Chenzi Xu ‘12
Harvard University, Department of Economics
The United States Congress has 535 members, and in 2011, only 88 of them were women. Historically, only 2% of the U.S. legislative body has been female. This paper builds on the political science and labor market literature to empirically investigate whether discriminatory behavior against female politicians in the House of Representatives partly explains this gender-gap. Specifically, it employs Gary Becker’s taste-based model for discrimination, which says that people are willing to sacrifice output in order to satisfy their preferences. In this model, the target of discrimination would have to be more productive than average in order to overcome these barriers. By proxying legislative ability with the amount of federal funding secured to representatives’ home districts, this paper assesses whether women in the House are underrepresented. This analysis finds that (1) on average female representatives are more productive than men, suggesting that the gender gap is partly explained by voter discrimination, (2) Republican female representatives outperformed their male counterparts to a greater extent than Democratic females did, meaning that Republican voters are more discriminatory, and (3) freshman female representatives suffer a greater loss to output than new male representatives, suggesting discrimination by other members of the House.
Introduction: why are so few women in Congress?
In the United States legislature, women are a clear minority in both the House of Representatives and the Senate. Historically, only 2% of the members of Congress have been women, with only 270 of them compared to the 11,743 men, and as of 2011, women hold only 16% of Congressional seats and 23% of seats in state legislatures (CAWP, 2011). Currently, there are 71 female representatives and 17 female Senators out of the 535-member Congress. The large gender disparity in the Congressional labor market could be explained by two broad categories of labor market explanations: human capital theory and discrimination theory.
In the discrimination literature, surveys have found that biases against women seem to exist in the electorate: a considerable proportion of them would hesitate to vote for a female candidate in various state and national elections (Newport & Carroll, 2007; Milyo & Schosberg, 2000). Indeed, more people are unwilling to vote for a female candidate (11%) than a black candidate (5%). This suggests discrimination in the political labor market. Yet studies of election outcomes and campaign funding have shown that female candidates are as successful as male candidates, which have led some scholars to conclude that political gender discrimination no longer exists (Burrell, 1994; Fox R., 2006).
This paper uses Federal Assistance Award Data (FAADS) that attributes discretionary federal spending dollars to the 435 different U.S. Congressional districts. These data proxy for the observable “productivity” of representatives and are used to assess the existence and nature of discrimination. This approach measures ex-post outcomes rather than ex-ante qualifications and builds on Gary Becker’s theory of discrimination that has been applied to a variety of labor markets. In particular, this paper attempts to distinguish between the co-worker and employer types of taste-based discrimination in the Congressional labor market and finds that both are plausible explanations for the gender gap in the Congressional labor market.
Theoretical background: human capital and discrimination explanations
Human capital theory argues that differences in professional outcomes among people are due to variations in their qualification levels, which may stem from health, knowledge, intelligence or a variety of other factors that are sometimes difficult to observe and quantify. However, it assumes that if all these characteristics were accurately quantifiable, they would accurately predict a person’s labor market outcomes. Discrimination theory, on the other hand, falls broadly into the two categories of statistical and taste-based: in the former, the inability to perfectly observe a person’s productivity will lead to “rules of thumb” based on inexact correlations between observable characteristics, such as gender, race, or age, and outcomes. In the latter, productivity is knowingly forsaken in order to satisfy personal biases. The rest of this discussion will describe the elements of these theories relevant to the Congressional labor market.
In the rest of this paper, I define the “labor market” as male or female politicians who are seeking election or re-election in the U.S. House of Representatives. “Employment” is being voted into office by the public, which makes the “employers” the voting members of the representative’s district. “Co-workers” are the other representatives in Congress, and especially those who serve on the same legislative committees. The “product” of interest is the federal funding the legislator secures for the district. The “wages” paid to the legislator are fixed at the standard level, and outside endorsements and wages for other jobs are not included in this analysis. It is important to note that insofar as “employees” deliver a “product” to their customers, in this Congressional labor market, the employers are effectively indistinguishable from the customers. While voters choose the politician (employment), they are then bound (as customers) to that representative until the next election cycle. This formulation is shaped by the theoretical background that follows, and it motivates the empirical methods of this paper.
