July 30, 2006 - August 5, 2006
By now it seems that the entire political blogosphere has taken note of the latest Quinnipiac poll of Connecticut primary voters showing Net Lamont leading Joe Lieberman by 13 points (54% to 41%). Much of the commentary I have seen has been appropriately cautious. As noted here previously, polling in this sort of primary is difficult and results may vary widely depending on the way the pollster samples and identifies "likely voters." If it's not too late, I want to point out a few things about the Quinnipiac poll that -- caveats aside -- bode well for Ned Lamont.
If nothing else, Quinnipiac has been consistent in their methodology, while caution is in order regarding the current levels of candidate support, the tracking polls provide an excellent gauge of the trends in that support. And the trend of late is clearly all in Lamont: He has gone from trailing Lieberman by 15 points in June (40% to 55%), to 4 point lead two weeks ago (51% to 47%), to the 13 point (54% to 41%) lead on the current survey.
The aside from the overall numbers, two things stand out for me that have generally not been given much notice elsewhere:
First, Lamont's favorable rating has increased even more in the last two weeks than his share of the vote. It jumped ten points, from 36% to 46%, while his unfavorable rating remains quite low at 14% (up just one point from 13%):
This is a remarkable trend given the conventional wisdom that "Lamont has built his lead almost entirely on dissatisfaction with Lieberman," as the Washington Post's Dan Balz put it this morning. It is certainly true that on another question, 44% of Lamont's supporters say that "Lieberman's support for the war in Iraq" is their "main reason" for their choice. But Lamont has clearly made a positive impression in his own right. Consider also the negative advertising run by the Lieberman campaign, arguing among other things, that Lamont has put forward little more than a "No Joe" slogan. Despite the negative ads and Lieberman's aggressive debate performance, Lamont's unfavorable rating remains at just 13%, while his favorable rating (46%) is now nine points higher than Lieberman's (37%).
The second notable finding is the surprising collapse of Lieberman's support among moderate to conservative Democrats. Among liberals, Lamont is currently trouncing Lieberman by a more than two-to-one margin (66% to 31%). But as my friend Charles Franklin points out, Lieberman's support has also dropped sharply among moderate to conservative Democrats, falling from 61% in June to just 49% on this latest survey. Lieberman's lead among moderate and conservative Democrats has fallen from 35 points in June to a statistically insignificant 4 points (49% to 45%).
Consider that finding in light of what Chris Barnes, associate director of the Center for Survey Research and Analysis at the University of Connecticut, had to say about primary polls in his state. According to Hartford Courant columnist, Barnes believes that Connecticut primary polls "exaggerate the number of moderate Democrats who will vote and understates the liberal turnout."
Perhaps. While the Quinnipiac poll release does not tell us the precise liberal percentage of their sample, their cross-tabulations imply that it is somewhere in the mid 40s, probably about 44%. The exit polls** conducted for recent Democratic presidential primaries in Connecticut indicate a more liberal electorate: 56% in 2004 and 46% in 2000. Turnout for those primaries was not high -- 130K (or 19% of registered Democrats) in 2004 and 177K (or 26%) in 2000. Keep in mind, however, that these results may differ depending on the way pollsters ask the ideology question.
However, given the state of the race as measured by Quinnipiac, the size of the liberal electorate may not affect the outcome. Lamont's lead will be much wider, of course, if the Quinnipiac poll is under-sampling liberals. However, Quinnipiac now has the race so close among moderate and conservative Democrats that (by my own calculations) Lamont still leads by eight points even if we assume that liberals make up only 30% of the primary electorate. In fact, given the Quinnipiac results, the race would still be a 47-47% tie even if we assume liberals make up only 10% of the electorate!
Again, caution is always in order when it comes to polls for this sort of primary. The turnout puzzle should make us all skeptical of any poll on this race. Other polls that have been conducted this week by the campaigns or other pollsters may yield different results. But consider the news Dan Balz managed to extract from one member of the Lieberman campaign:
Another campaign adviser, who spoke on the condition of anonymity in order to discuss strategy, said the public poll tracked internal campaign surveys. "The race has been headed in that direction for a while," the adviser said. "It's a fairly accurate reflection of where the race is."
If so, things are looking very promising for Ned Lamont.
**My source for the past exit poll results are the codebooks available for download through the Inter-University Consortium for Political and Social Research (ICPSR).
UPDATE - Tim Tagaris on the official Ned Lamont blog has this to say about their campaign's internal polling: "I can tell you one thing, our numbers don’t show us up 13 – and that is no spin."
UPDATE II - In a comment below, DemfromCt passes on word of another poll of likely Democratic primary voters in Connecticut, this one conducted for Theday.com by Research2000 that has Lamont ahead of Lieberman by 10 points (53% to 43%).
UPDATE III (8/5) - More details this morning in The Day on the /Research 2000 poll. They conducted 600 interviews of likely Democratic primary voters using random digit dial (RDD) sampling. The article includes no further information on how likely voters were selected or how many adults were interviewed in order to select the 600 likely voters.
I am hiring to fill an immediate opening for a full-time, entry-level analyst/researcher/assistant to work with me in Washington exclusively on MP. Applicants need to have good writing and proofreading skills, strong attention to detail and, ideally, a junkie's obsession with political polling. Previous experience in survey research, statistical analysis or web site development/administration is a strong plus but not a requirement.
If you are interested and would like more details on this unique opportunity, please email me and attach a resume.
And yes, regular readers, this move implies bigger and better things coming soon for MP. It's still a little early for a formal announcement, but stay tuned...
UPDATE (8/17): The job opening described above has been filled. A sincere thank you to those who expressed an interest.
The Pew Research Center released results a few days ago from its biennial study of American's media consumption habits (report, printable PDF, topline questionnaire). Although it may have been lost amidst a rush of major news stories and other polls released in the last week, two things make the Pew survey report a must read for political polling junkies. First and most important, the ongoing study (the latest in a series conducted regularly since 1996) is the most comprehensive data available on the ways Americans get their news. Second, the survey is the first major public poll I am aware of to incorporate interviews conducted by mobile phone among Americans from mobile-phone only households [for some questions].**
The focus of the latest report - not surprisingly - is the way the Internet continues to change the way Americans get their news:
A decade ago, just one-in-fifty Americans got the news with some regularity from what was then a brand new source - the internet. Today, nearly one-in-three regularly get news online. But the growth of the online news audience has slowed considerably since 2000, particularly among the very young, who are now somewhat less likely to go online for news than are people in their 40s. For the most part, online news has evolved as a supplemental source that is used along with traditional news media outlets. It is valued most for headlines and convenience, not detailed, in-depth reporting.
The study goes into remarkable detail on how, where and when Americans get their news and what they like and dislike about the news sources they use. It asks about a wide variety of different news outlets and modes, including how many Americans get news by cell phone, PDA or podscasting. A follow-up survey probed the perceived credibility of the national television networks and various national news brands, and the report provides tabulations of those ratings by party identification.
If you have a question about how new modes of communication are affecting the way Americans get their news, the odds are that this survey asks about it, one way or another. Here is an interesting example: The number of Americans that reports having a "digital video recorder like TiVo that automatically records TV programs you select," has sharply increased from 3% in 2002, to 13% in 2004 to 23% on the current survey. Nearly one in four Americans how has the ability to record programs and - take note media consultants - fast forward through paid advertising
The noteworthy twist regarding the methodology of the survey involves the way the Pew researchers opted to check on the roughly 7%-9% of the public that cannot be reached by a landline telephone (individuals that are systematically excluded from traditional telephone surveys):
To evaluate the news usage of the cell-only population, Pew conducted a shorter version of the media consumption survey with a sample of 250 people who have a cell phone but no landline telephone. Respondents were interviewed on their cell phones, using a sample drawn from a nationally representative cell telephone number database. The interviews were conducted May 15 - June 3, 2006.
News consumption for some sources especially newspapers was lower among cell-only respondents than among those with a landline phone. However, when the cell-only respondents are included with the respondents reached on a landline, and this blended sample is weighted to match the full U.S. population demographically and with respect to telephone status, overall estimates of news consumption are affected by an average of less than one percentage point.
The report provides a detailed comparison of data from the main (landline) sample and the "blended sample" that comes from both landline and cell-phone-only households. It is all of interest to political poll junkies and worth reading in full.
**An earlier Pew survey interviewed respondents by cell phone in order to the study the cell-phone only household phenomenon. This is the first major media study I am aware of on more general topics to incorporate a cell-phone-only sample.
Also, to clarify: Most of the tabulations in the Pew report, including the results I cited above and all of the results in the filled-in questionnaire are based on the conventional sample of landline telephone households. The only tabulations based on the blended sample of cell-phone and wired households appear at the end of Section 6 of the report under the heading "Cell-Only Household's News Usage."
Of the many national polls released last week, the most intriguing was the survey released by NPR of likely voters in the 50 competitive congressional districts (story, report). Two campaign pollsters -- Republican Glenn Bolger and Democrat Stan Greenberg -- conducted the survey. Their creative design is helpful in moving away from a purely "generic" congressional ballot question asked of all voters nationally to a survey of voters competitive districts that asks about the actual candidates whose names will be on the ballot.
Using polling data to follow the battle for control of the House is a tough task. While a number of different polling organizations will track the competitive races for the U.S. Senate, public polls on individual House Races are relatively few and far between. National polls tend to focus on the so-called "generic vote" that asks about voting for the "the Republican candidate or the Democratic candidate" rather than candidate names. For now, these surveys typically report results among registered voters in all 435 House districts nationally rather than the most competitive seats that will decide the outcome.
That's what makes the design of the NPR survey so helpful. According to the summary on the Greenberg Quinlan Rosner web site, they selected 50 House districts as most competitive based on their rankings by the best known congressional handicappers (Charlie Cook, Stu Rothernberg, Larry Sabato and the Hotline). Using random digit dial sampling, they interviewed 1,000 adults who, according to the questionnaire, said they were registered voters, cast ballots in 2004 (if eligible) and reported they are "almost certain" to vote or "will probably vote" this fall. [Clarification: NPR did one sample of 1,000 likely voters spread across 50 districts].
Glenn Bolger summed up the importance of that design:
This is where the effort's going to be made [. . .] This is where the money's going to be spent, and this is where the messages are going to be sharpest ...This is where the House hangs in the balance.
The NPR questionnaire also asked about actual candidates rather than the party label featured in the "generic" ballot question. They asked the generic ballot question (Q12), but then followed it with a vote preference question (Q15) featuring the actual names of the candidates in most instances (using the "most likely" challenger in districts with upcoming primaries). [A Greenberg spokesperson tells me that footnote #2 on the version of the questionnaire currently posted to their web site is incorrect and reflects the procedure used on a survey conducted earlier in the year].
The results: "While republicans do a little bit better with these voters than they do in a nationwide sample," according to the NPR story, "the numbers still point to trouble for the party in power." More specifically:
- The Democratic candidate leads by six points (49% to 43%) on the question that includes candidate names
- The Democratic candidate leads by a bigger margin (+10, 52% to 42%) in the 23 seats classified as most competitive than in the other 27 lower tier seats (+3, 47% to 44%)
- Consistent with various national surveys, the Democrats interviewed expressed significantly more interest in the election (66% very interested) and enthusiasm about voting (62% more enthusiastic than usual) than Republicans ( 56% and 49% respectively)
I saw these results and wondered: What was the actual Congressional vote in these districts in 2004 and 2002? The report posted on the Greenberg Quinlan Rosner website listed the specific districts. Through their spokesperson, I also obtained a list of the districts they considered top and bottom tier in terms of competitiveness (see below) and spent an hour or so entering the vote return data from America Votes into a spreadsheet.
The results are fascinating. If the election were held today, the results of the NPR survey indicate a much different overall outcome than in 2004 or 2002. Republicans won 55% of the vote cast in these districts in 2004 and 58% of the votes cast their in 2002, but the Republican candidates are currently preferred by only 43% of the voters in the NPR survey. The gap is similar across various sub-groupings of districts reported in the survey:
Now these numbers require a few caveats. First, keep in mind that the turnout in these districts in 2004 was considerably greater (roughly 14 million) than in 2002 (roughly 9 million). This year's turnout will likely be much closer to 2002 than 2004. Second, the 2002 totals do not include the two competitive Texas districts (17 & 22) because redistricting changed their boundaries after that election.
Third and probably most important, the gap between current polling and past vote returns probably varies widely across individual districts. Some of the districts now considered "in play" were not at all competitive in 2002 or 2004 and saw lopsided wins by the incumbent in those years. Given that 40 of these 50 seats are now held by Republicans, some of the gap between current polling and past support for Republicans is about the unique dynamics of individual races (an open seat, a seat held by a freshman member, a previously unopposed incumbent now facing a real competitor, etc.).
Regardless, this survey design has great potential to give us a much better view of the developing race for Congress should NPR conduct tracking polls featuring the same design in coming months. Let's hope they do.
More details on the NPR sample on the jump
Continue reading "The NPR Survey and the Race for Control of Congress"