Poll statistics regarding Taiwan's independence vs. unification

For readers convenience, this is being posted backdated here, it was originally cross-posted at the Taiwan Matters blog on December 14, 2009.

Another look at the Global Views survey of October 2009

For several weeks, I have owed my readers the second part of my analysis of the recent Global Views Magazine opinion poll. The poll asked the usual questions about the attitudes of Taiwanese towards "independence versus unification." I've finally gotten around to putting this analysis into a presentable format.

At the bottom of the poll is a description of the methods used [sic throughout]:

This survey was conducted by GVSRC from 6.20 p.m. to 10.00 p.m. from Oct. 14-16, 2009. It was conducted with random-digit-dial sampling and computer-assisted telephone interviewing methods. 1006 Taiwanese people who are 20 or older completed the interview. One can say with 95% confidence level that the theoretical margin of sampling error is ±3.1%. Gender, living areas, ages, educational level and other features of the interviewees have undergone weighting procedure and test of the sample's representativeness in the survey results. The funding of the survey came from the Global Views Monthly.

I wasn't sure if each person interviewed had gone through the whole set of questions, or only a portion of the questions. If each person had completed the entire questionnaire via telephone interview (which seemed to be the case), I wondered why there were some inconsistent outcomes within their own survey.

I don't want to blame the pollster for this or to suspect data manipulation right away because sometimes people are tired of answering the questions at the end of a telephone interview, and they may provide inconsistent answers. However, from my experience, I can conclude that the survey gave inconsistent results -- assuming there was no data and/or sampling manipulation at all such as removing samples which gave unfavorable results or including additional samples (from the original randomly selected samples) which gave favorable results.

The abbreviations used here will be the same as those in my earlier analysis:

SD: Maintain the status-quo now, decide later
SF: Maintain the status-quo forever
SI: Maintain the status-quo now, declare independence later
SU: Maintain the status-quo now, proceed with unification later
II: Independence immediately (or as soon as possible)
UI: Unification immediately (or as soon as possible)
NA: No answer or don't know
TT: Total percentage who responded

The time when the poll was taken will be denoted with MM/YYYY

M: Mainland [sic] Affairs Council (MAC) statistics
G: Global Views Magazine statistics
AV: Average statistics for the year-to-date

The percentage sign is omitted throughout.

	2007	2008	2009-AV-M	09/2009-M	10/2009-G
SD	39.90	38.750	35.20	35.4	40.7
SF	19.25	23.350	27.50	28.0	11.0
SI	14.15	15.875	15.00	14.9	10.3
SU	10.70	8.050	8.15	8.7	4.3
II	8.80	8.850	5.85	5.0	19.0
UI	2.70	1.750	1.45	1.7	4.0
TT				93.7	89.3

Global Views Magazine conducted an earlier poll in May 2009 (refer to page 5 of the linked PDF file) which tells us:

SD = 44.9%, SF = 11.5%, SI + II (as in that survey it was not distinguished between the two) = 25.4%, and SU + UI = 8.3%, a total response rate of 90.1%.

The first thing I did was to add up the figures to reveal the total percentage of those who responded to find out the percentage who refused to respond or didn't know the answer. In the October 2009 Global Views survey, there was a total of 10.7% who either refused to answer or didn't know the answer, whereas the figure for the MAC survey was 6.3%. Generally speaking, a greater response rate gives more accurate poll results.

The next thing to do was to compare past trends with present figures. The SI of 10.3% is low compared to past trends as well as to the MAC result of 14.9% in September 2009, this coupled with the UI of 4% -- which is unusually high, as I mentioned in my previous post that this figure had never been as high before (only 3 out of 28 times had this figure ever exceeded 3%) -- made me suspect that maybe there was some attempt here to give an overall false impression (since most people will not compare the October 2009 survey with past trends) that people are shifting from the "status quo first then independent later" to the "unification immediately" group. But read on, and you will find out why this is not true when compared with the statistics of past MAC surveys.

The next task was to compare Global Views' results from the above chart with the different questions within the same Global Views poll.

The first question was whether one would agree (A) or disagree (D) with eventual unification with China, and here are the results (from page 3 of the link at the top of this post):

	02/2006	09/2008	05/2009	10/2009
A	28.7	19.5	16.2	15.7
D	54.5	67.5	69.0	69.0
TT	83.2	87.0	85.2	84.7

Notice that the people who want independence (SI + II) would answer "disagree" to this question while the people who would like to maintain the status quo forever (SF) would either answer "disagree" or would not give an answer to the above question. So, the percentage of "disagree" here contains (SI + II + some SF), and the "no answer" figure actually contains (NA + some remaining SF).

The second question in this survey asked whether one would agree or disagree with Taiwan's eventual (formal) independence. The results were as follows:

	02/2006	09/2008	05/2009	10/2009
A	44.3	50.6	48.5	47.2
D	40.3	34.1	35.0	34.1
TT	84.6	84.7	83.5	81.3

Notice that the people who want unification (SU + UI) would answer "disagree" to this question while the people who would like to maintain the status quo forever would answer either disagree or would simply not give an answer to this question. Therefore, the percentage of "disagree" here contains (SU + UI + some SF), and the "no answer" would contain (NA + some other SF).

