Polarization in American Politics

12th September, 2019

Recent U.S. media coverage has been saturated with partisan squabbles. While browsing through the news, it is not so hard to find an article covering how Democrats and Republicans are divided on countless issues. Public Religion Research Institute (PRRI) reports that in 2018, 91% of Americans agree that the country is divided on politics, while 83% agree that the country is divided on race and ethnicity, and 77% agree that the country is divided on religion. Has it always been like this? In order to understand whether political polarization is a recent trend or a phenomenon inherited from the past, we must turn to data.

Generally, political polarization is defined as the divergence of ideology between two different parties, and the convergence of ideology within each party. There are two aspects of political polarization that we must consider: Elite polarization and popular polarization. Elite polarization refers to the political polarization of the elected officials and party elites, while popular polarization refers to such polarization in the general public. In this post, I'll attempt to show a recent evolution of ideology among the party elites and the general public, and note some important trends.

Elite Polarization:

The most popular measure for the ideology of the party elites is DW-NOMINATE (Dynamic Weighted Nominal Three-step Estimation) scores by Keith T. Poole and Howard Rosenthal. To put simply, they have taken advantage of all the roll-call votes of the congress to pinpoint the relative ideal points of the members of the Congress on economic issues. Figure 1 shows the evolution of ideology of the member of the congress according to the DW-NOMINATE scores. Higher DW-NOMINATE scores indicate that the member is more conservative on economic issues, compared to other members in the Congress (for more details on DW-NOMINATE scores, see voteview.com).

It is noteworthy that the gap between the median Democrats and median Republicans decreased near 1940s, and has been on increasing trend afterwards. To see this more clearly, Figure 2a shows the distance between the median Democrats and median Republicans in the Senate and the House of Representatives. From 1900s to 1950s, there has been a convergence of ideology between the two parties, followed by sharp divergence of ideology from 1970s to 2010s in both Congress. This recent increase in elite polarization over the past few decades has been noted by many scholars (see for example Poole and Rosenthal, 1984; Coleman, 1997; Fiorina, 1999; Roberts and Smith, 2003 among many others). To observe the convergence of ideology within each party, Figure 2b plots the variance of ideology in Democrats and Republicans. While the variance of ideology for Republicans has mostly remain stable for the past few decades, Democrats have become more homogeneous in their ideology from the 1960s.

Popular Polarization:

To measure the popular polarization, I used the questionnaire data provided by ANES (American National Election Studies). I followed the approach taken by Ansolabehere et. al. (2008) and Zingher and Flynn (2015) to retrieve the ideology scale using factor analysis for the year 1996 and 2016 (I've only chosen those two years for the sake of time, but in the future I may come back to analyze more years). Using the responses for a wide set of questions, ranging from issues like government spending, insurance, foreign aid, immigration, abortion, affirmative action, etc., I ran a factor analysis to retrieve one underlying factor that represents the ideology of the respondents. Standard statistical analysis before running factor analysis all showed that the dataset is suitable for factor analysis. Bartlett's test of sphericity resulted in the p-value less than 0.05 for both year 1996 and 2016, rejecting the hypothesis that the correlation matrix is an identity matrix. Kaiser-Meyer-Olkin measure of sampling adequacy indicated that the strength of the relationships among variables was high (0.82 and 0.95 for year 1996 and 2016 respectively).

Factor loading calculated from the factor analysis was then multiplied with the responses that has been ordered and standardized to have mean 0 and variance 1, with higher number indicating a more conservative response. Figure 3a and 3b shows the distribution of the ideology uncovered through the factor analysis. Note that the percentage of self-claimed Republicans who were more ideologically liberal than median Democrats has decreased from 11.1% in 1996 to 5.2% in 2016, while the percentage of self-claimed Democrats who were more ideologically conservative than median Republicans has decreased from 13.6% in 1996 to 5.1% in 2016. In the appendix, I have added additional graphs for the distribution of ideology calculated by simply adding the ordered and scaled survey answers. The result is not very different from the ideology calculated from factor analysis. Does this mean that the distribution of ideology for Democrats and Republicans have shifted away from one another? The answer is not so straightforward.

