Friday, March 18, 2016

National polls

As I think I have said before, these national polls measuring which primary candidate is leading nationally are silly. The primaries are not a national vote, they are a race for delegates collected in a series of states and U.S. territories. But despite that objection polling firms just keep releasing national poll data. So here's my next question:

Do national polls of the primary race exclude states that have already voted? Are they still including people in New Hampshire who won't get any more votes in the primary season? Or people in Illinois? Or Florida?

If the already voted states are excluded, it's not really a national poll. That would also mean that each poll that is released across the primary/caucus calendar would gradually become less and less of a truly national poll. Plus, if different populations are sampled in different polls for the ever-evolving national poll, you can't really fairly compare the results to make a claim about changing levels of support.

For example, if states are progressively excluded from the polls and Trump's percentage seems to go up, that wouldn't necessarily mean that Trump's support is really going up. It could just mean that the remaining parts of the country that haven't caucused or primaried yet just happen to skew more pro-Trump.

If, on the other hand, these national polls don't exclude the already-voted states and territories, then that means that data collected from those polls is from a sample that is increasingly irrelevant to the remaining primary race. By now, more than half of the states and territories have held their primaries or caucuses (by my count there have been 29 Democratic primaries/caucuses and 37 Republican primaries/caucuses). Which means that if the polls are truly national more than half of the respondents are not eligible voters in the remaining race. So why do their responses matter?

Either way, consider this another reason that national polls of the nomination process are stoopid.