Should we trust the polls? (Warning: This is a long discussion piece and I don't have time to edit it now so - I'm sure there are typos, logical breaks, and generally odd bits in there. You know, like most of my posts. Proceed at your own risk!).
If you're anything like me, 16 days out of this thing I'm looking for reassurance that the 2020 elections are really going like the I think they're going. While a month ago I was nervous but optimistic, I've become more confident as the month has gone on and every major event ends with Trump failing to capitalize on his opportunities and the polls remaining very stable. Senate races continue to move in Democrats favor, and overall the House continues to see suburban Republicans pushed into ever more competitive races. Today, in other words, I'm pretty confident that Dems will win the White House, the Senate, and add seats in the House. Yes, I know, some of you will curse me for saying this out loud - but you're looking at a guy who's going to get put under to get a wisdom tooth removed on Friday the 13th. I don't believe in curses.
Of course, not everyone agrees with this view. You can find any number of rosy Trump scenarios on sites like Real Clear Politics. Typically these articles point to several things: (1) polling inaccuracies in 2016 and 2018 ; (2) greater enthusiasms at rallies for Trump than Biden; (3) a continued insistance of the existance of "Shy Trump" voters; and (4) lots and lots of anecdotal feedback.
I'm mostly interested in #1 because the rest of the beliefs are just that - beliefs - and people hold them simply because they are hoping their guy is ahead. I'm not going to do a scientific deep dive in the polling - I'm not qualified to do that, for one thing, and for another lots of other people have done that. Instead, I'm just going to give my gut feel on this issue and how I read the situation as someone who has closely followed polls and elections since I was about 12 years old.
First, I believe strongly in poll aggregation. It tends to iron out the wrinkles in individual pollsters sampling, modelling, and weighting methodologies. If you look at the averages, it is very clear that Joe Biden is winning the race. If you assign all electoral votes to the current leaders in the RCP average, Biden would win 357-181. That's a pretty substantial victory and is the first reason why I am pretty comfortable that this shows something realistic about the race. Look at the battleground states that the RCP currently has going for Biden vs. going for Trump:
For Biden: NV, AZ, FL, GA, NC, WI, MI, PA, NH, IA, NE-2, ME-2
For Trump: OH, TX
First, the fact that states like Texas and Georgia are on this list (and Georgia is currently polling blue!) while Minnesota is not (my threshold is a 5 point split in the polling) says a great deal. So does the fact that the battlegrounds are all going in the same direction. The sheer volume, coupled with the locations this race is being fought out (and just look at where the campaigns are spending money and sending candidates and you will see that this list is indeed the list of competitive locations) give me a great deal of confidence. Biden could lose a great deal of the states on the list above and still be fine.
The most likely pro-Trump scenario, for instance, would see him winning the southern trio of Florida, Georgia, and North Carolina (which is a realistic possibility), as well as winning in Arizona. If he does that and also takes Wisconsin and Iowa suddenly things would be very very close ... but if Biden wins Omaha (Omaha? Yes Omaha!) the race finishes 270-268.
Bottom line, pollsters can be wrong - but I believe they've adjusted their models since 2016 and that *all* of the pollsters aren't all going to be quite this wrong - it would be historically bad, far worse than any suggested polling errors from 2016 and 2018.
Now, is it possible the world just looks completely different from the way the vast majority of these pollsters see it? One pollster is betting yes - the Trafalgar Group. This pollster is notable on RCP for essentially always being contrary. Everyone else shows Biden up 5-6 in a state, Trafalgar will show Trump up 1-2. It'd be easy to dismiss them as gadflys, like the awful Rasmussen Reports polls are, but Trafalgar actually has a pretty solid history in 2016 and 2018.
In some sense, they're worth watching. If we wake up on November 4 (or whenever the race is decided) and Donald Trump was re-elected, then the Trafalgar Group will probably succeed in toppling the polling industry entirely. I'm skeptical of them for several reasons, however.
First, they bill themselves as disruptors. Lots of people like disruptors, but that's the problem. Being a disruptor is how you make yourself known and how you make lots of money. When you start calling yourself a disruptor, it suggests you're a bit cocky and think you've hit on something that no one else that does something for a living has ever thought of before. Is it possible? Sure. It's also possible that they've gotten lucky. A high success rate over two cycles when you don't do a lot of polls is relatively easy to attain through luck. If they do it again this year it starts to look real.
Second, I've listened to Trafalgar's owner discuss their methodology. Some of the stuff they do is very smart, but an awful lot of it sounds like attempts to take anecdotal evidence and shove it into numerical form. This makes me think their modeling shows what they want it to show. That means their modeling is heavily susceptible to their biases. That may have played in their favor in 2016, but I suspect it is going to play against them in 2020.
The bottom line - I think the race is probably pretty close to what we think it is, but what really matters is not going to be the polls but who actually shows up to vote. I'm thrilled with the early voting numbers we've seen so far and believe the electorate is shaping up to look like what it needs to look like to secure a Biden win. We'll find out in a couple of weeks.
If you're anything like me, 16 days out of this thing I'm looking for reassurance that the 2020 elections are really going like the I think they're going. While a month ago I was nervous but optimistic, I've become more confident as the month has gone on and every major event ends with Trump failing to capitalize on his opportunities and the polls remaining very stable. Senate races continue to move in Democrats favor, and overall the House continues to see suburban Republicans pushed into ever more competitive races. Today, in other words, I'm pretty confident that Dems will win the White House, the Senate, and add seats in the House. Yes, I know, some of you will curse me for saying this out loud - but you're looking at a guy who's going to get put under to get a wisdom tooth removed on Friday the 13th. I don't believe in curses.
Of course, not everyone agrees with this view. You can find any number of rosy Trump scenarios on sites like Real Clear Politics. Typically these articles point to several things: (1) polling inaccuracies in 2016 and 2018 ; (2) greater enthusiasms at rallies for Trump than Biden; (3) a continued insistance of the existance of "Shy Trump" voters; and (4) lots and lots of anecdotal feedback.
I'm mostly interested in #1 because the rest of the beliefs are just that - beliefs - and people hold them simply because they are hoping their guy is ahead. I'm not going to do a scientific deep dive in the polling - I'm not qualified to do that, for one thing, and for another lots of other people have done that. Instead, I'm just going to give my gut feel on this issue and how I read the situation as someone who has closely followed polls and elections since I was about 12 years old.
First, I believe strongly in poll aggregation. It tends to iron out the wrinkles in individual pollsters sampling, modelling, and weighting methodologies. If you look at the averages, it is very clear that Joe Biden is winning the race. If you assign all electoral votes to the current leaders in the RCP average, Biden would win 357-181. That's a pretty substantial victory and is the first reason why I am pretty comfortable that this shows something realistic about the race. Look at the battleground states that the RCP currently has going for Biden vs. going for Trump:
For Biden: NV, AZ, FL, GA, NC, WI, MI, PA, NH, IA, NE-2, ME-2
For Trump: OH, TX
First, the fact that states like Texas and Georgia are on this list (and Georgia is currently polling blue!) while Minnesota is not (my threshold is a 5 point split in the polling) says a great deal. So does the fact that the battlegrounds are all going in the same direction. The sheer volume, coupled with the locations this race is being fought out (and just look at where the campaigns are spending money and sending candidates and you will see that this list is indeed the list of competitive locations) give me a great deal of confidence. Biden could lose a great deal of the states on the list above and still be fine.
The most likely pro-Trump scenario, for instance, would see him winning the southern trio of Florida, Georgia, and North Carolina (which is a realistic possibility), as well as winning in Arizona. If he does that and also takes Wisconsin and Iowa suddenly things would be very very close ... but if Biden wins Omaha (Omaha? Yes Omaha!) the race finishes 270-268.
Bottom line, pollsters can be wrong - but I believe they've adjusted their models since 2016 and that *all* of the pollsters aren't all going to be quite this wrong - it would be historically bad, far worse than any suggested polling errors from 2016 and 2018.
Now, is it possible the world just looks completely different from the way the vast majority of these pollsters see it? One pollster is betting yes - the Trafalgar Group. This pollster is notable on RCP for essentially always being contrary. Everyone else shows Biden up 5-6 in a state, Trafalgar will show Trump up 1-2. It'd be easy to dismiss them as gadflys, like the awful Rasmussen Reports polls are, but Trafalgar actually has a pretty solid history in 2016 and 2018.
In some sense, they're worth watching. If we wake up on November 4 (or whenever the race is decided) and Donald Trump was re-elected, then the Trafalgar Group will probably succeed in toppling the polling industry entirely. I'm skeptical of them for several reasons, however.
First, they bill themselves as disruptors. Lots of people like disruptors, but that's the problem. Being a disruptor is how you make yourself known and how you make lots of money. When you start calling yourself a disruptor, it suggests you're a bit cocky and think you've hit on something that no one else that does something for a living has ever thought of before. Is it possible? Sure. It's also possible that they've gotten lucky. A high success rate over two cycles when you don't do a lot of polls is relatively easy to attain through luck. If they do it again this year it starts to look real.
Second, I've listened to Trafalgar's owner discuss their methodology. Some of the stuff they do is very smart, but an awful lot of it sounds like attempts to take anecdotal evidence and shove it into numerical form. This makes me think their modeling shows what they want it to show. That means their modeling is heavily susceptible to their biases. That may have played in their favor in 2016, but I suspect it is going to play against them in 2020.
The bottom line - I think the race is probably pretty close to what we think it is, but what really matters is not going to be the polls but who actually shows up to vote. I'm thrilled with the early voting numbers we've seen so far and believe the electorate is shaping up to look like what it needs to look like to secure a Biden win. We'll find out in a couple of weeks.
