Competing for the platform

Abstract

What explains which groups are included in a party coalition in any given election cycle? Recent advances in political party theory suggest that policy demanders comprise parties, and that the composition of a party coalition varies from election to election. We theorize three conditions under which parties articulate an interest group’s preferred positions in its quadrennial platform: when groups are ideologically proximate to the party median, when groups display party loyalty, and when groups are flush with resources. Using computer-assisted content analysis on a unique and rich data source, we examine three cycles of testimony that 80 organized groups provided to the Democratic Party. The analysis compares group requests with the content of Democratic and Republican National Committee platforms in 1996, 2000, and 2004. Results show that parties reward loyal groups and those that are ideologically proximate to the party but offer no confirmation of a resource effect.

Keywords

content analysis Democratic Party interest groups platforms United States

During the 1996 Republican Convention, presidential candidate Bob Dole met conflict with pro-life advocates. The platform committee had adopted a strongly pro-life plank in the platform, but Dole favored a statement of tolerance that might appeal to party moderates. Although trying to foster a sentiment of compromise, Dole confronted a platform committee stacked with grassroots activists from the Christian Coalition, and his efforts ultimately failed (Rozell et al., 2012).

Studying the complexities of the relationship between political parties and organized interests has generated scholarship that describes these entities as distinct (Key, 1964), competing (Sabato, 1981), dependent (Kolodny and Dulio, 2003), or fully enmeshed (Cohen et al., 2008). These arguments, in light of stories such as above, lead us to explore the US platform-drafting process and ask: Under what conditions will a party include a group in its coalition via the party platform?

American political parties publicly articulate their ideological positions in quadrennial platforms, adopted in presidential nominating conventions. Organized groups seek to join the party coalition by having their interests articulated via platform planks. Parties may use these platforms to appeal to undecided voters (ideological moderates) or to boost enthusiasm (and therefore turnout) from a party’s ideological base. While Downs (1957) tells us parties use platforms only for the former, the “extended party network” (EPN) theory of parties, and a variety of anecdotal evidences, suggests parties regularly use conventions and platforms for the latter. We contend that platforms can be tools for either, depending on group characteristics.

We seek to further the understanding of the relationship between groups and parties by exploring their interaction during the platform-drafting process. We hypothesize conditions under which parties respond to organized groups and test these expectations using platform-drafting hearing testimony. Using automated text analysis, we find evidence of parties negotiating with groups for inclusion in their coalition. We find political parties responding to organized groups that show loyalty and ideological similarity to the mainstream of the party, but no evidence that parties bend to those with greater resources.

Our work contributes to and furthers our knowledge of the party–group interaction empirically, theoretically, and substantively. Empirically, we utilize exclusive transcripts of platform hearing testimony given to the Democratic Party in 1996, 2000, and 2004. These data are not public and are unavailable to the academic community except through our research. They provide a window into the direct negotiations between parties and groups. We have empirical evidence of the interaction between these potential coalition partners.

We also employ recent methodological advances in the form of automated text analysis. This technique is relatively new and commonly utilized by a select group of scholars of comparative party preferences (Laver et al., 2003; Slapin and Proksch, 2008). By employing these tools to scale differences in platform text with automated text analysis, we facilitate the discussion between comparative and American politics.

Theoretically, our findings help advance debate on the birth order of parties and groups (e.g. which came first?). We provide an important update to the group–party literature by demonstrating that money and resources are less influential in group–party relationships than one might expect. We also speak to ongoing concerns raised in popular literature about the corruptive power of wealthy groups on political action. Our results are consistent with recent evidence that shows ideas precede parties (Noel, 2014), and our findings dampen heightened anxiety that political parties are purchased by highest bidders. Instead, we show that loyalty to a party gains more attention than absolute dollars.

Substantively, we demonstrate how party platforms can matter to both parties and groups. Although criticized for lacking efficacy (Truman, 1993) or embodying “nothing more than the momentary sentiments of a majority of party activists” (Rozell et al., 2012), the very fact that groups mobilize to testify at platform hearings indicates that American party platforms are valuable. Platforms may not be the most important documents in American history, but they are possibly the single most representative document of group influences on the central parties.

Political parties and organized interests

Scholars often conceptualize organized interests as groups external to political parties. Thought of as natural combatants in some instances, or “symbiotic adversaries” (Fine, 1994) in others, their goals may vacillate between competing and complimentary. More recent research reconceptualizes this framework by acknowledging that political parties are ultimately, at least in part, comprised of organized interests (Bawn et al., 2012; Cohen et al., 2008; Karol, 2009; Noel, 2014). In this view, parties are loose coalitions of intense policy demanders who seek some common goal (Masket, 2016). Prior work has established the overlapping individual memberships of groups and parties as ranging from intertwined to indistinguishable (Grossmann and Dominguez, 2009; Heaney et al., 2012; Herrnson, 2009; Koger et al., 2009; Kolodny and Dwyre, 1998; Skinner et al., 2012).

Our idea of a dynamic negotiation between parties and groups is consistent with the EPN theory, which assumes parties are comprised of groups. Noel (2014) argues that parties and ideologies are distinct but related phenomena, where “parties choose the ‘who’ that they unite by considering what those people want” (p. 24). Ultimately, a party’s primary objective is to maximize the number of seats won in a given election. A party therefore strategizes to create a coalition that will maximize the probability of attaining that goal.

Organized interests, on the other hand, have multiple goals, and success is often self-defined and dynamic (Brunell, 2005). Groups demonstrate partisan affiliation (Koger et al., 2009, 2010) with respect to legislative activity (Grossmann and Dominguez, 2009; Masket, 2011) and candidate selection (Cohen et al., 2008; Miller and Schofield, 2003). Some groups seek specific policy implementation and would be unsatisfied with anything less than policy or social change. Other groups, however, may be pleased with the opportunity to voice, to clearly and vigorously express, their preferred policies; taking advantage of the opportunity to be heard would be sufficient cause for celebration.

What does this mean about the interaction between parties and groups? A group’s ability to mobilize and convince blocks of individual votes gives that group bargaining leverage with the party and its positioning (Harvey, 1998; Miller and Schofield, 2003; Truman, 1993). A group can promise to support a party or candidate or to mobilize out-group voters in supplement to the party campaign (Esterling, 2007). A group might also credibly threaten to exit a two-party system, either by not voting or by voting for a third party. These organized groups of ideological thinkers have little reason to compromise (Noel, 2014).

Parties, meanwhile, have many incentives to compromise. One party’s calculus can change from one election to the next, and it is not uncommon for parties to change issue positions over time. Karol (2009), for example, explains how US Democrats and Republicans switched positions on trade policy in the 20th century because party leaders valued organized labor as coalition partners more than they valued a consistent position on trade. Wolbrecht (2000) shows that the parties switched positions on women’s issues during the same time period largely because of the groups included in each party’s coalition—women for Democrats and religious conservatives for Republicans. In both cases, parties opted to maintain coalitions by changing policy positions.

This literature suggests elasticity in party coalitions, with organized groups moving in and out. We see bargaining across issues and groups where coalitions (e.g. parties) and their potential partners negotiate about which interests (e.g. groups) to include and which issues to favor. Party leaders settle on an arrangement of interests that maximizes the ability to win elections. This process is fluid and results in party coalitions that change between election cycles. We suspect it does so in predictable ways.

Why platforms?

The platform-drafting process is only one element of party building, and some regard it as entirely symbolic. The extant literature on the value of party platforms in the United States contradictorily finds platforms to be both vapid (Truman, 1993) and useful (Maisel, 1993; Pomper and Lederman, 1980; Snyder and Ting, 2002). We argue that it is a suitable juncture at which to examine group–party coalition formation for several reasons.

First, the platform-drafting process is time and labor intensive for both parties and groups. In 2004, for example, the Democratic Party’s platform-writing process spanned 6 months and four cities. It began with the appointment of a small platform drafting committee, which drafted an initial document. Any individual or group could submit statements, comments, or requests to testify in response to the document (Democratic National Convention, 2004). An official platform committee, made up of party leaders and “rising stars,” then conducted hearings in Portland, OR, Baton Rouge, LA, and Santa Fe, NM. Ultimately, 193 entities from around the country testified in person at the hearings.¹ The committee then revised the platform and submitted it to the full convention for discussion, debate, and adoption by delegate vote.

This activity highlights the utility of platform drafting to parties and groups, which is the second reason for our focus. Groups see platforms as a tool to voice interests and broadcast preferences nationally, while parties use platforms to bargain with groups for the ultimate electoral prize: votes. Rozell et al. argue groups “assign symbolic importance to the creation of platforms” and “believe that candidates do ultimately heed platforms to some degree” (2012: 46–8). They support their view with evidence from 2008, when the Republican platform offered immigration and abortion positions more conservative than McCain’s and included his name only once, due to group influence (Rozell et al., 2012).

Finally, platform drafting is a direct and observable negotiation between those drafting an official statement of the party and those seeking to influence that statement. It demonstrates the bargaining process leading to a party coalition, wherein groups commit support in exchange for interest articulation, and party leaders offer platform input based on the ability to deliver votes. Whether resulting in policy change or not, the interaction captures the behavior of “intense policy demanders” (Bawn et al., 2012; Cohen et al., 2008), those who have a strong preference for particular policy outcomes and, in this case, influence over platform creation.

Hypotheses

We assume that parties are instrumental and strategic about building coalitions, and that parties are dominated by elites who drive their broad decision processes. We theorize that parties court groups into their coalition in an attempt to maximize votes in an election. If we consider parties and groups to be related, overlapping, and perhaps even wholly enmeshed, we can think of the negotiation over platform content as a strategic game. The final outcome (i.e. the platform) is a Nash equilibrium representing each entities’ best response to the other (see Reinhardt and Victor, 2013). These observations beg our question: Under what conditions will a party include a group in its coalition via the party platform? Based on the above insights, we see three characteristics of groups as critical determinants of the final outcome: party loyalty, ideological proximity, and group resources.

First, we expect parties to include groups that can either promise or mobilize voters (Adkins and Dowdle, 2004). Groups with a large membership base or substantial financial resources (Gilens, 2012; Schattschneider, 1975) may find it easier to mobilize large numbers of voters. We therefore expect that all else being equal parties will seek to incorporate more resourceful groups into their coalitions.

