How Facebook Has Become A Political Battleground In Bangladesh

View Authors

Sheikh Waheed Baksh
Dhara Mungra
Nazmul Ahasan
Veer Pariawala
Raghav Jain
Aditya Surve
Pratyay Banerjee
Subinoy Mustofi Eron
Swapneel Mehta
Sabhanaz Rashid Diya

We analyze nearly half a million posts shared across 14.4 million followers, resulting in 97.5 million interactions, to understand how political parties leverage interconnected networks to disseminate their messages in preparation of the country’s national election. This is the first of a 3-part series on Bangladesh’s information ecosystem.

With over 44 million active monthly users in the country, Facebook (owned by Meta) has emerged as a vital battleground for political parties to win over public approval and mobilize public opinion ahead of Bangladesh’s parliamentary election in January 2024. Drawing from publicly accessible data, our investigation explores how the two leading political parties, the Bangladesh Awami League (the ruling party) and the Bangladesh Nationalist Party (the largest opposition to the ruling party), are using Facebook to scale their political messaging organically ahead of the election.

Although there is anecdotal evidence of how a political party in Bangladesh is using social media platforms, there are no comprehensive studies at present that systematically analyze how parties organize themselves online. Similar to prior work in different geopolitical contexts¹²³ , we investigate the behaviors of over 600 partisan Facebook Pages and Groups likely affiliated with the Bangladesh Awami League and the Bangladesh Nationalist Party. We gather nearly 500,000 posts with over 14.4 million followers and 97.5 million engagements, focusing on their interconnected networks, tactics used to distribute messages, and the differential approach in establishing their political agenda. We filtered this from a much larger dataset by setting a 500-follower threshold on the minimum reach of Pages and Groups to qualify for in-depth analysis.

Nothing Exists In a Vacuum

To better understand how and why the two leading parties organize themselves online in specific ways, it is imperative to first recognize the larger domestic and geopolitical contexts within which they exist and their significance in the upcoming election. The Bangladesh Awami League (Awami League) and the Bangladesh Nationalist Party (BNP) have dominated the country’s politics since the early 90s when mass public protests helped end more-than-a-decade-long military rule. The two parties alternated as the ruling and main opposition parties until 2007 when a military-backed technocratic government seized power in response to political violence and went on to rule the country for over a year. In 2009, the Awami League returned to power through an election considered credible by local and international observers. The party has been in power ever since.

The BNP reportedly faces significant obstacles as an opposition party, with many of its top leaders ailing, in exile, or imprisoned. According to a recent report in The New York Times, about half the party’s estimated five million activists are embroiled in numerous court cases.

In Bangladesh’s polarized society, both parties maintain a degree of organic popular support, but it is impossible to determine to what extent. The last two elections were contentious: in 2014, the opposition boycotted the election, allowing the ruling coalition to win mostly uncontested, while in 2018, the ruling party and its allies clinched more than 85% of the parliamentary seats. There are no notable public opinion polls that have definitively measured the extent of support enjoyed by political parties.

Recent estimates suggest that a large share of the 100+ million Internet users in Bangladesh also use social media with Facebook in the lead (followed by TikTok and YouTube). Both major political parties have sought to tap into this vast user base in recent years.

For the BNP, social media offers an alternative mode of communication and political outreach as it faces severe restrictions offline. Although the BNP formally launched its official, verified Facebook handle, bearing the party’s full name, in 2019—years after the Awami League did so in 2013—the opposition party already had a sizable presence due to unaffiliated influencers with large followings.

Activists of the ruling party are known to possess sophisticated resources, which include support from influential public voices. In 2021, the party’s official think-tank reportedly aimed to train 100,000 ‘cyber warriors.’ In this election cycle, its campaign seeks to reach 20 million active Facebook users. The party’s electoral campaigns often feature endorsements from prominent celebrities and public figures.

Supporters of both political parties tend to use Facebook Groups, among other mediums, to organize and mobilize support. Facebook Pages are used to engage the public and influence opinion. In our analysis of Facebook Groups and Pages featuring variations of the Awami League and BNP names—resulting in a total of 660 Pages and Groups in our final dataset—we sought to uncover the patterns, engagement, and content behaviors of their members and supporters — one of the first reports of its kind in the context of Bangladesh, to our knowledge.

The Building Blocks

Our data consists of a randomized set of top 260 Pages and 70 Groups bearing the name of each party, respectively. Our analysis focused on Pages and Groups that were active during the period spanning from December 1, 2022, to November 30, 2023. We define ‘activity’ to include at least one piece of content shared every month during the timeframe of the study.

We find that the creation of new Pages related to the Awami League and BNP surged in 2013 and 2014, years that were among the most tumultuous in Bangladesh’s recent history.

The year 2013 marked the rise of the Shahbagh Movement, a popular uprising demanding death penalties for Islamist politicians accused of collaborating with the Pakistani military during Bangladesh’s Liberation War in 1971. This movement, which preceded an online cultural war, was closely aligned with the political and cultural objectives of the Awami League. In 2014, a reactionary backlash supported by the BNP emerged against the Shahbagh Movement, sweeping across the country. During this period, the latter also led separate protests advocating for the restoration of the caretaker government system.

The 2013-2014 period was, therefore, marked by intense online wars across party lines, which likely contributed to the rise in Page creations by supporters of both parties.

Additionally, there was a noticeable increase in the creation of ‘BNP’ Pages in 2017 and 2022, the years leading up to the respective elections. The ‘Awami League’ Pages saw its biggest jump in 2020, coinciding with the COVID-19 pandemic.

We zoom in on the top 20 Pages of the Awami League and the BNP respectively using a combination of the number of followers (>=500) and the highest volume of content shared on a monthly activity (highest activity). We find that despite there being 135 highly active Pages throughout this past year bearing the party’s name, only two Pages of the Awami League are most prominent and responsible for driving the bulk of the content: the Bangladesh Awami League and the Bangladesh Awami Jubo League.

The Bangladesh Awami League is the party’s official verified Page with 3.4 million followers that shared nearly 9,500 pieces of content during the timeframe of our analysis (excluding third-party links). To put this in perspective, the party shared content every hour of every day, and at least 25 pieces of content for every 24 hours in a day for the past year. Actual data will show variations in sharing behavior that we will discuss in later parts of this report.

Similarly, despite 158 highly active Pages, the Bangladesh Nationalist Party (BNP) primarily relies on two Pages to drive the bulk of its content: BNP Media Cell, its media Page managed by its press team, and the Bangladesh Nationalist Party – BNP. BNP Media Cell has 2.5 million followers and has shared more than double of Awami League’s official Page—nearly 20,000 posts in the last year.

The Tactics

Political parties use an intricate web of Pages, Groups and verified Profiles to disseminate their political messaging. Groups are primarily used to build and organize communities around the party’s agenda. At the same time, they are a launching pad for establishing coordinated and organic distribution for other Groups, Pages and Profiles.

Photos are the most common types of posts for both parties, followed by videos directly uploaded on Facebook (‘native videos’). The dependence on sharing photos and videos – as opposed to text – could be attributed to a likely prevalence of low literacy rates among account admins and users in Bangladesh. Photos are often easier to replicate and redistribute while being effective at simplifying complex messages. Photos and byte-sized videos are seen as more engaging and easier to consume, a more accessible form of content for a wide range of audiences. Anecdotal evidence of comments supplementing photos and videos frequently include stickers and GIFs. These capture the public sentiment quickly while accumulating user interactions in hopes of gaining more traction on the newsfeed.

Facebook’s algorithm is also known to reward high interactions, which likely and typically end up being photos and native videos, by placing them higher on people’s feeds. In recent years, its parent company, Meta, has prioritized products like Reels and Stories, noting the former is its ‘fastest growing content format’.

