THE U.S. ranks 72nd in the world for its percentage

A New Landscape: Gender and Campaign Finance in U.S. Elections Olivia Bergen NYU Abu Dhabi, Class of 2015 olivia.bergen@nyu.edu Abstract Research on Congressional races of the 1980s and 1990s has indicated that men and women raise equal campaign funds and that a gender gap in the value of those funds has been closing since the 90s. Significant changes have taken place in the campaign finance landscape since 2000, however, with campaign finance regulations loosening and spending reaching higher levels each election. This paper updates the literature with analysis of House and Senate races from 1980 to 2012, inclusive, determining that since 2000, women candidates for the Senate have not only reached parity, but also have also raised and spent greater funds on average than men candidates. For House races, over time finance variables have become weaker predictors of electoral success for women. I. Introduction THE U.S. ranks 72nd in the world for its percentage of women in the national legislature, with only 19.4 percent of seats in the 113th Congress occupied by women. (Inter-Parliamentary Union 2015). At the state level, women are not much closer to political parity, holding only 24.2 percent of statewide executive positions and 24.2 percent of state legislature seats (Center for American Women in Politics 2015a,b). Why does women s representation in U.S. political institutions remain so low? Scholars have investigated a multitude of factors relating to women s candidacies, including voter attitudes, political ambition, and media coverage, each with their own literatures. One critical dimension to candidate emergence and success is campaign finance. Raising enough funds to support communications, organizing, and other costs to persuade and mobilize voters is a greater challenge each election cycle as spending levels have continued to increase. Regulation shifts in 2002 s Bipartisan Campaign Reform Act and the Supreme Court s 2010 decision in Citizens United v. Federal Elections Commission have only opened the doors to higher giving. National races are the most expensive: the cost of winning a Congressional election in 2012 was $1.6 million for the House of Representatives and $11.4 million for the Senate. The total cost of that year s Congressional races was $3.7 billion (Center for Responsive Politics). Figures 1 and 2 show that campaign receipts and disbursements have risen for both Senate and House races over the past 30 years, with the highest historical peaks in 2010 and 2008, respectively. In those peak years, the average level of fundraising and spending was over 9 million dollars for Senate candidates and 1.2 million dollars for House candidates. Figure 1: Mean Campaign Receipts and Mean Disbursements, U.S. Senate Races 1980-2012 Unseating incumbents is especially expensive. In the 2012 election, the re-election rate for incumbents was 90 percent for House candidates and 91 percent for Senate candidates. Figure 3 shows the rising cost of beating a House incumbent over time. In 1980, a challenger had to spend an average of $471,000 to win. In 2012, it took an average $2.4 million (Center for Responsive I am deeply appreciative of the help my advisor, Rebecca Morton, and seminar professor, Adam Ramey, gave me as I completed this project. Special thanks are also due to Adam Bonica for his Database on Ideology, Money in Politics, and Elections, the foundation on which my work has rested. The Center for Responsive Politics data on U.S. campaign finance and the UCLA Institute for Digital Research and Education s helpful pages on the many features of Stata were also essential to completion of my project. 1

Politics). With most seats still occupied by men, fewer women run as incumbents and more run as challengers, faced with the task of raising large amounts of money to replace inert incumbents. Figure 4: Number of Total Female Candidates and Number of Victorious Female Candidates: U.S. Senate Races, 1980-2012. Figure 2: Mean Campaign Receipts and Mean Disbursements, U.S. House Races, 1980-2012. Figure 5: Number of Total Female Candidates and Number of Victorious Female Candidates: U.S. House Races, 1980-2012. Figure 3: Cost of Beating a House Incumbent by Election Year, 1974-2012. Source: Center for Responsive Politics Nonetheless, over time, Figures 4 and 5 show that the number of female general election candidates has risen for the Senate and the House. The number of female Senate candidates has seen a more volatile increase than has the number of House candidates, but in 2012, both chambers saw record numbers of women candidates run. That year, 19 ran for the Senate and 150 for the House. The number of women candidates achieving victory has correspondingly risen. Though spending may be rising, so is the number of women declaring candidacy for and winning Congressional races. Research on gender and campaign finance goes back to the 1970s and 80s, with some of the most recent studies conducted in the mid-2000s. It is clear, however, that the landscape of campaign finance is changing with new regulations and spending that reaches higher levels each election cycle. Do men and women raise the same campaign funds? Do campaign dollars go as far for women candidates as they do for men? Has the influx in spending narrowed the gap between men and women candidates? This paper uses a new, highly comprehensive dataset that brings together information on campaign finance for candidates in state and federal races from 1980 to 2012, inclusive. It investigates the same time periods as previous research on gender and campaign finance in the 1980s and 1990s, but also analyzes aggregate trends since 2000 to update the literature and determine whether changes in the campaign finance landscape post-2000 have had an impact on electoral 2

outcomes by gender. II. Literature Review My research intersects three campaign finance literature areas: campaign finance regulation, gender and elections finance, and value of campaign spending. The finance regulation literature describes the effects of campaign finance regulation on electoral outcomes. The gender and campaigns literature explores genderdistinctive fundraising and spending patterns. The campaign spending value literature analyzes how spending translates to votes and victory for candidates. I. Campaign Finance Regulation The campaign finance landscape in the U.S. is rapidly transforming. After the Bipartisan Campaign Reform Act (BCRA) in 2002, barriers began to lower for outside spending and electioneering communications. In the Supreme Court s 2010 decision in Citizens United v. Federal Elections Commission (FEC), the court ruled that corporations and unions have a First Amendment right to spend unlimited funds on campaign advertisements, as long as they are not formally coordinated with any candidate. In 2012, the court ruled in McCutcheon v. FEC to remove aggregate limits on individual donors giving to candidates, political parties, and political action committees (PACs), previously $123,200 (Center for Responsive Politics). Today super PACs, or independent-expenditure only committees, may raise unlimited sums from unlimited sources, whether corporations, unions, or individuals, and may overtly advocate for or against candidates. Election spending increased enormously after the 2010 court decision, with 83 super PACs registered in 2010 skyrocketing to 1,310 in 2012. Total super PAC spending was $62 million in 2010, but rose to $609 million in 2012. While outside spending is not the direct focus of this paper, it does have an indirect impact on the sums that candidates themselves must raise to preempt or respond to attacks and increased expenditures by their opponents and their supporters. The literature on the impact of campaign finance regulation is mixed. Stratmann and Aparicio-Castillo (2006) found that limits on direct contributions in state elections made lower house elections more competitive by increasing the number of candidates running, and Hamm and Hogan (2008) similarly concluded that limiting contributions reduced the probability that incumbents ran unopposed in state House races in 1994-1998, inclusive. However, Lott (2006) argued that donation regulations reduced the candidate pool by 20 percent and limited competition by increasing the probability that only one candidate would run. In a 2012 study of elections from 1968 to 2009 Raja & Schaffner found that corporate spending bans had limited effects on electoral results. Bans did not have an impact on the success of parties in winning seats, altering a partisan balance or vote share, or the likelihood of incumbent re-election. When analyzing state elections after Citizens United, Klumpp, Mialon, and Williams found that the ruling was associated with an increase in Republican election probabilities in state House races. Hypothesizing that Citizens United might shift the election balance toward business interests rather than labor interests, and that business interests favor Republican candidates over Democratic candidates, they argued that the loosening of restrictions on spending would disproportionately benefit Republicans. Their differencein-difference analysis compared states that had prior independent expenditure bans lifted with states that had never placed restrictions on such expenditures between 2000 and 2012, inclusive. Citizens United was associated with a six-percentage point increase in Republican victory probability, significant in House races, though not in Senate races. Republican incumbents became more likely to win reelection, while Democratic candidates were less likely to enter races at all (2014). Varying conclusions on the impact of regulation on candidate emergence and on electoral results means that questions remain as to how spending and its regulation affect electoral outcomes, and by extension, electoral outcomes by gender. Dittmar conducted an initial analysis of the impact of the Citizens United decision on spending through a gender lens in early 2014 (Re:Gender). She found that Democratic women still have an advantage over Republican women when it comes to PAC spending, though the men-women comparison did not yield a significant difference. Though total campaign receipts did not significantly differ between men and women, the report did not analyze the gender differences in actual electoral effects such as campaign outcomes or voting percentages as linked to finance. II. Gender and Campaign Finance Research has shown that since the early 1990s, women and men consistently raise equal campaign funds in federal and state races (Werner 1997, Burrell 2005, Hogan 2007). Where men and women candidates typically differ is the sources from which they raise campaign funds, PAC support, and perceptions of fundraising. 3

Campaign funding sources include political parties, PACs, individual contributions, and a candidate s own resources (Burrell 2005). Individual donors contribute a little over half of all campaign dollars, and while men raise a smaller number of larger donations, women tend to raise greater numbers of smaller individual contributions (Crespin & Deitz 2010). Women raise 57 percent of their total receipts from individual contributions, significantly more than men s 50 percent, and collect more from the lowest levels of giving (under 200 dollars). Organizations like EMILY s List, the WISH List, and the Susan B. Anthony List have been highly successful in supporting women candidates by bundling individual contributions. In 2008, there were 47 donor networks that gave predominantly to women candidates or had a primarily female donor base (Crespin & Deitz 2010). Campaign finance also has a distinctly partisan impact for women. Female non-incumbent candidates are often supported by women s PACs that organize early contributions to their campaigns. Such contributions provide critical seed money that help candidates develop campaign plans, prepare for attacks by opponents, and conduct research. These activities establish credibility and name recognition early on in the election and are highly important in fundraising and electoral success. However, these PACs give early money disproportionately to female Democrats. While Republican women must compete with male candidates for resources, Democratic women have their own PAC sources and do not face the same scarcity (Francia 2001). Furthermore, gender is a liberal signal to voters for women candidates, so while for Democrats, the effect enhances their attractiveness to liberal ideologue donors, for Republican women it moderates their ideology and diminishes their appeal, as donors reward extremity more than moderation (Crespin & Deitz 2010). Perception also plays a role in gender and campaign finance. Though women candidates do not have a proven disadvantage in collecting campaign receipts, candidates and voters still perceive gender inequalities in funding networks or comfort with asking for money. 56 percent of women think it is more difficult for them to raise money than it is for male candidates (Sanbonmatsu, Carroll, Walsh 2009). Women candidates use more fundraising techniques and rely on more sources of funding, suggesting greater effort to achieve the same receipts and expenditures as men (Jenkins 2007). While women candidates and especially Democratic women rely heavily on women donors, there is a significant gender gap at every level of giving. Women make up only 33.3 percent of total donors, and by dollar amount contributed represent 30.2 percent of giving (Center for Responsive Politics). Women are significantly underrepresented as mega-donors both individually and to super PACs, and the gender gap is greatest at the 95,000+-dollar level. Of the top 100 donors in 2012, only 11 were women (Re:Gender). This paper seeks to confirm longstanding conclusions that men and women raise similar campaign dollars by comparing receipts, disbursements, and other variables overall and by decade, especially post-2000. It also explores the finance elements of individual contribution totals and number of givers for men and women candidates over time. III. Campaign Spending Value Why could electoral outcomes be different because of increased levels of spending? There is a positive relationship between spending and the percentage of the vote garnered, especially for challengers (Jacobson 1985, Abramowitz 1988). The literature on aggregate effects of campaign spending is anchored in Gary Jacobson s 1978 conclusion that spending in congressional elections has a higher value for challengers than for incumbents, with whom voters are already familiar. Though men and women raise similar campaign receipts, the political literature is more mixed on whether they garner the same value from each campaign dollar raised. In a 1985 paper, Barbara Burrell proposed three effects for women candidates: women achieving the same value for each dollar spent because there are no electability differences by gender, less value per dollar because they are considered inferior candidates, or more for each dollar spent because they have some advantages over male candidates. Ultimately, she found that there was a weak correlation between gender and campaign financing, especially in the last elections of that decade. The gap between candidates appeared to be closing and was likely due to women s candidate status as challengers, not their gender. Rebekah Herrick s 1996 study analyzed the efficacy of four resources for candidates including spending, but also party strength, party identification, and previous experience. She estimated each resource s effects on male and female non-incumbent candidates in the 1988, 1990, and 1992 elections and created interaction variables by multiplying gender by each of the independent variables. Using individual candidates as the unit of analysis, she investigated the impact of the four variables on the percentage of the vote received by an individual candidate. Campaign dollars had slightly greater value for men than for women, the opposite of 4

