Let's Talk Sense...

--------------------------------------------------------------------------------
Sunday, October 15, 2000 Volume XXV, No. 36
Roswell, New Mexico

Quote of the Day:

" Al Gore blaming 'guns' for Columbine is like Rosie O'Donnell, blaming 'forks and spoons' for being fat."
--------a New Mexico Republican

In this issue:

"Statistical Dead Heats" NOT!

"On Friday, a new Marist poll showed Clinton and Lazio in a 'statistical dead heat.' [Emphasis mine.] The poll showed Clinton at 47 percent to Lazio's 43 percent among registered voters, which falls within the poll's margin of error of plus or minus 4 percentage points."
---AP Story, Albuquerque Journal, October 14, 2000

[This story is inaccurate, and it has nothing to do with the pollster, his sample (you may notice "registered voters" as the sample, which is useless at this point, but that is not being discussed here) or his results. Rather it has to do with the journalist's conclusions which he has written into the story. They are wrong.]

--------------------------------------------------------------------------------

The Truth about so-called "Statistical Dead Heats"

If you stop to think about it, the oft-repeated phrase "statistical dead heat" is at best a redundancy. More accurately it is a malapropism, a solecism, or some other misusage of plain language. The term "dead heat" in simple language means a tie. I believe it comes from horseracing, and the post-race photographs which show zero separation at the finish line between the frontrunners.

When used in elections it is understood to include the only "statistic" being measured, i.e. votes---or in the case of polls, the percentages indicating support for one candidate or the other.

A tie, whether based on counting votes or counting responses to a poll question is still a tie. Putting the word "statistical" in front of it is meaningless. It is the definition of redundant. (It is reminiscent of baseball announcers: "The White Sox have won eleven straight, consecutive, games in a row without a loss." Any one of the four terms alone would be enough.) Put another way, you never hear a baseball announcer say the White Sox and Yankees are in a "statistical" tie after 7 innings of play, the score is 4-4. They are merely tied---that is all the explanation that is needed.

The problem in the media right now, and the reason they have conjured up this phraseology, is that they are trying to tell you something that isn't true. They are positing the notion that a race is "tied" in the polls when it in fact is not. Thus the mumbo jumbo with the word "statistical."

Throughout this poll-driven process of media campaign coverage you have no doubt heard, or read, the term "statistical dead heat" on any number of occasions. I have. The news copy reader, whether Wolf Blitzer, Tom Brokaw or Dan Rather, will look into the camera and intone, "the presidential candidates (or some other set of candidates in some other race) nationwide (or in some state) are separated by only two points---that is a statistical dead heat."

But it isn't. If Gore is leading Bush 43-41, or if Bush is leading Gore (as most now think) 44-40, guess what? Those respective polls are NOT telling you it is a "statistical dead heat" or a "tie." One is saying Gore leads by 2 and the other is saying Bush is ahead by 4.

But wait, you say. Don't they usually say the margin of error is plus or minus 4 points, or three points, or something in that range? Yes. But the idea, pushed by the press, that you simply add the margin of error to the trailing candidate (or subtract from the leading one) and voila! they are tied, is an erroneous one at best, and an extremely misleading one at worst.

What the "margin of error" means is this: If you were to conduct the same poll, among the same number of randomly-selected participants at the same time, 19 times out of 20 (all major polls are at the 95% level of validity) you will get responses that are within the stated margin of error for each candidate. (The margin of error for a nationwide poll can be 5 to 7 points for samples of 400-800, and it can be from 3 to 4½ points for samples ranging from 900 to 1800.)

So if Smith leads Jones 44 to 43, (with a +/- 4% margin of error) what does that mean? What it means is that based on the very best job the pollster can do, Smith has a one point lead over Jones. It does NOT mean that the difference between Smith and Jones is meaningless----as the press (and the Jones spin team) will have you believe. By the way, both the Jones camp and the Smith camp know this. But both spin operations they know it can be to their advantage to let this story go without challenge----plus it is way too close to worry about it, and, besides, they have 13% undecided to go after.

What about the margin of error? In this case, the pollster is saying that if you were to take the poll 20 times, in at least 19 of them the Smith total would be between 40 points and 48, and the Jones total would be between 39 and 47. Oh, so Jones could be ahead 47 to 40? Or Smith could be leading by 9 points, 48 to 39? Theoretically, yes. But this is unlikely. And certainly just jumping to the conclusion that that is the case is plain wrong. (This is something I can guarantee you neither side will do, not even their spin machine.)

In fact, assuming anything---other than the result shown by the pollster--- is wrong. The pollster's best random sampling, best question technique and best internals (the demographic balances which must be achieved, sometimes by weighting) says Smith leads by 1.

Believe me, that is the conclusion that the actual professional consultants reach. They are not standing around musing about being in a tie, both sides know that one is up by one and the other is down by one. No one in the Lazio campaign, for example, believes his candidate is in a "statistical tie." The Lazio people know they are down by 4. Period.

Here in New Mexico, Research and Polling in 1986 published its last sampling of the Attorney General's race showing Republican Hal Stratton leading by some 15 hundredths of one percentage point. While one could definitely conclude the race was simply too close to call---and one would be right----the fact is the BEST DATA AVAILABLE led to the conclusion that Stratton would win. And it was accurate. Stratton won the race by 690 votes out of some 370,000 plus.

From the What it's Worth Department

Speaking of Lazio, at least a half dozen polls now show Lazio behind Clinton in the New York senate race by three to six points. At least one shows Hillary leading 50-43. This represents a steady drift toward the First Lady over the last six weeks.

I for one still believe Lazio will win---as I have since he got into the race. True, Hillary has finally reached 50% in a least one poll, something I predicted she would not do. But I do not believe that is a solid "definitely vote" for Hillary figure. I also believe current events worldwide will subtly work against her---and her non-New Yorker status will help undermine her. I also find it significant that Lazio has a $6 million to $3 million lead in available funds in the last month.

In the final analysis, I am trusting him and his campaign. If he doesn't have the best campaign advice available in America, then I would be both surprised and disappointed. I have to believe they can craft a message and develop an approach which will result in his prevailing by about a 2 point margin on November 7th.