Calling his effort "Operation Chaos," Rush Limbaugh has been urging Republicans to cross over and vote for Hillary. There are competing claims about how successful his effort has been (here and here), so I decided to exploit variation in the Indiana and NC primary rules to see how much influence Limbaugh had on yesterday's results.

Here's the key idea--Indiana has an open primary but NC does not permit Republicans to vote in the Democrat primary (unaffiliated voters can). Moreover, NC had a contested primary for the GOP nomination for governor that would serve to keep NC Republicans in their own election.

So I estimated a regression model for the percent of the vote received by Hillary in NC and IN counties. RHS variables include the black percent of the population, the percent of the population between ages 16 & 24, the percent of the population over 65, the percent of the population that is male, and per capita income. The model also includes a dummy variable taking a value of 1 for IN counties--this variable should pick up any support for Hillary that is not explained by the other variables thereby making it a crude measure of the Rush effect.

So what do the results find? The Indiana dummy has a coefficient of 0.53 meaning that on average Hillary got a about one-half percentage point larger share in Indiana than would be explained by the control variables. The point estimate is not statistically significant (t = 0.43). The regressors perform as one would expect, except the percent male has no effect (either in magnitude or significance).

My student worker Katie compiled data for me and is compiling more as I type. Look for updates later.

BTW, Limbaugh has just come on. He is claiming credit for tilting IN to Hillary and playing audio to that effect from John Kerry. My results suggest otherwise.