Burrell s conclusion ten years earlier, but Herrick notes that this value difference might have been a sign of current events and likely to vary with the nature of the times. In a 1998 critique of Herrick s methodology, Joanne Green argued for district and candidate-paired analyses of elections rather than aggregate analysis. A district analysis would take into account the price of media markets, the ideological leaning of the district, whether it was urban or rural, and other factors. The candidate v. candidate analysis considered the opponent s spending, occupational background, and gender. She conducted a multivariate analysis looking at campaign expenditures, elective experience, party strength, party identification, and the percentage of the district considered urban, ultimately finding a spending disadvantage that appeared to be closing between the 1980s and 90s (Green 1998). Her methodology also looked at differential winning rates, finding that between 1982 and 1988, women who outspent opponents won 56 percent of the time, while men who outspent female opponents won 83 percent of the time. A later analysis she conducted on races between 1986 and 2000, inclusive, found that differential returns for campaign resources persisted (Green 2003). Since these studies, there has been little scholarship into comparative spending value for men and women candidates. Gender effects are worth investigating post-2000, after the large changes in the campaign finance system described previously. I will explore campaign funding s relationship to electoral outcomes and whether it differs as a predictor of success for men and women candidates. III. Hypotheses Analysis of the effect of campaign spending is challenging because a number of interacting variables contribute to election outcomes, including incumbency, previous electoral experience, whether the seat is open, the partisanship of the district, the competitiveness of the race, and other factors. My analysis first uses difference of means tests on the critical finance variables of receipts, disbursements, individual contributions, and number of givers to determine whether men and women s finance profiles are significantly different. I then construct a binary logistic regression model that incorporates the above variables to isolate the finance variables and compare their effects among male and female candidates. Do receipts and other finance variables, when controlling for incumbency, experience, and other factors, have the same predictive power for winning an election for men and women candidates? I hypothesize that the finance totals, as well as the value gained from those variables expressed in electoral outcomes, will be the same for both men and women candidates when controlling for these other factors. IV. Data and Methodology Stanford s Database on Ideology, Money in Politics, and Elections has contribution information for 51,572 candidates and 6,408 political committees as recipients in state and federal races from 1980 to 2012 (N=158,188). Candidates have a unique ID number that follows them across elections in the database. I analyze data for federal races, 1980-2012 inclusive, and split the dataset into House and Senate candidates to isolate differences in those races. I only include general election candidates, ensuring that finance variables can be more easily compared and that party identification will be more meaningful, given that most races will be between one Democrat and one Republican candidate. The database has multiple variables describing campaign finance data for candidates, including total receipts, total disbursements, individual contributions, and number of donors. I examine each of these as primary independent variables in difference of means analysis and regression analysis. As shown in Figure 6, the distributions of receipts and other finance variables are skewed to the left, with most candidates raising sums on the lower end. I therefore use the natural logarithm of each of these variables in my analyses. This is a best practice in the literature that avoids a linear association and accounts for diminishing returns from campaign spending (Jacobson 1990). All finance variables are also adjusted for inflation. Other independent variables in my regression model are included per typical practice in the campaign finance literature. I incorporate dummy variables for party (Democrat or Republican), whether the candidate is an incumbent or not, whether they are running for an open seat, and whether they are running in a midterm election year. Ideology, a measure created by Adam Bonica, is a continuous variable and a lifetime weighted average of the ideology of the people who donate to the candidate (Bonica 2014). District partisanship is a measure of the percentage of the vote that the Democratic presidential candidate received in the candidate s district in the last presidential election. Experience is a count variable measuring the number of previous state or federal elections the candidate has won. 5

Figure 6: Distributions of Total Receipts and Natural Logarithm of Receipts: U.S. House Races, 1980-2012. Male candidates All Candidates Female candidates My dependent variable is the election outcome, a binary variable with value 1 for victory and 0 for loss. Two separate binary logistic regressions, one for men, and one for women, express the likelihood of winning an election based on explanatory variables: primarily the finance variables, but also the control variables described above. Dividing the dataset by gender allows me to compare the similar or contrasting influence the finance and control variables may have on the probability of winning an election. The regressions are defined as below: Pr(Y = Win) ln Pr(Y = Lose) = (1) β (1) 1logreceipts + β(1) 2party + β(1) 3incumbency + β(1) 4openseat + β (1) 5ideology + β(1) 6districtpartisanship + β(1) 7experience + β (1) 7experience + β(1) 8midtermyear + β(1) 9shareo f spending The comparison candidate for the model had an average receipts total, was a Democratic challenger (nonincumbent, not seeking an open seat), had an average ideology score, was running in a district with an average partisanship level, had the average level of experience (previous elections won), ran in a presidential election year, and had an average share of spending in their race. 6

