6

Suppose you are using a survey to measure voter preferences towards binary policy issues (e.g. whether to increase taxes, whether to legalise gay marriage, whether to start a war...)

The simplest approach, which I suspect is the dominant approach in political science, is simply to ask voters what they want. For example: "Do you want to legalise gay marriage?" This approach makes a lot of sense to me.

In economics, however, there has been a long standing view that one needs to 'incentivise' accurate answers. This is not easy to do in the context of voter preferences. However, there are recent papers that try to do something along these lines. For example, to measure voter attitudes towards gay marriage, you can check if they want to donate some money that you have just given them to an organisation that lobbies for (or against) gay marriage.

Q: Are there methodological papers on the benefits/costs of such incentivisation? (Note: I am asking specifically about the case of voter preferences; methodological papers on incentivisation in general abound.) One question that is especially relevant, although perhaps hard to study, is whether incentivised measures predict actual voting behaviour better or worse than simple unincentivised measures.

Improve this question
11
  • 1
    What exactly is different between asking if someone supports a cause and if they are willing to donate to that cause? If you are talking about needing to actually donate to the cause itself that brings up many different issues from not being able to afford it to not trusting random people calling and asking for donations with all the different donation related scams these days.
    – Joe W
    Commented Dec 12, 2023 at 14:32
  • 2
    @JoeW people who support a cause are not always ready to do anything for promoting it - so one may question whether their support is genuine or just a knee-jerk reaction, or consequence of loyalty/opposition to a certain political party. E.g., many in this community support Ukrainians fighting to the bitter end, whatever are the human costs... but very few of these have actually volunteered to fight in the Ukrainian army.
    – Wag the mainstream media dog
    Commented Dec 12, 2023 at 15:12
  • 1
    @Joe W The existing wording was not optimal; hopefully, it's now clearer.
    – afreelunch
    Commented Dec 12, 2023 at 15:38
  • 1
    @JoeW I'm talking about the latter possibility ("actually asking them to donate to something"). Usually, this is done by first giving respondents some cash, and then asking them if they want to donate some of that (as stated in the question). Though you are right that this could also pick up how much a respondent needs the money, so might be a noisy proxy for the underlying variable of interest.
    – afreelunch
    Commented Dec 12, 2023 at 15:49
  • 1
    Good Q. Voters often say one thing and do another. For example, Washington state is fairly dependably Dem, but when push comes to shove has repeatedly nuked carbon taxes. The same problem is also faced by green policies in the EU. And, yes, there is often a direct linkage to people's wallets, so better knowledge would assist making sustainable (in the duration sense rather than green sense) policies. But, as you you note, would the results be representative?
    – Italian Philosopher
    Commented Dec 12, 2023 at 17:01

1 Answer 1

Reset to default
5

The purpose of the typical political "opinion poll" is different from a customer survey for market research. In an opinion poll, what you want to find out is not "how do people really feel" but rather "how will people vote". This means that you want your survey to be as representative of the voters as possible, and not representative of the general population. Similarly, you want to find out what opinions people would express in an election, and not what their (perhaps hidden, but true) beliefs are.

The financial deal you describe is very different from the situation that faces a voter when they cast a ballot, and so probably wouldn't give predictive results. Moreover the costs of doing such a survey would mean that only a very small sample would be possible, and there are significant issues of ensuring representativeness in a small sample.

That's not to say that such a survey can't be done, nor that it has never been done. But rather I say that as an "opinion poll" such a survey would have limited applicablity.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .