Day three of the hearing into alleged voter irregularities in the Saint John Harbour riding during last fall’s provincial election focused on what could be done with votes.
Liberal candidate Gerry Lowe won the riding by just 10 votes, which was upheld after a recount. Quickly afterwards, PC candidate Barry Ogden applied to have the result thrown out due to alleged voter irregularities, pointing to 78 ballots where information was not filled out correctly, voters were not struck off, potentially voted in the wrong riding, or even voted twice.
In other similar cases, something called the “magic number test” has been applied, where the result is thrown out if the number of rejected votes is greater than the margin of victory.
READ MORE: Hearing begins into voter irregularities in Saint John Harbour
Political scientist Peter Loewen believes he has found a better way. He feels statistical models can be applied to the data in order to determine the probability that the result would be overturned because of rejected ballots.
The first method, called the “naive method,” involves calculating the possible distribution of votes from rejected ballots between candidates, then determining how many of those vote distributions would change the result, which then gives a probability of the election being overturned.
For example, if there are 30 possible vote distributions and 10 of those distributions would lead to a different result, then there is a 66 per cent chance that the election would stand.
But Loewen has taken it a step further in what he calls the “informed method” which endeavors to fill out the result with more information, namely what voter preferences were in the riding.
The informed method first looks at the percentage likely to be cast for the two candidates, then determines the likelihood that a single ballot would have been cast for the winner, then looks at how many available votes there would be from the rejected ballots based on the two candidates share of the vote overall, and then used a binomial distribution calculation to determine the probability that the result would remain the same.
Get breaking National news
In this case, where the margin of victory was 10 votes, Loewen looked at the probabilities of a Lowe win with 10, 20, 30, 40, 50 and 60 rejected ballots. By his calculations Lowe would win 100 per cent of the time with 10 rejected ballots, and 92.5 per cent of the time with 60.
WATCH: Investigation into alleged Saint John Harbour voter irregularities drags on
But whether Loewen is qualified to even give that analysis was a matter of lengthy debate on Thursday. Lowe’s lawyer Tom O’Neil looked to have Loewen qualified as an expert witness in political science, econometrics, statistics and elections.
Representing Barry Ogden, Matthew Letson spent a significant amount of time questioning Loewen about his credentials. He asked about Loewen’s education, pointing out that he does not have a degree in statistics, mathematics or any related field. Letson asked if the methods developed by Loewen had been peer-reviewed, or if they had been looked over by statistician, to which he said no.
Justice Hugh McLellan eventually granted Loewen expert qualification in political science with a specialization in elections, but not in the realm of statistics.
READ MORE: Hearing over voter irregularities in Saint John Harbour set to begin next week
While much of the debate surrounded Loewen’s ability to create these methods, they were done so in a answer to a line from a Supreme Court decision dating back to 2012.
In Opitz v. Wrzesnewskyj the majority decsion mentioned the “magic number test” as the best available method for solving these issues, but admits that it is not without its faults.
“The ‘magic number’ test is simple. However, it inherently favours the challenger. It assumes that all of the rejected votes were cast for the successful candidate. In reality this is highly improbable. However, no alternative test has been developed. No evidence has been presented in this case that to support any form of statistical test that would be reliable and that would not compromise the secrecy of the ballot,” reads the decision.
But in the very next paragraph the decision leaves the door open for a better test to come along.
“However, we do not rule out the possibility that another, more realistic method for assessing contested election applications might be adopted by a court in a future case,” it says.
Loewen wrote his report as an answer to that sentence and believes himself to have found a better test.
Now that the evidence is finished lawyers for each side will have about a month to prepare for arguments on July 30. As he has done many times throughout the nine months that this case has dragged through the courts, Justice McLellan took the opportunity to ask counsel to do what they can to simplify proceedings.
McLellan said he would like the court to have the cleanest record possible of just the relevant issues should the case go through appeal and asked lawyers to be especially thorough in order to limit the possibility of one arising.
Because this case deals with the Elections Act, McLellan’s judgement will be automatically stayed until all avenues of potential appeal are exhausted.
Comments