According to the polls, who is winning the elections in the United States?

What does this mean? The two figures above represent probabilities: the ones that say which candidate has the best chance of winning. You can interpret that information as a warning: the forecasts say that Biden is the favorite, but when they were as sure in the past as they are now, they ended up being wrong 20% ​​to 25% of the time.

Another way to look at it is to imagine a tree of alternatives: out of every 100 possible futures, the numbers above say how many Biden wins and how many Trumps. What we do not know is which of those futures will be ours.

Where does the prediction come from: polls

It is an average of four predictions that we will see below: two statistical models, betting and expert forecasts. But all of them feed off the electoral polls. The graph shows the two basic figures to follow the race: 1, the percentage of vote of each candidate (an average of polls), and 2, their probability of victory (our prediction).

The graph shows how polls and forecasts have changed since January. Hovering over the graph, it’s easy to see, for example, that Trump lost options when it was confirmed that he had been infected with coronavirus.

It is important to understand the difference between probes and probabilities. The first figure (the polls) tells how many people will vote for Biden or Trump. Probabilistic prediction goes further. It is an attempt to estimate the chances of winning that each one has. The predictions are supported by the polls, but they also consider other factors, such as the performance of the economy, the characteristics of the electoral system (they know that Biden will need more votes than Trump to win), and a fundamental one: the possible error of the polls.

The situation State by State

The map represents the advantage in points that one candidate has over another in the polls and their odds of winning according to Fivehtirtyeight. Each hexagon represents one delegate.

On the map we see the importance of the presidential election mechanism. Each territory distributes a certain number of delegates —hexagons—, all of whom go to the candidate with the most votes: the winner takes everything. In total, 538 are distributed and the presidency is won with 270. In the graph we classify the States into five groups: those “safe” for Biden or Trump, those with favorites and those tied.

More about our probabilistic prediction

It is the result of averaging the forecasts from four different sources, which we consider the most reliable. Each has its advantages: statistical models are clearer and more systematic, but experts can react to complex or unique events. Taking an average is a consensus.

The forecasts offered by the political betting market, collected by Real Clear Politics.

The predictions of the quantitative prediction experts at Good Judgment. It is a project based on the investigations of Philip Tetlock and Barbara Mellers.

The weekly model in collaboration with statistician Andrew Gelman. it uses surveys, but gives considerable weight to other fundamental variables (eg economic).

The FiveThirtyEight statistical model, developed by the statistician Nate Silver for a decade. It mainly feeds on surveys.

Here’s how the forecasts from the four sources have evolved. The graph represents the odds of winning for Biden and Trump on each date:

These are the most recent predictions:

Why probabilities besides polls?

These four probabilistic predictions are fed by polls, which are the primary source of information for forecasting electoral results. But they also enhance them with three advantages:

  • They predict more than votes. These forecasts take into account other keys, such as the distribution of delegates or the possibility that a handful of States end up turning the election around. Simulation models are especially useful for exploring all of these scenarios.

Credits They have also collaborated on this information Mariano zafra, Daniele Grasso Y Borja Andrino.

Read More