This narrative is badly skewed. In fact, scholars do not agree that the American public has actually become more polarized. Our new research, forthcoming at the Journal of Politics, shows that although Americans disagree about many political issues, they have not become more polarized over the past 60 years.
Our research takes advantage of everything we know about people’s issue positions from the longest running and most respected survey in political science, the American National Election Studies. Unlike previous research, which tends to focus on changes in a few policy attitudes between two time periods, we use all of the survey questions about domestic policy that are available to us in 27 different surveys over 60 years.
We applied the same widely accepted method that is used to measure polarization in Congress to average Americans. The method is called DW-NOMINATE, and the researchers who developed it have created some of the most famous graphs in political science. These graphs show us how roll call voting in Congress has polarized over time.
Like DW-NOMINATE, our approach takes account of several key facts: some political issues are more indicative of people’s ideology than others; some responses are more ideological than others; the types of issues asked about in surveys change over time; and we can be more certain about the ideology of people with consistently conservative or liberal responses.
The figure below shows just how little the public has polarized compared to U.S. senators. It shows trends in the dispersion of people’s policy views (specifically, the standard deviation). If polarization were getting worse, people should be moving away from the center toward the ideological poles — that is, they should be dispersing. So if the public is polarizing, the standard deviation of their views should be growing over time.
But for the public (left side of the graph), the level of polarization shows no apparent trend over time. Pick your years carefully and you can get the conclusion you want, but the overall picture is stability. Compared to the swiftly polarizing Senate, the public is standing still.
This result may be surprising. If the people’s representatives are polarized, shouldn’t the people, who elected those polarized representatives, be polarized as well?
Consider one possible explanation: what political scientists call “party sorting.” Party sorting means that the composition of the parties in the American public has changed. In the middle of the 20th century, each party was relatively diverse. Both parties had liberals and conservatives, so a voter’s policy views did not say much about their preferred party.
Over the past 50 years, however, the parties have sorted so that now most liberals are Democrats, and most conservatives are Republicans. The result is much more homogeneous parties, even though the ideology of the public as a whole has not changed very much.
The graph below shows how much the differences between people (or senators) reflects differences between Republicans and Democrats. “1” would mean perfect sorting into two distinct parties, and “0” would mean that the parties are no different from one another.
However, the Republican and Democratic Parties in Congress have sorted much faster than the public. Moreover, even back in 1950, they were already better sorted. Parties in Congress are more homogenous than parties in the American public.
So why are so many convinced that the American public has become more polarized? One reason is that people mistake party sorting for polarization. This was true in the reactions to the Pew report, most of which actually concerns party sorting, as political scientist Morris Fiorina noted.
Another is that people mistake ideological consistency for polarization. In the Pew report, the authors pointed out — but many commentators apparently failed to observe — that the key metric in the report actually measures “consistency,” not polarization. Since 1994, Pew’s surveys have asked 10 policy questions and allow respondents to choose between two options on each question. The more liberal or conservative options you choose, the more ideologically consistent you are.
Our analysis is different in that our timeframe precedes 1994, we include a much larger number of questions, and most of these questions allow respondents to pick policy views from more than only two options. In other words, our data typically allow respondents to choose options that are “in the middle,” and often they do.
Thus, our data and methodology can map where respondents are located along the liberal-conservative continuum. When we do so and extend our view back into the 1950s, it is clear that polarization changes from year to year but has not moved in any particular direction.
In the meantime, every Congress appears to be more polarized than the last one. There are many potential explanations for this important phenomenon, but our research suggests it is not due to polarization in the policy preferences of the American public.