Editore"s Note
Tilting at Windmills

Email Newsletter icon, E-mail Newsletter icon, Email List icon, E-mail List icon Sign up for Free News & Updates

January 20, 2006
By: Kevin Drum

POLLS....This is pretty much unrelated to anything, but I just wanted to point out a tidbit from that Hotline poll I linked to yesterday:

This is one of my favorite examples of why poll results should be taken with many large grains of salt: respondents lie. The actual turnout for the 2004 election among registered voters was about 70%, which means that a full 20% of the respondents to this poll flatly lied to the polltaker. This dwarfs anything related to sampling error, the only kind of error that's ever routinely cited in news reports on polls, so instead of the usual "this poll has a margin of error of 3.5%," how about the following standard disclaimer instead:

Due to well known problems of statistical error, question wording, question order, respondent dishonesty, and poor survey design, the results of this poll are probably not accurate to within more than 10 percentage points. Caveat emptor.

Just a thought.

UPDATE: Apologies. The Hotline poll was conducted only among registered voters, not all adults. I've corrected the numbers to reflect this.

Kevin Drum 1:21 PM Permalink | Trackbacks | Comments (83)

Bookmark and Share
 
Comments

Isn't it possible that the sort of person who has trouble voting also has trouble answering poll questions- i.e., the elderly, the infirm, young folks without landlines, etc? That's not to say that the poll is right of course, but there may be a better explanation than dishonesty.

Posted by: bryguy on January 20, 2006 at 1:26 PM | PERMALINK

Bryguy: Nice try....

Posted by: Kevin Drum on January 20, 2006 at 1:27 PM | PERMALINK

Were they talking to registered voters, likely voters, or the general population. Not everyone who can vote is registered and wouldn't be addressed in this survey. If it was "likely voters" there may not be anything wrong with this survey.

Posted by: gq on January 20, 2006 at 1:27 PM | PERMALINK

Isn't it possible that the sort of person who has trouble voting also has trouble answering poll questions- i.e., the elderly, the infirm, young folks without landlines, etc? That's not to say that the poll is right of course, but there may be a better explanation than dishonesty

That would point to problems in the sample selected for the poll, which would also mean that the results of the poll aren't representative of the population at large.

Posted by: Doctor Gonzo on January 20, 2006 at 1:28 PM | PERMALINK

Is it possible that this was a poll of registered voters, as these polls often are? I bet more than 60% of registered voters voted. Not that people still aren't liars, but that might reduce the amount of lying somewhat.

(The link to the poll didn't work for me.)

Posted by: BRussell on January 20, 2006 at 1:31 PM | PERMALINK

What about all the people who tried to vote but had their ballot mysteriously disappeared?

;-)

Posted by: Gore/Obama '08 on January 20, 2006 at 1:32 PM | PERMALINK

What's almost amusing about the question is how honesty-friendly it is.

I mean, it gives all kinds of morally permissible outs to allow people all the comfort in the world in admitting they didn't vote.

And they still fucking lie through their teeth to a fucking stranger.

Man. Gotta love people.

Posted by: frankly0 on January 20, 2006 at 1:32 PM | PERMALINK

As someone who has never been polled, I'm always suspicious of where they dig up these people who respond to poll questions, or even if they really exist.

Posted by: skeptic on January 20, 2006 at 1:32 PM | PERMALINK

What does "(DO NOT READ)" mean?

2% of the respondents couldn't read the question?

Also, notice that Independents and Republicans are bigger liars than Democrats. Just like real life!

Posted by: craigie on January 20, 2006 at 1:35 PM | PERMALINK

I'd use this for the disclaimer:

Polling is an art, not a science. Have fun.

Posted by: Petey on January 20, 2006 at 1:35 PM | PERMALINK

Maybe that 30% are the ones Diebold didn't count.

Posted by: tomeck on January 20, 2006 at 1:37 PM | PERMALINK

The real problem with that question is that you can't tell if the problem is lying or bad wording: the questions specifically mentions inability by sickness or lack of time -- and reinforces them by asking the respondent to keep those factors in mind -- and asks were you able to vote in the 2004 election -- encouraging the voter to answer on the basis of their ability to vote, not their choice to vote or not -- but the answers are "Yes, voted", "No, did not vote".

There's also ambiguity in the "2004 election for the President of the United States and other offices", since it most states there were least two such elections (primary and general), with different though overlapping pools of voters, plus its easy for someone to misinterpret that as any election for any office in 2004.

Before resorting to lying as an explanation, I'd like to see a poll conducted with a question well-written to avoid accidentally mistaken responses due to misunderstanding of the question.

Posted by: cmdicely on January 20, 2006 at 1:39 PM | PERMALINK

Telephone polls necessarily suffer from validity threats due to design flaws. I've actually emailed pollsters asking about this, and the ones that have responded, responded honestly.

Polls cannot properly randomize (cell phone only users, screening answering machine users, etc), they suffer from people refusing to answer (mortality), and they ask multiple questions without controlling for treatment interaction.

There is simply nothing they can do about the first two, and the third would be prohibitively expensive to control.

Bottom line is: they have no idea what their margin of error is. This is not to say that it is always large, but it is always unknown.

Posted by: conspiracy nut on January 20, 2006 at 1:39 PM | PERMALINK

Hi,
Check out the latest Political Comic from H.L.

