June 26, 2014

Q/A: How Pew Research created the political typology

Scott Keeter, Director of Survey Research, Pew Research Center

The Pew Research Center has been doing its Political Typology reports every few years since the first one in 1987. The typology study published today is a follow-up to our report on political polarization in America and draws on the sample of 10,013 adults nationwide surveyed earlier this year. We asked Scott Keeter, Director of Survey Research, to explain why — and how — we do this particular analysis.

Why did you create the political typology?

The goal of the political typology is to sort people into homogeneous groups, based on their political values and attitudes. It’s an effort to categorize people politically to help us better understand the complexities of the current political landscape. This kind of “clustering” gives us a way to quickly summarize a large amount of complex information, to understand differences among people, and to help predict how people are going to act in the future.

The urge to categorize is so strong that it is almost a part of being human. People have long tried to classify the things they deal with in life, whether we’re talking about animals, vegetables, minerals, diseases, songs, celestial objects – but we especially like to classify humans. 

Isn’t the distinction between Democrats and Republicans all we need?

Political parties are one kind of grouping of like-minded people. But parties – at least in the U.S. – tend to be large agglomerations, with schisms and factions. Moreover, a growing number of people do not think of themselves in partisan terms, though they still have strong political views. Groups in the political typology may have a recognizable partisan tilt, but they are more internally consistent than the major U.S. parties. Understanding this pattern may help us understand the tensions in the two-party system and where there are political agreements or disagreements that are masked by the partisan lens through which most politics is viewed.

Do the groups in the political typology correspond to real-world groups?

There is considerable overlap between the most ideologically distinct groups in the typology and the two major political parties: 86% of Steadfast Conservatives and 84% of Business Conservatives think of themselves as Republicans or lean Republican, and 89% of Solid Liberals think of themselves as Democrats or lean Democratic. But many of the groups who occupy the middle of the political spectrum contain a mix of Republicans and Democrats, and don’t necessarily correspond to organized groups in the political world. One reason for this is that politics is a peripheral concern for most people, especially for people who hold a mix of liberal and conservative views. The lack of interest in politics, as well as a sense of futility regarding the ability to influence political decisions, reduces the incentive to organize and act, even when a relatively large number of people may have similar political views. But this can be a vicious cycle – people who feel they are “outside” of the red vs. blue debate may feel that nobody shares their views. The typology shows that that is not the case.

How do you find the groups in your data?

If we were trying to group people on the basis of only two or three issues or values, it would be relatively easy. But the complexity increases as the number of attitudes and values we want to consider increases. And so we use a statistical technique to identify groups of people who hold similar values and opinions across a large set of questions asked in the survey – 23 in the current typology – each of which gets equal weight in the process. This technique is called “cluster analysis.” Given the complexity of the data, it’s possible to sort respondents into many different combinations, depending on how large and how homogeneous we want the groups to be. Unlike many other statistical techniques, cluster analysis does not produce a single “correct” result. Instead, we run numerous versions of it (e.g., asking it to produce different numbers of clusters) and judge each result by how analytically practical and substantively meaningful it is. Fortunately, nearly every version we produced had a great deal in common with the others, giving us confidence that the pattern of divisions we found were genuine.

Does this typology differ from prior ones? Has the ideological makeup of the U.S. public really changed that much, or are the criteria different, or is there some other reason?

There’s actually more continuity than change in the typology when you look back over the past 27 years. We’ve nearly always found two strongly conservative groups that are divided over social issues. Similarly, we have usually found a group of across-the-board liberals at the other end of the spectrum. In the middle, a group of financially hard-pressed and sometimes disaffected individuals usually emerges, along with one or two younger, more upbeat groups.

Most of the questions that went into this year’s typology are the same as in the past. But not everything is the same. We added a few items to reflect the fact that the political agenda has changed somewhat over time. These include questions on marriage and family, the fairness of the economic system, personal privacy vs. protection from terrorism, and the importance of U.S. involvement in international affairs.

Two other changes were made in the statistical analysis. One is that we used individual survey measures rather than scales as inputs to the typology. Another is that we omitted measures of party affiliation and financial stress, which were used in previous typologies.

You are missing my political group. Why is that?

If your group is distinctive because of issues and values other than the ones we asked about, we probably couldn’t find you. Similarly, if your group is smaller than our average group (between 10%-15% of the general public), you wouldn’t have been discovered by our method.

One group that many people have asked about are libertarians. We actually devoted a great deal of effort in trying to identify libertarians, and in an upcoming report we will discuss this effort and describe how libertarians fit into the typology. The shorter answer is that many members of the Business Conservative group would qualify as libertarian, if by libertarian we mean a preference for minimal government involvement in economic and personal affairs. But unlike many libertarians, most Business Conservatives favor the use of overwhelming force against terrorism, think the U.S. should be active in world affairs and believe that military strength is the best way to achieve peace. Similarly, a plurality of Business Conservatives opposes allowing gays and lesbians to marry legally. For these and other reasons, we decided not to describe this group as libertarian. Similarly, it’s likely that there are libertarians among the Young Outsider group, who are socially liberal and fiscally conservative. But many of them favor regulation to protect the environment, and thus don’t fit easily into the libertarian category.