Table 1: Candidate and Campaign Receipts Descriptive Statistics, U.S. Senate Races 1980-2012. Variable Total Democrat Party Republican Incumbent Status Challenger Open Seat Lost Outcome Won Total (percent of total) All Candidates Female Candidates Male Candidates Mean Total Mean Total Total Receipts (percent Total Receipts (percent (SD) of female) (SD) of male) (SD) Mean Total Receipts 1062 6,104,689 148 7,323,898 914 5,907,267 (7,543,848) (9,292,234) (7,207,813) 527 7,323,898 90 7,707,585 437 5,631,667 (49.6%) (9,292,234) (60.8%) (8,922,982) (47.8%) (6,801,533) 535 5,907,267 58 6,728,523 477 6,159,757 (50.4%) (7,207,813) (39.2%) (9,887,836) (52.2%) (7,559,202) 447 7,292,136 44 10,178,891 403 6,976,957 (42.1%) (7,332,009) (29.7%) (7,039,070) (44.1%) (7,302,787) 422 4,041,105 70 4,097,351 352 4,029,920 (39.7%) (6,147,936) (47.3%) (6,725,680) (38.5%) (6,036,716) 193 7,866,572 34 10,272,095 159 7,352,183 (18.2%) (9,537,721) (23.0%) (13,549,137) (17.4%) (8,407,710) 522 4,548,680 86 4,959,155 436 4,467,715 (49.2%) (6,480,975) (58.1%) (8,570,473) (47.7%) (5,992,855) 540 7,608,830 62 10,604,026 478 7,220,332 (50.8%) (8,173,897) (41.9%) (9,322,381) (52.3%) (7,941,388) There are no district partisanship measures for the Senate candidates, so this is only incorporated in the House models. Additional models replicate the binary logistic regression executed above, but replace total campaign receipts with total disbursements, number of individual contributions, and number of givers. I. Summary Analysis V. Analysis Summary statistics for the 17 election cycles from 1980 to 2012, shown in Table 1, highlight some important characteristics of men and women candidates overall. Women have been only 13.9 percent of Senate candidates and 13.4 percent of House candidates over the past three decades. For the Senate, more incumbent candidates have been men: 44.1 percent of men to 29.7 percent of women, who more frequently run as challenger and open seat candidates. Female candidates are more often Democrats than Republicans in both chambers. While female Republican House candidates raise more money on average than female Democratic House candidates, this is reversed for female candidates in the Senate. Women of either party, running for either chamber, raise more on average than their male counterparts. Strictly compared by gender, women raise higher average campaign receipts than men: $7.3 million to $5.9 million in the Senate and $890,000 to $803,000 in the House, though with larger variances. However, their success rate in campaigns is lower than men: for Senate seats, women win 41.9 percent of the time to men s 52.3 percent, and for House seats, women win 48.2 percent to men s 58.3 percent. Are women really raising more funds, but winning less frequently? II. Difference of Mean Analysis HYPOTHESIS I: Men and women s campaign receipts should be equal. My first hypothesis examines men s and women s campaign receipts and other finance variables to determine if candidates of different genders raise equal funds. Tables 3 and 4 present the results of T-tests of means on several finance variables. In both the Senate and House data over the entire 1980-2012 time period, women s raw receipts and disbursements are greater than those of male candidates. However, when these variables are logged and normalized, the difference shrinks and is no longer statistically significant. Individual contributions and the number of givers are significant, however. Women Senate candidates raise over a million dollars more and women House candidates raise over $93,000 more from individual contributions than men on average. Many more donors give to female candidates on average than to male candidates: an average 4,932 donors give to women candidates for Senate, compared to an average 2,055 giving 7

Table 2: Candidate and Campaign Receipts Descriptive Statistics, U.S. House Races 1980-2012. Variable Total Democrat Party Republican Incumbent Status Challenger Open Seat Lost Outcome Won Total (percent of total) All Candidates Female Candidates Male Candidates Mean Total Mean Total Total Receipts (percent Total Receipts (percent (SD) of female) (SD) of male) (SD) Mean Total Receipts 12,062 814,966 1,615 890,937 10,447 803,222 (1,072,693) (1,091,643) (1,069,305) 6,244 783,862 1,027 861,016 5,217 768,673 (51.80%) (894,164) (63.60%) (840,623) (49.90%) (903,632) 5,818 848,348 588 943,197 5230 837,684 (48.20%) (1,235,140) (36.40%) (1,427,302) (50.10%) (1,211,323) 6200 1,056,115 673 1,213,027 3,907 1,037,008 (51.40%) (1,199,944) (41.70%) (1,324,473) (37.40%) (1,182,596) 4,646 441,898 739 523,204 5,527 426,519 (38.50%) (797,378) (45.80%) (732,342) (52.90%) (808,260) 1,216 1,010,816 203 1,161,814 1,013 980,557 (10.10%) (882,082) (12.60%) (901,267) (9.70%) (875,512) 5,200 500,380 837 618,922 4,363 477,639 (43.10%) (940,806) (51.80%) (1,158,989) (41.80%) (891,235) 6,861 1,053,127 778 1,183,580 6,083 1,036,443 (56.90%) (1,104,550) (48.20%) (929,882) (58.20%) -1,123,910 to men Senate candidates. In the House, it is an average 528 donors to men s 294. When the datasets are broken up into three time periods: 1980-1988, 1990-1998, and 2000-2012, the results become more nuanced. For both Senate and House candidates, women candidates raised equal or lesser funds than men candidates in the 1980s, narrowed the gap in the 1990s, and by the 2000s outperformed men on every finance measure on average. For the House, this trend is not significant for receipts and disbursements, but for the Senate in the 2000s, female candidates raised and spent over two million dollars more on average than male candidates. Of the top 50 campaigns for Senate with the highest funds raised, only eight have been women candidates. All but one of those women, Dianne Feinstein in 1994, ran post-2000. In the House, only five of the top 50 campaigns in terms of fundraising were by women candidates. The recent advantage women seem to have over men in fundraising is therefore not likely due to outliers on the high end skewing results, but a systemic change that has occurred since the new millennium. The overall advantage women candidates have in number of givers and individual contributions has mostly been driven by races in both the House and Senate since 2000: Senate women raise from an average 8,201 donors to Senate men s 3,608. Women galvanize a much larger donor base on average than do men, especially in recent years. III. Regression Analysis HYPOTHESIS II: Men and women with the same campaign receipts should have equal probabilities of winning elections. I first ran binary logistic regressions on four distinct datasets: Senate men, Senate women, House men, and House women over the entire 1980-2012 time period. The results of these regressions are presented in Tables 7-10. Predictably, the variables with the strongest effect regardless of chamber or gender were incumbency, whether the seat was open, experience level, and share of spending. Party, ideology scores, and whether the race was in a midterm year were not generally significant predictors of electoral outcomes. For all candidates, receipts and the other campaign finance variables had their own significant influence on outcomes separate from these other predictive variables. Each of these variables predict the election outcome more strongly for women than for men, however, both in the House and Senate. In the Senate, a one-unit increase in log campaign receipts means a 0.499 increase in the log odds of victory for women but only a 0.266 increase for men. A one-unit increase in log individual contributions increase the log odds of victory by 0.576 for women and by.155 for men. The difference between women and men House candidates is smaller, but consistently, finance variable increases predict victory more strongly for women than for men. 8

