Polls: Much Ado About Margins of Error

Polls - love them or hate them - are here to stay. Polls may be statistically significant ("outside the margin of error") or statistically not significant (within the margin of error).

Often people calculate statistical significance as twice the reported margin of error of the poll. This is done following the logic that if one candidate is at 55% and the other at 45% and the poll has a ±5% margin of error, then the first candidate could be as low as 55−5 = 50% and the second could be as high as 45+5 = 50%. In this case reporters would say the race was a statistical dead heat because the gap between the candidates (55 − 45 = 10%) is not more than two times the margin of error of the poll (5%).

There are other factors to consider: how recent was the poll, how large was the sample size, does the pollster have a reputation for bias, and are there other polls with which to compare it. If there are more than two candidates in the race or a substantial number of undecided respondents the results may be confounding. In general, however, where the results of one candidate versus the other are more than twice the margin of error, there is considerable likelihood of statistical significance.

Polls for the Senate race in New Jersey have consistently shown Senator Bob Menendez in the lead over State Senator Joe Kyrillos. Real Clear Politics provides him with an average 18.5 point lead based on four recent polls which have Menendez up 22, 20, 18, and 14 points. Reviewing the margins of error of these particular polls suggest they are all statistically significant.  

Polls for the Presidential race in NJ have also consistently shown President Obama in the lead over Mitt Romney. The five most recent polls have Obama up 14, 15, 7, 8, and 11 points. Real Clear Politics provides him with an average 11 point lead. All the polls are statistically significant, except for the one with Obama leading by 7 points.

In our congressional districts there are two strongly contested races which are leaning Republican. In CD 3 I know of only one poll, the the Stockton poll of October 4, which indicates incumbent Republican Jon Runyan leads Democratic challenger Shelley Adler by a margin of 49 percent to 39 percent among likely voters. The margin of error is +/- 4% suggesting it is statistically significant. Unfortunately I know of no other poll with which to compare it. In CD7 where incumbent Republican Leonard Lance is challenged by Democrat Assemblyman Upendra Chivukula, I have seen no poll.

Regarding the presidential tracking polls nationally, today's Real Clear Politics (RCP) has  47.1% for President Obama and 46.9% for Mitt Romney. The most significantly outlying poll, the Gallup Poll, as of today had Obama 45% and Romney 52% - a 7 point Romney lead with a +/- 2% margin of error - rendering it statistically significant. The other eight most recent polls in the RCP computation have Romney ahead by 2, 1, and 1 points; Obama ahead by 6, 3, 3 and 1 points; and one tie. Only the Gallup Poll holds statistical significance. The Presidential Electoral College Maps, based on state polls, show considerable variance. RCP has Obama with 201 likely/leaning votes and Romney with 206 likely/leaning votes and 131 toss up votes. POLITICO apportions the toss up states and has Obama with 277 votes (enough to win) and Romney at 261. Anyway you look at it, this race is a cliffhanger filled with statistical uncertainty.

Another pesky thing about polls is they can change. So with little more than two weeks to elections, stay tuned and check those margins of error.  

Go to NJ State Page
Category: 
origin Blog: 
origin Author: 
Comments Count: 
0
Showing 0 comments