A couple of days ago I assimilated data from a bunch of on line polls where people could informally and unscientifically express their opinion about who won the GOP debate (the big boy debate only, with ten candidates). I suggested a series of hypotheses to isolate the idea that this sort of on line unscientific effort might reflect reality, with the idea of testing the results of those polls with upcoming formal polls.
Now we have a couple of formal polls to test against. I took the raw percentages for the ten GOP big boy debate candidates, recalculated the percentages, and came up with the standings of those candidates in the more recent scientifically done polls. The polls are by Bloomberg and WMUR. The former is national, the latter pertains to New Hampshire, which will have a key early primary. Here is the relevant graphic:
We see verification of Trump being in the lead. His performance during the debate was liked by a large majority, and he is the leader of the pack, still by a large majority, by those subsequently polled. What appears to be a drop is more a factor of the difference between asking who won the debate vs. who one would vote for.
There is a big difference, though, in the back field. Bush and Walker were in the lower tier of the back field in people’s response to the debate, but are moving into a shared second place.
So, two things. First, Trump is still winning, and really is winning, the GOP race. Second, unscientific online polls seem in this case meaningful. The polls initially gave uncannily similar (not random) results, and the application of a more scientific methodology verifies them.
I quickly add this. This is not a prediction of who will win the GOP nomination, or who will win the election for President.
Nate Silver makes some excellent points about this question in this blog post. The bottom line is that polling at this stage, or even well into the primary process, does not predict either outcome very well. But I think Silver also misses an important point. These polls are not meaningless. If you view them as having only one function, predicting primary or general election outcomes, they are useless. But they do something else.
Polling at this stage in a presidential race is not about who is going to be President. Rather, such information is a good indicator of what people are thinking, how the politics are operating, how campaigns are doing, what issues are motivating people, and all that stuff. If you see polls early in the process this way, they are interesting. If you want to know who will be on the ballot in November (next November, not this November) or who will win, then … well, no.