Monday, December 10, 2012

The Curious Case of those Infamous Robo-Calls

One of the things I learned from watching the old Perry Mason TV show is that to win a court case you often need what jurisprudence experts usually refer to as “evidence.”

And it’s this legal technicality which could prevent the free-trade-hating, Stephen Harper detesting, left-wing fringe club, otherwise known as the Council of Canadians, from winning its highly publicized “robo-call” court challenge which starts this week.

If you haven’t heard of this case, here’s the low down: The Council is backing the legal challenge of six citizens who allege “fraudulent robo-calls” discouraged non-Conservative voters in their ridings from casting ballots in the last federal election.

Accordingly, they want the courts to overturn the election of six Conservative MPs and order new by-elections.

(Originally there were seven challengers but one complainant dropped out after it was discovered she actually didn’t live in the riding she claimed to live in. Way to go, Council of Canadians crack research team! Although in fairness to the Council perhaps this woman was misled by a robo-call.)

Anyway, to win its challenge the Council will need to prove in court that robo-calls misled citizens, who were all set to vote Liberal or NDP or Green or Other, into spending Election Day sitting on the couch eating potato chips.

That’s a pretty high threshold of proof. How can you prove that people changed their voting patterns because of a robot phone call that took place more than a year ago?

Answer: It isn’t easy.

Mind you, some people have tried using fancy, shmancy statistical models to gauge the impact of robo-calls.

For instance, back in March, Professor Anke Kessler of Simon Fraser University published a statistical analysis which seemed to show that robo-calls suppressed non-Conservative supporters in the last election.

More specifically, Kessler did a poll-by-poll comparison between the 2008 election and the 2011 election, and found "a statistically significant effect of the alleged demobilization efforts" in the 27 ridings where robo-call complaints emerged compared to all other ridings.

Mind you, Kessler also cautioned that her "analysis and the corresponding results are not suited to bring the outcome in a particular riding into question" and that her “findings in no way can 'prove' whether misconduct or an illegal act has occurred."

Naturally, the media took these caveats into consideration when reporting on this story.

Ha! Just kidding!

What actually happened, of course, was the media reported this study with sensationalistic headlines like: “Study supports vote suppression allegations” and “ELECTION FRAUD! Robo-Calling May Have Significantly Impacted Voting, Says SFU Study” and “Did Robo-calls affect the last election? Apparently so, says new study.”

Me, I registered my skepticism about this study on Twitter, calling it “Statistical nonsense”. I would have offered a deeper analysis, but the word “gobbledygook” took up too many characters.

And for this I was immediately assailed by academics and economists who accused me of being “anti-science.”

Me, anti-science!!?

Full disclosure: It's true, at one time science and I didn’t really get along all that well (I barely passed grade 11 physics) But these days, now that I am no longer required to actually write exams, I am very much pro-science.

Besides, Kessler’s study wasn’t really science, but more like social science hokum dressed up as true science.

For one thing, her econometric regression analysis model didn’t and couldn’t factor in the myriad of different variables that can and do impact on voting behaviour. Or to put it another way, some things in politics just can’t be quantified or fed into a mathematical equation. For example, how do you compute the fact that former Liberal leader Michael Ignatieff had the charisma of a moldy kumquat?

To my mind, such a critical failing made Kessler’s study more useless than two tickets to the NHL Winter Classic.

Fortunately, a bright young man named Michael R. Smith, who witnessed my Twitter battle on this issue, agreed with me and what’s more, he helped prove my point in a brilliantly funny way.

Acting in a completely scientific manner, Smith managed to show how Kessler’s formula could also be used to provide statistically significant evidence to suggest  a secret, Conservative mind control device may also have demobilized the anti-Conservative vote.

Yes, Smith’s “study” was just for fun, but I am half-expecting the Council of Canadians to use it as the basis for a new court challenge.

Or maybe it won’t, since the Council is not using Kessler’s model in its court challenge.

But it is using for evidence something we all know is absolutely 100 percent infallible: polls.

In fact, last spring, the Council commissioned Ekos Research to do a survey on the impact of robo-calls in the last election.

(Ekos, by the way, is the same polling firm that confidently predicted that the last election would result in an NDP-Liberal Coalition government replacing the Conservatives.)

At any rate, the resulting Ekos survey showed Liberal, NDP and Green party supporters in the ridings involved in the legal challenge were much more likely to report receiving a misleading telephone call in the final days of the election than Conservative supporters in the same ridings.

In other words, if the survey is to be believed fraudulent robo-calls were “widespread” and targeted in the last election.

Sounds damaging for the Conservatives and good for the Council’s case, right?

Well, before passing judgment on this poll, a few key points must be considered.

