Sunday, May 4, 2014

45 Years Of House Ballot Polling Finds Pollsters Typically Overstate Democrats Actual Performance

Image on the left is courtesy of Alaskans For Begich TV ad, and highlights the wealthy, mega-donors Koch Brothers support of Republican candidates. Image on the right comes from an anti-Obamacare ad sponsored by Generation Opportunity.

Glancing at the current HuffPollster and RCP polling averages, Democrats and Republicans are essentially tied on the generic House ballot question. The former finds Republicans ahead by less than half-a-percentage point, while the latter finds them down by a roughly equal margin. But don't expect the same polling dynamic that exits today to be present on election day - at least not if past generic House ballot polling dating back to 1970 has anything to do with it. And further, I'd caution against relying too much on the surveys taken in the final week before the election, too, as they've been pretty hit-or-miss themselves.

Hundreds of generic ballot surveys taken over the last 45 years and eleven midterms finds polls conducted in the Spring of an election year (April, May, and June) have overstated the eventual Democratic margin in nine of them, or 82% of the time (including the last five midterms in a row). Surveys taken in the final week(s) of a campaign are no exception, understating the Republican margin in eight of eleven elections. Consider the tables below, the first of which provides polling data in the spring of each midterm since 1970 compared to the final result, the second of which compares polling data taken in the final week(s) of the campaign with the actual result.

Polling data pre-2006 was provided by the Roper Center's IPOLL databank. * Indicates the cutoff date for polls to be included in the average was TWO weeks prior to the actual election day. ** Indicates the cutoff date for polls to be included in the average was THREE weeks prior to the actual election day. Dates were only altered if necessary to obtain a large enough sample to average.

Surveys conducted in the Spring of the last eleven midterm elections tended to overstate the eventual Democratic margin over Republicans by an average of five net points. Looking at specific elections, such as 2010, you can see Republicans were polling an average 43% in the Spring of that year, though they eventually won 51% of the popular vote. Democrats, meanwhile, were polling an average of 42% that Spring, and wound up with 45% that November. The Republican advantage over Democrats grew from just one point in April, May, and June, to six points in November.

In 2006, a Democratic wave year, the Democratic polling advantage shrank from ten points that Spring to the eventual eight point final margin. Republicans actually trailed Democrats in the Spring of 2002, as the memory of 9/11 remained strong. But by that November, Republicans pulled off a five point popular vote victory. The largest shift seen in Spring polling vs. actual results came in 1994, the year of the Republican Contract with America. In the Spring of '94, Democrats actually averaged a four point polling advantage over Republicans across seven surveys. By November, Republicans had turned their four point deficit into a seven point popular vote win.

House ballot surveys conducted in the final week(s) of midterm elections, obviously (& thankfully), are a bit more on the mark than surveys conducted the prior Spring, when compared with actual results. But pollsters have still typically overstated the Democratic margin, despite the close proximity to election day. In 2006, an average of thirteen polls taken in the final week of the campaign estimated the Democrats' winning margin to be eleven points. The actual winning margin? Eight points.

In 2002, an average of seven surveys taken in the final seven days of the campaign found Republicans with a one point lead over Democrats on the generic House ballot. They actually won by five, however. The same scenario is seen in the 1998 midterm, where an average of ten surveys in the final week found Democrats winning the popular vote by one point. Republicans actually carried the popular vote that year, by one point. The 1994 midterm was the biggest generic ballot polling misfire of the last 45 years. Republicans only managed to tie Democrats, 44-44%, across five surveys taken in the final week of the campaign. But on election day, they managed a seven point, 52-45% popular vote victory. Yeah...oops.

Interestingly, one notable exception to pollster's routine overstatement of Democrats' margin on the generic House ballot in the final week(s) of the campaign came in 2010, when an average of thirteen surveys taken the final week showed Republicans winning by an even more impressive margin than they actually did (eight vs. six points).

Eleven elections across 45 years doesn't provide enough data to say with certainty that Democrats are likely over-performing their eventual margin in polls being taken today. But if you made such a prediction, you wouldn't be without some level of evidence to support it.

Other notable generic ballot polling misfires, 1970-2010
  • USAToday/Gallup really blew it in their final 2010 generic ballot poll, predicting Republicans would win 55% of the vote on election day, with Democrats winning just 40%. In reality, the Republicans margin of victory was just over 1/3 of that amount (or six points). Only slightly better than USAToday/Gallup in 2010 was Fox News and Rasmussen Reports, which found Republicans leading the generic ballot 50-37% and 51-39%, respectively. Again, not even close.
  • 2006 was a banner year for polling misfires, with the worst showing coming from CNN's final survey taken November 3-5. The poll showed Democrats leading Republicans 58-38% on the House ballot. The actual outcome of the election was 12 net points lower! The political magazine polls conducted by Newsweek and Time Magazine weren't far behind, with the former finding Democrats ahead on the generic ballot by 16-pts, and the latter finding them ahead by 15-pts. Again, Democrats' margin of victory was only eight points. Fox News wasn't so hot either, finding a 13-pt lead for the Democrats.
  • In 2002, Pew Research and Zogby share the dubious honor of being the most off from the actual result in their final generic ballot surveys. Both predicted two point popular vote wins for the Democrats. Republicans won 50-45%. 
  • In 1998, six days before election day, the Democratic Leadership Council, through pollster Penn, Schoen, found a 10-pt Democratic lead in their final pre-election survey. Republicans actually won the popular vote that year by one point. On the other side of the political spectrum, Fox News was greatly exaggerating the size of the Republican popular vote victory, finding them ahead of the Democrats by ten points.
  • In 1994, the final ABC News poll  predicted a five point Democratic lead on the House Ballot. The Republicans won by seven. 
  • A 1986 CBS/NYT poll found Republicans with a two point, 47-45% lead over Democrats on the generic ballot. Democrats actually won that year by 10 points, or 54-44%. 

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.