Methodology of PSI (2023-07-04)

Basic Concepts

In 2012, HKPORI compiled the “Public Sentiment Index (PSI)” with an aim to quantify Hong Kong people’s sentiments, in order to explain and predict the likelihood of mass movements. PSI comprises 2 components: one being Government Appraisal (GA) Score and the other being Society Appraisal (SA) Score. GA refers to people’s appraisal of society’s governance while SA refers to people’s appraisal of the social environment. PSI comprises 10 public opinion indicators, with data collected since July 1992, meaning over 30 years of accumulated data.

For “Government Appraisal”, there are 4 indicator questions, as follows:

GA1: Please use a scale of 0-100 to rate your extent of support to Governor Chris Patten / Chief Executive (CE) Tung Chee-hwa / CE Donald Tsang / CE Leung Chun-ying / CE Carrie Lam / CE John Lee, with 0 indicating absolutely not supportive, 100 indicating absolutely supportive and 50 indicating half-half. How would you rate the Governor Chris Patten / Chief Executive (CE) Tung Chee-hwa / CE Donald Tsang / CE Leung Chun-ying / CE Carrie Lam / CE John Lee?
GA2: If a general election of the Chief Executive were to be held tomorrow, and you had the right to vote, would you vote for Tung Chee-hwa / Donald Tsang / Leung Chun-ying / Carrie Lam / John Lee?
GA3: Are you satisfied with the performance of the HKSAR government? (Interviewer to probe intensity)
GA4: On the whole, do you trust the Hong Kong/Hong Kong SAR government? (Interviewer to probe intensity)

For “Society Appraisal”, there are these 6 indicator questions:

SA1: Generally speaking, how much are you satisfied or dissatisfied with the current political condition in Hong Kong? (Interviewer to probe intensity)
SA2: Generally speaking, how much are you satisfied or dissatisfied with the current economic condition in Hong Kong? (Interviewer to probe intensity)
SA3: Generally speaking, how much are you satisfied or dissatisfied with the current livelihood condition in Hong Kong? (Interviewer to probe intensity)
SA4-1: Please rate on the scale of 0-10 the importance of political condition in your overall satisfaction with Hong Kong’s societal condition, with 0 meaning absolutely not important, 10 meaning absolutely important, 5 meaning moderately important. How would you rate the importance of political condition?
SA4-2: Please rate on the scale of 0-10 the importance of economic condition in your overall satisfaction with Hong Kong’s societal condition, with 0 meaning absolutely not important, 10 meaning absolutely important, 5 meaning moderately important. How would you rate the importance of economic condition?
SA4-3: Please rate on the scale of 0-10 the importance of livelihood condition in your overall satisfaction with Hong Kong’s societal condition, with 0 meaning absolutely not important, 10 meaning absolutely important, 5 meaning moderately important. How would you rate to the importance of livelihood condition?
Computation Method

Step One is to quantify the data from the 10 questions into numbers using the following method:

GA1 (unstandardized):

Calculate the mean of valid cases for this question, resulting in a number with initial value ranging 0~100.

GA2 (unstandardized):

Subtract the “No” percentage from the “Yes” percentage to obtain the net support value among valid cases for this question, which is a number with initial value ranging -100 ~ +100.

GA3, GA4, SA1, SA2, SA3 (unstandardized) [1]:

Quantify the individual responses into 1, 2, 3, 4, 5 marks according to their degree of positive level, where 1 is the lowest and 5 the highest, and then calculate the means of valid cases for each of these questions, resulting in numbers with initial values each ranging 1~5.

SA4-1, SA4-2, SA4-3 (unstandardized and transformed values):

First calculate the mean value of each question for valid ratings for each of these questions separately, ranging 0~10, then divide each of them by the sum of the three mean values, ranging 0~30, to obtain 3 transformed values each ranging 0~1, with their total sum equal to 1.

[1] Prior to 2012, if the 6 indicators of unstandardized SA score had not been updated, HKPORI would use simple linear regression to extrapolate the unstandardized SA score from the unstandardized GA score of the same time period. Starting from 2013, this method has been replaced by the direct adoption of the most recent announced data instead.

Step Two is to obtain the standardized and final scores from the numbers obtained from the initial quantification process:

GA1, GA2, GA3, GA4, SA1, SA2, SA3 (standardized):

Each of the transformed numbers was standardized according to a scheme derived from previous findings obtained since 1992 up to the month before and transformed to a normal distribution with the mean value set at 100 and standard deviation set at 15, meaning that each number was transformed into another number fitting the normal curve described.

Unstandardized GA:

An unstandardized GA score was calculated by simply taking the mean of the transformed values of GA1, GA2, GA3 and GA4, each fitting the normal curve with mean value set at 100 and standard deviation set at 15.

Final GA:

Unstandardized GA was then standardized according to a scheme derived from previous findings obtained since 1992 up to the month before and transformed to a normal distribution with the mean value set at 100 and standard deviation set at 15, to obtain the final GA score.

Unstandardized SA:

The transformed SA4-1, SA4-2, SA4-3 each ranging 0~1 were used as weights to calculate an unstandardized SA score using this formula:

(Standardized_SA1 × Transformed_SA4-1) + (Standardized_SA2 × Transformed_SA4-2)
+ (Standardized_SA3 × Transformed_SA4-3)

Final SA:

Unstandardized SA was then standardized according to a scheme derived from previous findings obtained since 1992 up to the month before and transformed to a normal distribution with the mean value set at 100 and standard deviation set at 15, to obtain the final SA score.

Final PSI:

An unstandardized PSI score was calculated by simply taking the mean of the final GA and final SA, and then standardized according to a scheme derived from previous findings obtained since 1992 up to the month before and transformed to a normal distribution with the mean value set at 100 and standard deviation set at 15.

Handling of Missing Data and Revision of Computation Method

Since some survey series were not yet started in 1992, those items would be excluded as missing data in that stage, while the value of SA4 was assumed to be one-third. After the commencement of those survey series, if some data was not updated when calculating the indices, their values would be imputed from the most recent data. As for the standardization of various values, for the first generation of PSI, HKPORI basically takes July 1992 as a starting point, and then takes the end date of certain CE’s term of office as the end point to generate the standardization database. The following table briefly explains:

CE and term time Period of PSI calculation Covered period of standardization database Years covered in the database
Chris Patten
(1992-1997)
July 1992 to June 1997[2] July 1992 to June 2012 20 years
Tung Chee-hwa
(1997-2005)
July 1997 to March 2005[2] July 1992 to June 2012 20 years
Donald Tsang
(2005-2012)
June 2005 to June 2012[2] July 1992 to June 2012 20 years
CY Leung
(2012-2017)
July 2012 to June 2017 July 1992 to June 2012 20 years
Carrie Lam
(2017-2022)
July 2017 to June 2022 July 1992 to June 2017 25 years

[2] As the PSI was used only after 2012, the earlier values need to be computed in retrospect.

When it comes to the second generation of PSI, HKPORI still takes July 1992 as a starting point, but will take the first five years of data to generate the standardization database, and then keep it growing month by month. The following table briefly explains:

CE and term time Period of PSI calculation Covered period of standardization database Months covered in the database
Chris Patten
(1992-1997)
July 1992 to June 1997[3] July 1992 to June 1997 60 months
Tung Chee-hwa
(1997-2005)
July 1997[3] July 1992 to June 1997 60 months
August 1997[3]July 1992 to July 1997… 61 months…
Donald Tsang
(2005-2012)
June 2005[3] July 1992 to May 2005 155 months
July 2005[3]July 1992 to June 2005… 156 months…
CY Leung
(2012-2017)
July 2012 July 1992 to June 2012 240 months
August 2012… July 1992 to July 2012… 241 months…
Carrie Lam
(2017-2022)
July 2017 July 1992 to June 2017 300 months
August 2017… July 1992 to July 2017… 301 months…
John Lee
(2022- )
July 2022… July 1992 to June 2022… 360 months…
June 2023 July 1992 to May 2023 371 months

[3] As the PSI was used only after 2012, the earlier values need to be computed in retrospect.

Understanding the Index Values

PSI, GA and SA values are all standardized to a normal distribution with the mean value set at 100 and standard deviation set at 15, similar to that of Intelligence Quotient (IQ), meaning that each number was transformed into another number fitting the normal curve described. The lower the value, the poorer the public sentiment is. The higher the value, the better the public sentiment is, while 100 means normal. Specific values can be interpreted using this table:

Value Percentile Value Percentile
140+ Maximum 1% 60- Minimum 1%
125 Maximum 5% 75 Minimum 5%
120 Maximum 10% 80 Minimum 10%
110 Maximum 25% 90 Minimum 25%
100 being normal level, meaning half above half below
Click to access the login or register cheese
Google Translate »