Content-Type: text/html
Media-Sponsored Opinion Polls
Media-Sponsored Opinion Polls
Media-Sponsored Opinion Polls
The Use and Abuse of Media-Sponsored Opinion Polls
in Two Presidential Campaigns:
A Critical Analysis of Network TV News and Six Prestige Print Media
by
Dennis T. Lowry and Josephine Nio
School of Journalism
Southern Illinois University
Carbondale, IL 62901
[log in to unmask]
Dennis T. Lowry is professor of journalism and Josephine Nio is a doctoral
student in journalism at Southern Illinois University. The authors are grateful
to graduate students Cynthia Bond Hopson and Young Soo Shim for assistance with
library research, and to Charles Kingsley for compiling video tapes.
ABSTRACT
The Use and Abuse of Media-Sponsored Opinion Polls
in Two Presidential Campaigns:
A Critical Analysis of Network TV News and Six Prestige Print Media
This study is the first longitudinal critical analysis of media-sponsored
opinion polls by three different categories of prestige news media---network TV,
newspapers and news magazines. Poll stories from Campaign '92 and Campaign '96
were analyzed for conformity to AAPOR disclosure standards, and also for eight
additional types of reporting/interpretation errors. Both the number of poll
stories and the quality of poll stories was down in Campaign '96. Newspapers
were the clear winner in conforming to AAPOR disclosure standards, and TV
networks were the clear loser.
The Use and Abuse of Media-Sponsored Opinion Polls
in Two Presidential Campaigns:
A Critical Analysis of Network TV News and Six Prestige Print Media
Polls have become a staple of American political news coverage---and more so in
presidential races than at any other time. Even one political pollster himself
has acknowledged, "The news media are obsessed with polls; almost every major
newspaper and television station is conducting its own. Unfortunately, the
increase in quantity has not produced better quality."[1]
To what extent have the major news media used opinion polls in recent
presidential campaigns? To what extent have they abused those opinion
polls---i.e., used them in fragmentary ways or used them incorrectly? And, if
the media abuse of polls is common, to what extent is this important to the
democratic process? This study is the first longitudinal critical analysis of
media-sponsored opinion polls by three different categories of prestige news
media---network TV, newspapers and news magazines. Cantril stated that ". . .
polls sponsored by news organizations often get caught up in the swirl of events
and too seldom are subjected to the scrutiny they may deserve."[2] This study
is an effort by independent scholars to subject these polls to the scrutiny they
deserve.
The most infamous media-sponsored opinion poll in a presidential campaign was
the 1936 reader presidential preference survey conducted by Literary Digest.[3]
This polling debacle, which predicted that Alfred Landon would win over
President Franklin Roosevelt, has been used as a case history of what not to do
by two generations of students in journalism, political science, and sociology.
In most presidential campaigns since then, the major polls have been reasonably
accurate, and in some cases highly accurate, in predicting electoral
outcomes.[4] However, the accuracy of final polling results and election
outcomes are only one small part, and not even the most important part, of the
controversy surrounding media-sponsored opinion polls---i.e., polls conducted on
behalf of and paid for by the media themselves.
The AAPOR Standards
In any type of scholarly critical analysis the first difficult task is to
determine which evaluative criteria to use. Against which standards should the
news media be judged? And who says so? In the context of the present study,
however, this problem is considerably reduced. In 1968 the American Association
for Public Opinion Research (AAPOR) published its "Standards for Reporting
Public Opinion Polls." The basic standards were somewhat modified in 1986 and
were listed in AAPOR's Code of Professional Ethics and Practice, Section III,
titled Standard for Minimal Disclosure:
At a minimum, the following items should be disclosed:
1. Who sponsored the survey, and who conducted it.
2. The exact wording of the questions asked . . .
3. A definition of the population under study . . .
4. A description of the sample selection procedure . . .
5. Size of samples . . .
6. A discussion of the precision of the findings, including, if
appropriate,
estimates of sampling error, and a description of any weighting or
estimating
procedures used.
7. Which results are based on parts of the sample, rather than on the
total
sample.
8. Method, location, and dates of data collection.[5]
Even though, strictly speaking, the AAPOR standards apply only to the full
technical research reports produced by AAPOR members, the standards have been so
influential that even The Associated Press Stylebook and Libel Manual states
that essentially the same points "should be in every story based on a poll."[6]
Likewise, the standards have been so influential that recent reporting textbooks
have recommended that today's journalism students---the journalists of
tomorrow---include almost the identical standards in their news stories.[7]
Review of Past Research
Scholars have criticized how the prestige press has been reporting public
opinion polls for more than three decades.[8] One of the first content analyses
to measure most of the variables suggested by the AAPOR standards was conducted
by Paletz, Short, Baker, Campbell, Cooper and Oeslander.[9] They studied the
reporting of polls by CBS-TV, NBC-TV, and the New York Times during 1973, 1975,
and 1977, and concluded that
". . . the flawed public opinion polls presently purveyed by the media are
inappropriate guides for public policy."[10]
Miller and Hurd focused specifically upon how well the Chicago Tribune, the Los
Angeles Times, and the Atlanta Constitution conformed to the AAPOR standards in
their reporting of opinion polls between 1972 and 1979. They found that
conformity was highest for the standard of sample size (85%) and lowest for
margin of error (16%).[11]
Salwen found "dramatic improvement" in conformity to AAPOR standards in two
major newspapers (the Detroit News and the Detroit Free Press) in a longitudinal
study of presidential campaign years from 1968 to 1984.[12] Smith and Verrall
conducted a critical analysis of Australian television coverage of opinion polls
in the 1980 Australian election. They studied conformity to the standards of
three different professional organizations: AAPOR, the National Council on
Public Polls, and the Australian Press Council. The authors concluded: "The
results of this study confirm earlier American findings that poll coverage is
extensive, superficial, and inaccurate."[13]
Lowry conducted a critical review of media-sponsored opinion polls during the
1984 presidential campaign to determine conformity to AAPOR standards. The
media studied were the three major network TV newscasts, three major
newsweeklies, as well as the New York Times and Washington Post. His major
conclusion was that ". . . all of the news media were deficient in following the
AAPOR standards in some respects, and some of them were deficient in most
respects."[14]
Demers carried out a qualitative analysis of the reporting of opinion polls in
15 large daily newspapers, one magazine, and one news service from 1980-1987.
Instead of an improved picture over time, he found four shortcomings: "1)
problems in survey design, 2) lack of background information, 3) problems in
analysis and 4) lack of insightful analysis."[15] Based upon their content
analysis of six newspapers in Campaign '88, Rollberg, Sanders and Buffalo also
ended on a critical note: "It is disturbing that the increase in the number of
polls reported in these newspapers was accompanied by a decrease in thoroughness
of reporting."[16]
The overwhelming conclusion from all of the above studies is that most of the
news media are doing a relatively poor job in presenting the results of public
opinion polls to the public in a way which (a) satisfies AAPOR's minimal
disclosure standards, or even (b) provides educated viewers and readers with the
basic information needed to interpret and evaluate the poll findings.
Research Questions
The present study was guided by five research questions.
1. Overall, did the number of media-sponsored polls increase between Campaigns
'92 and '96? Impressionistic evidence and anecdotal comments indicate that each
presidential campaign season produces more polls than the past campaign, but are
these impressions actually supported by the data?
2. Overall, did the media do a better job conforming to AAPOR disclosure
standards in Campaign '96 than in Campaign '92? Again, the purpose is to look
for change, for improvements, over time.
3. Which of the three categories of media (network TV, newspapers, and
magazines) did the best job of conforming to AAPOR disclosure standards?
Network TV of course operates under the severe limitations of a much smaller
"news hole" than do newspapers and news magazines. Therefore, do the print
media demonstrate higher levels of conformity than network TV news?
