26 October 2016

Should the MSU and Mason-Dixon polls be averaged?

That’s a judgment call. At Flathead Memo, the judgment is No. Averaging can make sense when the polls to be averaged:

1. Sampled the same population.
2. Occurred during approximately the same time frame.
3. Asked the functional equivalent of the same question.

If those three tests are met, averaging is a valid procedure, provided the average is weighted by sample size.

The MSU and Mason-Dixon meet only two of the three tests. As noted earlier today, the polls did not sample the same population. Averaging polls that do not sample the same populations amounts to averaging apples and oranges, and yields fruit salad, not additional clarity.