About That Questionable Delaware Poll Claiming That Castle Is “Vulnerable”
Supporters of Christine O’Donnell are likely to be crowing about this poll over the weekend:
Christine O’Donnell is trailing Rep. Mike Castle by only 2% among voters most likely to turnout in the Delaware Senate GOP primary, according to a Tea Party Express poll.
The poll, obtained by Hotline On Call, found that among GOP voters who indicated they are most likely to vote — or “10s” in polling lingo — Castle leads O’Donnell 43% to 41%.
The Tea Party Express released the top line of the poll on Thursday. That showed O’Donnell trailing Castle by nearly 6%, 43.7% to 38%.
These new numbers suggest that Castle is significantly more vulnerable than originally believed.
However, it is worth taking these numbers with a grain of salt. The poll was conducted by NSON Opinion Research with live interviewers. It surveyed 300 DE Republican voters and has a margin of error of +/- 5.7%.
I’ve known about this poll since last night thanks to emails from TPE that have been circulating around the Internet, but we still don’t have any links to the poll cross tabs or methodology. Moreover, a poll of 300 people in a state with an estimated voting age population of 657,000 people strikes me as incredibly small, and a 5.7% Margin of Error essentially means that the results are meaningless.
I’d love to see a poll of this race from a reliable outfit to see if O’Donnell is gaining any traction, but this ain’t it folks.
Well actually, it is a lot larger, relatively speaking (about 140 times larger), than a poll of 1000 people in a voting age population of, what is it, 280 million or so – i.e, the standard sample size of national polls.
It actually doesn’t matter how large the actual population is – the margins of error are solely a function of the sample size, not the population size. But of course there are far more factors than just “margins of error” that can contribute to a poll’s inaccuracy.
That said, if it is a well-done poll (no idea if it is), and it finds a gap of 6 points, with roughly a 6 point margin of error, then there is certainly considerable overlap in the probable values for the two candidates, But i wouldn’t call it meaningless.