4. To what extent did these major media voices provide their readers and
viewers with additional cautionary information beyond the basic disclosure
standards recommended by AAPOR? More than two decades ago, Wheeler cautioned:
In sum, though the theory of polling is scientifically sound, the actual
practice is not . . . . Polling is not the pristinely pure scientific
process the
pollsters want us to believe it is . . . . But as serious as these
technical
problems are, as suspicious of polls as they should make us, they are not
the only
source of polling error. The fundamental problem of polls lies in their
analysis.[17]
Pollster Burns W. Roper also stated that the news media frequently mislead
readers and viewers---not by the facts they report, but by implication:
Partly because of their lack of deep expertise, reporters have overstressed
sampling error and understressed the other more important and considerably
greater
sources of error. And, in the process of stressing sampling error, they
have not
warned the reader or viewer of error as they have intended. Instead they
have
implied an unwarranted degree of accuracy.[18]
Miller, Merkel, and Wang state that the AAPOR standards provide only a starting
point for reporters when preparing poll stories.[19] Also, Brady and Orren
state that, "Measurement and specification errors . . . are often five or ten
times greater than sampling error."[20] Cantril stated, "The larger---and much
more important---problem with reports of sampling error is that they imply that
sampling is the primary, if not the only, source of possible error in a
poll."[21] Thus, for all of the above reasons, it becomes very important to
determine to what extent these prestige news media provided additional
cautionary information beyond the basic AAPOR disclosure standards.
5. Apart from the AAPOR disclosure standards, what were the most common
reporting/interpretation errors introduced by the three categories of media?
The importance of this question was suggested by Smith and Verrall's critical
analysis of factual and logical errors in Australian television coverage of
opinion polls.[22] The types of reporting/interpretation errors are defined
below.
Methods
Sample. The content universe for the TV portion of this study consisted of all
weeknights from August 24 through October 30, 1992, and from September 1 through
November 1, 1996. For each campaign, we drew a simple random sample of 25
network newscasts from ABC, CBS, CNN, and NBC.
The 199 newscasts (one was lost from Campaign '92 due to a recording error)
were viewed in entirety in two screenings. The purpose of the first screening
was to locate every presidential and vice presidential campaign story. A
campaign story was defined as being primarily about the national campaign. The
next step was to view all of the campaign stories to locate opinion polls
sponsored by the four networks.
For the three news magazines (Newsweek, Time, and U.S. News & World Report),
poll stories were located simply by reading all of the issues for the periods
defined above for the TV networks. Poll stories for the New York Times, the
Washington Post, and the Wall Street Journal were located by consulting the
index published by each paper.
Coding categories. The coding categories closely followed the recommended
AAPOR standard:
1. Sponsor---organization for which the poll was conducted
2. Question wording---complete wording provided for question(s) referred to in
the news report
3. Population---definition of population from which sample was drawn, i.e.,
specific type(s) of respondents and where located
4. Sample size---total number of persons interviewed
5. Interviewing method---e.g., in person, telephone, or mail
6. Interviewing date(s)---i.e., when interviewing was done
7. Sampling error---estimated percentage of error (+/-) specified
8. Partial results specified---identification of and basis for results that use
less than the total sample
In addition to the basic AAPOR standards, this analysis also coded the
following variables:
9. Additional cautions---any cautionary references to additional sources of
error or bias besides sampling error
10. Factual inaccuracy---clear misreporting of a factual detail
11. Overgeneralization---standard inductive fallacy, e.g., invalid
generalization across polls, or generalization to a population different from
that sampled
12. Modal force---any overstatement of the strength of a claim, especially
drawing certain conclusions from probabilistic poll data
13. Conclusive swing---a species of model force error, the claim that some
consequence would certainly (rather than probably or possibly) follow from
a swing of
some magnitude
14. Temporal transposition---reporting poll results in present (or future)
terms, e.g., "The poll shows candidate X is ahead."
15. Speculation---fallacy of drawing a conclusion from a hypothetical
situation---e.g., "If the election had been held last week, candidate X
would have
won."