Harry Reid is Truly Sorry

The Hollywood Liberal

If you like it there is lots more at Theres lots more at H.L. Comics Links
Thank You

Posted by: HL on January 20, 2006 at 1:41 PM | PERMALINK
What does "(DO NOT READ)" mean?

It means that the telephone interviewer is not to read that as an available option, only to record any responses in that category.

(Which is yet another built-in source of error, as not reading a "Don't Know" response encourages respondents who are in fact uncertain as to whether they voted or not in that particular election to guess, which will tend to skew significantly false "Yes".)

Posted by: cmdicely on January 20, 2006 at 1:42 PM | PERMALINK

86% of registered voters voted in the 2000 election according to the census bureau, and turnout was a bit higher in the 2004 election, so the number almost makes sense for a percentage of registered voters. The Jan 2006 hotline data doesn't list who was polled, but the Dec 2006 data was of registered voters, so it's probably reasonable to conjecture that the January data was as well.

Posted by: me2i81 on January 20, 2006 at 1:43 PM | PERMALINK

To respond to cq's question - the poll was of REGISTERED voters (that's why the poll able to is categorize answers by party affiliation but does NOT ask about party). [The poll is available via the front page of the website that Keven linked to, but a direct link seems to fail.]

So do 90% of registered voters actually vote. Here's what the Census Bureau said in 2002 (see my "www" URL above:

The voting rate of all citizens both registered and unregistered rose from 58 percent to 60 percent and their registration rate dropped slightly, from 71 percent to 70 percent, between the 1996 and 2000 elections.

If 70 percent of potential voters do register, and 60 percent of potential voters do vote, then the percentage of registered voters who do vote is about 86%.

Posted by: skeptic2 on January 20, 2006 at 1:43 PM | PERMALINK

not so clear that there is any lying going on.

perhaps the people who selected themselves to respond to this poll were the group that actually voted. The fact that 60% of the eligible voters voted does not imply that only 60% of the people who chose to respond to this poll voted. It could be (actually, it makes some sense) that people who agree to answer polls (get it? polls) are the same people who vote.

I for one, never respond to polls (although I do vote). But I suspect that if I did not vote, I would be even less likely to waste my time responding to a poll.

It's important not to get confused about that margin of error thing. It just means if you asked the same questions again of a similar sized group of people who agreed to answer the questions at the same time, then it would be epsilon close. Margin of error says nothing about the majority of people who simply refuse to answer the poll, or weren't contacted in the first place
.

Posted by: morris on January 20, 2006 at 1:46 PM | PERMALINK

cmdicely,

I think you're overinterpreting the question. I doubt that most people would find the tone and content of the question unclear. It's doing everything it can to give them an out, and I think it comes through pretty clearly.

They don't take the out. They lie.

Posted by: frankly0 on January 20, 2006 at 1:46 PM | PERMALINK

Or, OTOH, we've learned that the kind of people who are willing to answer 34th quesions in a telephone poll are the same sort of people who vote.

I think your point is right, but there are far better examples. Like the percentage of teens who say they have safe sex compared to the STD rates, the people who say they hate negative advertising and would punish a candidate who did it and not the one who is the victim, but who when actually shown a negative ad, don't, etc.

Posted by: theorajones on January 20, 2006 at 1:47 PM | PERMALINK

Thanks skeptic, the link didn't open for me and I didn't try.

Posted by: gq on January 20, 2006 at 1:47 PM | PERMALINK

Among public opinion pollsters, this issue always sparks interest. For example, a good poll has a high response rate, right? But actually, increasing response rates too much includes people who are less likely to participate in public activities, e.g., voting. In other words, a high response rate might lead to misleading results.

Years ago, public opinion researchers estimated that about 1 in 7 people misreport either voting registration or voting. Social desirability is a very strong influence on poll results (Katosh & Traugott, 1981.). I'd be willing to bet that finding could be replicated today.

I'd even go further and say that if our current president gets any lower ratings in polls, we'd find fewer people who'd say "yes, I voted for Bush/Cheney."

Posted by: Lee on January 20, 2006 at 1:48 PM | PERMALINK

My favorite polling BS was on the cover of Newsweek about a decade ago. They trumpted the finding of their sex poll, in which they found that the average man had twice as many hetrosexaul sex partners that did the average woman. And yes, they meant average, as in mean, not median.

Of course we know that's impossible - the mean had to be the same. What they had proven is that on average men lie about sex by exagerating (as SNL has riffed on memorably) while women lie about sex by minimizing. "Oh, it was just an innocent kiss..."

What was completely hysterical was that Newsweek put this statistical nonsense on the front cover - and claimed that they had found that men and women act differently.

Posted by: Samuel Knight on January 20, 2006 at 1:49 PM | PERMALINK

OK, if it really is only registered voters who were asked, and registered voters vote about 85% of the time, my faith in humanity is officially restored.

For now.

Posted by: frankly0 on January 20, 2006 at 1:52 PM | PERMALINK

reminds me of the polls of teachers in which 90% of teachers think they're in the top 50% of teachers nationwide. Of course, these are examples of different phenomena for the most part (in general, the teachers don't think they're lying), but still...

Posted by: Goldberg on January 20, 2006 at 1:53 PM | PERMALINK

I have to admit, the times I did get polled, I have always told them I was for leftist causes and against the Republican. I am sure that's what they wanted to hear anyway.

