Liberal media bias? The hell you say.Follow

When did I state or even suggest that liberal media bias is new? I'm quite sure that I've been saying this for years now.

Huh? But not 46%.



Any poll in which 46% of the respondents identify themselves as Democrats is going to tend to be skewed in favor of the leader of the Democratic Party, right? I mean, it's not like what I'm saying is even particularly controversial or anything. It's pretty simple math and a dollop of common sense.

You didn't actually think that 46% of the population identifies themselves as Democrats, did you?

Edited, May 18th 2011 8:34pm by gbaji

Apparently when AP-GfK calls people on the phone, they have been, and for years now. So, still not really seeing a problem. It's also one of many polling organizations, and I don't know much about their history of polling, but I'd imagine they're not much different than others. Maybe if you could show that AP-GfK polls are consistently ~13% in favor of Democrats or liberal concepts, you'd have a point?

Just like if, say, Rasmussen (which you site) consistently polled out of sync with every other polling organization in favor of Republican/conservative concepts?




Edited, May 18th 2011 11:28pm by bsphil

These polls are going to ruin us, clearly. Being that so many people decide who to vote for based on them...



You seem like the sort of fellow to enthuse over Ayn Rand. You follow her line of thought well enough, anyway. Were you aware that she survived largely by government benefits at points in her life?

You also seem to be under the impression that wealth is a gauge of a person's inherent value. It isn't.



I think he's pretty well aligned towards it. You'd be hard pressed to find any modern president who didn't compromise his ideology in favor of pragmatics.



There's a few things you're not accounting for. EG, people who are bothered about receiving such calls during dinner and intentionally skew the results (Not overly common, one assumes, but it happens). The people who answer these polls are going to be both politically inclined and have enough time to answer some stupid survey. It's not exactly representational of the general population.

I have to wonder why people make these sorts of idle semi-ad hominum comments when they don't have a strong argument to make for themselves.



Yes. I'm well aware of this. However, if you do that 50 times and every single time you get 10%+ more yellow marbles than green, you'd have to suspect that something is causing a skew in the results. Perhaps there are more yellow marbles in the part of the jar you are pulling from, so perhaps you should shake things up so that the results are more representative.



No. They really don't. Most polling services maintain a list of numbers that they've succeeded in getting polling results from in the past and re-use them. Worse, some of them share those lists (kinda like marketing). That's why some people get calls during dinner all the time for various political polls, market research, etc all the time and other people *never* get called.



I used to take part in surveys back when. The kind where they pay you 20 bucks to tell them which beer tasted better or something. Trust me, I know that they put your name/number on a list because I used to get calls for random surveys all the time, including political polls. After I moved and changed my number, I never got called. I haven't taken part in any sort of phone survey in 15 years now. It is *not* random. And skews in those non-random sets of people polled will tend to perpetuate over time and continue to skew the results.


Good poll analysis accounts for that. It's why they ask the distribution of conservative/moderate and democrat/republican. They can adjust the results by relating those to the actual distribution of those groups. Taking the unadjusted poll results as fact is a mistake. But lots of people do it anyway.

All reasonable points. Weighting polls is a contested subject. Whether it's the right thing to do IMO really depends on what you intend to do with the poll results.



Yes. But we need to be careful to restrict our analysis of the import of those numbers to comparing deltas to past polls using the same methodology and similar sampling. You can't just say "Obama has a 60% approval rating", since that suggests that 60% of the actual population approves, which is clearly not the case. You can only really say that his approval rating is higher today than it was when the AP-Gfk polling had him at 55%.


Unfortunately, as you well know, lots of people will assume that those polls are really (and even perfectly!) representative of the population as a whole. I could talk about how skewed polls can shift the perception of things, which in turn can actually influence peoples real positions, but that would be crazy motive speculation on my part, right? It's a bit more subtle than the basic push-poll concept, but it's there. Never underestimate the power of a bandwagon.

Edited, May 19th 2011 3:34pm by gbaji

And why it's also correct to check and see if said poll is representative of the population as a whole, right?

Gbaji, are you saying that the AP-GFK poll that's being discussed here is not random?


Here's the link.

Just for fun I looked up the Fox News polls. Guess what, even Fox says Obamas numbers are looking good, in fact, their numbers are right in line with the AP's numbers.

Here is the link.

Fox's poll gives a margin of sampling error of plus or minus 3 percentage points and the AP-Gfk poll gives a margin of error of 4.2 percentage points. Between the two polls that is up to a 7.2% possible difference that is being accounted for. Now, look at the 5% difference between what Fox says and what AP-Gfk says and explain where the bias comes in.


As far as you suggesting that some places keep lists of whom to call, I will agree, some do, but AP-Gfk does not.

Hahaha, I missed this gem.

So you assume that it's a 50/50 split between yellow and green marbles. Over and over you keep seeing that it's not the case through random sampling, so your conclusion is "this can't be representative!"

Yes, keep changing the research methodology until you get data that matches your preconceived notions. I believe that's the christian approach to scientific research - coming up with the answer first and hunting for the proof after.

Gbaji attended the Almalieque School for Statistical Research.

Why in the world are you guys talking about 50/50 splits?

Except that the same method you're castigating for not matching your assumed numbers is the way those numbers are derived in the first place. Frankly, I see the partisan divide as a distant second in importance to the ideological divide as people are more likely to shed party ID than basic ideology. The only time the partisan divide is really worth anything is when you can't/didn't measure for ideology or are trying to determine direction for upcoming elections. For "How are you feeling today?" style polling such as this, it's a footnote. Especially in cases where the numbers have been stable since you're better off determining direction with a stable line of results.

Well, no.


Admittedly, I'm not crying about a poll I don't like and insisting it must be unfair based on a slipshod understanding of polling. So I guess we're not talking about the same thing.