One has to do a bit of clicking and scanning to discover the actual poll’s methodology, but here it is:
Public Policy Polling surveyed 678 Arizona voters on December 19th and 20th with a margin of error of +/-3.8%. Half of the interviews were conducted by phone and half were conducted by text.
So they phoned and texted (far) more than 678 numbers and asked them if they were registered Arizona voters and would they take a survey. It should always be noted when the number of rejected efforts is not included. Yet i’d assert that considerably more rejected than accepted. The resulting sample of 678 who answered their phone to an unknown number and took a poll are thereby declared to be a proper random sample of all Arizona voters. How does one know they’re declaring that? Because they use a “margin of error” calculation which requires that be the case: 1/sqrt(678) → 0.0384 for their stated 3.8%. Just something tedious to keep in mind when weighing whether a poll is actually reflecting reality or not. Particularly given points like @ClutchLinkey has wisely raised.