Predicted probability graphs in Figures 7-10 illustrate how changes in the campaign finance values impact the probability of winning an election. The House graphs for both men and women candidates have a more pronounced S-shape that indicates beyond a minimum threshold, electoral success probability rises steeply and then levels off with diminishing marginal returns of each additional dollar. For the Senate, these graphs are flatter, indicating a more linear relationship between funds and probability of winning. Especially for women Senate candidates, the threshold at which receipts begin to influence probability is much higher than for men. The graphs of individual contributions of men and women Senate candidates also show a greater predictive power of individual contributions for women than for men. I split the House men and women datasets each into three chronological sections: 1980-1988 inclusive, 1990-1998 inclusive, and 2000-2012 inclusive. The Senate data did not have sufficient women candidates in the 1980s and 1990s to run similar logistic regressions. Tables 11 and 12 present the coefficients for the split regressions. Between the three decade-periods, the predictive power of incumbency has fallen for both men and women from the 1980s to the 2000s, but has decreased more sharply for women than for men. While in the 1980s and 1990s women s victory was strongly influenced by finance and other variables, this is less so the case in the 2000s. In the 1980s, a one-unit increase in log campaign receipts increased the log odds of victory by 1.131, but in the 2000s, it was only a 0.473 increase. The impact of finance variable increases on men s likelihood of winning an election has fluctuated, but not dropped as dramatically as for women. VI. Conclusion Since 2000, there have been shifts in several patterns of men and women s campaign finance. Spending growth has accelerated each election cycle for all candidates for all seats. Meanwhile, there are more female candidates running for both the House and Senate. Only 26 women ran for Senate and 246 ran for the House in the 1980s, but from 2000-2012, 80 women ran for the Senate and 891 ran for the House. Since 2000, women candidates for Senate are raising and spending more than men Senate candidates, and female candidates for both chambers consistently raise greater sums from individual contributions and from higher numbers of givers. Success in raising funds from individuals and from a greater number of donors is consistent with research by Crespin & Deitz, who in 2010 found that women raise more funds from individuals than do men. Women s PACs and organizations like EMILY s List, with increasing clout as campaign spending has risen in recent years, have bundled individual contributions highly successfully for Democratic women s campaigns. Such bundling is a likely driver of women s receipts advantage post-2000 and the strong lead women candidates enjoy in individual contributions and number of givers. Because women are more likely to run as Democratic candidates, there are also possible spending differences based on district costs. Advertising, the prevailing expense in most congressional campaigns, is more expensive in urban media markets in New York, California, Massachusetts, and other Democratic states and contributes to higher campaign spending totals for candidates running in those states and districts. Other campaign costs such as staff salaries and real estate are similarly higher in these urban areas. Further research using the Database on Ideology, Money in Politics, and Elections could break down this paper s analysis by party to investigate whether there is a partisan discrepancy in women s fundraising success. This could support or discredit the hypothesis that it is largely Democratic women candidates behind women s overall funding advantage in Senate races post-2000. In the past, fundraising and disbursements had a sharper effect on electoral outcomes for House women candidates, but as more women candidates are entering races and spending is increasing, it is becoming a weaker predictor of electoral success for women. Incumbency, experience, and finance, once less accessible resources to the field of female candidates, are now more common advantages for women and are less salient predictors of whether they will claim election victory. The rising number of women candidates may also mean there are more dispersed receipts among them as donors and organizations who give primarily to women face more choices and spread their resources more widely. Critically, increased outside spending is flooding races and possibly muting the electoral impact of candidates own fundraising. Especially after Citizens United in 2010, spending not coordinated with candidates campaigns but nonetheless persuasive to voters has become as important as candidates own messaging. Outside spending is not directly measured in this paper s models, which limits their power in describing and the full campaign landscape. Especially for the 2012 races and beyond, it is necessary to incorporate independent expenditures in the models, adding it to the list of primary independent variables and analyzing its effects by gender. In a period of rising spending, women seem to be excelling at massive fundraising and spending ef- 9

