WOMEN'S PACS AND EMILY'S LIST

Sanbonmatsu suggests that “[the] relationships among funds, qualifications, and elections merit further investigation” (Reference Sanbonmatsu2006b, 200). An endorsement by EList is certainly about funds, but it is also something of a proxy for qualifications, since the organization determines who it will endorse based on “viability”—the candidate's experience, her existing network of supporters, access to money, and an existing campaign organization (Day and Hadley Reference Day and Hadley2005, 25). In this article, we attempt to identify the relationship between endorsement by EList and electoral success using propensity score matching, which is explained in more detail in the following sections. The present puzzle in the literature regarding EList's impact is whether the endorsement is essential to pro-choice, Democratic women's electoral success, or whether EList merely picks the winners. It may also be that EList's impact was significant in the 1980s and early 1990s but is now waning in importance, given that the Democratic party is now less likely to overlook viable women candidates than it was in the past. To date, none of these notions have been backed by empirical investigation. As an introduction to why EList would be so essential in the first place, it is important to briefly provide background information on women's PACs.

Women's PACs are unique in that they are founded by women and focus almost exclusively on electing women to Congress and a range of state offices. “EMILY” stands for “Early Money Is Like Yeast” (it makes the dough rise) and was founded in 1985, when women held only 5% of congressional seats (Day and Hadley Reference Day and Hadley2005). The founders of EList had played active roles in women's PACs in the 1970s but felt that the bipartisan nature of these organizations was inhibiting women's electoral success. By 1992, EList had achieved considerable notoriety by helping elect 24 Democratic, pro-choice women to Congress and boasting 23,000 members on its list.Footnote ¹

Other women's PACs emerged as partisan or ideological counters to EList (e.g., the WISH List and the Susan B. Anthony List), although none have achieved the same level of success as EMILY (Burrell Reference Burrell, Carroll and Fox2006, 155). Women's PACs run the ideological spectrum and are as varied in their objectives as are female citizens, activists, and candidates (Day and Hadley Reference Day and Hadley2005, 1–2; see also Burrell Reference Burrell, Lovenduski and Norris1993; Burrell Reference Burrell, Carroll and Fox2006; Hartmann Reference Hartmann1989).

The early moneyFootnote ² that EList provides is largely in the form of bundled checks that come from its network of members (Pimlott Reference Pimlott2007, 8). Because EList can collect contributions from its members and forward them to the candidates, official reports of what the PAC gives to an individual candidate may underestimate the actual financial impact (Day and Hadley Reference Day and Hadley2005, 5–6; Burrell Reference Burrell, Carroll and Fox2006, 156–57). EList members are prompted to send checks via press releases, e-mails, or postal mailings announcing that a particular candidate has been endorsed. As a result of this “early money” strategy, as well as its leverage of the power of its membership, EList “has been able to provide its endorsed candidates with hard contributions that far exceed the limits afforded to PACs” (Pimlott Reference Pimlott2007, 6). Additionally, EList is the largest and wealthiest of the women's PACs. It is not just wealthy in comparison to other women's PACs—its total contributions to candidates stand at or near the top of all PACs (Day and Hadley Reference Day and Hadley2005, 24; Burrell Reference Burrell, Carroll and Fox2006). For example, in 2006, EList's total receipts were $34,118,930, and in 2008, the total receipts were $35,232,112.Footnote ³

In its first four election cycles (1986–1992), EList focused on raising money and contributing to candidates. Beginning with its success in 1992, it evolved as an organization and expanded to also conduct campaign-related research; train candidates and campaign staff; identify and contact voters; and conduct surveys, focus groups, and advocacy campaigns (Day and Hadley Reference Day and Hadley2005, 27; Pimlott Reference Pimlott2006, 64–81). Despite the expanded role of the PAC and the complexities of electing women to Congress, EList continues to assert that supporting women in primaries by providing them with early money is essential to electoral success. In her empirical examination of the impact of EList on elections between 2000 and 2006, Pimlott (Reference Pimlott2007) found a significant positive relationship between EList endorsement and primary victory. She observes that “EMILY's List candidates are more likely to win their primaries than candidates not endorsed by EMILY's List,” but also notes that other factors such as incumbent/challenger status, total money raised, and the political atmosphere of the district or state are likely to be critically important (9).

EList has championed recruitment, nomination, and financial support to expand the role of the typical PAC and has not taken for granted that female Democratic candidates will win if they run. Sanbonmatsu (Reference Sanbonmatsu2006b) found that organization outside of the Democratic party has advantages for women, particularly in cases in which the party is less likely to select women as candidates. In fact, women's PACs such as Hope Chest in Ohio and Lillian's List in North Carolina “were modeled self-consciously after EMILY's List” in an effort to elect women outside of state Democratic party support (194). Since the party organizations are a patchwork among the states, it cannot be assumed that female Democratic candidates will receive party support, despite research that suggests the party is improving their support of women candidates (Burrell Reference Burrell1994; see also Sanbonmatsu Reference Sanbonmatsu2002, Reference Sanbonmatsu2006a; Niven Reference Niven2006; Duerst-Lahti Reference Duerst-Lahti, Thomas and Wilcox1998).

Party leaders ought to be interested in finding the strongest candidate when the two parties are competitive in the particular district or state, and the office itself should likewise have an impact on the extent to which party leaders are selective among a pool of candidates (Aldrich Reference Aldrich1995; Aldrich Reference Aldrich2000). Sanbonmatsu (Reference Sanbonmatsu2006b) observes that, in the past, voters have been more supportive of women as candidates than have party leaders (see also Niven Reference Niven2006). Changing attitudes over time may reflect the changing role of the party with regard to the recruitment and election of women. The notion that EList has become somewhat of a party adjunct in recent years (Pimlott Reference Pimlott2007, 1) seems to fit with research suggesting that women's political groups that ally themselves with the Democratic party are much stronger partisans than organizations that endorse Republican women (Cooperman Reference Cooperman2001) and thus able to wield more influence in the context of an election.

One of the recurring questions regarding ideological PACs such as EList is how the endorsements are decided. Aside from the requirement that the candidate be a pro-choice, Democratic woman, EList determines the viability of a candidate based on her experience, existing network of supporters, access to some money, and existing campaign organization (Day and Hadley Reference Day and Hadley2005, 25; Pimlott Reference Pimlott2007). In other words, EList does not endorse candidates who are starting from scratch. This strategy may provide a framework with which to view EList endorsees and their likely electoral success, but it also begs the question of the role actually played by EList in the electoral success of the candidates it selects. Day and Hadley's (Reference Day and Hadley2005) assessment of EList suggests that it does not merely pick the winners, but also makes their success possible by identifying and contacting donors and voters, and providing research and advocacy on behalf of women as candidates. This assertion has not been tested, however. In short, there appears to be a problem of causal inference regarding endorsements predicting electoral success or likely winners receiving endorsements. The role of EList and other women's PACs could be better understood with further empirical investigation.

HYPOTHESES AND DATA

Does an endorsement by EList predict electoral success? Previous empirical studies have tested for a linear relationship with traditional controls. The most well-established variable predicting electoral success is incumbency (Mayhew Reference Mayhew1974). Incumbency has been a barrier to women achieving widespread electoral success (Duerst-Lahti Reference Duerst-Lahti, Thomas and Wilcox1998), and because of this structural barrier, open seats are also of particular interest to women candidates (Sanbonmatsu Reference Sanbonmatsu2006b). In addition to incumbency, other traditional covariates have included the amount of money raised, candidate gender, the political ideology of the district, and political culture, typically controlled for as percent voting Democratic in either the district or the state for the last presidential election. The particular year may also have an impact on women's electoral success (Burrell Reference Burrell, Lovenduski and Norris1993; Burrell Reference Burrell1994; Delli Carpini and Fuchs Reference Delli Carpini and Fuchs1993; Lawless Reference Lawless2004) so the year of the election has been a traditional covariate as well.

