The National Telegraph - Wyatt Claypool - April 12, 2026


Pollster ADMITS Mark Carney Liberals aren't doing that well!


Episode Stats


Length

19 minutes

Words per minute

176.80359

Word count

3,373

Sentence count

171

Harmful content

Toxicity

2

sentences flagged


Summary

Summaries generated with gmurro/bart-large-finetuned-filtered-spotify-podcast-summ .

In this episode, we talk about polling in Canada and why the Liberals are not leading in the polls by as much as some pollsters are claiming. We also talk about why the polls are biased in favor of the Liberals and why they should be leading by much more than they are.

Transcript

Transcript generated with Whisper (turbo).
Toxicity classifications generated with s-nlp/roberta_toxicity_classifier .
00:00:00.200 Ahoy everyone, Wyatt Claypool here, back again to talk about Canadian national polling.
00:00:06.080 And today on the show, we're not talking about any specific poll that has come out.
00:00:11.060 We're actually going to be doing a bit of a lesson on polling in Canada,
00:00:15.140 and talking about why the Liberals are not leading by nearly as much as many of the firms are claiming.
00:00:21.540 The Liberals are probably leading from anywhere from about 5 to 8 percent.
00:00:25.860 It's within striking distance if the Conservatives start to make up a little bit of ground, but it's not plus 15 like some of these pollsters like ECOS or Leger are saying.
00:00:37.520 They are not doing proper sampling.
00:00:39.940 And I have to hand it to David Coletto, who runs Abacus Data, because he actually came out and admitted that a lot of these pollsters are not doing proper sampling.
00:00:50.940 They are simply just sampling a thousand people and putting up the results, not actually disclosing that they didn't check if many of them were partisan liberals taking the polls and they were heavily undersampling conservatives from the last election.
00:01:05.240 What you should be doing in a poll is trying to find about the mix of people who voted for the different parties last election. So 6% of your people in your poll should have voted NDP last time, 6% for the bloc, 41% for the conservatives, and 44% for the liberals, plus whatever the Greens and the PPC got.
00:01:25.680 That should generally be the mix, plus some people who didn't vote last time and maybe don't usually vote or sometimes voted.
00:01:34.320 You understand what I'm saying here.
00:01:36.360 So what I want to take you guys through is what David Coletto actually put on the Abacus data substack,
00:01:43.040 showing just how bad it has gotten in terms of the liberal bias in the samples that the pollsters are using.
00:01:50.740 Abacus actually tries to correct for the problems.
00:01:54.180 Other pollsters are not, and David Coletto is currently calling this out.
00:01:58.840 So in just a second here, I want to take you guys through this chart of him showing the weighting problems in these polls
00:02:04.660 so we can get a better picture of where the Liberals and Conservatives actually are in the polls these days.
00:02:11.800 But before we get into this, I just want to remind you guys, if you like the show, make sure to leave a like on this video.
00:02:18.540 Subscribe if you are not yet a subscriber. Hit the notification bell.
00:02:21.920 leave a comment on what you think about all this and if you want to become a channel member and
00:02:27.080 help incentivize videos like this to be made that are not as flashy a little bit more technical
00:02:32.800 consider hitting the join button and becoming a monthly contributor so i can be less reliant on
00:02:38.320 the youtube algorithm and make videos like this where it's not kind of the polya versus carny
00:02:43.900 type narratives that do well on youtube it's talking about polling numbers and biases in
00:02:49.460 different polling firms. Anyways, so we're going to go through the chart that Apicus data put out.
00:02:54.800 I'm just going to kind of write down the numbers as we go. And so we're just going to go through.
00:03:00.720 So David Coletto in this chart ends up showing in red what the unweighted sample would say a party
00:03:07.680 was at. And then the green is the weighted sample. And then the yellow is what parties were actually
00:03:14.420 So, yeah, basically, David has proven that he actually tries to wait for partisanship, and it's made the Abacus data polls far more accurate than other firms over time.
00:03:23.920 But what I want to focus on is when you just do a general poll of people, because he released what his general poll numbers show when he doesn't add weightings and when he doesn't try extra hard to find extra partisan conservatives to take the poll, people who voted conservative last election to take his polls, what the numbers are like.
00:03:42.500 So we're going to go through both of the parties, and then we are going to show what the actual number is. So I'll write down the parties in their colors, and then I'll write down the actual on the right. Or I guess I can just stick to the party colors, but this will be kind of unweighted, and this will be the actual.
00:04:01.300 So in 2019, Abacus data in that election, and I believe these are based on the averages, in that election, they had the Conservative Party in their polls only getting 30% in that election.
