County-level results from Texas suggest that Republican cross-over voting may have well given Hillary Clinton a victory in the Texas primary, thus enabling a narrative that has allowed to her to remain in the race. This has been reported in the WSJ, but my number crunching indicates the so-called Limbaugh effect may be significantly larger than is being estimated. First, from Washington Wire:
Conservative talk show host Rush Limbaugh led a campaign to have his Republican followers in Texas cross party lines and vote for Sen. Hillary Clinton in the state’s open primary last Tuesday. Why? Because Limbaugh thinks Republicans can defeat Clinton in a general election. Plus, watching Clinton and Sen. Barack Obama bloody each other in a nomination fight is pure sport for Limbaugh conservatives.
According to exit polls, Clinton won a notably higher number of Republican voters than she has in past open primary contests. Of the 9% of voters who identified themselves as Republicans in the Democratic Primary, Obama still edged Clinton 53%-46%. However, that margin is significantly slimmer than earlier contests. In Wisconsin’s open primary, for instance, Republicans broke 72%-28% for Obama. Similarly, in Virginia’s open primary, Obama was favored 72%-23%.
Clinton unquestionably secured a Texas victory, but some locals are convinced it was a false win bolstered by dirty politics. Laura Jean Kreissl, an accounting professor at West Texas A&M University, served as an election official in Canyon, Texas on Tuesday. She contacted the Wall Street Journal to report the hijinks she observed at the four precincts that voted at her polling location.
Of the 181 voters she personally dealt with, 70 offered that they were “Rush Limbaugh voters” who were there to cast ballots for Clinton. “I’m here to vote for Hillary Clinton, I want to see the Democratic Party implode,” one voter told Kreissl, she recounted in an interview. “I was just stunned,” she said. “As an election official we can’t say anything. We just jot them down and let them vote.”
So let’s turn to the Texas Secretary of State web site for a look at the county by county results to see if there are any interesting patterns. If Republican voters went en masse for Hillary, we should see a strong Clinton performance wherever John McCain, or any other Republican, got relatively few votes compared to what would be expected via the number of registered voters.
For example, Borden county registered zero votes for Republicans. Odd, huh? Oh well, only 432 registered voters there, so I guess it’s possible. Those who did vote came out for Clinton 93 to 32; John Edwards even got 9 votes. And then there’s Cottle County with 1230 registered voters. No Republican/McCain votes there. Clinton won, 276-113. Still just coincidence? Duval County has 9331 registered voters, and almost 4000 came out for Clinton; McCain, 0, and Obama, about 1000. 4:1 Clinton. So maybe this is just a rural, small county thing, right? Webb County has 100, 606 registered voters: Clinton 32,383, Obama 8615, McCain, 832. Democrats are either going to win Texas in November or something fishy is going on.
Total, there were 21 counties that registered zero votes for McCain. In those counties, Clinton won 20 times, with a vote total of 24, 982 compared to Obama’s 9056. Again, these are all small counties, so let’s look at a broader set of data. There were 46 counties in which McCain got less that 2% of registered voters: combined these counties went for Clinton 201,289 to 76,133. That’s a 125,126 vote difference. Clinton won the state by just over 101,000 votes.
All this is “opportunistic” reasoning, so I took a more analytical approach. Using statistical modeling, I posed that the ratio of Clinton to Obama votes was a function of the percent of McCain votes, a lo and behold, there was a moderately strong negative correlation which was statistically significant–in counties where McCain pulled in very low numbers relative to the number of registered voters, Clinton outpaced Obama by 2:1, 3:1, and 4:1 margins. Was it a county size effect? I threw that into the model, and got the same results. Maybe it’s simply that these are heavily Democratic counties that just happened to be for Clinton… so I controlled for these number of Democrats by using the number of John Edwards votes, and the result was the same. Thus, even after controlling for county size and Democratic strength in the county, Clinton did well when McCain didn’t.
I am sure there are other possible explanations, but the data seems to strongly suggest that Limbaugh’s campaign was effective and that a huge number of Republicans crossed over for Clinton, possibly saving her campaign. If Wyoming gives Clinton a win today, I’ll be looking for the same pattern in their data.