HKU POP releases results on the feature study of Legislative Council ElectionBack

 

Press Release on August 26, 2016


Special Announcements

1. The Public Opinion Programme (POP) of The University of Hong Kong has submitted its application to the Registration and Electoral Office by the deadline yesterday to conduct exit poll in 50 polling stations on the Legislative Council Election Day on September 4. Different from the previous operations, the exit poll this year is only a voluntary student project. There will be no data analysis carried out on the election day, no predictions of the election results, also no collaboration with any non-academic organizations and media.

2. This year, POP is conducting the Legislative Council election rolling survey for 35 consecutive days, which is co-sponsored by a number of media organizations and a civic organization “Power for Democracy”. Yesterday some of the media sponsors decided to stop reporting the rolling survey figures, their rationale is it is inappropriate for media to co-sponsor the survey with a political organization. POP responded that as long as the sponsoring organizations agree to our service policies, which include giving POP full autonomy in its operation and also agreeing to release all survey data for public consumption, POP would not concern their background and stance. POP also treats all sponsors on an equal footing. However, POP would remind Power for Democracy that it has explained publicly that their reason for sponsoring this rolling survey is solely to provide reference information and original data to all candidates. POP respects the decision made by some of our media sponsors that they will not report our rolling poll survey result anymore. However, POP thinks that such act is not conducive to facilitate free information flow and to construct a civil society.

3. As the Legislative Council election draws near, please note that, starting from today, POP will follow its general practice to include the projection of winning chances of all candidates in the rolling survey results.

4. The feature study released by POP today was self-funded by POP. The survey questions are similar to the previous election surveys. POP hereby thanks all the sponsors of our rolling survey, which supports POP to develop this platform to conduct such feature study.


Abstract

POP conducted two surveys on the feature study of Legislative Council election in between July and August, 1,056 and 2,157 registered voters were interviewed by means of random telephone surveys conducted by real interviewers. The latest survey shows that after the nomination period for this year’s Legislative Council election, nearly 70% of all registered voters were able to tell us the exact month of the election, namely, September. This is much lower than four years ago. As for split-voting, only about 10% of the voters said they are willing to adopt strategic voting, which is also much lower than that registered four years ago. Besides, in terms of priority, voters’ concerns with respect to the 6 attributes prompted in the survey are candidates’ livelihood policies, economic policies, political issues, relationship with the Central Government, council experience and fame. Among them, voters are giving more weight to candidates’ relationship with the Central Government, and less importance to their council experience and livelihood policies. This probably reflects the changes in Hong Kong’s political ecology in the past 4 years. Regarding different geographical constituencies, voters in all constituencies are giving more importance to candidates’ relationship with the Central Government this year. Voters in Hong Kong Island and Kowloon West show significantly lower concern to candidates’ livelihood policies. Meanwhile, voters in Hong Kong Island show significantly higher concern to candidates’ political policies than 4 years ago, those in Kowloon West show significantly lower concern to candidates’ economic policies, those in Kowloon East show significantly lower concern to candidates’ fame, while those in New Territories West are giving less importance to candidates’ council experience. As for “Super DC”, since most voters interviewed belong to this constituency, findings are highly comparable to those of the overall sample. The maximum sampling error of all percentage figures is +/-6 percentage points at 95% confidence level, while the response rate for both stages is 75% and 72% respectively.

Points to note:

[1] The address of the “HKU POP SITE” is http://hkupop.hku.hk, journalists can check out the details of the survey there.

[2] The sample size of the first survey is 1,056 successful interviews, not 1,056 x 74.6% response rate, while the sample size of the second survey is 2,157 successful interviews, not 2,157 x 72.4% response rate. In the past, many media made this mistake.

[3] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified. When quoting these figures, journalists can state “sampling error of percentages not more than +/-6% at 95% confidence level”. Because POP introduced “rim weighting” in 2014, during the transition period, whether changes in various figures are beyond sampling errors are based on tests using the same weighting methods. That is, to test whether the figures of this survey are significantly different from those of the previous survey, both sets of data are rim weighted before testing, instead of using simple computation of the published figures.

