A Tale of Three Primaries

With the most recent primaries in Indiana and North Carolina the pundit class has closed the books on the Democratic nomination, crowning Senator Obama with an unassailable lead. Of course, this is based on strategic calculations of the super delegates’ political logic rather than the clear preference of the voters. In this they are probably right, baring some unforeseen disaster for Obama. But there are some interesting details hidden in these votes we should carefully note before we leave the scene.

Most of our received wisdom on elections comes from exit poll sampling, often a highly subjective and error prone art. Exit polls also tend to bias explanations toward identity characteristics of voters—things like race, age, gender, income, and religion—because these data are more easily collected in surveys. A more fruitful and complementary approach is to compare votes to census data at the county level. This gives us a complete population dataset, rather than a sample approximation.

The last three primaries give excellent insights into the political mood of the country when compared to the last two presidential elections. In addition, they’ve occurred at a turning point in the Democratic primary season when both candidates have been tested and voters are paying close attention. While none of the individual states of PA, IN, or NC are truly representative of the nation (they all tip Republican), taken together their county profile is remarkably in tune with the average profile of all three thousand-plus US counties. This applies to mean population density (240 vs. 245 inhabitants/sq. mi.), median family income ($45K vs. $42K) percentage of white (86% vs. 85%) and black households (10% vs. 9%), median age (37 vs. 37), and percentages of married households (56% vs. 55%) and female heads-of household (10% vs. 10%). In other words, the 259 counties in these three states look remarkably similar to the national profile. So, taken together, they make an ideal sample for the nation as a whole.

What does this method of analysis tell us? First, it tells us that voting preferences are unfolding in patterns uncannily similar to the past two polarized elections, with a few notable, though not necessarily encouraging, differences.

In the 2000 and 2004 elections we had the red-blue pattern of county results that largely owed to divisions between rural and urban, married and unmarried or female heads-of-household, and a residual attributed to ideology (article here). In the tri-state primary votes these divisions still hold but have been overwhelmed by divisions of class, race and gender. The same old red-blue pattern holds fairly true. (In case you haven’t guessed, Clinton is the red candidate and Obama the blue—another reason we should expect Obama to decisively close out the nomination.)

This interpretation is confirmed by many analytics and comparative statistics. Clinton counties are strongly and positively correlated with Bush counties, while Obama counties are strongly correlated with Gore and Kerry counties. A simple regression equation matching county characteristics against vote outcomes across all 259 counties shows almost all of the voting choice can be explained by white and black households, married households, female heads-of households, median age, median family income and population density. The most significant variables were black households, median family income and female heads-of-household. These results are confirmed by exit polling showing that Clinton captured lower and middle class whites, women, and older married voters living in rural and outer suburban communities. Obama won among young voters, blacks, and single urbanites.

So, what does this mean in terms of the general election? One, though Democratic primaries only capture Democrats, it seems they vote in many ways similar to the entire electorate. Two, Democratic voters are largely divided by identity politics: race, gender, age and marital status. This makes for a hazardous coalition in national elections where the white, married, middle class is the dominant group. Three, Obama’s weakness was most glaring among working class married whites. Four, Clinton’s strategic error was to look past the primaries and gear up for the general election by moving towards the center on the issues. This left her vulnerable in primaries that are driven by more pure ideological partisanship. Both Clinton and Obama are perceived as far to the left of the general electorate, but without the Obama challenge Clinton could have portrayed herself more left-of-center against McCain’s right-of-center. Five, despite the rhetoric of post-partisanship and unity from both presumptive nominees, one wonders if we’re actually headed for the same polarized 50-50 election where electoral politics once again trumps more important issues of governance.