forts. Will these trends continue, however, with women pulling ahead of men in the fundraising game? It will be necessary to continue studying both chambers to understand if the 2000s have been a fluke decade, soon to even out between genders as more diverse candidates continue to run for office, or whether women candidates will continue to outspend men. Furthermore, other differences between men and women candidates exist beyond fundraising and spending that need to be addressed, from the supply and quality of candidates to the gender gap in political ambition. Continued research in these areas and understanding how finance interacts with these other candidate emergence factors is crucial to reaching political parity. References Abramowitz, Alan I. 1988. Explaining Senate Election Outcomes. American Political Science Review 82.02, 385-403. Bonica, Adam. 2013. Database on Ideology, Money in Politics, and Elections: Public version 1.0 [Computer file]. Stanford, CA: Stanford University Libraries. Bonica, Adam. 2014. Mapping the Ideological Marketplace American Journal of Political Science 58(2), 367-387. Burrell, Barbara C. 1985. Women s and men s Campaigns for The US House of Representatives, 1972-1982 A Finance Gap American Politics Research 13.3: 251-272. Burrell, Barbara C. 2005. Campaign Financing: Women s Experience in the Modern Era. Women and Elective Office: Past, Present, and Future. 3-22. Columbus: Ohio State University Press. Carroll, Susan J. and Richard Logan Fox. 2005. Gender and Elections: Shaping the Future of American Politics Second Edition. Cambridge: Cambridge University Press. Center for American Women in Politics (a). 2015. Fast Facts: Statewide Elective Executive Women 2015. Eagleton Institute of Politics, Rutgers University. Center for American Women in Politics (b). 2015. Fast Facts: Women in State Legislatures 2015. Eagleton Institue of Politics, Rutgers University. Center for Responsive Politics. 2015. Historical Elections. Crespin, Michael H. and Deitz, Janna L. 2010. If You Can t Join Em, Beat Em: The Gender Gap in Individual Donations to Congressional Candidates. Political Research Quarterly. Berkely, USA: University of California Press. De Vries, Catherine and Hobalt, Sara. 2012. When Dimensions Collide: The Electoral Success of Issue Entrepreneurs. European Union Politics 63(3): 581-593. Dittmar, Kelly. 2014. Money in Politics with a Gender Lens Re:Gender: http://www.regender.org/moneypoliticsgenderlens. Douglas M. Spencer and Abby K. Wood. 2014. Citizens United, States Divided: An Empirical Analysis of Independent Political Spending. Indiana Law Journal 89.1:315-372. Francia, Peter L. 2001. Early Fundraising by Nonincumbent Female Congressional Candidates: The Importance of Women s PACs Women in Politics 23.1/2, 7-20. Green Joanne Connor. 1998. The Role of Gender in Open-Seat Elections for the U.S. House of Representatives: A District Level Test for a Differential Value for Campaign Resources. Women & Politics 19.2, 33-55. Green, Joanne Connor. 2003. The Times...Are They A- Changing? An Examination of the Impact of the Value of Campaign Resources for Women and Men Candidates for the U.S. House of Representatives. Women & Politics 25.4, 1-29. Hamm, Keith and Robert Hogan. 2008. Campaign Finance Laws and Candidacy Decisions in State Legislative Elections. Political Research Quarterly 61, 485-467. Herrick, Rebekah. 1996. Is there a Gender Gap in the Value of Campaign Resources? American Politics Quarterly. 24:68-80 Hogan, Robert E. 2007. The Effects of Candidate Gender on Campaign Spending in State Legislative Elections. Social Science Quarterly 88.5: 1092-1105. Inter-Parliamentary Union. 2015. Women in National Parliaments. Jacobson, Gary C. 1978. The Effects of Campaign Spending in Congressional Elections. The American Political Science Review 72.2:469-491. Jacobson, Gary C. 1985. Money and Votes Reconsidered: Congressional Elections, 1972-1982. Public Choice. 47.1: 7-62. 10

Jacobson, Gary C. 1990. The Effects of Campaign Spending in House Elections: New Evidence for Old Arguments. American Journal of Political Science. 34.02: 334-62. Web. Jenkins, Shannon. 2007. A Woman s Work is Never Done? Fund-raising Perception and Effort among Female State Legislative Candidates. Political Research Quarterly. 60.2:230-239. Klumpp, T., Mialon, H.M., & Williams, M.A. 2014. The Business of American Democracy: Citizens United, Independent Spending, and Elections. Independent Spending, and Elections. July 2014. La Raja, R.J., & Schaffner, B.F. 2012. The (Non-)Effects of Campaign Finance Spending Bans on Macro Political Outcomes: Evidence From the States. Available at SSRN 2017056. Lott, John. R., Jr. 2006. Campaign Finance Reform and Electoral Competition. Public Choice. 129 3/4: 263-300. Political Parity. 2015. Money and Women Candidates. http://www.politicalparity.org/researchinventory/money-and-women-candidates/. Sanbonmatsu, Kira, Susan J. Carroll, and Debbie Walsh. 2009. Poised to Run: Women s Pathways to the State Legislatures. Center for American Women and Politics, Eagleton Institute of Politics, Rutgers University. Stratmann, Thomas and Francisco Aparicio-Castillo. 2006. Competitions Policy for Elections: Do Campaign Contribution Limits Matter Public Choice. 127:177-206. Werner, Brian. 1997. Financing the Campaigns of Women Candidates and their Opponents: Evidence from Three States, 1982-1990. Women & Politics 19: 81-97. 11

VII. Appendix Table 3: T-Tests of Means for U.S. Senate Races, 1980-2012. Variable Female Male Difference Candidate Mean Candidate Mean (T) Total receipts 7,323,898 5,907,267 1,416,631* (1.77) Log total receipts 14.948 14.961-0.014 (-0.10) Total disbursements 7,283,126 5,959,580 1,323,547* (1.65) Log total disbursements 14.932 14.926 0.006 (0.042) Individual contributions 4,928,400 3,774,663 1,153,737** (2.11) Log individual contributions 12.348 12.235.112*** (2.671) Number of givers 4,932 2,055 2,878*** (2.771) *** p<0.01, ** p<0.05, * p<0.1 Table 4: T-Tests of Means for U.S. House Races, 1980-2012. Variable Female Male Difference Candidate Mean Candidate Mean (T) Total receipts 890,937 803,222 87,715*** (3.013) Log total receipts 12.98 12.953 0.0268 (0.673) Total disbursements 862,884 761,519 101,365*** (3.551) Log total disbursements 12.931 12.876 0.0555 (1.378) Individual contributions 522,114 428,750 93,364*** (4.312) Log individual contributions 12.348 12.235.112*** (2.671) Number of givers 528 294 233.4462*** (6.3435) *** p<0.01, ** p<0.05, * p<0.1 12

Table 5: T-Tests of Means for U.S. Senate Races by Decade. SENATE 1980-1988 1990-1998 2000-2012 Variable Female Mean Male Mean Total receipts Log total receipts Total disbursements Log total disbursements Individual contributions Difference (T) Female Mean Male Mean Difference (T) Female Mean Male Mean Difference (T) 2,803,983 4,629,702-1,825,719** 4,416,232 5,437,651-1,021,419 10,319,396 7,188,408 3,130,988** (-2.594) (-1.187) (-2.358) 14.1 14.855-0.754** 14.559 14.928-0.369 15.427 15.064 0.363* (-2.507) (-1.476) (-1.804) 2,766,635 4,581,247-1,814,612** 4,364,342 5,464,953-1,100,610 10,283,347 7,332,493 2,950,855** (-2.539) (-1.258) (-2.207) 14.06 14.823-0.762** 14.54 14.902-0.362 15.421 15.018 0.403** (-2.49) (-1.445) (-1.989) 2,058,973 3,188,066-1,129,093* 3,205,601 3,305,695-100,094 6,765,433 4,558,259 2,207,175** (-1.9924) (-0.154) (-2.464) Log individual contributions 13.596 14.397-0.801** 13.949 14.345-0.396 14.978 14.548 0.430** (-2.280) (-1.105) (-2.13) Number of givers 328 535-206 * 1,557 1,458 99 8,201 3,608 4,593** (-2.022) (-0.334) (-2.485) 13

