Back

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

Predicting election outcomes is a nasty job. Evading it altogether, however, is somewhat irresponsible. Perhaps the best approach is to explain the process and its limitations, so that everyone learns from everybody else.

This article focuses on the analysis of voters' intention to vote, at the final stage of election campaign. It also describes one method of predicting candidates' chance of winning the election, matched to their political alignment.

Voting intention cannot be taken at its face value

Every researcher who studies Hong Kong's elections knows that voters' intention to vote in a certain election as measured in random surveys cannot be taken at its face value. There is usually a wide gap between the number of people who said they would vote, and those who actually voted. Looking at a decade's record of our HKU POP pre-election surveys, only 48% of those who said they would vote in District Council elections actually voted. For Legislative Council elections, the figures ranged from 52% to 69%.

There could be a number of reasons for such discrepancies:

(1) Due to various reasons, voters who said they intended to vote had eventually not voted. Common reasons include their busy schedule, elections not attractive enough, or bad weather, just to name a few. Many researchers would, therefore, probe into respondents' intensity of commitment to vote, and whether they had already decided whom to vote for, in order to work out the discount rate.

(2) Voters' intention to vote has not yet stabilized when the survey was completed. The only solution to this problem is to constantly monitor the latest development of voter's propensity to vote.

(3) The respondents lied, or were unwilling to tell the truth due to social desirability effect.

(4) Bias occurred in the sampling process, and the survey sample became unrepresentative.

(5) Special factors occurred, like the issuing of election souvenir cards in a certain year, or significant revamp of the electoral roll, thereby causing structural changes to the discount rate.

As for this year's elections, when compared to 1999, voters' intention to vote began to level off about three weeks before the elections, and then fell. Projecting from this set of figures along, voters' intention to vote on election day may be 3 percentage points lower than that of 1999, which is a negative factor on the turnout rate. Yet, the percentage of registered voters who have made up their mind on how to vote has increased, compared to the corresponding figure of 1999. Based on the assumption that this figure would stand stable at the current level of 37% on election day this year, and the corresponding figure for 1999 had followed the trend and rose to 31%, the positive push factor would be 6 percentage points. All in all, there may be a net gain of 3 percentage points. Starting from the actual turnout rate of 35.8% in 1999, and taking into consideration the increase of 3 percentage points, this year's turnout may well be within the 36% to 41% region, depending on the final atmosphere of the election campaign.

The above project has only taken care of the first two factors aforementioned. Other factors, like the social desirability, sampling bias and outdated electoral roll, can only be taken as unchanged.

Candidates' chance of winning according to track record

Prior to each District Council Elections in the past, POP always tried to work out the chance of winning for candidates from different political camps across various districts. The methodology described below has been fairly accurate in 1994 and 1999. The projection method is summarized as follows:

1. The 837 candidates (including those elected unopposed) in all 400 geographical constituencies across 18 districts were first grouped into 4 camps according to their political affiliations, namely, "Pro-China", "Democrats", Liberal Party and "Independents/others". Voters' support to the candidates of different political groups was then gauged.

2. Restricted by the length of the questionnaire, voters were prompted only with the names of political groups which sent most candidates to the election. The "Pro-China" camp only included Democratic Alliance for Betterment of Hong Kong, Hong Kong Progressive Alliance and Civil Force. "Democrats" only included Democratic Party, Hong Kong Association for Democracy & People's Livelihood, Frontier, Neighbourhood and Worker's Service Centre and Civic Act-up. Liberal Party formed its own group, while all other candidates were grouped under "Independents/Others".

3. Using the latest poll figures, we then worked out a political strength index for each of the political camps in each constituency. The number of seats won by each political camp in each district, and hence the success rate for each of them, were obtained by multiplying the number of candidates from different political camps by the political strength index of their group, followed by slight adjustments for errors registered in the previous projection exercise.

Using this method, among the 400 directly-elected seats in the coming District Council, we project that the Pro-China camp would get about 116 seats, nearly 107 seats would go to the democrats, around 9 would go to the liberals, while the remaining 169 seats or so would be shared by other candidates. Notwithstanding its higher rate of success, the democrats would won fewer seats than the Pro-China camp. Together with seats appointed by the Chief Executive, the coming District Council would still be dominated by the conservatives

The error of projection could be fairly big for individual districts, but in terms of the overall success rates of different political camps territory-wide, the average error registered in 1994 was less than 5%, while that of 1999 was below 3 percentage points. The accuracy of this year's projection is yet to be seen.