Interest group support for non-group issues

Abstract

Organized interest groups tend to focus on a narrow set of issues that promote the common interests of their members. They support political candidates who are favorable toward the group’s interests. But whereas interest groups support politicians based on a narrow set of issues, politicians have platforms that cover the entire political spectrum, so supporting a politician implies supporting all of that politician’s positions. A secondary effect of interest group support for politicians on one issue is that they are also supporting positions on other issues that are well outside the scope of that group’s interests. This analysis shows that the systematic relationships among politicians’ political platforms result in interest groups supporting issues that are well outside the stated common interests of the groups.

1 Introduction

Buchanan (1954) notes that one of the differences between individual choice in politics versus the market is that the alternatives among which voters can choose are different. One of the biggest differences is that in markets, individuals are able to make marginal adjustments in the purchase of individual goods, whereas in politics individuals must choose between complete market baskets of goods that are represented in candidates’ platforms. Consider grocery shopping as an analogy. In a supermarket shoppers can place as much or as little of any of the goods in the store in their baskets, and are free to choose goods from competing producers. Someone who prefers Pepsi as a cola drink and Sprite (a product of the Coca-Cola company) as a lemon-lime drink can buy both, purchasing from both companies. In politics, people can choose all of the policies of the Republican candidate or all of the policies of the Democrat, but cannot choose, for example, the military policies of the Democrat and the health care policies of the Republican. To continue with the supermarket analogy, competing candidates fill up shopping carts and individuals must choose a complete shopping cart full of goods rather than choosing the individual items to be placed in the cart.

The choice political shoppers face is far from transparent. Again extending the supermarket analogy, individuals might be able to see the top items in the cart but only have the vaguest idea what lies at the bottom of the cart. But choosing a candidate’s cart means choosing everything in it, lending support to all of the policy positions in that candidate’s platform. Sometimes the lack of information about the cart’s contents may be due to vague policy positions by politicians,¹ but it also may be that voters do not invest the resources to discover its entire contents because, as Downs (1957) notes, they are rationally ignorant. They choose based on just a few items they see that are important to them.

One possible result is that by choosing based on a few items, they end up with a cart that costs more than it is worth.² Sowell (1980:119–120) notes that because voters cannot see the whole cart, politicians often claim there is more in the cart than it possibly could contain. The nature of individual choice in politics means that supporting one candidate over another means supporting everything in that candidate’s platform, unlike in the market where individuals are free to buy their colas from one company and their lemon-lime drinks from another.

The aggregated nature of political choice is especially relevant to interest group politics. The National Rifle Association (NRA) supports political candidates who have platforms that actively promote the rights of individuals to keep and bear arms, while the National Education Association (NEA) supports candidates who have platforms that promote educational issues and support teachers. The platforms of the candidates they support also have positions on other issues, so when interest groups support candidates based on one issue, they also have the secondary effect of supporting all the other issues in those candidates’ platforms. In generic terms, consider an interest group that supports a candidate because that candidate is favorable to issue A. The candidate is favorable to many other issues as well and the candidate’s platform also supports issues D and F, and opposes H. Thus, when the interest group supports the candidate because the candidate is favorable to the group’s issue A, that interest group is also unintentionally supporting issues D and F, and opposing H.

Public choice has often analyzed politics as exchange and described a process in which political bargaining is analogous to market exchange,³ but political processes and market processes are not entirely analogous and this paper demonstrates one case in which the analogy breaks down. As the analysis that follows shows, there are often systematic relationships among politicians who support particular issues. Candidates who are favorably inclined toward supporting Second Amendment issues also tend to be favorably inclined on anti-abortion right to life issues. So, a secondary effect of people’s contributions to and support of the NRA is that they are also supporting an anti-abortion agenda.

One might think that interest groups tend to be aligned with the interests of political parties, so some interest groups are “Republican” and others are “Democratic.” The empirical work that follows shows that even when controlling for political party, there are significant links among interest groups that support political candidates that are not party-related. Political parties tend to be identified by ideologies, as Hinich and Munger (1994) describe them, but interest groups build their support based on issues, not ideologies.⁴

Interest groups can organize and be effective when they have narrow concentrated interests, as Olson (1965) shows. Congleton (2015) notes that people will participate in group activities when they share the group’s norms, so member preferences beyond those narrow common interests are largely irrelevant to the group’s activities. Some people may be single-issue voters who follow an interest group’s recommendation at the ballot box, but voters who consider a broad array of issues when voting may contribute financially to a particular interest group help support the group’s agenda. Supporters tend to believe that their interest group policy positions are in the public interest, Congleton (1991) notes, but those supporters also effectively support other policy positions they may not recognize they are supporting. This paper examines the degree to which interest groups unintentionally but systematically support issues that are outside of their group’s stated purpose.

2 An example

Interest group ratings and/or endorsements provide a measure of the degree to which political candidates support that group’s interests. The website votesmart.org provides aggregated data collected from interest groups which rated and/or endorsed 680 politicians in elective office or running for office in 2014. This analysis uses 2014 data and all data comes from that website. An endorsement is binary—either the politician is endorsed or not—and interest group ratings of candidates consistently appear on a 0–100 scale making it easy to compare the ratings of different interest groups. Note that ratings and endorsements are two different measures of interest group support for candidates. Candidates might be highly rated by an interest group, but not endorsed, for example. Endorsements and ratings provide two separate measures of the degree to which candidates are supported by interest groups.

The US Chamber of Commerce, which rates many politicians is an interesting example because it represents business interests. Their website says “The US Chamber of Commerce is the world’s largest business organization representing the interests of more than 3 million businesses of all sizes, sectors, and regions. … We advocate for pro-business policies that create jobs and grow our economy. Key issues range from smart tax policy and regulatory relief to legal reform and trade promotion.” Presumably, those three million members have diverse interests in other areas, but the Chamber supports politicians based on the single issue of the degree to which they are business friendly. By supporting business-friendly politicians, they also unintentionally lend their support to the other issues those business-friendly politicians support, and unintentionally work against the issues those business-friendly politicians oppose.

Table 1 shows the results of linear regressions that use the Chamber rating of politicians as the independent variable and the ratings of nine other interest groups as dependent variables. Where the Chamber rating is represented as USCOC and the other i interest groups are represented as GROUPi, the regressions in Table 1 are of the form

$${\text{GROUPi}} =\upalpha +\upbeta{\text{USCOC}} +\upvarepsilon$$

(1)

Eight of the nine show very high correlations in the ratings. The only one that does not is the National Rifle Association. All interest groups do not rate all politicians, and the Obs. column in the table shows the number of observations in the regression, which is the number of politicians rated by both the Chamber and the NRA. The USCOC column shows the coefficient and t-statistic from the regression. Both the t-statistic, which is less than 1 and the R² of 0.0147 show that there is very little correlation between the USCOC and NRA ratings.

Table 1 US chamber of commerce rated politicians bivariate linear regressions

All of the other groups in the table show very high correlations. Perhaps surprisingly, in light of the low correlation between the Chamber ratings and NRA ratings, the Gun Owners of America ratings are strongly correlated with the Chamber ratings. There were 409 politicians rated by both interest groups, and the coefficient on USCOC has a t-statistic of 17.7 and an R² of 0.4351. The correlations between the Chamber ratings and other interest group ratings are similarly high. By supporting the Chamber, those three million businesses are supporting the interests of National Right to Life, and opposing the interests of the National Abortion Rights League, the National Education Association, the Alliance for Retired Americans, and the other interest groups shown in the table. The Chamber is an interesting example because one would think that these other issues are largely unrelated to business interests, and Chamber members would likely have varying views on those issues.

Voting models (such as the median voter model) often collapse issues into a single-dimensioned left-to-right continuum with Republican positions tending toward the right and Democratic positions tending toward the left. Table 2 shows the results of linear regressions that add a binary variable, REP, which is one for Republican politicians and zero for Democrats, to the model specification to estimate

$${\text{GROUPi}} =\upalpha +\upbeta1{\text{USCOC}} +\upbeta2{\text{REP}} +\upvarepsilon.$$

(2)

when accounting for party affiliation, the relationships between the Chamber ratings and the ratings of the Gun Owners of America and the National Right to Life are no longer statistically significant at the .05 level but all of the other relationships remain statistically significant. The signs on the coefficients remain the same except for the National Education Association, which flips from negative to positive. Table 2 shows that the results from Table 1 are not due to party affiliation. The results in Table 1 provide the direct evidence of the paper’s claim that when interest groups that support a candidate based the candidate’s position on their narrow issues, they also unintentionally lend their support to other issues unrelated to the group’s collective interests, even when controlling for party affiliation.

Table 2 US chamber of commerce rated politicians linear regressions controlling for party affiliation

3 Other interest groups

The same analysis that was shown for the Chamber of Commerce was done for the other nine interest groups in the dataset. To conserve space, the results for the other groups from regressions like Eq. (1), shown in Table 1, are shown in Table 3 and results from regressions like Eq. (2) are shown in Table 4. When β from Eq. (1) is statistically significant at the .05 level or greater, Table 3 enters a + when the sign is positive, a − when the sign is negative, and enters a 0 when the coefficient is not statistically significant at the .05 level.

Table 3 Correlation of interest group positions significant bivariate regression results

Table 4 Correlation of interest group positions significant regression results controlling for party dependent variable on left

Interpreting the magnitude of the coefficients is problematic because while the ratings give a good ordinal measure of which candidates more closely correspond to the policy preferences of interest groups, they are a relative ranking and not a cardinal measure of the group’s interests. Information about the degree to which group rankings are correlated is shown by the t-statistics and R²s, and the signs on the coefficients indicate whether one group’s rankings are in support or in opposition to another’s, so this is the information summarized in the table.⁵

Table 3 shows that the only coefficients not statistically significant are the one from Table 1 showing NRA support for the USCOC, and Stand Against Spying’s support for the NRA and NEA, and their inverses. The insignificant coefficient is somewhat surprising considering that the table shows positive coefficients relating the Chamber with the Gun Owners of America, and also positive coefficients showing shared support between the NRA and GOA. The regression that produces the positive coefficient in the NRA column is

$$\begin{array}{*{20}l} {{\text{NRA}} = 0.224\left( {4.24} \right) + 0.826{\text{GOA}}\left( {11.73} \right) + \varepsilon ,} \hfill \\ {{\text{R}}2 = 0.6792\quad {\text{t{-}statistics}}\;{\text{in}}\;{\text{parentheses}}} \hfill \\ \end{array}$$

(3)

so as expected, those with high NRA ratings also have high GOA ratings.

Table 3 shows that for all of the ten interest groups in the sample, supporting one of them also brings with it support or opposition for policy positions that likely have only a peripheral relationship, at best, with the stated goals of that group. Consider the National Education Association, for example. There is a negative correlation between the NEA ratings and the ratings of both the NRA and the GOA, two Second Amendment rights groups. There is a positive correlation between the National Abortion Rights Action League and a negative correlation between the National Right to Life. Support for the NEA also, as a secondary effect, opposes gun rights groups and supports a woman’s right to choose abortion. There is a positive correlation with the Chamber and the League of Conservation Voters, so supporters of the NEA also support a pro-business agenda and a pro-conservation agenda. That same support, as a secondary effect, also applies to human rights, retirees, and the interest group Stand Against Spying. Support of the NEA means supporting a market basket of issues that are not closely related to the stated agenda of that group, and that undoubtedly are at odds with the political views of some of its members and financial contributors.

Table 4 shows the results when the binary variable for party affiliation is included in the regression. For almost all groups, there remain strong correlations. After accounting for party affiliation, the League of Conservation Voters shows statistically significant relationships with every other interest group. With other groups there are only a few cases in which relationships become insignificant with the party variable. Stand Against Spying has four statistically significant relationships, and most groups lose only two or three statistically significant relationships.

The glaring exception is the NRA, which is negatively correlated with the League of Conservation Voters but not correlated with any other groups when a party variable is included in the regression. Examining this relationship in more detail, a regression like Eq. (1) was run with the NRA rating as the dependent variable and the GOA rating as the independent variable, but separately for Republicans and Democrats. In both cases the coefficients were not statistically significant, with t-statistics of 1.50 for Republicans and 1.65 for Democrats. Both groups tend to rate Republicans higher than Democrats, but after accounting for party affiliation, their ratings within one party are essentially uncorrelated. The groups’ stated agendas are the same, so it is interesting to see that after accounting for party affiliation, there is no correlation between their rankings.

As noted earlier, the reason for running the regressions in Table 4 is only to show that the conclusions from Table 3 remain after taking into account the party affiliation of politicians. The primary results are those in Table 3 that show what other issues an interest group supports or opposes by its support of politicians sympathetic to the group’s stated interests.

4 Partisan leanings of interest groups

Table 4 shows that partisan leanings of interest groups do not explain all of the correlations among interest group ratings of candidates, but the interest groups in this paper’s data set do have strong partisan leanings, as Table 5 shows. The table shows the results of regressions using the group’s rating, GroupRating, as the dependent variable and a binary variable R given the value of 1 for Republican candidates and 0 for Democrats, estimating

$${\text{GroupRating}} =\upalpha +\upbeta{\text{R}} +\upvarepsilon$$

(4)

The number of observations listed in the table is the number of candidates that each interest group has rated. For every group, the t-statistics are highly significant and with the exception of Stand Against Spying, the R²s show that party affiliation is closely correlated with interest group ratings.

Table 5 Effect of party on interest group rating

There is no denying that many interest groups, including all those listed here, have strong partisan leanings. The previous section shows, however, that even with these very partisan interest groups, the relationships among their ratings of candidates are based on more than just partisanship.⁶ This makes sense when one considers the motivation for forming interest groups. If all interest groups were strictly partisan, individuals could just support parties rather than special interest groups to promote their interests. The fact that interest groups are not strictly partisan is the motivation for supporting the interest group rather than a political party.

5 Endorsements

This section uses the same methodology used in the previous section to analyze interest group ratings to look at the relationship of endorsements among interest groups. Endorsements are binary—either a candidate is or is not endorsed, so the results in Tables 6 and 7 are from probit regressions. Table 6 looks at the relationships among endorsements in bivariate regressions, again signifying relationships positive and significant at the .05 level by + and negative and significant with a −, with 0 representing not significant at the .05 level. In most cases, there are statistically significant relationships that, not surprisingly, are similar to those in Table 3 which examines relationships among ratings. Note that Stand Against Spying is not included in the table because they rate candidates but do not endorse any.

Table 6 Correlation of interest group endorsements significant bivariate regression results

Table 7 Correlation of interest group endorsements significant regression results controlling for party dependent variable on left

There are a few interesting differences between the results in Tables 3 and 6. For example, there is no statistically significant relationship between endorsements by the National Rifle Association and the Gun Owners of America, although there is a positive statistically significant relationship in their ratings. Meanwhile, the endorsements of the NRA and the Chamber of Commerce have a positive and statistically significant relationship, although the relationship between their ratings is not significant.

Table 7 adds a second binary variable that controls for party affiliation, and the statistical significance of most of the interest group coefficients disappears. Comparing Tables 7 with 4 shows that endorsements are much more closely aligned with party affiliation than are ratings. The majority of the ratings coefficients remain statistically significant after accounting for party affiliation. While Table 7 is included to show the effect of controlling for party affiliation, the results in Table 6 are more relevant to the paper’s thesis that the supporters of one interest group are supporting a market basket of other interest group issues that have no direct relationship with each other. Table 7 is included to show that interest group endorsements, a binary variable like party affiliation, often do not retain statistical significance when accounting for party affiliation.

6 When interest groups endorse a candidate, they support many issues

Interest groups rate candidates to provide a measure of how closely candidates’ positions conform to the goals of the group. An endorsement is stronger than a high rating, because it means the interest group supports the election of that candidate. Perhaps the best measure of the support an interest group gives to non-group issues would be to look at the correlation between that group’s endorsements—candidates they actively support—and the ratings given to those candidates by other interest groups. Table 8 shows the results of probit regressions using the binary variable of an endorsement by a group as the dependent variable, listed in the far-left column, and the rating given by the interest group to that candidate as the independent variable. As above, separate regressions are run for all pairs of groups, so where Endorse is the endorsement by an interest group and Rating is the rating by a group,

$${\text{Endorse}} =\upalpha +\upbeta{\text{Rating}} +\upvarepsilon.$$

(5)

Rather than listing a +, −, or 0 as in the above tables, each entry is the z-statistic on β. This makes the table more difficult to quickly read and interpret, but gives a better indication of the statistical significance of the relationships.

