This was a random telephone survey conducted by telephone interviewers under close supervision. To minimize sampling bias, telephone numbers were first drawn randomly from the residential telephone directories as "seed numbers", from which another set of numbers was generated using the "plus/minus one/two" method, in order to capture the unlisted numbers. Duplicated numbers were then filtered, and the remaining numbers were mixed in random order to produce the final telephone sample. When telephone contact was successfully established with a target household, one person of age 18 or above was selected. If more than one subject had been available, selection was made using the "next birthday rule" which selected the person who had his/her birthday next from all those present.
The fieldwork was conducted during the period of January 30-31, 2008. A total of 514 qualified Hong Kong citizens who spoke Cantonese were successfully interviewed. The overall response rate of this survey was 73.7%, and the standard sampling error for percentages based on this sample was less than 2.2 percentage points. In other words, the sampling error for all percentages was less than plus/minus 4.4 percentage points at 95% confidence level.
To increase the representation of the survey, the data collected have been adjusted according to provisional figures obtained from the Census and Statistics Department regarding the gender-age distribution of the Hong Kong population in mid-2007. Data quoted in this report were in provisional figures.
This survey adopted 4 out of the 6 questions that had been used in the 2006 survey, as well as 2 new questions to gauge respondents' knowledge on Carbon Dioxide emission and stance over legislation to regulate electricity company over emission of greenhouse gases.
For comparison purpose, statistical tests of "difference-of-proportions" have been employed whenever applicable, in order to check for significant changes in between the past 2
surveys. Figures marked with double asterisks (**) indicated that the variation has been tested to be statistically significant at p<0.01 level, whereas those with single asterisk (*) denoted statistical significance at p<0.05 level.
