SP&R analysis of its pre-election polling and post-election outcome

Posted on November 12, 2012. Filed under: General Surveys, In The News |

SP&R strives to achieve the most accurate polling whether we are polling for private clients or media-sponsored groups.  In our last statewide poll on behalf of the Pittsburgh Tribune Review showing Obama and Romney tied at 47%-47% (5% undecided), SP&R was the only firm that pegged Romney’s 47% support precisely at his margin on Election Day, whereas according to realclearpolitics.com most other firms showed Romney polling in the low to mid forty percent range.  Plus, most undecided voters in our last poll were Democrats and Independents, and they ultimately broke disproportionately for the incumbent President (based on exit polling).  This was an anomaly since historically undecided voters tend to break for the challenger, which is why Obama over performed our Pitt Trib poll by 5 points on Election Day (getting 52%).  If there was any sampling error in our polling, it was in our basic assumption that turnout among 18-29 year olds would not match 18% of all votes cast like in the last presidential election. This cycle, SP&R polled 18-29 year olds at approximately ten percent of all interviews based on the assumption that 18% turnout in ’08 was a high water mark due to historic popularity for then-candidate Barack Obama.  Exit polling shows 18-29 year olds voted for Obama by a 2:1 margin in 2008 and by a similar margin this year, even despite higher unemployment among this group in comparison to other age groups.   In terms of party affiliation, in all our polls Democrats were sampled at approximately 48%-49% and Republicans at 42% (+6 or +7 point Democrat advantage), whereas initial exit polling shows Democrats with a 45-35 advantage in self party identification.  This small difference was therefore unlikely to have had much of an impact on sampling error.

Lastly, suggestions by some that our polling isn’t to be trusted because we poll mainly for Republican clients is untrue, and conveniently ignores our past record of accurate public polling in past election cycles when Democrats did well.  In 2010, SP&R was the only firm to accurately predict a narrow, 2-point victory margin for now Sen. Toomey against Democratic candidate Joe Sestak in our last poll before the election sponsored by the Pittsburgh Tribune Review showing Toomey leading 48-46.  Most other firms according to realclearpolitics.com showed Toomey with a much bigger lead.  In this same year, SP&R was perhaps the only “Republican” firm with publicly-released polling (again, on behalf of the Pittsburgh Tribune Review) showing then-Democratic candidate Mark Critz with a substantial lead over Republican Tim Burns in the May 2010 special election to fill the vacancy in the 12th Congressional District.  This poll showing Critz with a 7-point lead was widely criticized by top Republican officials at both the Burns campaign and in Washington, DC.  Critz went on to win the special election by 8 points over Tim Burns.

Good polling means all pollsters have to carefully weigh the percentages of interviews being conducted based on varying demographic groups (age, party affiliation, geography, etc.) using not only objective data points like past voting performance and actual voter registration figures, but also more subjective measurements like quantifying enthusiasm levels among various segments of the electorate to predict future voting trends.  This is becoming increasingly more difficult, due to things like non-response error, higher refusal rates and the difficulty of reaching certain segments of the population not easily reached via traditional telephone methods like cell-phone only respondents. Plus, all pollsters generally agree that misreading the way undecided voters will cast ballots means actual vote margins between the candidates in pre-election polling will tend to differ from Election Day-outcomes, as in the case of races for both President and US Senate this year.  SP&R will continue to do the best job it can to carefully weigh all these factors so we can continue to conduct the most accurate polling we can on behalf of all clients in future elections.


Liked it here?
Why not try sites on the blogroll...

%d bloggers like this: