In a horse race, the term "photo finish" is used whenever two or more horses are so close together at the finish line that only a photograph (and not the naked eye) can determine which horse won the race. It sometimes happens--but not very often--that the horses involved in the photo finish will cross the finish line simultaneously. When this occurs, the term "dead even" is used to describe how the horses finished the race. In a more general sense, the term "dead even" means that two or more contestants (whether in a swimming race, a car race, or any other kind of race) finish the race in a tie.
The term "dead even" appeared in yesterday's newspaper in reference to a recent poll of voters. As you know, lots of polls are being taken these days as we approach this year's elections. Most of them, of course, center on the presidential election, and that was the focus of the article I read in yesterday's paper.
The opening sentence of that Associated Press article stated that "the Republican presidential candidate is dead even with his Democratic foe." This statement was based on a poll of registered voters who said they were likely to vote in the upcoming election. Over 600 voters participated.
Based on the statement that the poll showed the two candidates "dead even," most readers would surmise that they both had support of the same number of the people polled. However, the 2nd sentence in the AP article indicated that one of the candidates was ahead of the other candidate "46 percent to 43 percent." It my guess that not just a few readers scratched their heads after seeing these two percentages immediately after being told that the two candidates were "dead even." Are you scratching your head?
Instead of scratching your head's exterior in dismay, it's my hope that the term "confidence interval" popped into consciousness inside your head! If you did, in fact, think of this statistical concept as way to reconcile the "dead even" and "46 vs. 43" sentences, give yourself a big pat of the back.
Each candidate's support level (46 percent for one, 43 percent for the other) come from a sample of likely voters. Hence, instead of treating those percentages as population parameters, it's important to view them as sample statistics. Because of the likelihood of sampling error being present in each candidate's percentage, neither one should be trusted as indicating the level of support among all likely voters (on the day the poll was taken. Each percentage is probably a close guess as to its corresponding population parameter, but neither one is likely to "hit the nail on the head."
In the newspaper article, I read that the "margin of error" was plus or minus 4 points. If you build a confidence interval around each candidate's percentage by adding and subtracting 4 points, you'll end up with one interval that extends from 42 to 50 and another interval that that extends from 39 to 47. Because these confidence intervals "overlap" each other, the two presidential candidates were said to be "dead even."
On some radio and TV reports, I've heard the phrase "statistically tied" to describe a situation like the one I've described here. I prefer that term to the phrase "dead even." Your preference may be the other way around. Be that as it may, I hope you can see why a political poll can be interpreted as indicating the candidates are "even" despite the fact that the poll showed them to have different levels of support.
Copyright © 2012
Schuyler W. Huck