Although these are the typical controls that would be considered in empirical studies examining the impact of PACs on election outcomes, our research question requires us to leverage on causality, and, therefore, a traditional linear model is insufficient. In the following section, we present a propensity score–matching model to test our research question. Propensity matching reproduces an experimental condition in which no controls are needed because cases are matched based on the predictive condition or independent variable(s) instead of randomizing the experimental condition or independent variable(s).Footnote ⁴

MODELS

Using propensity score matching requires us to engage in a three-step modeling process. The first step is to run a logistic regression with endorsement as the dependent variable and the following independent variables (representing all our available data): incumbency (2 variables: challenger and incumbent, with open seat as the reference), gender of the opponent in the general election, total money raised by the candidate, population of the candidate's district, and ideology of the district (operationalized by the percentage of the vote received by the Democratic presidential candidate in the previous presidential election). We also controlled for the election year through dummy variables, and our reference category was 2008. This model produces predicted group membership (endorsed or not) for each candidate and provides us with the probability of group membership (the probability of endorsement).

Table 1 shows the results of the model for predicting the probability of endorsement. It appears that the propensity of endorsement is related to the amount of money a candidate previously raises, the partisan composition of the district, and the election year. The propensity of endorsement increases with the amount of money raised; it is the highest for open seats and lowest for incumbents. Furthermore—and as expected, given the mission of the organization—the more Democratic a district is, the higher the propensity of endorsement. It also appears that the propensity of an EList endorsement decreases over time. The candidates in our sample who ran in elections from 1996 to 2002 were more likely to receive an endorsement than candidates in subsequent election years. The highest likelihood of endorsement occurred in 1996. This finding is consistent with what is known about the evolution of the organization. The organization went through significant changes in the mid- to late 1990s, including the creation of several new programs that greatly expanded the services provided to candidates. Part of this transformation included EList's increased focus on a more strategic endorsement process, leading to fewer candidates overall being endorsed each year; however, those candidates receiving endorsements were identified by EList as strategic recipients in targeted races.

Table 1

Predicting the Probability of Endorsement

Note. N = 504.

* p < .05,

** p < .01,

*** p < .001

Because the model is an acceptable fit, we can trust its predicted probabilities for propensity matching purposes. We can now use these probabilities to divide cases into quartiles for final analysis and testing of our research question. We selected quartiles because of the relatively small sample size (quintiles or larger splits are more commonly used). The second step of our analysis involved separating our original cases into subsamples based on the predicted probabilities produced above. The first subsample (first quartile) is composed of the candidates with the lowest probabilities of endorsement (0%–25%). This group contained 285 candidates, 28 of whom were endorsed by EList. The second subsample is the second quartile (25%–50%). This group contained 70 candidates, and 29 were endorsed by EList. The third quartile (50%–75%) contained 49 candidates, 37 of whom were endorsed by EList. The fourth quartile (75%–100%), comprising candidates with the highest propensity of endorsement, contained 100 candidates, 86 of whom were endorsed by EList. Assuming no mistakes were made in the variable selection process, these quartiles contained candidates who did not differ on relevant characteristics that predict electoral success other than EList endorsement. For this reason, the cases within each quartile can be treated as if they were randomly selected to be endorsed or not—just as if we had experimental data.

To complete our final step in testing the relationship between EList endorsement and electoral success, we set up three bivariate models for each quartile. The first model was a logistic regression in which electoral success was predicted by the single dichotomous independent variable (endorsement). The logistic regression model estimation included a simple chi-square test of the 2 × 2 table reported along with the exponents of the endorsement coefficients. To examine the impact of an EList endorsement on the margin of victory, we set up a two-sample t-test to explore the differences between the endorsed candidates endorsed and the nonendorsed candidates. The same test was also conducted to examine the relationship between an EList endorsement and the percentage of votes received. Results of these analyses are presented in the following section.