Politweet.Org

Observing Malaysian Social Media

Archive for the ‘Statistics’ Category

Recent Trends in Political Party Interest on Facebook in Malaysia (Aug 2016)

1. Introduction

This document provides a measurement of the political party interests of Facebook users in Malaysia. This is based on public information collected from Facebook.

Some important notes to remember when interpreting Facebook figures:

  1. Total population refers to Facebook users aged 13 years and above.
  2. Potential voters refer to Facebook users aged 21 years and above.
  3. Youth refers to Facebook users aged 13 – 20 years.
  4. Gender breakdown figures do not add up to the total. This may be due to Facebook users not sharing their gender, and also due to rounding errors by Facebook when dealing with specific age groups. State breakdown figures also do not add up to the total, due to the same rounding errors.
  5. Figures provided by Facebook are estimates. Some inaccuracies are to be expected.
  6. Facebook users residing in Malaysia are not necessarily Malaysian citizens.
  7. Interest in a topic is equal to the number of users expressing interest in a topic.
    1. To measure interest we used a combination of Facebook Interests (a collection of interests, activities, groups, pages, status updates and job history identified by a common term determined by Facebook e.g. ‘United Malays National Organization’) and specific Group and Page names (e.g. Friends of BN).
    2. These are used to collect the number of users interested in a given party/coalition/politician/group. For example, a user mentioning a party name in a status update; sharing a news link related to the party or sharing content from a party-affiliated page would count towards the total interest in that party
    3. Interest in a political party does not indicate support for the party, only awareness
    4. It is assumed that interest in PAS includes interest in AMANAH as PAS leaders migrated to AMANAH
  8. Audience refers to the population of users that express interest in a topic.
  9. Based on our research to date, Pages that are of type ‘politician’ are not always included under related Facebook Topics. For example, not all ‘Tony Pua’ (MP, PJ Utara, DAP) Page likes are included under interest in ‘DAP’. However because Facebook does not make Topic details available we cannot easily determine which politicians, if any, were included.
  10. Statistics on the Opposition primarily refer to component parties of the former Pakatan Rakyat – PKR, PAS and DAP. Interest in PSM is included in total statistics for the Opposition, but is not listed separately due to its small audience.

 

2. Interest in Political Parties on Facebook

The following graph shows the partisanship of interest in political parties by Facebook users in Malaysia aged 21 years and above. Interest in PAS is assumed to include interest in AMANAH because Facebook has not made separate AMANAH figures available yet.

FBPartisanship_Aug15_Aug16

Out of 8.4 million users in Malaysia (aged 21 years and above) that are interested in BN or Opposition parties:

  • 54% are male and 46% are female
  • 3 million are interested in Opposition parties
  • 8 million are interested in BN parties
  • 76% (400 thousand) are exclusively interested in Opposition parties
  • 52% (2.9 million) are interested in a mix of Opposition and BN parties
  • 71% (5.1 million) are exclusively interested in BN parties

As of August 2016 the level of exclusive interest in 60.71% for BN and 4.76% for the Opposition. This is a record high for BN and a record low for the Opposition since we began tracking these statistics in December 2012.

Read the rest of this entry »

Advertisements

Written by politweet

September 1, 2016 at 9:44 am

Analysis of Opinions on Lim Guan Eng’s Corruption Charges by Twitter Users in Malaysia

1. Background

On March 17th 2016, Shabudin Yahaya (BN MP for Tasek Gelugor) alleged that the Chief Minister of Penang, Lim Guan Eng was involved in the sale of two plots of land in Taman Manggis, Penang to a company whose owner was connected to the owner of a bungalow purchased by Lim Guan Eng. The purchase of the bungalow was alleged to have been below the market-price [1].

Comparisons were made between these allegations and former Menteri Besar of Selangor, Khir Toyo’s case where he was convicted for abusing his power to acquire land and property below the market-price.

Following these allegations, a series of exposes and more allegations surfaced online. MACC conducted investigations and on June 29th Lim Guan Eng was arrested and held overnight.  On June 30th he was charged with corruption in the Penang High Court. His former land-lady, Phang Li Koon was also charged for her involvement.

