Pew Center and Electorate Groupings

Pew uses opinion polling to identify clusters of like-minded voters, splitting the Republican supporters into "Enterprisers, Social Conservatives, Pro-Government Conservatives", Democratic supporters into "Liberal Dems (formerly 60s Liberals), Disadvantaged Dems, and Conservative Dems", and Independents into "Upbeats, Disaffected and Bystanders". This splitting is essential for us to understand who makes up the Rep & Dem coalitions, and where the fault lines are within the parties.

The Pew typologies give us a more detailed underlying structure to the political spectrum. Pew Typologies are also designed to have long-term stability so they can track political values over the years. One of the most intriguing results from the Pew study is the growth of their category "Liberal Democrats" (aka 60s Democrats), which have doubled from 9% to 18% of the populace in the past 5 years.

Beyond "Beyond Red vs Blue"

I've been carefully re-reading the Pew Typology Report. The report shows clearly how aggregate political affiliation gives us much less information than if you split the electorate by Demographics, Values and Geography. Particularly important in the cross-tabs are the "Three Rs": Race, Religion and Region. Another interesting point is that class or economic status in isolation doesn't explain much (i.e. there are plenty of well-off Democrats, and not-so well off Republicans), but if you dis-aggregate the electorate by the typologies and/or combine income with one of the Three Rs, then you start to see interesting patterns.

I wanted to look in more detail at Pew's regional cross-tabs, in order to answer questions like "what percentage of Westerners are "Liberal Democrats", so I used Excel to recalculate the Pew Typologies by Region table. Instead of looking how each typology splits vertically across the regions, I made a table showing how each Region splits horizontally across the typologies. Second, I wanted to know which regions had more or fewer people within a given Typology Group from what you would expect from the national averages. In the Pew Typology Differences table, I've flagged big negatives with red, big positives with blue and medium differences with bold.

Here are the tables:

Three huge things stand out for me:

(1) Outside the South, the Republican Party is not that strong. To win elections they need a highly motivated base, un-motivated opposition, national crisis to whip up patriotism. It is admirable how carefully the Republican Party nurtures their coalition: Religious rhetoric for the social conservatives, Corporate give-aways to the business conservatives, Aggressive foreign policy for the Pro-government conservatives and the Democratic Conservatives.

(2) Liberal West: Look at the huge block of Liberals. Also, the three "conservative" categories (Rep & Dem) are much lower than nationally. As a result, 27% of the Western electorate and 2/3 of Democratic Party supporters are strong Liberals. This is also the Pew Typology Group that doubled in the past 5 years. I wonder why the Party hasn't caught on or caught up with these voters? Why are they running from the Liberal label?

(3) Conservative South: The Southern electorate is much more conservative than the National average for both parties. If I were a Southern Democratic leader, I would be concerned about national vs regional messaging. How would I appeal to Upbeat Independents?

Wonk Alert

Note: I didn't compute the Pew Typ Diffs table directly from the Typ by Region table. It is calculated as a percentage of measured population compared to the expected, i.e. the difference from the national average. Since Pew interviewed 2,000 people their polling accuracy is below 2 percent; simply transforming from vertical to horizontal has good accuracy; the difference table (expressed as a percentage of "measured" from "expected") is a lot less accurate, so you should only view the percentages in terms of big, medium or small differences.

The Pew polling data is normally presented by column percentages (each column adds up to 100%). To get the percentages across the rows, I first calculated the "Measured table": multiply each Typ-group by its (columnar) regional percentages; now adding up all the data in all bins of the table gives you 100%. The Expected population table (E) is easy to calculate from the Regional populations (R) and the national Typology distribution (T): R x T = E.