Human capital theory
Human capital explanations are based on the impact of differences in ability, education, experience, training, and health on labor market outcomes among individuals. These differences themselves may arise from endowments (genetics) or investments (schooling, training). In terms of endowments, men may be more charismatic and thereby able to represent their constituents better by pushing for legislation more effectively. However, even if men and women are equally endowed with the traits that would make them successful politicians, they may invest in public service training differently based on their individual investment returns functions and therefore accumulate different amounts of human capital (Mincer, 1958). Investments such as educational paths (for instance, more men choosing law school or business school) might make them more qualified to hold Congressional office. In addition, years of experience are a significant contributor to labor market outcomes, and on average, women have fewer prior political experiences due to the family obligations of raising children. This is evidenced by the tendency for women to enter politics at a later stage in life than men. The feedback loop of expecting to spend less time in the labor market would also lead women to invest less in skills ex-ante (Altonji & Blank, 1999).
Another facet of the human capital explanation is that self-selection based on personal tastes could partly account for the decreased presence of women. Some occupations attract one particular sex because of their demands on the individual. In this case, women might on average be less willing to take on a consistent public role that requires living away from home for much of the year, or engage in the public eye of politics due to family responsibilities. Experimental evidence has also been presented that women tend to prefer occupations with more career stability and lower levels of personal and professional risk, which hardly characterizes national politics (Bertrand, 2010).
It is important to distinguish here between occupational self-selection out of the labor market for reasons due to individual tastes versus those due to perceived barriers to entry. Studies have shown that holding objective qualification measures constant, women are less likely to consider themselves qualified for office (Lawless & Fox, 2005). While personal life circumstances and psychology may partly account for their hesitation to run (for instance, women appear more concerned than men about their ability to fundraise and campaign effectively, Fowler & McClure, 1989), the majority of both men and women in the likely candidate pool believe bias against women exists in elections (Lawless & Fox, 2005). If women choose not to pursue political candidacy or drop out after early failures because of the latter reason, then the root cause of occupational self-selection is still discrimination. This leads to various theories of discrimination.
According to Heckman (1998), “discrimination is a causal effect defined by a hypothetical ceteris paribus conceptual experiment—varying [gender] but keeping all else constant”—i.e. a female politician who is otherwise identical to a male politician would be treated differently by voters and other politicians. Discrimination can be part of either or both of the labor market and pre-market conditions (Altonji & Blank, 1999). The former refers to different treatment of the sexes upon entering the political arena and can be categorized as employment, wage, or occupational discrimination. The latter describes unequal opportunities to cultivate the human capital necessary to be successful in the labor market. Both of these types of discrimination will be discussed in further detail. More generally, all of them can take on the form of statistical or taste-based discrimination, and the model for analyzing Congressional outcomes will be built upon the latter.
Types of pre-labor market and labor market discrimination
Pre-market discrimination generally refers to access to education and training, but in the case of political office, can also refer to resources that directly or indirectly support political participation. These include the extent to which party leaders encourage and support female versus male candidates, which has been shown to significantly influence the number of women who run for office (Fox & Oxley, 2003). In addition, studies about gender bias by party elites, especially in more conservative geographical areas, helps to explain the significant party gap among Congresswomen: Republicans have historically nominated about half as many female candidates as Democrats, although they have won elections at roughly equal rates (Fiber & Fox, 2005; Niven, 1998; Fox & Oxley, 2003). The differences between parties will be examined in detail later, but pre-market discrimination is not the focus of this paper.
