Who will decide this election?


I know exactly how to get you to stop reading right now: We’re going to do a little arithmetic!

It’s simple, so I hope you’ll bear with me. 

Each cycle, pollsters and pundits alike seem to derive great pleasure from anointing some segment of the electorate as the decider of the outcome. For months, perhaps years, we’ll be hearing that some group of voters decided this election.  

I’m here to suggest that such an approach is often meaningless and incoherent, masking several different questions.  

Let’s start with an imaginary electorate made up of 80 moose and 20 bears who must choose between two candidates: Bullwinkle Moose and Yogi Bear.

Begin where the case for a decisive segment is obvious. Say the moose go 75 percent for Bullwinkle, yielding 60 votes. Moose clearly decided the election. Even if 100 percent of the bears went for Yogi, it would not have altered the outcome. Only a different result among moose could. 

But let’s complicate the story a bit.  

Say Yogi Bear got 45 percent of the moose votes (36 votes) and 75 percent of the bears (15 votes). Who decided the election then? Though Yogi lost the moose vote in a landslide, the largest number of Yogi voters came from the moose (36 votes vs. 15 from bears), for a 51 to 49 Yogi victory. 

Pundits may opine that bears made the difference because Yogi did so well among them. Another could argue that the moose provided the margin of victory.  

But the important fact here is that if any two voters — two bears, two moose or one of each — flipped their vote from Yogi to Bullwinkle, the latter would be the winner instead of the former. So, neither group “decided” the election; the interaction of all the individuals in both groups did. 

No pundit is reading the exit polls that way or analyzing the data from that perspective.   

Another way to think about who decided the election is to assess who moved between elections. Say Yogi Bear only got 40 percent of the moose last cycle and lost. One might attribute his victory this time to his 5-point gain among moose and conclude they decided the election.   

Well, he wouldn’t have won without the movement among moose, but it’s still the case that any two voters switching would have cost him this race as well.   

Turnout complicates the picture further. 

Go back to our earlier partisan division when Yogi Bear won by garnering 45 percent of the moose and 75 percent of the bears. Let’s say in the next election, Yogi got a lower proportion of bears, say 70 percent, but rather than comprising 20 percent of the total vote, bears benefitted from a turnout surge. Instead of 20 bear voters, 40 turned out.   

With 120 voters now (80 moose and 40 bears), the minimum it takes to win is 61, and Yogi garners a more comfortable 64 votes. Changing the composition of the electorate can more than make up for declining levels of support. And here again, if any combination of any five voters from either segment flipped to Bullwinkle, he would be the victor.

Even assuming exit polls are largely accurate — a big assumption — they have never been effective in pinpointing the composition of the electorate. Yet, if one is trying to determine which group “decided” the election, it is vital though rarely considered information. 

Obviously, these are highly contrived, just-so examples, but I hope their point is evident.  

First, it’s often unclear what we mean by saying some group “decided” the election. Are we asking which group provided the most total votes or the largest percentage of votes? Those can be quite different.   

Are we asking whether the votes of one group on their own made all the difference? Or are we asking whose votes and whose turnout changed over time?  

Often, the fundamental question — which group decided the election — is unanswerable, at least with the tools we have. 

Post a Comment

Previous Post Next Post