I thought polling, of all things, was a relatively objective measure of peoples’ opinions and leanings; “Do you like Candidate A or Candidate B?” So how is it that there are “Democratic” pollsters and “Republican” pollsters (Rasmussen Polling is a prominent Republican Poll that comes to mind)?
This seems to undermine the very notion of polls as being objective indicators of the public’s pulse. A Republican poll undoubtedly will “find” results that show a public much more sympathetic to Republican ideals (whatever those might be); and vice versa for a Democratic Poll. My question is how can polling work this way? Why aren’t all pollsters simply neutral, non-political observers? There must be a reason why things are this way but the answer eludes me. Unless the reasons are for each side to use these “polls” as political tools to further there causes.
Well, you can have “push-polling”= "if you learned that candidate xxxx was a well-known puppy kicker, would you be more or less likely to vote for him?"
And then loaded questions (In this case I can give you a real question, iirc) =“Are you in favor of reducing violent crime by requireing handgun registration?”
And why aren’t they neutral? Well, you hire whatever pollster you will think will give you the answer you want. Thus some pollsters lean the way the money comes.
Another way bias can creep in (intentionally or otherwise) is if some portions of the public grow to trust some pollster more than another. Once there’s a public perception of a pollster having a bias, people who disagree with that bias are more likely to hang up when they hear who’s calling, and people who agree are more likely to stay on the line. And so, perception becomes reality.
Polls are not always about who you are going to vote for. Let’s say the Republicans are committed to policies X, Y, and Z. They might commission a poll to find out what swing voters think about X, Y, and Z. Maybe they find out that lots of people like the sound of X, but not many care much about Y, and many do not like Z. Then they know to focus on X in their campaigning, and downplay mention of Z.
The Democrats might care more about issues P, Q, and R, and what people think about them. They will need to hire a different pollster to find out about these issues so that they can adjust their own campaigning appropriately.
When someone is called a Democratic or a Republican pollster, it does not mean that the pollster (necessarily) supports that party, it means that they have been hired by that party to find things out things that that party’s campaign strategists want to know.
In the case with Rasumussen, it’s an question of how they sample. Truly random samples are difficult, and there are always tweaks you need to do in order to get an accurate poll.
Rasumussen chooses tweaks that tend to favor Republicans. It’s not that they are fake poll that deliberately chooses to favor Republicans; they are making perfectly legitimate choices within the scope of what they’re doing.
As a hypothetical example, say in their polling 15% of the respondents are over 60. But say that nationally, 17% of the population is over 60. They would have to adjust for the difference. But let’s also say another source says 15.5% of the population is over 60. So they pick that (legitimate) source and run the numbers with that assumption. The actual number of the population could be 15.5%, or it could be 16%, or it could be some other percentage. Pollsters try to get the most accurate number, but it’s hard to be precise.
Rasmussen does have a long and consistent history of giving better poll numbers to Republicans, but their data and methods are considered well within standard industry practice.
To use one example, pollster Lou Harris was notorious for phrasing questions in such a way as to make a “liberal” answer almost certain.
During the Reagan years, I constantly saw articles based on Lou Harris polls saying “Public Wants Tax Increases to Support The Arts.” If you looked closer, the question was always something like, “If keeping arts programs for children alive would cost you only 75 cents a year in taxes, would you support it?”
When pollsters are editorializing, the respondent feels pressured to say, “Why, of COURSE I’d gladly pay a little more to support such a worthy cause.”
Every political poll comes down to sampling unless its a poll of registered voters. If they want to poll the general public, they have to decide the makeup of the population between D I and R. When it comes down to “likely voters” it gets even trickier.
Polling is not cheap. It easily costs 5 figures to run a decent poll. Just like how businesses form relationships with each other to control quality and cost, news companies are no different.
Sorry, I only answer questions with three or fewer question marks.
Sometimes sampling is definitely non-random and results are published in such a skewed way they almost tumble down. If you take a poll in Strawberry Fields County, publishing the results as “100% of Americans think the Beatles were the greatest musicians ever” is, at the very least, misleading; 100% of respondents in Strawberry Fields County said they think that - but how representative is Strawberry Fields County of people who live in Graceland County? And how many of the “doesn’t answer” didn’t want to get in trouble with the neighbors, because the poll was taken at the supermarket’s door where anybody could hear?
The example is exaggerated; sadly, there’s been actual examples in Spanish politics where choosing to perform the poll in a town or another within the region that was supposed to be polled produced completely different numbers - something which could have been guessed simply by looking at the last electoral results.
The views of people on particular issues are certainly of public interest. I fail to see a big difference between a newspaper commissioning a poll and a newspaper interviewing members of the public at a city council meeting about their thoughts on various issues. If anything, a responsible poll is a much more accurate reflection of the public’s thoughts on an issue.
It’s actually extremely easy to influence the results of polls, and surprisingly difficult to make them completely neutral. This is a piece by Nate Silver explaining how the wording of a Rasmussen poll, which wasn’t ostensibly biased, managed to produce results significantly more skewed towards the answer Republicans were looking for in the given situation. There’s a number of ways to do it:
Question wording: “Do you think firefighters should be allowed to go on strike, short-staffing the emergency service they provide?” might produce wildly different results from, “do you support the right of firefighters to go on strike to defend their salaries from spending cuts?”. Working out how to phrase these things “neutrally” is surprisingly difficult, because people can be prompted to give such different answers depending on the exact wording of what they’re asked.
Question order: If you produce a poll where the first question says, “Did you support President Obama’s handling of the killing of Osama bin Laden?” and then ask as the next question, “Do you approve or disapprove of the way Barack Obama is handling his job as President?”, that will probably skew Obama’s approval rating upwards compared to asking the approval question first. It’s well-founded that people can be “primed” to answer later questions differently, depending on the tenor of the questions they’ve already been asked.
Sampling: Samples can also significantly affect the results. Internet polls often tend to skew the results towards being more liberal, because old people (who are generally more conservative) don’t use the internet as much. Same goes for mobile phone polls. You really need landline contact as well. And obviously trying to get a sample of 1000 people that exactly matches the demographic make-up of the country as a whole is a tricky task that can easily be manipulated - all you need to is set up your poll so that there’s an extra 5-10% old people compared to the country at large, and you could skew the results conservative quite effectively, and it’ll be hard for anyone to pick up on what you’ve done.
There’s other ways to manipulate the polls as well.
Another example I remember from the healthcare reform debate in 2009/2010 was that “Do you approve or disapprove of President Obama’s proposed health care reform bill?” produced higher approval ratings than “Do you approve or disapprove of the proposed health care reform bill currently being discussed in Congress?”, because Obama’s approval ratings were much higher than Congress’s at the time (and still are), even though the question was asking about the same bill. Really subtle things can produce a significant effect on these polls.
A city council meeting is news (to some degree or another) in the first place. I’d liken it more to those bogus “town hall” meetings that politicians and news organizations often put together. The ones where the audience and the questions are pretty tightly controlled.