[4] Because of sampling errors in conducting the survey(s) and the rounding procedures in processing the data, the figures cannot be too precise, and the totals may not be completely accurate. Therefore, when quoting percentages of the survey(s), journalists should refrain from reporting decimal places, but when quoting the rating figures, one decimal place can be used.

[5] The data of this survey is collected by means of random telephone interviews conducted by real interviewers, not by any interactive voice system (IVS). If a research organization uses “computerized random telephone survey” to camouflage its IVS operation, it should be considered unprofessional.


Latest Figures

Whenever there are large-scale elections, POP would conduct different surveys to track opinion changes. POP recently conducted a pre-election survey with the following contact information. From 2014, POP enhanced the previous simple weighting method based on age and gender distribution to “rim weighting” based on age, gender and education (highest level attended) distribution. The latest figures released today in relation to geographic constituencies have been rim-weighted according to the age-gender distributions of registered voters in each geographic constituency based on the 2016 final register of electors provided by the Government. Herewith the contact information of various surveys:

Date of survey

Overall sample size
(registered voters)

Effective response rate

Maximum sampling error of percentages[6]

30/7-3/8/16 (First stage opinion survey)

1,056

74.6%

+/-3%

8-18/8/16 (Second stage opinion survey)

2,157

72.4%

+/-2%

[6] Errors are calculated at 95% confidence level using full sample size. “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.


Major findings are summarized below, together with parallel figures obtained in the last Legislative Council elections of 2004, 2008 and 2012:

Date of survey

18-22/8/04

1-3/8/08

1-8/8/12

30/7-3/8/16

Latest change

Sample base (registered voters)

1,882

771

1,266

1,056

--

Effective response rate

65.4%

65.7%

69.0%

74.6%

--

Max. sampling error of percentages (at 95% conf. level)

+/-2%

+/-4%

+/-3%

+/-3%

--

Latest finding

Finding

Finding

Finding

Finding & error[7]

--

Could tell the exact date or month of the election

--

69%

75%[8]

69+/-3%

-6%[8]

Inclined to follow instructions to the electorates on the allocation of votes to another list belonging to the same camp

19%

20%

15%[8]

11+/-2%

-4%[8]

Inclined not to follow instructions to the electorates on the allocation of votes to another list belonging to the same camp

56%

64%

72%[8]

76+/-3%

+4%[8]

[7] All error figures in the table are calculated at 95% confidence level. “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified. Media can state “sampling error of percentages not more than +/-3% at 95% confidence level” when quoting the above figures.

[8] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.


One month before this year's election, 69% of the registered voters responded were aware of the coming Legislative Council election in September. With respect to whether they would follow split-vote instructions for strategic voting, 11% inclined to do so, which means a decrease of 4 percentage-points compared to the same time in 2012. On the other hand, 76% inclined not to do so, which means an increase of 4 percentage-points compared to the same time in 2012.

Regarding the voters’ considerations in casting their votes, recent figures on Territory-wide survey are tabulated as follows, in descending order of importance:

Territory-wide survey

Date of survey

10-14/8/04

18-23/8/08

16-20/8/12

8-18/8/16

Latest change

Sample base (registered voters)

1,817

1,996

1,786

2,157

--

Effective response rate

73.9%

69.9%

68.6%

72.4%

--

Max. sampling error of percentages (at 95% conf. level)[9]

+/-2%

+/-2%

+/-2%

+/-2%

--

Latest finding/sampling error

Finding

Finding

Finding

Finding

Error

Regarded candidates’ livelihood policies important[10]

85%

87%

89%[11]

87%

+/-1%

-2%[11]

Regarded candidates’ livelihood policies not important[10]

4%

4%

4%

4%

+/-1%

--

Regarded candidates’ economic policies important[10]

74%

77%[11]

75%

73%

+/-2%

-2%

Regarded candidates’ economic policies not important[10]

6%

6%

8%[11]

9%

+/-1%

+1%

Regarded candidates’ political policies important[10]

65%

60%[11]

64%[11]

66%

+/-2%

+2%

Regarded candidates’ political policies not important[10]

12%

13%

14%

12%

+/-1%

-2%[11]

Regarded candidates’ relations with the CPG important[10]

