A Better Way to Report Election Results: Part 2 - Single-Winner Contests and Election Integrity
By Neal McBurnett and Celeste Landry
In our previous article, we highlighted the misleading nature of reporting multi-winner city council contests, and advocated for a shift to presenting a candidate’s support as a percentage of ballots rather than a percentage of votes. Continuing this discussion, we delve deeper into how our current reporting approach can hide serious ballot design flaws, which can be responsible for changing the outcomes of elections! These examples underscore the crucial need for consistent and transparent reporting of election results.
The LWVCO Voting Methods Position Statement notes the need to conduct post-election analysis to evaluate voters’ usage of the voting method. This analysis can also reveal ballot design problems. Timely access to data on all the ballots, including undervotes and overvotes, is important for putting election results in context while public interest remains high.
Mistaken Undervotes
A striking example of the impact of ballot design was observed in the 2018 US Senate election in Florida. Broward County relegated the race to the bottom of a column filled with instructions. This design evidently caused many voters to overlook the contest entirely, since the undervote rate was 4.5% county-wide, compared to 0.7% in the rest of Florida, and was nearly 12% for one ballot style in which the contest was most easy to miss. Notably, the county is disproportionately Democratic, and by some analyses, the effect was large enough to flip the race statewide, helping Scott to win instead of Nelson! The effect was unintended, but if better reporting of undervotes had been the norm in elections, the ballot designers might have been more aware of and more attentive to the implications of their designs.
Undervotes due to Top-Two Primary in California
The voting method itself can also lead to large undervote rates. California uses a Top-Two Primary, and in 2018, the top two vote-getters in the blanket (all-candidate) primary for US Senate were both Democrats, and only they advanced to the general election. In some conservative counties, over 20% of the voters declined to vote in the general election contest.
Mistaken Overvotes
The problem of overvotes, where voters select too many candidates, invalidating the entire contest on their ballot, is likewise concerning. For example, in certain contests in recent years, ballot design choices and voting system deficiencies led to tens of thousands of lost votes in New York and Florida, and over 200,000 in California, with rates of up to 14.8% of ballots in Lake County for US Senate in the 2016 primary.
Exhausted and Ignored Votes affect “Majority” in Instant-Runoff Voting
In our January Daily Camera guest opinion, we showed 3 different ways to measure majority in the recent Boulder IRV mayoral contest. We list them here and add a fourth way to measure majority that was newsworthy in the Alaska 2022 special congressional IRV election.
- Majority of active votes in the final round - This is the smallest type of majority, and is the standard definition used by IRV advocates. It ignores the people who didn’t support one of the final two candidates, whose ballots were “exhausted”. Critics call it a “False Majority” when the next type of majority in this list is not also achieved.
- Majority of valid votes cast in the contest, based on the first round of counting - This is what most people think of as a “majority of votes.”
- Majority support of all the people who cast ballots in the election - This includes the voters who left the targeted contest blank or who cast an invalid vote in a contest. (No candidate in the Boulder mayoral contest achieved majority support from all the ballots.)
- Majority in all head-to-head contests between candidates in pairwise contests - This is called the Condorcet winner. In the Alaska special election, Mary Peltola was declared the winner, but Nick Begich was the majority (Condorcet) winner as we discussed in our article, “The ‘Correct’ Winner is Squeezed Out in the Alaska Special Election”, since when you look at all the rankings on all the ballots, he was preferred by a majority when compared to each other candidate, including Peltola, the declared winner.
The LWVBC Voting Methods Team has looked at the ballot-level data for the Instant-Runoff Voting (IRV) results of the August 2022 Alaska special congressional contest, and the Boulder 2023 mayoral contest. Many rankings of the voters were ignored, because only the top remaining choice on each ballot is considered. Examining the ignored rankings is what allows us to compare the IRV algorithm with other algorithms which some experts consider to be preferable. A small number of ballots in each contest were invalid, due to skipped rankings, duplicate rankings, etc. Both of the contests were close enough that a small swing of votes could have changed the outcome.
Conclusion and Recommendations
These examples emphasize the importance of making all the ballot data available, and of reporting election results based on ballots, with complete data, including undervotes and overvotes, and rankings which may be ignored by a particular tabulation algorithm. These details help us understand the results, understand our voting method and algorithm choices, and understand voter preferences.
Finally, in terms of timing, we note that the official Boulder County 2023 Statement of Votes does include undervotes and overvotes, but doesn’t appear until after certification. Instead, the data should be presented starting on election night, when public interest is high, and to make it useful for determining auditing and recount needs.
If elections officials don't report all the ballot data, we can’t fully understand the voters’ experience. If we don’t pay attention to the right data, we won't even know when there's a problem to be solved. Understanding the number of ballots cast in each contest is essential for comprehending the electorate's opinions and driving critical aspects of election integrity like good ballot design and voter education. This practice also aids voters and candidates in better understanding election results. Moreover, this approach aligns with the ballot-centered practice of our post-election risk-limiting audits, ensuring a more robust and transparent electoral process. So in 2023, LWVBC updated its Voting Methods position to include:
For all election contests, except for uncontested races, election officials should report the number of undervotes, overvotes, and invalid votes.
The data on how many people undervote, especially in the down-ballot contests, may shock or delight political analysts, and might lead to voter education efforts to decrease the roll-off (aka ballot fatigue) rate, or alternatively may spur efforts to move some contests to a different time in order to shorten ballots.
In our Part 1 article we also recommended that all election night reporting should include the number of ballots for each contest, and use these counts in determining the reported percentages. Implementing our recommendations will provide a more accurate and insightful representation of voter intent and help identify any underlying issues with ballot design or voter engagement.