[Insert Figures 3.3 & 3.4 About Here]

where all terms are previously defined. Similarly, then, Figure 3.4 depicts (3.8) as a measure of colleague valuation among out-group members. This equation behaves symmetrically in relation to (3.7). The Technical Appendix shows not only that marginal utility decreases as PD increases for all m, but also that decreases in utility from PD are more dramatic when m is further from the inflection point, m*. Figure 3.4 shows a graphical portrait of the effect of preference divergence on the marginal utility a member of Group B receives from a fellow Group B member. When Group B is very small, token Group B members do not value each other highly. This devaluation is exacerbated when preference divergence is high. Yet when Group B is very large, a Group B member receives very little benefit from a colleague’s status as a Group B member. The only possible benefit, then, is preference proximity. Absent that, utility falls dramatically. Therefore, as preference divergence grows between any pair of members, all members’ marginal utility valuation of colleagues, both in the majority and the minority group, drops at a steeper rate as the degree of minority group tokenism increases, thus implying the following hypothesis:

Descriptive representation is an essential aspect of democratic theory. Political institutions with sufficient descriptive representation reflect an accurate portrayal of the plurality of interests within a given polity. Hence, descriptive representation leads to an increased public perception that institutions are legitimate (Mansbridge 1999; Mill 1861: Chapter 3; Phillips 1991; Pitkin 1967). The legitimizing function of descriptive representation, however, necessitates the equitable treatment of underrepresented or minority groups once they enter the political institution. Clearly, if under-represented or minority groups face serious discrimination in the institutions they seek to legitimize and thus cannot fully contribute to lawmaking, institutional legitimacy is merely a chimera.

Past research on representative institutions has been rooted in Rosabeth Moss Kanter’s (1977) foundational tokenism logic that is based on two groups – a majority group and a minority group (Heath, Schwindt-Bayer and Taylor-Robinson 2005; Kathlene 1994). As noted in the previous chapter on the partisan implications of tokenism behavior in the U.S. House of Representatives, past studies focus on the institution (i.e., chamber), rather than the organization (i.e. party) as the unit of analysis at the group level. This distinction is of critical importance in many elected assemblies, including the U.S. House of Representatives, since it is partisan members and caucuses that are largely responsible for determining both the treatment and status of majority and minority group members – it is not conferred by the opposition party or individual members in a legislative institution. Therefore, analysis of tokenism logic must center on political parties as the unit of analysis, and hence, should not aggregate members across parties within the chamber. Furthermore, these extant studies do not fully explore the logical consequences of tokenism theory for understanding colleague valuation in representative institutions. Kanter’s logic, for instance, examines only the consequences of minority group size, without considering how the preferences or values of organizational members may color how they esteem one another as colleagues. Our aim has been to extend Kanter’s logic of tokenism applied to the theoretical study of political organizations by examining how minority groups receive different sanctions in relation to ideological-based preference divergence among same group (party) members. The major testable implication derived from the theory is that increases (decreases) in minority group size attenuate (exacerbate) the negative impact preference divergence exerts on individual-level colleague valuation decisions.

The advantages associated with our unified theory of colleague valuation in political organizations are threefold. First, an individual-level theory of organizational tokenism that departs from the aggregate-level focus of past accounts is offered. This departure is critical since relationships among members of the organization are ultimately rooted in how individuals treat one another. Second, our theory allows us to examine the conditions under which tokenism logic exerts varying effects on colleague valuation decisions through the mechanism of ideological-based preference divergence. This is critical to understanding how member diversity resulting from descriptive characteristics (e.g., gender, race) has distinctive consequences from those that can be attributed to diversity arising from different perspectives (i.e., preference divergence). Finally, our theory considers not only how men treat women colleagues, but also how men treat one another and how women treat both men and women members, respectively. This is an important theoretical innovation since past studies applying tokenism logic to understand issues of representation focus exclusively on how majority group members treat minority group members. Instead, our unified theory of colleague valuation explicitly incorporates all potential majority-minority group member combinations that comprise a functioning political organization, and do so at the individual level, thus allowing us to ascertain different patterns from different evaluators.

The next chapter empirically tests the pair of theoretical propositions derived from our model by using data on member-to-member leadership PAC contributions in the U.S. House of Representatives for the 105^th-108^th Congresses. As shown in the previous chapter (Chapter 2), this recent historical period reflects the greatest overall combined variation in the proportion of women members of the House over the past forty years. Nonetheless, the variation in this period is still relatively modest in absolute terms. Furthermore, Republicans as the majority party for the entire sample period (105^th-108^th Congresses) do not necessarily have to value women more highly in order to broaden their appeal to voters. Furthermore, Republicans were also highly ideologically unified during our sample period, which creates a relatively conservative test because little ideological variance results in little ideological variance to explain. For both reasons, this provides a conservative empirical test of the unified theory of colleague valuation in political organizations. Attention now turns to this empirical analysis.

APPENDIX TO CHAPTER 3

Proofs of How Member’s Marginal Utility Declines in Preference Divergence

I. Between-Group Valuation Case

First, Group A’s marginal utility decreases as PD increases. In other

words, when , then the marginal utility derived from colleague p is greater than the marginal utility derived from colleague q. One can express this inequality as: (A3.1)

where the expression in (A3.1) comes directly from (3.7). Multiplying the PD through, cancelling like terms and rearranging yields the following expression:

Cancelling like terms again and multiplying through by -1yields the following inequality:

which is true by assumption. Therefore, (A3.3) directly implies (A3.1), meaning that the marginal utility Group A members derive from Group B members decreases when preference divergence increases for all w.

Next is shown that when w is further from w*, decreases in utility from preference divergence are greater. In other words, suppose that there are two values for w, and , where and therefore, is closer to w* than is. Also, marginal utility is greater at than at for all values of PD. Showing this is true implies by symmetry that the same is true for values greater than w*. Substituting and into (7), one can show that implies that the following is true:

(A3.4)

Both rearranging and cancelling terms yields:

which is necessarily true if and are both true. The first expression is true by assumption and the second reduces to the first by cancelling like terms, so it is therefore also true by assumption.

A numerical illustration highlights the theoretical relationships among preference divergence, conditioned by gender group size, and colleague valuation decisions. Let us consider two values of w, where w^H = 0.2 and w^L = 0.1 and parameter values are those depicted in footnote 3. When Group B’s size is 0.2, Group A members’ valuation of a Group B member exhibiting zero preference divergence is 0 (PD = 0, or both members agree perfectly) is 0.75. When preference divergence between group members increases to 0.5, Group A’s valuation decreases to -0.125. When PD = 1, Group A’s valuation decreases to -1. But consider what happens when Group B’s size is 0.1. When PD = 0, the Group A’s valuation is 1.29, but when PD = 0.5, Group A’s valuation falls to 0.145. Note that an increase in PD increases from 0 to 0.5 represents a decline in utility of 0.875 when Group B’s size is 0.2, but the same increase in PD represents a decline in utility of 1.145 when Group B’s size is 0.1 – the latter representing a larger precipitous decline in marginal utility for Group A members.
II. Within-Group Valuation Case

Again, analysis begins by showing that when, then the marginal utility derived from colleague p is greater than the marginal utility derived from colleague q. One can express this inequality as:

where (A3.6) comes directly from (3.8). Multiplying through by PD, cancelling like terms and rearranging yields the following expression:

Cancelling like terms again and multiplying through by -1 yields the following inequality:

which is true by assumption. Therefore, (A3.8) directly implies (A3.6) is true, thus Group B members’ marginal utility decreases when preference divergence increases for all m.

