Sentiment Analysis on Selangor Menteri Besar Crisis

We performed opinion-based sentiment analysis on 2,062 users who tweeted about the Selangor Menteri Besar crisis from August 9th – August 15th 2014. These users were randomly selected based on tweets we collected since the crisis began. Spammers, news agencies and accounts with automated tweets were not included in the sample.

Our goal initially was to gauge public opinion primarily on whether to keep or remove Tan Sri Khalid Ibrahim. However mixed opinions emerged that allowed us to gauge responses to other parties involved e.g. Pakatan Rakyat (PR), Wan Azizah and Anwar Ibrahim. We were also able to gauge support for the State Assembly to be dissolved and state elections called.

Himpunan Kebangkitan Rakyat #KL112 Geo-located Sentiment Analysis

Himpunan Kebangkitan Rakyat (People’s Uprising Rally / #KL112) was a protest rally held at Stadium Merdeka, Kuala Lumpur on January 12th, 2012. It was organised by Pakatan Rakyat (PR) and various NGOs.

At its peak the crowd is estimated to have been between 63,976 – 78,193 people. For details on how this figure was calculated please read this post.

The purpose of this report is to show:

• Where Twitter users in Malaysia are concentrated
• How many of them tweeted about the event
• Where positive / neutral / negative sentiment was concentrated
• Which areas were devoid of users tweeting about the event

We collected 103,343 tweets from 25,901 users tweeting about the event. A detailed analysis of that along with a comparison with Bersih 2 and Bersih 3 will come in a future report.

In addition to collecting tweets about the event, we also collected geo-located tweets from Malaysia. This was to show areas where Twitter users were concentrated, and visualise the percentage of them tweeting about the event. From our experience 5% -10% of Twitter users use geo-located tweets.

We ran an opinion-based sentiment analysis on users tweeting about the event, using the same methodology previously used on tweets regarding free education and the abolishment of PTPTN. The methodology is:

Sentiment Analysis on the Abolish PTPTN movement

Sentiment analysis was conducted on 31,639 tweets by 5,627 users regarding the abolish PTPTN movement (#mansuhPTPTN) from 5th – 21st April 2012. The majority of those who opposed abolishment cited the ethical issue of borrowing money and not repaying. An estimated 280 users (5% of total) who did not support the cause, were willing to support reducing or writing-off the interest.

Sentiment (users)

Support: 723 (12.85%)
No Opinion: 1,000 (17.78%)
Oppose: 3,904 (69.3%)

Reasons for supporting

1. Most gave no reason for their support, other than ‘Why not?’
2. If the govt can afford PKFZ, NFC scandal, it can afford this
3. High cost of living makes repayment difficult
4. PKR/PR supporters
5. Anti-UMNO/Anti-BN sentiment
6. Problems with converting the loan to scholarship; blacklisting too harsh; interest rate too high

Reasons for opposing

1. Taking a loan carries the moral obligation to repay it
2. There is a sense of pride in taking a loan and working to settle it. Write-off insults those who worked hard.
3. The loan is converted to a scholarship for students who get good grades. Free education must be earned.
4. Taxpayers will have to pay for it, and taxes may increase in future to support free education
5. Why PTPTN and not other education loans too?
6. It does not solve the problem of lack of skills,job opportunities and poor grades
7. Poor competition if education is free.
8. Taxpayers should not carry the burden of student living allowance, which makes up 50-80% of the loan
9. Living allowance is spent on personal luxuries such as gadgets, motorbike,etc.
10. Low interest rate and long repayment period is a good deal
11. Country is not rich enough to support free education
12. PTPTN is currently being abused as it can be taken together with other loans
13. Bad behaviour by students – using the grave marker (batu nisan), funeral wreath on Najib Razak and Khaled Nordin’s photo was very offensive. Chants of ‘Reformasi’,’Najib-Altantuya’ and ‘Musnah Rosmah’ were seen as political and irrelevant to the cause
14. Small number of student activists at Dataran Merdeka show their movement does not represent the majority
15. PTPTN should be restructured instead of abolished, perhaps with alternate payment plans or change in interest

Methodology for sentiment analysis

Because Malay tweets have alternate spellings for most words, a phrase-based approach (where phrases/tweets are assigned negative/positive values)was not practical. Instead, timelines for each user containing keywords related to PTPTN were read and analysed to determine a user’s stand on the issue.

This proved to be more accurate than an automated approach because quite often a user would change his/her opinion after learning more about the issue.

Users who were retweeting very strong opinions on the issue, while offering no opinion of their own, where considered to have that opinion.

In cases where a user would tweet/retweet conflicting opinion without making their own opinion clear, the user would be categorised as having ‘No opinion’. The same applies to users who opposed the issue because they thought it was about ending scholarships, or writing-off the loan for debtors.

Analysis

Users tweeting about PTPTN mostly tweeted in Malay (83.5% of tweets), and 46.3% are apolitical and do not follow politicians on Twitter. This implies that the opinions expressed are mostly by young Malays, most of whom are not interested in politics (at least on Twitter).

The initial negative feedback was due to the student activist’s focus on abolishment of PTPTN and writing-off the loan. Free education was not a widespread message even by the supporters. Some users misunderstood the issue to be an end to all scholarships, a misunderstanding that was played up by BN supporters. Users who had this misunderstanding were counted as having ‘No opinion’.

What was surprising was asking for a write-off was seen by the majority as insulting. Many users took this personally, and followed-up with an expression of pride in their ability to convert the loan due to good grades from hard work or getting a job and repaying. Some users stated that even if PTPTN were to be abolished, the loan should be repaid. Religious arguments were used as well, with expressions that demanding others to suffer (tax payers) for what you enjoyed (education) is a sin.

When supporters countered this by bringing up the govt. settlement of suits between GLCs and Tajuddin Ramli, this was not effective. Using one wrong to justify another wrong was not a good approach. The suggestion to repay the amount paid by those who settled their loans was also offensive.

It was as though the student activsts lost all credibility the moment they asked for their loan to be written-off. This potentially closed people’s minds to any future argument regarding free education or any payment plan that came after. As some users later remarked, they did not understand what the students were fighting for as free education and abolishment PTPTN were seen by many as separate issues.

The relatively small number of student activists involved in the #mansuhPTPTN march on Saturday, 14th April 2012 reflect the small number of supporters on Twitter.

When looking at the partisanship (stats to come in a future graph), 49% of opposers are apolitical, whereas 32% of supporters are apolitical. There is greater political influence on supporters from both PR and BN. Another interesting fact is that for users who exclusively follow PR politicians, 108 oppose and 107 support. Even among partisan PR supporters there is lack of agreement on this issue.

Based on the response in the real world and the Twitter, it can be said that the move to propose abolishing PTPTN was not popular. However writing off the interest or reducing it would win a lot of support from existing PTPTN debtors.

Ahmed Kamal
25th April 2012

Summary of data used
====================

Sentiment (users)

Support: 723 (12.85%)
No Opinion: 1,000 (17.78%)
Oppose: 3,904 (69.3%)

Following politicians (users)

Political: 3,024 (53.74%)
Apolitical: 2,603 (46.26%)

Language (tweets)

Malay: 26,422 (83.51%)
English: 5,198 (16.43%)
Chinese: 19 (0.06%)