Representation and Legislative Roles in Congress: Evidence from the 1950s

Abstract

Classic and contemporary research assumes legislators subjectively adopt a role orientation. This role indicates the degree to which they attempt to provide delegate representation for their constituents with the policies they support. Existing research, however, is divided on the causes that influence the choice between a relatively delegate or trustee role. I hypothesize that district heterogeneity affects this choice. Using data from the American Representation Study in the 1950s, I assess how a U.S. representative’s selection of a delegate or trustee role is affected by district heterogeneity. The findings indicate that greater ideological heterogeneity generally increases the likelihood of delegate representation, depending on a legislator’s electoral marginality. Greater partisan heterogeneity usually decreases the likelihood of delegate representation, except among electorally marginal legislators. I also find that marginality, seniority, and party affect legislators’ role choices, providing evidence that in some ways Congress has remained the same over time.

Disclosure statement

No potential conflict of interest is reported by the author.

Notes

1 Massa retired early in March of 2010, and did not serve the remainder of his term. In November of 2010, Reed won the election to replace Massa.

2 I use the term "district" here to indicate a legislator’s geographic constituency, the people contained within the legal boundaries of the locality he or she represents. Focusing on geographic constituency avoids controversy. Scholars have identified many kinds of constituencies (see Fenno 1977, Bishin 2000, etc.), but everyone can agree on what constitutes a legislator’s geographic constituency.

3 Heterogeneity may affect behavior in other ways too; see Ashworth & de Mesquita (2006).

4 Kingdon (1989) found that legislators can sometimes contradict the desires of constituents like this.

5 Canes-Wrone, Herron, and Shotts (2001) discuss this form of representation for presidents, as does Bianco (1994) for House members on the Ways and Means Committee.

6 Implausible as this may seem, it does occur. U.S. Senator Joe Manchin (D-WV) is a great example. Manchin, a moderate Democrat, has represented West Virginia in the Senate since 2010, despite the fact that West Virginia has voted for the Republican candidate for president in the last 3 presidential elections by 25% or more. One example from the U.S. House of Representatives was Minnesota’s 7th District, a rural, fairly conservative district that voted for the Republican presidential candidate in the last 6 presidential elections and voted for Republican Donald Trump in 2016 and 2020 by an average of 30 percentage points. From 1991-2021, it was represented by moderate Democrat Collin Peterson.

7 This article chiefly considers what is often referred to "policy responsiveness" (Eulau and Karps 1977, Griffin and Flavin 2011) or "policy congruence" (Harden 2016), or the degree to which legislators attempt to pass policies consistent with what their constituents want, due to the unique look that the American Representation Study provides at representatives’ self-described representation style. Griffin and Flavin (2011) note that across demographics, a majority of constituents rank "policy representation as the most important aspect of representation," even though "policy responsiveness cannot provide a complete explanation of representation" (Harden 2016).

8 This idea, true in general, may not hold for specific issues. For example, one can easily imagine a usually heterogeneous district that holds a large military base having very homogeneous preferences about the future and funding of that base. Its representative’s decisions regarding this base will be salient and most likely satisfy Arnold’s (1990) requirements for traceability.

9 Fenno (1977, 913) describes a similar argument.

10 Assuming, of course, that the legislator believes constituents can be persuaded to care (see Fiorina 1974). A legislator’s ability to shape constituent opinion is also dependent on the kind of issue in question, which can affect the direction of representational linkages (Hurley and Hill 2003).

11 Other factors contribute to legislator uncertainty about constituent preferences as well, including the arrival of new voters in a district (Bertelli and Carson 2011, Grose and Yoshinaka 2011).

12 Available here: http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/7293.

13 Pitkin (1967, 210-214) herself observes that representatives in the real world are rarely the extreme of either type. The median category of this ordinal scale could be thought of as "politicos" (see Kuklinski and Elling 1977), who blend delegate and trustee roles. This analysis subsumes them and determines the conditions under which they are most likely to adopt each role.

14 This measure does exhibit meaningful variation: 52.94% of representatives describe their districts as "safe," 19.33% describe their districts as "fairly safe," and 27.73% describe their districts as "fairly close." By this author’s count, there were at least 107 House districts actually decided by 10 points or less in the 1958 U.S. House elections (24.60% of the 435 total). For a modern comparison, 74 of the 435 (17.01%) U.S. House elections in 2022 were decided by 10 percentage points or less. This comparison highlights the greater number of close, party-heterogeneous districts in the 1950s.

15 This measure of variance is commonly used in related research (Levendusky and Pope 2010).

16 In contrast, foreign policy issues, a subject that the American Representation Study also measured constituent opinions on, are not uncomplicated issues. They are harder for the public to understand, something that the public is not as familiar with hearing arguments about from politicians. Erikson (1978) also expected foreign policy issues to show weaker constituent-representative linkages than social welfare or civil rights attitudes. Since they are a different kind of issue, what Hurley and Hill (2003) would call a complex issue, it is not surprising that district variance on foreign policy attitudes is not as related to variance on social welfare issues or variance civil rights issues as the variance of attitudes on those two issues are to each other. Variance on district foreign policy attitudes is only correlated with variance on social welfare attitudes at 0.15, which is not significant at conventional levels of statistical significance. Variance on district foreign policy attitudes is only correlated with variance on social welfare attitudes at 0.32 (which is significant at p < 0.001)

17 For one illustrative quote, see V.O. Key’s (1949, 41) Southern Politics in State and Nation: ""A more or less totally irrelevant appeal - back the hometown boy - can exert no little influence over an electorate not habituated to the types of voting behavior characteristic of a two-party situation."

18 DW-NOMINATE scores are available for this period, but the American Representation Study preserves anonymity of respondents and I am unable to match legislators with DW-NOMINATE scores. Furthermore, a DW-NOMINATE score is not a direct measure of ideology, but a proxy based on voting record that may be contaminated with the influence of constituency preferences and party effects.

19 It is worth noting that some of the contradictory findings about the marginality hypothesis may be due to research being done with data from different eras of Congress, some which are more ideological and partisan than others. Gulati (2004) recognizes this when he suggests in his conclusion that legislators may choose different strategies when control of the legislature is very competitive and partisan than when it is not. This is another reason it is valuable to extend research on the marginality hypothesis back to the 1950s, a distinctly less ideological era than the 1990s (which provide most of Gulati’s data) or the hyperpartisan era of today (2023).

20 The percentage of the total vote that the representative received is divided by 100 to create a decimal number that ranges between 0 and 1. This number is then subtracted from 1. The remainder is used as the legislator’s electoral marginality score. Marginality Score = 1 - $\frac{\mathit{\text{Legislator}}\text{'}s\hspace{0.17em}\mathit{\text{Percent}}\hspace{0.17em}\mathit{\text{of}}\hspace{0.17em}\mathit{\text{Total}}\hspace{0.17em}\mathit{\text{Vote}}\hspace{0.17em}\mathit{\text{in}}\hspace{0.17em}\mathit{\text{Last}}\hspace{0.17em}\mathit{\text{Election}}}{100}$.

21 These bootstrapped coefficients and standard errors are calculated based on B = 1000 bootstrap iterations. Bootstrapping can reduce the bias on estimated standard errors by re-sampling from the empirical distribution and is helpful in small-sample data analysis; see Efron and Tibshirani 1986.

22 Displaying the 95th percentile bootstrap confidence interval for a coefficient, rather than a t-statistic, is recommended by Jung, Lee, Gupta, and Cho (2019) because it delivers more information about the precision of the estimate.

23 For the most part, the explanatory variables in this model are not characterized by high multicollinearity. However, the Southern Democrat dummy is correlated with marginality at −0.71, Ideological Heterogeneity is correlated with the interaction between Ideological Heterogeneity and Marginality at 0.81, and Partisan Heterogeneity is correlated with the interaction between Partisan Heterogeneity and Marginality at 0.97. This last correlation is very high, which is another reason to bootstrap the confidence interval on this coefficient. Bootstrapping can help with calculating parameter estimates in cases of multicollinearity (Zahari, Ramli, and Mokhtar 2014).

24 As discussed earlier, representatives placed themselves on a 5-point scale of representation that ranged from "should vote the way his district wants," which is referred to as "Pure Delegate," to "should vote the way he thinks best," which is referred to "Pure Trustee." These are the primary representation categories of interest. I refrain from discussing the others, which is why the sum of the profile representative’s probabilities of choosing Pure Delegate or Pure Trustee does not equal 100%.

25 District Ideological Heterogeneity and district Partisan Heterogeneity are only positively correlated at about 0.20 in the American Representation Study, which not quite a statistically significant correlation at p < 0.06. They would almost certainly be much more highly correlated in contemporary U.S. House districts.

26 This is especially true given the low N of the survey samples in each U.S. House district, discussed earlier, but might still be true even if each district survey contained considerably more respondents.

27 These survey measures have the same virtues and challenges as those measuring social welfare issues.

28 Converse and Pierce (1986) find no evidence of this in France, but this could occur in America (e.g., Eulau 1987).