First, to conduct its poll Ekos used (ironically) robo-calls. Yup that’s right. Rather than talking to a real live person, respondents to the Ekos poll answered questions posed by a soulless machine (And no I don’t mean Stephen Harper!).

Such a survey might go a little something like this:

Ring, Ring, Ring

Human: Hello

Robot Pollster: Hello sir and or madam. We would like to ask you a simple survey question.

Human: Well, I don’t live actually here, I’m from Alabama; I’m just visiting my brother-in-law so …..

Robot Pollster: If you received a fraudulent robo-call in the final days of the last election press “one” on your phone; if you didn’t receive a fraudulent robo-call press “two”; if you are not sure if you received a robo-call press “three”, if you’re sitting on the couch eating potato chips press “four”.

Human: What’s a robo-call?

Robot Pollster: If you need me to repeat your choices, please press “seven”.

Human: OK … Oops I hit one by mistake.

Robot Pollster: Thank you for completing our survey.

Human: But I didn’t….


The polling industry is divided as to whether or not this sort of polling method can achieve reliable results. Darrell Bricker of Ipsos Reid, for instance, has slammed this sort of polling as “tremendously biased in terms of sample coverage” he also noted that  in the last federal election such polls were “MASSIVELY (capitalized emphasis his)off on the final vote.”

Second and more importantly, the Ekos poll assumes people can accurately remember details of a phone call that took place several months in the past.

That’s an awfully huge assumption since humans, generally speaking, have extremely unreliable memories.

I can’t even remember why I’m writing this!

And it’s not just me.

Just consider a recent story that ran in Post Media newspapers concerning donations to the Conservative Party.

According to the story, reporters had uncovered the names of several people who said they did not donate to the Conservative Party even though party financial filings indicated they did.

In fact, eleven people contacted by Post Media claimed they definitely, positively, absolutely did not make any donations to the Conservatives and they wanted to know how their names ended up on the list.

As one of them put it, “I have nothing to do with the Conservatives. I want to find out who the guy was doing the fundraising because I have a few words to say to him.”

Seems like another sure-fire scandal, not to mention another possible Council of Canadians’ lawsuit.

Yet, a few days after this story appeared, we learned it was all just an innocent misunderstanding -- it turns out the people cited simply forgot they had made donations to the Conservatives.


So my point is, if people can forget about writing checks (that were for hundreds of dollars) isn’t it reasonable to assume they can also forget the details of a phone call that might have taken place nearly a year ago?

Plus all the publicity about robo-calls has probably tainted any survey results. People might genuinely misremember getting a robo-call just because they kept hearing about them on the news.

(For a more technical shredding of the Ekos poll, check this out and this and this.)

Certainly the Ekos poll would have been much more persuasive had it been conducted a few days after the election, instead of several months later.

I realize, of course, this analysis will disappoint those people who desperately want to believe the only reason the Conservatives prevailed in the last election was because they cheated.

But like it or not, there is no real scientific evidence to support the thesis that misleading robo-calls swayed any election results.

And if a judge is going to take the drastic step of overturning an election result and in effect disenfranchise thousands of Canadian voters, he or she will need some solid evidence of wrong-doing.

In my view the only true, scientific way to settle this would be to re-run the entire federal election under the exact same conditions as in 2011, except the robo-call variable would be eliminated.[1]

And since that’s unlikely to happen, it’s hard for me to see how the Council of Canadians will ultimately win its robo-call challenge.

But what do I know? Maybe legal rules have changed since Perry Mason went off the air.

Now if you will excuse me, a robo-call recently instructed me to sit on the couch and eat potato chips.

[1] This would make a great plot for the next Matrix movie!


WE Speak said...

Not to mention that the EKOS poll was conducted after several months of intensive media on this issue as well as the NDP and Council of Canadians running a marketing campaign soliciting people to file complaints. They then used the numbers of those solicited complaints as one of their justifications for filing the court case, regardless of the fact that none of them have been investigated as yet and that many of them were just as likely to trace back to Liberal and NDP campaigns as Conservatives. They asked for people to complain about any sort of problem call, not just the Election Day, Election Canada phony calls. Anyone who has ever worked on an election knows what kind of flood of complaints that was going to generate. In fact, they're solicitation of complaints did lead to a finding of a campaign breaking the law. The Liberal campaign of Frank Valeriote in Guelph was found to have twice breached the Telecommunications act and was fined $10,000.00 for doing so.

Anonymous said...

Hope you're right. It makes perfect sense to me. Polls are not facts, therefore not evidence.

Anonymous said...

I've just sent $45 to that left-wing fringe group known as the CoC. Apparently they are standing up to a pattern of democratic abuse that the writer appears not to have noticed.