Table 6: T-Tests of Means for U.S. House Races by Decade. SENATE 1980-1988 1990-1998 2000-2012 Variable Female Mean Male Mean Difference (T) Total receipts Log total receipts Total disbursements Log total disbursements Individual contributions Female Mean Male Mean Difference (T) Female Mean Male Mean Difference (T) 525,438 560,874-35,436 676,523 690,822-14,299 1,106,877 1,075,880 30,996 (-1.062) (-0.449) (-0.62) 12.577 12.764-0.187** 12.792 12.901-0.109 13.189 13.137 0.0523 (-2.141) (-1.534) (-0.916) 503,070 519,234-16,164 646,120 656,677-10,557 1,078,515 1,028,349 50,166 (-0.495) (-0.342) (-1.025) 12.532 12.661-0.129 12.733 12.831-0.0977 13.145 13.074 0.0712 (-1.485) (-1.370) (-1.23) 292,424 307,467-15,043 364,614 352,473 12,141 670,025 580,513 89,511** (-0.710) (-0.61) (-362) Log individual contributions 11.948 12.071-0.123 12.122 12.111 0.0107** 12.575 12.456 0.120** Number of givers (-1.355) (-0.144) (-1.988) 60 61-1 235 196 39 814 549 264*** (-0.0964) (-2.305) (-4.003) 14

Table 7:. Binary Logistic Regressions: Female Candidates, U.S. Senate Races, 1980-2012. Variables 1 2 3 4 Disburse Individual Receipts Number of Givers -ments Contributions Log receipts 0.499* (0.257) Log disbursements 0.446* (0.246) Log individual contributions 0.576** (0.263) Number of givers 4.66E-05 (5.29E-05) Republican 1.285 1.235 1.512 0.511 (1.248) (1.239) (1.276) (1.202) Incumbent 3.310*** 3.348*** 3.245*** 3.607*** (0.848) (0.846) (0.848) (0.839) Open seat 1.378** 1.398** 1.410** 1.691*** (0.678) (0.673) (0.676) (0.654) Ideology -1.288** -1.259** -1.331** -0.854 (0.6) (0.592) (0.61) (0.587) Experience 0.485*** 0.489*** 0.468*** 0.545*** (0.166) (0.166) (0.166) (0.173) Midterm year -1.725*** -1.736*** -1.752*** -1.616** (0.657) (0.656) (0.662) (0.654) Share of spending 3.341*** 3.385*** 3.518*** 3.388*** (1.291) (1.279) (1.3) (1.21) Constant -11.96*** -11.15*** -13.06*** -4.342*** (4.226) (4.054) (4.319) (1.147) Observations 148 148 148 148 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 15

Table 8: Binary Logistic Regressions: Male Candidates, U.S. Senate Races, 1980-2012. Variables 1 2 3 4 Disburse Individual Receipts Number of Givers -ments Contributions Log receipts 0.266*** (0.0996) Log disbursements 0.199** (0.0944) Log individual contributions 0.155* (0.0856) Number of givers -2.10E-05 (2.55E-05) Republican 0.239 0.259 0.293 0.393 (0.488) (0.487) (0.484) (0.482) Incumbent 2.390*** 2.417*** 2.430*** 2.516*** (0.233) (0.233) (0.233) (0.231) Open seat 0.841*** 0.873*** 0.898*** 1.003*** (0.25) (0.25) (0.249) (0.247) Ideology -0.142-0.147-0.164-0.214 (0.297) (0.296) (0.294) (0.291) Experience 0.348*** 0.354*** 0.357*** 0.390*** (0.0556) (0.0557) (0.0558) (0.0568) Midterm year 0.199 0.208 0.223 0.243 (0.193) (0.192) (0.192) (0.192) Share of spending 2.598*** 2.640*** 2.668*** 2.733*** (0.372) (0.367) (0.364) (0.354) Constant -7.305*** -6.344*** -5.635*** -3.512*** (1.489) (1.409) (1.252) -0.362 Observations 913 913 910 914 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 16

Figure 8: Predicted Probabilities: Female Candidates, U.S. Senate Races, 1980-2012. Figure 9: Predicted Probabilities: Female Candidates, U.S. Senate Races, 1980-2012. 17

Table 9: Binary Logistic Regressions: Female Candidates, U.S. House Races, 1980-2012. 1 2 3 4 Variables Disburse Receipts Individual Contributions Number of Givers -ments Log receipts 0.567*** (0.109) Log disbursements 0.508*** (0.104) Log individual contributions 0.497*** (0.0977) Number of givers 7.01E-05 (4.91E-05) Republican -0.113-0.0753-0.284-0.0572 (0.594) (0.59) (0.599) (0.568) Incumbent 2.580*** 2.614*** 2.697*** 2.866*** (0.285) (0.284) (0.284) (0.282) Open seat 1.224*** 1.252*** 1.302*** 1.589*** (0.244) (0.243) (0.242) (0.233) Ideology 0.166 0.142 0.233 0.0766 (0.312) (0.31) (0.314) (0.294) District partisanship -0.548*** -0.538*** -0.538*** -0.459*** (0.12) (0.119) (0.12) (0.111) Experience 0.310*** 0.312*** 0.316*** 0.357*** (0.0774) (0.078) (0.077) (0.0799) Midterm year -0.236-0.239-0.238-0.213 (0.2) (0.2) (0.201) (0.197) Share of spending 3.944*** 3.973*** 3.944*** 4.027*** (0.399) (0.394) (0.394) (0.352) Constant -11.50*** -10.74*** -10.24*** -4.194*** (1.508) (1.446) (1.297) -0.397 Observations 1,567 1,567 1,562 1,576 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 18

