washington, dc

The Democratic Strategist

Political Strategy for a Permanent Democratic Majority

Lessons Learned From the State of the Art in Local Polling

By Thomas Riehle
Majority Watch made history in 2006. Before this, no one has ever set out to track district-by-contested-district the race for control of the U.S. House. It was never done before on such a systematic basis because the cost was prohibitive. Constituent Dynamics (a Seattle-based recorded-voice interviewing firm) and RT Strategies (a bipartisan polling firm in Washington, D.C.) developed a methodology using

  1. Long-time analysts (Charlie Cook, Stuart Rothenberg, Chuck Todd and Congressional Quarterly), the insights from a newer generation of analysts (MyDD.com, Real Clear Politics and others) and evidence of DCCC/NRCC investments in select races to determine which races were in play,
  2. Census data (compiled by Polidata) to model likely voters in each district,
  3. Voter lists (much improved and standardized nationwide since passage of the Help America Vote Act of 2002), and
  4. Recorded interviews (taking advantage of rapid developments in recorded calling software, as well as the years of training American consumers have endured as they grow accustomed to a recording asking them to punch 1 for one thing and punch 9 for another).

MW tracked the race in 60 districts. Final MW polls predicted a 49%-46% Democratic victory of 3 percentage points in these 60 districts. The actual vote was 50.2% – 48.5%, a Democratic victory margin of 1.7 points, just 1.3 points lower than MW’s prediction.
The MW performance was all the more remarkable in that all MW polling ended October 26, when MW conducted final polls in 41 races a week and a half before Election Day. MW final polls in the other 19 districts were conducted in early October, and in some cases in August. The difference between MW polling and actual election returns: persuadable and undecided voters who trended toward Republicans in the final 10 days of the campaign, according to national polls conducted after the final MW polling was completed.
MW polling offers a number of broad lessons for the Democratic Party, but before getting to that, consider how MW did on a state-by-state basis.
A few of the insights Majority Watch delivered:
Arizona. In AZ-08, MW generic polling identified this district as favorable for a Democratic take-over in August. MW did not need to go back to get the result right. AZ-01 was trickier: Incumbent Rick Renzi (R) was behind in an early October MW poll in the wake of the Mark Foley scandal and subsequent related disclosures about Jim Kolbe… but MW found Renzi marginally ahead in its final poll, and Renzi won.
California. MW polls showed in CA-04 Republican incumbent John Doolittle ahead by a significant margin in early October, then almost exactly predicted the final result in a late-October follow-up poll. But in CA-11 MW showed Republican Richard Pombo failing to win re-election, at 46 percent (he wound up with 47 percent and lost).
Colorado. In CO-07, three MW tracking polls starting in August showed successful Democratic challenger Ed Perlmutter move from marginally behind to significantly ahead. Final vote, a week and a half after our last survey: Perlmutter won, 55%-42%. In CO-04, MW showed Republican incumbent Marilyn Musgrave ahead in August, but a late October follow-up showed Democratic challenger Angie Paccione with a statistically insignificant lead. The actual vote favored Musgrave, 46%-43%.
Connecticut: MW polled the races of three Republican incumbents, predicting that Nancy Johnson was in the most trouble, Chris Shays was likely to win re-election, and Rob Simmons would lose a close race. Exactly right.
Illinois. In the race where the DCCC spent more money than anywhere else, IL-06, MW polls in August and twice in October showed a very close race, with Democrat Tammy Duckworth ahead by one, then Republican Peter Roskam ahead by one, then Duckworth ahead by one. Roskam won, 51%-49%. In IL-08, MW polls in August and then twice in October accurately tracked a growing lead for successful Democratic incumbent Melissa Bean. But in IL-10, two October MW polls overstated the vote of the Democratic challenger and understated the vote of the Republican incumbent. Finally, MW polls in mid-October, in the immediate wake of the Foley scandal, demonstrated that in IL-14 and IL-19 neither Speaker Dennis Hastert nor Page Board Chairman John Shimkus were in any trouble as a result of the scandal. Both Republicans were easily re-elected.
Indiana. September MW polls determined that Republican incumbents in IN-02, IN-08 and IN-09 were in serious trouble, and subsequent October MW polling confirmed that. All three lost.
Iowa. MW polls in IA-02 in early October showed incumbent Republican Jim Leach in serious trouble, with a late October follow-up confirming this race as a statistical dead heat. When Leach lost, 49%-51%, on Election Night many political commentators said that was a shocker. Not to those who followed MW! Elsewhere in Iowa, in IA-01 and IA-03, an August MW poll suggested Democratic candidate Bruce Braley was en route to a relatively easy victory, and incumbent Democrat Leonard Boswell was far ahead, and both breezed to victory.
Florida. On October 1 (two days after the resignation of Republican Congressman Mark Foley in FL-16), MW conducted two parallel surveys in his district of 1,000 voters each–one poll that informed respondents that a vote for Foley would be counted as a vote for a still-to-be-determined Republican nominee, and one that simply confronted voters with Foley’s name on the ballot and no other information. In fact, voters were informed at the polling place that a vote for Foley was vote for Republican Joe Negron, and the “informed voter” MW poll predicted that would lead to a narrow 3-point Democratic victory. Successful Democratic candidate Tim Mahoney won 50%-48%, as MW predicted would happen under those ground rules. Before the MW polls were conducted, almost all analysts had jumped to the conclusion that the Democrat would easily romp to victory because Foley’s name was still on the ballot, after his disgraced resignation from Congress. But the closeness of the race was no shocker to MW poll watchers! Elsewhere in Florida, in FL-13, the MW poll in August showed this district to be generically favorable to Republicans, but a very marginal race once the nominees were known (tested in two follow-up MW surveys). MW showed the race to be within the margin of error and as of November 20, a winner had still not been declared. In FL-22, an August MW poll showed incumbent Clay Shaw (R) significantly ahead, but a late October tracking poll put Democratic challenger Ron Klein ahead. Klein won 51%-47%.
Minnesota. In another race that seemed close to a foregone conclusion prior to MW polling, in MN-01, two October MW polls showed incumbent Republican Gil Gutknecht struggling in a close contest, with statistically insignificant leads of between one and three points. Prior to that, only true believers in the Democratic camp believed Gutknecht could lose. A week and a half after the last MW poll, Gutknecht lost 47%-53%. Meanwhile, in MN-06, a generic August MW poll demonstrated that the district favored Republican candidate Michele Bachmann. At the height of the Foley scandal, the Democratic candidate moved into a lead in a MW poll, but the Democrat’s lead disappeared in the final MW October poll. On Election Day, Bachmann won.
New Hampshire. In NH-02, a late October poll showed Democratic challenger Paul Hodes ahead with at least 50 percent of the vote. Hodes won, 53%-46%. If only MW had thought to poll in NH-01 as well! We might have predicted what came to pass on Election Day–namely, the need to move all meetings of the New England House Republican Caucus into the offices of Chris Shays for the foreseeable future.
North Carolina. In NC-11, MW polls showed Democratic challenger Heath Shuler winning decisively. The actual vote favored Shuler, 54%-46%. In NC-08, MW polling showed incumbent Republican Robin Hayes in trouble in early October, then showed a statistically insignificant Democratic lead in late October tracking. As of November 20, Hayes leads challenger Larry Kissell by fewer than 400 votes in a race headed for a recount.
Pennsylvania. Two very different, but solidly Republican districts, PA-07 and PA-10, were shown in MW polling to be easy Democratic victories over entrenched Republican incumbents. When the August MW poll showed incumbent Republican Don Sherwood behind his Democratic challenger, few believed the poll’s prediction! In fact, Democrats won both races easily. In PA-04, PA-06 and PA-08, MW polling predicted close races with mixed results. That’s what occurred on Election Night.
…and WI-08: Three MW polls showed a small but steady advantage for Democrat Steve Kagen. The final survey estimated Kagen’s support at 51 percent. Kagen received 51.19 percent of the actual vote.
For a complete rundown on the performance of MW polling, go here. For a rundown on House race polling by another active pollster using recorded-interviewer technology to accomplish large sample-size polls at a fraction of the cost of live interviewers, go here.
What was learned from this exercise?