Posted by: Freedom Fighter on January 20, 2006 at 1:54 PM | PERMALINK

The mystery pollster covered this a long time ago. If asked who they voted for in the last election the incumbant always get a ~ 10% bump over their actual count in the previous election.

The mystery pollster is charitable and attributed this to people misremembering. There is apparently a tendency for people to believe they voted for the winning side or did "the right thing". Peer pressure aparently. In non-secret exit polls -- there is also a tendency for people to say they voted for the candidate that they think that particular pollster would vote for. People trying not to be confrontational??

Posted by: B on January 20, 2006 at 1:54 PM | PERMALINK

I saw all those liars at Woodstock.

Posted by: Michael7843853 G-O in 08! on January 20, 2006 at 1:59 PM | PERMALINK
I think you're overinterpreting the question. I doubt that most people would find the tone and content of the question unclear. It's doing everything it can to give them an out, and I think it comes through pretty clearly.

Written properly to do that, the call of the question would be: "With that in mind, did you vote in the 2004 election for President of the United States and other offices?" (which retains the ambiguity problems that result from the definition of the election -- a small problem -- and from not reading the "Don't Know" response -- a bigger problem, because it encourages guessing if people don't remember, and guessing about that is likely to skew false positive.)

Given that its a poll of registered voters, I suspect the problems with the question are enough to explain any real inaccuracy (rather than merely slight distortions due to sampling error, which will happen.) I don't think lying is a significant factor at all.

Posted by: cmdicely on January 20, 2006 at 1:59 PM | PERMALINK

I am sure that's what they wanted to hear anyway.

Wow, you really are paranoid, aren't you?

Posted by: craigie on January 20, 2006 at 1:59 PM | PERMALINK

Wow, you really are paranoid, aren't you?
I am, but that's because they are out to get me.

Posted by: conspiracy nut on January 20, 2006 at 2:02 PM | PERMALINK

I always lie to market researchers, who are trying, I think, to obtain information in order to manipulate me and my peers. I used to do a lot of market research back in the Seventies for Ford and other large corps. The old saw that market research is used to improve products is basically a lie. Market research is used to manipulate consumers to purchase products they do not need or want.

Lying on political surveys is also a good idea because the results are going to be used to manipulate constituents rather than make better policy. I have been receiving a lot of push poll calls lately, especially about immigration, where I was asked if I agreed with amnesty for 'illegal' immigrants. Since I think we should open the borders and let people pass through without documentation, I answered no, which probably was the conservative/fascist answer they were looking for. So, I did not lie, but did give them an answer that had nothing to do with my opinion on the subject.

Here is an old joke from my marketing research days- Marketing question: What do you use shake and drink yogurt in place of?

A: My IUD.

Posted by: Hostile on January 20, 2006 at 2:03 PM | PERMALINK

The "Don't Know" are not really people who forgot, mostly people who are not comfortable discussing that information. I knocked on thousands of doors and made more phone calls in the 2004 election and people are happy to give opinions about some things but not others.

There doesn't seem to be anything sinister or horrible about this.

Posted by: gq on January 20, 2006 at 2:03 PM | PERMALINK

Someone is talking about someone other than MEEEEEEEEEEEEE!

Posted by: conspiracy nut on January 20, 2006 at 2:03 PM | PERMALINK

I am, but that's because they are out to get me.

Yes, you're probably wiretapped as we speak. And yet you still carry their water.
Heh. Black helicopters indeed

Posted by: ckelly on January 20, 2006 at 2:09 PM | PERMALINK
My favorite polling BS was on the cover of Newsweek about a decade ago. They trumpted the finding of their sex poll, in which they found that the average man had twice as many hetrosexaul sex partners that did the average woman. And yes, they meant average, as in mean, not median.

Of course we know that's impossible - the mean had to be the same.

Well, no, it doesn't, if its a poll of people who are currently alive.

If everyone was immortal, the means would have to be the same. If women with large numbers of sex partners tend to die younger more than men with large number of sex partners (despite the general trend that, overall, women live longer), then its quite possible that an accurate poll of living people would find that men, on average, have had more heterosexual partners than women.

One way in which this would be the case is, for instance, if most women had below the mean number of partners, but a small number of, e.g., prostitutes who are also intravenous drug users, have very high numbers of partners, but also tend to die much younger, but the number of men having sex with them was a much larger number (each having sex with a smaller number of people), with less mortality effect.

Then you'd have a lot of men responding who had sex with women who were no longer in the population studied. (Of course, that still wouldn't support the ideas that poll was cited as supporting, it just illustrates how the results could be accurate.)

Posted by: cmdicely on January 20, 2006 at 2:09 PM | PERMALINK

Or it could be that a handful of women are having so much sex they don't have time to answer the phone...

Posted by: conspiracy nut on January 20, 2006 at 2:14 PM | PERMALINK

Well, Kevin's linked to a study showing that a full 25% of Americans think the sun revolves around the earth, so...

Posted by: Steve Brady on January 20, 2006 at 2:16 PM | PERMALINK

The link seems to be dead... but it's available here (see "Press Release" and "Data" at the lower right).