H1: Groups with greater resources for mobilization are more likely to be included in the party coalition than groups with fewer resources.

Second, Karol (2009) and Wolbrecht (2000) show that parties are willing to shift issue positions to appeal to loyal groups. Those groups that have expressed interest in a variety of parties across time may not be able to compel one party to switch positions, while consistently loyal groups may be more likely to persuade a party to reconsider its position. We therefore expect parties to reward and protect groups showing loyalty to their coalition, as opposed to that of another party.

H2: Groups that display greater loyalty toward one party are more likely to be included in the party coalition than groups that show less loyalty.

Third, the basic Downsian model emphasizes the importance of spatial proximity (Downs, 1957). Groups are not likely to seek out party coalitions ideologically distant from their ideal point, and parties are unlikely to court groups ideologically opposite from the rest of the party coalition. We therefore expect parties to seek to include those groups in their coalition that are nearest to the median position of the party coalition. While that median point may shift as groups move in and out of the coalition, all else being equal we expect the probability of a group being included in a party’s coalition to increase as the ideological distance between the party median and the group median declines.²

H3: Groups with median ideology closer to the existing party coalition median are more likely to be included in the party coalition than groups that are ideologically more distant.

Data and methods

Platform hearings provide a venue for groups to articulate their views of the party platform and offer an excellent source from which to estimate the positions of groups trying to influence platform creation.³ Our unit of analysis is a group year, accounting for each group that testified at a Democratic National Committee (DNC) platform hearing in 1996, 2000, or 2004 (N = 80). Our empirical strategy is to examine what groups’ request and compare this to what parties produce in platform negotiations. We expect to systematically explain the group interests that are incorporated into the party coalition (platform) using group ideology, loyalty, and resources.

Our dependent variable is the overlap between the group’s testimony and the final DNC platform. We derive this by analyzing the content of the testimonies and platform using Wordscores, described fully below (Laver et al., 2003). Our data includes the complete population of groups that jockeyed for platform position during these three election cycles; however, groups that do not participate are unobserved. We therefore note that our data has a selection bias and do not draw inference on the population of all groups, but only those in a position to be included by parties in the first place. While this limits our inference, our approach increases our understanding of the competition for party attention among this class of groups, which has not been previously understood.

Using text as data

We began with the full (hardcopy) texts of the Democratic Party platform hearings from 1996, 2000, and 2004.⁴ We digitized the information into plain text, extracting testimony and question responses for each participating interest group. We obtained the full text of final party platforms from the websites of the DNC and Republican National Committee (RNC).

Analyzing text to proxy actors’ positions relative to one another generally includes two possible strategies. One approach is to hand code text. This strategy involves developing a series of categories and codes from the substance of the text and using humans to apply codes to segments of the text. Error can be minimized by using intercoder reliability checks, though the increase in reliability comes with increases in labor costs as well (see e.g. Budge, 2001a, 2001b).

An alternative approach is to use computer-assisted coding. This strategy involves some human coding to develop categories and codes on a sample of text, and then using software to apply these codes to remaining text. It allows one researcher to code large quantities of text (see e.g. Laver and Garry, 2000). Each of these modes of text analysis involves trade-offs. Generally, the researcher trades some measurement error in exchange for efficiency.

A third strategy is computer-assisted text analysis. Developed by Laver et al. (2003), it does not treat text as substantive prose meant to be dissected, analyzed, and comprehended. Rather, these scholars approach text with naïve agnosticism and assume that individual words, and short strings of words, convey information. Their approach treats texts as a chaotic “bag of words” where information about speakers’ positions is contained in the frequency of word choice, rather than in the meaningful substance of sentences and paragraphs. This strategy is efficient and reliable. The approach is easy to implement, fully replicable, and not prone to measurement error, or to human coding error or variance. It does, however, trade some precision and nuance in favor of efficiency and replicability.

Laver et al. (2003) argue that one can ascertain the policy positions of political actors by analyzing texts the actors have written. They show that treating text as data for the purpose of revealing actor’s policy positions is sometimes advantageous over alternatives. For example, let’s say one seeks to evaluate congressional policy preferences. Legislative roll call voting indicates one’s choice from dichotomous options, but is only observed at the end of a long process, and likely masks strategic behavior. Alternatively, one might use surveys or interviews, but costly methods can include measurement error due to participant guardedness. Analyzing the texts produced by political actors can provide an objective, easy to replicate, inexpensive, and transportable metric of actors’ policy positions.

There are two key requirements to make the Laver et al. approach work. First, the technique requires a known and quantified “reference text” against which to measure all other texts. The reference text must have a known position in ideological space. The original research on this technique, for example, uses the Comparative Manifestos Project, which hand codes multiple party manifestoes in several countries to develop a common ideological score for each. This score, and the word frequencies associated with it, acts as a quantitative anchor from which to compare new unscored “virgin” documents about which no ideological information is known. The algorithm in this technique, known as Wordscores, compares the word frequencies in the virgin text to those in the previously scored reference text to produce a score for the virgin text. Therefore, a known, verifiable, meaningful, and quantifiable score for the reference text is fundamentally necessary.