16. Ambiguous comparison---reporting results in reference to an unspecified
baseline---e.g., "The results show a 4% swing to candidate X." Erroneous
because
unverifiable and essentially ambiguous.
17. Evaluative description---reporting poll results in evaluative (rather
than
factual) terms---e.g., "The latest poll is more good news for candidate X."
Categories 10 through 17 were taken from Smith and Verrall.[23] All coding was
done by the authors, working from video tapes of the network TV news and from
copies of articles from the print media.
Results
Although, as noted by research question 1, impressionistic evidence and
anecdotal comments indicated that each presidential campaign season produces
more polls than the past campaign; however, we found that there were fewer
media-sponsored polls stories in Campaign 96'. There were 70 poll stories in
Campaign '92 and 48 in Campaign '96. At first this seemed surprising, but in
light of the fact that other independent research has indicated that at least
the TV networks had reduced the total amount of their campaign coverage by
almost half from 1992 to 1996, this finding makes more sense.[24] If the print
media likewise reduced their total amount of political coverage, then the
reduction in poll stories, although not expected, would be easy to understand.
TABLE 1 ABOUT HERE
Research question 2 asked whether the media did a better job conforming to
AAPOR disclosure standards in Campaign '96 than in Campaign '92. Was there
improvement over time? Although some categories of the media did improve in
meeting the eight disclosure standards, in general the media did not do a
better job in conforming to the AAPOR standards. Of the 24 cells (8 variables X
3 categories of media), there were no changes in five cells, improvements in
seven cells, and lower scores in 12 cells. Furthermore, the magnitude of the
improvements tended to be smaller than the magnitude of the lower scores.
Identification of sponsor. As Table 1 indicates, all of the media received
perfect 100% scores on this standard, but these scores were an artifact of how
the study was conducted. Since we analyzed only poll stories that were
identified as being sponsored by the ten prestige media selected for analysis,
these scores had to total 100%.
Question wording. This standard refers to whether the news media provided
complete wording for questions reported in their poll stories. Table 1 provides
very mixed results on this standard. It shows that, on one hand, the prestige
newspapers improved their compliance from 44.4% to 50%. The networks doubled
their compliance, but still remained at only 18.2%. The news magazines, on the
other hand, dropped in compliance from 80% in Campaign '92 to 20% in Campaign
96.
Population specified. This standard required a definition (type of people
sampled and geographic location) of the population from which the sample was
drawn. News reports without information about the specific type(s) of
respondents and from which geographic area, did not meet this standard. Table 1
indicates that newspapers reported this information in half of their stories in
Campaign '92, but that the trend was in the wrong direction---down from 66.7% in
Campaign '92. The prestige news magazines and TV networks provided this
information in only three to five percent of their poll stories. To put it
differently, they failed to meet the AAPOR standard in at least 95% of their
poll stories in both campaigns.
Sample size. This standard requires the news story to report the total number
of persons interviewed in the poll. As Table 1 indicates, all three categories
of media performed worse in meeting this standard in Campaign '96 than in
Campaign '92. Once again, newspaper poll stories were far superior to news
magazine stories, and the news magazine stories were far superior to the TV
networks. In fact, the networks reported sample size in none of their Campaign
'96 stories.
Interviewing method stated. Network television news failed to meet this
standard completely. In both campaigns, not a single poll story on the networks
met this standard. News magazines were worse in 1996 than in 1992 . Again
newspapers did by far the best job in conforming to this standard, with about
two-thirds of their poll stories in each campaign reporting this information.
The following statement from the New York Times is a good example of a paragraph
explaining the interviewing method and selection of respondents:
The sample of telephone exchanges called was selected by a computer from a
computer list of exchanges in the country. The exchanges were chosen so as
to assure
that each region of the country was represented in proportion to its
population. For
each exchange, telephone numbers were formed by random digits, thus
permitting access
to both listed and unlisted numbers. Within each household one adult was
designated
by a random procedure to be the respondent for the poll.[25]
Interviewing date(s) stated. This standard requires that the date(s) when the
interviewing was done be reported. The prestige newspapers improved on this
standard, and ended with 75% conformity in Campaign .96. Both the news
magazines and network TV declined seriously in Campaign '96, and ended with 20%
and 40.9% compliance, respectively.