Topics: Research Methods, Political Attitudes and Values, Political Typology, Political Polarization

  1. Photo of Scott Keeter

    is a senior survey advisor at Pew Research Center.


  1. David Niday2 years ago

    I consider myself a Libertarian, and yes this survey did label me as a “young outsider”. Still it was an interesting survey and it makes us think about multiple issues.

  2. JASON WILLIAMSON3 years ago


  3. Kevin Bess3 years ago

    I really liked this analysis. I first learned about cluster analysis when I read the MIT sloan sports analytics paper “from 5 to 13”. It said there were actually 13 different playing styles (positions) in the nba as opposed to the normal 5 positions. This is really useful to show how a person falls based on multiple spectrums not just on the liberal – conservative scale. I am most similiar to the Next Generation Left. I calculated my distance from each group based on my responses and I fell in between Next Generation Left, Solid Liberals, and The Young Outsiders. I think it would be interesting if you guys would try to place political figures into the categories. Or maybe in a year you could place the 2016 presidential candidates into the categories. Very good work!

  4. Bill Sauber3 years ago

    I was very surprized that the quiz did not include any questions about taxation. It asked about corporate profits and about social safety net spending versus going further into debt, but nothing about revenue. Taxation, and the change in attitude by the Republican Party was the single largest reason that drove me from that party. It is perhaps Republican politicians’ single most rigid ideological stance, and the one that keeps them most frimly in the pockets of the folks that are bribing them (Oh, excuse me, I meant to say “making campaign conturbutions.”)

  5. Mark E Miller3 years ago

    It is interesting to compare the 2011 and 2014 typologies side-by-side:

    Pew Typology

    2011 2014

    Mostly Republican
    Staunch Conservatives 9 Steadfast Conservatives 12
    Main Street Republicans 11 Business Conservatives 10

    Mostly Independent
    Libertarians 9 Young Outsiders 14
    Disaffecteds 11 Hard-Pressed Skeptics 13
    Post-Moderns 13

    Mostly Democratic
    New Coalition Democrats 10 Next Generation Left 12
    Hard-Pressed Democrats 13 Faith and Family Left 15
    Solid Liberals 14 Solid Liberals 15

    Bystanders 10 Bystanders 10

    100 101

    Now I understand the cluster boundaries will not be exactly the same, that is the whole point of re-doing the typology every few years. The Post-Modern cluster has disappeared, most of them divvied up, I guess, between Next Generation Left and Young Outsiders.

    But it is interesting to see the Dem-leaning groups have gone up 5% over three years, and the R groups up 2%. Of course, two of the previous Dem-leaning groups have been put in the middle category now, so they are not as strongly Democratic, but still, the most noticeable trend here is the increasing polarization of the electorate.

    I answered two of the questions opposite to the “liberal consensus”, but ended up plotted at the extreme left edge of the ideological spectrum, (correctly) classified as a Solid Liberal. I guess this means that very few must be answering ALL of the questions in the same ideological direction?

  6. Joe Stewart3 years ago

    Poll also divides conservatives with arbitrary categories – and refuses to do the same with liberals. It diminishes the ability to see equity in the process.

    1. Scott Keeter3 years ago

      Thanks for your comment. Just a couple of thoughts. One is that the statistical procedure we used (cluster analysis) creates the groups without our intervention. It’s true that we looked at a lot of different versions of the cluster analysis, but the consistent pattern across them was that we found one solid liberal group and two strong conservative groups, divided by the social issues and attitudes about business. The other thought is that while the divisions among the conservatives have attracted a lot of attention from columnists and bloggers, we did try to describe the cleavages on the left as well, since there are three left-leaning groups; see the section in the overview entitled “Divisions on the Left.”

  7. Richard Patton3 years ago

    The two responses allowed for each section were far left or far right and I assume that most respondents wee not comfortable with either selection. Most would select the response closest to their personal views and this would tend to make them more liberal or conservative than they actually are. This leads me to believe that the poll was designed to produce a divisive outcome which only serves to perpetuate the idea that most Americans hold extreme views.

    1. Vince Clancy3 years ago

      It also perpetuates the pollster’s idea that we need polls, period! The fact that they may be seeking the extremes is a good point. Unfortunately, the lazy American voter goes along with polls and uses that lazy way of not having to check out candidates and issues on their own.

    2. Scott Keeter3 years ago

      We know that a lot of people feel frustrated by the “forced choice” format of the questions used in the quiz. Here’s a link to blog post we wrote about why the questions are asked as they are: pewresearch.org/fact-tank/2014/0…