In the span of a day, a pair of polls came out documenting the presidential preferences of the citizens of Florida (and of the 29 votes in the Electoral College they hold).
One poll shows a nearly definitive win for the presidential nominee of the Democratic Party, Hillary Clinton.
The other poll shows a statistical tie between Clinton and her opponent, the presidential candidate of the Republican Party, Donald Trump.
The first poll, from Gravis Marketing (2,542 registered voters, margin of error 2 percentage points, full results here), shows Clinton besting Trump by a margin of 46%-42% (12% undecided).
The second poll, from CBS News and YouGov (995 likely voters, margin of error 4 percentage points, full results here), shows the Democratic Party candidate beating the Republican Party candidate by a margin of 43%-42% (14% saying “Someone else/Don’t know” and 1% replying “Probably won’t vote”).
We have two polls, a day apart, asking the citizenry of Florida the same question of who would they vote for in a two-person presidential contest. Those two polls almost come to the same conclusion (a Clinton victory), but not quite.
Why the difference (albeit a small difference)?
The answer lies within the questions, “Who did you ask?” and “How many did you ask?”
I have written earlier that a voter demographic that support Trump is older voters (examples are here and here). In the Gravis survey that showed a 46-42 Clinton victory, 19% of the respondents are over the age of 65. However, in the CBS News/YouGov survey, almost a quarter (24.9%) of the poll participants were over the age of 65. While the Gravis poll, sadly, does not break out presidential preference by demographics, the CBS News/YouGov poll does. That survey shows that those over the age of 65 prefer Trump by a margin of 49%-43%. If that margin holds true for the Gravis survey (and why shouldn’t it?) and fewer voters in that age range were polled in the Gravis survey, it would mean that the Gravis survey polled fewer elder voters as to their presidential preference. Fewer older voters surveyed would mean fewer Trump voters surveyed, which could help explain why the overall Trump numbers are lower in the Gravis survey than in the CBS News/YouGov poll.
A similar thing occurs with race. One of Clinton’s core demographics is the non-white voter (examples in my blog of this statement can be found here and here). In the CBS News/YouGov survey, 12.5% of the respondents are Black. In the Gravis survey (the one where Clinton does better), 16% of the respondents are African-American. More Black voters being polled would translate into a higher number of Clinton voters. This could explain why Clinton does better in the Gravis survey.
Another demographic that does not have equal representation among the two polls are Tea Party supporters. I have not mentioned in this blog whether Tea Party supporters lean towards Trump or Clinton, but in this posting here and now, I will state that Tea Party supporters would line up for Trump. The CBS News/YouGov survey validates my assertion as 76% of those who claim to be supporters of the Tea Party would vote for Trump (as opposed to the 17% of Tea Party supporters who would vote for Clinton). In that survey, 21% of the respondents identified themselves as supporters of the Tea Party. In the Gravis poll, only 14% of respondents saw themselves as supporters of the Tea Party. Again, if Tea Party supporters lean towards Trump and if fewer Tea Party supporters (as what happened in the Gravis poll) are asked who their presidential preference is, this means that fewer Trump supporters will be counted. The opposite is also true. If more Tea Party supporters are polled (as what happened in the CBS News/YouGov poll), then more Trump voters will be counted.
Sometimes the headline numbers of Trump over Clinton 45-42 or Clinton over Trump 44-41 is not enough. You need to look at who was asked and how many of them were asked.