By Jim Ellis
July 5, 2016 — Often, people want to know which are the most accurate pollsters within the huge volume of data being released into the political marketplace. Bloomberg Politics reviewed the 2016 presidential pollsters who routinely release their numbers into the public domain and publicized their ratings late last week. They isolated a dozen polling firms and corresponding media organizations that picked the presidential primary winner in at least 70 percent of their studies. Polling frequency and reliability stats were also recorded.
The firm projecting the primary winner most often — 95 percent of the time — was Gravis Marketing, which conducted 19 surveys in 12 different states. The students who comprise the Emerson College Polling Society in Massachusetts, who have fared very well in previous performance review studies despite being amateurs, placed a close second in win predictability, proving correct 94 percent of the time. They polled 16 times in eight states. Completing the top five are Opinion Savvy (91 percent winner accuracy; 11 polls; eight states), NBC/Wall Street Journal/Marist College (90 percent winner accuracy; 29 polls; 15 states), and the Survey Monkey (90 percent winner accuracy; 10 polls; eight states).
But capturing the correct victory margin proved to be another story for all 12 pollsters. None of them came within the average polling error margin pertaining to victory spread projection in relation to their cumulative polling average.
The most accurate entity predicting the margin between the candidates was Fox News, whose polls are jointly conducted by Robbins Research and Shaw & Company Research. Their margin factor was an average of 6.3 percent away from the actual result in 16 polls from nine states while correctly predicting the winner 88 percent of the time. The worst in the victory margin category was the Survey Monkey who missed the spread by an average of 10.9 percent. CNN/Opinion Research was second to last, missing the win split by an average of 10.7 points per study.
Interestingly, despite having the best win predictability rating, Gravis fared particularly poorly within the Democratic contests. Here, their miss ratio was 12.3 percent, the second-worst number in the entire field. They rebounded well on the Republican side, however, recording a miss factor of only 4.1 percent, which was the single best score.
Quinnipiac University had the most accurate projections for their Democratic polls, falling only 6.1 percent away from the final result, but their 10.7 percent miss rate in Republican polls kept them from doing better cumulatively. Overall Quinnipiac had a 77 percent win projection rating from 13 polls in seven states.
The cycle’s most prolific pollster is Public Policy Polling. They conducted 37 polls in 26 states, with an 84 percent win predictability rate. They were an average 9.7 percent away from the actual victory spread, which ranks ninth of 12, but they were consistent between the Democratic and Republican races. Their miss rate was 9.8 percent for the Democratic polls versus 9.5 percent on the Republican side.
The remaining firms are Monmouth University (88 percent win predictability; 7.3 percent miss rate; 25 polls; 13 states), YouGov (80 percent win predictability; 7.6 percent miss rate; 30 polls; 16 states), and the American Research Group (71 percent win predictability; 10.6 percent miss rate; 24 polls; 13 states).
These 12 pollsters, according to the Bloomberg performance review, were rated as the most reliable so far in the presidential polling season. Like most modern pollsters, dealing with polling sample challenges involving the land line to cell phone ratio, we see firms hitting certain races within a point or two, and then missing others by very wide margins. Thus, we continually see a void in finding a very few firms that can always be counted upon as being consistently strong.