Table 8 Effect of other group ratings on interest group endorsement probit estimation: dependent variable on left

The table shows that most of the relationships are statistically significant at the .05 level. Looking at the first row, for example, candidates endorsed by the NRA all have ratings from other interest groups that are highly statistically significant, except for Stand Against Spying. People who support the NRA by voting for endorsed candidates or by joining and financially supporting the NRA also are supporting the anti-abortion movement and business interests, and are opposing teachers’ unions, the League of Conservation Voters, and the Alliance for Retired Americans. Supporting the NRA means supporting a broader package of issues.

Note that an endorsement by the Gun Owners of America, shown in the second row, has no statistically significant relationship with a candidate’s NRA rating, which is surprising, as one would think that candidates endorsed by the Gun Owners of America would be highly rated by the NRA. Endorsements from four of the nine groups—the National Right to Life, the League of Conservation Voters, the Human Rights Campaign, and the Alliance for Retired Americans—have statistically significant relationships with the ratings of all the other groups. The table compares each group’s endorsements with the group’s own ratings, and another curious result is that an endorsement from the National Education Association does not have a statistically significant relationship with the NEA’s ratings of candidates.

Table 8 has 90 coefficients in it, and 73, or 81% of them, are statistically significant at the .05 level. Two interest groups, the National Education Association and Stand Against Spying, account for ten of the 17 insignificant coefficients, so dropping the SAS column (SAS rates but does not endorse) and the NEA row leaves 72 coefficients with 65 of them—90%—significant at the .05 level. The market basket analogy holds less for the NEA and SAS, but for the other interest groups, supporting endorsed candidates also means supporting (or opposing) many issues that are peripherally related, at best, with the group’s stated goals.

7 Conclusion

One way that choice in the political marketplace is different from choice in markets for goods and services is that support of a politician means supporting a platform that consists of a bundle of policy positions. Supporting a politician on one issue also furthers other issues in that bundle, regardless of whether supporters intend to support the whole bundle. The evidence presented here shows that not only is this a theoretical possibility, it is an empirical reality. There is a very strong correlation between interest group ratings and endorsements of political candidates, both positive and negative. Interest groups that support politicians based on that group’s interests also, as a secondary effect, support and oppose other interests unrelated to the groups’ stated purposes.

Table 3 shows the close relationship among the ratings of different interest groups for political candidates. Support for one interest group’s highly-ranked candidates really does imply support for a market basket of issues not closely related to the interest group’s stated goals. Table 4 shows that this relationship remains even when party affiliation is taken into account. Interest group ratings do tend to follow party lines, however, as Table 5 shows. There is also a strong correlation among endorsements of different interest groups, Table 6 shows, although Table 7 shows that the relationships between endorsements weakens after taking into account party affiliation. Table 8 shows that supporting a candidate based on the endorsement of one interest group also implies supporting some issues and opposing others that are at best only peripherally related to that group’s stated agenda.

The interest group ratings and endorsements examined here do line up on a one-dimensional left-to-right issue space to a large degree, and Table 5 shows the strong correlation between interest group ratings and the party affiliations of the rated candidates. While this suggests the possibility that interest group members might be sympathetic with the larger market basket that they are secondarily supporting—which is undoubtedly true for some members—the effect remains after party affiliation is taken into account, and undoubtedly there are some members who contribute to an interest group but do not share all of the political views they unintentionally support as a secondary effect.

The paper began with a shopping cart analogy suggesting the theoretical possibility that by supporting an interest group on one issue, supporters also support other issues as a secondary effect. The primary purpose of this paper is to show that there is strong empirical evidence that this theoretical possibility is an empirical reality.