Table 10: Binary Logistic Regressions: Male Candidates, U.S. House Races, 1980-2012. 1 2 3 4 Variables Disburse Receipts Individual Contributions Number of Givers -ments Log receipts 0.510*** (0.0368) Log disbursements 0.442*** (0.035) Log individual contributions 0.472*** (0.0356) Number of givers 8.06e-05** (3.14E-05) Republican -0.255-0.23-0.207-0.0935 (0.181) (0.18) (0.193) (0.174) Incumbent 3.030*** 3.078*** 3.126*** 3.297*** (0.101) (0.1) (0.108) (0.0997) Open seat 1.510*** 1.549*** 1.609*** 1.885*** (0.0939) (0.0937) (0.102) (0.0897) Ideology 0.245** 0.230** 0.156 0.169* (0.104) (0.103) (0.113) (0.0985) District partisanship -0.100** -0.0903** -0.0247-0.023 (0.0462) (0.046) (0.0499) (0.0444) Experience 0.176*** 0.178*** 0.173*** 0.216*** (0.024) (0.0242) (0.0252) (0.0247) Midterm year -0.0935-0.0902-0.0633-0.0514 (0.0728) (0.0726) (0.0787) (0.0716) Share of spending 3.949*** 3.987*** 4.073*** 3.992*** (0.143) (0.141) (0.155) (0.129) Constant -10.49*** -9.632*** -9.800*** -4.103*** (0.499) (0.476) (0.472) -0.129 Observations 11,823 11,820 10,221 11,934 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 19

Figure 10: Predicted Probabilities: Female Candidates, U.S. House Races, 1980-2012. Figure 11: Predicted Probabilities: Male Candidates, U.S. House Races, 1980-2012. 20

Table 11: Binary Logistic Regressions for Male Candidates, U.S. House Races by Decade. Variable 1980-1988 1990-1998 2000-2012 1 2 3 4 1 2 3 4 1 2 3 4 Log receipts.618***.560***.673*** (-0.0856) (-0.0811) (-0.0687) Log disbursements 0.535***.472*** 0.607*** (-0.0812) (-.0764) (-0.0654) Log Individual Contributions.537*** 0.545*** 0.653*** (-.0755) (-0.0728) (-0.0642) Number of Givers.00425*** 0.00260*** 0.000111*** (-0.00125) (-.000404) (-4.22E-5) Republican -0.552-0.583-0.456-0.409 0.23 0.287 0.109 0.178 0.523 0.51 0.357 0.398 (-0.374) (-0.37) (-0.372) (-0.361) (-0.346) (-0.343) (-0.344) (-0.34) (-.373) (-.369) (-0.373) (-0.334) Incumbent 2.893*** 2.954*** 2.944*** 3.127*** 2.274*** 2.362*** 2.317*** 2.439*** 2.134*** 2.189*** 2.230*** 2.670*** (-0.192) (-0.191) (-0.192) (-0.19) (-0.225) (-0.223) (-0.223) (-0.224) (-0.188) (-0.187) (-0.187) (-0.18) Open Seat 1.951*** 2.002*** 2.035*** 2.265*** 1.467*** 1.521*** 1.478*** 1.677*** 1.139*** 1.176*** 1.208*** 1.578*** (-0.194) (-0.194) (-0.192) (-0.189) (-0.175) (-0.175) (-0.174) (-0.171) (-0.181) (-0.18) (-0.18) (-0.169) Ideology 0.124 0.158 0.0373 0.157 0.00682-0.0246-8.65E-05 0.0609-0.0263-0.0254 0.0372 0.0438 (-0.236) (-0.233) (-0.236) (-0.226) (-0.213) (-0.211) (-0.211) (-0.208) (-0.197) (-0.194) (-0.197) (-0.173) District Partisanship -0.0829-0.0783-0.0718-0.0472-0.0489-0.036-0.0402 0.0134 0.0451 0.0567 0.0567 0.196** (-0.0999) (-0.0996) (-0.0999) (-0.0975) (-0.0864) (-0.0859) (-0.086) (-0.0834) (-0.0864) (-0.0857) (-0.086) (-0.0807) Experience 1.634*** 1.648*** 1.698*** 1.675*** 0.369*** 0.358*** 0.374*** 0.377*** 0.189*** 0.185*** 0.199*** 0.205*** (-0.22) (-0.22) (-0.223) (-0.223) (-0.0598) (-0.0596) (-0.0593) (-0.0608) (-0.0313) (-0.0313) (-0.0312) (-0.0317) Midterm Year -0.435*** -0.434*** -0.458*** -0.334** 0.0668 0.0819 0.0795 0.052-0.0625-0.0695-0.06-0.0131 (-0.162) (-0.162) (-0.162) (-0.158) (-0.137) (-0.137) (-0.137) (-0.137) (-0.132) (-0.131) (-0.132) (-0.128) Share of spending 4.054*** 4.065*** 4.066*** 3.863*** 3.540*** 3.617*** 3.706*** 3.715*** 4.637*** 4.686*** 4.754*** 4.413*** (-0.301) (-0.297) (-0.299) (-0.272) (-0.277) (-0.273) (-0.28) (-0.261) (-0.274) (-0.272) (-0.277) (-0.228) Constant -11.84*** -10.77*** -10.55*** -4.261*** -11.02*** -9.949*** -10.49*** -4.401*** -13.46*** -12.56*** -12.84*** -4.493*** (-1.157) (-1.094) (-0.992) (-0.275) (-1.066) (-1.01) (-0.927) (-0.263) (-1.009) (-0.964) (-0.919) (-0.242) Observations 3,102 3,100 3,089 3,138 3,089 3,090 3,080 3,128 4,065 4,063 4,052 4,092 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 21