Pollster.com

Articles and Analysis

 

Bush Approval: Four polls, Trend at 36.2%

Topics: George Bush

BushApproval20050620061022small.png

Four more polls in today revise the trend estimate of approval of President Bush to 36.2%. CNN/ORC taken 10/20-22/06 has approval at 39%, disapproval at 58%. ABC/Washington Post, taken 10/19-22/06 has approval at 37%, disapproval at 60%. Newsweek, 10/19-20/06 found approval at 35%, disapproval at 57%. The new Cook/RT Strategies poll, taken 10/19-22/06, got approval at 37%, disapproval at 53%.

The trend estimate is revised up from 35.8% as of polling through 10/15. The trend line (the blue line) remains down because the new data revises the rate of decline, rather than demonstrating (so far) a reversal of the decline. It typically takes 6-12 polls for a change of direction to be clearly revealed by the trend estimate. So far, the data do not hint that a reversal of direction has occurred. For example, the "sensitive" estimator shows a hint of flattening, but no upturn. Since the sensitive estimator is easily fooled, it is too soon to say approval has stabilized. More data this week should help clarify this, however.

The figure below shows the four new polls. All results are reasonably close to the trend estimate. In response to a comment on the earlier item, I reran the analysis of Newsweek's previous poll, which I showed was an outlier, beyond the 90% confidence interval. Repeating that analysis with the new polls included does not change that result. For the time it was conducted, the previous Newsweek poll is still unusually low. Their new poll, in contrast, is well within the 90% CI, as are the vast majority of their polls.

FourPanelApproval20061022small.png

Note: This entry is cross-posted at Political Arithmetik.

 

Comments

I still do not understand why these polls are lower than the daily RasmussenReport tracking polls which consistently show Bush's aproval ratings at 41% to 43%.

____________________

Ben Ross:

Rasmussen gives an explanation on their web site. They ask the question differently than some other pollsters, with four options (strongly or somewhat approve or disapprove). They have tested four-option and two-option questions using the same polling methodology on the same night and found that the four-option question gives about 3% or 4% higher approval.

Apparently, some people who answer the four-option question with "approve somewhat" answer disapprove when there are only two options. Here it is:
http://www.rasmussenreports.com/ElectionPollsPollingMethodology.htm

On another topic, has anyone else noticed that the new WaPo poll sample contained 30% Democrats, while their poll sample two weeks ago contained 38%? And the new sample had 6% more independents. While there are certainly good arguments for and against party-affiliation weighting, it seems obvious in this case that any differences between the two polls (such as the slight uptick in Bush approval) have more to do with random changes in sample composition than with real trends in public opinion.

The Post's coverage I thought was responsible, since they focused on the opinions of the independents they polled rather on trends.

____________________

Charlie:

Both Real Clear Politics and the Washington Post poll site have their Pres Approval at 40%, Dissapproval 58% for the Oct 19-22 poll. How is it you are reporting 37% - 60%?

____________________

Charlie,

Real Clear uses the registered voter numbers, while I use all adults. The vast majority of approval polling is based on samples of adults rather than registered voters so I prefer to keep the sample constant, rather than confound change in sample population with change in opinion.

The Post gives both in their data:

http://www.washingtonpost.com/wp-srv/politics/polls/postpoll_102306.htm

I DO include pollsters who consistently use RV (Fox, is an example) and even those who quit reporting Adults and switch to RV or LV only close to elections-- but if the Adult number is available that is what I use.

Charles

____________________

wayne:

The GOP has been heavily pushing the idea that polling is biased in favor of Democrats. Hugh Hewitt (or maybe it was Barnett), for example, says he views a tied poll as a GOP lead and a narrow Democratic lead as a tie. The example typically used to support this is the famous 2004 exit polls that supposedly favored Kerry. They also mention Zogby, who allegedly has not been accurate. However, I looked at a historical sample from the 2004 Senate races at SurveyUSA's website. In 15 Senate races, SurveyUSA actually averaged a 2% error in the margin of victory in favor of the GOP. Strategic Vision (in only 5 races listed) was more tilted to the GOP. This is a pretty small sample to generalize from but would not appear to support the conservative meme. Has anyone done a statistical analysis of past polls and if so, how do they compare and who has been the most accurate historically?

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR