Prior to Super Tuesday, Zeeto ran three polls in North Carolina, California, and Texas that surveyed registered Democrats who expressed being likely to vote in their primary for the 2020 presidential race. This report assesses how well our polls fared in predicting the top three candidates of each race.
In North Carolina, we accurately predicted the top three candidates in the correct order: Joe Biden in first, Bernie Sanders in second, and Michael Bloomberg in third.
We predicted Joe Biden would receive 41.8% of the vote, which was extremely close, off by 1.2 percentage points.
We predicted Bernie Sanders would receive 22.7% of the vote, which was also extremely close, off by 1.4 percentage points.
We predicted Michael Bloomberg would receive 19.1% of the vote, which was overshot by 6.1 percentage points.
In California, we accurately predicted the winner Bernie Sanders as well as the runner up Joe Biden. We incorrectly predicted Michael Bloomberg in third who was slightly beaten by Elizabeth Warren.
We predicted Bernie Sanders would receive 31.1% of the vote, which was relatively close, off by 2.9 percentage points.
We predicted Joe Biden would receive 22.9% of the vote, which was undershot by 4.3 percentage points.
We predicted Elizabeth Warren would receive 7.2% of the vote, which was undershot by 6.1 percentage points.
In Texas, we accurately predicted the winner Joe Biden. We predicted Michael Bloomberg slightly beating Bernie Sanders, but the order was switched with Sanders beating Bloomberg by 14 percentage points.
We predicted Joe Biden would receive 34.4% of the vote, which was nearly exact, off by .1 percent.
We predicted Bernie Sanders would receive 23.7% of the vote, which was undershot by 6.3 percentage points.
We predicted Michael Bloomberg would receive 24.9% of the vote, which was overshot by 10.5 percentage points.
We accurately predicted the winners in California, Texas, and North Carolina. With the exception of Texas, we also accurately predicted Joe Biden and Bernie Sanders at the top of each race and in the right order. Across the three races for the top three candidates, we had one relatively close prediction that was off by less than 3%, two extremely close predictions that were off by less than 1.5%, and one nearly exact prediction that was off by .1 percent. Across the three races for the top three candidates, our combined predictions were off by 41.5%. However, in just accounting for the top two across the three races, our combined predictions were off by only 18.8%. This means that we were much more accurate in predicting how much first and second place would win by than we were third place, i.e. as much as political pundits, we were surprised by Michael Bloomberg’s poor showing on Super Tuesday. Looking at the predictions by state, we were most accurate in predicting North Carolina, with two extremely close predictions and off on our combined predictions for top three candidates by 8.7% (off by 2.6% for the top two candidates ). We were second most accurate in California, with one relatively close prediction and off on our combined predictions for the top three candidates by 13.3% (off by 7.2% for the top two candidates). As a whole, while we were least accurate in Texas (off on our combined predictions by 16.9%), it’s where we made our single most accurate prediction of the winner Joe Biden, which was nearly exact, off by .1 percent.