In terms of the labor market discrimination, not all of the different types of labor market discrimination are relevant to the political market. First, wage discrimination at different levels of office is minimal because of the standardization of salaries paid to public servants. Occupational discrimination would emerge if women who would like to seek public office are instead forced into other forms of political participation despite having equal levels of talent. Finally, “employment” discrimination—or candidate success in elections—has been the focus of most studies of the political arena. While these conclude that women have been making much progress and are winning elections at record rates, there is still evidence that they face higher barriers to entry (Fiber & Fox, 2005).
Statistical vs. taste-based discrimination
Two types of discrimination have been distinguished: statistical and taste-based. The former explains why discrimination may be an outcome even among rational employers. The lack of perfect information about employee quality leads firms to rely on correlations between easily observable characteristics and productivity, in which they proxy for more precise measures with the easily observable measure (Arrow, 1973). Since relatively few people witness the legislative process and know which qualifications really matter in Congress, it is likely that statistical discrimination exists. Candidate qualifications are not usually directly comparable, and campaign platforms and promises can diverge dramatically from actual performance and policy. In addition, the “human component” to policy-making, such as charisma and likability, is undeniable, and media coverage of debates and interviews hardly capture the experience of interaction. The extent that women are perceived to be less adept at policy-making due to pre-conceived notions of gender roles and differential media coverage, and the difficulty of assessing the true value of significant characteristics will lead to statistical discrimination.
Political theory models of elections diverge on the existence of statistical discrimination. In moral hazard models, politicians are incentivized to exert effort and perform well from their desire to gain re-election. Voters cannot perfectly observe candidate qualifications ex-ante, so they choose politicians through retrospective voting and keep an incumbent in office only if he meets a threshold level of performance (Barro, 1973; Ferejohn, 1986). These theories suggest the presence of statistical discrimination that is corrected once voters receive better information. Other literature argues that elections and the campaign process are mechanisms for distinguishing future performance. Hence, they assume that human capital characteristics relevant to policy-making are apparent and that candidate performance is based on these innate qualities so that statistical discrimination would not exist (Zaller, 1998).
On the other hand, taste-based discrimination as developed by Becker occurs when customers, co-workers, or employers prefer to interact with one group or another, holding productivity constant, and would be willing to pay to do so (Becker, 1971). Unlike rational agents, they are not maximizing their choices purely for returns. Indeed, the key component of this discrimination is that the agent willingly pays to discriminate. The implications of the three types of taste-based discrimination in the Congressional market are considered below.
Customer discrimination would exist if constituents preferred a male representative over a female representative of equal qualification and ability to the extent that they would physically move away in order to be represented by a man. This possibility is not considered here because election cycles occur biennially, and the decision to change address is most likely due to a host of other, more pressing considerations. In other words, it is unlikely that the preference for males would be priced at tens or hundreds of thousands of dollars in foregone investment. Employer discrimination would occur if voters prefer being represented by a male representative and would, to a certain quality differential, choose to elect a lower quality male candidate than a higher quality female candidate. Lastly, co-worker discrimination would occur if other members of the House prefer to work with men and would sacrifice legislation-making productivity in order to avoid working with women. More than one type of taste-based discrimination can exist, and my analysis considers both of the latter types.
It is necessary to note that taste-based discrimination is often difficult to disentangle from statistical discrimination, and in the case of the Congressional labor market, can coexist with it. Due to the fact that the voters are not themselves involved in the legislative process, they may statistically discriminate if they believe that female representatives are less productive due to taste-based discrimination by other members of the House. Their expectation of a lower level of productivity, though rational, does not lead them to vote according to the candidates’ true qualities. In addition, there may be heterogeneity among voters in their reason for electing a male candidate instead of an equally-qualified female candidate, whether it is because of their own tastes or expectations of others’ tastes.
A model of taste-based discrimination
Method of accounting for outcomes
In this paper, discrimination is measured by comparing the marginal revenue products among workers (representatives) with equal wages. This method of measuring the quality of a person’s output has been used in many disparate labor markets, such as that of professional sports, to argue for the presence of discrimination when it is difficult to measure human capital resources (Pascal & Rapping, 1972). It has also been extended to the Congressional labor market with respect to female performance by Anzia and Berry (2010), which found that women consistently outperform men in terms of the federal dollars they bring to their districts. This paper extends their results by considering the specific models of discrimination.