Sampling error specified. Network television news performed well in meeting
this standard. In fact, this was one of the few bright spots for the networks,
with 95.5% compliance in Campaign '96. Usually the sampling error was included
as part of the graphic, but often in difficult-to-read fine print. The news
magazines were worst, down from 45% compliance in Campaign '92 to only 20%
compliance in Campaign '96---75% lower than the networks.
Partial results unexplained. This variable in Table 1 is the only variable
where a low score is desirable. Since all of the percentages are low, this
means that all of the media essentially were in compliance on this standard.
These results are also an artifact, however, because almost none of the poll
stories reported scores for sub-populations (e.g., men vs. women, or blacks vs.
whites vs. Hispanics). Therefore the reason the media were in almost 100%
compliance is that the standard simply didn't come into play; it was not
applicable. One of the few good examples was the following:
One registered voter was selected from each household by a procedure aimed
at
providing appropriate proportions of male and females respondents.
Certain questions were asked of all respondents and certain ones were asked
of
half the respondents. Among those questions asked of everyone, chances are
19 of 20
that if all registered voters with telephones in the U.S. had been
surveyed, the
findings would differ from these poll results by no more than 3.2
percentage points
in either direction. The margin of error for questions asked of half the
sample would
be 4.5 percentage points. The margin for a subgroup would depend on the
size of that
group.[26]
Research question 3 asked which of the three categories of media (newspapers,
news magazines or network TV) did the best job of conforming to AAPOR disclosure
standards. The overall results of Table 1 indicate that the prestige newspapers
were the clearly in first place in compliance, and the networks were in last.
Research question 4 asked to what extent the prestige media provided their
readers and viewers with additional cautionary information beyond the basic
disclosure standards recommended by AAPOR. In absolute terms, all three
categories of media did a poor job of supplying additional cautionary
information. However, in relative terms, newspapers did much better than the
other two categories of media, providing additional cautionary information of
33.3% of their poll stories in Campaign '92 and in 43.8% in Campaign '96. The
corresponding scores for news magazines were 25% and 0%. The TV networks were
again in last place, supplying additional cautionary information in only 9.4% of
their poll stories in Campaign '92 and in none of their poll stories in Campaign
'96.
One useful example of additional cautionary information was provided by the New
York Times:
In addition to sampling error, the practical difficulties of conducting any
survey of public opinion may introduce other sources of error into the
poll.
Variations in question wording or the order of questions, for instance, can
lead to
somewhat different results.[27]
TABLE 2 ABOUT HERE
Research question 5 asked about the extent to which the prestige news media
introduced eight different types of reporting/interpretation errors apart from
the AAPOR standards. Table 2 indicates that the biggest problem area was with
temporal transpositions, and the TV networks were by far the most frequent
violator of this type of error. Polls always report the state of affairs at the
time the polls were taken (past tense). The networks anchors and reporters
tended to use present tense when reporting the results, far more so than did the
prestige print media. However, it should be pointed out that the networks did
make a 48% improvement in this respect between Campaigns '92 and '96.
The networks also had the worst scores in terms of making evaluative statements
about the poll results of the various candidates---e.g., reporting the results
as "good news" or "bad news" rather than simply as factual statements.
Likewise, the networks had the worst scores in terms of making ambiguous
comparisons of polling results.
The results for factual inaccuracies were somewhat mixed---ranging from a low
of zero for newspapers in Campaign '96 to a high of 30% for magazines in the
same campaign. Most factual errors resulted from discrepancies between
statistics reported in stories versus in tables or on-screen graphics.
Discussion and Conclusions
Contrary to our expectations, we found that the total number of poll stories
was down, not up, in Campaign '96. Of much greater importance, though, was our
finding that the ten prestige media we studies did a poorer job of complying
with AAPOR standards in Campaign '96 than in Campaign '92. And the Campaign '92
scores were not very high to begin with.
The news media in election years are frequently criticized of placing too much
emphasis on "horserace journalism," of reporting in terms of winners and losers.
If the same words were applied to this study in terms of between-media
comparisons, the prestige newspapers were the clear overall winners, and the TV
networks were the clear losers. This does not mean, however, that the prestige
newspapers did not have considerable room for improvement.
The standard we used for coding the media's reporting of the population
standard was quite rigorous, and did not give credit for implied geographic
locations. What this means then, is that the scores on this standard,
especially for the TV networks, were somewhat suppressed. For example, if the
anchors or correspondents referred to "our latest survey of likely voters," with
no additional information given about where the likely voters lived, then this
was coded as not specifying the population. On the other hand, if the phrase
had been "our latest nationwide survey of likely voters," this would have been
coded as specifying the population, because if provides both the who and where
of the population. Therefore, if one were to give the networks credit for
implied information (e.g., nationwide), then the networks would look better on
this standard.
On a different standard, reporting margins of error, however, the score of
95.5% for the networks in Campaign '96 actually makes the networks look better
than they really were. The reason for this is that (a) the error margins were
usually presented only on-screen as part of a graphic, and (b) most of the +/-
scores were so small, or buried among stars, stripes, donkeys, elephants, and
the much larger percentages being reported, that only a dedicated scholar
analyzing video tapes would be likely to comprehend them. A typical news viewer
does not analyze news graphics this closely, and therefore, without the anchor
or correspondent actually mentioning the margin of error, this information would
in the vast majority of cases not be perceived by the viewer. It would be on
the screen, and the networks would be technically in compliance, but very few
viewers would be better-informed. This would be a worthwhile topic for future
experimental research dealing with (a) the perception of and (b) comprehension
of the information that is presented in on-screen graphics.
It is also ironic that the high network compliance on the standard of margin of
error is also the standard which, when presented without additional disclosure
information, can be misleading to news consumers. The reason, as pointed out
above, is that it implies more precision for the overall survey than is actually
warranted. Most viewers are not researchers or polling specialists, and do not
realize that the total error of a survey is the sum of sampling error and
measurement error. One of the most important conclusions of this critical
analysis, then, is that all of the prestige media were deficient in reporting
additional relevant disclosure information---i.e., deficient in not going beyond
sampling error.
Although the news magazines were not the worst among the three categories of
media, they were the most disappointing. The reason is that all three news
magazines for this study were of course weeklies; therefore, in terms of news
writing, layout and any other editorial decisions, the editors had the luxury of
more editorial time than the newspaper editors or TV news producers to include
"the right stuff." However, the news magazines, with weekly deadlines, did not
outperform the newspapers, with daily deadlines, in meeting the AAPOR standards.
Can the media do a better job of reporting opinion polls? Yes. Should the
media do a better job? Yes. Cantril stated, "As reading and viewing audiences
are exposed to more polls, and as political analysts and commentators draw on
polls with growing frequency, it would seem that poll consumers could be treated
to more sophisticated forms of poll analysis."[28]
Will the media do a better job in Campaign 2000? Probably not---at least not
without additional encouragement from AAPOR, from NCP, and possibly even from a
special outside committee of independent scholars and journalism educators,
perhaps created under the auspices of the Association for Education in
Journalism and Mass Communication.
Crespi stated, concerning the apparent lack of concern of those who are
involved in media-sponsored polls, ". . . that the prospects for improvement in
the near future must be rated as dim."[29] In light of the decrease in
compliance we found in this analysis between Campaign '92 and Campaign '96, we
share Crespi's pessimistic view.
While we are not arguing that our democratic form of elections and
self-government will stand or fall depending upon what the prestige media do in
this respect, we are arguing that these media do have a responsibility to raise
the level of campaign reporting by providing reasonably full disclosure of
polling methods. Along with the privilege of reporting timely polling
information to millions of readers or viewers should go the responsibility of
providing those increasingly better-educated readers and viewers with the basic
information necessary to interpret the poll stories for themselves.