45%

48%[11]

38%[11]

53%

+/-2%

+15%[11]

Regarded candidates’ relations with the CPG not important[10]

23%

22%

32%[11]

24%

+/-2%

-8%[11]

Regarded candidates’ experiences important[10]

55%

50%[11]

45%[11]

41%

+/-2%

-4%[11]

Regarded candidates’ experiences not important[10]

20%

22%

28%[11]

33%

+/-2%

+5%[11]

Regarded candidates’ fame important[10]

35%

32%[11]

33%

31%

+/-2%

-2%

Regarded candidates’ fame not important[10]

35%

33%

36%[11]

38%

+/-2%

+2%

[9] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.

[10] Collapsed from a 5-point scale.

[11] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.


Regarding the voters’ considerations in casting their votes, the survey on August 8 to 18 showed that, 87% of the voters interviewed claimed they would consider the livelihood policies proposed by candidates, such as problems on housing, education, medical, environment and public order, etc., important. Besides, 73% and 66% said economic policies and political policies proposed by candidates, such as constitutional affairs, democratic development, etc., were important to their voting decision respectively. The corresponding figures for candidates’ relations with the Central Government and their experiences in the Council were 53% and 41%, while that for candidates’ fame was 31%.

Regarding the voters’ considerations in casting their votes in individual geographical constituencies, the relevant survey results revealed that, out of the 6 items we put to test, the livelihood policies proposed by candidates was still the most important consideration factor while their fame was the least. The figures in each constituency are summarized as follows:

Hong Kong Island

Date of survey

10-14/8/04

18-23/8/08

16-20/8/12

8-18/8/16

Latest change

Sample base (registered voters)

323

360

288

424

--

Max. sampling error of percentages (at 95% conf. level)[12]

+/-6%

+/-5%

+/-6%

+/-5%

--

Latest finding/sampling error

Finding

Finding

Finding

Finding

Error

Regarded candidates’ livelihood policies important[13]

86%

89%

91%

87%

+/-3%

-4%[14]

Regarded candidates’ livelihood policies not important[13]

2%

4%

2%

4%

+/-2%

+2%

Regarded candidates’ economic policies important[13]

75%

79%

74%

76%

+/-4%

+2%

Regarded candidates’ economic policies not important[13]

7%

6%

10%

10%

+/-3%

--

Regarded candidates’ political policies important[13]

71%

64%[14]

61%

70%

+/-4%

+9%[14]

Regarded candidates’ political policies not important[13]

7%

13%[14]

16%

12%

+/-3%

-4%[14]

Regarded candidates’ relations with the CPG important[13]

48%

51%

43%[14] [15]

57%

+/-5%

+14%[14]

Regarded candidates’ relations with the CPG not important[13]

22%

21%

29%[14]

24%

+/-4%

-5%[14]

Regarded candidates’ experiences important[13]

52%

50%

43%[14] [15]

42%

+/-5%

-1%

Regarded candidates’ experiences not important[13]

23%

23%

29%[14]

36%

+/-5%

+7%[14]

Regarded candidates’ fame important[13]

39%

34%

33%

36%

+/-5%

+3%

Regarded candidates’ fame not important[13]

33%

32%

35%

35%

+/-5%

--

[12] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.

[13] Collapsed from a 5-point scale.

[14] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.

[15] Based on the figures in 2012, in two decimal places, the percentage of people that Regarded candidates’ relations with the CPG important is 43.04% and that of people that Regarded candidates’ experiences important is 42.66%.


Kowloon West

Date of survey

10-14/8/04

18-23/8/08

16-20/8/12

8-18/8/16

Latest change

Sample base (registered voters)

259

394

264

285

--

Max. sampling error of percentages (at 95% conf. level)[16]

+/-7%

+/-5%

+/-6%

+/-6%

--

Latest finding/sampling error

Finding

Finding

Finding

Finding

Error

Regarded candidates’ livelihood policies important[17]

81%

87%[18]

91%

84%

+/-4%

-7%[18]

Regarded candidates’ livelihood policies not important[17]

4%

5%

4%

5%

+/-3%

+1%

Regarded candidates’ economic policies important[17]

