Maine’s approval of a statewide ranked-choice voting system being used for the first time in four Tuesday primaries required big changes in how the Bangor Daily News reports on Election Day.
It could be the only ranked-choice election in Maine if Question 1, a people’s veto effort to nix a law passed last year to delay the method, fails. Whether you want it to survive or not, it is probably going to lead to delays in finding out who wins the Republican and Democratic nominations for governor.
We wanted to explain why that is and how the BDN is expanding its Election Day efforts — which will include exit polling for the first time. We also have a new national partner for town-by-town results presentation. It’s all in an attempt to compensate for this uncertainty.
We will know first-round choices in certain races on election night, but the state needs time to retabulate ranked-choice races. In the past, you have generally known winners in high-profile races on election night because of unofficial results collected from Maine clerks by the Bangor Daily News or the Associated Press.
The reason why is rooted in data science, but the idea is simple. Historic data on how cities and towns have voted can be compared in real time to tallies coming in on election night. Once we are confident that margins won’t change dramatically, we can project a winner and tell you.
That’s fine in races decided with a plurality of votes, but not as valuable under ranked-choice voting, in which a winner is only declared if a majority picks a candidate as their first choice. If not, the candidate with the lowest number of first-place votes is eliminated and second-place votes for that candidate are reallocated. The process is repeated until someone wins a majority.
With seven Democrats and four Republicans running for governor, a first-round majority is unlikely. Clerks will report first-round votes to us on election night as they have in the past, but further rounds of ranked-choice voting will be run centrally by the secretary of state’s office.
The secretary of state has said that this will happen during the week after the election. State-hired private couriers must collect ballot information and bring it to Augusta for ranked-choice tallies. That’s a long period of uncertainty about crucial elections.
Exit polling will help us access those later-round choices, but it’s an experiment and we’re being cautious. An analysis of 2010 gubernatorial primary results found with good certainty that a small, random group of Maine cities and towns served as a good predictor of the statewide returns.
Using that principle, we have designed a basic exit polling operation only for the Republican and Democratic governor’s races. (We will not be polling the Democratic race for Maine’s 2nd Congressional District or a three-way legislative primary to be decided by ranked-choice voting.)
We’re sending people to a more-or-less random group of cities and towns above a certain size. They will try to hand primary ballots to a random distribution of voters — perhaps one of every five leaving the polls. Interested voters will fill out an anonymous primary ballot corresponding to their party that also asks for age and gender. We will collect apparent age and gender data for people who decline the poll.
This will be put in a database. Once actual first-round results come in, we will evaluate polling data to see if it is close to those totals. If it is, we will have more confidence in our later-round information. If not, we will have less confidence and be more cautious in reporting on the poll.
If this information is good, we will use real first-round results and the exit polling data for later rounds to simulate a ranked-choice voting race like we did in a re-run of the 2010 and 2014 governor’s races with reader-submitted data. But our model will also incorporate the kind of historical data that we typically use to predict races on election night.
This could be powerful and allow us to forecast the election and show you our math. It could also go poorly and if we think it is, we’ll pull the plug on it. We’ll be making judgment calls on that as the polls close and our final exit polling results start rolling in.
We have a new national partner for results collection. The BDN’s Election Day results page will be operating as readers are used to — from the gubernatorial and 2nd Congressional District primaries to legislative, district attorney, county and even local races.
But for the June primary, we’re partnering with Decision Desk HQ, a new company that delivers and interprets election results and also works with Buzzfeed and Vox. We’ll be putting Maine results into its system; it will be calling races for us if they don’t require ranked-choice finagling.
Maine is experimenting with ranked-choice voting and so are we. We’re telling you about it in the spirit of transparency and we’re excited to show you what we’ve come up with for Tuesday. We’ll be even more excited if it works the way we think it can.