Second, the texts used in Wordscores must be appropriate texts for the technique in that they can be interpreted on a common spatial scale. Such a scale is common in the study of politics, legislatures, and parties where for several decades scholars have used Euclidean distances to estimate ideological positions. In such cases, the technique is flexible enough to analyze texts and actors such as activists, judges, commentators, bloggers, lobbyists, legislators, candidates, and so on. Yet, one would be misadvised to compare a party manifesto to a series of appliance manuals.

We employ the Wordscores technique for three reasons. First, the party platforms provide known and quantified reference texts to which we can compare the virgin texts of interest group platform testimonies. Since the groups try to influence the platform, the texts have a direct relationship. Second, the texts we have are appropriate. It is common to quantify party platforms and to assume that they can be placed in a Euclidean ideological space. It is equally common to assume that political actors, such as those that represent organized interests, can be placed in a similar space. Third, we are ultimately only interested in estimating the overlap in policy positions between the party and groups and have no expectations regarding specific planks or policies. We seek efficiency, accuracy, and scientific replicability over substantive nuance of particular groups’ requests. Wordscores therefore provides the best option for evaluating our expectations about the conditions under which parties will incorporate group interests into their platforms.

More technically, we give Wordscores two collections of reference texts—a series of Democratic Party platforms (from 1996, 2000, and 2004) and a series of Republican Party platforms (from 1996, 2000, and 2004). We concatenate each party’s platforms into one document, yielding one Democratic reference text and one Republican reference text. Next, we provide an a priori score to these reference documents of −1 (Democrat) and 1 (Republican), akin to the scale commonly used in ideological space.⁵

The algorithm in Wordscores observes a word (or two-word or three-word clusters, depending on the parameters provided) in a reference text and calculates the probability that the word came from a particular reference text based on word frequencies. This probability is multiplied by the a priori ideological anchoring score we provide, and the result is a “score” for each word in the reference texts. Then, we can compare individual words in the virgin texts to the scored words in the reference texts using the same logical procedure. First, the software calculates the relative frequency of each word in the virgin texts and multiplies that frequency by the weighted score of the word from the reference texts. An overall wordscore for the entire virgin text is then the average weighted scores of its words. For our purposes, individual words uttered by testifiers are weighted and aggregated to produce a testimony score that places the group in the same ideological policy space as the party platform (see Online Appendix 1 for a technical description of Wordscores). As Laver et al. (2003: 313) state:

Scoring words in this way replaces the predefined deterministic coding dictionary of traditional computer-coding techniques. It gives words policy scores, not having determined or even considered their meanings in advance but, instead, by treating words purely as data associated with a set of reference texts whose policy positions can be confidently estimated or assumed.

If we take the predefined ideological or policy space score of the reference texts as meaningful, the Wordscores technique allows us to use the frequency with which groups used particular words to estimate party proximity without evaluating their meaning or intent.

A skeptical reader may question the wisdom of stripping words of meaning and calculating organizational policy positions based on word frequencies. Yet counterintuitively, the blindness of this technique gives us confidence in its estimates. The procedure is immune to bias, personal interpretations, preset expectations, or other cognitive limitations of human coding, which makes our work both replicable and falsifiable. We therefore accept the trade-offs.

The Wordscores technique has been utilized (Baum and Zhukov, 2012) and criticized (Lowe, 2008). A primary critique is that Wordscores generates outputs that do not have interpretable magnitude. This is a serious shortcoming, if one is interested in generating point estimates. In our case, however, we seek only to draw an inference on the relationship between group inclusion in a party platform and the characteristics of a group that are predictive of its inclusion. The Wordscores provides us the invaluable tool of quantifying party and group interests in a single space, with information about the relative positions of each player rather than the magnitude of causal effects.⁶

Dependent variable

To calculate the extent to which an interest group is included in the platform, we use the transformed Wordscore for each group year as an indication of group proximity to the Democratic and Republican platforms. Values closer to −1 are nearer the Democratic platforms, while those closer to 1 are nearer the Republicans’ platforms. We make this calculation on “unigrams,” or the overlap in single words, as well as “bigrams” and “trigrams,” the overlap in groups of two and three words, respectively. The literature suggests that counting small groups of words, rather than single words, provides a greater degree of sensitivity of overlap between texts; however, one must trade-off degrees of freedom and power to make this exchange. Figures 1 and 2 show the relative positions of the groups in our sample to the party platforms for unigrams (Figure 1) and bigrams (Figure 2). The reader may notice the DNC position (at −1) and the RNC position (at 1), which more or less splits the population into thirds, with most groups falling between the two parties. There is similarity between the arrangements of groups between the two figures, giving us confidence in the method.

Figure 1.

Wordscores estimates for unigrams.

Figure 2.

Wordscores estimates for bigrams.

Substantively, the transformed Wordscore value for each group represents the group’s proximity to each party, or the extent to which each party included the group’s interests in their platforms, as measured by the overlap in word frequencies. The arrangement of groups along this scale shows a reasonable left–right logic. On the left, we see traditional Democratic groups such as labor unions and civil rights organizations, while on the right, we see veterans’ organizations and business interests. Table 1 shows the summary breakdown of the texts used to create the figures.

Table 1.

Summary of texts.

