3

The Jackie (and Jill) Robinson Effect:

Why Do Congresswomen Outperform Congressmen?

Sarah Anzia

Stanford University

Department of Political Science

[email protected]

and

Christopher Berry

The University of Chicago

Harris School of Public Policy

[email protected]

This Draft: May 24, 2010

Abstract: We argue that the process of selection into political office is different for women than it is

for men, which results in important differences in the performance of male and female legislators

once they are elected. If voters are biased against female candidates, only the most talented, hardest

working female candidates will succeed in the electoral process. Furthermore, if women perceive

there to be sex discrimination in the electoral process, or if they underestimate their qualifications

for office relative to men, then only the most qualified, politically ambitious females will emerge as

candidates. We argue that when either or both forms of sex-based selection are present, the women

who are elected to office will perform better, on average, than their male counterparts. We test this cen-

tral implication of the theory by using legislators’ success in delivering federal spending to their

home districts as our primary measure of performance. We find that congresswomen secure rough-

ly 9 percent more spending from federal discretionary programs than congressmen. This amounts

to a premium of about $49 million per year for districts that send a woman to Capitol Hill. Finally,

we find that women’s superiority in securing particularistic benefits does not hurt their performance

in policymaking: women also sponsor and cosponsor more bills per congress than their male col-

leagues.

Women are a minority in legislatures across the U.S. In 2010, women hold only 17 percent

of the seats in each chamber of Congress and 24.3 percent in state legislatures (CAWP 2009).

Granted, women have made advances in politics in recent decades, but even today, 11 percent of

American adults openly admit that they would not vote for a woman for president (Newport and

Carroll 2007). Moreover, qualified women express greater hesitation about running for office than

similarly qualified men (Fox and Lawless 2004).

In this paper, we draw a connection between models of political agency and the economics

of discrimination, linking both to the vast literature on women in politics. We propose that the

process of selection into office is different for women than it is for men, resulting in important dif-

ferences in the performance of male and female legislators once they are elected. This phenomenon,

which we call ―sex-based selection,‖ can occur in one or both of two ways: First, if voters discrimi-

nate against female candidates, only the most talented, hardest working female candidates will win

elections. Second, if women in the political eligibility pool underestimate their qualifications for of-

fice relative to men, or if women perceive there to be sex discrimination in the electoral process,

then only the most qualified, politically ambitious females will emerge as candidates. We argue that

when either or both forms of sex-based selection are present, the women who run and win office will per-

form better, on average, than their male counterparts. We test this proposition by evaluating the

success of congresswomen relative to congressmen in delivering federal dollars to the home district.

The paper proceeds as follows. Section one reviews the related literature. In section two, we

describe our sex-based selection theory and its implications for the quality and effectiveness of

women in legislatures. Section three details the methods and data we use for the empirical analysis.

The fourth section presents the results of a series of empirical tests of the relationship between legis-

lator sex and job performance. Section five concludes.

1. Related Research and Background

In the empirical literature on the distribution of federal spending across congressional dis-

tricts, little has been done to estimate differences in distributive spending by legislator sex. Mean-

while, the literature on legislative productivity has focused almost exclusively on general conditions

within the legislature rather than the characteristics of legislators themselves.

Within the literature

on women in politics, however, there is a great deal of relevant scholarly work.

To start, scholars have amassed evidence that men and women of equal political qualifica-

tions do not entertain the possibility of running for office in equal frequencies. Lawless and Fox

(2005) find that politically eligible women with the same objective qualifications as men are less like-

ly to consider themselves qualified to run for public office. Moreover, women express greater con-

cern than men about their ability to raise the necessary financial support and win elections (Duerst-

Lahti 1998; Fowler and McClure 1989; Fox and Lawless 2004; NWPC 1994). The differences in men

and women’s political ambition might be the result of differences in male and female socialization,

psychology, and personal life circumstances (Burrell 1994). In addition, women’s political ambition

might be dampened by the perception of sex bias in politics: over 90 percent of women and 75 per-

cent of men in the candidate eligibility pool (e.g., attorneys, business people, educators, and political

activists) believe that there is bias against women in elections (Lawless and Fox 2005).

Some scholars have argued that such concerns are unwarranted. One of the most well

known findings in the literature on women in politics is that female candidates raise as much money

and win general elections at the same rate as male candidates (Burrell 1994; Fox 2006; Newman

1994; Seltzer, Newman, and Leighton 1997; Uhlaner and Schlozman 1986). Based on this evidence,

many scholars have concluded that discrimination against women in politics is a phenomenon of the

past (e.g., Fox 2006, Seltzer et al. 1997, Smith and Fox 2001).

However, a recent Gallup survey showed that 11 percent of both men and women said they

For important contributions, see Howell, Adler, Cameron, and Riemann (2000) and Clinton and Lipinski (2006).

would not vote for a female presidential candidate even if she were qualified for the job, and another

11 percent said they would vote for a qualified woman only ―with reservations‖ (Newport and Car-

roll 2007).

This almost certainly understates the prevalence of sex bias among voters, since pressure

to provide socially desirable responses to public opinion polls likely prevents some respondents

from admitting their underlying prejudices, if they are even aware of them (see Fox and Smith 1998).

Even so, this figure is more than double the percentage of respondents who said they would not vote

for a black candidate for president. In an innovative paper, Mahoney (2007) revisits the question of

sex bias in elections and finds that candidates with male-sounding first names win an additional 15

percent of the vote in state legislative elections, controlling for the candidates’ true sex.

Moreover, a series of experimental studies find that voters do harbor bias against female can-

didates. Rosenwasser and Dean (1989) find that voters prefer ―masculine‖ traits in candidates for all

levels of public office, and Huddy and Terkildsen (1993a, 1993b) show that voters’ gender stereo-

types are most harmful to female candidates running for national office. Fox and Smith (1998)

present subjects with a series of hypothetical male and female House candidates and find that signif-

icantly fewer subjects choose to vote for female candidates (see also Dolan 1997).

The path to congressional office also presents more hurdles to women than to men. Law-

less and Pearson (2008) find that congressional primary elections in which at least one of the candi-

dates is female tend to attract larger numbers of contenders. Also, Palmer and Simon (2006) show

that female incumbents are significantly less likely than male incumbents to face uncontested prima-

ry and general elections. Similarly, Milyo and Schosberg (2000) find that female candidates are signif-

icantly more likely to face high-quality challengers than male candidates. Moreover, political party

leaders believe that there is generally more uncertainty about a woman’s electability than a man’s;

Due to its timing, the survey might conflate voters’ opinions about women with their views on Hillary Clinton’s candi-

dacy, but the aggregate responses do not differ dramatically from surveys conducted in the 1990s (Fox and Smith 1998).

There is some evidence that female candidates for the U.S. House actually have an advantage in gaining support from

female voters (see Smith and Fox 2001), but conclusions on the subject are mixed (see Philpot and Walton 2007).

hence, they are less likely to recruit women to run for office (Sanbonmatsu 2006). Notably, the

women who do emerge as congressional candidates tend to have greater political experience than

male congressional candidates (Pearson and McGhee 2007).

Another large literature examines the differences between men and women once they are in

office. There is evidence that female legislators direct more of their attention to policy areas thought

of as ―women’s issues‖ (e.g., Norton 1999; Thomas 1991; Swers 2002). In addition, the presence of

women in legislatures has been shown to influence the nature of policy outcomes (Besley and Case

2003; Chattopadhyay and Duflo 2004; Rehavi 2007). For the most part, however, this literature is

not well integrated with work that examines the performance of women in pre-election politics. In

the next section, we propose a theory of political selection that connects the performance of women

in campaigns and elections with their performance once in office.

2. A Theory of Political Selection: The Jackie (and Jill) Robinson Effect

In 1947, Jackie Robinson became the first African American to play Major League Baseball.

He is widely revered as one of the greatest players in the history of the game. This is no coincidence.

If Robinson could have been easily replaced by a white player, no team would have been willing to

take a chance on him, given the widespread bigotry of the time. Robinson had to be better than al-

most any white player in order to overcome the prejudice of owners, players, and fans.

