A month to go and the winner is?

Image by WordPress.com

In this post, I present hybrid and pure polling model forecasts for the president and the Senate.  The embedded hyper-links allow you to click on the organizations to check whether prediction change significantly over the next couple of weeks.    

Recall, hybrid prediction models rely heavily on state and national surveys, and then factor in certain economic and political conditions.  Hybrid models also apply advanced statistical reasoning to generate predictions for states, the nation, the Senate and House.  

Pure polling models are based entirely on surveys.  The models either report their own survey outcomes or combine available polls from other organizations and report an average.      

Because both hybrids and pure polling models are based on a rolling average of polls, predictions can fluctuate significantly – sometimes daily.  Remember, polls vary in their sampling approach, definition of voters, time of surveys, questionnaire, mode of communication – phone, internet, human interviewers, and automation. 

Therefore, be aware that by combining surveys, the accuracy of predictions depends on the quality of those surveys.   

Let’s first look at the forecasts of two well-known hybrids models. 

FiveThirtyEight Forecast

Biden is strongly favored to win.  On Sunday October 4^th, Biden’s chance of winning is 81%.  Trump’s chances are slim at 19%.    Across the entire time series provided, FiveThirtyEight projected Biden as clear winner.  It has never been close.  Trump’s best chance of winning occurred on August 31 at 32%.

	Biden	Trump
wins president	81%	19%
popular vote %	53%	45.8%
electoral college vote	335	203

Chances Democrats win Senate = 64%

The Economist Forecast

This is perhaps the most sophisticated and comprehensive model.   The Economist thinks Biden is very likely to beat Trump.  In fact, popular vote projections since June are remarkably stable – Biden in mid-50% and Trump consistently hovering around 45%.  On March 2^nd, before the pandemic shutdowns in the United States, Biden was still projected to win 53.1% of popular vote and Trump 46.9%.  On October 4th, the Election day prediction has Biden at 53.8% and Trump 46.2%. 

	chances
Biden wins president	81%
Biden wins popular vote	98%
Biden wins electoral college	89%
Democrats win Senate	69%

Now, let’s examine 3 pure polling models.

Real Clear Politics Average

Journalist often refer to the Real Clear Politics average.  Hundreds of poll results are tabulated, and an average published every day.  Instead of probabilities for the Electoral College or Senate winner, Real Clear Politics uses terms like “solid” Democrat or “leans” Republican or “likely” Democrat, and “toss ups” for state contests. Remember a candidate requires a majority of electoral college vote to win – 270.     

	Biden	Trump	Toss up
popular vote	50.6%	42.5%
electoral college	226	125	187
Senate	46	46	8

YouGov

YouGov does use some statistical modeling.  However, I categorize YouGov as a pure polling model because predictions are derived from over 60,000 survey interviews. 

	Biden	Trump
popular vote	53%	44.9%
electoral college	350	188

Rasmussen Reports

Rasmussen is one of the few polling firms that predicted a close race in 2016 and forecasted a Trump victory.   Their latest national survey shows Biden leading Trump 51% to 43%.     

Bottom line

The hybrid and the pure polling models draw generally the same conclusion.  Joe Biden wins comfortably.  And, the Senate flips to Democrats. In fact, the polling data for the presidential race exhibits a stability notably absent four year ago.  Forecasts from this spring and early summer are nearly the same today. The presidential race shows very little movement – voters are dug in.    

Of the models presented, only Rasmussen and Real Clear Politics open the door for a possible Trump win.  Prior Rasmussen forecasts showed a much closer race than other models, and at times Trump led Biden.  Real Clear Politics shows Biden does not yet have enough Electoral College votes to win. Likewise, there are 8 ‘toss up’ races that make the Senate to close to call.

We shall see whether events this past week alter the race.  However, most of the prediction modelers consider the contest effectively over, regardless of what may transpire.