In an efficient labor market, people will be paid according to their marginal product. Determining an accurate proxy measure for the marginal product has been based on the political science literature which has found a significant relationship between the amount of discretionary federal spending awarded to a district and the representative’s perceived success. Thus, among representatives, if one gender group receives less than their marginal revenue products in wages while another group receives full payment, then this is evidence of discrimination against the first group. Since there is no variation in wages among representatives, in the absence of discrimination, there should not be significant differences in their output. Assuming that taste-based discrimination exists, if p is the monetary value of a unit of output produced by this group, then this output is valued at p (1 – d) where d is the discrimination coefficient. The larger the coefficient, the more productivity is sacrificed in order to discriminate. The presence of gender discrimination in Congress is tested empirically using federal outlays data.
Taste-based employer discrimination
The total profits attained by a district are equal to the amount of federal outlays (total revenues) it receives minus the costs of procurement (total costs). Total revenues in a fiscal year is modeled as a function of the quality of the representative (the “price” he/she can charge) and a number of district-level characteristics (the “quantity” of funding needed). Total costs would be the wages paid to the representative.
π = p(q)*d – w
π indicates profits, total revenues is represented by p, and the amount of funding is a function of q (the representative’s quality) scaled to a vector of district characteristics d. The cost, w, is the wage paid to the representative.
Candidate quality is proxied by profit, so if voters care only for candidate quality, then they will vote as if according to π. When confronted with two candidates, a male and female who will secure πM and πF respectively, they will compare these pure profits. If πM > πF, they will vote for the male; if πM < πF, they will vote for the female. Assume that the quality of the candidates is the same: qM = qF = q. The vector of district characteristics d does not change for a voter, and w is fixed by law. Given this set-up and these assumptions, discrimination by voters is modeled as follows:
Assuming that voters derive utility not only from the federal spending in their district, but also from the gender of their representative, then they act as if female representatives cost more. Let w be the real wage and w’ be the taste-adjusted-wage. Then w’M = wM and w’F = wF (1 + de) where de is the coefficient of discrimination.
The voters act as if they receive profits of the form:
π’ = p(q)*d – w’
Given that wM = wF = w, then the previous equation simplifies into:
π’M– π’F = w*de
Therefore, the expected difference in performance between male and female representatives is proportional to the amount of discrimination felt by the voters. This model is assessed empirically in Part A of Section V.
Taste-based co-worker discrimination
As in the previous case, the total profit attained by a district is equal to output scaled by district characteristics minus wages:
π = p(q)*d – w
Co-worker discrimination would exist if male representatives preferred not to work with female representatives and thereby provided less support for their female colleagues than their male ones. This lack of support would manifest itself in reduced output by female representatives: pF(qF) < pM(qM). Assume that the quality of male and female representatives is the same: qM = qF = q, and as before the vector of district characteristics d does not change, and w remains fixed by law.
In the firm, co-worker discrimination would be compensated by paying additional wages to the discriminating group in order to induce them to work with the minority group. However, as wages in Congress are fixed, men cannot charge more for working with women. Instead, since legislation requires consensus, male representatives can express their dissatisfaction by lowering the output of female representatives. Let p(qF) be the output without discrimination and p(qF)’ be the depressed output. Then p(qF)’ = (1 – de)*p(qM) where de is the coefficient of discrimination.
The total profits by women in the presence of discrimination will therefore be less than profits without discrimination, where .
π’F = p(qF)’*d – w
Substituting from above for p(qF)’ and qM = qF = q gives:
π’F=(1 – dε)*p(q)*d – w
The difference in a female representative’s total profits in the presence of co-worker discrimination is thus:
∆πF= πF– π’F=(dε – 1)*p(q)*d
Notice that de is scaled so that 0 ≤ de ≤ 1 , which means when de > 0. In other words, the female representative’s output will be lower than a comparable male representative’s in the presence of co-worker discrimination. This model is assessed empirically in Part C of Section V.
Empirics: testing for discrimination from federal funding
The empirical analysis is based on the theory of comparing marginal productivity of equally paid Congressmen and Congresswomen. The mandate for a Congressman is to represent his constituents in a national setting, and one of his explicit goals is to bring federal dollars to the district. While this “pork” legislation has often been derided as a source of inefficiency in the government, there is evidence that doing so bolsters the representative’s chances of reelection (Ferejohn, 1986; Levitt & Snyder, 1995). In addition, the level of government projects in a district is more easily observed by constituents than the outcomes of the legislative process on Capitol Hill. Hence, as in Anzia-Berry’s analysis of Congressional outcomes, the underlying assumption is that Congressmen are elected in order to produce funding for their constituents. Therefore, their ability to do so will reveal their human capital qualifications to build coalitions, set agendas, and make deals that will benefit their districts.
Federal outlays and Congressman characteristics data
The Federal Assistance Award Data System (FAADS) captures and records over one thousand programs that disperse funds throughout the country. These include long-term spending such as Social Security and Medicare benefits as well as short-term grants for research and education purposes. Bickers and Stein (1991) present a methodology for compiling and studying FAADS data. FAADS records program spending in various ways that does not always provide a clear district allocation. Their procedure provides a way to estimate allocations to certain districts. Levin & Snyder (1995) and Anzia & Berry (2010) also utilize their method for assigning federal outlays to Congressional districts. The details concerning their data compilation, as well as observations about the reliability of the U.S. Census Bureau’s coding techniques can be found in Appendix A.
The FAADS data for this analysis include the years 1984-2004, adjusted to 2004 dollars. Average outlays per district over the entire sample of 9,135 observations are $236 million with ranges spanning $398 million to $753 million. One important characteristic about FAADS data that Levin & Snyder (1995) accounted for in their analysis is that different programs display different amounts of spending variation. The largest outlays come from entitlement programs which apportion very similar amounts of funds to districts across time periods. On the other hand, some programs are much more discretionary in nature, and Congressmen have much more influence over their allocation. Hence, a more accurate measure of a representative’s productivity is not the total amount of outlays per district, but the outlays that stem from discretionary spending. Further details about the variations in spending are described in Appendix B.
Since the federal budget is determined the year prior to outlays, the spending allocation to a particular district is matched to the representative from the year prior. Hence, the representatives during this period come from the 98th through 108th U.S. Congresses, which span the years 1983-2005. Key characteristics of the data are summarized in Table A.
Two measures for the outlays assigned to a district are used in a fixed effects linear regression model in order to answer the questions of whether females are more productive than their male counterparts, as proxied by their ability to bring profits to their constituents. First, the log-transformed estimate of the high-variation federal program spending is used to estimate whether having a female representative significantly increases a district’s allocations. Secondly, the rate of growth in a district’s federal outlays is used to assess how representative characteristics influence the trend of productivity while in office. This value is constructed by differencing log-outcomes for each representative, and so the sample size decreases by one year for each unique representative.