Of course,

this story is not unique to Robinson. Pascal and Rapping (1972) found that black Major League

Baseball players in 1967 outperformed white players in every position. Nor is the story unique to

baseball. There is widespread evidence that black athletes have historically faced higher performance

standards for entry into professional sports than white athletes (see Kahn 1991). More generally,

This view is explained by Hank Aaron, himself an African American former ballplayer and erstwhile holder of the ma-

jor league career home-run record. According to Aaron (1999), ―Jackie Robinson had to be bigger than life. He had to

be bigger than the Brooklyn teammates who got up a petition to keep him off the ball club, bigger than the pitchers who

threw at him or the base runners who dug their spikes into his shin, bigger than the bench jockeys who hollered for him

to carry their bags and shine their shoes, bigger than the so-called fans who mocked him with mops on their heads and

wrote him death threats.‖

Becker (1957) pioneered the idea that workers who face discrimination in the labor market must per-

form better in order to earn the same wage as other workers.

We suggest that a similar performance premium is demanded of female politicians when

there is sex discrimination in the electorate.

If voters are prejudiced against women, then a woman

must be better than the man she runs against in order to win.

Moreover, if women anticipate dis-

crimination by voters, or simply underestimate their own qualifications, then only the most formida-

ble women will run for office to begin with. In either case, our prediction that sex-based selection

will lead the women in office to perform better, on average, than the men flows naturally from the

literature on political agency, which focuses on two issues: moral hazard and selection.

In moral hazard models of elections, originating with Barro (1973) and Ferejohn (1986), the

desire to be reelected in the future motivates politicians to exert effort while in office. Citizens vote

retrospectively, reelecting the incumbent only if his performance is above a threshold level chosen to

maximize the incumbent’s incentive to work and hence the voter’s ex ante expected utility. Other

contributions that focus on elections as sanctioning devices for inducing effort from politicians in-

clude Austen-Smith and Banks (1989), Seabright (1996), and Persson, Roland, and Tabellini (1997).

A second body of theory conceives of elections as devices for selecting high-quality politi-

cians, or ―good types,‖ into office (e.g., Zaller 1998, Gordon, Huber, and Landa 2007). In this view,

voters use information gleaned from campaigns and from incumbents’ performance in office as sig-

nals about intrinsic characteristics of candidates, such as talent or honesty. Elections select good

types and filter out bad types, but they do not alter politicians’ behavior in office.

For a recent survey of the economics of discrimination, see Rodgers (2006).

It may seem that a more obvious analogy is with racial discrimination in politics. However, the use of race-conscious

districting confounds the problem. We return to this issue at the end of the paper.

Technically, some voters might reverse discriminate, that is, give preference to female candidates. Our prediction still

holds as long as the proportion that discriminates is greater than the proportion that reverse-discriminates. It is worth

noting that the greater the level of discrimination by voters, donors, and gatekeepers, the greater should be the observed

quality differential for women who win elections. Of course, if discrimination is strong enough, it is possible that no

quality advantage will be sufficient to overcome it, in which case we should not observe women winning elections.

There have been attempts to adjudicate between the electoral selection and moral hazard

models (e.g., Fearon 1999) as well as attempts to unify them (e.g, Ashworth 2005, Banks and Sunda-

ram 1998, Besley 2006). We do not stake out a position on such issues but rather emphasize that the

implications of sex-based selection are the same in either framework. We assume that performance

is a function of a candidate’s innate ability and her effort. In other words, a candidate will perform

better in office if she is more able, works harder, or both. Therefore, if voters discount the ability of

female candidates (electoral selection model), or if voters demand a higher performance threshold

for women (moral hazard model), then the women who win will perform better in office than the

men who win, on average. If female candidates anticipate that they will face discrimination in the

election or otherwise underestimate their chances for electoral success, the women who do run for

office will be those who expect to exceed the higher performance threshold demanded by voters.

We note that in order for this prediction to hold, the attributes that make someone a high

quality candidate must be related to the attributes that make her a high quality legislator. If the two

were uncorrelated, then we would not expect to observe a difference between the performance of

male and female legislators in office. However, as long as candidate quality and legislator quality are

based on similar traits, the process of sex-based selection should result in a legislature in which the

average female representative outperforms the average male representative.

Importantly, our theory of sex-based selection can accommodate two apparently conflicting

strands of evidence from the existing literature discussed above. On one hand, scholars who have

examined female candidates’ vote totals and success rates in House general elections have found that

they do just as well as male candidates (e.g., Burrell 1994). On the other hand, a sizeable proportion

of voters are biased against female candidates in state legislative elections (Mahoney 2007), in presi-

dential elections (e.g., Newport and Carroll 2007), and in hypothetical House election candidate

match-ups (Fox and Smith 1998). Our theory shows that there is no inconsistency between these

sets of findings. If only higher-quality female candidates will actually run for office, then we would

not necessarily expect to observe a vote or campaign funding differential between male and female

candidates even if there is, in fact, discrimination by voters and donors. Yet, if the average female

candidate is of higher quality than the average male candidate but receives the same amount of fund-

ing and wins the same number of votes, she is clearly not on equal footing with the man.

There-

fore, existing studies that compare women’s and men’s vote shares are not directly informative

about the presence or absence of discrimination by voters, since the workings of the candidate selec-

tion stage might mask the presence of voter discrimination at the electoral stage.

Theory aside, the existing evidence suggests that both female self-selection on quality and

voter discrimination are at work. If the average woman running for office were of higher quality

than the average man and voters did not discriminate, then we should observe female candidates

winning at higher rates than men. But they do not. If voters discriminate but women do not self-

select based on quality—implying that the average female candidate is equal in quality to the average

male—then we should see women losing more often than men. But they do not. If the two occur in

combination, such that voters discriminate against female candidates and female candidates self-

screen in anticipation of that discrimination, we would observe fewer but more qualified women

running for office and possibly equal electoral success rates for male and female candidates. This last

set of circumstances is the one most consistent with existing empirical evidence (Pearson and

McGhee 2007). Importantly, however, our theoretical prediction holds regardless of whether dis-

crimination by voters occurs alone or in combination with self-screening by candidates: in either

Again, we find an analogous situation in professional sports. On average, black players in the NBA earn salaries equal

to those of white players. Some see this pay parity as evidence that discrimination has been overcome. Others suggest

that black players are better on average than white players and that salary equality is evidence of discrimination rather

than its absence. See Kahn and Sherer (1988).

case, the women who run and win will perform better, on average, than the men who run and win.

We emphasize that we are not arguing that women have more innate political talent than

men, nor do we claim that all female candidates outperform their male counterparts. Our theory

simply identifies a connection between the economics of discrimination and models of political

agency: when sex discrimination is present among voters, women must be better than their male

counterparts to be elected. If women anticipate such discrimination, or if they underestimate their

chances for electoral success, then only the most qualified women will run in the first place. There-

fore, on average, the women we observe in office will perform better than the men, all else equal.

3. Empirical Strategy and Data

As our primary measure of a legislator’s performance, we look to her success in delivering

federal program spending to her home district. The use of spending as an indicator of incumbent

performance has strong empirical and theoretical foundations. Empirically, congressional scholars

have long observed that a fundamental and explicit goal of members is to bring home federal dol-

lars, and this observation has been a central theme in the classics on Congress (Fenno 1966, 1978;

Ferejohn 1974; Fiorina 1981; Mayhew 1974). There is evidence that such efforts bolster an incum-

bent’s reelection prospects (Alvarez and Saving 1997; Bickers and Stein 1996; Levitt and Snyder

1997; Sellers 1997; Stein and Bickers 1995). Moreover, members of Congress themselves appear to

believe that they must serve their constituents through both casework and project work to build the

reputation necessary for future electoral success (Cain, Ferejohn, and Fiorina 1987).

There is also a strong theoretical motivation for using district spending as an indicator of leg-

islator performance. In particular, Ashworth (2005) presents a model in which reelection-minded

incumbents face a fundamental tradeoff between allocating their resources toward producing dis-

There is, of course, another scenario to consider. In the case that there is no discrimination by voters and potential

female candidates do not self-screen, we would expect that female candidates would win at rates equal to male candi-

dates and that there would be no performance premium on the part of female politicians.

trict-specific benefits, such as federal program spending, or national public goods, such as legislation

or bureaucratic oversight. Voters learn about the ability of incumbents by observing two signals,

which are a function of the politician’s effort on the two tasks, and reelect those politicians whom

they believe are of high ability. A central result from Ashworth’s model is that politicians have an

incentive to bias their effort toward tasks that voters observe with less noise. This logic favors the

dedication of effort to securing district-specific projects, which are more informative signals of the

incumbent’s ability than are national public goods and hence receive greater weight when voters up-

date their beliefs. In other words, it is the observability of program spending that makes it the most

efficient pathway for politicians to signal their quality to constituents.

With these empirical and theoretical motivations, we adopt the not unfamiliar assumption

that legislators are universally motivated to direct projects and funding to their districts (e.g., Evans

2004). Furthermore, while some program spending is formulaic, we assume that a representative’s

talent and effort play an important role in the logrolling, agenda setting, coalition building, and other

deal-making activities that characterize distributive politics. Of two legislators who come from dis-

tricts with similar characteristics (or who represent the same district at different times), the one who

succeeds in directing more spending to her district can be deemed to have performed better in the

context of this fundamental political pursuit. Of course, we recognize that delivering federal benefits

to the home district is only one aspect of a legislator’s job. Therefore, we round out our analysis of

legislative performance by examining legislators’ bill sponsorship and cosponsorship activity.

3.1 Federal Outlays Data

To compare federal program spending in congressional districts represented by men and

women in the U.S. House of Representatives, we use data from the Federal Assistance Award Data

System (FAADS).

FAADS is a comprehensive source for federal domestic spending programs and

See Appendix A for a detailed description of the data.

reports expenditures of about 1,000 programs, including agricultural programs, education grants,

research grants, large entitlement programs, and many others.

We aggregated the FAADS records

to produce a data set that includes 9135 federal outlays observations for congressional district and

fiscal year combinations, tracking approximately $20.8 trillion in federal expenditures from 1984 to

2004 (in 2004 dollars). We attribute the federal outlays for each fiscal year to the member of Con-

gress who represented the congressional district in the calendar year prior.

Congressional district boundaries are redrawn every ten years due to decennial reapportion-

ment and redistricting, and therefore we had to trace districts over time in constructing our panel.

After a decennial redistricting, some districts remain essentially intact while others change beyond

recognition. We consider a district to be a continuous entity across redistricting periods if the major-

ity of the land area of the post-redistricting district is made up of pre-redistricting district land area.

Otherwise, when we could not match a new district clearly to a pre-existing district, the new district

is treated as a new unit following the decennial redistricting. Consequently, the panel includes 733

unique district entities over the 21-year time period.

FAADS reports award transactions and recipient congressional districts according to the ini-

tial recipient. This poses a problem for awards made to states for redistribution throughout the state:

FAADS grossly inflates federal outlays to the congressional districts that contain state capitol build-

ings. We improve a bit upon Levitt and Snyder’s (1995) treatment of this issue by including a control

variable equal to the fraction of the state capitol county contained in each congressional district,

weighted by the state population.

More importantly, we include district fixed effects, as explained

below, which control for time-invariant factors, such as being part of the state capitol county.

Finally, we note that the FAADS data include a great deal of federal spending by broad-

The major omission is defense: military spending and defense procurement programs are not included.

The results presented in the following pages are not highly sensitive to the land area threshold used for matching dis-

tricts across years. See Appendix A.

Details are provided in Appendix A. We transform this variable to a natural log.

based entitlement programs, such as Social Security and Medicaid, the distributions of which are de-

termined by formula. It hardly seems appropriate to attribute this kind of spending to the political

skill and effort of a district’s representative. In order to separate broad-based entitlement programs

from programs that represent discretionary spending, we adopt a tactic used by Levitt and Snyder

(1995, 1997). Specifically, we calculate coefficients of variation in district-level spending for each

program contained in the FAADS data and use the coefficients to separate programs into two cate-

gories: low variation programs have coefficients of variation less than 3/4, and high variation programs

have coefficients of variation greater than or equal to 3/4.

The low variation category includes 26

programs, most of which are programs within the Veterans Benefits Administration, the Centers for

Medicare & Medicaid Services, and the Social Security Administration, which make up 76 percent of

the spending in our data. The high variation category comprises hundreds of smaller programs. In

the analysis that follows, we examine spending from high variation programs only, in expectation

that legislator ability and effort play a larger role in the distribution of high variation spending.

We adjust the spending data to 2004 dollars. The mean value of district-level high variation

program outlays ranges from $398 million in 1984 to $753 million 2003. The median value increases

from $151 million to $361 million. Of the 9135 congressional district and fiscal year combinations,

8307 observations represent annual outlays to districts led by male legislators, and the remaining 828

are for women-led districts. There are 112 unique women in the dataset. Thirty-eight states had at

least one female member of Congress during this time period. Of the 733 unique districts we ob-

serve across the three redistricting plans, 133 had a female representative for at least one congress.

3.2 Identification Strategy

For our main analysis, we use a differences-in-differences approach, based on district and

year fixed effects, to identify the effect of having a female representative on a district’s receipt of

We also tried four thresholds greater than 3/4, none of which produced notably different results. See Appendix A.

See Appendix B for a discussion of results using spending from low variation federal programs.

federal program outlays. Essentially, we ask whether a district receives more federal spending during

the years in which it sends a woman to Washington compared to the years when it sends a man. Be-

cause we rely on variation within districts over time for identification, we can eliminate any (observ-

able or unobservable) time-invariant attributes of a district that could influence both the likelihood

of electing a woman and the flow of federal spending. Importantly, the district fixed effects also

subsume time-invariant heterogeneity across states, such as the well known result that smaller states

receive greater federal outlays, on a per capita basis, due to malapportionment in the Senate (e.g.,

Lee 1998). We specify the following basic model:

ln(outlays

) = β

+ β

) + δ

+ ψX

+ ρZ

+ α

where subscript i denotes congressional districts and t denotes time. The variable of interest is F

which is a binary indicator variable coded one if the person representing district i at time t is female,

zero if male. We include year indicators, δ

, to control for general changes in spending over time.

The vector X

denotes other legislator characteristics that may influence spending. We con-

trol for party, which is expected to account for the traditional Republican preference for fiscal con-

servatism which conceivably tempers the push for more spending to the home district.

In anticipa-

tion that legislators in the majority party are better positioned to secure money for their districts, we

include an indicator for majority party status during years in which Congress and the presidency are

controlled by the same party. Unified government occurs in only 5 fiscal years within our study pe-

riod: in 1994, 1995, and 2002 to 2004.

We also add a measure of seniority: the number of terms a

legislator has served as of the year of the outlays. Finally, we introduce a variable that equals a mem-

ber’s two-party victory margin in the preceding congressional election, which controls for the possi-

bility that electorally vulnerable members receive priority in discretionary spending (Shepsle 1978).

Alvarez and Saving (1997) find that Democrats reap greater electoral benefit from funneling pork to their districts

than do Republicans.

Recall that FAADS data are reported in fiscal years and that legislator characteristics are lagged by one year.

The vector Z

captures a fairly rich set of observable attributes of congressional districts:

population living in urban areas, African American population, population 65 years of age or over,

number of farmers and farm managers, foreign-born population, median family income, unem-

ployed population, population in the armed forces, population in public school, and population em-

ployed in manufacturing and construction.

Although a supplementary analysis shows that district

demographic characteristics are not important predictors of the presence of a female legislator (see

Appendix B), we remain concerned that unmeasured district characteristics predict both legislator

sex and the amount of federal spending received by a district. We therefore include congressional

district fixed effects, α

, to account for unobservable, time-invariant district characteristics.

Finally,

, ψ, and ρ are regression coefficients, β

is a constant, and ε

is an error term.

Even with a broad set of control variables, the unobservable, time-variant predictors of fed-

eral spending within a particular district are likely to be correlated across time periods. Furthermore,

the geographic distribution of federal spending likely reflects the effects of senators as well as the

quality and effort of House members, suggesting that there may be correlation across districts within

a state. Consequently, we use robust standard errors clustered by state throughout our analysis.

4. Analysis and Results

Table 1 presents the results of our fixed effects models of high variation program spending.

Model (1) includes district characteristics, legislator characteristics, and district and year fixed effects,

as described above. The main result is clear: within districts over time, roughly 9 percent more federal spending

is brought home when there is a woman representing the district in Congress than when the district is represented by a

man.

