Polling companies and media organizations sometimes argue that including third-party candidates in a poll produces an inaccurately high result for that candidate.
In the era of modern polling, there have been seven third-party presidential candidates included through the final polls in early November. Only one time did the difference between the final poll and the actual election result exceed 2%, and that was when Perot over-performed the final poll by five points in 1992.
Not once has a third-party candidate been included in polls, and then under-performed the final poll by more than the margin of error.
2000 Nader -Final Poll: 4% Result: 3%, Error +1%
1996 Perot -Final Poll: 7% Result: 8%, Error -1%
1992 Perot -Final Poll: 14% Result: 19%, Error -5%
1980 Anderson -Final Poll: 8% Result: 7%, Error +1%
1968 G. Wallace -Final Poll: 15% Result: 14%, Error -1%
1948 H. Wallace -Final Poll: 4% Result: 2%, Error +2%
1948 Thurmond -Final Poll: 2% Result: 2%, Error 0%
en.wikipedia.org/wiki/Historical_polling_for_U.S._Presidential_elections
Looking at that link, there does appear to be a pattern of collapsed performance. John Anderson’s peak polling was three times his final result, while Perot and Nader’s runs saw peaks of twice the actual result (although we can point to Perot’s failings). We don’t know if Johnson and Stein have peaked yet. But either way it appears we’ll have a record breaking result for Libertarians and a strong showing for the Greens. Best guess is Johnson around 7% and Stein below Nader around 2%.
“I wonder if others have seen the 10% plus that Johnson’s experiencing and seen it vanish.”
I’m not sure any have seen 10% to begin with. There was Ross Perot, but I think he held onto it in the two elections. Between then and now, I don’t think any third party candidates got to 10% to begin with.
There’s also a difference between early polling and the final poll before an election. It would be interesting to see that comparison for third party candidates. I wonder if others have seen the 10% plus that Johnson’s experiencing and seen it vanish.
That’s a good point, Shawn. I’m not sure if that stuff makes any difference, or if anyone even knows, one way or the other.
Of course this is all before the current era of polling, where people actively resist unsolicited phone calls, have unlisted numbers, and get oversaturated with political phone calls, both robotic and human.