With the election season heating up (why are we the only civilized country that spends over a year promoting candidates???), you’ll be hearing more and more about what the surveys are showing. As I mentioned previously, the news media is all about marketing drama, and surveys create some of that drama.
You’ve heard plenty about how Bernie was streaking ahead of Hilary until she did well in the debates and the Bengazi hearing. Then you heard that Hilary was surging back into the lead again.
You’ve also heard how Ben Carson had moved ahead of Donald Trump despite the fact that only a third of the registered voters in the country are registered Republicans, and Ben has 24% and Donald has 22% in the latest polls. Two points is not a statistically significant difference….and what about the other 54%?
What you haven’t heard much about is how the polls are conducted or how the questions are framed.
Depending on what the latest “news” item is, the way questions are worded, and whether a valid random sample process was used, it’s likely that much of what passes as a significant survey result is meaningless.
To be a random sample survey with a measurable degree of accuracy, the sample must be pre-selected using a random number system. If someone chooses to respond to an on-line survey (like Survey Monkey), by definition this is not a statistically valid process.
So, be aware when you look at the results of any survey. Especially at this point in the election process, survey results are likely to have little or no meaning.