Analyzing the data, we find photos are more prevalent in the Awami League’s Pages, accounting for 63.75 percent of the content shared, in comparison to 45.47 percent of the BNP’s. Native videos, a close second, are more common in BNP’s Pages and Groups, especially in BNP Groups, indicating a stronger focus on video content. This trend may be driven by the belief among BNP supporters that mainstream television channels do not adequately or fairly cover their events. Consequently, Facebook has gained popularity as an alternative platform for disseminating their activities.

Links such as news or articles are shared more by Awami League’s Pages and Groups, suggesting a greater emphasis on directing traffic to external sites. BNP’s Pages and Groups use live video more frequently than Awami League ones, particularly BNP Pages. This finding is consistent with the pattern of opposition party supporters streaming party events on Facebook. The BNP has leveled accusations against the government for deliberately slowing down internet speeds to hinder the broadcasting of live videos.

The Behaviors

For the Awami League, the Pages “Bangladesh Awami League” and “Bangladesh Awami Jubo League” have been the most active, each posting around 9,500 times in a year. However, during October and November, amid escalating election tensions, two Pages emerged with notably high post counts: “Bangladesh Awami Olama League” and “Awami League Media Cell.”

In terms of engagement, the “Bangladesh Awami League” Page, the party’s official Page, consistently outperformed others. It accumulated 6.6 million likes on its posts over the year, far surpassing the “Bangladesh Awami Jubo League,” which received 1.1 million likes. Other Pages received less than 100,000 likes each. The “Bangladesh Awami League” Page also maintained a dominant share of comments and shares, matching or exceeding its lead in likes.

Although the BNP’s media and official Pages took the lead for posting consistently and in high volumes, there were occasions when other Pages, such as “Natore Zela B.N.P-নাটোর জেলা বিএনপি,” led in terms of post count. From September 2023 onwards, coinciding with the party’s escalating political programs, other smaller Pages such as “My BNP” and “লিডার আসছে BNP” began to make frequent posts. When it comes to likes, comments, and other forms of engagement, the two official Pages significantly outshone the others.

From June 2023, there was a substantial surge in the number of posts being made on BNP Pages, which skyrocketed starting in September. In contrast, the activity on the Awami League Pages remained relatively stable, with only a modest uptick in April and September.

In July, the BNP Pages had amassed over 4,500 posts, marking a significant rise from fewer than 3,000 posts in the preceding month. In September, these Pages added just under 4,000 posts, but in November, this number almost tripled, surpassing 11,000.

A similar, though more linear, trend was observed in BNP Groups. The number of posts in these Groups doubled from 16,000 in December 2022 to over 30,000 in November 2023. Meanwhile, the Awami League Groups performed relatively better than their Pages, with the number of posts peaking at nearly 22,000 by the end of November compared to 12,000 in December last year.

The surge in BNP’s online activities in June and October appears to be linked to two major political events.

In late May, the United States government announced a visa restriction policy related to Bangladesh’s election. This policy, intended to penalize anyone undermining the democratic election process in Bangladesh, was largely seen as a relief for the BNP, which had been facing severe operational constraints and a harsh government crackdown.

Subsequently, the BNP scheduled its grand rally in Dhaka for October 28. The rally later descended into violence when law enforcement tried to disperse the crowd. This led to the BNP calling for general strikes and transportation blockades, resulting in several deaths – mostly among BNP activists – and the burning of hundreds of vehicles.

The increased activity on pro-BNP Pages and Groups during these periods seems to correlate with the escalating tensions on the ground.

Unlike their Pages, the BNP and Awami League Groups exhibited varied behaviors.

Our analysis of the top 20 Groups for each party, based on post counts, revealed distinct patterns. For Awami League Groups, a substantial portion of the posts – between 19% and 50% – were concentrated in just two Groups: “BANGLADESH AWAMI LEAGUE” and “বাংলাদেশ আওয়ামী অনলাইন লীগ (Bangladesh Awami Online League).” However, in terms of likes, shares, and comments, the “বাংলাদেশ আওয়ামী-যুবলীগ (Bangladesh Awami Jubo League)” Group also stood out as a key player.

