Polls Are Stupid Things

The current election narrative - Republicans lose their way, Democrats stumble into near-certain victory - is getting a little dull, so bloggers with some GOP sympathizers are scouting for polls that could prophesy a Republican comeback. It's a good time to point out two things everyone seems to be ignoring in the partisan heat wave.

- Sample size. A SurveyUSA phone poll in Virginia gives George Allen a 5-point lead in his Senate race. But according to the pollsters he's winning 21 percent of the black vote (although that's not what he calls it) and a whopping 74 percent of the Hispanic vote. Why so high? Because the pollsters called small, volatile samples of minority voters. Only 88 black and 19 Hispanic Virginians were asked if they'd vote for Allen or Webb, and they happened to be disproportinately Republican. Or does someone want to bet that Allen wins a 3-1 landslide of Hispanic votes in six weeks?

- State-by-state trends. Shockingly, Tim Graham alleges media bias in a story about Bush dragging down Ohio Republican candidates.

The Post says: "The president's approval ratings may be in the drain, but the first lady still regularly scores in the high sixties." In the drain?? The ratings are not spectacular, but they're rising. Look at the trend. USA Today has him up to 44 percent, L.A. Times-Bloomberg at 45. Try May or June if you want to locate "the drain," when it was 10 points or so lower.

Little-known fact: The USA consists of 50 states and the District of Columbia, all of which are holding separate elections in November. In the last Survey USA poll of Ohio roughly two weeks back, Bush's approval clocked in at 34 percent, which is basically where other state polls put him. That's one of the GOP's structural bummers this year. Since they fluffed challenges to incumbent senators in pro-Bush states like North Dakota, Nebraska, and West Virginia, they're trying to win Senate seats in places like Maryland (Bush 34 percent), New Jersey (Bush 32 percent), and Michigan (Bush 33 percent). Bush's approval has partially rebounded in deep red states, which has a nice positive effect on the national polls. But a fat lot of good that does Republican candidates in the states where he's losing a popularity contest to John Mark Karr. (Remember him?)

Comments

[Gilbert Martin]

The only people who'se opinion actually matters are those are actually going to vote.

I take all polls with a big grain of salt as they are prone to manipulation but I would pay more attention to polls of likely voters than most of the rest of them.

[kevrob]

National preference polling in the off-year elections are generally crap. If you really want to predict the House and Senate outcomes, you need reliable polls in the competitive House districts and in the states with competitive Senate races. One way to do that is to aggregate the best local polls, usually the ones commissioned by a state's best newspaper or a local think tank. The trouble there is that you are relying on others for the sample size, screen and timing of the poll, so you don't get a perfect "snapshot" of the situation for a particular day or week. The polls, especially tracking polls, paid for by the campaigns are often better, but you only get access to those when they are leaked. Bad news is kept close to the vest, unless some operative fears it will get out anyway, and wants to spin it. Good news is trumpeted far and wide, in order to encourage supporters to write checks. Whenever you see a poll in the press not commissioned by a media outlet it should be subjected to skepticism, and even when it is you should squint hard at the fine print.

Kevin
(not a professional pollster, but I did study it in college)

[Alberto Gonzales]

Or does someone want to bet that Allen wins a 3-1 landslide of Hispanic votes in six weeks?

Mr. Weigel, you're not trying to take bets on the internet...are you?

[Ron Hardin]

Polls try to make what never shows itself, public opinion, show itself, and succeeds about as well as a philosopher's analysis of women.

A vote on the other hand involves a decision, not public opinion, and has no such difficulty. A completely different animal.

In fact a vote is now immediately followed by Democrats denying that the winner has a mandate.

Nice analysis by Derrida in _The Other Heading_ ``Call It a Day for Democracy''

[Brian24]

Not for nothing, but I believe a big Republican talking point after the elections of 1992 and 1996 was that Clinton didn't have a mandate because he didn't win 50% of the vote.

The lesson: all politicians try to minimize their opponents' victories.

[[e-mail]]

For what it's worth, tradesports.com, where people can bet on the outcome of raises, has Republicans narrowing the Democrats' earlier advantage re: control of the house. Republicans are still holding strong in controlling the Senate. Of course, given how little money is at stake there, tradesports.com is not the best indicator. But I sure would like someone to come in and bet heavily on a Democratic takeover of the Senate as a shorted the Republicans three weeks ago at 75 and it's been consistently over 80 since then.

[Brian]

I dunno what to think about tradesports. It seems like a cool idea. It also seems like the parties would spend a little bit of money to get tradesports to say what they want it to say so they can spin stories about how they're turning the corner and gaining momentum.

Republicans are good at winning elections for two reasons: 1) Democrats and 2) they act like winners and everyone wants to get with the winning team. Even if they only really have that magic 35% behind them.

[buck smith]

How would one evaluate Tradesports data to determine if it had a good record for predicting elections? Maybe the ratio of the integrals of the times it was correctly predicting the winner to the times it was not?

[Robert]

Polls may be stupid, but I'm looking pretty smart: Lieberman is stretching out his lead in Conn. as I predicted.

[plunge]

Something tells me that any guy who calls Doug Wilder that word that rappers use to describe each other is not going to pull many African American voters. That Allen's press conference of the 3 African American ministers who support his stance against gay marriage included one who was just booted for sexual harrasment is not exactly a winning strategy either.

[joe]

Robert,

Still too early to tell. Most polls have the CT race within the MOE, but that one recent poll shows Lieberman with about a ten point lead. At this point, there's no way to tell if it's an outlier or a trend.

[crackpot]

Lieberman?

Didn't Reason write Lieberman's obituary several weeks ago?
Has he rediscovered his MoJoMentum?

[Nike Dunk Low]

thanks