On Monday, I posted a piece where I looked at polls from 5 different states holding primaries on Super Tuesday (I didn’t do caucus states cause there wasn’t polling data out there for most of them).
So, the time has come to take a look at how accurate the polls were, state by state.
On Monday, the Real Clear Politics (RCP) average had “Gingrich leading with 39.3%, Romney second with 23.8%, Santorum third at 20.5%, and Paul trailing with 6.0%.”
The results: Gingrich 47%, Romney 26%, Santorum 20%, and Paul 7%, so we got the order right, and given that the polls only accounted for a little under 90% total, meaning there were about 10% undecided or naming another candidate, I’d say the percentages were in the ballpark as well, especially Santorum and Paul.
Monday, the RCP average had “Santorum leading with 41.0%, Romney second with 20.5%, Gingrich third with 20.0%, and Paul trailing the field with 7.5%.”
When all the ballots were counted, we had Santorum with 34%, Romney 28%, Gingrich 27%, and Paul with 10%. Again, the order of finish was correct, and the final percentages were within shooting distance of the polls, though Santorum’s drop of 7 points wasn’t hinted at by the data I saw from RCP.
The state that likes to keep pundits up late had an RCP average showing “Santorum in the lead with 34.0%, Romney a close second with 31.3%, Gingrich a distant third with 15.7%, and Paul in last with 12.0%” on Monday.
The polls, honestly, got this one wrong, but there were things they got right even while not correctly predicting the winner. The final results were Romney 38%, Santorum 37%, Gingrich 15%, and Paul 9%. So, while the polls were correct that this was a very close race, they didn’t pick up on who was going to end up on top.
Like Georgia, this one was absurdly easy to call. RCP had “Romney way out in front with 64%, Santorum a very distant second with 16%, Paul and Gingrich bringing up the rear with 7 and 6% respectively.” The results were Romney 72%, Santorum 12%, Paul 10%, and Gingrich 5%.
The last of the 5 states I reported on Monday, on Monday RCP predicted “Romney ahead with 34%, Santorum second with 27%, Paul in third with 14%, and Gingrich in last with 10%.”
The actual results were Romney 40%, Paul 25%, Santorum 24%, and Gingrich 8%. So while the polls got the winner and loser right, they got second and third places wrong.
What can we make of this? Seems to this humble blogger that we can trust poll averages in blowout races to some extent, but the closer the race gets, or the smaller the number of voters or poll respondents (as in Vermont), the harder it becomes to accurately predict, so I’m going to maintain my skepticism of polls for a little while longer. I guess we haven’t perfected the science of predicting mass human behavior, called “psychohistory” by Isaac Asimov, quite yet.