We note that the inclusion of district fixed effects and the control variables works to dampen

Demographic variables for 1984 to 2001 come from Scott Adler’s ―Congressional District Data File.‖ All demograph-

ic data for 2002 to 2004 come from the 2000 U.S. Census. Details are provided in Appendix A.

In some instances, explained below, we use state fixed effects because of data limitations.

The models in table 2 have 9,067 observations, rather than 9,135, due to missing values for electoral margin and terms

in office. We also exclude 5 observations that recorded negative high variation outlays. See Appendix for details.

the magnitude of the coefficient on legislator sex relative to a model that includes only the female

indicator and year fixed effects (see Appendix B), but the effect of sex in column (1) is substantively

large and highly statistically significant.

The district fixed effects subsume any attributes that do not change over time, including the

unchanging attributes of the states in which they are located. However, a lingering concern may be

that there are unmeasured trends within districts over time that make them both more likely to elect

a woman and more likely to receive federal spending. To explore this possibility, we compare the

rates of change in the federal spending received by a district before and after it elects a female repre-

sentative. Specifically, in model (2), we use pre- and post-female linear time trends for the 3 terms

before and the 3 terms after the election of a woman. We find that the rate of increase in spending is

higher after a woman is elected than it was before. Using an F-test, we reject that the two trends are

equal (p=.06). Thus, we find no evidence that female representatives merely inherit an already favor-

able trend in spending; instead, the trend changes after a woman is elected.

Next, in column (3), we present a model that includes state rather than district fixed effects.

This approach allows us to take advantage of more variation in the data, as 38 states had at least one

woman in Congress during our study period. A disadvantage is that the state fixed effects do not

account for unmeasured within-state, between-district heterogeneity. That the results of the state

and district fixed effects models are so similar, therefore, is reassuring and strengthens our belief that

district-level attributes do not explain the connection between legislator sex and federal spending.

Finally, in the spirit of a regression discontinuity (RD) design (Thistlewaite and Campbell

1960, Lee 2008), we estimate changes in district-level spending following close elections in which a

male candidate ran against a female candidate and in which the election resulted in a change in the

sex of the district’s representative. Based on the closeness of the elections, we can infer that each

district was roughly equally likely to have elected a woman or a man. When we restrict our analysis

to mixed-sex races in which the winning candidate garnered less than 55 percent of the vote, there

are 39 instances in which the sex of a district’s representative changed as a result of the election.

Model (4) is a first-differences regression in which the change in spending from the year before to

the year after the election is regressed against changes in legislator sex and other covariates. The fe-

male effect from the close elections sample is 7 percent. Note that because we have only 39 observa-

tions in model (4), we do not attempt to control for the full set of district covariates. This is not a

major concern, since we do not expect district attributes to change significantly in two years. We do

control for legislator characteristics that may change along with the sex of a district’s representa-

tive—party, majority status, and seniority—and the coefficients for these variables are comparable to

their fixed effects counterparts. While we admittedly have few instances of close elections that pro-

duce a change in the sex of a district’s representative, and therefore do not put much stock in these

results taken in isolation, the findings from model (4) comport with those from the fixed effects

models and provide a useful complement to them.

Among the remaining variables included in table 1, only a handful demonstrate a robust rela-

tionship with federal spending. Democratic districts appear to garner more federal money in the

state fixed effects model, but the result dissolves when district fixed effects are introduced. In other

words, it would appear that Democrats come from districts that are otherwise prone to receive fed-

eral largess, but within-district changes in legislator party are not significantly associated with

changes in spending to the district. Membership in the majority party appears to be uncorrelated

with district spending.

Tenure in office, while positive in every specification, fails to attain statis-

tical significance. Furthermore, the size of a congressperson’s victory margin does not appear to in-

fluence the allocation of federal spending. Among the district attributes, an increasing number of

Such races are close by congressional standards, where the average election is decided by a 40-percent margin and only

7 percent of elections are decided with the winning candidate earning less than 55 percent of the vote.

Appendix B provides additional RD analysis and discusses the strengths and weaknesses of our data for this design.

While contrary to popular wisdom, these results are consistent with prior studies, such as Knight (2005).

African Americans is associated with increased federal spending over time. Lastly, as expected, dis-

tricts that contain more of the population of the state capitol county receive more federal spending.

In summary, the unambiguous result is that female legislators succeed in directing more dis-

cretionary spending to their home districts than male representatives. A spending advantage of 9

percent amounts to approximately an extra $88 per capita per year for districts represented by wom-

en. Given that the average district has 563,732 residents, the aggregate spending increase for the dis-

trict is roughly $49 million when it sends a woman to Capitol Hill.

4.1. Is It Sex-Based Selection?

The results of the previous section provide strong evidence that congressional districts re-

ceive more federal funding when they are represented by women than when they are represented by

men. However, we have not established that the source of the spending difference is that congress-

women are more able legislators, nor that the reason for their differential success in office is what

we refer to as sex-based selection. The results thus far leave open the possibility that women are

stronger legislators simply because they are more attuned to their constituents, more dedicated to

procuring funds for so-called women’s issues (e.g., Swers 2002), or more collaborative and coopera-

tive in their legislative and leadership style (Carey et al. 1998, Kathlene 1994, Rosenthal 1998).

In order to demonstrate that the mechanism responsible for the female spending advantage

is the one we have proposed, we would like to be able to measure either variance in sex-based selec-

tion or variance in candidate quality across districts and time. With respect to the latter, we know

that female House candidates tend to be more qualified than male candidates on the basis of raw,

formal qualifications such as prior office-holding experience (Pearson and McGhee 2007). However,

quality is much more than formal qualifications, and it is only measurable through performance. In

baseball, for example, we do not know what qualities cause one player to hit more home runs than

another, and it is certainly something more than just the players’ experience, but we are comfortable

concluding that the player who hits more home runs is a better hitter. Candidate quality is similar in

nature. Of course, if we cannot measure quality as distinct from performance, we cannot hope to

exploit variation in candidate quality to isolate its effect on performance.

Alternatively, to measure sex-based selection, we would want to quantify either the degree of

sex discrimination in the district’s electorate or the extent to which higher quality women self-select

into politics relative to men. We would expect to find a positive relationship between congresswo-

men’s spending advantage and the level of sex-based selection in the district, conditional on a wom-

an being elected. Unfortunately, we know of no such measures at the district level, much less the

district-by-year level. Instead, we use average constituent ideology in the district as a proxy, albeit a

somewhat crude one, for the prevalence of sex-based selection in the district. We also examine fed-

eral spending outcomes for women who, we believe, faced fewer barriers to entry to politics because

of their sex than other female legislators.

First, we take advantage of the fact that attitudes about women in politics are correlated with

the ideology of constituents in a district.

We use Clinton’s (2008) survey-based measure of district-

level constituent ideology, which ranges from -1 (most liberal) to 1 (most conservative). This meas-

ure does not capture variation in constituent ideology over time within districts, but it does allow us

to estimate the extent to which the female spending advantage varies systematically with a time-

invariant measure of constituent ideology. If more conservative districts also tend to be those where

average sex discrimination levels are higher or where qualified women are more reluctant to enter

politics, then our theory would predict that the spending advantage achieved by female legislators in

more conservative districts will be greater than the advantage received by those in liberal districts.

Table 2 presents the results of the main models with an interaction between legislator sex

Of the 11 percent of Gallup respondents who reported that they would not vote for a well-qualified female candidate

for president, 63 percent identified themselves as either very conservative or conservative. Only 36 percent of those who

said they would vote for a woman were conservative or very conservative (USA Today / Gallup Poll, February 9-11 and

March 2-4, 2007).

and district-level constituent conservatism, the latter of which is centered around its mean. The main

effect of district ideology cannot be directly estimated since it is constant across time periods and is

therefore subsumed within the district fixed effects. The table is truncated to preserve space; all four

models include the covariates whose coefficients are presented in table 1 as well as district and year

fixed effects.

In column (1), the coefficient on legislator sex represents the spending advantage

that accrues to districts of average ideology (because the ideology measure has been mean deviated)

when they have female representatives. Since the average district is slightly conservative according to

Clinton’s measure, we conclude that female legislators elected from a moderately conservative con-

gressional district deliver approximately 13 percent more federal spending to their constituents than

male legislators. More importantly, however, the coefficient on the interaction term is positive and

statistically significant at the 1 percent level. Thus, more conservative districts – when they elect

women to represent them – receive a larger increase in spending than districts that have more liberal

constituents. The magnitude of the coefficient implies that a one standard deviation increase in aver-

age constituent conservatism is associated with an additional 10 percent boost in federal spending.

We might suspect that since constituent ideology is likely to be positively correlated with leg-

islator ideology, the interaction presented in column (1) picks up the ideological leanings of the con-

gresswomen themselves. Since we are concerned here with general views about women in politics in

the district, column (2) enters the legislator’s ideology as a separate regressor, measured by his or her

NOMINATE score (Poole and Rosenthal 1997). The inclusion of the individual members’ ideology

changes the coefficients on the female indicator and the interaction term only modestly. Notably,

the coefficient on the interaction term is still large, positive, and significant at the 1 percent level.

These results are consistent with our argument that the mechanism driving the spending ad-

We lose a substantial number of cases due to the fact that Clinton’s measure of constituent ideology only exists ac-

cording to the congressional district boundaries of the 1990s. Where possible, we used the same values of this measure

for corresponding districts in the 1980s and 2000s. See Appendix A for details.

vantage is a talent and effort differential induced by sex-based selection. We see from columns (1)

and (2) of table 2 that the positive effect of female representation on spending is considerably larger

in districts where public attitudes are likely to be less friendly to the idea of women in politics. Of

course, district ideology is a rough measure for sex-based selection, so we bolster these results by

comparing two groups of women for whom, we assume, the political selection process differs.

One route by which women have historically entered Congress is by succeeding their hus-

bands who passed away while in office (Burrell 1994). If widows benefit from outpourings of public

sympathy surrounding the deaths of their husbands, they are unlikely to be subjected to the same

degree of electoral scrutiny as other women. Moreover, since they have closely followed their hus-

bands’ tenure in office, widows may be less inclined to think themselves insufficiently qualified for

political office than other women. In other words, we would expect that widows are free of many of

the hurdles other women must clear on the way to office; therefore, they would not need the same

edge in quality or effort in order to become candidates and get elected. In fact, widows may even be

able to win with a quality disadvantage relative to male candidates thanks to public sympathy. If sex-

based selection is the mechanism that causes women to perform better in office than men, then wi-

dows should have a smaller spending advantage than other women, and possibly even a spending

disadvantage relative to men. Thus, we compare spending outcomes for districts represented by wi-

dows who succeeded their husbands in office with districts represented by other women as a win-

dow onto female performance in environments with and without substantial sex-based selection.

Column (3) of table 2 presents a test of these predictions. We note upfront that our power

to conduct this test is limited because we have only eight widows in our data set, accounting for a

combined 57 years of presence in the legislature. Nevertheless, the results confirm our expectations.

We create separate binary indicator variables for widows and female non-widows; males are the

We are grateful to Linda Fowler for suggesting this idea.

omitted category. The model includes the full set of legislator characteristics, all the district-level

demographics, and state and year fixed effects. The small number of widows prohibits us from run-

ning district fixed effects models.

We cannot reject the null hypothesis of no spending advantage for widows, while the female

non-widow effect is large, positive, and statistically significant. In fact, the widow coefficient is nega-

tive, suggesting that widows deliver less spending than male legislators, although this difference is

not significant. An F-test allows us to reject the hypothesis that the coefficients for widows and non-

widow females are equal at p = 0.054. This result lends support to the hypothesis that sex-based se-

lection explains the female spending advantage. If widows are not held to a higher standard than

male candidates by voters and not likely to underestimate their qualifications for politics, then we

should not expect female legislators who succeed their late husbands to perform better than male

legislators. The results presented in column (3) of table 2 show that this is the case. Of course, we

recognize that there may be other reasons why widows are less effective in office. These results,

while consistent with our theory, are not dispositive.

4.2. Alternative Explanations

The preceding results show that congresswomen’s spending advantage cannot be explained

by the districts they represent and is even larger in districts where women are elected amidst chal-

lenging conditions. In this section, we address the question of whether there is some other correlate

of being female, apart from the sex-based selection, that can explain congresswomen’s success in

garnering federal spending for their districts.

It is well known that electorally vulnerable members of Congress seek additional spending

for their districts (e.g., Cain et al. 1987). Is it possible that women respond disproportionately to

electoral vulnerability by seeking more federal spending for their districts? We test for this in model

(1) of table 3 by estimating a district fixed effects model that includes an interaction between the fe-

male indicator and the candidate’s electoral margin in the preceding election. (To conserve space,

only the coefficients for the primary independent variables of interest are reported, although the full

set of control variables is included in the models reported in table 3.) If it is true that women re-

spond disproportionately to electoral vulnerability, we should find a negative coefficient on the inte-

raction term. In fact, however, the coefficient is positive and insignificant. We can therefore dismiss

the possibility that electoral vulnerability is at the source of women’s spending advantage.

Next, we investigate the role of partisanship and ideology. The women in Congress during

our study period are more likely to be Democrats (65 percent) than the men are (50 percent). Wom-

en are also more ideologically liberal: the average NOMINATE score (Poole and Rosenthal 1997)

for a female member of Congress is -0.15, while the average for congressmen is 0.05.

In model (2)

of table 3, we estimate an interaction between the female and Republican indicators. While female

Republicans demonstrate a modest edge over female Democrats, the difference is not statistically

significant. In model (3), we estimate the interaction between the female indicator and the NOMI-

NATE scores. Again, the point estimate suggests that conservative women garner more spending

than liberal women, but the interaction is not significant. Based on these analyses, we reject the idea

that partisanship or ideology can explain the female spending differential.

As a next step, we explore one avenue through which women may attain their added spend-

ing: committee assignments. Observing that women achieve more desirable committee assignments

would be consistent with our theory of sex-based selection. However, the observation would also be

equivalent to the alternative explanation that parties display favoritism toward women in the com-

mittee assignment process, perhaps because there are few female members and their presence on

top committees is valuable for other reasons, such as public relations. We use the Groseclose and

Stewart (1998) House committee desirability scores to place values on the committee portfolios of

The NOMINATE scale ranges from -1 (most liberal) to 1 (most conservative).

individual legislators in each year. We find that, controlling for seniority, women have slightly less

desirable committee portfolios, although the difference is not statistically significant (not shown).

In any case, when we control for a complete set of committee indicator variables as well as indicator

variables for committee chairs, committee ranking minority members, and party leaders, which we

do in model (4) of table 3, the estimated female spending advantage is essentially unaffected. Wom-

en do not attain their spending advantage merely by securing better committee assignments.

In analysis not shown, we also investigate the nature of the spending that women bring

home to their districts.

Women in politics scholars have found that female politicians are more

active in areas considered to be ―women’s issues‖ (e.g., Swers 2002). If women derive their advan-

tage in spending primarily from federal programs that reflect traditional ―female‖ legislative priori-

ties, we might be dissuaded that it is women’s talent and effort that drives the spending effect.

To the contrary, we find that the female spending advantage is present across a diverse set of

federal programs. We estimate fixed effects models of spending from each of the four agencies re-

sponsible for the greatest amount of high variation program spending from 1984 to 2004: the De-

partments of Agriculture, Health and Human Services, Transportation, and Education. Only Agri-

culture fails to demonstrate a spending advantage for women. Women have a clear advantage in se-

curing funds for their districts from Transportation, Health and Human Services, and Education. In

fact, the coefficient for Transportation is largest in magnitude, a result that is particularly suggestive

since transportation is the area identified by congressional scholars as especially amenable to pork

barrel politics (e.g., Ferejohn 1974). It is therefore not the case that women only have an advantage

in securing spending for programs related to ―women’s issues.‖

These results are presented in Appendix B.

The null findings for committee chairs, ranking minority members, and party leaders are contrary to expectations.

However, we note that these positions change relatively infrequently, and so the estimates are very imprecise in the con-

text of district fixed effects.

The following results are detailed in Appendix B.

Lastly, we investigate whether congresswomen’s success in securing federal funding for their

constituents comes at the expense of attention to another important aspect of their job – policymak-

ing. As a test of whether congresswomen are less effective policymakers than congressmen, we ana-

lyze bill sponsorship and cosponsorship patterns for male and female members of the House. We

use data made available by Fowler (2006), who compiled sponsorship information for every piece of

legislation proposed in Congress since 1973.

We examine the number of bills sponsored and cosponsored by women relative to men

from 1984 to 2004, modeling each as a function of legislator sex, party, and majority status and in-

cluding indicator variables for committee chairs, ranking minority members of committees, and par-

ty leaders. To account for the possibility that electorally vulnerable members are spurred into action,

we also control for the members’ vote margin in the preceding election. In addition, because some

committees provide more opportunities for legislative activity than others, we include a full set of

committee membership indicator variables. We also include the demographic variables from table 1

to control for the possibility a member might sponsor more bills if she is from a district where con-

stituents are particularly attentive to legislative behavior. We expect that it is easier for ideologically

moderate members to work with larger numbers of their colleagues than more extreme members, so

we include each legislator’s ideological distance from the legislator with the median NOMINATE

score in each congress. Finally, we include congressional term fixed effects to account for general

changes in sponsorship and cosponsorship over time. To conserve space, we do not report the coef-

ficients for the demographic variables, committee variables, or fixed effects.

Clearly, it is not the case that women neglect their roles as policymakers. In fact, model (5) of

table 3 demonstrates that congresswomen sponsor more legislation than congressmen. On average,

See Appendix A for details.

There are 4752 observations in models (5) and (6), rather than the full 4785, due to missing values for Terms (24 miss-

ing) and Margin (9 missing). See Appendix A for details.

women sponsor about three more bills per congress, which is a difference of roughly 17 percent rel-

ative to the member average of 18 bills. Women are also more active in supporting the legislation of

their colleagues through cosponsorship. Congresswomen cosponsor about 26 more bills per con-

gress than congressmen, as seen in model (6). In results presented in the Appendix, we find that

women also garner cosponsorship support from a greater number of their peers, which suggests that

women have stronger networks of collaboration with their colleagues than congressmen.

There are obvious limitations to counting sponsored and cosponsored bills as a measure of

legislators’ attentiveness to policymaking. In particular, the decision to cosponsor a bill is relatively

costless. In more comprehensive examinations of congressional policymaking, however, Volden and

Wiseman (2010) and Volden, Wiseman, and Wittman (2010) track each bill introduced in the 97

110

congresses through all stages of the legislative process – from introduction to signing – and

find that women score significantly higher on their measure of ―legislative effectiveness‖ than men

do. Not only does this evidence refute the argument that women pay close attention to distributing

district-level benefits at the expense of policymaking, but it is consistent with the idea that policy-

making is yet another area in which congresswomen outperform congressmen.

5. Discussion

If we believe the evidence that the average woman underestimates her qualifications relative

to the average man, then it is reasonable to conclude that a woman who identifies herself as a candi-

date for national office is more qualified than the average male candidate. If it takes more talent and

greater effort for female candidates to be taken seriously by voters, campaign contributors, and party

gatekeepers, then the women who succeed in the electoral process are likely to be more talented and

hardworking than the men who do the same. Because of this, the women who are elected to Con-

gress are actually poised to be more effective legislators than their male counterparts.

See Appendix B for additional results.

Our theory of sex-based selection makes precisely this point. It does not matter whether

women are elected to public office at lower rates than men because they perceive their own qualifi-

cations differently or because bias against women in the electorate produces a barrier to entry for

them. The central implication of sex-based political selection is that the women we observe in office

will, on average, outperform the men.

We test this implication using legislators’ success in directing funds to their home districts as

our primary measure of performance. The federal spending data provide strong empirical support

for the prediction that women outperform men. All else equal, congressional districts receive rough-

ly 9 percent more high variation federal program spending when they are represented by women.

This spending bonus amounts to approximately $88 per capita, or $49 million in total, for districts

that have a woman in Washington in a given year. According to the estimates contained in Levitt

and Snyder (1997), the addition of $88 per capita in high variation program spending produces an

electoral reward for the incumbent of almost 2 percent of the popular vote.

However, our results are not invulnerable to criticism. Without a direct way to measure legis-

lator ability or effort, we cannot definitively show that these factors explain female success in office.

In section 4.2, we considered a set of competing explanations for the spending differential and

brought each one to the data. The results allow us to reject the possibility that women’s electoral

vulnerability, differing ideology or partisanship, or advantageous committee assignments can explain

the connection between legislator sex and spending. Moreover, it is not the case that female House

members manage to excel in securing federal spending for their districts by neglecting policymaking:

they actually sponsor and cosponsor more bills per congress than their male counterparts.

While our evidence cannot substitute for a direct test of the relationship between legislator

sex and ability or effort, it dispels several reasonable competing explanations. For example, one

might conjecture that political party leaders intentionally channel disproportionate funding to wom-

en’s districts, either to protect their relatively small cadres of female representatives, or simply to

make it obvious that they do not discriminate against them. Alternatively, perhaps female legislators

feel the need to work harder in order to prove themselves to their colleagues in the male-dominated

House. While these are all plausible explanations for a female spending advantage in general, they

cannot account for why it does not apply to women who succeed their late husbands in office or

why it is greater in districts where constituents are more conservative. Any alternative explanation

for our findings would have to account for all of these patterns, as well as the fact that women spon-

sor and cosponsor more bills than their male counterparts. We believe that our theory of sex-based

selection provides the most logical and parsimonious explanation for these findings.

In closing, we note that our theoretical contribution does not apply uniquely to women or to

the measures of performance that we have chosen. Future research might look for other areas in

which females excel in office. In addition, our theory suggests that members of other groups that

suffer from discrimination by the electorate also must perform better in order to be elected. Future

research might apply a similar analysis to African Americans or Latinos in Congress. However, we

anticipate that the use of race-conscious districting, in particular majority-minority districts, will se-

riously confound testing of the theory. If racial districting makes it easier for minorities to be elected,

then there is no reason to expect that those in office will perform any better than average. Of

course, political selection is not based solely on candidates’ personal attributes. We might expect, for

example, a Republican elected from an historically Democratic district to demonstrate a similar qual-

ity advantage. These are empirical questions that we may explore in future research.

At the most general level, our results highlight the importance of connecting research on

women in politics, models of political agency, and the economics of discrimination. Women are

some of the most effective politicians in Congress. One only has to look to the political selection

process to understand why.

REFERENCES

Aaron, Henry. 1999. ―Jackie Robinson,‖ in TIME Magazine’s TIME 100: The 100 Most Influential

People of the Century, June 14.

Adler, E. Scott. 2003. ―Congressional District Data File, 98

to 105

Congresses.‖ University of

Colorado, Boulder, CO.

Alvarez, R. Michael and Jason L. Saving. 1997. ―Deficits, Democrats, and Distributive Benefits:

Congressional Elections and the Pork Barrel in the 1980s.‖ Political Research Quarterly 50 (4):

809-831.

Ashworth, Scott. 2005. ―Reputational Dynamics and Political Careers." Journal of Law, Economics and

Organization 21 (2): 441-466.

Austen-Smith, David and Jeffrey S. Banks. 1989. ―Electoral Accountability and Incumbency,‖ in P.

Ordeshook, ed., Models of Strategic Choice in Politics. Ann Arbor: University of Michigan Press,

121-148.

Banks, Jeffrey S. and Rangarajan K. Sundaram. 1998. ―Optimal Retention in Agency Problems.‖

Journal of Economic Theory 82 (2): 293-323.

Barro, Robert J. 1973. ―The Control of Politicians: An Economic Model.‖ Public Choice 14 (1): 19-42.

Becker, Gary S. 1957. The Economics of Discrimination. Chicago: University of Chicago Press.

Besley, Timothy and Anne Case. 2003. ―Political Institutions and Policy Choices: Evidence from the

United States.‖ Journal of Economic Literature 41 (1): 7-73.

Besley, Timothy. 2006. Principled Agents? The Political Economy of Good Government. Oxford: Oxford

University Press.

Bickers, Kenneth N. and Robert M. Stein. 1991. Federal Domestic Outlays, 1983-1990: A Data Book.

Aramonk, NY: M.E. Sharpe.

Bickers, Kenneth N. and Robert M. Stein. 1996. ―The Electoral Dynamics of the Federal Pork Bar-

rel.‖ American Journal of Political Science 40 (4): 1300-1326.

Burrell, Barbara C. 1994. A Woman’s Place Is in the House: Campaigning for Congress in the Feminist Era.

Ann Arbor: University of Michigan Press.