Year	DNC platform word count	RNC platform word count	Number of groups	Group testimony average unigrams (percent scored)	Group testimony average bigrams (percent scored)	Group testimony average trigrams (percent scored)
1996	18,032	27,160	24	800 (95%)	454 (54%)	132 (16%)
2000	23,964	34,503	26	852 (93%)	458 (50%)	132 (15%)
2004	17,821	42,076	30	1332 (94%)	712 (51%)	199 (15%)
Total	59,817	103,739	80	1016 (94%)	288 (52%)	157 (15%)

Independent variables

Interest group ideology is operationalized using Bonica’s (2014) measure of group ideology. This score is developed from observations of campaign contribution behavior from groups, PACs, and associated individuals (e.g. employees) and has been shown to provide a good approximation of group ideology. We must accept some missingness to use this variable because not all of our group testifiers appear in the Bonica data (Table 2 offers summary statistics).

Table 2.

Summary statistics.

Variable	Mean (standard deviation)	Range	Definition	Source
Platform inclusion—unigrams (DV1)	0.0185 (1.41)	−2.7 to 3.43	Transformed Wordscores values, where DNC (−1) and RNC (1) platforms, concatenated for 1996, 2000, and 2004, are reference texts, and group testimonies are virgin texts.	Interest group testimony to DNC platform drafting committees and texts of 1996, 2000, and 2004 DNC and RNC platforms.
Platform inclusion—bigrams (DV2)	0.0199 (1.41)	−2.6 to 4.2	Transformed Wordscores values, where DNC (−1) and RNC (1) platforms, concatenated for 1996, 2000, and 2004, are reference texts, and group testimonies are virgin texts.	Interest group testimony to DNC platform drafting committees and texts of 1996, 2000, and 2004 DNC and RNC platforms.
Ideology	−0.437 (0.597)	−1.19 to 1.02	Ideological spatial position estimated from campaign contributions of group’s PAC or donations of its members.	Adam Bonica Database on Ideology, Money in Politics, and Elections (2013).
Loyalty	10.6 (3.02)	5.3 to 14.6	The natural log of the sum of all donations made to candidates by the group’s PAC or its employees, multiplied by the percentage of donations made to Democrats.	Federal Election Commission campaign contribution data, as compiled by Opensecrets.org.
Resources	0 (1)	−0.75 to 5.2	Standardized index for each group based on two characteristics known to affect their ability to mobilize voters (Kollman, 1998): number of members and reported budget. Each measure is standardized among groups, then summed and standardized again.	Galenet’s Encyclopedia of Associations (1996, 2000, 2004); Colgate, National Trade and Professional Associations of the United States (1996, 2000); Gale Encyclopedia of Business and Professional Associations 1996–1997; Congressional Quarterly Inc., Public Interest Group Profiles (1996–1997, 2000–2001); GuideStar (2008); Foundation Center (2008); Associations Unlimited (2008); OpenSecrets.org (2008); Campaign Money (2008); the Federal Elections Commission (2008); and groups’ webpages and archived webpage.

In measuring groups’ resources, we seek an indicator of potential to mobilize voters. We create a standardized index for each group based on two characteristics known to affect their ability to mobilize voters (Kollman, 1998): number of members and reported budget. Each measure is standardized among groups, summed, and standardized again. Formally, resources are the sum of the standardized count of the number of members a group has and the group’s standardized budget. The sum is standardized again:

Resources = std (std (M_{t}) + std (B_{t}))

where M_t is the reported number of members in the group in electoral cycle t, B_t is the group’s reported budget in cycle t, and std redistributes the variable to a standard normal distribution.

We measure loyalty as the percentage of campaign contributions made by each group’s political action committee, or by individual employees of the group, to the Democratic Party during the 2-year campaign cycle of the testimony. We weight the percentage by the total amount of money the group gave to Democratic candidates in the same cycle and take the natural log of the result. Most of the testifying groups give heavily to Democrats. Twenty-four have no PAC or employee contributions tracked by the Federal Elections Commission. Formally:

Loyalty = ln ((\frac{C_{D t}}{C_{t}}) \times A)

where C_t is the total contributions made by a group in cycle t, C_Dt is the group’s total contributions to Democratic candidates in cycle t, and A is the total amount donated.⁷

Loyalty and ideology are both derived from campaign contribution behavior, though this should not cause concern. While the initial data is the same, the algorithms convert the observed behavior into two different metrics. They are correlated at r = 0.2.

Results and analysis

We estimate a linear regression model. We cluster the error term on groups, as several groups are represented multiple times in this short cross-sectional time series. With 80 observations⁸ spanning three time periods (N = 80 and T = 3), our data do not fit the T > N restriction desired for standard cross-sectional time series analysis (Beck and Katz, 1995). We therefore insert timewise fixed effects for each platform year, excluding 1996, the reference year (appropriate for short panels, see Arellano, 2003; Wooldridge, 2010). We estimate the standard errors with bootstrapping because our dependent variable is itself the product of an estimation procedure. The statistical model is⁹:

{Platform inclusion}_{i t} = α + β_{1} {ideological congruence}_{i t} + β_{2} {loyalty}_{i t} + β_{3} {resources}_{i t} + β_{4} D_{i t_{2}} + β_{5} D_{i t_{3}} + ∊_{i t}

We estimate this model separately for unigrams, bigrams, and trigrams (Table 3 provides the results).

