Conditions for “Primarying”: Sorting and Congressional Polarization in US

• In Event: Poster Session: Formal Political Theory; Political Methodology & Canadian Politics
  In Poster Session: Poster Session: Political Methodology

Abstract

This paper examines the impact of electorate sorting on congressional polarization in US. First, I identify the conditions under which candidates are “primaried” at the district level. (1) district ideology and (2) district share of partisan identifiers are the two observable parameters that characterize the conditions under which more or less extreme candidates are nominated from the House primary elections in a given district. In particular, the average (or median) ideology of the general electorate in the district determines the interdistrict extremity of the House primary nominees. The perceived majority status of a party in a district, on the other hand, should decrease the strategic voting incentive among the primary electorate of the party, thereby enabling them to nominate a more partisan (extreme) candidate than when the status of the party in the district is otherwise. Therefore, the share of the partisan identifiers in the district determines the intradistrict extremity of the primary nominees from each district party. I provide support for how well the two electorate parameters - district ideology and district share of partisan identifiers - predict the extremity of the primary nominees, by examining House primary candidate ideologies against Bayesian multilevel regression and poststratification (MRP) estimates of district ideology and partisanship (2006-2016).
Next, I examine how the electorate sorting observed at the national level is associated with these district level conditions. Specifically, the steady increase in electorate sorting, construed as a dual process of stratification of partisan identification (partisan sorting) and geographical segregation by ideology (geographical sorting) among the voters, has fostered the conditions favorable for extremist nomination in US districts. The geographical sorting is the direct manifestation of the increase in interdistrict extremity of the district ideologies. Partisan sorting, provided some level of geographical sorting, then fortifies the majority status of a party in each district, i.e., increases intradistrict extremity of the nominees. Thus, the interplay of these two sorting processes has contributed to increased level of polarization between the nominees from the Democratic and Republican parties at the national level. The paper develops measures of partisan and geographical sorting and shows this relationship from the data.
The estimation of causal effects of electorate characteristics on political candidates, however, is frequently recognized as challenging due to vastly potential omitted variable and simultaneity biases. This paper develops and applies a particular blip-down estimator to evaluate the impact from potential sources of bias on the estimates and provide robustness to the conclusions of the paper. In the first step, I decompose the causal variables (e.g., district’s average ideology and partisanship) into exogenous and endogeneity-ridden components by manipulating the poststratification process in multilevel regression and poststratification (MRP) estimation using population and migration data. In the second step, I apply a blip-down estimator coupled with sensitivity analyses to these decomposed variables. The procedure retrieves a narrower range of bias-corrected estimates than other existing procedures.
A notable conclusion of the paper is that the data generating process for the effect of electorate sorting on congressional polarization is more complex than commonly understood. Of particular note is that the impact of electorate sorting is not linear. Thus, even in the steady trend of electorate sorting, the rate at which congressional polarization progresses is expected to decrease over time.

Author

• Aram Kim