With actual votes cast, it’s now possible to understand the relationship between support for a candidate and demography. One way we can visualize this is a correlation matrix. The following visualizes the correlation between the percentage support for a Republican candidate at the county level and various characteristics of that county found in the Census’s American Community Survey.

The matrix naturally clusters correlated variables together. If the intersection of two variables on the X and Y axis is blue, those variables are highly correlated. If it’s red, they’re negatively correlated. If it’s white or a very light color, there is little relationship. Clusters can be visualized both by groups of blue boxes, and the borders that are drawn at various points throughout the matrix. Candidate support variables are interspersed with demographic variables; this means we can see both which other candidates and which demographics a candidate is correlated with.

The upper left corner captures the conservative lane of the primary, most prominently Ted Cruz and Ben Carson. Key variables included the percentage of married families, the percentage of people born in state, higher median age, and more high school graduates.

The large cluster in the center is the center-right, establishment-lane. Support for Marco Rubio is more correlated with a higher number of incomes above $200,000, a higher number of people who moved in the 2000s, more new home construction, more postgraduate degrees and college graduates.

Donald Trump support in Iowa was associated with counties with more unemployment and more who list their ancestry as Scots-Irish or “American.”