_________________________________________________________
Table 1: Percentage Conformity with AAPOR Standards by Year and by Three
Categories of Media
Campaign '92 Campaign '96
Papers Magazines TV Papers Magazines TV
(N = 18*) (N=20) (N=32) (N=16) (N=10) (N=22)
____________________ _____________________
AAPOR Standards
Identification of Sponsor 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%
Question Wording Provided 44.4 80.0 9.3 50.0 20.0 18.2
Population Specified 66.7 5.0 3.1 50.0 0.0 4.5
Sample Size Given 77.8 45.0 12.5 75.0 20.0 0.0
Interviewing Method Stated 66.7 40.0 0.0 68.9 10.0 0.0
Interviewing Date(s) Stated 66.7 60.0 71.9 75.0 20.0 40.9
Sampling Error Specified 77.8 45.0 81.3 75.0 20.0 95.5
Partial Results Unexplained 5.5 0.0 0.0 0.0 0.0 4.5
*Column N's refer to the number of poll stories reported.
_________________________________________________________________
_________________________________________________________________
Table 2: Percentage of Reporting/Interpretation Errors Introduced by Year and by
Three Categories of Media
Campaign '92 Campaign '96
Papers Magazines TV Papers Magazines TV
(N = 18*) (N=20) (N=32) (N=16) (N=10) (N=22)
____________________ _____________________
Types of Errors
Factual Inaccuracy 11.1% 15.0% 3.1% 0.0% 30.0% 9.1%
Overgeneralization 11.1 10.0 6.3 0.0 0.0 0.0
Modal Force 0.0 10.0 0.0 18.8 0.0 0.0
Conclusive Swing 22.2 10.0 0.0 6.3 0.0 0.0
Temporal Transposition 5.6 10.0 84.4 6.3 0.0 36.4
Speculation 0.0 10.0 3.1 12.5 0.0 4.5
Ambiguous Comparison 11.1 15.0 34.4 0.0 0.0 13.6
Evaluative Description 11.1 10.0 28.1 12.5 0.0 36.4
*Column N's refer to the number of poll stories reported.
_________________________________________________________________
[1] Frank Luntz, "All Polls Are Not Created Equal," U.S. News & World Report,
Sept. 28, 1992, p. 24.
[2] Albert H. Cantril, The Opinion Connection: Polling, Politics, and the
Press, Washington, DC: Congressional Quarterly Press, 1991, p. 237.
[3] See Irving Crespi, Public Opinion, Polls, and Democracy, Boulder, CO:
Westview Press, 1989, p. 15.
[4] For examples of this body of literature, see: Mack C. Shelley II and
Hwarng-Du Hwang, "The Mass Media and Public Opinion Polls in the 1988
Presidential Election: Trends, Accuracy, Consistency, and Events," American
Politics Quarterly, 19: 59-79 (1991); John Bare, "Truth about Daily Fluctuations
in 1992 Pre-election Polls," Newspaper Research Journal, 15: 73-81 (1992);
Andrew Gelman and Gary King, "Why Are American Presidential Election Campaign
Polls so Variable When Votes Are so Predictable?" British Journal of Political
Science, 23:409-451 (1993); Richard R. Lau, "An Analysis of the Accuracy of
'Trial Heat' Polls During the 1992 Presidential Election," Public Opinion
Quarterly, 58:2-20 (1994).
[5] American Association for Public Opinion Research, Code of Professional
Ethics and Practices, 1986.
[6] Norm Goldstein, ed., The Associated Press Stylebook and Libel Manual,
Reading, MA: Addison-Wesley, 1996, pp. 161-162.
[7] See Donald L. Shaw, Maxwell McCombs and Gerry Keir, Advanced Reporting:
Discovering Patterns in News Events, Prospect Heights, IL: Waveland Press, 1997,
p. 133.
[8] See Gerhart D. Wiebe, "The New York Times and Public Opinion Research: A
Criticism," Journalism Quarterly, 44: 654-658 (Winter 1967).