The charges faced by Lim Guan Eng are described below:

“Lim is facing two charges for corruption – one under Section 165 of the Penal Code and another Section 23 of the Malaysian Anti-Corruption Act (MACC) 2009 — over his approval of an application from Magnificent Emblem to convert a piece of land from agricultural to residential use, as well as over his purchase of a house from the firm’s director, Phang, for RM2.8 million, which was below the property’s market value of RM4.27 million.” [2]

Since the story broke in March we tracked mentions of Lim Guan Eng, Taman Manggis, Khir Toyo and other related terms to gauge the response to the initial story and on-going exposes.

2. Initial Analysis (March 17th – April 30th)

We initially examined tweets by 10,627 users from March 17th – April 30th 2016 mentioning the keywords related to allegations against Lim Guan Eng.

What we found was the topic was popular mainly with users with a strong partisan interest in Malaysian politics. This issue did not draw enough interest from the general public – it was not worth talking about, and those who did tended to express disinterest or only retweet news articles.

The topic also drew more interest from users based in Kuala Lumpur, Selangor and Penang. 59% of users tweeting about the topic (not including retweets) were based in these 3 states. The highest drop in interest was from users in Johor, which made up 8.63% of the local population (1.96 points lower than the proportional average).

There was also a high degree of spammed tweets, with spammed tweets outnumbering non-spammed tweets on some days. This can be seen in the chart below:

lgearrest_mac_interest

1,358 users spammed 49,223 tweets. In other words, 12.8% of the users spammed 41.8% of the total tweets.

From a manual reading of non-spammed tweets during this period, we found that tweeted opinions about the scandal fell mainly into the following categories:

  • Users not interested in Lim Guan Eng’s scandal
  • Users complaining about excessive media coverage. Most complaints implied users were bored or not interested in listening to the repeated allegations.
  • Users wanting Lim Guan Eng to be investigated
  • Users comparing Lim Guan Eng’s case with Khir Toyo’s case
  • Users criticising Lim Guan Eng’s responses to the allegations
  • Users critical of BN and DAP, equating both to be corrupt
  • Users defending Lim Guan Eng. Among the more popular reasons were:
    • BN / UMNO / PM Najib are considered to be worse
    • The discount isn’t that big / there is nothing wrong with a good deal
    • The 1MDB scandal is much bigger and more important than Lim Guan Eng’s scandal
    • Khir Toyo’s house is bigger

Users defending Lim Guan Eng were a small minority. There was little evidence of pro-DAP or pro-Opposition users being mobilised to defend Lim Guan Eng.

Out of 44 DAP politicians actively tweeting in this period, only 27 politicians tweeted/retweeted tweets mentioning Lim Guan Eng or keywords related to the allegations. This does not include images or tweets not mentioning related keywords. By not talking about the allegations the 17 politicians missed an opportunity to contribute to Lim Guan Eng’s defence on Twitter.

Because of the low level of interest from the general public and the high degree of spam, we could not do a detailed opinion analysis at the time.

3. Analysis of Opinions on Lim Guan Eng’s Corruption Charges (June 29th – July 6th)

We examined tweets by 8,365 users from June 29th – July 6th 2016 mentioning keywords related to allegations against Lim Guan Eng. The daily interest is shown in the graph below.

lgearrest_jun_interest

We then performed opinion-based analysis on 520 users based in Kuala Lumpur, Selangor and Penang. The margin of error is +/- 4.3%.

Read the rest of this entry »

How Barisan Nasional and Pakatan Rakyat Performed With Voters in Sarawak (GE13)

1. Background

Prior to the 13th General Election (GE13) we came up with a methodology of predicting election results based on voting patterns in previous elections.

Our method relied on mapping polling lane results to individual voters. This process assigned probability values (chance of turnout; chance of voting for each coalition) to the voter that was not affected if they migrated to another constituency. This is important because between GE12 and GE13 527,849 voters migrated to different constituencies.

The impact of voter migration cannot be measured for a single seat just by comparing results of GE12 and GE13 for that seat. An analysis of the whole country needs to be performed. New voter registrations, voters passing away and voters no longer eligible to vote are other factors that require deep analysis.