Table 3.

Linear regression results.

Regression Estimates for Interest Group Wordscores Relative to Party Platforms (−1 = Dem/+1 = Repub)
	Unigrams	Bigrams	Trigrams
	Model 1	Model 2	Model 3
Ideology (−1 = D/+1 = R)	1.131** (0.3461)	1.324*** (0.367)	1.021** (0.3751)
Dem. Loyalty	−0.245** (0.0781)	−0.210** (0.0787)	−0.095 (0.0663)
Resources	0.06 (0.1868)	0.066 (0.2022)	−0.097 (0.1815)
Year 2000	0.269 (0.5391)	0.052 (0.3702)	−0.306 (0.5201)
Year 2004	0.697 (0.382)	0.526 (0.3457)	0.041 (0.4865)
Constant	2.660** (0.9366)	2.635* (1.0275)	1.589* (0.801)
Adj. R ²	0.243	0.253	0.13
N	45	45	45
Pr (Chi)	0	0.003	0.043

***p(z) < 0.001; **p(z) < 0.01; *p(z) < 0.05; Bootstrapped standard errors clustered on groups in parentheses.

We note that the unigram and bigram models perform similarly to each other, while the trigram model performs poorly. We suspect the increased sensitivity of the trigram is not worth the sacrifice of statistical power, so we focus instead on the bigram model (model 2), combining statistical power and sensitivity.

We find no support for H1 that group resources will be positively associated with platform inclusion. Although unexpected, this finding is consistent with previous research suggesting that group monetary resources are not particularly predictive of group behavior or success (McKay, 2012). Moreover, it suggests that parties do not necessarily rely on constituent groups for external voter mobilization. This may be because parties have sophisticated internal machineries using microtargeting and “smart” campaigning (Issenberg, 2013). The platform-drafting process and coalition building between parties and groups sit external to this process, occur in a different timeframe, or remain unobserved using our methods. Either way, our finding dampens concerns about groups buying their way into the Democratic Party’s favor.

Loyalty is negative (−0.775) and statistically significant (p < 0.01) in model 2. Results for loyalty are thus consistent with our expectation (H2) that groups showing more loyalty to the Democratic Party are more likely to be included in the Democratic platform (recall that inclusion is represented by a transformed Wordscore nearer to the Democratic platform (−1)).

The positive and statistically significant coefficient on ideological congruence is consistent with our expectation (H3) that groups ideologically closer to the party’s status quo position are more likely to be included in the platform. Recall that the dependent variable and ideology are measured on the same scale (−1, 1), which is equivalent to (Poole and Rosenthal, 2011) NOMINATE scales. The coefficient is positive, relatively large given the distribution of the data (1.1), and statistically significant (p < 0.05).

Discussion

We have used the Wordscores technique to gauge the relative positions of parties and groups and to engage in a basic test of our hypotheses. From our results, we infer that ideology and loyalty are significant predictors of group inclusion in a party coalition (as measured by the platform), while resources are not. Overall, our analysis casts a shadow on the role of group resources in influencing party platforms, though we do not interpret these findings to suggest generally that group resources do not influence party behavior.

Our inference is limited in three important ways. First, we have temporal and ideological limitations because our specialized data includes only the Democratic Party’s platform-drafting process in 1996, 2000, and 2004. In addition to being denied access to Republican platform testimony in 1996–2004, both parties changed their platform-drafting process in 2008, so comparable data for later elections does not exist.

Second, our sample only includes groups that testified before the platform drafting committee, who are presumably a subset of groups who requested to testify. It is likely that the Democratic Party would seek input from their most loyal activists, and if so, our loyalty measure is biased due to the sample on which we make observations. We have no information on groups that might have sought to influence the process but were not granted access, and it is possible that well-funded groups influenced the process without testifying at the hearings.

Yet, these are limitations of the research design, not of the empirical model. We have opted to utilize this rare and valuable data to gain insight into the Democratic Party’s process during the time period under study and to conduct an analysis with novel and superior measures of group–party interactions. Even with limitations, our results show that among testifying groups, group resources do not predict inclusion in a platform, while group loyalty does.

Third, we have limited ability to substantively interpret coefficients or relative magnitudes. The substantial transformation of text to data is compelling because it allows us to quantify a previously unmeasured process, but also somewhat dissatisfying because the values of the measures are not necessarily substantively meaningful. If one is persuaded by the critiques, a downside of our empirical strategy is that we cannot reasonably interpret the relative size of the effects of group characteristics on platforms—only whether such associations exist.

We measure the groups of policy demanders who attempt to join the party coalition through the platform-drafting mechanism. Limitations notwithstanding, our work represents the first systematic window into this closed-door interaction between groups and parties. Our approach allows us to measure the characteristics of groups that seek to become part of a party coalition in a given election cycle. The opportunity to observe and directly compare organized groups’ requests about the party and its platform to the final document is unique and valuable.

