Biden campaign made decisions on ‘probability’ of a polling error: Annie Duke

Former professional poker player Annie Duke joins 'Influencers with Andy Serwer' to break down Joe Biden's successful presidential campaign.

Video transcript

ANNIE DUKE: So sort of taking this election aside, which I think is still very hot in people's mind, we can think about Hillary Clinton as well. So you know, again, we don't really think about process. We don't really think about what people knew at the time as we're sort of evaluating people's decision making. We really just think about the result.

So we know that Hillary Clinton did not spend a ton of time campaigning in Pennsylvania, Wisconsin, and Michigan. And we also know that she lost those three states by a combined total of somewhere around 80-ish thousand votes across the three states. So it was a very narrow margin.

But you could see the takes start coming out, you know, oh, you know, this was such a terrible decision. She-- such horrible strategy. Why wasn't she in Pennsylvania, Wisconsin, and Michigan? You know, this was her fault that she lost, because she made such bad decisions.

So here's the thing about that, and this is how we can sort of see that this is a resulting problem as people have these takes, is that we need to think about what is the information that she had at the time of the decision, right? So could she have known that she should have been in those states? Well, the input into that decision is going to be the polls. And what we found out after the fact that there was quite a large polling error, that that polling error happened to go in Donald Trump's favor in 2016.

And-- and so when you think about the input into her decision making, which is where am I ahead, where am I close, you know, she was very close in Arizona. She was very close in Georgia, very close in North Carolina, very close in New Hampshire, very close in Florida, and that's where she was spending a lot of her time. And it looked like she was very far ahead in Pennsylvania, Wisconsin, Michigan.

We find out after the fact, by definition, that there's a polling error. And so can we really expect her to have known that beforehand? Well, there's an easy way for us to figure out whether that was a mistake and she should have known about it beforehand.

You can look on a search engine, like Yahoo. You can say, what would happen if we search to see whether people were writing about this, what feels like an obvious polling error after the fact? And the answer was nobody was.

In fact, almost all of the writing, and there was very little about it, were questioning why Donald Trump was campaigning in Pennsylvania at all and saying, why would he be there? He should be in places where the polls are much closer. And in fact, they were suggesting he should be spending more time in Ohio.

So this is a really good case of resulting. She lost, therefore the decision making must have been bad, as opposed to she lost because there was a problem with the polls that nobody could have known about except after the fact, because that's when you see the barrage of-- of articles appearing. So that's one of the ways that you can figure out that this resulting problem is happening.

So in the Pete Carroll case, we can go back and do the math. In the Hillary Clinton case, we can go back and say, was it reasonable for her to know? And what you can see pretty quickly is no, it certainly was-- wasn't at all reasonable for her to know because nobody knew. And every Silicon Valley data analyst was going to be all over that, and political observer, and whatnot. But this is really where we see that we take very bad lessons from history.

ANDY SERWER: Pollsters still have some work to do, apparently, if you look at this election, Annie. I don't know if you have any thoughts on that or if you wanted to talk about--

ANNIE DUKE: Yeah, so--

ANDY SERWER: --sort of another problem.

ANNIE DUKE: Sure. I mean--

ANDY SERWER: But go ahead.

ANNIE DUKE: You know, I think that this is a good example, right, of what do you know and what don't you know. So we know, for example, that Biden was polling very far ahead in Pennsylvania, Wisconsin, and Michigan. And yet when you look at his ad spending, that's where most of the spending was, including the PACs, because I think that at that point they were taking into account that there was-- that there could possibly be a polling error again.

And I think that that shows the difference between what can you know and what can't you. So having seen that polling error in 2016, I think that what they said was, OK, there's a non-zero probability that this polling error is occurring again. And you can see that their-- their resource allocation actually-- actually took that into account. So that's a very good example, actually, of resulting versus not resulting, right, which is you see that you learned something new.

And this is the only problem that you can get into in decision making. You learn something new and then you don't take it into account for the next decision. So learn something new in 2016, something that Clinton couldn't have known about it. But as we go into this, Biden can know about it. And you see that the resource allocation was taking into account a possible polling error.