After GE13 we were able to apply the same estimation method to voters based on GE13 results. By comparing the shift in probabilities we are able to calculate the swing in support for each coalition. Because we base our calculations on individual voters, we are able to calculate shifts in support based on combinations of the following dimensions:

  • By Age
  • By Race
  • By Gender
  • By Urban Development Category (rural / semi-urban / urban)
  • By Parliament/State Assembly Seat
  • By Polling District
  • By Locality
  • By Seats Won by Specific Parties

Any voter whose level of support cannot be determined is assigned a probability of 50% and categorised as a fence-sitter. The most reliable metric is age because voters are separated into polling lanes based on age. Additionally we have also categorised the 222 Parliament constituencies as rural, semi-urban or urban based on satellite imagery. The descriptions of each category are:

Rural = villages (kampungs) / small towns / farmland distributed within the seat. Rural seats tend to be physically large with a low population.

Semi-urban = larger towns and/or numerous small towns, may include villages as well

Urban = cities where a majority of the seat is covered by some form of urban development

For this report we will focus on how Pakatan Rakyat (PR) and Barisan Nasional (BN) performed with regular voters (pengundi biasa) in Sarawak. 31 of the total 222 Parliament seats are in Sarawak.

Our analysis will focus on Malay, Chinese and Bumiputera Sarawak voters. Other ethnic groups such as Indians, Orang Asli and Bumiputera Sabah voters will be counted under the ‘Others’ category unless otherwise specified. This is due to their low numbers within the electorate and the lack of detail within the National Census data.

Postal and early voters are not part of this analysis, other than the section on polling lanes. Postal voters need to be analysed separately due to their different voting process and difficulties in campaigning to both groups.

The predicted support for PR based on GE12 was estimated to be low. This is because in GE12 the PR component parties did not contest all seats. SNAP and Independents contested BN in some seats with no PR candidates. There were also seats that were won by BN uncontested. PR was effectively untested in Sarawak.

We tested analysis using SNAP and Independent results as ‘pro-Opposition’ in place of PR. However this approach made little impact on the analysis. A vote for SNAP or Independents also cannot be assumed as a vote for PR. To keep analysis consistent with a ‘BN versus PR’ perspective we did not treat SNAP and Independent candidate results as PR results.

We will also present analysis of seats at the state (DUN) level based on individual voting at the Parliament level. It is not as accurate as performing analysis based on state-level results but it should be applicable for constituencies where voters voted for the same coalition (BN / PR) for both state and Parliament.

Please remember that unless otherwise stated, all statistics in this analysis refer to regular voters in Sarawak only.

Read the rest of this entry »

Written by politweet

April 16, 2016 at 9:12 pm

Facebook Census of Political Interest in Malaysia, August 2015

1. Introduction

This document provides a measurement of the political interests of Facebook users in Malaysia and a brief analysis of recent trends. This is based on public information collected from Facebook. Characteristics and interests were chosen based on what would be most relevant to political analysts.

Statistics cover users aged 21 years and older unless otherwise specified.

Some important notes to remember when interpreting figures:

  1. Total population refers to Facebook users aged 13 years and above.
  2. Potential voters refer to Facebook users aged 21 years and above.
  3. Youth refers to Facebook users aged 13 – 20 years.
  4. Gender breakdown figures do not add up to the total. This may be due to Facebook users not sharing their gender, and also due to rounding errors by Facebook when dealing with specific age groups.
  5. Figures provided by Facebook are estimates. Some inaccuracies are to be expected.
  6. Facebook users residing in Malaysia are not necessarily Malaysian citizens.
  7. Interest in a topic is equal to the number of users expressing interest in a topic.
    1. To measure interest we used a combination of Facebook Topics (a collection of interests, activities, groups, pages, status updates and job history identified by a common term determined by Facebook e.g. ‘United Malays National Organization’) and specific Group and Page names (e.g. Friends of BN).
    2. These are used to collect the number of users interested in a given party/coalition/politician/group. For example, a user mentioning a party name in a status update; sharing a news link related to the party or sharing content from a party-affiliated page would count towards the total interest in that party
    3. Interest in a political party does not indicate support for the party, only awareness
  8. Audience refers to the population of users that express interest in a topic.
  9. Based on our research to date, Pages that are of type ‘politician’ are not always included under related Facebook Topics. For example, not all ‘Tony Pua’ (MP, PJ Utara, DAP) Page likes are included under interest in ‘DAP’. However because Facebook does not make Topic details available we cannot easily determine which politicians, if any, were included.
  10. Statistics on the Opposition primarily refer to component parties of the former Pakatan Rakyat – PKR, PAS and DAP. Interest in PSM is included in total statistics for the Opposition, but is not listed separately due to its small audience.