Carey, John M., Richard G. Niemi, and Lynda W. Powell. 1998. ―Are Women State Legislators Dif-

ferent?‖ In Women and Elective Office, Sue Thomas and Clyde Wilcox, eds. New York: Oxford

University Press: 87-102.

Cain, Bruce, John Ferejohn, and Morris Fiorina. 1987. The Personal Vote. Cambridge: Harvard Uni-

versity Press.

Center for American Women and Politics (CAWP). 2005. ―Fact Sheet: Women Who Succeeded

Their Husbands in Congress.‖

Center for American Women and Politics (CAWP). 2006. ―Fact Sheet: Women in the U.S. Congress

1917-2006.‖

Center for American Women and Politics (CAWP). 2009. ―Facts on Women in State Legislatures.‖

Chattopadhyay, Raghabendra and Esther Duflo. 2004. ―Women as Policy Makers: Evidence from a

Randomized Policy Experiment in India.‖ Econometrica 72 (5): 1409-1443.

Clinton, Joshua, and Jon Lipinski. 2006. ―Measuring Legislative Accomplishment, 1877-1994.‖

American Journal of Political Science 50(1): 232-249.

Dolan, Kathleen. 1997. ―Gender Differences in Support for Women Candidates: Is There a Glass

Ceiling in American Politics?‖ Women and Politics 17: 27-41.

Duerst-Lahti, Georgia. 1998. ―The Bottleneck: Women Becoming Candidates.‖ In Women and Elec-

tive Office, Sue Thomas and Clyde Wilcox, eds. New York: Oxford University Press: 15-25.

Evans, Diana. 2004. Greasing the Wheels: Using Pork Barrel Projects to Build Majority Coalitions in Congress.

New York: Cambridge University Press.

Fearon, James D. 1999. ―Electoral Accountability and the Control of Politicians: Selecting Good

Types Versus Sanctioning Poor Performance,‖ in A. Przeworski, S.C. Stokes and B. Manin

eds., Democracy, Accountability, and Representation. Cambridge: Cambridge University Press.

Fenno, Richard F., Jr. 1966. The Power of the Purse: Appropriations Politics in Congress. Boston: Little,

Brown and Company.

Fenno, Richard F., Jr. 1978. Home Style: House Members in Their Districts. Boston: Little, Brown and Co.

Ferejohn, John A. 1974. Pork Barrel Politics. Stanford: Stanford University Press.

Ferejohn, John. 1986. ―Incumbent Performance and Electoral Control.‖ Public Choice 50 (1-3): 5-25.

Fiorina, Morris P. 1981. ―Some Problems in Studying the Effects of Resource Allocation in Con-

gressional Elections.‖ American Journal of Political Science 25: 543-68.

Fowler, Linda L. and Robert McClure. 1989. Political Ambition. New Haven: Yale University Press.

Fowler, James H. 2006. ―Connecting the Congress: A Study of Cosponsorship Networks.‖ Political

Analysis 14 (4): 456-487.

Fox, Richard L. 2006. ―Congressional Elections: Where Are We on the Road to Gender Parity?‖ In

Gender and Elections, Susan J. Carroll and Richard L. Fox, eds. New York: Cambridge Univer-

sity Press, 97-116.

Fox, Richard L. and Jennifer L. Lawless. 2004. ―Entering the Arena? Gender and the Decision to

Run for Office.‖ American Journal of Political Science 48 (2): 264-280.

Fox, Richard L. and Eric R.A.N. Smith. 1998. ―The Role of Candidate Sex in Voter Decision-

Making.‖ Political Psychology 19 (2): 405-419.

Gordon, Sanford, Gregory Huber, and Dimitri Landa. 2007. ―Challenger Entry and Voter Learn-

ing.‖ American Political Science Review 101 (2): 303-320.

Groseclose, Timothy and Charles Stewart III. 1998. ―The Value of Committee Seats in the House:

1947-91.‖ American Journal of Political Science 42 (2): 453-474.

Howell, William, Scott Adler, Charles Cameron, and Charles Riemann. 2000. ―Divided Government

and the Legislative Productivity of Congress 1945-94.‖ Legislative Studies Quarterly 25:285-312.

Huddy, Leonie and Nayda Terkildsen. 1993. ―Gender Stereotypes and the Perception of Male and

Female Candidates.‖ American Journal of Political Science 37: 119-47.

Huddy, Leonie and Nayda Terkildsen. 1993. ―The Consequences of Gender Stereotypes for Women

Candidates at Different Levels and Types of Office.‖ Political Research Quarterly 46 (3): 503-25.

Jeydel, Alana and Andrew J. Taylor. 2003. ―Are Women Legislators Less Effective? Evidence from

U.S. House in the 103

-105

Congress.‖ Political Research Quarterly 56 (1): 19-27.

Kahn, Lawrence M. 1991. ―Discrimination in Professional Sports: A Survey of the Literature.‖ In-

dustrial and Labor Relations Review 44 (3): 395-418.

Kahn, Lawrence M., and Peter D. Sherer. 1988. ―Racial Differences in Professional Basketball Play-

ers’ Compensation.‖ Journal of Labor Economics 6 (1): 40-61.

Kathlene, Lyn. 1994. ―Power and Influence in State Legislative Policymaking: The Interaction of

Gender and Position in Committee Hearing Debates.‖ American Political Science Review 88 (3):

560-576.

King, Gary. 1994. ―Elections to the United States House of Representatives, 1898-1992 (computer

file). ICPSR version. Cambridge, MA: Gary King (producer). Ann Arbor, MI: Inter-

university Consortium for Political and Social Research (distributor).

Knight, Brian. 2005. ―Estimating the Value of Proposal Power.‖ American Economic Review 95 (5):

1639-1652.

Lawless, Jennifer L. and Kathryn Pearson. 2008. ―The Primary Reason for Women’s Underrepresen-

tation? Reevaluating the Conventional Wisdom.‖ Journal of Politics 70 (1): 67-82.

Lawless, Jennifer L. and Richard L. Fox. 2005. It Takes a Candidate: Why Women Don’t Run for Office.

New York: Cambridge University Press.

Lee, David S. 2008. ―Randomized Experiments from Non-Random Selection in U.S. House Elec-

tions.‖ Journal of Econometrics 142 (2): 675-697.

Lee, Francis E. 1998. ―Representation and Public Policy: The Consequences of Senate Apportion-

ment for the Geographic Distribution of Federal Funds.‖ Journal of Politics 60 (1): 34-62.

Levitt, Steven D. and James M. Snyder, Jr. 1995. ―Political Parties and the Distribution of Federal

Outlays.‖ American Journal of Political Science 39 (4): 958-980.

Levitt, Steven D. and James M. Snyder, Jr. 1997. ―The Impact of Federal Spending on House Elec-

tion Outcomes.‖ The Journal of Political Economy 105 (1): 30-53.

Mahoney, Neale. 2007. ―Gender Discrimination in Elections: The Impact of Gender-Misleading

First Names.‖ Working paper.

Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale University Press.

Milyo, Jeffrey and Samantha Schosberg. 2000. ―Gender Bias and Selection Bias in House Elections.‖

Public Choice 105: 41-59.

National Women’s Political Caucus (NWPC). 1994. Why Don’t More Women Run? A Study prepared

by Mellman, Lazurus and Lake. Washington, D.C.: National Women’s Political Caucus.

Nelson, Garrison. 2005. Committees in the U.S. Congress, 1947-1992, House Committees 98

-102

Congresses.

Newman, Jody. 1994. Perception and Reality: A Study Comparing the Success of Men and Women Candidates.

Washington, DC: National Women’s Political Caucus.

Newport, Frank and Joseph Carroll. 2007. ―Analysis: Impact of Personal Characteristics on Candi-

date Support.‖ Princeton, NJ: Gallup News Service.

Norton, Noelle H. 1999. ―Uncovering the Dimensionality of Gender Voting in Congress.‖ Legisla-

tive Studies Quarterly 24 (1): 65-86.

Palmer, Barbara and Dennis Simon. 2006. Breaking the Political Glass Ceiling: Women and Congressional

Elections. New York: Routledge, Taylor and Francis Group.

Pascal, Anthony H. and Leonard A. Rapping. 1972. ―The Economics of Racial Discrimination in

Organized Baseball.‖ In Racial Discrimination in Economic Life, edited by Anthony H. Pascal.

Pearson, Kathryn and Eric McGhee. 2007. ―Strategic Differences: The Gender Dynamics of Con-

gressional Candidacies, 1982-2002.‖ Working paper.

Persson, Torsten, Gérard Roland, and Guido Tabellini. 1997. ―Separation of Powers and Political

Accountability.‖ Quarterly Journal of Economics 112 (4): 1163-1202.

Philpot, Tasha S. and Hanes Walton, Jr. 2007. ―One of Our Own: Black Female Candidates and the

Voters Who Support Them.‖ American Journal of Political Science 51 (1): 49-62.

Poole, Keith T. and Howard Rosenthal. 1997. Congress: A Political-Economic History of Roll Call Voting.

New York: Oxford University Press.

Poole, Keith T. and Howard Rosenthal. 2005. DW-NOMINATE data: 98

Congress to 108

Congress.

Rehavi, M. Marit. 2007. ―Sex and Politics: Do Female Legislators Affect State Spending?‖ Working

Paper.

Rodgers, William, ed. 2006. Handbook on the Economics of Discrimination. Northampton, MA: Edward

Elgar Publishing.

Rosenthal, Cindy Simon. 1998. When Women Lead. New York: Oxford University Press.

Rosenwasser, Shirley M. and Norma G. Dean. 1989. ―Gender Role and Political Office: Effects of

Perceived Masculinity/Femininity of Candidate and Political Office.‖ Psychology of Women

Quarterly 13: 77-85.

Sanbonmatsu, Kira. 2006. Where Women Run: Gender & Party in the American States. Ann Arbor, MI:

University of Michigan Press.

Seabright, Paul. 1996. ―Accountability and Decentralization in Government: An Incomplete Con-

tracts Model.‖ European Economic Review 40 (1): 61-89.

Sellers, Patrick J. 1997. ―Fiscal Consistency and Federal District Spending in Congressional Elec-

tions.‖ American Journal of Political Science 41 (3): 1024-1041.

Seltzer, Richard A., Jody Newman, and M. Voorhees Leighton. 1997. Sex as a Political Variable.

Boulder, CO: Lynne Reinner.

Shepsle, Kenneth A. 1978. The Giant Jigsaw Puzzle: Democratic Committee Assignments in the Modern

House. Chicago: The University of Chicago Press.

Smith, Eric R.A.N. and Richard L. Fox. 2001. ―The Electoral Fortunes of Women Candidates for

Congress.‖ Political Research Quarterly 54 (1): 205-221.

Stein, Robert M. and Kenneth N. Bickers. 1995. Perpetuating the Pork Barrel: Policy Subsystems and Amer-

ican Democracy. New York: Cambridge University Press.

Stewart, Charles III and Jonathan Woon. 2005. Congressional Committee Assignments, 103

108

Congresses, 1993-2004: House of Representatives.

Swers, Michele L. 2002. The Difference Women Make. Chicago: University of Chicago Press.

Thistlethwaite, D., and D. Campbell. 1960. ―Regression-Discontinuity Analysis: An alternative to the

ex post facto experiment.‖ Journal of Educational Psychology 51: 309-17.

Thomas, Sue. 1991. ―The Impact of Women on State Legislative Policies.‖ The Journal of Politics 53

(4): 958-976.

Uhlaner, Carole Jean and Kay Lehman Schlozman. 1986. ―Candidate Gender and Congressional

Campaign Receipts.‖ Journal of Politics 52: 391-409.

U.S. House of Representatives. 2005a. Office of the Clerk. ―Women in Congress.‖

U.S. House of Representatives. 2005b. Office of the Clerk. ―Election Statistics, 1920-2004.‖

Volden, Craig and Alan E. Wiseman. 2010. ―Legislative Effectiveness in Congress.‖ Working paper.

Volden, Craig, Alan E. Wiseman, and Dana E. Wittmer. 2010. ―Why are Women More Effective

Lawmakers in Congress?‖ Working paper.

Zaller, John. 1998. ―Politicians as Prize Fighters: Electoral Selection and the Incumbency Advan-

tage.‖ In Politicians and Party Politics, John G. Geer, ed. Baltimore: Johns Hopkins University

Press.

Table 1: Legislator Sex and Discretionary Federal Domestic Spending

(1)

(2)

(3)

(4)

Female

0.091

0.12

0.069

(0.043)**

(0.067)*

(0.033)**

Pre-Female Trend

0.013

(0.008)

Post-Female Trend

0.03

(0.012)**

Republican

-0.001

-0.002

-0.079

0.043

(0.022)

(0.023)

(0.032)**

(0.033)

Majority

0.003

0.005

-0.042

0.025

(0.028)

(0.031)

(0.082)

Terms

-0.001

0.002

0.009

(0.003)

(0.004)

(0.008)

Margin

0.003

.000

0.004

(0.023)

(0.059)

Population

0.35

0.361

2.846

(0.730)

(0.732)

(0.604)***

State Capitol

0.062

0.121

(0.017)***

(0.007)***

Age 65 and Older

-0.033

-0.029

-0.45

(0.211)

(0.189)**

Black

0.161

0.16

0.061

(0.062)**

(0.024)**

Construction

0.015

0.014

-0.588

(0.171)

(0.173)

(0.119)***

Public School

-0.382

-0.391

-1.296

(0.299)

(0.295)

(0.273)***

Farmers

0.049

0.045

0.161

(0.029)

(0.030)

(0.053)***

Foreign Born

-0.014

-0.017

-0.103

(0.051)

(0.052)

(0.071)

Manufacturing

-0.187

-0.189

-0.402

(0.133)

(0.090)***

Median Income

-0.075

-0.371

(0.151)

(0.152)

(0.157)**

Unemployed

0.1

0.103

0.055

(0.111)

(0.121)

Armed Forces

-0.012

-0.022

(0.034)

(0.016)

Urban

0.005

0.006

-0.015

(0.006)

(0.015)

Constant

18.837

18.813

13.745

0.003

(2.327)***

(2.342)***

(1.588)***

(0.050)

Observations

9067

R-squared

0.9

0.66

0.063

Model Specification

District & year fixed

effects

District & year fixed

effects

State & year fixed

effects

First-differences

Notes: Robust standard errors clustered by state in parentheses. The dependent variable in models (1) through (3) is ln(high-variation federal out-

lays). In model (4), the dependent variable is the difference in logged outlays between two consecutive years. Model (4) includes only observa-

tions in which a mixed-sex close election (where "close" is defined as a winning vote share of less than 55 percent) results in a change in the sex

of a district’s representative (see text). Outlays are in constant 2004 dollars. Female = 1 if legislator is female. Pre- and Post-Female Trends are

linear trends for the 6 years before and 6 years after the election of a woman, respectively. Republican = 1 if legislator is Republican. Majority = 1

if legislator is a member of the House majority party when Congress and the presidency are controlled by the same party. All demographic va-

riables are transformed as natural logarithms. * significant at 10% level; ** significant at 5% level; *** significant at 1% level.

Table 2: Evidence of Sex-Based Selection

District Ideology

District &

Member Ideology

Widows

(1)

(2)

(3)

Female

0.12

0.106

(0.048)**

(0.046)**

Female * Constituent ideology

0.584

0.656

(0.205)***

(0.193)***

Member ideology

-0.426

(0.098)***

Female non-widows

0.138

(0.068)**

Widows

-0.104

(0.119)

Constant

18.817

18.764

13.814

(2.388)***

(2.409)***

(1.633)***

Observations

7404

9067

R-squared

0.89

0.66

Fixed effects

District & year

State & year

F-test: Widows = Non-widows

p = 0.054*

Notes: Models include all control variables reported in table 1. Robust standard errors clustered by state in

parentheses. The dependent variable is ln(federal outlays by congressional district by year) from high varia-

tion programs, 1984-2004. Outlays are in 2004 dollars. Female = 1 if legislator is female. Constituent ideology

is the average constituent ideology in the district as measured by Clinton (2008). Member ideology is the

legislator’s DW Nominate score. Widows = 1 if legislator is female who succeeded her late husband in office.

Female non-widows = 1 for all other female legislators.

* significant at 10% level; ** significant at 5% level; *** significant at 1% level.

Table 3: Alternative Explanations

High Variation Program Spending

Bills