On the other hand, ‘BNP’ Groups showed a more decentralized trend. No small set of Groups dominated in terms of post numbers; instead, posts were more spread out across a wider range of Groups. Similarly, when looking at likes, shares, and comments, there was a more even distribution of engagement across five to six Groups, with none significantly outperforming the others. This likely indicates that BNP supporters are more organically organized and share content to affirm their party’s ideological position.

The Web of Networks

Our network analysis finds that content posted on a Group often is widely shared across other Groups, Pages, and Profiles, either affiliated with the political party by name or are alleged supporters. To understand this hypothesis further, we analyzed the distribution of the highest-circulating links from both Awami League and BNP over the one-year period, focusing on how these links were shared across various Pages, Groups and verified Profiles.

Our findings revealed that only 19% of the Pages, verified Profiles, or Groups that shared the Awami League’s top links were those that had ‘Awami League’ in their name, in either Bengali or English. In stark contrast, for the BNP’s highly shared link, this figure was 64%.

This suggests that the Awami League’s distribution networks are more varied than those of the BNP, aligning with our initial hypothesis. It appears that Awami League’s content is more frequently shared on Groups or Pages with names that reflect cultural or political affiliations other than the party’s own name.

The networked behavior raises important insights into how political parties organize themselves online. We find Groups share content that often is widely disseminated across other Groups, Pages, and Public Profiles (collectively called ‘mediums’). In the case of the Awami League, 91 mediums shared the same content within a span of 13 days of initial posting, while for BNP, 50 mediums shared within the same time span. We could not analyze data about individual profiles at scale because of privacy settings. Based on ground truth, we believe individual Profiles constitute a sizable share of mediums for both political parties.

In studying networked behavior in Bangladesh, it is important to therefore go beyond superficial indicators of Group (or actor) affiliation—same name, shared admins, or the same content. Rather, the analysis should include a combination of explicit and implied indicators to establish coordination.

Impact on Online Constituencies

Our analysis of the two leading political parties in Bangladesh indicate that they use an interconnected web of Pages, Groups and Profiles (verified, public and individual) to establish their messaging. There is some evidence of coordination among these mediums by both parties with the intent to compete against each other in ‘flooding’ people’s newsfeeds. Based on how the Pages, Groups and Profiles are organized, we find that Awami League has a more centralized (where bulk of the content is driven by a very small number of mediums) dissemination strategy relative to the BNP. The latter tends to have a more decentralized approach, likely driven organically by their supporters.

Additionally, the analysis shows that BNP cumulatively shares a higher volume of content, and subsequently, attracts more reach and engagement. However, this does not necessarily mean that the Awami League is lagging behind. Contrarily, we find the Awami League leverages a more sophisticated network of Pages, Groups and Profiles that do not bear the party’s name. Limitations around the safe access of privacy-preserving Facebook data make it a challenge to rigorously assess the expansive nature of Awami League’s online presence.

Considerations for Platforms

Our investigation raises several critical questions on political engagement on social media platforms and expectations around transparency. Many of the largest platforms provide some degree of transparency on political advertisements, including disclosing to the public on who paid for the ads and archiving relevant information on their ad library. Meta’s Ad Library provides information on spend, reach and source of funding, as well as whether the ads are visible or active for up to 7 years.

A key consideration for platforms is whether there needs to be comparable transparency for organic partisan content. Our analysis of political parties in Bangladesh using Pages and Groups outside of their verified and named accounts reveals three risk vectors:

Online constituencies do not know whether a Page or Group bearing the party’s name is formally affiliated with a political party — operated professionally by an official member or paid by the party.
Partisan content is often shared by unassuming Groups and Pages that do not bear the party’s name or disclose their association. Group or Page names can, hypothetically, include Good Morning Bangladesh or Our Dhaka. They typically tend to be partisan actors that are misleading users about the fact that they are partisan and sharing hyperpartisan content.
There are no discernible factors to determine what constitutes “authoritative information” coming from a political party, which in turn, increases the likelihood of amplification of misinformation.

