Pundits and Polls
From the idiots covering politics for ABC News to the more respected pundits and bloggers like Taegan Goddard, everyone misses a key point about polls. Take Goddard’s latest post about a Zogby poll:
A Zogby tracking poll in Pennsylvania shows Sen. Hillary Clinton “had a good day” following a contentious debate on Wednesday night and now holds a 47% to 43% advantage over Sen. Barack Obama — just inside the poll’s 4.1 point margin of error.
Key finding: “There was a shift in the genders. Among men, Clinton made up seven points in the last 24 hours against Obama, who still holds a 49% to 41% edge. But Obama also made up a little ground among women, where Clinton now leads by 13 points, down from 15% in yesterday’s tracking poll.”
A 4.1 margin of error works both ways (plus or minus) for both candidates. Any first-year statistics student or research methods student could tell you that. So the pollster is confident Hillary’s support is anywhere from 42.9% to 51.1% in their poll, and that Obama’s somewhere between 38.9% and 47.1%.
So Obama could be anywhere from 12 points behind or 4 points ahead of Hillary. Yet, they treat this poll as if Hillary’s lead is “just under” being statistically secure. That’s mislead at best. It’s more likely that she’s ahead, but since there is a possibility Obama’s ahead, we can’t pretend this poll tells us any more than the trendline it might us from the previous Zogby poll.
I think any journalist working professionally on a political/pollster beat should take an actual course in statistics before getting paid to report any poll-related news.