Skip to content
 

Horse-and-buggy era officially ends for survey research

Peter Enns writes:

Given the various comments on your blog about evolving survey methods (e.g., Of buggy whips and moral hazards; or, Sympathy for the Aapor), I thought you might be interested that the Roper Center has updated its acquisitions policy and is now accepting non-probability samples and other methods. This is an exciting move for the Roper Center.

Jeez. I wonder what the President of American Association of Buggy-Whip Manufacturers thinks about that!

In all seriousness, let’s never forget that our inferences are only as good as our data. Whether your survey responses come by telephone, or internet, or any other method, you want to put in the effort to get good data from a representative sample, and then to adjust as necessary. There’s no easy solution, it just needs the usual eternal vigilance.

P.S. I’m posting this one now, rather than with the usual six-month delay, because you can now go to the Roper Center and get these polls. I didn’t want you to have to wait!

2 Comments

  1. Stacy says:

    “you want to put in the effort to get good data from a representative sample, and then to adjust as necessary”

    .

    sounds like the oh so simple recipe for Unicorn Stew — all you need are carrots/potatoes/salt and some good fresh unicorn fillets.

    How do you consistently get representative samples without true random sampling?
    Sampling is easy, but getting samples with a high-confidence as representative …ain’t.

    And ‘adjustments as necessary’ are usually difficult to distinguish from subjective fudge-factors.

    • Andrew says:

      Stacy:

      As the saying goes, there is no alternative. Random or representative samples are not available, so we have to adjust. If you don’t want to adjust, then fine, but then you’re not in the game at all.

Leave a Reply