海角大神

In a recent post, I explored whether it鈥檚 possible to quantify Bill鈥檚 value to Hillary Clinton鈥檚 campaign (see ). In other words: Do voters reward (and punish) Mrs. Clinton based on their opinions of her husband? If so, it could have major implications for the outcome of the 2016 presidential election.

Looking at survey data from the American National Elections Study, the answer would seem to be an emphatic 鈥測es.鈥� Even when we take into consideration various reasons why voters would hold a favorable or unfavorable opinion of Hillary Clinton (such as party identification, ideology, gender, etc.) we find that Bill has a discernible positive effect on Hillary鈥檚 approval rating.

Perhaps most notable is just how much Bill Clinton matters according to the results. In particular, the data suggest that when Bill鈥檚 approval rating increases by 1 unit, Hillary鈥檚 approval increases by just under 1/2 in the same direction.聽

For today鈥檚 post, we鈥檙e going to reexamine the same topic with additional data. Indeed, the prior post left some important questions unanswered.

First, the聽original analysis聽examined opinions of Bill and Hillary Clinton in 2000. It鈥檚 certainly possible that the so-called 鈥淏ill effect鈥� is unique to this particular time period (perhaps because Bill Clinton was president at the time and the economy was performing quite well). Simply put, we want to know if the finding is generalizable to more recent conditions.

Fortunately, the ANES asked the same questions in 2008 (when Hillary Clinton ran for the Democratic nomination) as it did in 2000 (when she was first lady). So, we can easily run the same analysis as in the prior post and see if the 鈥淏ill effect鈥� holds. You can see the results in the chart above.

I鈥檓 skipping the statistical details, but the regression model confirms the existence of the 鈥淏ill effect鈥� in 2008. As with the prior results (see ), it would seem that that there is a very strong positive relationship between opinions of both Clintons. It鈥檚 also the case that this effect is statistically significant.聽

But what about the act of voting?聽 Isn鈥檛 this the most important outcome?聽 So far, we鈥檝e only looked at opinions of both Clintons.

Fortunately, the ANES asked respondents about their vote choice in the 2008 presidential primary. So, we can once again run the exact same analysis expect that, this time, we鈥檙e modeling whether someone voted for Hillary Clinton. (Click on the right arrow on the chart above to see the second chart.)

According to the results, the 鈥淏ill effect鈥� exists with respect to voting behavior as well. In fact, the results indicate that Bill Clinton鈥檚 effect is larger in magnitude than any other predictor. Keep in mind that this is specific to the 2008 presidential primary, where we would expect the effect of economic conditions and party identification to be lower.

According to the results, a 25-point increase in Bill鈥檚 favorability聽rating (out of 100) increases a respondent鈥檚 probability of voting for Hillary Clinton by about 10 percent. Substantively, therefore, the effect is meaningful (as it was earlier).

Lastly, perhaps there are differences in who is susceptible to the 鈥淏ill effect.鈥� For example, it may be interesting to know whether Democrats, Republicans, or Independents are most likely to be persuaded to vote for (or against) Hillary Clinton based on their opinions of Bill Clinton.

I created the third chart above by simply interacting Bill Clinton鈥檚 favorability with a respondent鈥檚 party identification and re-running the analysis.聽 Higher values in the figure (in red) indicate a larger 鈥淏ill effect鈥� for respondents in the respective group. We see two interesting patterns.聽

First and foremost, Bill Clinton鈥檚 effect on a respondent鈥檚 probability of voting for Hillary Clinton is strongest for Independents. Roughly speaking, there appears to be a curvilinear effect. 聽For 鈥渨eak鈥� and 鈥渟trong鈥� partisans on both sides of the aisle, the effect is smaller in a relative sense.聽 聽聽

Second, Bill Clinton鈥檚 effect on Hillary鈥檚 vote probability is larger for Republicans than it is for Democrats. In fact, for 鈥渟trong Democrats,鈥� Bill Clinton has 鈥渘o effect鈥� at the 0.05 level (as revealed by the fact that the confidence interval just crosses the zero line).聽

Based on the results, we might conclude that Hillary Clinton鈥檚 campaign should get Bill in front of independents and Republicans rather than steadfast Democrats. Indeed, Bill could be a potent weapon in 2016鈥� if used properly.