Notice also that the total percentage of respondents to the second question dropped slightly -- from 84.7% to 81.3%. One explanation for this could be that some respondents thought that once they had provided their answer to the first question, it would not be necessary to provide their answer to the second question, viewing the answer to the second one as obvious or redundant once they had answered the first question. The immediate effect of this is that the second question got a higher non-response rate.

Since the same people (samples) had answered both questions plus the question contained in the chart at the very beginning, we can analyze further.

If we take the percentage who disagree with eventual unification (69%) minus the percentage who agree with the eventual independence (47.2%), it will give us the percentage of people who want SF (status quo forever) since the same people (samples) had answered both questions:

69 - 47.2 = 21.8%

i.e. (SI + II + some SF) – (SI + II) = some SF, and the remaining SF are hidden among the non-responses (NA)

The SF should therefore be greater than 21.8%

On the other hand, if we take the percentage who disagree with eventual independence (34.1%) minus the percentage who agree with eventual unification (15.7%), it will also give us the percentage of people who want SF (status quo forever) for the same reason stated above:

34.1 – 15.7 = 18.4%

i.e. (SU + UI + some SF) – (SU + UI) = some SF, and the remaining SF are hidden in the non-responses (NA)

The SF should therefore be greater than 18.4%.

In either situation, the SF (those who wish to maintain the status quo forever) cannot be less than 18.4%, but the chart at the very beginning has the SF at 11%.

This shows us that the survey actually gives three widely-differing figures for SF: 11%, >21.8%, and >18.4%. It is quite inconsistent to have three different figures in one survey obtained from the same samples. Could it be because this figure was intentionally kept low by the Global Views pollster in order to give a higher percentage to the SD (thereby creating the false impression that many people want to keep the status quo now and decide later, including the option of unification), but in reality the 47.2% (assuming this percentage is not an underestimate) itself tells us that the SI + II = 10.3 + 19 = 29.3% from the first chart is an underestimate of people who want eventual independence. The rise from 29.3% (the chart at the very beginning) to 47.2% (the second question) is a difference of 17.9%, hidden in the SD.

The SU + UI = 4.3 + 4 = 8.3% from the first chart compared to the MAC result of 10.4% in September and the earlier trend suggests that there is a slight drop in the percentage of people supporting eventual unification (Could recent cross-strait exchanges have made some pan-blue supporters want to maintain the status quo forever?), but when given no such choice of maintaining the status quo, the percentage of people who want eventual unification rises from 8.3% to 15.7%, the difference in the rise is 7.4%, hidden in the SD.

Comparing the 17.9% (two paragraphs above) with that 7.4%, one can see that the increase in support for independence is more than twofold that of the increase in support for unification (hidden in the SD).

If we add the percentage of people who agree with eventual unification (15.7%) and the percentage who agree with eventual independence (47.2%), it is 62.9%, and if a person neither agrees with unification nor agrees with independence, the person must be SF or NA, meaning 37.1% (100 – 62.9 = 37.1) of the people who either want to maintain the status quo forever or who don't know or refuse to answer. Since the percentage of people who either don't know or refuse to answer does not usually exceed 15% (as a matter of fact, under normal circumstance and from past trend it ranges from about 7% to 12%), that leaves the percentage of people who want to maintain the status quo forever to be at minimum 22.1% (37.1% - 15% = 22.1%). Comparing this figure to the first chart of SF 11%, the figure of 11% is clearly an underestimate. This conservative figure of 22.1% can be compared with the MAC figure of 28% in September 2009 or with past trends ranging from 19.25% in 2007 but increasing yearly. This group (possibly wanting to avoid war with China) has the biggest percentage increase and the implication has been discussed in my previous post on this subject.

Finally, let's look at yet another question on the same survey (also from page 3 of the survey linked at the very top of this post). If the two sides [Taiwan and China] come to have similar conditions, 68.3% still deemed it unnecessary to unify with China while only 11.7% said it would be OK to be unified. And the total response rate for this question is only, 68.3 + 11.7 = 80%. It's interesting that the response rate nicely and coincidently adds up to the round figure of 80% and that the response rate is even lower than that of the second question above, which was 81.3%. This 68.3% figure may likely have suffered an underestimate because the response rate was limited to 80% when in reality the actual response rate was probably higher. How could anyone explain this 11.7% (yet another inconsistency) in light of the SU + UI = [only] 8.3% in the first chart?

To sum it up, there is no consistency in the figures in this survey.

The trend can only be observed when we compare the statistics over a number of years. There is no actual increase in the percentage of people who want eventual unification -- as a matter of fact, that number is dropping. And most people who answered that they wish to maintain the status quo now and decide later (SD) are actually people who want to choose independence later.

References:
* My previous post on Talk Taiwan: "Statistics on Taiwan's Independence or Unification opinion polls - Part 1"
* American Association for Public Opinion Research: "Do Response Rates Matter?"
* Oxford University Press' Public Opinion Quarterly: "A Comparison of Address-Based Sampling (ABS) Versus Random-Digit Dialing (RDD) for General Population Surveys"

(This post was edited by Tim Maddog.)