NETSURFER DIGEST

Lies, Damned Lies and Statistics: Analysis of our Starr Poll Results

In Netsurfer Digest 04.27, we ran an item about the Starr Report and made the rather grandiose claim that just about no one outside Washington, D.C., gave a rat's ass about this whole Clinton issue. We got a lot of flak over that, with numerous letters complaining that we were way out of line in our opinion.

Being fans of empirical data, we ran a silly little poll in NSD 04.28 which asked our readers to make the deceptively simple choice between two answers:

We collected 204 hours of raw data from our poll. The raw results give us the following:

Seemingly a rousing defeat for our editorial intuition. Or is it?

Here is the graph of the raw poll data:

Now, you can see immediately that a suspiciously large number of votes showed up during one 25-hour period.

Let's look at some more statistics.

These two graphs show the deviations in total votes from the average number of votes per hour and from the median number of votes per hour. Clearly, the 25-hour period from about hour 89 to hour 114 had a number of votes way out of line with the voting pattern during the other 179 hours. Voters deliberately hit our Web site with large numbers of votes during those 25 hours.

What's more, during that suspicious period, the number of "Do give a rat's ass" votes far outnumbered the "Don't" total:

What happened? We detected no particular pattern in the IP addresses that spawned all those "Do" votes. We think that someone posted the poll on a Web site or newsgroup and encouraged people to vote the way those behind this little manipulation wanted them to vote - telling us that they do give a rat's ass - to imply that our opinion was way wrong.

So can we really discover how the readers of Netsurfer Digest really feel about this issue? Yep. Let's throw out the 25 hours of suspicious data, data that's out of line with the background voting patterns of our readers. Were there no manipulation, we'd theoretically be punishing both factions equally.

We're left with a sample of 179 hours of voting, with no discernible manipulation one way or the other that we can detect.

The results then become:

And we breathe a sigh of satisfaction, having proven that our initial instincts were right - most NSD readers don't really give a rat's ass about this whole Clinton mess.

But wait, there's more!...


Letters to the Editor #4.29

Two crumby choices. What about the rest of us, who think "I do give a rat's ass but wish Congress would not obsess over Clinton/Lewinsky to the exclusion of regular business?"

Larry Sherman - Bronx, New York


I voted in the polls just now, but I have to tell you that your poll's questionnaire betrays the same bias as your original perverted opinion. I do want to know what goes in the highest echelons in the government and I do think Congress should act on the Starr Report. But that should not preclude either the President or Congress discharging their normal day-to-day routine. You poll question strongly suggest that if Congress decides to addresses this problem, everything else will come to halt. Is it your lack of intelligence, or perverted morals that prompt you make suggestions?

Sid Ghosh


I give a rat's ass, but have you guys ever heard of the excluded middle? In other words, how the hell am I supposed to vote if (a) I do give a rat's ass - in fact, I care a lot - about things like sexual highjinks in the Oval Office and such, lying under oath, attempting to influence witnesses, and so forth, but (b) I want this investigated by Congress as provided in the Constitution, which doesn't mean suspending everything else (such as getting a budget for next year or having elections) until it is. The choices you provided were either the Charles Rangel (drop it now) or the Dick Gebhart (do it in the next two weeks). Not a wide spectrum of opinion there!

David M. Kreps

Clever of us, wasn't it? :)

If you give it some thought - and I'll grant you this is way beyond obscure in terms of the kind of inferences we can expect from a lot of people - our poll is a multi-level commentary on the silly media polling practices.

First, everybody thinks polls accurately represent what the country really thinks. I'm not sure I buy that in this day and age of media savvy consumers. Certain polls can be easily manipulated (somebody tried to manipulate ours), lots of people lie, and a large majority won't give the time of day to the pollsters. I've never seen any statistics on the number of people who refused to answer any questions and just hung up. Our poll is not particularly accurate since it's not polling a truly random sample of the country - it's essentially meaningless other than to show what a tiny, self-selected sample of our readers think.

Second, it's well known that polls often turn strictly on the question that you ask. To some extent good pollsters can get any result they want by using the right questions. It's amazing how many polls give contradictory results around election time, entirely depending on which party is releasing the results. As you correctly point out, our poll very neatly obscures the complexity of the situation by giving you just the yes/no choice on a vague statement. Clever, no?

Finally, the whole point of our poll was to ask the question which no one in the media has bothered to ask - how many people even give a damn? All the media polls assume that every respondent has an opinion on the question (impeach/don't impeach, support/don't support), but no poll I've ever seen - and I watch way too much TV news - gives the respondents the third choice, "Leave me alone, I don't care, and don't ever call me during dinner again!" I suspect that would be an eye opener for the pollsters. :)

So there you have it. And you thought it was just a stupid, biased, little poll! Heh! :)

As always, thanks for caring enough to write. - AB


Publisher: Arthur Bebak
Editor: Lawrence Nyveen

Address your letters to editor@netsurf.com.
Letters and signatures edited for clarity and brevity.


NETSURFER DIGEST © 1998 Netsurfer Communications, Inc. All rights reserved.
NETSURFER DIGEST is a trademark of Netsurfer Communications, Inc.