The first set of data—the natural log of annual district outlays—is very close to being normally distributed and is much more tractable to work with than direct measures:
Models and results
The basic model of gender significance
The first basic model using outlays data is as follows:
ln(outlaysit) = α0+ β1 (Femaleit)+ β2*Rit+ β3*Dit+ δt+θi+ єit
The second basic model using the growth rate of outlays is:
α0+ β1 (Femaleit)+ β2*Rit+ β3*Dit+ δt+θi+ єit
The key coefficient of interest is β1 of Femaleit, which is an indicator variable for whether the representative of district i in time t was female. Ritis a vector of representative characteristics that includes the number of terms he/she has served, the Committees he/she is part of, and whether he/she chairs any Committees or holds a party leadership role in the House. These variables control for factors that may affect a representative’s effectiveness in Congress. The vector Dit contains a number of district-level characteristics which may affect the amount of funding they receive. For instance, smaller districts tend to receive more funding allocations on average than larger ones due to representation in the Senate. In addition to population itself, the composition of population within a district influences the type of projects that might be directed toward it. Therefore, elements such as median income, the percentage of African-American population, and the unemployed population are included. Another key indicator included in the vector of district characteristics is whether the district contains the state capital. Many federal programs do not directly disperse funds but rather reroute it through the state government. Therefore, any such district has significantly higher outlays receipts.
The coefficient θi captures the time-invariant characteristics of the district that would systematically influence either the probability of having a female representative or the amount of federal outlays it receives. While the district fixed-effects model is more precise, a state fixed-effects model is also shown in order to allow for more within-state variation while accounting for broader geographic trends. The year indicators δt are included to account for time-trends, which include general spending increases over the 21 -year period covered by the data.
The key results on the importance of being female can be seen graphically below. The peak of the distribution outlays for district with a female representative follow a similar distribution but at higher levels:
The regression results confirm these findings and are listed as Table 1 in Appendix C. The coefficients indicate that having a female representative increases a district’s annual outlays by around 10%. These results are significant at the 5% and 10% levels when clustering by district and state fixed effects, respectively.
Augmenting the basic model with party differences to show voter discrimination
The earlier discussion of discrimination earlier mentioned that Democrats are a much larger proportion of the female Congressional labor force than Republicans. While they win elections at similar rates, many more Democratic women run for office than Republican women, and one explanation for these differences is that Republicans are more conservative and hold to traditional gender roles to a greater extent than Democrats (Niven, 1998). Given that Congressmen tend to garner the votes of their party’s constituents and are less likely to appeal across gender lines, the lower female Republican entry rates into the political labor market suggest that they face higher barriers to entry than Democratic females. If that were the case, Republican women would be of even higher quality relative to Republican men than both Democratic women are to Democratic men and of women in general compared to men.
The graphical evidence for this phenomenon is suggested below:
In the first graph of outlays data, there are fewer Republican women than Democratic women, and the peak of the female Republican distribution is at a higher value. This result is particularly compelling given the ideological differences between Republicans and Democrats, which would suggest lower levels of outlays in Republican districts. Indeed, this appears to be the case. There are more Democrats in the sample than Republicans, but the mean of their distribution is visibly farther to the left.
Hence, the inference that Republican women would outperform Democratic women due to the higher barriers to entry into Congressional office is supported by both discrimination and political economy models. The implication for discrimination theory is that Republican women face higher “employer,” or voter discrimination.
This model thus accounts for the additional party-based voter discrimination by introducing a Party interaction term:
The outlays model changes accordingly as follows: in which the β1, β2, and β3 coefficients are of particular interest:
ln(outlaysit)= α0+ β1 (Femaleit)+ β2 (Republicanit)+
β3 (Femaleit )*(Republicanit)+ β4*Rit+ β5*Dit+ δt+θi+ єit
The second formulation with the growth rate of outlays indicates how gender and particular party affiliation affect the performance trend during a representative’s tenure.
ln(outlaysit/ outlaysit-1)= α0+ β1 (Femaleit)+ β2 (Republicanit)+
β3 (Femaleit )*(Republicanit)+ β4*Rit+ β5*Dit+ δt+θi+ єit
The regression results are given in Table 2 of Appendix C. Although the district fixed effects models are not significant in this case, they do support the direction of the effect that is seen in the state fixed effects models. The direct outlays data reaffirm what was shown earlier that being represented by a female significantly increases a district’s outlays, and it also confirms the hypothesis that being represented by a Republican decreases them. While the nonlinear effect of being both female and Republican is not significant the outlays model, it is positive as predicted.
