Joined 2005 and asking that? Is this the real cardsharkk04 or a troll account?

Why didn't you just ask voting preference as part of the Census? We just had a full enumeration this year! could have settled this months ago imo.

Underrated post right here.

I didn't work on Census design. I only worked on survey design.

I was sent out West to do statistical quality control work at our Western Processing center for the 1980 Census. But this had nothing to do with what the questions were.

Also, I think it's against the law to ask about voting preference on the Census.

Best wishes,
Mason

Mason, no comment on the Trafalgar guy's offhand claim that Trump needs to win Pennsylvania "by 4 or 5, to overcome the voter fraud that's gonna happen there"? Does that affect your opinion of his political acumen at all?

My opinion has not changed, but you have no understanding as to what my opinion is. That's too bad because you might learn something about how a professional statistician trained in this stuff goes about forming an opinion.

Specifically, when someone goes against what everyone else is saying, they're not likely to be correct. However, in this case, the Trafalgar people have already been right when most everyone said they had it wrong.To me, that implies that more data needs to be collected before a strong opinion should be formed.

It's also the reason why in other fields when a statistical test shows a surprise result, before it's accepted, statisticians want to see the result duplicated in another test. Put another way, when your Bayesian information says that something can't be right, it's time to retest.

The reason you're unhappy is that someone is getting some positive publicity when he takes a position whose outcome is something you don't like. A good statistician would never think like that.

Mason

Is Cahaly a "professional statistician"? His Wikipedia page says he has a BA in Political Science. It may not be exhaustive but a quick search showed no other degrees.

If you're just talking about yourself, I guess a "professional statistician" doesn't let his opinion of people be influenced by those people saying things that are batshit crazy on their face? If so, I did learn something new! Thanks.

Also, how much have you invested into taking a full accounting of Trafalgar's results vs. other pollsters? Not noted in that fawning RCP article posted earlier is the fact that he bucked the trend in Nevada polling in 2018, coming up with Heller +3 in a race that Heller lost by 5, and also being wrong in calling Arizona for McSally. The argument seems to be "he has a lot of results that were right", but is it known that other outlets are not as right?

Similarly, you called them "the most accurate pollster of 2016" or something like that, when it seems like they just had a lot of wildly pro-Trump polls that got lots of attention in the 2 states where being biased towards Trump by a few points (which they were in almost every state) ended up looking the most prescient. I hear much less about their Colorado D+1 poll! Shouldn't an honest assessment of their results involve looking at the totality of them, relative to other pollsters?

Mason seems awfully invested in knowing exactly one thing about Trafalgar and then talking down to people who know more things as if the one thing is all anyone could know and therefore their added certainty is unfounded.

I know very little about Trafalgar and suspect you know more than me. So, let me ask a question. How many employees does Trafalgar have and do you think any of them are statisticians?

Mason

Looks like it comes down to what % of the electorate has already voted.

Hi Mason,

I want to hear your views on the recent phenomena of this notion of the quiet or silent conservative as discussed in the below article of the conservative voter who is not willing to divulge their true voting intent to a pollster compared to their actual voting intentions on the day:

https://www.politico.com/news/magazi...te-2016-433619

Is this something that you have noted in your analysis of recent polls especially in 2016? Also internationally as we saw a landslide to the conservatives in the last UK general election that the polls didn't predict and a win to the conservatives in Australia in the last federal election when polls were predicting a slight progressive win? Also does COVID dilute the full effect of this that we saw in recent elections?

Thanks.

Would also like to hear this. "Shy Tories" was the explanation for the UK conservatives' win in 1992 so this goes back a long way.

IMHO the best defence against it is to weight by previous vote and that's what we do in Britain. Someone who's ashamed to say they're voting conservative is also not likely to openly say they voted conservative in the previous election either, so in that case they aren't going to skew the picture of where last times' voters for various parties are going this time.

The polls did predict this. The ones taken the same week of the election (12 Dec was a Thursday, so any taken on or after 9 Dec) were basically dead on.

https://en.wikipedia.org/wiki/Opinio...neral_election

(obv being Australian you know, but for those that don't know, the UK election is for the lower house of parliament and the majority leader in the house becomes prime minister)

The polls taken earlier in the campaign showed a lead of about 10 points all the way through. Also the cross-tabs of the polls saying that "Brexit" would be a big election issue for a disproportionately high number of people whose 2016 vote was "Leave" rather than "Remain" made it clear that the 3rd party voters would be squeezable for pro-Brexit tactical voting (in favour of the Conservatives) and not for pro-Remain tactical voting. That's not hindsight either: https://forumserver.twoplustwo.com/s...&postcount=144

The really interesting thing about the election was that the media and the betting markets were so convinced the polls were wrong:

(NOC means no overall control, i.e. no majority, 2.18 means +118 in US odds - i.e. you give 100 pounds to the betting shop and you come and back and collect 218 pounds if the selection wins)

When this was posted the most recent polls were Conservative +12, Conservative +16 and Conservative +12.

That's your mistake. Trafalgar's results aren't a surprise. Anybody could have told you months ago that everything they release would be significantly more republican than the averages. Just by scanning their twitter page and seeing that their founder wrongly thinks there is massive democratic voter fraud anybody with any background in American politics could predict that they will release polls titling R.

This is nothing more than your buddy who won a huge bet last week picking the Bengals to beat Tennessee telling you the Bengals are favorites to beat Pittsburgh this week. And oh yeah, by the way he's a huge Bengals fan. It's kind of silly to take a need more data approach on his "system" and that is not the same thing as saying the Steelers can't lose this week.

First, I haven't analyzed any polls. Second, the idea of a shy responder who will answer certain questions incorrectly is something that I have no familiarity with.

Mason

Their polling results aren't surprising because they showed similar results before and were right. And since they were right, why would they want to change things.

I think it's clear they're doing some different things in their data collection and perhaps their weighting. That's why I brought them into this conversation. We'll know shortly whether they got things right.

Mason

They've also shown similar results and were very, very wrong. Why do you keep insisting that the one outcome you know about is the only one anyone knows about?

It's like you're not even reading the posts you respond to.

Does vegas have odds on who's going to win the election? If so, who's in the lead?

Not vegas, but the prediction markets have Biden at 64%.

This, rather than Trafalgar et al, is the best reason to think a Trump victory is underrated imo.

You keep saying this but I linked to an article above which says the opposite.

MM

Hmm:

It was a stupid article. Let me ask you this. Two pollsters offer up polls of equal samples sizes and stated confidence interval, one says the Democrat leads by a point. The other says the Republican leads by three. The final result is the Republican by 0.5 points. Which pollster was more accurate?

Your link is just wrong.

The Arizona race is now decided and they missed. They also missed Nevada and that was curiously left out. Any other misses fortuitously left off? Nobody has really checked. And in some of the ones they got "right" they were way off, like overstating Cruz's win by 7 points. If you just look at the least squared error for Trafalgar vs the RCP average in 2018 they lose. And they've never successfully called a state race with as much difference as they are showing right now vs the RCP average in a few of the states they have Trump up in; Michigan is one example.

I don't have a problem with Trafalgar Group. They have a different read on the electorate than I do, but they are putting their reputation on the line behind it, so good on them. However, they are one of the worst-rated pollsters at 538, so I would be leery of relying on a few cherry-picked high-profile outcomes in a news article to support a claim to their overall accuracy.