September 14, 2000
7:31 AM   Subscribe

The lead that Al Gore once enjoyed in most presidential polls in wake of the Democratic convention has all but disappeared, according to the latest Battleground 2000 poll released Thursday. The survey reports that the vice president’s numbers are slipping and George W. Bush is back on top for the first time in weeks.
posted by aaron (22 comments total)
For another interpretation of what's happening, check out (evil) Slate and (unscientific but amusing) IEM.
posted by johnb at 10:58 AM on September 14, 2000

There have been around a dozen polls released in the past two weeks that have Gore leading the race. Hmm. I wonder why Aaron didn't post a story about a poll until he found one where Bush was leading?
posted by rcade at 11:20 AM on September 14, 2000

Frankly I'm kind of embarrassed for the GOP this year. There are plenty of smart, able conservatives. I had some respect for Dole and for McCain. I didn't want them in the White House, but I was interested in what they had to say. Bush the 2nd doesn't seem to have anything to say. His speech at the GOP convention was pure helium-- lighter than air. He and his staff keep shooting themselves in the foot admitting things like he doesn't like to read reports longer than 2 pages and he doesn't like meetings and gets bored easily. The debate dispute is comical; the more Bush tries to squirm out of the party-arranged debates, the worse he looks. I'd love to hear some actual reasons why conservatives think Bush should be president, because the only positive thing even the GOP could come up with when he was nominated was "he's electable", which is looking less likely all the time. He's as vacant as Reagan at his worst, but at least Reagan hit his marks and got his lines right.
posted by wiremommy at 11:40 AM on September 14, 2000

I kind of admire the republicans for nominating GWB. They are usually against affirmative action, but in this case they overlooked more capable candadites in favor of one with a mental handicap. Good for you, Republicans! Why should smart people be the only ones in politics?
posted by Doug at 12:02 PM on September 14, 2000

Nice bait, rcade. I don't post polls that aren't news. A switch is news.
posted by aaron at 12:14 PM on September 14, 2000

I don't think any online poll is accurate. It is skewed because of the people who use the internet: people with enough money to have a computer and an ISP.

I didn't see a link showing the margin of error, the sampling size or anything else.
posted by terrapin at 12:48 PM on September 14, 2000

If one goes to the "poll center' the first headline is "Gore Ahead by Seven" So which it? ;)
posted by terrapin at 12:52 PM on September 14, 2000

Where did it say it was an online poll? And in fact it did show the sampling size and margin of error. You did follow the link didn't you?
posted by gyc at 1:42 PM on September 14, 2000

The survey, based on a four-day rolling sampling of a total of 1,000 likely voters, was conducted by Democratic pollster Celinda Lake of Lake, Snell, Perry & Associates and Republican pollster Ed Goeas of the Tarrance Group.

The most significant finding from the poll was that among a select group of voters deemed most likely to cast a ballot if the election were held today, 50 percent said they would pull the lever for Bush, 44 percent for Gore, 4 percent for Green Party presidential nominee Ralph Nader and 1 percent for Reform Party candidate Pat Buchanan. The polls margin of error is 3.1 percent.

Polling. A sample size of only 1,000 and the numbers are expected to have ANY accuracy? Let alone a margin of error of just 3.1 percent?

Mind you, I don't necessarily doubt that people are for Bush. He's succeeded at keeping his mouth shut lately, he comes off as dumb but affable, and I guess there at least 500 people who would like to have another vacuous Reagan type rather than the annoying psuedo-liberalism of Gore.
posted by wiremommy at 2:12 PM on September 14, 2000

That's roughly the standard size for a poll. Statistically, once you poll above a certain number of people - I forget the number, but it's not much above 1000 - the ratio of margin-of-error to # polled goes through the roof. In other words, you could poll three times as many people and only get a couple of tenths more accuracy at best. And polling is very expensive, so there's not much to be gained by going beyond 1000 or so.
posted by aaron at 2:31 PM on September 14, 2000

Sure, but IMO generally there's not much to be gained by going below 1000 or so either. Did polls accurately predict what a huge chunk of the '92 vote Perot would take? (No sarcasm, I'm honestly asking because I don't know.)
posted by wiremommy at 2:39 PM on September 14, 2000

Give Aaron that one: sample sizes of ~1000 are the norm, no matter how big the electorate. But it proves above all that:

1. The polls are all over the place
2. A "national" poll doesn't necessarily correlate to the vagueries of the electoral college.
3. Now's a good time to be a pollster

and probably

4. People lie in opinion polls. Very very often.
posted by holgate at 2:42 PM on September 14, 2000

I honestly can't possibly see how a poll of 1,000 could predict national sentiment in a country with 260 million or so. I have lived in several places, each with a completely distinct political sentiment, and there are probably dozens of variations of what I've seen across the country. Do the math -- a few oddballs here, a couple there, and your regional variations could get very screwy. And in Minnesota, where I now live, I recall a poll with around 3.5% margin of error saying the week of the election Jesse Ventura was down by 8.
posted by norm at 2:54 PM on September 14, 2000

Aaron: I don't think any of these polls are newsworthy enough for an item here. They're all over the map and the press is too obsessed with the horse-race aspect of the election. I think it's extremely cheesy to trumpet the results of one poll on MetaFilter because it plays to your own politics.
posted by rcade at 3:58 PM on September 14, 2000

Nah, rcade, you don't need to slam him. Metafilter is open to all. There's no reason one poll is by itself more accurate than any other poll. And in many ways this remains a wide-open contest; I'm concerned that the possibility exists, however remotely, that we could see the popular vote and the electoral college disagree.
posted by dhartung at 5:15 PM on September 14, 2000

So why doesn't someone gather information from all the poll sites and create some sort of uber-poll? With fifteen or twenty polls of 1000 each, we might manage a couple more tenths of a percent in accuraccy. I would be interested in seeing it.
posted by Lirp at 5:27 PM on September 14, 2000

Lirp: See Orvetti.
posted by rcade at 9:01 PM on September 14, 2000

I think I've read somewhere that many undecideds wait till the closing weeks (maybe even the week) of the election before finally making a decision? Plus, how can someone determine who's a "likely" voter?
posted by gyc at 1:44 PM on September 15, 2000

According to Orvetti, the poll Aaron is quoting here is "was based on an equation assuming 70% turnout. The actual poll results were Bush 41%, Gore 38%," which is probably within the margin of error.
posted by rcade at 1:49 PM on September 15, 2000

I honestly can't possibly see how a poll of 1,000 could predict national sentiment in a country with 260 million or so.

That's where statistics classes help out ;)

There's an interesting explanation of this wrt Gallup's polling methodology including an explanation of 1996 polling practices with plenty of historical numbers to support the way they work:

The process of polling is often mysterious, particularly to those who don't see how the views of 1,000 people can represent those of hundreds of millions.

Worth reading.
posted by holgate at 8:55 PM on September 15, 2000

Rogers, and Orvetti, are incorrect about the poll "assuming a 70% turnout." The following is from National Journal:

The Tarrance Group's Ed Goeas expressed concerns about the following "Swizzle Sticks" in the 9/14 Last Call! of National Journal:

1.) "Memo to Battleground pollsters: if you are going to do a 'turnout model' in your tracking poll, why not use a more realistic percentage. 70%? Come on! Since 1960, the highest turnout we've hit is 63%."

2.) "Memo to news orgs.: the supposed 50-44% Bush lead in the Battleground poll is a 'projection' based on a turnout model, not the poll's result. Today's Battleground RESULT is Bush 41-38%."

Goeas writes:

"While I always appreciate press attention of the Battleground survey, I am also very proud of the record Celinda Lake and I have developed over the last nine years, and feel that your comments need to be corrected. On the first point you make reference to and mock the seventy percent (70%) turnout model and point to the fact that turnout has not been over sixty-three percent (63%) since 1960. While that is true of Voting Age Population (VAP), what we are dealing with in our surveys is turnout of registered voters. In the case of registered voters, turnout has averaged around seventy percent (70%) over the last several decades. As a point In fact, in the 1992 presidential election 105,750,424 of 137,848,178 registered voters (or 76.7%) went to the polls. In the 1996 election 96,456,345 of 146,211,960 registered voters (or 65.97%) voted. As far as your comments to news organizations about the 'projection' of the turnout model, Celinda and I certainly agree the media focus should be on the overall numbers until the final weeks of the election. However, the term 'projection' in not accurate. What we use is a matrix that creates a model of 'likely' presidential voters using the components of age, education, intensity to vote, and intensity for the presidential candidates. In effect, it is another form of 'screening' for likely voters using four component parts rather that using a single screening question. More to the point, it was this 'likely' presidential model or screening that resulted in the Battleground poll being the most accurate poll of both the 1992 and 1996 presidential election's" -- Ed Goeas (e-mail, 9/14).
posted by aaron at 12:21 AM on September 16, 2000

More on how the Battleground Poll is operated, from the American Spectator:

First, Goeas and Lake asked 1,000 registered, likely voters the following question: "If the election for President were being held today, and you had to make a choice, for whom would you probably vote?" The pollsters named no names; respondents had to come up with Bush and Gore on their own. In the most recent "unaided ballot" -- through Wednesday, September 13 -- Bush is ahead with 38 percent to Gore's 33 percent. Ralph Nader has two percent, and Pat Buchanan commands just one percent (a variety of even smaller candidates account for another two percent). The remaining 24 percent told pollsters they were unsure how they would vote.

And they are the key. Many analysts seem to think that they are a large mass of voters who will eventually choose one candidate or the other, but, according to Goeas, most of them won't vote at all. "Right now there are a lot of younger voters who are saying they're definitely going to vote," Goeas says. "But a lot of them will be non-voters." To try to determine which of today's "likely voters" will actually cast a ballot, Goeas and Lake work with something called a "turnout model," a formula that gives extra weight to those undecided voters who are statistically most likely to vote. The pollsters take four factors into account: 1) how definite respondents are in saying they will vote; 2) how strongly they favor a particular candidate; 3) their age; and 4) their education level. It's all fairly complicated, but in short, the older, more educated, and surer the respondent is, the more likely he or she is to actually vote.

Given all that, the Battleground people put together a turnout model from their polling data. They came up with the following results, as of September 12. If the election were held that day, Bush would have received 50 percent of the vote and Gore would have gotten 44 percent, with Nader at four percent and Buchanan at one percent.

The numbers are strikingly different from most of those being reported in newspapers and on television. And they're particularly encouraging for Republicans because the survey was done after Bush had suffered two straight weeks of negative reporting. But the results should be viewed with great caution; the election is still a long way off, and a million things could change before November 7. Because of that, Goeas says he and Lake will not publish the turnout model results again until a couple of weeks before Election Day, when daily fluctuations will take on a critical importance. In the meantime, he simply stresses that the race is very close.
posted by aaron at 12:28 AM on September 16, 2000

« Older   |   Has MacOSRumors lost it's edge? Newer »

This thread has been archived and is closed to new comments