It seems like gq and BRussel have hit on the correct explanation. According to the linked site, the poll was of registered voters. According to the U.S. Census Bureau 89% of registered voters actually voted in 2004 -- well within the error bars of the 91% found by the Diageo/Hotline Poll.

The 63% turnout you're thinking of was of all those eligible to vote -- a very different population indeed.

Posted by: Clint on January 20, 2006 at 2:19 PM | PERMALINK

"Or it could be that a handful of women are having so much sex they don't have time to answer the phone..."

Silly, what are headsets for?

You guys are amateurs!

Posted by: Call Me Madam on January 20, 2006 at 2:20 PM | PERMALINK

Skeptic, I used to feel the same way, but last week I was polled for the first time. They askes some questions about ID vs. Evolution. It was some firm I'd never heard of, so who knows if it was legit.

At the end of it, they told me I could log in to their website and get payed to take online polls, which makes zero sense to me unless they are in the business of freeping other sites polls are generating their own fake polls.

Posted by: Boronx on January 20, 2006 at 2:25 PM | PERMALINK

cmdicely

The prostitutes also wouldn't even have to be dead to be missing from the population sample. Chances are, you would have a hard time getting them to answer polls.

Another factor of course is that the distribution of ages in the male sample and female sample would have to be the same and it would need to cover all age cohorts.

I have to admit that the assertion that the average number of partners must be the same SEEMED like it was incorrect, but I couldn't put my finger on why.

That said, I would not discount the effects of lying on a poll like this.

Posted by: BradtheDad on January 20, 2006 at 2:32 PM | PERMALINK

Well at least you guys didn't make fun of my mis-spellings. Great points though, this is what makes these threads so interesting..

You're right prostitutes could account for the difference and they're not likely to get polled. But that just reinforces Kevin's point - they obviously didn't get a statistically valid sample. Because if they did get a statistically valid sample - the mean number of hetrosexual partners for men and women would have to be pretty close together. If they aren't you have something wrong with the data - either you're missing people or people are lying.

Same with the age cohort idea. If you're just looking at the total numbers it doesn't matter what the ages are - by definition hetrosexual partners would have to be even (OK - I just realized that one would have to put in a question to clarify threesomes - and make sure that we get the gender count correct.)

And of course, if you wanted to control for either age or pay factors you would need to ask about it. But my guess, and probably that of most of you is that people would really lie about that stuff too. Which is what the Newsweek article really should have been about.

Posted by: Samuel Knight on January 20, 2006 at 3:18 PM | PERMALINK

Voting rolls include "deadwood", people who have moved or died. This complicates the analysis about what percent of eligible voters are voting.

Posted by: Carl Nyberg on January 20, 2006 at 3:18 PM | PERMALINK

Kevin has no idea what he is talking about.

Political pollsters often sample from lists of registered voters with actual vote history. This alows them to sample people who are known with certainty to have voted in the past.

Of course polls should be taken with a grain of salt, just like anything, but not quite as large and not for the reason Kevin says.

Posted by: The Fool on January 20, 2006 at 3:26 PM | PERMALINK

Kevin:

That Powerpoint link you gave has a lot of other factors thrown in, not least of which is a division between "active" and "inactive" registered voters in some states. I don't think simply dividing the number of registered voters with turnout to get 70 percent is all there is to that number. Does that include "inactive" registered voters?

Add to that the idea that those who respond to polls among registered voters are also the ones most likely to vote (tend to be more active in politics), and I don't think you have quite the error margin you think you do.

Posted by: tbrosz on January 20, 2006 at 3:31 PM | PERMALINK

Can't help but notice that Republican voters are bigger liars than independents or Democrats... though not (statistically)significantly so.

Posted by: Nell on January 20, 2006 at 3:40 PM | PERMALINK

Sorry, craigie beat me to it. Tired of Chinese spam and trolls, for once I didn't even skim comments, just jumped to the bottom and posted.

Posted by: Nell on January 20, 2006 at 3:43 PM | PERMALINK

Why is the left and the media OBSESSED with POLLS, rather then actual votes?

You have no clue how people will vote or how they will feel on a given day about an issue depending alot on how the pollster words the questions and who they can find at home and who will anser their stupid questions.

By the way, I hear since the NYT publish their article on NSA wiretaps, Al Queda and related terrorist groups have significantly reduced their electronic communications. WAY TO GO GUYS...IF WE GET HIT AGAIN, IT WILL MOST LIKELY BE BECAUSE THE NYT TOLD THE TERRORISTS TO KEEP QUIET ON UNSECURE LINES AND USE COURIERS, CODES IN E-MAIL ETC.

Posted by: Patton on January 20, 2006 at 4:10 PM | PERMALINK

Kev, I'll bet you a million dollars that a hell of a lot of those people read "were you able to vote" as "were you too sick or too busy to vote."

Posted by: Swan on January 20, 2006 at 4:11 PM | PERMALINK

There is simply nothing they can do about the first two, and the third would be prohibitively expensive to control. Bottom line is: they have no idea what their margin of error is. This is not to say that it is always large, but it is always unknown. Posted by: conspiracy nut on January 20, 2006 at 1:39 PM

Very good post CN. You've obviously have training in survey research.

Now why can't you alway post as informatively?

Posted by: Dr. Morpheus on January 20, 2006 at 4:11 PM | PERMALINK

That's to say, they may not have read the question as, "Did you vote?"

