On election night, the Santa Cruz County Supervisors race in the 5th District was a virtual tie. Bruce McPherson led by 26 votes over Eric Hammer – with 6,500 absentee votes to count. Sentinel reporter Jason Hoppin ran some numbers and made a prediction that McPherson would win by 231 votes. Turned out he won by 245.
Jason was asked to share some of his homework on how he become the Nate Silver of Santa Cruz County’s biggest political race. Here’s what he said:
Not sure if you’d categorize this one as dumb or smart, but the result worked out so we’ll call it smart. Here’s how we “called” (OK, we said the apparent eventual winner had “reasons for optimism”) an election with 26 votes separating two candidates and 6,500 absentees still to tally. Before you scoff that it’s too simple, we were off by a grand total of 14 votes.
It was actually pretty basic, with an assist from an elections office that has come to expect detailed requests from us. When the dust settled on election night, we had results by polling place, broken out into several categories, including the number of in-person ballots and results and the number of vote-by-mail ballots and results.
Deciding that the absentees to be counted would break much like those absentees already cast (instead of in-person voters, which one campaign somewhat wishfully expected), we created a spreadsheet. For each polling place we analyzed the split of already-counted absentees. Two days after the election, the county clerk was able to supply us with the raw numbers of ballots from each precinct (in many cases, I had to combine precincts to get the total number for each polling place, since several precincts often make up one polling place).
From there, we ran the numbers. Applying ratios from already-counted absentee ballots to pending absentee ballots, we came up with expected results for each polling place. Then we added those together. We also reduced both totals by 8 percent — the difference between all counted ballots within this supervisorial district and the total number of votes in this particular race (in other words, the dropoff rate).
That’s it. We predicted that when the counting was over, Candidate A would have a lead of 231 votes over Candidate B. It ended up being 245. I’m sure part of this is luck, but we had a pretty big sample size on which to run our projections. In the end it worked out. -Jason