By day, Philippe J. Fournier teaches physics and astronomy at Cégep de Saint-Laurent in Montreal. His evenings and weekends, nevertheless, are dedicated to a really totally different type of experience—election modelling.
Fournier’s status is constructed on his creation of 338Canada, a platform that has change into a go-to useful resource for political junkies. By aggregating polling knowledge, demographic developments, and electoral historical past, Fournier creates subtle projections that routinely outperform conventional punditry. Since launching in 2016, his mannequin has known as outcomes accurately in over 1,600 electoral districts, reaching an astonishing 90 % accuracy charge.
Whereas there are different outstanding polling analysts, akin to Darrell Bricker, what units Fournier aside isn’t simply his monitor report however how he blends statistical precision with insights into dynamics akin to voter psychology, capturing nuances uncooked polling usually misses. This has made him a fixture in Canadian media. It’s additionally what led us to ask him to change into a contributing author.
We spoke just lately over Zoom. In our dialog—which has been edited right here for size and readability—we mentioned his course of, his dedication to accuracy, the upcoming federal election, and the worth of making use of scientific rigour to the unpredictability of politics.
I wish to begin by speaking a couple of correction you had us make if you first joined us. We launched you as a “pollster,” and also you swiftly emailed saying, no, “I’m a polls analyst.” For the report, are you able to clarify the distinction?
I don’t conduct polling. That is completed by skilled companies which have an infrastructure that I don’t have and don’t want to have. As an alternative, I take their knowledge and attempt to make one thing out of it—calculate seat projections and odds of profitable. I additionally charge the companies. That’s the opposite factor. I can’t take part within the polling since I charge the polling. It’s like if Customary and Poor opens a financial institution. It doesn’t work. So the analogy I like to make use of is—they develop the wheat and I attempt to make bread.
How do you belief the standard of that wheat? What’s your standards? I think about not all polling companies are created equal.
The easy reply: I examine the monitor report. If a brand new polling agency pops up, I’m going to incorporate their polling however give it very low weight till they present that they’ll get near election outcomes, both a by-election or a normal election. A agency with a fantastic monitor report can have extra weight in my calculation—and in addition extra leeway in the event that they pull out numbers which can be a bit unusual. If Leger immediately exhibits the federal Liberals tied with the Conservatives, I might in all probability ask some questions, however I’ll go—Yeah, it’s Leger, proper? It’s not just a few obscure pollster.
What does it take to change into a “Leger poll”? What does it take to change into a gold commonplace within the polling trade?
Consistency. It’s important to present that your pattern might be consultant in lots of locations—provincially, municipally, and federally. It’s important to present that if you get it proper, it’s not a fluke. Within the case of Leger, in case you take a look at their monitor report previously decade, it’s virtually straight “A”s all through. Within the large provinces, within the smaller provinces, on the federal stage, they’re by no means too far off. That is what it takes to be an A or A-plus pollster in my score. It’s important to present that you could constantly name elections inside the margin of error.
How do you go about your day as an analyst? Do you examine polls very first thing?
Day by day, I am going by the polls from virtually each province. I’ve some assist with this, as a result of there’s numerous knowledge on the market. I cowl 9 provinces. I don’t cowl Prince Edward Island. It’s not that it’s not attention-grabbing, however there are only a few polls there. Additionally, it’s a small place. I additionally don’t cowl the territories—once more, only a few polls. However I examine the numbers from the federal stage and, for the larger cities, the municipal stage. I crunch these numbers and, federally, I replace each Sunday. It’s good to have common updates, as a result of it drives site visitors, and polls are not often launched on Sunday. Throughout a marketing campaign—particularly a federal marketing campaign, just like the one we are going to in all probability have within the spring—I’m going to replace each evening, as a result of there can be polls on a regular basis. If you wish to have image of what’s happening, particularly if there are vital actions within the polls, it’s important to put within the work to replace each day.
Wait, you mentioned spring. You assume an election will occur within the spring?
I don’t assume the opposition can have a chance to make the federal government fall sooner. I don’t see how the Liberals can go one other finances. We’ll see. Some politicians are Teflon. However I feel the probably situation can be late spring, possibly round Could.
Why do you even do that? You’re an astrophysicist! How did you get into this recreation?
I did a grasp’s degree in astrophysics at Laval College round 2004 and obtained myself a part-time job instructing physics at Cégep de Saint-Laurent. I used to be twenty-four, and I made a decision that I ought to get pedagogical expertise to final on this job. So I signed up for the training certificates at Concordia and took a political science class. I assumed it was very attention-grabbing. Utterly totally different from physics and astrophysics. Completely different setting as effectively. Normally, I sat down with my calculator, my pens and pencil case, able to take notes. However right here it was simply individuals speaking. I got here to know that I actually, actually love numbers in politics—greater than I like politics itself. I completed my training degree and went again to instructing. Then, in the course of the 2016 US presidential election, I adopted the marketing campaign on a day-to-day foundation. And I observed an enormous discrepancy between what the speaking heads and the columnists have been saying and what the numbers have been saying. The numbers have been suggesting a detailed race, however many individuals simply couldn’t consider a so-called simpleton like Trump may win. That’s after I realized that the media, together with Canadian media, wanted extra scientists to speak about politics.
Because the time you began doing this, has polling modified, or has the way in which of accumulating info modified? And have you ever wanted to adapt to these modifications?
The variation is extra on the facet of pollsters. Polling is an ever-changing trade. There was a golden age within the ’70s, ’80s, and ’90s, when all people had a landline and you can simply name them. Cellphones have now made it more durable to achieve individuals, particularly younger individuals. However pollsters must discover a method—in any other case they exit of enterprise, proper? Nonetheless, there’s no query there’s been a Darwinian winnowing. For those who take a look at the companies we had possibly fifteen years in the past, half don’t exist anymore. A lot of them have been actually good, however they might not adapt to the brand new actuality.
Polling now occurs on-line.
It’s increasingly more of an internet factor, sure. Polls are nonetheless completed over the cellphone, and the companies that conduct them are fairly good. There’s Janet Brown, in Alberta, who nonetheless does cellphone polling, and her numbers are at all times actually shut. Nanos Analysis additionally does cellphone polling on the federal stage. So it nonetheless exists. It’s important to have individuals in name centres sitting at their desks with their headphones and calling numbers.
I assume I ought to have began with this query, however—so simple as you can also make it, what’s polling?
Think about you have got a giant reservoir, and picture that reservoir is stuffed with colored balls—inexperienced, blue, yellow, pink. And let’s say, at random, you begin selecting out balls. One is pink, one is blue, then a blue one once more. Sooner or later, since you might be selecting at random, the pattern that you’re constructing will appear to be the whole lot of the reservoir. The percentages that you’ll maintain selecting, say, pink balls at random are nil. It doesn’t occur. That is the character of statistics. And so, in case you’re a pollster and also you’re making an attempt to take a consultant pattern of the inhabitants, then it’s important to make the pattern as random as attainable. If randomness is concerned, your pattern can be near the whole inside a margin of error that may be calculated.
What’s the trick for reaching that randomness?
You choose respondents in a method that ensures each voter inside the goal inhabitants has an equal likelihood of being included within the pattern. That is usually completed by strategies like random digit dialling for cellphone surveys or utilizing algorithms in on-line surveys which pick respondents from a big pool of potential voters. However true randomness is unimaginable to attain with perfection, so pollsters additionally apply statistical weighting to regulate for underrepresented teams, primarily based on demographic components like age, gender, and area. This helps make the pattern extra reflective of the broader inhabitants.
And with out randomness, findings change into skewed.
Sure. That’s why polling depends on an infrastructure that wants sustaining. Folks transfer, die, are born, change into eighteen. In case your pattern stops being consultant of the general inhabitants, your numbers fail. And in case your numbers fail throughout an election, that’s dangerous publicity, and immediately you get fewer contracts. Pollsters make cash by showcasing how effectively they do in an election and turning to, say, an airline firm and saying—Look, you’ll be able to belief our numbers. We known as it accurately.
Leger makes use of its success in political polling as a calling card for different industries?
Oh, 100%. They don’t do political polling for revenue. There’s little or no revenue in it, except you get to ballot for a political occasion, and even then, you’ll be able to’t rely solely on that. Political polling is like promoting. Leger invests in it to show their salt as market researchers. There was a brand new ballot from Leger just lately, and within the first pages, they put the results of the 2021 federal election subsequent to their last numbers and reminded readers that they have been the closest agency within the nation. And it’s true. They have been the closest.
Has your method of analyzing polls modified over time? Are you slightly bit bolder now that you just really feel snug studying the indicators?
I don’t assume my methodology has modified a lot. I make changes frequently. Each election, I attempt to study with each missed name or mistake. We had the BC election just lately, and my projections have been fairly good, contemplating there was a brand new electoral map, a brand new occasion, a brand new contender. However there have been some surprises. And I at all times ask myself: Was there a method within the numbers to make that decision precisely? If sure, then I attempt to right the course.
What makes the mannequin that you just’ve constructed so correct? Is there a secret sauce?
There’s no secret sauce. My methodology is on the web site, and all people can learn it. It’s fairly clear. You’ll by no means see me brag on-line about how effectively I did in an election, as a result of if the info is nice, the outcomes must be okay. I additionally at all times do not forget that lots of my college students, and now former college students, are watching, and if I begin to be the hot-take man who is aware of that one thing is totally true—that so and so will lose or their rival will win—I cannot be true to the method. I attempt to see polling by the eyes of a scientist. All the time be humble earlier than the info, and do not forget that we’re coping with human beings. As an astrophysicist, I’m used to coping with complicated stuff. However people, being irrational, might be much more sophisticated than electrons or galaxies. When you have got a ballot, at all times keep in mind: that is an approximate factor; this isn’t an ideal factor.
However what are you in search of within the numbers? What do it is advisable see to start out making projections or predictions?
My first thought is: The place does this knowledge come from? Then I search for comparability between the brand new knowledge and previous developments. Is there continuity within the numbers, or are new tendencies rising? Clearly, taking a look at horse-race numbers or authorities satisfaction is totally different than crunching knowledge on extra complicated points, akin to tariffs or NATO. Then if I see surprising developments, I ask myself whether or not that is merely noise or a part of a sustained shift? Good polling requires common polling. We received’t know whether or not opinions are shifting or bettering or eroding if we take a look at a snapshot. We’ve got to interpret each ballot inside context—and that is typically what’s lacking from mainstream media evaluation of political polls.
Do you are likely to get into fights about your polls with conservatives or liberals? Do you get pushback from political events?
Actually, I don’t give a shit what individuals assume. I’ve good connections with political events in Quebec on the federal stage—some in Ontario as effectively. We change info typically, and it’s very well mannered and respectful. But when some randos on-line are sad with my numbers? So be it. Normally, it’s as a result of their occasion is trailing and so they don’t prefer it.
The frustration, I think about, has to do with the sense that you’ve got sway in how individuals really feel about an election.
Look, I’m simply offering info. Do polls affect voters? That’s query, and the reply for me is: info influences voters, however lack of understanding additionally influences voters. Say I present that the NDP vote would possibly maintain the important thing to the Liberals beating the Conservatives—if sufficient individuals change. If it’s true, it’s info, and other people can resolve to do no matter they need. There are lots of NDP voters that can by no means vote for the Liberals, even when they know they {split} the vote. That’s not my downside. Individuals are free to do no matter they need.
What’s essentially the most helpful method for somebody to know what polls characterize?
In an election, events will do the whole lot they’ll to persuade you that they’ve the momentum. But when a ballot is completed proper, if it’s completed by professionals, it’s the one goal info you have got in regards to the state of a marketing campaign. Every part else is spin. And so after I hear that polls have affect, it drives me nuts. Pollsters are merely providing you with the rating of the sport. Might you think about going to a Habs–Bruins match, and it’s three–two for the Bruins after two durations, and Habs followers say, Shhh, don’t inform me the rating; you’ll affect the result!
What’s the story of Canadian politics proper now? Is it as apparent because it appears? Liberals sinking, Conservatives rising, NDP struggling?
It depends upon how a lot decision you need your image to have, however typically Occam’s razor enters the fray, and the only rationalization makes essentially the most sense. In Canada, in contrast to the US, we don’t have time period limits, however we now have pure time period limits. Normally, it’s a decade in energy, after which individuals get sick of you and vote you out. Proper now, the Liberals have been in energy virtually a decade—a decade that included crises we hadn’t seen earlier than. I imply, what was the worst factor Pierre Trudeau needed to cope with? The October Disaster? I don’t wish to diminish Quebec separatism, however you actually wish to evaluate that to battle within the Center East, battle in Ukraine, and a world pandemic? Positive, possibly Liberals made self-inflicted errors, however the remainder of it’s about being in energy a decade, a tough ten years. Generally it’s so simple as that.
What’s the model with a bit extra decision? What will we see once we zoom in?
That Conservatives are a lot, significantly better campaigners than Liberals—and have been for some time. That Trudeau has needed to cope with an rising variety of conservative premiers who disagree on many points, and it makes the whole lot more durable. However when Pierre Poilievre turns into prime minister, his base will love him, others is not going to as a lot, and after a number of years, the tables will flip. It’s a pure cycle in Canadian politics.
Carmine Starnino (@cstarnino) is editor-in-chief of Ontario Chronicle.
fbq(‘monitor’, ‘PageView’);