74%

77%

78%

72%

+/-5%

-6%[18]

Regarded candidates’ economic policies not important[17]

6%

7%

7%

8%

+/-3%

+1%

Regarded candidates’ political policies important[17]

66%

61%

63%

67%

+/-6%

+4%

Regarded candidates’ political policies not important[17]

15%

14%

12%

10%

+/-4%

-2%

Regarded candidates’ relations with the CPG important[17]

50%

49%

39%[18]

58%

+/-6%

+19%[18]

Regarded candidates’ relations with the CPG not important[17]

19%

21%

30%[18]

17%

+/-4%

-13%[18]

Regarded candidates’ experiences important[17]

58%

48%[18]

45%

46%

+/-6%

+1%

Regarded candidates’ experiences not important[17]

18%

25%[18]

29%

28%

+/-5%

-1%

Regarded candidates’ fame important[17]

30%

34%

34%

35%

+/-6%

+1%

Regarded candidates’ fame not important[17]

37%

35%

36%

33%

+/-6%

-3%

[16] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.

[17] Collapsed from a 5-point scale.

[18] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.


Kowloon East

Date of survey

10-14/8/04

18-23/8/08

16-20/8/12

8-18/8/16

Latest change

Sample base (registered voters)

310

319

286

320

--

Max. sampling error of percentages (at 95% conf. level)[19]

+/-6%

+/-6%

+/-6%

+/-6%

--

Latest finding/sampling error

Finding

Finding

Finding

Finding

Error

Regarded candidates’ livelihood policies important[20]

84%

87%

88%

86%

+/-4%

-2%

Regarded candidates’ livelihood policies not important[20]

4%

2%

3%

6%

+/-3%

+3%

Regarded candidates’ economic policies important[20]

73%

77%

75%

72%

+/-5%

-3%

Regarded candidates’ economic policies not important[20]

6%

5%

8%

9%

+/-3%

+1%

Regarded candidates’ political policies important[20]

61%

59%

64%

64%

+/-5%

--

Regarded candidates’ political policies not important[20]

11%

12%

13%

14%

+/-4%

+1%

Regarded candidates’ relations with the CPG important[20]

38%

46%[21]

38%[21]

50%

+/-6%

+12%[21]

Regarded candidates’ relations with the CPG not important[20]

25%

22%

32%[21]

25%

+/-5%

-7%[21]

Regarded candidates’ experiences important[20]

49%

52%

52%

45%

+/-6%

-7%

Regarded candidates’ experiences not important[20]

21%

20%

24%

29%

+/-5%

+5%

Regarded candidates’ fame important[20]

37%

31%[21]

37%

30%

+/-5%

-7%[21]

Regarded candidates’ fame not important[20]

33%

29%

30%

36%

+/-5%

+6%

[19] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.

[20] Collapsed from a 5-point scale.

[21] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.


New Territories West

Date of survey

10-14/8/04

18-23/8/08

16-20/8/12

8-18/8/16

Latest change

Sample base (registered voters)

447

461

491

545

--

Max. sampling error of percentages (at 95% conf. level)[22]

+/-5%

+/-5%

+/-5%

+/-4%

--

Latest finding/sampling error

Finding

Finding

Finding

Finding

Error

Regarded candidates’ livelihood policies important[23]

87%

83%

87%[24]

87%

+/-3%

--

Regarded candidates’ livelihood policies not important[23]

3%

5%

5%

3%

+/-2%

-2%

Regarded candidates’ economic policies important[23]

76%

75%

75%

72%

+/-4%

-3%

Regarded candidates’ economic policies not important[23]

5%

6%

5%

8%

+/-2%

+3%[24]

Regarded candidates’ political policies important[23]

63%

56%[24]

66%[24]

64%

+/-4%

-2%

Regarded candidates’ political policies not important[23]

12%

13%

13%

13%

+/-3%

--

Regarded candidates’ relations with the CPG important[23]

43%

44%

38%[24]

48%

+/-4%

+10%[24]

Regarded candidates’ relations with the CPG not important[23]

26%

22%

32%[24]

25%

+/-4%

-7%[24]

Regarded candidates’ experiences important[23]

60%

52%[24]

47%

41%

+/-4%

-6%[24]

Regarded candidates’ experiences not important[23]

17%

21%

27%[24]

35%

+/-4%

+8%[24]

Regarded candidates’ fame important[23]

34%

33%

32%

29%

+/-4%

-3%

Regarded candidates’ fame not important[23]

34%

34%

35%

40%

+/-4%

+5%[24]

[22] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.

[23] Collapsed from a 5-point scale.

[24] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.


New Territories East

Date of survey

10-14/8/04

18-23/8/08

16-20/8/12

8-18/8/16

Latest change

Sample base (registered voters)

477

462

457

583

--

Max. sampling error of percentages (at 95% conf. level)[25]

+/-5%

+/-5%

+/-5%

+/-4%

--

Latest finding/sampling error

Finding

Finding

Finding

Finding

Error

Regarded candidates’ livelihood policies important[26]

85%

89%[27]

91%

88%

+/-3%

-3%

Regarded candidates’ livelihood policies not important[26]

6%

2%[27]

4%

3%

+/-2%

-1%

Regarded candidates’ economic policies important[26]

74%

77%

76%

72%

+/-4%

-4%

Regarded candidates’ economic policies not important[26]

7%

6%

10%[27]

9%

+/-2%

-1%

Regarded candidates’ political policies important[26]

64%

62%

66%

67%

+/-4%

+1%

Regarded candidates’ political policies not important[26]

12%

14%

14%

12%

+/-3%

-2%

Regarded candidates’ relations with the CPG important[26]

46%

49%

35%[27]

53%

+/-4%

+18%[27]

Regarded candidates’ relations with the CPG not important[26]

22%

23%

35%[27]

25%

+/-4%

-10%[27]

Regarded candidates’ experiences important[26]

55%

50%

41%[27]

37%

+/-4%

-4%

Regarded candidates’ experiences not important[26]

19%

23%

32%[27]

33%

+/-4%

+1%

Regarded candidates’ fame important[26]

34%

30%

32%

29%

+/-4%

-3%

Regarded candidates’ fame not important[26]

36%

35%

41%[27]

42%

+/-4%

+1%

[25] “95% confidence level” means that if we were to repeat a certain survey 100 times, using the same questions each time but with different random samples, we would expect 95 times getting a figure within the error margins specified.

[26] Collapsed from a 5-point scale.

[27] Such changes have gone beyond the sampling errors at the 95% confidence level under the same weighting method, meaning that they are statistically significant prima facie. However, whether numerical differences are statistically significant or not is not the same as whether they are practically useful or meaningful.


Commentary

Frank Wai-Kin Lee, Research Manager of POP, observed, “The latest survey shows that after the nomination period for this year’s Legislative Council election, nearly 70% of all registered voters were able to tell us the exact month of the election, namely, September. This is much lower than four years ago. As for split-voting, only about 10% of the voters said they are willing to adopt strategic voting, which is also much lower than that registered four years ago. Besides, in terms of priority, voters’ concerns with respect to the 6 attributes prompted in the survey are candidates’ livelihood policies, economic policies, political issues, relationship with the Central Government, council experience and fame. Among them, voters are giving more weight to candidates’ relationship with the Central Government, and less importance to their council experience and livelihood policies. This probably reflects the changes in Hong Kong’s political ecology in the past 4 years. Regarding different geographical constituencies, voters in all constituencies are giving more importance to candidates’ relationship with the Central Government this year. Voters in Hong Kong Island and Kowloon West show significantly lower concern to candidates’ livelihood policies. Meanwhile, voters in Hong Kong Island show significantly higher concern to candidates’ political policies than 4 years ago, those in Kowloon West show significantly lower concern to candidates’ economic policies, those in Kowloon East show significantly lower concern to candidates’ fame, while those in New Territories West are giving less importance to candidates’ council experience. As for ‘Super DC’, since most voters interviewed belong to this constituency, findings are highly comparable to those of the overall sample.”


Future Release (Tentative)

  • August 30, 2016 (Tuesday) 1pm to 2pm: Popularity of CE and HKSAR Government