Friday, April 06, 2007

Angus Reid Really Loves Me

My latest Angus Reid persona got an email yesterday morning:

Dear SuperSecretPersona, Angus Reid wants to know WHAT DO YOU THINK?

Please take a quick break to answer a few questions on topics ranging from Canada's justice system and the RCMP to household pets.

So my boat-owning, B.C. Conservative answered a number of questions re. his political beliefs, the pet-food scandal (my fictitious cat became seriously ill), and a harmless but rather unsightly condition known as "skin tags".

Now, one of my anonymice claimed earlier that my obsession with pranking Angus-Reid makes me like the guy who orders a pizza using a false name, or who short-changes the TTC by eight cents every time he rides the bus and thinks that makes him a rebel.

Well, for one thing, I don't know how he found out that I do all that stuff. For another, it isn't what I'm up to here.

I am arguing that, using their current methodology, Angus-Reid is nothing more than a glorified quicky-poll like the type you find at (for example) the CTV website.

AR allows you to create multiple profiles from a single IP address, so while I currently have two of them active, that could be twenty, or 200, as many as the number of hotmail accounts I could create. Furthermore, AR has no means of determining whether or not the demographic information I have given them is correct (in fact its mostly false). And this is quite different from a phone poll. Given a phone-book, or better a reverse directory, a polling company can match the phone-number of a potential contact to a street, and therefore to a postal code or Stats-Canada census tract. Thus they have some basic information about the contact's likely age, ethnicity, social status, and so forth. And of course, its difficult to lie about your gender over the phone!

Now, I'm not saying on-line polls can't be designed to work as well as a phone-poll. But, in my experience of such things, its necessary to make real-world contact with your potential participants first (by, for example mailing something to an address and asking them to mail it back) in order to pin down their location in physical space, at which point you can make assumptions about their background, social status, and etc. Of course, that involves spending money on postage and paper. Angus Reid seems to be going the El Cheapo route, and the credibility of their polling suffers as a result.


Anonymous said...

"AR allows you to create multiple profiles from a single IP address, so while I currently have two of them active, that could be twenty, or 200, as many as the number of hotmail accounts I could create."

So when they see multiple profiles coming from one IP address they exclude those responses from the sample. You may be the one who is getting pranked.

bigcitylib said...

Not impossible, but why allow them to be created in the first place? And, if you saw these different profiles coming in from the same IP, why would you not kill one of them, or both of them? Yet I am now getting email to both.

I have also written earlier that my first profile is demographically incoherent, yet they keep asking me to fillout surveys (four so far).

Anonymous said...

Oh who knows, this is new to them so maybe they are building there own profile of respondents to determine what a likely prankster would look like. That's how spam filters are designed. I think it's more likely you're playing there game rather than the other way around.

bigcitylib said...

But its not a new technology. All they have done was done, god, over ten years ago, including possible methods of trying to rig an online poll. There are certain ways of building up an online database of polling subjects. They usually, as I say, involve phone or mail contact with the potential survey subject, which costs money, and then you start saving with online followup.

None of this should be new to AR. I am still convinced they're just being sloppy.

(I actually spent a couple years researching some of these issues)

UWHabs said...

They may not be perfect, but I would argue that as long as the vast majority of people are honest and truthful (seeing as most people really don't care about pranking them), the results should be fairly valid.

I mean the best thing would be if there was a way they could ascertain your true identity, but that would usually involve giving your credit card number for verification, and I bet you 99% of people wouldn't be willing to submit to that.

In general, I'd trust an online poll result maybe not as much as a phone result, but for most of the time, sure. Their methodology isn't perfect, but the main point of online surveys is that they're faster for you, you can fill them out when you want. Adding any sort of mailing or phone verification slows it down (plus it means that I can't fill out the survey as if I were still at home. I may live in California for these few months, but my survey results are good for a voter in Waterloo).

Ti-Guy said...

I think these poll's are solely for the manipulation of perception and not for anything truely useful or insightful.

I think it's naive to think of them in any other way.

Anonymous said...

Did you see the results of their Quebec election poll? Angus Reid was the most accurate polling firm in the country. In fact, the only firm to peg the ADQ as the Official Opposition. They must be doing something right despite the efforts of certain idle hands.

bigcitylib said...

Anon 8:58,

As far as I can see the Quebec result was more a case of being less wrong than the other agencies rather than actually right. Nobody did very well with that election.

Anonymous said...

Well, hate to say it, but you're totally wrong about that. They had a total differential of 3.5 percentage points - that's 3.5% off the actual results for all three main provincial parties combined. Their biggest miss was in lowballing the Liberals by 2.2% - but the margin of error for the poll was +/- 3.5%. So in other words, they were right about the results for all three parties. None of the traditional polling firms (including Quebec experts Leger Marketing) can claim that. See for yourself...