Another week, another dramatic Republican primary, this time a runoff. Senator Thad Cochran eked out a narrow win against Chris McDaniel by appealing to Democrats in general and African-American voters in particular. How effective was this strategy? What changed between the primary and the primary runoff?
First, mapping the changes. If you want to see maps of the results, they're here. Below, on the left, we have turnout increase from the primary to the runoff. In the middle, how Cochran's share of the vote changed from the primary to the runoff. And on the right, the results of the 2012 presidential election. These are all repeated in the second set of maps in cartogram form, where the size of each county is proportional to the total number of votes in the runoff election.
Click to enlarge.
Click to enlarge.
It's clear that McDaniel took partial advantage of the favorable landscape the state presented to him. Turnout went way up in the northern part of the state, and so did his vote share.
Sadly for McDaniel, however, Cochran's strategy paid off big time. A massive increase in voters in Hinds County—the dark purple part of the blob in the middle of the cartogram—along with a decent increase in Cochran's vote share provide an increase in margin of 5,300 votes in that one county alone. In other counties where Obama had more than 60 percent of the vote, turnout also soared. Only a handful of counties—all of them red—had reduced turnout. Runoffs are typically lower turnout than the initial go-around, but not this time.
Below, some calculations that show at least 10,000 Obama voters took part in this runoff primary election.
Doing the math
1. The simple way.
Assume no Democrats (Obama voters = Democrats for the sake of this post) voted in the runoff in counties where Obama got less than 30 percent (a terrible assumption!). These counties on average increased turnout by 13 percent. If we extrapolate to the entire state, that's about 41,500 additional Republicans. Turnout increased by about 56,000, so that leaves about 14,500 Democrats. This is our floor for the number of crossover Democrats.
2. More complicated.
Is the data consistent with Democrats voting? If they voted, we should see an increase in turnout as Obama's vote share increases, roughly. We can see that in the map, but can we confirm it with a graph? Yes, here, although it's not a very strong relationship. Meanwhile, the first primary did not show this relationship.
We can eliminate some of the noise on that graph by plotting instead the increase in turnout. This incorporates the known turnout from the initial primary. We can calculate the runoff turnout based on the initial primary turnout assuming a uniform increase in Republicans and a uniform percentage of Obama voters joining in. The percentage of Obama voters determines how steeply the data rises; the uniform increase in Republicans moves the curve straight up or down. Somewhere between 8 percent and 12 percent works best for the Republican increase. Here's something close to the best fit, although it's not an exact science here:
Here's two more graphs with
2 percent and
6 percent Obama voters. It's a judgement call which might be best; I would think somewhere between 4 percent and 6 percent, but closer to 4 percent. (In case you're worried this method of estimation is a little too wishy-washy, something similar
worked pretty well in VA-07.)
Math check: 10% increase in Republicans = 31,900; 4% of Obama voters = 22,100; sum = 54,000. Actual primary turnout increase: 56,000.
Ranges:
Low—12% Republican increase, 3% of Democrats—17,000 Democrats crossover
High—8% Republican increase, 5% of Democrats—28,000 Democrats crossover
3. Checking the complicated way.
If these numbers are approximately correct, we should be able to use them to predict the increase in Cochran's vote share based on the initial primary results and Obama's vote share in a county. This time, the blue dots show the calculated change in Cochran's support, using the turnout numbers from part (2) above. I assumed that all additional Republican voters voted for McDaniel, and all crossover Democrats voted for Cochran. These assumptions certainly aren't true.
The blue dots follow the data reasonably well. On the left side of the graph, a fair number of data points fall above the blue dots. This is telling us that, if the 10 percent and 4 percent numbers are correct, some proportion of the additional Republican voters actually voted for Cochran, not McDaniel, which makes sense.
What if?
What if Democrats hadn't voted? It's clear Thad Cochran would have lost. Using the most conservative estimate from above of 14,500 crossover Democrats, and assuming only 80 percent of them voted for Cochran, they still would have contributed 8,700 votes to Cochran's margin of victory.
Any way you look at it, it's clear that Thad Cochran owes his career to thousands of crossover Democrats. Only time will tell what difference that may make.