Next, it is shown that when w is further from m*, decreases in utility from
preference divergence are greater. In other words, suppose that there are two values for w,
and, where , and therefore, is closer to

m* than is. It now must be shown that marginal utility is greater at than at for all values of PD. Showing this implies that the same is true for values greater than w* by symmetry. Substituting and into (3.8), one can show that implies that the following is true:

Both rearranging and cancelling terms yields:

which is necessarily true when and are both true. The first expression is true by assumption and the second reduces to the first by cancelling like terms, so thus it is also true by assumption. Furthermore, these decreases in Group B members’ utility nearly exactly mirror the decreases depicted for Group A members previously noted. Consider, for example, the more precipitous declines in utility when Group B’s size is 0.2 versus when it is 0.1. At 0.2, Group B’s utility when PD = 0 is 1.25, and 0.875 when PD = 0.5, for a decrease of 0.375, just as with Group A’s valuation decisions. Similarly, when Group B’s size is 0.1, this group’s members utility when PD = 0 is 0.71, and 0.06 when PD = 0.5, for a decrease of 0.65.

Note: Consult equations 3.1 and 3.2 for derivations of the utility and marginal utility calculations, respectively.

Note: Consult equations 3.4 and 3.5 for derivations of the utility and marginal utility calculations, respectively.

“It is important to recall that in 1992, Senator Carol Moseley-Braun was not in the U.S. Senate. She came here in 1993 with me, Senator Murray, and Senator Feinstein, and we joined Senator Mikulski. Senator Hutchison was not here, either. So I do not know what would have happened in 1992. … I do not know, but I will tell you one thing. It is 1994, and here we are, and we are making this an issue with good reason.” – Senator Barbara Boxer (D-CA), on the floor of the U.S. Senate during debate on the retirement of Navy Tailhook scandal figure Frank Kelso.

The “Year of the Woman” stories detailed above represent real-life illustrations of how an increase in the number of women in Congress raises the level of threat men legislators feel from their women colleagues. As the ranks of women grow in Congress, men legislators feel their own foothold on power weaken, thus leaving them ready to lash out and attempt to minimize the power their women colleagues have, by calling them names and demeaning their ability to participate in governing. Furthermore, three of the four stories documented here represent women expressing views different from those of the men who attacked them. This preference divergence, then, merely adds to the ire and discontent the men feel.

On the other hand, these stories may be outliers. Representatives Wilson and Stark, for example, are both particularly well-known in Washington for their off-the-cuff, often impolitic remarks. However, this ‘outlier’ explanation is not borne out in committee chair assignment patterns: There were only two women committee chairs in the U.S. House during this period. In the 103^rd Congress (1993-1994), for example, no Democratic woman held a full committee chair. In the 104^th Congress (1995-1996), two Republican women, Nancy L. Johnson (R-CT) and Jan L. Meyers (R-KS), held committee chairs.²² But again, this fact can be attributed to the dearth of committee chair slots, rather than to the lack of respect or esteem for women legislators. Furthermore, more women in plum assignments might indicate esteem from women colleagues, men colleagues, or both, but it is impossible to discern which, if any, simply by looking at whether or not a legislator received such an assignment. Did no woman in the 103^rd Congress receive a chair because they were not valued by their colleagues? Or was it simply that no chair positions were available? Or was it that seniority norms dictated that chair positions must go to someone else? Did Johnson and Meyers receive their chairs due to the value their fellow legislators placed on them, or despite it?

The current chapter seeks to answer these questions in a systematic manner by providing an empirical test of the theory we advanced in Chapter 3. Considering the U.S. House of Representatives allows us to analyze the theory in an institution with both a great deal of variation in the percentage of women in the legislature and a means of glimpsing members’ latent private valuations of each other, using data gleaned from patterns of contributions from member-to-member leadership PACs. If tokenism theory is indeed valid, we should expect to see men valuing women less and women valuing women more as the proportion of women increases. Furthermore, we should also observe legislators devaluing their colleagues when those colleagues are ideologically further from them, an effect that is exacerbated as the size of the minority decreases. More specifically, the pair of testable predictions we derived from the unified theory of colleague valuation in political organizations is as follows:

These hypotheses have direct implications for the link between descriptive and substantive representation. If men truly devalue women as women’s numbers increase, it would imply that the descriptive-substantive link is frayed. This is because these devaluations may lead to women having less, rather than more, power as their numbers increase. At the same time, though, if women truly do value one another more as their numbers increase, this could indicate that the link between descriptive and substantive representation is intact and healthy. As women’s numbers increase, so, too, does their ability to join forces and increase their own institutional voice vis-à-vis men. Last, the effect of preference divergence is also key to the link between descriptive and substantive representation. Considering H3.1 in isolation may lead practitioners to think, wrongly, that they ought to decrease the proportion of women in the legislature in order to assure that women’s voices are best heard, since smaller numbers mean greater consideration from their majority men colleagues. But H3.2 reveals why this prescription is misguided. Since preference divergence matters – and matters most when the minority group is at its smallest – tokens lose their high valuation should they attempt to use their voices to represent issues important to women. This is because doing so would reveal differences of opinion between the valuator and the valuatee that would prompt a decrease in valuation. Tokens can have power, but only under the proviso that they never actually use that power to strengthen the link between descriptive and substantive representation.

To test our unified theory of colleague valuation, we analyze gender-based colleague valuation decisions in the U.S. House of Representatives. Focusing on gender is an appropriate venue for our test because women comprise roughly 50 percent of the population, yet remain minorities in virtually every legislature in the world.²³ In fact, women have held less than 15 percent of the seats in the U.S. House of Representatives over the past decade. Further, studies of the U.S. House of Representatives have shown that the ability of women to work together is central to successful policy enactment (Bratton 2002; Thomas 1991). Although the theory is generalizable to other minority groups, the status of women as a minority group is special because it is both large enough and provides sufficient variation that is conducive to valid statistical analysis.²⁴ Similarly, focusing on the U.S. House allows us to take advantage not only of variation in both the proportion of women across parties and across time, but also of measurable variation in preferences, even among legislators of the same political party.

The U.S. House of Representatives is viewed as “the people’s body,” more closely tied to their constituents than the Senate through both more frequent elections and smaller (in most cases) constituencies. Furthermore, the U.S. House is elected directly by eligible citizens that are represented in single-member districts in which voters select their one most preferred candidate. Voters, therefore, know the personality, the policies, and the demographic characteristics of the individual candidate for whom they are voting. In this way, voters can choose a woman or a minority candidate in a way that is much more direct than that seen in most party list systems. For these reasons, the U.S. House is the ideal empirical venue to test theories relating to the link between descriptive and substantive representation, since the link between legislator and constituent is more direct and personal than that of many other electoral systems.

The U.S. House is also ideal for the current application because it is likely the single most studied legislative institution in the world, thus providing us with a solid base of knowledge that can inform testing the unified theory of colleague valuation in political organizations. For example, how parties work in the House (e.g. Cox and McCubbins 1993) informs our understanding of inter-party gender relations. Because we know and can quantify the individual ideological voting behavior of each Member of Congress (Poole and Rosenthal 1997), this allows for providing a direct test of how ideological proximity affects colleague valuation decisions within the context of testing this theory. Because the extant literature on how congressional elections and campaign finance (e.g. Jacobson 2008), as well as leadership PACs (e.g. Cann 2008) operate, is well developed, we have a solid roadmap for properly controlling for electoral variables external to our theoretical model. Furthermore, we have specific information about member valuations through leadership PAC contributions, information that is available to us only because laws passed for wholly different reasons mandate that all campaign contributions to federal offices be made public.

In the case of gender groups in the U.S House of Representatives, valuation of women House members by men and women colleagues is simply a function of the percentage of minority (women) members of the political party of the woman being valuated. We take the political party as the organizational unit of analysis, reflecting that most work in the U.S. Congress takes place within, rather than across, parties (e.g., Cox and McCubbins 1993; Poole and Rosenthal 1997). Moreover, the parties provide variation in their respective proportion of women members during the period under consideration, from a high of 20 percent for Democrats in the 107^th Congress, to a low of 8 percent for Republicans in the 105^th and 106^th Congresses. Because clear partisan differences exist in the proportion of women members of the House, we should expect to observe clear differences in House members’ colleague valuation decisions.