Read the rest of this entry »

Written by politweet

October 8, 2015 at 4:39 am

Posted in Census, Statistics

Tagged with , , , , ,

Twitter Statistics and Crowd Measurement of the Bersih 4 Rally

1. Background

On July 29th Bersih 2.0 announced that a rally entitlted ‘Bersih 4’ would be held on the streets of Kuala Lumpur, Kuching and Kota Kinabalu from August 29th, 2pm to August 30th. The demands of the rally are for Prime Minister Najib Razak to step down and the following institutional reforms to be implemented:

  1. Clean Elections
  2. Clean Governments
  3. Saving Malaysia’s Economy
  4. Right to Dissent
  5. Strengthening Parliamentary Democracy (added on August 14th)

On August 14th Bersih released a statement adding a demand for a transitional government to be formed after Najib’s resignation. This government would need to implement 10 institutional reforms within the next 18 months to ensure the next General Election would be conducted in a clean, free and fair manner:

  1. Reform of electoral system and process
  2. Reform of the Election Commission (EC)
  3. Separation of Prime Minister and Finance Minister
  4. Parliamentary Reform
  5. Separation of the functions of Attorney General and Director of Public Prosecution
  6. Reform of the MACC
  7. Freedom of Information laws
  8. Asset declaration by Ministers and senior state officials
  9. Abolishment of/Amendment to draconian laws
  10. Establishment of the Independent Police Complaints and Misconduct Commission (IPCMC)

On August 27th Opposition MPs from PKR, DAP and GHB released a joint statement declaring they would work with BN MPs to form a new government provided that Anwar Ibrahim and other prisoners of conscience be released; and political reforms be the core agenda of the new government.

2. Twitter Statistics

All statistics referring to tweets include retweets unless otherwise stated.

2.1 Basic Stats

111,879 users made 583,338 tweets about Bersih from July 28th – August 30th 2015.

96,890 users made 446,967 tweets about Bersih during the rally period (August 29th – August 30th 2015). In other words, 86.6% of the total users and 76.6% of total tweets were made during the rally.

The chart below shows a comparison in the number of users tweeting about the Bersih 2 (July 2011), Bersih 3 (April 2012) and Bersih 4 (August 2015) rallies in the days leading up to the event.

Read the rest of this entry »

Written by politweet

September 2, 2015 at 9:44 pm

Teluk Intan Social Media Stats

As a follow-up to our previous analysis on voter sentiment, we have collected social media statistics that are relevant to the current by-election in P76.Teluk Intan. The Democratic Action Party (DAP) under Pakatan Rakyat (PR) and GERAKAN under Barisan Nasional (BN) are contesting in the by-election, so our analysis will focus on these parties.

From what we have so far it is clear that social media usage will not have a big impact on the by-election results. Popularity on Facebook or Twitter are not going to be an indication of which party wins.
Read the rest of this entry »

Urban Development Categorisation of Parliament Seats in Malaysia

The list of Parliament seats in Malaysia by urban development category can be found here.

For further reading on the methodology please read this blog post.

The categories are:

  • Rural = villages (kampungs) / small towns / farmland distributed within the seat. Rural seats tend to be physically large with a low population.
  • Semi-urban = larger towns and/or numerous small towns, may include villages as well
  • Urban = cities where a majority of the seat is covered by some form of urban development

This classification was done by us based on Google Maps satellite imagery and SPR maps. This is not the same as SPR’s own internal classification of seats. You can find our reference maps here (link).

We have obtained a gridded map of Malaysia that defines persons per square kilometre, based on a grid cell size of 19 square kilometres. This population estimate is based on an extrapolation of the 2000 National Census. This map will enable us to :

  • Estimate the size of the constituency
  • Estimate the min, max and average population density
  • Define a range of rural, semi-urban and urban categories, instead of just three
  • Perform seat classification at the State Assembly Seat (DUN) level

However this is time consuming and will likely take months to prepare. A sample of the Peninsular Malaysia map is shown below, rendered as a heatmap ranging from Green (low population density) -> Yellow -> Orange -> Red -> Purple (highest population density). The SPR maps are overlaid over it, and by toggling visibility we can identify which cells belong to which constituency. For area calculation we can subdivide the 19 square kilometre grid into 4.75 or 1.1875 square kilometre grids, depending on how small the related constituency is.

GriddedMsia_SPR1

Written by politweet

May 30, 2013 at 1:57 pm