The specification using data on outlays growth does find significance in the interaction term that is positive as predicted. In addition, while the Republican term is negative, the effect of being a female Republican clearly dominates it. Interestingly, the growth trends for female representatives is negative, which suggests that while women do bring in significantly more funds than men, the rate at which they do so declines over their tenure in office. Although this result does not directly bolster the discrimination model, it does not detract from it. Many logical explanations exist, including the fact that attracting federal dollars is an imperfect proxy for legislative quality, and a longer tenure in office affords Congressmen other positions and responsibilities. While they substitute their efforts away from securing funding for their constituents, they substitute into other less easily observable functions of legislation, such as leadership and bill sponsorship.
The implications of freshman status on co-worker discrimination
Having considered ex-ante voter discrimination by the electorate and shown that women, and particularly Republican women, bring in higher profits to their districts, it is now time to consider the possibility of co-worker discrimination as modeled by the effect of being a freshman Congressman. The large incumbency advantage in House elections is a well-known phenomenon which suggests that any representative who is voted out of office would be replaced by someone of higher quality. A graphical analysis of this hypothesis does not appear to confirm it:
The distribution of outlays follows almost an identical pattern between old and new representatives. However, this phenomenon may be due to the disadvantages that the freshmen face, in which any human capital advantage they have over their predecessor in the form of innate quality or ability is balanced by their relative dearth of accumulated training an experience. They do not have the social capital and legislative practice that would bolster their performance in Congress beyond that of their colleagues. Separating the sample between men and women also does not render clearly distinguishable differences, as they both seem to follow the pattern of their gender cohorts.
Yet having already established that women are more successful than men at producing funds for their districts, it is likely that freshmen females outperform everyone else. Strong graphical evidence of this hypothesis is shown in the second graph above in which their mean performance is almost one log point higher. An empirical analysis of this question an indicator for freshman status:
ln(outlaysit) = α0+ β1 (Femaleit)+ β2 (Freshmanit)+
β3 (Femaleit)*(Freshmanit) + β4*Rit+ β5*Dit+ δt+θi+ єit
The parallel analysis based on growth trends would not help to answer this question since by definition, freshman status and its labor market implications due to inexperience diminish with time and indeed no longer holds within two years. Therefore it is unlikely to have an impact on the trends in spending procurement. The results of the empirical analysis are presented in Table 3 of Appendix C. The “freshman effect” is negative, although not significant, as predicted and indicated by the graphical analysis, and the female outperformance endures. However, the combined effect of being both freshman and female is negative, indicating that women face more difficulty upon entering the Congressional labor market than men do. Freshman female outperform every other group so the female quality advantage dominates, but the significance of this interaction suggests the presence of co-worker discrimination. In particular, it indicates the unequal treatment of women in a cohort of relatively comparable market experience.
This paper has used a price theory structure to argue that discrimination may be the cause of some of the considerable gender gap among legislators in the U.S. House of Representatives. It bases these conclusions on widely accepted models of discrimination, and in particular, it goes beyond previous work on this question by attributing the particular combination of outcomes to certain types of taste-based labor market discrimination. Although the proxy variable used to measure a representative’s human capital qualifications has its advantages in the readily available and easily quantifiable nature of the data, it also has several drawbacks. The first lies in the other responsibilities that representatives must address in addition to securing monetary funds for their district. As hypothesized with respect to the result concerning the trend of decreasing output for female-led districts, the measurement of funding secured can only partly account for certain trends. More broadly, while the legislative labor market in the United States is relatively unique, the models, methods, and insights are broadly applicable and can be applied to a variety of questions that suffer from the difficulty of measuring human capital.
I would like to thank Professors Claudia Goldin and Larry Katz of the Harvard Economics Department for inspiring and supporting this paper, originally written for their class “History and Human Capital,” offered in the spring of 2011. I would also like to thank Professor Chris Berry of the University of Chicago’s government department and Sarah Anzia of Stanford’s political science department for sharing their federal outlays data with me. In addition, I am grateful to Lawrence Li and Gerardo Flores for their feedback, support, and humor during the writing process.