00:04:18.780 Now, remember, obviously, the polls for the two parties were not as high in those elections because there was the NDP was doing far better and the bloc was doing far better.
00:04:28.320 So we weren't having those like plus 40 percent results like liberals and conservatives got last time.
00:04:33.440 But they found they had the conservatives getting 30 percent and they had the liberals getting 36 percent, which it should go without saying that would have been a big Justin Trudeau majority government if that was true.
00:04:45.920 Now, when they applied weightings, the conservatives were doing better. They were at like 33 percent and the conservatives ended up getting 34.3 in that election. So they did really well with their weighting at Abacus with trying to make sure that they have a better previous election partisan balance to their polls.
00:05:04.580 and the actual, or the conservatives here, grabbed the wrong color there. So they actually
00:05:12.300 ended up getting 34. I'll just round down here just to make it simple. And the liberals who were
00:05:19.960 predicted to get 36 based on the unweighted sample ended up getting 33. So we see here,
00:05:27.600 we have an undersampling of conservatives and we have an oversampling of liberals.
00:05:32.820 And remember, this is not something that David Coletto is like unaware of. He corrected for this. His actual numbers when he did the proper weightings were 33 and 34. So he was pretty much almost right on that election.
00:05:45.580 In 2021, he actually had the conservatives in that election, sorry, Ron Culler again, only winning 26% of the vote, based on who answered with no weightings.
00:05:59.200 The liberals had 33% of the vote.
00:06:03.620 Now, that was actually pretty accurate when it came to the liberals.
00:06:07.100 The liberals in that election ended up getting 32.6, so maybe let's just go, you know, I'll write that down because it's right smack in the middle between 32 and 33.
00:06:18.840 So he had 32.6 for them, but of course the conservatives in that election actually got 33%.
00:06:26.800 So they were under-polled by seven points when David Coletto didn't go the extra mile to make sure that he was getting enough partisan conservatives from the previous election answering the poll.
00:06:39.360 In 2025, the trends actually keep getting worse over time.
00:06:44.560 So we have the liberals in this election being polled at getting 45% in the unweighted sample, which is actually pretty close to what they got.
00:06:53.180 they ended up getting 43.7% of the vote in that election, this past one. But the conservatives
00:07:01.580 in the unweighted sample for that election, before Coleto started making the corrections
00:07:07.020 needed to make the poll more accurate, only had the conservatives getting 35% of the vote. And
00:07:12.320 even when he weighted it up, he only could get them to 39. But of course, the conservatives
00:07:16.260 of this last election got 41.2% of the vote. So this is a great example of why you cannot trust
00:07:29.400 certain pollsters, why Spark Insight sucks, why Eco sucks, why Leger usually cannot be trusted.
00:07:36.540 Some of them will have bad numbers from time to time, and it's because they're not correcting for
00:07:40.640 this, but they do eventually sort it out. I find Main Street Research, they'll put out some bad
00:07:45.820 polls, but then they will have corrected for the next time and their numbers generally improve
00:07:50.260 as they learn lessons. But so many of the pollsters are incentivized to not learn lessons.
00:07:56.100 Liaison is never going to learn a lesson because they want to show big numbers for the liberals.
00:08:00.700 They want to show plus 15. But when you don't correct for the fact that conservatives are
00:08:06.400 less likely to take a poll than a liberal, and if you don't try harder to try and find more
00:08:11.380 conservatives to take your poll, you're going to have worse numbers. And that's not saying you look
00:08:15.260 at a poll result and say, hmm, I think that's too low, so I'm going to just go and find people
00:08:19.520 who I know are going to stay conservative. The whole point is that if you're polling the next
00:08:23.780 election, you should try and make this the split in terms of outside of the margin of people that
00:08:30.580 you poll who didn't vote last election, of your partisan people who did vote, this should generally
00:08:36.880 be your baseline. You should try and get to this as closely as possible. When you ask people who
00:08:42.060 they voted for last time, you should want around 41% and around 43% of people say they voted
00:08:47.660 conservative and liberal answering. Because let's put it this way, to make a logical,
00:08:53.880 to put some logic into this. When you see these pollsters showing that the liberals are at 51%
00:09:01.020 nationally, would it change the poll a little bit? Would it change how you would react to the poll
00:09:06.620 if you found out that 48% of the people who took that poll voted Liberal last time.
00:09:12.320 So they are literally maybe over-polling the Liberals by maybe 5% of Canadians that they polled
00:09:21.020 were more Liberal than the last election was.
00:09:25.280 And so when they're doing 51%, well, they're maybe doing better in the polls these days,
00:09:32.280 but they're only at 51% because you mostly polled partisan Liberals.
00:09:35.860 And if the conservatives are only at, let's say, like, I don't know, like these days, they're saying they're at 35, but if you only polled 37% of people who said that they voted conservative last time, they're not doing that much worse than last time.
00:09:51.600 they're in fact only doing maybe one point worse because you polled 5% fewer conservatives than
00:09:59.400 actually voted that way in the previous election. This is the thing that drives me up the wall.
00:10:05.520 People will act completely credulous at any numbers that get reported by pollsters. They'll
00:10:12.020 just assume that, well, they must be correct because a polling firm said so. I know some
00:10:17.760 people now distrust all pollsters, and in part, you do it because they don't report stuff like
00:10:23.120 David Coletto is doing at Abacus Data. He is opening up the back rooms of his own polling
00:10:29.360 firm and showing you, if I didn't make the corrections I was doing, like other polling 0.97
00:10:33.580 firms don't do, I would be putting up these crap numbers showing the liberals winning by 0.95
00:10:37.720 10 points. If you don't make these corrections and make sure that you are polling people
00:10:43.420 based on how in proportion to how many what who voted what last time you will have crazy results
00:10:49.860 like plus 10 when in fact was only a when it was only like a plus 2.5 percent victory
00:10:57.920 you overestimate the liberal win by four times when you don't actually make these corrections
00:11:05.860 And so when we see ECOS saying, oh, plus 20 liberal, is it actually plus 20? Or did you do something that caused the lead to be exaggerated by four times as big as it currently is?
00:11:19.740 So in theory, this is why I conclude the conservatives are behind right now.
00:11:24.320 Mark Carney is in the lead.
00:11:25.900 He has a certain sort of, I guess, establishment advantage right now because he's currently in office.
00:11:31.500 People are giving him a chance right now.
00:11:33.440 Every once in a while, he gets into a fight with Donald Trump, which a lot of liberals like, and it keeps them loyal to him, even though he hasn't accomplished anything.
00:11:41.180 But when I see plus 15, plus 12, plus 19, plus 18, it's probably this coming into effect, that the people who take the polls, when you don't do proper weighting, when you do an unweighted poll, are significantly different than what actual turnout says.
00:11:59.720 Look at that. Would have been in 2019, based on unweighted polls, plus six for the liberals. It would have been in 2021, plus seven for the liberals, when it was literally a plus one for the conservatives on election day in 2019, as well as basically a plus, you know, 0.4 or whatever, plus half a point for the conservatives.
00:12:22.120 of. I didn't put the decimal in. It could be a little higher. But it was like one-point advantages
00:12:26.340 in both of these elections. And in this one, it went from plus 10 victory for the Liberals
00:12:31.380 to just two and a half. You're getting my point here. So I don't want to just kind of rant on
00:12:39.220 about the same thing over and over again. But this is more so to make the point that
00:12:43.340 the Conservative Party has actually a good way of winning this next election. They're behind.
00:12:49.840 We shouldn't discount that they're behind. We shouldn't go and find fake numbers that tell us the Conservatives are ahead like some Twitter poll. Conservatives are probably anywhere from five to eight points behind. If Poliev and the Conservative Party start making the right moves, they can absolutely reverse that trend, catch up, get within the margin of error, and beat the Liberals on turnout. They need a bolder platform. They need a way of getting attention. They need to make the election about themselves.
00:13:16.080 Now, this is just going to turn into an election. We did our polling lesson, and now this is an election question. I agree with my friend Chris, who runs the Great Canadian Polling Channel. I actually am going to do a video on a poll that he ended up doing for British Columbia that had 1BC at 9.7%. And he's a very, very serious pollster. He does not skew polls in favor of parties he elects or not.
00:13:41.560 he is very if anything very more conservative in how he polls and i mean that from like
00:13:46.740 a temperamental basis he doesn't like he doesn't he makes sure to tamp down on sudden changes
00:13:52.880 because people don't just change how they vote overnight it's usually gradual rises and falls
00:13:58.420 and he makes sure that his polling doesn't just find people wanting to you know take the polls
00:14:03.320 for fun he wants the people who usually wouldn't take a poll to take the polls but he points out
00:14:08.300 that you always want the election to be about yourself. And I think right now the Conservative
00:14:14.340 Party needs to make the election about Pierre Polyev having bigger and bolder plans than Mark
00:14:21.000 Carney. Now, he's adjusted a little bit this way on trade, and his proposal to eliminate the gas
00:14:27.160 taxes is not bad. But I would say it's a bit tepid at the moment. Polyev, if he wants to build a
00:14:33.860 reverse the five to eight point gap the Liberals have, which if we went to an election right now
00:14:38.460 and they got that plus five national vote result, they would win a majority, probably 190 seats,
00:14:44.600 200 seats. You're going to reverse that. The election has to stop being about Mark Carney
00:14:49.820 being the man in the arena against Donald Trump, because that is what the current ballot box
00:14:54.560 question is for Canadians. If an election happened right now, just like it happened last time,
00:14:59.740 You can't simply sit there and be the government in waiting that a lot of conservatives like to be. That means you're going to lose. Government in waiting means that we're going to sit here quietly and take over if he screws everything up.
00:15:12.340 I can tell you, Carney is not going to screw everything up. He's not doing a good job, like at all. But he's not going to burn everything down to the point where people are going to go and search out another party to take over. You need to grab that brass ring really hard.
00:15:28.660 You need to run in there and say, I'm going to cut taxes 20% across the board, including corporate, and take a point off the GST for good measure. And Carney is letting the Americans out-compete us because he refuses to do this serious tax reform I'm proposing. That's what the conservatives need on social policies, on cultural policies, on fiscal policies. You need something that is so transformatively different.
00:15:51.460 It makes people look away from the Mark Carney show and towards the Pierre Polyev show.
00:15:57.660 Make the narrative not about Carney being the serious man standing up to Trump.
00:16:01.460 Make it about Polyev being the guy with big ideas and Carney being the guy who's puttering along, just leading us into the climb.
00:16:08.900 That's what the next election needs to be about.
00:16:11.240 But I think conservatives, when they see, frankly, bad polling numbers, polling numbers that aren't even accurate, they get tight.
00:16:17.760 They get uncomfortable.
00:16:18.920 They stop moving.
00:16:19.680 They stop making choices.
00:16:20.940 And it's like politics is a battle of movement. It's a battle of momentum. You've got to be making big, you've got to be doing big things and getting a lot of attention and proposing big changes. Who shows up to the ballot box to vote for small change? Most, us as conservatives, we'll show up no matter what, probably next election, because we hate the liberals.
00:16:40.300 But if you're a guy who doesn't really watch politics, you're a plumber who gets tired at the end of the day, he works hard, and it's 7 p.m., and you're driving home, do I really want to stop the car and vote?
00:16:53.500 You'll want to run on something so big, that guy's not going to think that, he's just going to go and vote.
00:16:58.580 And that's why the Conservative Party needs to stop simply playing prices right with the Liberals.
00:17:04.220 It's a myth whenever people say, oh, the differences between the conservatives and liberals are so small. They're basically no different than each other. I hate that stupid uniparty narrative. They are significantly different. If you follow politics, you know the differences.
00:17:19.200 If you're a member of the general public who doesn't understand the ideological differences or foreign policy differences, and you don't take politics that seriously, you might be tricked into thinking there's not much of a difference and there's no point in voting.
00:17:33.660 So the conservatives need to run on something that makes the differences extremely clear.
00:17:38.880 If you're going to be the pro-business party, the free market party, be that party in extremely neon colors.
00:17:48.680 Do it hardcore, or you're not going to get noticed by the person who usually opts out of even voting.
00:17:55.500 Anyways, so that should be it for today's video, guys.
00:17:59.020 Hopefully you appreciated this little lesson on the board on why pollsters so often overestimate liberals and underestimate conservatives.
00:18:06.940 take this into account whenever you look at pollsters in the future. Don't always assume
00:18:12.440 they're over-polling something by four times, but that can kind of give you a sense of magnitude.
00:18:18.380 If a pollster is saying plus 20 for the liberals, probably more like a plus 5, plus 7. If it's a
00:18:23.400 plus 12, probably more like a plus 4 for the liberals or so. These days, when pollsters are
00:18:28.480 not, you kind of have to unskew the poll a little bit by taking into account that when pollsters
00:18:34.860 are not checking to make sure that their polls are not overrun by partisan liberals who voted
00:18:40.240 liberal in the previous election, that conservatives are probably being underpolled by three or four
00:18:46.320 points, and liberals are probably being overpolled by three or four points. And if it's a really bad
00:18:51.400 pollster like Ecos and Spark, probably like underpolling the conservatives by five and
00:18:56.460 overpolling the liberals by six or seven. Anyways, so with that all being said, thank you guys for
00:19:02.260 watching and I'll see you all next time.