2016 Postmortem
In reply to the discussion: Page 7 New CNN poll ..NOBODY POLLED UNDER 50yrs Old [View all]Gormy Cuss
(30,884 posts)It's hard for me to describe this without jargon so I'll use a very simplistic and not totally accurate way to describe it. The poll has a relatively small number of respondents but enough to make reasonable national estimates of opinion by applying weights to adjust for small discrepancies in the number of respondents by category. However, once the data is segmented into subcategories (like a cross tabulation of support for each candidate by age group of the respondent), the actual number of people polled may be too low to report as representative of that group. Best practice is to state as ORC did that the sampling error was too large to report results. It does not mean that they did not interview people under age 50. If the poll had more responses, say 1000, there would be a better chance that cross tabs could be weighted to be representative.