skip to main content
Recent Variations in Bush Approval Ratings

Recent Variations in Bush Approval Ratings

Early January can be a busy time for pollsters, especially since little polling is done in late December during the holidays. Several firms conducted polls in the first week of January, and as occurs from time to time, their results do not entirely agree. Specifically, the estimates of George W. Bush's job approval rating from these polls range from a low of 38% in recent Pew poll to a high of 46% in an ABC News/Washington Post poll. 

Firm

Dates

Approve

Disapprove

DK/REF

 

%

%

%

ABC/Washington Post

Jan 5-8

46

52

2

CNN/USA Today/Gallup

Jan 6-8

43

54

3

CBS

Jan 5-8

41

52

7

AP/Ipsos

Jan 3-5

40

59

1

Pew

Jan 4-8

38

54

8

This is not necessarily new, we've referred to a similar occurrence in an earlier blog.

In theory, all results should be similar because each firm is attempting to estimate the same population value of job approval. It is difficult to pinpoint why poll estimates might disagree because there are many possible explanations, and some are difficult to quantify.

A first place to look is sampling error. But given sample sizes of roughly 1,000 in each poll, it is relatively unusual to have polls with this large a range. 

Another reason estimates can vary is because of the timing of the polls. But as the tables show, these were all conducted about the same time; in fact, only the CNN/USA Today/Gallup poll and the AP/Ipsos poll did not overlap at least one day.

A possible explanation for the poll differences is some "house effect" -- which basically means differences in the way a firm handles the nuts and bolts of its polls. A house effect does not necessarily imply a problem with the way a poll is conducted. All of the above polls are conducted by reputable firms with well-trained survey research professionals overseeing them. 

But even well-trained professionals can sometimes have different approaches to the nuts and bolts of polling. These include (but are not limited to) the means of drawing sampling, selecting respondents once household contact is made, the placement of the job approval questionnaire in the survey, differences in interviewer training, the way "no opinion" responses are handled, and they way firms weight and round their data once they are collected. 

What is perhaps more interesting is that -- despite the range in Bush ratings -- all the polls would say that Bush's approval rating is flat, when compared with their prior poll. The early January estimates for the polls are all within a point or two of their December readings. That makes the case for "house effects" as the explanation for the differing poll results more persuasive. 

Firm

Jan Dates

% Approve

Dec Dates

% Approve

ABC/Washington Post

Jan 5-8

46

Dec 15-18

47

CNN/USA Today/Gallup

Jan 6-8

43

Dec 19-22

43*

CBS/New York Times

Jan 5-8

41

Dec 2-6

40

AP/Ipsos

Jan 3-5

40

Dec 5-7

42

Pew

Jan 4-8

38

Dec 7-11

38

 

 

 

*The Dec. 19-22 was a Gallup Poll, not conducted in conjunction with CNN/USA Today. The methods on Gallup and CNN/USA Today/Gallup polls are the same.

Usually when a poll tends to be on the high or low end of other estimates, they will tend to converge with the rest of the pack on the next reading. At least that has been our experience at times we've been on the high or low end of the spectrum. But in this case, each firms' polls have maintained their differences. 

In can be somewhat troubling to the poll consumer to see differences such as this, and to see them persist over a period. That is probably due to someone putting too much emphasis on the actual estimate for each firm, rather than looking at the results in a larger context. 

While the polls do not necessarily agree on the precise approval estimate, all show more Americans disapproving than approving of Bush. They also suggest his actual level of approval is likely in the low 40% range. Additionally, all would indicate that Bush's approval ratings have held steady over the past month. That is notable as well because the polls showed some improvement in Bush's ratings from November to December. 

To some degree, the analogy can be made to taking one's temperature using different thermometers. Most people wouldn't be surprised if they got different readings from different thermometers. But the results would likely tend to tell similar (if not identical) stories, and they would tend to converge on one conclusion. 


Gallup https://news.gallup.com/poll/20941/Recent-Variations-Bush-Approval-Ratings.aspx
Gallup World Headquarters, 901 F Street, Washington, D.C., 20001, U.S.A
+1 202.715.3030