Thursday, October 28, 2010

Polls Schmolls

When people don't like the implications of polls the response is frequently that polls can be made to say anything. Sort of like saying you've got your sources and I have mine and I guess it's too difficult to determine who is right. Is that fair?

So for instance yesterday Bob Dutko interview a guy by the name of Mickey Huff who contributed to this book. Huff and I see the media very much the same way. Corporate media naturally offers a product that is intended to maximize profits. This results in various skewing of the stories we hear. Dutko conceded that corporate interests are a factor but he said the media also has a left wing ideology. Huff said that's just inconsistent. Left means anti-corporatist/anarchist, which is fundamentally incompatible with being pro-corporatist. He said that studies show that in the mainstream media guests skew to the right according to studies done by media matters. Maybe he had in mind something like this.

Bob tries to blow that off by claiming that since media matters is a left wing source it is not to be trusted. He likewise has studies done by right wing groups such as the Media Research Center. Huff wouldn't accept those so Bob won't accept the claims coming out of media matters.

Huff's reply was interesting and something I agree with. No, he doesn't necessarily disagree with the studies done by MRC. Let's have a look and see. I find that it's a very rare study that I would just assert is flat out wrong. If proper scientific methodologies are followed then studies from right wingers can be trusted.

Bob was vague in his description of the so called studies, so I wasn't sure what he had in mind. But I went to MRC to see what it might offer, and they do have some polling data here. As I look through it I find I have no problem with it. I don't dismiss it because it's being reported by a right wing group. I think it's probably true.

Take this poll as an example. The public generally thinks the media is too liberal. I suspect they do. Doesn't mean they are right, but this probably is what the general public thinks. To determine if they are right you actually have to look at the content of the news. That's what the Media Matters study does. So both of these claims are correct. The majority of guests on the Sunday talk shows are conservative, but public perception is that the media is too liberal.

Just to pick another poll at random, this one shows that the public thinks media coverage of the war in Iraq is too negative and too liberal. I have no reason to doubt that this is what the public thinks.

Again, the implications of the poll can be wrongly extracted. The classic example involved New Coke. Taste tests indicated that the new flavor was preferred. But sales did not reflect that. Was the poll wrong? Not necessarily. One of the reasons the poll was misleading was that the sample size was too small. People did prefer the new Coke in small dosages because it was sweeter. But when you drink an entire bottle you might not prefer it.

Bob can't imagine that there is sometimes a sense in which the media does tilt to the right. So when presented with solid evidence he just wants to dismiss it, and he does so by pretending studies can be made to say anything. He probably rationalizes his positions on health care in a similar way. But while it is true that polls and studies can be misleading I think there is much to be learned from them and they shouldn't be dismissed so cavalierly. Bob should consider first that his intuitions are wrong before pretending that the result of any given study must be wrong.

4 comments:

Darf Ferrara said...

One of the difficulties with interpreting polls is that to really understand the results you need to understand the language that they are using. The argument about the you claim that left wing means anti-corporatist/anarchist, which is a valid use of the term, but I doubt that many people in the US that describe themselves on the left of the spectrum would consider themselves anarchists. I consider myself a classical liberal, but I usually have to take 25 minutes to explain to people what that means.

When I look at the further at Media Matters studies it gets hairier. I don't know who they define as conservative or liberal, although it is probably different from what I would define, and in fact it is probably impossible to define scientifically. Then there are results such as Getting Better or Getting Worse? Do you see that this question betrays an agenda? Apparently exact equality is an ideal to be achieved is implicit in the questions that they ask. You agree with their agenda, so you give the poll much weight.

Further, conservatives are at work during the week, and can't watch TV, while liberals are all on welfare and stay home and watch liberal CNBC on weekdays, getting their fill of punditry Monday through Friday.

Jon said...

I pretty much agree with you totally. The meaning of the word "liberal" has evolved, so the results must be interpreted in light of the way we would expect the general population to understand the term. I agree that often polling questions are termed in a leading manner. This is all important in evaluating the significance of some polling result.

My gripe is with Bob Dutko who is basically dismissing research because it's from a left leaning source. I think he can't do that. He has to explain the data. Part of that explanation may lie in the fact that a question is misleading or a term wasn't properly defined, but it's not like the data is wrong. Possibly useless if the questions aren't properly crafted, but he has to look at it and show it.

Darf Ferrara said...

I knew you'd come around to my way of thinking. Anyway, I think that he's right to do it, and you would probably be right to ignore MRC. You both suffer from Status quo bias and confirmation bias, so you probably will just believe what you did before you look at the data anyway, and then you would have spent all that time looking at poll results for nothing.

Not to mention that looking at raw polling data isn't very helpful. If you look at the presidential poll right after the republican convention you would have thought that McCain was going to win. After all, he was slightly ahead in the polls. If you had any knowledge of the history of presidential polling after conventions though, you would know that he had a very slim chance. Unless people have some skin in the game there isn't much reason the believe the raw numbers, it takes a skilled person to analyze them and give the right interpretation. If you know that the person quoting the number is someone with an agenda (in this case MM and MRC) it doesn't give you confidence that the numbers are being interpreted correctly.

Polling as a science is somewhere between alchemy and real chemistry. It can be suggestive and helpful, but it hasn't had its Lavoisier yet.

Jon said...

I didn't think McCain was going to win. I thought that if the election was held immediately after the convention he would have won. The poll was right. At that moment he had more support.