Our results suggest that groups and parties engage in a strategic interaction to negotiate a platform, where the platform represents the aggregation of group policy interests that dictate a party organization’s agenda in a given election cycle. By analyzing the population of testifying groups, and identifying the characteristics of those with successfully articulated interests, we provide a creative and conservative test of how groups achieve interest articulation in party platforms. Testifying groups that demonstrate higher degrees of party loyalty through campaign contributions, and those with ideological preferences closer to the party, are more likely to see their interests articulated in the platform.

Conclusions

We set out to investigate party–group coalition building during the platform-drafting process. We develop three testable expectations: (1) parties reward loyal interest groups with platform inclusion, (2) parties seek to include groups that are ideologically near the existing party median, and (3) parties are more likely to reward groups that have greater resources to mobilize voters. Our text analysis of party platforms helps probe the makeup of the final coalition of “policy demanders.” Our evidence furthers EPN theory’s view that parties represent extended coalitions of organized interests and that these interests jockey with one another to build a party coalition during each election cycle.

We use unique data to generate these results. Automated text analysis of the transcripts from three cycles of testimony provided to the DNC’s platform writing committee helps determine how much congruence there is between what groups’ request and what they receive in the final platform. Word frequencies and Wordscores technology produce reasonable estimates of overlap between groups and parties. Our empirical findings generally support our expectations: Groups ideologically closer to the party, and those more loyal to the party, are more likely to be included in the platform.

Substantively, our article offers first steps toward a greater understanding of party platforms, often thought to be useless, and neglected in the agenda of American politics. If the parties use platforms to mobilize voters vis-a-vis interest groups, the study of those platforms and the groups participating in their construction has implications for the analysis of networks and party building (Koger et al., 2009). Can weakly connected groups use platform-drafting hearings to gain better network positions? Do the relative connections between groups and parties predict meaningful behavior from parties (e.g. which candidates to nominate) and groups (e.g. which candidates to support financially)? Platform-drafting politics can provide insight into broader cleavages within a party that may reveal new insights about party positions, candidate strategies, and policy goals.

Despite shortcomings in our research design, we offer two clear, consistent findings: Political party platforms are responsive to organized interests that are ideologically similar to the party status quo and to those who have demonstrated loyalty to the party. Moreover, the composition of parties changes across time in systematic ways. We look to future scholarship to improve our ability to measure group actions, incentives, and resources with respect to parties.

Supplemental material

Supplemental Material, PPQ678888_Appendix_ed - Competing for the platform: How organized interests affect party positioning in the United States

Supplemental Material, PPQ678888_Appendix_ed for Competing for the platform: How organized interests affect party positioning in the United States by Robert Harmel, Svåsand Lars, Jennifer Nicoll Victor, and Gina Yannitell Reinhardt in Party Politics

Footnotes

Acknowledgements

A special thanks to Julia Azari, Janet Box-Steffensmeier, Randall Calvert, Kevin Esterling, Will Everett, Elisabeth Gerber, Zachary Greene, Seth Masket, Richard Skinner, Jonathan Slapin, and participants at the 2008 Visions in Methodology conference for their suggestions and feedback.

Author contribution

The authors contributed equally to this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplemental material

Supplementary material for this article is available online.

Notes

References

Adkins

Dowdle

(2004) Bumps in the road to the white house: how influential were campaign resources to nominating George W. Bush? Journal of Political Marketing 3(4): 1–27.

Arellano

(2003) Panel Data Econometrics. Oxford, UK: Oxford University Press.

Baum

Zhukov

(2012) What determines the news about foreign policy? newspaper ownership, crisis dynamics and the 2011 Libyan uprising. In: International Studies Association Annual Meeting, San Diego, CA. Available at: http://www.hks.harvard.edu/fs/mbaum/documents/BaumZhukov_ISA2012.pdf (accessed 23 November 2016).

Bawn

Cohen

Karol

. (2012) A theory of political parties: groups, policy demands and nominations in American politics. Perspectives on Politics 10(3): 571–597.

Beck

Katz

(1995) What to do (and not to do) with time-series cross-section data. American Political Science Review 89(3): 634–647.

Bonica

(2014) Mapping the ideological marketplace. American Journal of Political Science 58(2): 367–386.

Brunell

(2005) The relationship between political parties and interest groups: explaining patterns of PAC contributions to candidates for congress. Political Research Quarterly 58(4): 681–688.

Budge

(ed) (2001a) Mapping Policy Preferences: Estimates for Parties, Electors, and Governments, 1945–1998. Oxford, UK: Oxford University Press.

Budge

(2001b) Validating party policy placements. British Journal of Political Science 31(1): 210–223.

10.

Cohen

Karol

Noel

. (2008) The Party Decides: Presidential Nominations Before and After Reform. Chicago: University of Chicago Press.

11.

Democratic National Convention (2004) Frequently asked questions about the national platform and the convention platform committee. Available at: http://www.kintera.org/site/pp.asp?c=luI2LaPYG&b=97938 (accessed 23 November 2016).

12.

Downs

(1957) An Economic Theory of Democracy, 1st ed. New York: Harper and Row.

13.

Esterling

(2007) Buying expertise: campaign contributions and attention to policy analysis in congressional committees. American Political Science Review 101(1): 93–109.

14.

