Ideological Warriors and Constituent Servants: Understanding the Determinants of Participation on Health Care Legislation in the Polarized Senate

Abstract

Recent scholarship on the Senate indicates that partisanship and polarization have fundamentally changed the dynamics and the nature of policy making in the institution. To understand how senators balance their roles as constituent servants and partisan warriors, we examine senators’ participation in floor amending on major health legislation over a 10-year period (2000–2010). Health care is central to the ideological divide over the nature of the welfare state and has a significant impact on constituents. Unlike previous studies, we develop a fuller picture of the factors that motivate senators to participate by analyzing both amendments filed to a bill and amendments considered on the floor. We find that minority party status and ideology, both the liberal-conservative dimension and ideological extremism, are important indicators of participation. Furthermore, senators elevate ideology in comparison to constituent need when an issue is a presidential priority. Finally, the influence of constituency factors varies by party. For Democrats, their ideological commitment to the social welfare state makes them more likely to participate on health issues regardless of constituent need. However, Republican senators become more likely to sponsor amendments as the level of observable constituent factors increases.

Notes

¹Amendments considered on the floor are obtained from the bill summaries in Thomas, the legislative information website of the Library of Congress, www.thomas.gov. Amendments that are filed but not considered on the Senate floor are printed in the Congressional Record. We eliminate duplicate amendments that are both filed and later considered on the floor or filed multiple times.

²We focus on constituency characteristics that affect a broad swath of voters across states. We do not examine the impact on business interests in the state such as health insurance companies or medical device manufacturers. An examination of these interests is beyond the scope of this article and would require an examination of the lobbying efforts and campaign donations of specific groups.

³We utilize a negative binomial rather than a Poisson event count model because the negative binomial uses a dispersion parameter to account for the fact that events may not be independent of each other as members who sponsor one amendment may be more likely to sponsor additional health-related amendments. A dispersion parameter of zero indicates an absence of dispersion and independence of events while a dispersion parameter greater than zero indicates overdispersion (King 1989).

⁴DW-NOMINATE scores are available at http://www.voteview.com.

⁵We also ran models (results not shown) without the ideology scores to address concerns that the preferences measured by NOMINATE scores are endogenous because party strategy and constituency interests influence these measured preferences. Excluding NOMINATE does not yield substantive changes to the observed constituency variables of interest or change the significance of the party variable in the full models.

⁶Because many of these measures are state-level measures and consistent for both senators, we also ran models that clustered the standard errors by state. However, due to negative intra-cluster correlations, the clustered errors on the Patients Bill of Rights were smaller, generating more statistical significance on several variables (Arceneaux and Nickerson 2009). For the other two bills, Medicare Part D and the Affordable Care Act, the results did not change substantially. Therefore, we present unclustered errors that represent more conservative results.

⁷We also ran the models without the percent of African American and Hispanic population variables out of concerns about correlation with other variables in the model. The results for Medicare Part D and the ACA remain the same. However, the party-constituency interactions are no longer significant with the Patients’ Bill of Rights, yet constituent enrollment in HMOs remains significant. The Patient's Bill of Rights models are more sensitive to specification due to the lower number of amendments proposed and the removal of significant demographic variables.

⁸A variable measuring seniority was also tested and had no significant impact.

⁹We do not include the Patients Bill of Rights considered in the 106th Congress because there were only 21 amendments. We do not pool the sample because we cannot test our theories about the impact of majority and minority party status with the pooled sample since Republicans controlled the Senate in the 106th Congress and Democrats in the 107th Congress and there was a Democratic president in the 106th Congress and a Republican in the 107th. Therefore, we limit our analysis to the bill in the 107th Congress.

¹⁰A GOP-backed prescription drug benefit proposal passed the Republican-controlled House in both the 106th and 107th Congress. However, both bills died in the Senate.

¹¹The absolute value of DW-NOMINATE is used in the majority of the models for the ACA because it is consistently a more powerful predictor of sponsorship behavior while the raw measure is more important for the Patients’ Bill of Rights and Medicare Part D.

¹²Democratic vote share is a consistent negative predictor of sponsoring amendments. Interpretation of this constituency factor is complicated by its high correlation with partisanship and ideology. Yet the negative result holds even when ideology and party are excluded from the models. For the Patients Bill of Rights we expected Democratic vote share to be negative because Democrats are the majority party and senators from more Democratic states should be less likely to amend the bill. Similarly, the Democratic vote share is negative but not statistically significant as a predictor of which senators offer amendments to the Affordable Care Act, likely indicating that senators from more Democratic states are less likely to offer amendments after accounting for the amending activity of the extreme liberals. For the Medicare Prescription Drug bill, minority party Democrats are more likely to offer amendments; therefore, we can only speculate that once liberal ideology is accounted for, the state Democratic vote is picking up the fact that conservative Democrats and moderate Republicans were less likely to amend the bill.

¹³Figure 3, which plots the partisan differential in amendment sponsorship by the percentage of insured individuals in HMOs, is the only difference that fails to reach a conventional level of significance (p = 0.12). It is signifcant at the 0.10 level in the more parsimonious model. Given the sample and the small number of amendments offered overall, we include it in our results and argue this difference is in line with the other findings.

¹⁴Another trend worth noting is the consistent positive and significant impact of gender. Across the three bills women are more likely to offer amendments to health bills and the impact gains strength as women increase their presence in the Senate over time. The enhanced activism of women on health issues is consistent with research that finds that female senators are more interested in social welfare policy. Moreover, voters trust women on health policy, thus making them better messengers for their parties on health issues (Swers 2002, 2013). Since the majority of female senators are Democrats, the strong ideological commitment of Democrats to the social safety net likely also plays a role for these women.