As many observers have noted, support in public opinion polls for the overall health care reform bills before Congress has slipped several percentage points from October to mid-December. (Support for the public-option aspect remains high, but that will not be addressed in the present posting.)
Though the trend for support of overall reform bills is clearly down, not every poll follows this pattern. Take, for example, the FOX News and Economist-You Gov polls that were in the field right around the same time (December 8-9 and 6-8, respectively). FOX showed 34% support for reform, whereas the Economist showed 51%. That’s quite a difference!
My first inclination when seeing apparent oddities in polling results is to investigate the partisan composition in polls’ samples (i.e., what percentages of respondents are self-identified Democrats, Republicans, and Independents?). Indeed, the FOX poll contained the same percentages of Democratic and GOP respondents, whereas in the Economist poll, Democrats outnumbered Republicans by 14 percentage points. Pollster.com’s aggregate averages currently have the Democrats up approximately 9 percentage points in general-adult samples and roughly 4.5 points up in registered- and likely-voter samples. Thus, FOX and the Economist each appear to diverge from middle-of-the-road estimates, albeit in different directions.
To test my hunch that differences in party ID composition might account in good part for different polls’ health care reform support levels, I obtained a scatter plot and correlation coeffcient between the two variables. The resulting graphic is shown below. I had hoped to restrict the polls in the analysis to very recent ones, but in order to have enough data points, I had to go back almost to the beginning of November; some otherwise usable polls did not report party ID breakdown, about which I continue to be frustrated and disappointed.
In addition to FOX and the Economist, I included the most recent polls from Public Policy Polling, Rasmussen (results; party ID), Ipsos, Democracy Corps, CBS News, ABC-Washington Post (results, party ID), Pew, Public Opinion Strategies, and Associated Press-GfK. (I used a correlation-graphing website to generate the plot, then took a screen capture and annotated the display in PowerPoint.)
As can be seen, there is a clear positive, linear relationship (.61 correlation, where 1.0 is the maximum possible), indicating that the greater the Democratic edge over Republicans in sample composition, the higher the support for health care reform.
Party ID turns out to be a substantial, but by no means comprehensive, factor in accounting for the various polls’ differences in support for reform. Statistically, the .61 correlation must be squared, yielding the result that party ID accounts for 37% of the variance in support levels. More intuitively, one can see that even when holding partisan composition constant (such as in the two polls with +5 Democratic edges or three polls with +14 Democratic margins), support for reform still varies. Some of the predictive imperfection of party ID margin (as operationally defined as D minus R) might also stem from the fact that each poll’s percentage of Independents is not taken into account (e.g., a poll with 40% D, 35 R, and 25 I, would be treated identically, as D+5, to one with 30 D, 25 R, and 45 I).
I next examined each poll’s cross-tabs (where available) for reform support among Democrats, Republicans, and Independents, which are summarized in the following chart.
Support for reform among Republicans was very consistent from poll to poll, almost always at or near 10%. Support among Democrats varied widely (from 57-79%), as did support from Independents (25-47%). For whatever reason, the Economist-You Gov and ABC-Washington Post polls had not only the largest Democratic advantages in sample composition (along with Pew), but also the highest levels of support for health care reform among Democrats and Independents.
Regarding variation in reform support among self-identified Democrats, the issue of opposition to pending bills from the left — due to many liberals’ perception that their preferred reforms are being watered down in congressional negotiations — has gotten attention recently, both on the present blog and elsewhere. Perhaps different polls’ item wordings and question sequencing put Democratic respondents in different frames of mind, sometimes leading them to focus on what they would consider positive aspects of the pending bills (e.g., providing coverage to millions of uninsured) and sometimes making salient the negative aspects (e.g., the watering down).
I will leave this issue — and possible reasons for Independents’ varying levels of support in different polls — for other analysts to explore.