Fine

(1994) Interest groups & the framing of the 1988 democratic & republican party platforms. Polity 26(3): 517–530.

15.

Gilens

(2012) Affluence and Influence: Economic Inequality and Political Power in America. Princeton: Princeton University Press; New York: Russell Sage Foundation.

16.

Grossmann

Dominguez

CBK

(2009) Party coalitions and interest group networks. American Politics Research 37(5): 767–800.

17.

Harvey

(1998) Votes without Leverage: Women in American Electoral Politics, 1920-1970. Cambridge: Cambridge University Press.

18.

Heaney

Masket

Miller

. (2012) Polarized networks: the organizational affiliations of national party convention delegates. American Behavioral Scientist 56(12): 1654–1676.

19.

Herrnson

(2009) The roles of party organizations, party-connected committees, and party allies in elections. The Journal of Politics 71(4): 1207–1224.

20.

Issenberg

(2013) The Victory Lab: The Secret Science of Winning Campaigns, Reprint ed. New York: Broadway Books.

21.

Karol

(2009) Party Position Change in American Politics: Coalition Management, 1st ed. Cambridge/New York: Cambridge University Press.

22.

Key

(1964) Politics, Parties, & Pressure Groups, 5th ed. Georgia: Crowell.

23.

Koger

Masket

Noel

(2009) Partisan webs: information exchange and party networks. British Journal of Political Science 39(3): 633–653.

24.

Koger

Masket

Noel

(2010) Cooperative party factions in American politics. American Politics Research 38(1): 33–53.

25.

Kollman

(1998) Outside Lobbying: Public Opinion and Interest Group Strategies. Princeton: Princeton University Press.

26.

Kolodny

Dulio

(2003) Political party adaptation in US congressional campaigns why political parties use coordinated expenditures to hire political consultants. Party Politics 9(6): 729–746.

27.

Kolodny

Dwyre

(1998) Party-orchestrated activities for legislative party goals campaigns for majorities in the US house of representatives in the 1990s. Party Politics 4(3): 275–295.

28.

Laver

Benoit

Garry

(2003) Extracting policy positions from political texts using words as data. American Political Science Review 97(2): 311–331.

29.

Laver

Garry

(2000) Estimating policy positions from political texts. American Journal of Political Science 44(3): 619–634.

30.

Lowe

(2008) Understanding wordscores. Political Analysis 16(4): 356–371.

31.

Maisel

(1993) The platform-writing process: candidate-centered platforms in 1992. Political Science Quarterly 108(4): 671–698.

32.

Masket

(2011) No Middle Ground: How Informal Party Organizations Control Nominations and Polarize Legislatures. Ann Arbor: University of Michigan Press.

33.

Masket

(2016) The Inevitable Party: Why Attempts to Kill the Party System Fail and How They Weaken Democracy. New York: Oxford University Press.

34.

McKay

(2012) Buying policy? the effects of lobbyists’ resources on their policy success. Political Research Quarterly 65(4): 908–923.

35.

Miller

Schofield

(2003) Activists and partisan realignment in the United States. American Political Science Review 97(2): 245–260.

36.

Noel

(2014) Political Ideologies and Political Parties in America. New York: Cambridge University Press.

37.

Pomper

Lederman

(1980) Elections in America: Control and Influence in Democratic Politics, 2nd ed. New York: Longman.

38.

Poole

Rosenthal

(2011) Ideology and Congress. New Brunswick: Transaction Publishers.

39.

Reinhardt

Victor

(2013) Competing for the platform: the politics of interest group influence on political party platforms in the United States. In: Annual Meeting of the Southern Political Science Association, Orlando, FL. Available at: https://www.researchgate.net/profile/Jennifer_Victor/publication/228256117_Competing_for_the_Platform_The_Politics_of_Interest_Group_Influence_on_Political_Party_Platforms/links/00b7d52827fed13d33000000.pdf (accessed 23 November 2016).

40.

Rozell

Wilcox

Franz

(2012) Interest Groups in American Campaigns: The New Face of Electioneering, 3rd ed. New York: Oxford University Press.

41.

Sabato

(1981) The Rise of Political Consultants: New Ways of Winning Elections. New York: Basic Books.

42.

Schattschneider

(1975) The Semi-Sovereign People: A Realist’s View of Democracy in America, 1st ed. Hinsdale: Cengage Learning.

43.

Skinner

Masket

Dulio

(2012) 527 committees and the political party network. American Politics Research 40(1): 60–84.

44.

Slapin

Proksch

S-O

(2008) A scaling model for estimating time-series party positions from texts. American Journal of Political Science 52(3): 705–722.

45.

Snyder

JMJ

Ting

(2002) An informational rationale for political parties. American Journal of Political Science 46(1): 90–110.

46.

Truman

(1993) The Governmental Process: Political Interests and Public Opinion (1993, 2nd (reprint). Originally published: New York: Knopf [1971, 1951] ed.). Berkeley: Institute of Governmental Studies, University of California, Berkeley.

47.

Wolbrecht

(2000) The Politics of Women’s Rights: Parties, Positions, and Change. Princeton: Princeton University Press.

48.

Wooldridge

(2010) Econometric Analysis of Cross Section and Panel Data, 2nd ed. Cambridge: MIT Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB