There is a mystique that the most recent poll is the most important and most accurate, so the eager press reports it with much excitement. Baloney. Let’s not forget that Gallup, the most respected pollster in American history, projected in its last-minute final 2012 poll that Mitt Romney would win.
So let’s take a look at today’s Republican presidential nomination polls. Real Clear Politics this morning provides the six most recent national polls conducted with an ending date between November 30 and December 9. For Donald Trump his results range from 27% to 36%. Just in two polls ending one day apart Trump was at 35% in one poll and 27% in another. In the case of Ted Cruz his range was 13% to 22%, and for Marco Rubio it was 9% to 17%.
Which single poll mentioned above is important and worthy of being trumpeted? Probably none. The best approach is to look at the average of recent polls. It was correct in the 2012 presidential election and provides a better measure than just one poll.
Retrospectively we know pollster results vary in quality. In the 2013 NJ senatorial and gubernatorial races Monmouth University polling proved more accurate than Rutgers-Eagleton. Some polls vary in whom they query: registered voters, unregistered voters, people who say they are likely to vote, people who say they will vote, and end up providing different results. Some pollsters today are better at reaching those who do not have landlines. There are statistics, demography and other factors which can improperly skew the results.
There is always a margin of error, often of 4% to 5%, which confounds results particularly in the current Republican presidential race. Now at the bottom of the higher tier is Ben Carson at an average of 13.2%, immediately followed by nine candidates who range from 4.2% to 0%. Christie is at 2.7%, but a more accurate projection might just as easily have him ahead of Jeb Bush at 4.2% or below Rand Paul at 2.2%. Any of these candidates in a perfect poll could be several points higher or lower.
My advice is that when a news outlet boldly states on a given day that Trump is surging or tanking, or Christie is rising or sinking because of that day’s poll you should ignore the claim. Pay more attention to the recent average and realize that today’s hot news can turn cold quickly.
In spite of the above, I wish we had more polling on our NJ congressional and legislative elections. In races where a poll indicates a close one it might stimulate more donor, volunteer and voter participation. Some candidates can not afford expensive polls and their knowledge of what groups or regions support them and their expectations of results might be way off. Ideally to gain the best data possible there would be several polling groups presenting results on more key candidates and doing so on multiple occasions. In this year’s Assembly election with scant polls the perceived wisdom was that LD 1, 2, and 38 held the contested races. As it turned out LD 38 soon lost its contested status and LD 11 and 16 emerged with three upsets.