While I have tried to select similar sets of questions for both 1996 and 2016, inevitably there were some questions that were not present in both years that I have included in the analysis. This is simply because over the past two decades, political issues that divide the Democrats and Republicans have changed. Zingher and Flynn (2015) tackled this issue of cross year comparability by setting one specific question as a common benchmark, rotating the factor loading matrix through this benchmark question. However, since the survey answers have been scaled, it is hard to conclude that the distributions have absolutely shifted from one another. Rather, we can conclude that in 2016, general public were able to identify the party that fits their ideological interest better than the public in 1996. Fiorina (1999) commented that there is some evidence that party identifiers have moved apart, but no sufficient evidence has been presented to show that the population as a whole has become more polarized.

Discussion:

While many literature agree that elite polarization has been ongoing for the last few decades, evidences of popular polarization are not as clear. However, is political polarization necessarily evil? Looking back at the 1950s, the era of bipartisanship, many people decried the inability of political parties to present clear distinctions from one another. Taking centrist position in the American politics made sense in light of median voter theorem as mentioned by Downs (1957). However, the atmosphere in the American politics is not quite the same today. Rather than taking the middle-of-the-road approach, both parties have been fighting hard to create sufficient 'product differentiation' to gained the hearts of their party supporters. Hetherington (2001) and Zingher and Flynn (2015) asserted that such elite polarization has helped the general public to better understand the parties' stance on different issues, and to support the party that best suits their interests. Perhaps the shift of the ideology distribution of the general public in ANES reflects this argument.

However, increased elited polarization does not come without any cost. Downs (1957) argued that when the political parties are polarized, regardless of which party is in office, half the electorates would feel as though other half is imposing policies that are strongly repugnant to them. If one party keeps getting re-elected, other half would be disgruntled and may revolt. However, if the two parties alternate, government policies will keep shifting from one extreme to another, resulting in social chaos. Poole and Rosenthal (1984) also posited that such differentiation has caused interests of middle-of-the-road voters to be underrepresented as well.

References:

Ansolabehere, S., Rodden, J., and Snyder, J. M. (2008). "The Strength of Issues: Using Multiple Measures to Gauge Preference Stability, Ideological Constraint, and Issue Voting". The American Political Science Review, 102(2): 215-232.

Coleman, J. J. (1997). "The Decline and Resurgence of Congressional Party Conflict". The Journal of Politics, 59(1): 165-184.

Downs, A. (1957). "An Economic Theory of Political Action in a Democracy". The Journal of Political Economy, 65(2): 135-150.

Fiorina, M. P. (1999). "Whatever Happened to the Median Voter?" Manuscript prepared for the MIT Conference on Parties and Congress, Cambridge, MA, October 2, 1999.

Hetherington, M. J. (2001). "Resurgent Mass Partisanship: The Role of Elite Polarization". The American Political Science Review, 95(3): 619-631.

Poole, K. T., and Rosenthal, H. (1984). "The Polarization of American Politics". The Journal of Politics, 46(4): 1061-1079.

Roberts, J. M., and Smith, S. S. (2003). "Procedural Contexts, Party Strategy, and Conditional Party Voting in the U.S. House of Representatives, 1971-2000". American Journal of Political Science, 47(2): 305-317.

Zingher, J. N., and Flynn, M. E. (2015). "From on High: The Effect of Elite Polarization on Mass Attitudes and Behaviors, 1972-2012". British Journal of Political Science, 48(1): 23-45.

Appendix:

Figure 4a and 4b shows the distribution of the ideology in the general public in 1996 and 2016, calculated by ordering the survey answers such that higher number represents more conservative answer, and then scaling the responses to have mean 0 and variance 1. I then added all the scaled responses, and again scaled the sum to have mean 0 and variance 1. Percentage of Democrats who were more conservative than median Republicans decreased from 12.4% in 1996 to 4.6% in 2016, and percentage of Republicans who were more liberal than median Democrats decreased from 16.1% to 5.2%.