[9] David L. Paletz, Jonathan Y. Short, Helen Baker, Barbara Cookman Campbell,
Richard J. Cooper, and Rochelle M. Oeslander, "Polls in the Media: Content,
Credibility, and Consequences," Public Opinion Quarterly, 44:495-513 (1980).
[10] Ibid., p. 511.
[11] M. Mark Miller and Robert Hurd, "Conformity to AAPOR Standards in
Newspaper Reporting of Public Opinion Polls," Public Opinion Quarterly,
46:243-249 (1982), at p. 246.
[12] Michael B. Salwen, "The Reporting of Public Opinion Polls During
Presidential Years, 1968-1984," Journalism Quarterly, 62: 272-277 (Summer 1985).
See also Michael B. Salwen, "Does Poll Coverage Improve as Presidential Vote
Nears?" Journalism Quarterly, 62:887-891 (Winter 1985).
[13] Ted J. Smith III and Derek O. Verrall, "A Critical Analysis of Australian
Television Coverage of Election Opinion Polls," Public Opinion Quarterly,
49:58-79 (1985), at
p. 76.
[14] Dennis T. Lowry, "Media-Sponsored Opinion Polls: A Critical Review of
Campaign '84 and Recommendations for Campaign '88," paper presented at the
annual convention of the Association for Education In journalism and Mass
Communication, San Antonio, Texas, August 2, 1987.
[15] David Pearce Demers, "Commentary: A Qualitative Analysis of Newspaper
Polls," Newspaper Research Journal, Vol. 9, Spring 1988, pp. 105-114.
[16] Jeanne Norton Rollberg, Luther W. "Sonney" Sanders, and M.D. Buffalo,
"Down to the Wire: How Six Newspapers Reported Public Opinion Polls During the
1988 Presidential Campaign," Newspaper Research Journal, 11:80-93, Fall 1990, at
p. 91.
[17] Michael Wheeler, Lies, Damn Lies, and Statistics: The Manipulation of
Public Opinion in America, New York: Liveright, 1976, p. 103.
[18] Burns W. Roper, "The Media and the Polls: A Boxscore," Public Opinion,
3:46-49, 1980, at p. 46.
[19] Peter V. Miller, Daniel M. Merkle, and Paul Wang, "Journalism with
Footnotes: Reporting the 'Technical Details' of Polls," in Paul J. Lavrakas and
Jack K. Holley, eds., Polling and Presidential Election Coverage, Newbury Park,
CA: Sage, 1991, pp. 200-214.
[20] Henry E. Brady and Gary R. Orren, "Polling Pitfalls: Sources of Error in
Public Opinion Surveys," in Thomas E. Mann and Gary R. Orren, eds., Media Polls
in American Politics, Washington, DC: The Brookings Institution, 1992, pp.
55-94, at p. 68.
[21] Cantril, op cit., p. 119.
[22] Smith and Verrall, op cit.
[23] Ibid., p. 72.
[24] Dennis T. Lowry and Jon A. Shidler, "The Sound Bites, the Biters, and the
Bitten: A Two-Campaign Test of the Anti-Incumbent Bias Hypothesis in Network TV
News," presented at the annual convention of the Broadcast Education
Association, Las Vegas, Nevada, April 3, 1998.
[25] Robin Toner, "Survey Finds Hostility to Perot and no Shift in Race," New
York Times, Oct. 6, 1992, p. A:4:1.
[26] D. Shridman, "Home Stretch: Bill Clinton is Solidifying His Grip on the
Electorate, Latest Poll and Interviews Show," Wall Street Journal, Oct. 23,
1992, p. A:8:3.
[27] R.L. Berke, "Poll Finds Dole Has Cut Clinton's Lead in Half Since
Convention and Clarified Goals," New York Times, Sept. 6, 1996, p. A:14:1.
[28] Cantril, op cit., p. 214.
[29] Irving Crespi, Pre-Election Polling: Sources of Accuracy and Error, New
York: Russell Sage Foundation, 1988, p. 184.