  1. More research could better inform the conventional wisdom and lead to better targeting decisions. Any organization, whether an official party organization, a 527, or a netroots coalition of private individuals with a budget for research should be evaluating new technologies that could add greater breadth and efficiency to a sound research plan using traditional focus groups, dial sessions, mall intercepts and RDD live-interviewer polling. Recorded-voice interviewing works (and for the latest breakthrough in true Internet representative sample polling, go to http://www.polimetrix.com/news_20061106.html).
  2. The Republican “72-hour plan” doesn’t work everywhere, but where it has been tested three times or where the Republican Party chooses to focus on building it up quickly, it remains formidable. MW obviously did not get them all right, and we try to be cautious before we reach for rationalizations to explain away our clunkers. Our clunkers, however, were so geographically concentrated that we hypothesize our poll predictions may have been defeated by the “72-hour plan.” Other MW polls in New York State and Ohio were very much on target (as were MW polls in every other state), but not in western New York or in the Columbus and Cincinnati areas of Ohio (the most Republican urban areas in the state). With no way to prove it, we can only surmise that in western New York, early warnings caused NRCC Chairman Tom Reynolds to bring the best of the RNC’s turnout and grassroots efforts to bear on his district, possibly with positive effects for endangered Republicans in nearby western New York districts. As for Ohio, there is no state (with the possible exception of Florida) where the “72-hour plan” is so battle-tested as Ohio, and that may explain how Deborah Pryce, the number-4-ranking member of the House Republican caucus, came back to win a race she was in danger of losing.
  3. There is nothing wrong with the Democratic Party that a 10-percentage-point increase in Democratic vote from Independents won’t cure. In race after race MW polled in 2006, we found intense partisan loyalty on both sides, among Democratic voters and Republican voters. What changed since 2000 and especially since 2002, is the increasing willingness of Independent voters to support Democratic candidates. In national exit polls, Democrats lost Independents in five of the last six House elections from 1994-2004 (winning them only in 2004) after having won Independents in five of six House elections from 1982-1992 (losing only in 1984). This time, the margin was enormous–Democrats won Independents by 59%-41% nationwide.
  4. The monolithic Republican vote may be cracking, but so far only in the East. With 41,000 interviews conducted a week and a half before Election Day, we were able to find plenty of Republican voters who disapproved of the performance of President Bush. What we found was that the willingness of those disgruntled Republicans to vote Democratic was locally determined, not nationally uniform. In particular, in New York, North Carolina and Florida, majorities of the roughly 20% of Republicans dissatisfied with Bush were ready to vote Democratic in our final polls, while only about one-third of those similarly dissatisfied Republicans in other districts were willing to let national dissatisfactions affect their choices in local politics. Were that same crack-up in the Republican monolith developed in the Midwest, Rocky Mountain States and the West, there might be more Republican incumbents in danger in the 2008 House elections.
  5. The marriage gap may still be retarding Democratic advances in some suburban and small-town/rural districts where the marriage rate is very high. In the pre-election edition, MW polls were pointing at eight districts (Florida 16th, Illinois 6th, Illinois 8th, Minnesota 6th, New Jersey 7th New York 3rd, Washington 8th , and Wisconsin 8th) that were in play, but had marriage rates in excess of 63% which was a concern unless Democrats did better among married people than they have since 1992. Democrats converted only 2 of those Republican seats and held onto the contested Democratically-held seat in the group. In the national exit poll, evidence is that married men and married women both voted Republican in House races this time. In six House election exit polls from 1982-1992, married women supported Democrats, and Democrats even fought to a standstill for the votes of married men in half those elections. Since 1994, Republicans have benefited from a marriage gap in their favor. Married people represent one challenge where Democrats made little progress in 2006.

The purpose of the Majority Watch project of RT Strategies and Constituent Dynamics was to determine whether a wave election like 2006 could be accurately tracked through a comprehensive polling project. The capability of new, tested technology in tracking the horse races across all contested House elections is the bigger revelation.

Thomas Riehle is the co-founder of RT Strategies, a bipartisan polling firm in Washington D.C. Majority Watch is a joint project of RT Strategies and Constituent Dynamics, a non-partisan automated recorded-voice polling firm in Seattle Washington. Majority Watch is designed to track trends in the fight for control of the U.S. House of Representatives by means of polls of 1000 or more likely voters in each contested House race.

One comment on “Lessons Learned From the State of the Art in Local Polling

  1. www.policywank.com on

    This is a great effort, but I’m not too impressed by a prediction of a 3 point margin actually being a 1.7 point margin. Given the nature of your typical close race in your typical election year (which this certainly wasn’t) such a margin of error would be the difference between the results being correct or disastrously incorrect.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.