novice being swept along by the party, but in many instances the lamb may be in on the sacrifice, acting strategically for the party's interest and/or to accrue their own individual political capital.
The recently established Electoral Fund is jeopardized by laranjas; scarce state resources cannot be squandered on extreme non-viable candidates. Greater enforcement and accountability over the laranjas phenomenon will not only improve the gender quota but will also be essential if public financing of campaigns is to be a viable possibility for enhancing the representativeness of Brazil's formal political sphere.   Table A2 further assesses whether the difference in the average of laranja candidates by year is statistically significant. A significant F value of 137.42 tells us that the means are not all equal (or at least one election cycle differs from zero). We use Bonferroni, Scheffe, and Sidak multiple comparison tests, which examine the differences between each pair of means reported in Table A2 (Bonferroni tests are displayed). The difference between the average of laranja candidacies in 1994 and 1998 is 3.99, and all three corrections show this difference as significant at the .05 level or better. The same conclusion holds for the subsequent elections when compared with the 1994 elections, although the size of the difference in the averages increases up to 16.6 considering the 2014 elections. However, the difference in the averages is not quite significant at the .05 level between the 1998 vs. 2002, and 1998 vs. 2006 elections. The same conclusion holds between the 2002 vs. 2006 elections. What is most striking is that the 2010 and the subsequent 2014 elections are significantly different from all previous election cycles regarding the difference in the averages of laranja candidacies.