Back

Note: This article represents the view of the author and not the University of Hong Kong.

The United States presidential election is drawing near, opinion polls are again springing up everywhere in the whole of America. To the American and Hong Kong peoples, this is nothing special. Many in Hong Kong are actually enjoying the flames and fumes of American politics.

In the final countdown of the election, electronic and printed media in Hong Kong are constantly citing the result of opinion polls conducted by various American media, in order to explain the latest development. According to the author's rough estimate, polls conducted by Reuters, Zogby, CNN, Fox News, and Washington Post are most frequently quoted by Hong Kong's media.

Take the news report 3 days before the election as an example, the news reports of TVB, ATV and Cable TV that day have used charts to show the result of the public opinion poll jointly conducted by Reuters and Zogby, i.e. Bush drew 47% to 47% with Kerry. TVB and ATV also used charts to show the result of the poll conducted by Fox News, i.e. Bush was ahead of Kerry at 50% to 45%. ATV and Cable TV also used charts to show the result of the poll conducted by Washington Post, i.e. Bush was ahead of Kerry at 50% to 47%. In other words, the polls of Reuters/Zogby and Fox News were reported by TVB, the polls of Reuters/Zogby, Washington Post and Fox News were reported by ATV, while the polls of Reuters/Zogby and Washington Post were reported by Cable TV. Besides, all 3 television stations have reported the integrated analysis of CNN, revealing that if the election were held on that day, Bush could obtain 227 votes while Kerry only had 207 votes.

Now, let us take a look at the news reports of the radio stations. The result of the poll conducted by Reuters and Zogby was quoted by RTHK, Commercial Radio and Metro Radio. Thus, one can be certain that the poll jointly conducted by Reuters and Zogby was most respectable by the local media. This observation is also evident from the coverage of the local newspaper.

However, when reporting the above polls, the local media has generally missed out the basic information of the concerned polls. One possible reason was that the responsible media practitioners might not have enough understanding of the concerned poll. Another reason was that a set of code of practice for the public opinion polls reporting was lacking in the local media, resulting in these rough reports. As a result, it seems that the United States presidential election's public opinion polls could not fit into the local election polls. In fact, this is a misunderstanding. In this article, the author would like to have an illustration on the rolling polls of the United States media.

Since CNN's analysis of integrated public opinion figures belongs to another type of analysis, namely "poll of polls", details were skipped due to the limitation of space here. The author would like to focus his discussion on the pre-election surveys conducted by authoritative media bodies such as Reuters/Zogby and Washington Post /CNN. In fact, all of them were rolling polls. According to the professional code of ethics adopted by the United States media, important information such as the rolling method, target sample, sampling error and phrasing of questions have to be laid down in the websites in detail and mentioned in the original reports. It is better for the Hong Kong media to quote the result cautiously.

First, let us take a look at the survey jointly conducted by Reuters and Zogby. Information reveals that, when the local media quoted the Reuters/Zogby's survey in which Bush drew 47% to 47% with Kerry, in fact that was the nation-wide telephone survey conducted by Zogby. The date of survey was 26 to 28 October and calls were made from the headquarter of Zogby in New York. The date of release in the United States was 29 October. The sample size was 1,203 people intending to vote, with a sampling error of +/-2.9 percentage points.

The Zogby's website has clearly stated that the data have been adjusted according to the district, party sense, age, race, religion and gender of the voting population, and the sampling error of the sub-sample was higher than that of the overall sample. Besides, Zogby has also stated, while Reuters sponsored Zogby's presidential rolling survey exclusively, all figures, analyses and commentary were provided by Zogby.

More importantly, the releases of both Reuter and Zogby have clearly stated that the survey was a 3-day rolling survey, with an accumulation of 1,200 cases in 3 days. The concerned rolling survey started from 1 October to 1 November, the day before the election. Thus, the sample would move forward to include 400 cases and move backward to drop 400 cases every day. Hong Kong people should be very familiar with this kind of rolling method. Yet, it is strange that when reporting the survey, it seems that the local media did not know it is a rolling survey. Possibly it was because the media simply took no notice of the research design of the original source, or might have been distracted by the use of "daily tracking poll" in the original source and as a result, paid little attention to the illustration of "rolling sample" in the text. In the United States, rolling survey and tracking survey sometimes mean the same thing.

Now let us look at the joint poll conducted by Washington Post and CNN. Information reveals that, they were also rolling surveys which began from 1 October. The website of the 2 bodies explained that, the survey was conducted by a research agency in Pennsylvania. A certain number of self-reported registered voters and voters likely to vote were selected and interviewed every day. The daily sample size was controlled at around 350 people intending to vote. On or before 23 October, the results released by the 2 organizations were 3-day samples and from 24 October onwards, 4-day samples were released instead.

Although Washington Post and CNN co-owned the data, each of them developed their own analyzing method, especially regarding the definition of "voters intending to vote" and a series of filtering and weighting method. Therefore, the figures released by the 2 bodies on the same day would always be different. As per ATV and Cable TV news reports mentioned previously where according to the poll conducted by Washington Post, Bush was ahead of Kerry at 50% to 47%, this was in fact based on the result of the rolling poll conducted by Washington Post from 25 to 28 October. The date of release was 29 October and there were 2,820 U.S. citizens in the sample. Among them, 2,488 were self-reported voters and 2,047 of them claimed they would go to vote. The base of this 50% to 47% was 2,047 people, with a sampling error of +/- 3 percentage points.

Among the local media's coverage, the poll result of Reuters and Washington Post were different. The reason was that the survey dates, the number of rolling days, and their definitions on "voters intending to vote" were different.

Last of all, the survey of Fox News was briefly mentioned here. Information reveals that, as per TVB and ATV news reports mentioned previously where Bush was ahead of Kerry at 50% to 45% according to the poll conducted by Fox News, this was in fact the telephone survey conducted from 27 to 28 October. The date of release was 29 October and the sample size was 1,400 registered voters. Among them, 1,200 were voters intending to vote. The sampling error was alleged to be +/-3 percentage points. Apparently, the survey date of Fox News was again different from those of Reuters and Washington Post, and thus, direct comparison was impossible.

Since it was 10 days ago Fox released the result of a similar survey, it was clear that the design of rolling survey was not adopted here. However, when the election is drawing near, many US media bodies have speeded up the pace of their research and altered their strategy in surveys. Whether Fox News will adopt the rolling survey eventually is still under observation.

This article concentrates on discussing the US presidential rolling surveys popularly reported by the Hong Kong media, aiming at enabling the local media practitioners to handle and report the relevant figures more accurately in the coming 2 days, and allowing the public to have a deeper understanding on the public opinion figures' meaning and limitations. Besides, it is also hoped that those commentators and heads of the policy research agencies who have criticized the rolling surveys of the local Legislative Council elections as merely a repetition of figures could also broaden their horizon, put away their bias and collect more information, so as to have a more comprehensive discussion on the pros and cons of various public opinion polls.

List of major research agencies in the US presidential election

Name of research agencies	Name of executing agencies	Main research method
Reuters News Agency	Zogby International	Long-term rolling poll
Zogby International	Zogby International	Long-term rolling poll
CNN	CNN	Integrated analysis of polls from various bodies
ABC News	TNS of Horsham	Long-term rolling poll
Washington Post	TNS of Horsham	Long-term rolling poll
Fox News	Opinion Dynamics Corporation	Mainly single poll