Back

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

The British parliamentary election is all over. As predicted by the opinion polls, Tony Blair led the Labour Party to a historic third term in government. However, in line with the predictions, Labour Party's majority in the parliament has drastically reduced.

The race was tight, over the 30-day campaign period. Eclipsed by the historic visits of Lien Chan and James Soong to Mainland China, Hong Kong's media were less interested in the British election than in the United States presidential election last year.

The influence of Britain on Chinese affairs is gradually declining. In the mind of the Chinese and Hong Kong people, the influence of Britain is becoming less and less important.

Nevertheless, from the perspective of public opinions' studies, since both Britain and the United Sates are developed democratic countries, mastering the relations between public opinion surveys and democratic development in these countries will be rather useful to the development of the local civil society.

When one takes a look at the results, the forecasts of major pollsters in Britain were actually rather accurate. The BBC has kept on tracking and reporting the figures of the 6 major pollsters during the election period (first 6 items in Table 1), and the final forecasts of 5 of them fell within the error margins. If one takes the average of these 6 pollsters, results then were extremely accurate. Besides, the exit poll, jointly released by the BBC and the ITV for the first time, was also able to make an instant and accurate forecast on the changes of support ratings and seats obtained, saving face for exit polls.

Since pollsters in Britain have strictly followed the professional code of conduct, the public can check out the details of their research methodology and raw figures from the internet. The author will just give a brief introduction here to local readers.

In terms of research methodology, among the 6 surveys, most of them were random sampling telephone surveys, with a sample size of more than 1,000 cases targeting at citizens of age 18 or above. All data have been weighted and adjusted according to the national population structure. It is worth noticing that all 6 pollsters have adopted different methods to adjust their raw figures so as to come up with more accurate forecasts. Methods adopted by various pollsters are as follows:

The statistics of the Times only focused on analyzing the preference of those who would definitely vote. Yet, among those who have not decided how to vote, 60% were also selected and allocated to the party they voted for in the last parliamentary election.

The Financial Times required the respondents to rate their propensity to vote from 1 to 10 marks, and forecasts were made basing on the results from those giving 10 marks.

The Guardian also required the respondents to rate their propensity from 1 to 10 marks, and adjustments were made according to the respondents' previous voting record. They took 50% of those who possessed good voting record, intended to vote but were undecided who to vote for this time, and then allocated them to the party they supported in the last election.

The Independent invited the respondents to rate their propensity to vote from 0 to 10 marks. Results were then weighted proportionately. Those answering 0 marks were discarded, and people answering 1 marks were given a weight of 0.1, 2 marks with a weight of 0.2, and so on, while those giving an answer of 10 marks were allocated a weight of 1.0. Some answers were further adjusted by the respondents' past voting record.

The Independent on Sunday required the respondents to rate their propensity from 1 to 10 marks. While those giving 4 marks or below were discarded, the rest were weighted proportionately. Respondents who were uncertain whether they would vote were further "squeezed" to give an answer to their preference. These answers were then combined with the results of the other samples.

The research design and weighting method of the Telegraph were the most special. An online panel of around 70,000 subscribers was first recruited, and 2,000 of them were selected for regular online surveys. The results were adjusted according to the respondents' personal background, voting record, newspaper reading habits, etc, so that it could also reflect the structure of voters without internet access and make up for the deficiencies of an online survey.

The above introduction shows that, the forecast methods adopted by different pollsters were actually rather complex, but mainly practical in nature. Since around one-third of voters still have not decided who to vote for in the forthcoming election, in fact only several hundreds constituted the base of the above surveys. Yet, with the clearly stated methodology and highly civilized society, these surveys have not been wildly criticized.

Besides, it is worth mentioning that the opinion survey of the Times, which was frequently cited by the media in Hong Kong, in fact was exactly a rolling survey. From 20 April onwards, 350 cases were conducted every day. As a result, a total of 1,400 cases were accumulated in 4 days, while 350 cases were added and dropped on a "plus one, minus one day" basis. Results were released by the Times and the ITV every day. Although rolling surveys were less popular in Britain when comparing to the United States, they were also considered to be a reliable and widely accepted research method. Seemingly when reporting the polls of the Times, the media in Hong Kong has not paid attention to their research method. While some academics and experts in Hong Kong have expressed their opinions on the Legislative Council Election rolling surveys conducted by the author some time ago, the author suggests that they should also take a look at the development of rolling surveys in Britain and the United States.

Before the Legislative Council Election, the author has explained the rolling survey conducted by the University of Essex in the last British parliamentary election in 2001. As 150 voters were interviewed by the University every day, over 1,000 cases were accumulated over a 7-day period. The results were then released via the internet every day. This year, the University of Essex has conducted the rolling survey and results were again released on a daily basis, though some modifications were made in the research design.

First of all, 7,813 voters were registered via the research agency before the election. The University then interviewed 270 people every day via the internet during the election period, but only the answers from "definite" voters, i.e. respondents giving 10 marks when asked to self-assess their propensity to vote basing on a scale of 0-10 marks, were used for the main calculation base. Around 1,100 valid cases were accumulated after 6 days, and results were released on a daily basis. Table 2 shows the detailed figures of the rolling surveys, and the last round was released on the election day. Results can also be found in Table 1.

The University of Essex is the top university specializing in British elections and is responsible for designing and conducting the British elections research series which has more than 40 years. All research funding comes from the public money, including the pre-election and post-election large-scale interviews and tracking surveys which have not been mentioned above.

If Hong Kong society really wants to "overtake Britain and catch up with the United States", judging from the levels of election and policy research, seemingly there is still a long way to go. We'd also better not have high expectations that Hong Kong can become Manhattan in the United States or London in Britain if we are still stuck in questions like whether opinion surveys are academic, whether rolling surveys are repetitive, whether policy research should be government-oriented, whether academics should comment on political matters, etc.

Table 1: Final Forecasts of Various Pollsters