Altonji, J. G., & Blank, R. M. (1999). Race and Gender in the Labor Market. In O. Ashenfelter, & D. Card, Handbook of Labor Economics Vol. 3h (pp. 3143-3213). Elsevier Science B.V.
Anzia, S., & Berry, C. (2010). The Jackie (and Jill) Robinson Effect: Why Do Congresswomen Outperform Congressmen. working paper .
Arrow, K. J. (1973). The Theory of Discrimination. In O. edds Ashenfelter, & A. Rees, Discrimination in Labor Markets (pp. 3-33). Princeton: Princeton University Press.
Barro, R. J. (1973). The Control of Politicians: An Economic Model. Public Choice (14) , 19-42.
Becker, G. (1971). The Economics of Discrimination. Chicago: The University of Chicago Press.
Bertrand, M. (2010). New Perspectives on Gender. Handbook of Labor Economics vol. 4B , 1545-1592.
Bickers, K. N., & Stein, R. M. (1991). Federal Domestic Outlays: 1983-1990. New York: M.E. Sharpe.
Burrell, B. C. (1994). A Woman’s Place is in the House: Campaigning for Congress in the Feminist Era. Ann Arbor: University of Michigan Press.
CAWP. (2011). Fact Sheet: Women in the U.S. Congress 1917-2011. Center for American Women and Politics.
Ferejohn, J. A. (1986). Incumbent Performance and Electoral Control. Public Choice (50) , 5-25.
Fiber, P., & Fox, R. L. (2005). Tougher Road for Women? Assessing the Role of Gender in Congressional Elections. In J. J. Josephson, & S. Tolleson-Rinehart, Gender and American Politics: Women, Men and the Political Process, 2nd. ed (pp. 64-81). New York: M.E. Sharpe.
Fowler, L. L., & McClure, R. (1989). Political Ambition. New Haven: Yale University Press.
Fox, R. (2006). Congressional Elections: Where are We on the Road to Gender Parity? In S. J. Carroll, & R. L. Fox, Gender and Elections (pp. 97-116). New York: Cambridge University Press.
Fox, R. L., & Oxley, Z. (2003). Gender Stereotyping in State Executive Elections: Candidate Selection and Success. Journal of Politics 65 , 833-850.
Lawless, J. L., & Fox, R. L. (2005). It Takes a Candidate: Why Women Don’t Run for Office. New York: Cambridge University Press.
Levitt, S. D., & Snyder, J. M. (1995). The Impact of Federal Spending on House Election Outcomes. The Journal of Political economy 105 (1) , 30-53.
Milyo, J., & Schosberg, S. (2000). Gender Bias and Selection Bias in House Elections. Public Choice , 41-59.
Mincer, J. (1958). Investment in Human Capital and Personal Income Distribution. The Journal of Political Economy 66 , 281-302.
Newport, F., & Carroll, J. (2007). Analysis: Impact of Personal Characteristics on Candidate Support. Gallup News Service .
Niven, D. (1998). Party Elites and Women Candidates: The Shape of Bias. Women and Politics 19 , 57-80.
Pascal, A. H., & Rapping, L. A. (1972). The Economics of Racial Discrimination in Organized Baseball. In A. H. Pascal, Racial Discrimination in Economic Life. Lexington.
Zaller, J. (1998). Politicians as Prize Fighters: Electoral Selection and the Incumbencydvantage. In J. G. Geer, Politicians and Party Politics. Baltimore: Johns Hopkins University Press.
 However, political office usually allows and often leads to higher-paying opportunities in the private sector, and Congressional measures of wealth are widespread and disparate.
Arguably, politicians also collect another “wage” in the form of reelection. However, the incumbency advantage is so strong among representatives and the relative dearth of Congresswomen until the last two decades makes it difficult to measure the impact of quality on tenure.