Tuesday, September 18, 2007

Poll Troll

Most of you know that I am a quantoid sociologist meaning, among various other things that I have been told, that I am a colonizer, I am not a real sociologist, I demean women and people of color, and take money that would otherwise be spent on better research done by qualitative researchers. Now, while being called/told these things are true, I also know that qualitative researchers get their fair share of shit thrown at them, too. Luckily, I see this needless division between quantitative and qualitative reasoning becoming less of an issue.

But, I bring all of this up because the American Association of Public Opinion Researchers (AAPOR) launched a new website that describes the basics of polling. Since cleaning, formatting and analyzing survey data is what I do all day, it is difficult trying to describe to people what I do or why surveys should be critically reviewed (such as surveys saying that political bias of professors is a problem[1]). It has an explanation of margin of sampling error, what makes a good sample and a bad one along with all other cool kinds of information. They also launched a free online class for journalists on the basics of survey methods. Just thought that I would pass it along for anyone who might be interested.

I think that it also describes fairly well what polls and surveys are good for and what they're not good for. If you want to know the amount of support for the president or the association between certain types of social and economic characteristics on residential preferences, they're great. If you want to know detailed information about very small or specific populations, you might have more trouble and they might not be the best way to go. If you want to know the nuance of how those social and economic characteristics are intertwined and paint a story of a life lived in society, they're awful. All of these things are valuable contributions to research -- but they all require that the research be done well and that is what this site is great for. I wish the ASA could do something similar: What is sociology? What is good sociology? What is bad sociology? Of course, it would never happen, but we can all dream...

And, for those who don't care about how it is done but love looking at the numbers, pollster.com has some awesome graphs and analysis tracking the presidential primaries (nationally and by state - see image), presidential approval (spoiler alert: it's tanking), governor's races and all kinds of cool things that readers might find an interest.

* Graph via pollster.com and created by Charles at http://politicalarithmetik.blogspot.com/
[1] See cjg's excellent analysis of the problems with the survey here and here.


wobblie said...

I'm feeling a little bit oppressed by your numbers and graphs.

Had I chosen to be a quantoid sociologist, I might have made it through grad school, because then I could've just sponged off of GSS or CPS data sets like a sane person, breezed through the IRB procedures, written some SAS code, and badda-bing, badda-boom.

Stupid interpretive sociology, killer of dreams!

Mike3550 said...

Where's the fun in that? The real fun is cleaning data, making variables and then watching as your advisor uses them for papers that he's writing while you are still struggling to finish your second prelim...

Although, to be fair, the PI of the project that I work on suggested to an author that used these variables that I be included as an author on the paper, so I guess they are looking out for me.

BTW, I find it completely odd that I never have to learn anything about other non-quant methods in my training, but people using comparative or qualitative methods had to take not one, but two, semesters of statistics. That never seemed fair to me.

Post a Comment