Posted by: Swan on January 20, 2006 at 4:14 PM | PERMALINK

Social Science is impossible.

Posted by: cary on January 20, 2006 at 4:35 PM | PERMALINK
The prostitutes also wouldn't even have to be dead to be missing from the population sample. Chances are, you would have a hard time getting them to answer polls.

Well, yes, but that's a "non-sampling error" problem. I was trying to explain how the results could be an perfectly accurate picture of the population (not merely the sample), though of course, your explanation explains how it could be inaccurate for reasons other than differences in how people exaggerate.

That said, I would not discount the effects of lying on a poll like this.

Oh, certainly, I think that lying in fact likely explains at least part of the results. I just take issue with the idea that it is logically impossible for the results to be an accurate picture of the population.

Posted by: cmdicely on January 20, 2006 at 4:35 PM | PERMALINK

I have always said that presidential approval numbers mean different things, depending on the political context in which they are asked. For instance, Bush's numbers were very low months before the election, then rose as the election approached, because the real question was "do you want to remove Bush from office". It is the same with Clinton's numbers. Since there was a real chance of him being removed, people knew that they were answering the question "Do you want Clinton removed from office?", which is a very different question than "Do you approve of the man."

Posted by: tool of some sort on January 20, 2006 at 4:37 PM | PERMALINK

I would go so far as to say it is impossible to get an approve/disapprove number for a president whose job is in jeopardy, because everybody knows they are part of the game. Conservatives disapprove of Bush over border issues. Doesn't mean they would ever vote for Kerry. Although, Hillary seems to be reading those numbers.

Posted by: tool of some sort on January 20, 2006 at 4:41 PM | PERMALINK

So according to Kevin's update, we still have 20% of voters lying through their teeth.

You're back on my shit list, humanity.

Posted by: frankly0 on January 20, 2006 at 4:43 PM | PERMALINK
You're right prostitutes could account for the difference and they're not likely to get polled. But that just reinforces Kevin's point - they obviously didn't get a statistically valid sample.

No, its not a matter of them being "not likely to get polled", its a matter of it being conceivable that there is a distribution of sexual behavior and correlated mortality that would make the results perfectly possible as an accurate portrayal of the population. There other explanations which would work, too -- for instance, if men with lots of partners prior to immigrating are immigrating to the US, but the women who they had sex with aren't, you'd get the same results. If lots of men engaged in sex tourism outside of the US you'd have the same results. All these factors, and others, could combine.

What has to be equal is the number of men and women having heterosexual intercourse at any given time. (Although even then, the mean number of partners for each man and women -- a number between 0 and 1 -- doesn't have to be the same, and in fact, cannot be the same unless the population of men and women is equal.)

The average lifetime number of heterosexual partners of men currently living in the US and women currently living in the US have absolutely no required relation.

Because if they did get a statistically valid sample - the mean number of hetrosexual partners for men and women would have to be pretty close together. If they aren't you have something wrong with the data - either you're missing people or people are lying.

No, my point is, this is completely wrong. Even if you restricted the time of the sex about which you questioned to a narrow enough timeslice that people having sex outside the surveyed area, people moving into or out of the surveyed area, and people dying or being born wasn't a factor -- which is not the case, at all -- then the means wouldn't be equal, but would have an inverse relation to the male:female population ratio.

But once you are discussing lifetime partners, there is no necessary relationship at all between the means.

Posted by: cmdicely on January 20, 2006 at 4:46 PM | PERMALINK

Let's take a poll, how many of us were at Woodstock?

Posted by: Rick DeMent on January 20, 2006 at 4:48 PM | PERMALINK

I believe the technical term is "response bias". Lying is a big part of that.

Response bias is the reason that in California we don't list candidates alphabetically. (I think we randomly order them for each county.) Response bias skews the results (by recollection) by 3% in favor of the first listed candidate.

Something to consider before mocking people that don't properly fill out confusing ballots - even non-confusing ballots yield response bias which can tip elections one way or another.

Posted by: Saam Barrager on January 20, 2006 at 4:51 PM | PERMALINK

I totally agree with those who say that a great number of the people who answered "yes" despite not having voted probably were confused by the "were you able to vote" formulation of the question. For instance, if they had polled me on the recent local elections (Fall 2005) in which I did not vote, and asked me the same question, I might have been confused and said yes, I was able to vote, even though I did not, in fact vote. I chose not to vote because I didn't care about any of the races, and was tired when I got home that day and didn't feel like walking down to the polling place. The question, in trying to be all "nice" and make people feel more willing to say "no," is actually confusing people and making them less likely to do so.

Posted by: John on January 20, 2006 at 4:54 PM | PERMALINK

So according to Kevin's update, we still have 20% of voters lying through their teeth.

You're back on my shit list, humanity.

So many of you took your comedy pills today, for which I thank you, especially if they fell into the "donut hole" and you had to pay full price for them.

Posted by: shortstop on January 20, 2006 at 5:01 PM | PERMALINK

I have to admit, the times I did get polled, I have always told them I was for leftist causes and against the Republican. I am sure that's what they wanted to hear anyway. Posted by: Freedom Fighter on January 20, 2006 at 1:54 PM

Believe me, your snickering gives you away.

Posted by: Dr. Morpheus on January 20, 2006 at 5:14 PM | PERMALINK

fool of some sort: I would go so far as to say it is impossible to get an approve/disapprove number for a president whose job is in jeopardy, because everybody knows they are part of the game. Conservatives disapprove of Bush over border issues. Doesn't mean they would ever vote for Kerry.

Ahhhh, rationalization rears its ugly head in the conservative blogobrain yet again.

Posted by: Advocate for God on January 20, 2006 at 5:19 PM | PERMALINK

Or it could be that a handful of women are having so much sex they don't have time to answer the phone... Posted by: conspiracy nut on January 20, 2006 at 2:14 PM

Do you have their number? Wait a minute, nevermind, they wouldn't answer the phone anyway...

Posted by: Dr. Morpheus on January 20, 2006 at 5:21 PM | PERMALINK

The respondents didn't say they actually voted -- they said "they were able to" (if they wanted to)...

Posted by: Mike on January 20, 2006 at 5:24 PM | PERMALINK

Respondents lie to pollsters! No kidding! And respondents lie to each other as well. They tell pollsters and each other what they think they should say, not what they actually do.

I always thought that pervasive lying accounted for the misconception that during some eras (like the swingin' 60s and 70s) people are getting it on with extraordinary levels extramarital sex like rabbits, but in other eras (the Eisenhower 50s and the Reagan 80s), everyone was prim and proper and saving it for marriage.

During the 60s and 70s a lot of people who weren't getting laid very much pretended that they were getting it all the time because they thought that was the cool answer that everyone else expected to hear. And during the 50s and 80s all the folks who really were screwing around at every chance they got were busy pretending they were chaste because that was the cool expected answer.

Come on, real-life sexual behavior has not changed that much through most of human history. And it certainly didn't change as drastically from one decade to the next during the past half century as a lot of people would like to believe.

One of the jobs of growing up during adolescence is figuring out that most of your friends are lying to you most of the time about their sexual exploits or lack thereof.

Posted by: nemo on January 20, 2006 at 5:27 PM | PERMALINK

Need your experience & opinion: Online political news survey! ($10 reward possible)

Hello!
My name is Daekyung (danny) Kim, a Ph.D. student studying journalism and mass communication in Southern Illinois University at Carbondale. I am emailing to ask you to do me a favor. I am now working on my dissertation whose topic is about online political news and want to gather information about how and why Internet users are using news Web sites for political information based on an online survey.

This online survey will approximately take 15-20 minutes to complete. After collecting the data, I will draw to pick up 50 respondents among those who complete this survey and each will be given $ 10 gift card.

Would you please spend some time to fill out this survey? http://www.surveymonkey.com/s.asp?u=594061481532
(please click on the address, OR if not working, copy it into the URL address)
Your experience will be very useful in understanding how politically interested online users are using online news and the consequent effect on traditional news media.

You can withdraw the survey at any time you want. All responses will be kept confidential and only be used for academic purposes. This survey has been reviewed and approved by the SIUC Human Subject Committee. So, there are no questions that may identify personal information.

Thank you very much in advance for your cooperation. Please email me back if you have further concerns or questions. Happy New Year!!

Posted by: danny on January 20, 2006 at 5:30 PM | PERMALINK

There is no reason to believe that subjects of this survey lied.

It is very easy to read that question as:

did sickness or lack of time cause you to be unable to vote in the last election?

5%-10% sickness or completely constrained schedule in early November doesn't sound outrageous. If it's a little high, then it might be subjects bleeding in who read it the way Kevin did:

did you vote in the last election?

In favor of Kevin's interpretation, it is the only "did you vote" question in the survey.

It is simply a terribly worded question.

And that doesn't even get into selection bias, where this question comes at #34 in a long series of politics questions. Those who didn't vote or don't care might have been long gone by now.

41 questions in a cold-call phone interview.

Holy cow.

Posted by: mere mortal on January 20, 2006 at 5:31 PM | PERMALINK

Kevin-

Sorry, but there's still something wrong with your analysis of the data.

According to CNN 121.4 million people voted for President in 2004 while the Washington Post says slightly more than 122 million.

According to the Census Bureau 126 million people reported voting. It's conceivable that as many as four million people in fact did vote, but didn't pick a Presidential candidate, or for one reason or another had their votes not counted. It's hard to see massive lying on the part of the public in this statistic.

However...

According to Election Data Services there were 177 million people registered to vote in 2004.

But, according to the Census Bureau only 142 million people reported being registered to vote in 2004.

Obvious conclusion:

The misreporting on the polls isn't registered voters lying about whether they voted, it's registered voters who don't vote claiming they aren't registered. I'd suggest two likely reasons for this effect: (1) "We're conducting a poll of registered voters, it will only take an hour of your time." "Sorry, I'm not registered." (the polite white lie) and (2) Many people were registered automatically at the DMV or in high school, but don't realize it. These people wouldn't even be lying.

Posted by: Clint on January 20, 2006 at 5:32 PM | PERMALINK
According to Election Data Services there were 177 million people registered to vote in 2004.

But, according to the Census Bureau only 142 million people reported being registered to vote in 2004.

Obvious conclusion:

The misreporting on the polls isn't registered voters lying about whether they voted, it's registered voters who don't vote claiming they aren't registered. I'd suggest two likely reasons for this effect: (1) "We're conducting a poll of registered voters, it will only take an hour of your time." "Sorry, I'm not registered." (the polite white lie) and (2) Many people were registered automatically at the DMV or in high school, but don't realize it. These people wouldn't even be lying.

I think you left out "(3) Election Data Services counted people registered more than once, either in different places or under different namesm as distinct registered voters, while surveys, either the Census Dept. one or any other, only count actual people, and only once each."

Posted by: cmdicely on January 20, 2006 at 5:38 PM | PERMALINK

Ahem. I can settle this entire debate. kevin is wrong.

Kevin is wrong because political pollsters often use official registered voter lists which include vote history. They don't have to rely only on what people say to them. They can review their actual vote history.

The vote history, of course, doesn't say how registered voters voted but it says whether they voted in elections going back for decades. Using that information, along with a screen at the beginning that asks you whether you plan on voting in the next election, pollsters can make a very accurate inference about a given respondent's likelihood to vote in the next election. Now you may object that people will get through the screen by lying, but, since the bias is to say you vote more than you really did, it is a good bet that those who say they won't vote in the next election really won't. Those who never or rarely vote won't be called in the first place.

Posted by: The Fool on January 20, 2006 at 6:12 PM | PERMALINK
Kevin is wrong because political pollsters often use official registered voter lists which include vote history. They don't have to rely only on what people say to them. They can review their actual vote history.

This is how polls of likely voters are set up, sure. Polls of registered voters don't take into account vote history in selecting the sample, since that would create an invalid sample of that population.

This is a poll of registered, not likely, voters.

Posted by: cmdicely on January 20, 2006 at 6:18 PM | PERMALINK

Nice try Dicely, but no cigar.

Kevin made a blanket statement about polls in general. What's more, his point could ONLY apply to likely voter polls, as polls of registered voters don't have to rely on asking you whether you will vote or not, they just have to check the list to see if you're registered.

Its slightly more complicated than that because you also have to account for new registrants but Kevin's point about people lying about their intention to vote doesn't hold at all.

Posted by: The Fool on January 20, 2006 at 6:30 PM | PERMALINK
Kevin made a blanket statement about polls in general.

Yes, his blanket statement was that poll results should be taken with a grain of salt because respondents lie. The particular voting behavior question on the specific poll was cited as an illustration of this lying.

What's more, his point could ONLY apply to likely voter polls, as polls of registered voters don't have to rely on asking you whether you will vote or not, they just have to check the list to see if you're registered.

Er, no. Polls of registered voters -- like the one Kevin cites -- do, indeed, often ask people if they voted as one of the reported questions; polls exclusively of likely voters are less likely to ask this as a question, since its a screening factor and they use, as you note in your previous post, voting records to screen for that up front, and have no reason to ask.

Its slightly more complicated than that because you also have to account for new registrants but Kevin's point about people lying about their intention to vote doesn't hold at all.

Um, Kevin's point is not about people lying about their intention to vote. Its about people lying in general, and uses a supposed instance (highly debatable) of people lying about their actual past voting practices to illustrate this.

Intentions have nothing to do with it.


Posted by: cmdicely on January 20, 2006 at 6:36 PM | PERMALINK

Nice try Dicely, but no cigar.

You missed an opportunity to say, "Nice try, but no Dicely."

Not that I'd have agreed with The Fool's post had I read it.

Posted by: shortstop on January 20, 2006 at 7:00 PM | PERMALINK

The first thing I noticed was that the totals for yes, no , and don't know/refused come out to 101%. Now even with a margin of error you should never come up with a percentage that exceeds ALL respondants should you?

Posted by: Eric Paulsen on January 20, 2006 at 7:48 PM | PERMALINK

The first thing I noticed was that the totals for yes, no , and don't know/refused come out to 101%. Now even with a margin of error you should never come up with a percentage that exceeds ALL respondants should you?

Totals, when each category is rounded to whole numbers, that round to either 99% or 101% are fairly common.

Consider if , to tenths of a percent, the figures were 90.7%, 7.6%, and 1.7% -- adding to exactly 100.0%. They'd round exactly as shown.

Posted by: cmdicely on January 20, 2006 at 7:59 PM | PERMALINK

In grad school my main area of research was voting behavior in the US. Many studies over many years have established that those who register to vote turnout at a rate of 80 - 90%, which is right in line with this Hotline poll. So when the question comes up "why don't more people vote?" it's not quite the right question. The real question is, "why don't more people register to vote?," because once they register, they turnout at a very high rate. Voting is relatively easy, but dealing with the paperwork and deadlines involved with registering is harder (although this varies by state).

Posted by: Mike T on January 20, 2006 at 9:24 PM | PERMALINK

The "Do Not Read" 2%:Must be RepubliCon Presidential Candidates.

Posted by: FitterDon on January 21, 2006 at 12:36 AM | PERMALINK

Makes sense cmdicely, I guess I just figured one would always round any number over 100% back down to 100%. For appearances sake.

