Sigh. Despite pointing out the blatant statistical flaws in her overinterpretation of voter coalition preferences to Audrey Young yesterday, the same junk polling data is still being peddled by the Herald, and even leads today's editorial. To put things in perspective, I've added in the actual numbers of voters involved this time:
New Zealand First's voters [15 in total], for instance, would prefer National over Labour by 90.9 per cent [10] to 9.1 per cent [1; the other 4 or 5 didn't express an opinion]. All [three] voters for United Future would prefer the party to back National. Yet NZ First and United Future have kept Labour in power this term.Then there's John Armstrong:
Notably, the latest Herald-DigiPoll survey has 57 per cent of respondents who backed the Maori Party [4 people] still preferring to deal with Labour after the election, as against 43 per cent [3 people; 4 Maori Party voters didn't express a preference] who want the party to talk with National first.Now, journalists aren't statisticians, but the poll results will have the actual numbers of respondents in each category along with the percentage (at least if you read past the executive summary), and I'd have thought that even journalists would be aware that taking the opinions of 7 or 11 people as a reliable sample is charitably described as dubious. And OTOH, the polling company presumably does know these things; shouldn't it be warning the journos about what is and isn't reliable data? Or am I going wrong by assuming people are even marginally competent?