Posted by: Eric Paulsen on January 21, 2006 at 9:34 PM | PERMALINK

歌手,写真,图片
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map
map

Posted by: dfd2dfd on January 22, 2006 at 5:36 AM | PERMALINK

http://www.20six.co.uk/nwts-atk-hairy-black-girls/ - nwts atk hairy black girls http://www.20six.co.uk/nwts-atk-hairy-black-girls/
http://www.20six.co.uk/plCX-women-hairy/ - plCX women hairy http://www.20six.co.uk/plCX-women-hairy/
http://www.20six.co.uk/HugC-hairy-muscular-guys/ - HugC hairy muscular guys http://www.20six.co.uk/HugC-hairy-muscular-guys/
http://www.20six.co.uk/wJtk-pictures-of-hairy-armpits/ - wJtk pictures of hairy armpits http://www.20six.co.uk/wJtk-pictures-of-hairy-armpits/
http://www.20six.co.uk/BAiz-hairy-puss/ - BAiz hairy puss http://www.20six.co.uk/BAiz-hairy-puss/
http://www.20six.co.uk/Mnfh-super-hairy/ - Mnfh super hairy http://www.20six.co.uk/Mnfh-super-hairy/
http://www.20six.co.uk/LOog-hairy-asians/ - LOog hairy asians http://www.20six.co.uk/LOog-hairy-asians/
http://www.20six.co.uk/anPy-hairy-pussy-gallery/ - anPy hairy pussy gallery http://www.20six.co.uk/anPy-hairy-pussy-gallery/
http://www.20six.co.uk/bKMl-hairy-bitches/ - bKMl hairy bitches http://www.20six.co.uk/bKMl-hairy-bitches/
http://www.20six.co.uk/iQqR-hairy-bushy-armpits/ - iQqR hairy bushy armpits http://www.20six.co.uk/iQqR-hairy-bushy-armpits/
http://www.20six.co.uk/LHsb-gay-hairy/ - LHsb gay hairy http://www.20six.co.uk/LHsb-gay-hairy/
http://www.20six.co.uk/KghP-very-hairy-bears/ - KghP very hairy bears http://www.20six.co.uk/KghP-very-hairy-bears/
http://www.20six.co.uk/kwzy-hairy-and-homo-and-men/ - kwzy hairy and homo and men http://www.20six.co.uk/kwzy-hairy-and-homo-and-men/
http://www.20six.co.uk/jTyf-hairy-girl/ - jTyf hairy girl http://www.20six.co.uk/jTyf-hairy-girl/
http://www.20six.co.uk/RrmQ-hairy-young/ - RrmQ hairy young http://www.20six.co.uk/RrmQ-hairy-young/
http://www.20six.co.uk/qPsT-atk-natural-hairy/ - qPsT atk natural hairy http://www.20six.co.uk/qPsT-atk-natural-hairy/
http://www.20six.co.uk/okdA-fat-hairy/ - okdA fat hairy http://www.20six.co.uk/okdA-fat-hairy/
http://www.20six.co.uk/gOKe-old-hairy/ - gOKe old hairy http://www.20six.co.uk/gOKe-old-hairy/
http://www.20six.co.uk/akQF-hairy-thumbnail-galleries/ - akQF hairy thumbnail galleries http://www.20six.co.uk/akQF-hairy-thumbnail-galleries/
http://www.20six.co.uk/KNWR-older-hairy/ - KNWR older hairy http://www.20six.co.uk/KNWR-older-hairy/
http://www.20six.co.uk/bLMe-hairy-female-crotches/ - bLMe hairy female crotches http://www.20six.co.uk/bLMe-hairy-female-crotches/
http://www.20six.co.uk/jRQO-hairy-italian/ - jRQO hairy italian http://www.20six.co.uk/jRQO-hairy-italian/
http://www.20six.co.uk/RnPp-big-naked-hairy-bears/ - RnPp big naked hairy bears http://www.20six.co.uk/RnPp-big-naked-hairy-bears/
http://www.20six.co.uk/GyMk-hairy-tgp/ - GyMk hairy tgp http://www.20six.co.uk/GyMk-hairy-tgp/
http://www.20six.co.uk/QlBa-hairy-chest-females/ - QlBa hairy chest females http://www.20six.co.uk/QlBa-hairy-chest-females/
http://www.20six.co.uk/hBmN-free-pics-of-hairy-men-nude/ - hBmN free pics of hairy men nude http://www.20six.co.uk/hBmN-free-pics-of-hairy-men-nude/
http://www.20six.co.uk/mAkC-hairy-pussey/ - mAkC hairy pussey http://www.20six.co.uk/mAkC-hairy-pussey/
http://www.20six.co.uk/koxE-hairy-young-girls/ - koxE hairy young girls http://www.20six.co.uk/koxE-hairy-young-girls/
http://www.20six.co.uk/PwTk-hairy-male-armpits/ - PwTk hairy male armpits http://www.20six.co.uk/PwTk-hairy-male-armpits/
http://www.20six.co.uk/TUNg-hairy-pussies/ - TUNg hairy pussies http://www.20six.co.uk/TUNg-hairy-pussies/
http://www.20six.co.uk/xeLJ-men-with-hairy-backs/ - xeLJ men with hairy backs http://www.20six.co.uk/xeLJ-men-with-hairy-backs/
WBR LKvqlryQBmCsYUISz

Posted by: VlhzbCsWmI on January 23, 2006 at 2:39 AM | PERMALINK




 

 

Read Jonathan Rowe remembrance and articles
Email Newsletter icon, E-mail Newsletter icon, Email List icon, E-mail List icon Sign up for Free News & Updates

Advertise in WM



buy from Amazon and
support the Monthly