The aforementioned combination exacerbates risks of a sophisticated influence operation that is widely disseminating organic hyperpartisan content and swaying public opinion. Although Meta requires Pages and Groups to categorically disclose whether they are affiliated with a political party or not (“political organization”), the policies are not sufficiently nuanced to differentiate between affiliation (member of the party or a paid operator) and support. Further, enforcement of existing policies on Group and Page categories is inconsistent. People can follow a Page or join a Group without knowing its partisan affiliation, especially when the political party chooses to disseminate their messages through non-partisan names.

This leads to two considerations for social media platforms. First, a Group, Page or verified Profile sharing partisan content organically should provide some transparency about their association with the party, especially if they do not categorically bear the party’s name. Second, a Page or Group bearing a political party’s name should go through verification to reduce risks of impersonation and misrepresentation. Although anti-impersonation policies are widely prevalent across platforms, we suggest additional scrutiny and user support when it involves a political party, given higher risks to influence civic participation in low literacy environments.

The scope of this report is limited to understanding the actor and behavior of the Pages, Groups and Profiles without going into specific content. Therefore, we cannot confidently conclude whether there is a higher prevalence of misinformation, and plan to explore this hypothesis further in future studies.

Methodology

We primarily relied on the Crowdtangle UI and its API to collect data from public Pages and Groups within rate limits permitted by the platform. We collected interaction data on Facebook with posts within our filters. Specifically, we employed a keyword-based search to find political parties that led to the creation of a list of Groups and Pages that might be associated with either of the two major political parties in Bangladesh: the Bangladesh Awami League and the Bangladesh Nationalist Party. We worked with local analysts and information researchers to identify whether the retrieved Pages and Groups were germane to the analysis and expanded our search accordingly. Our dataset included public posts and related interaction activities from December 1, 2022 to November 30, 2023. We analyzed a total of 660 Groups and Pages, totalling 427,928 data rows, with each row representing a unique post identifier.

In the final report, we only included data from Groups and Pages that have more than 500 followers. As a result, our final dataset included 260 Pages and 70 Groups each for Awami League and BNP, respectively. These Groups and Pages were selected as per the ordered results returned by the CrowdTangle UI search. The data is stratified resulting in an equal number of Groups and Pages for Awami League and BNP. We narrowed this down to Groups and Pages that were active during the timeframe of our study. We defined Groups and Pages to be active if they have made at least one post from December 1, 2022, to November 30, 2023.

We selected Groups and Pages using the names of the political parties because they have a higher likelihood of being affiliated with the corresponding political party. We used a number of combinations of spellings in both English and Bangla to obtain a representative dataset from the Pages and Groups on the platform. We recognize that political parties often leverage Groups and Pages that do not contain the party’s respective name, and therefore, our analysis does not cover the full breadth of how political parties are directly and indirectly using social media. Despite this limitation, we find significant activity for each selected Group and Page affiliated with the political party by name, allowing us to to understand their behavior and information strategy over time.

We apply the Disinformation ABC framework to analyze the actors and behavior of the Groups and Pages. Specifically, we investigate activity in terms of the number of posts created over time, the distribution across actors, and how the activity of each of these Groups and Pages has changed in light of the upcoming election.

We analyzed data using Page or Group names, post links, and post timestamps. Each record includes a post link and timestamp with no missing values for these identifiers. We calculated the total monthly posts for all Awami League and BNP Groups and Pages. Additionally, we gathered the total monthly posts for the top 20 Pages and Groups with the most posts. Our rationale is to examine how these Pages’ activity changed over the 12 months, specifically focusing on the differences between the number of posts made by Awami League and BNP. We also aimed to identify the most active Pages and Groups, anticipating increased activity during the upcoming elections in Bangladesh.

During our analysis, we encountered several technical limitations:

Rate limits: There are API rate limits of a certain number of queries possible to run each minute that was a constraint in expanding our analysis further for each list created through the CrowdTangle UI, we collect historical data for each month, extending as far back as 12 months from the date of the analysis.
Search structure: Keyword-based search without query expansion techniques have a known set of limitations, primarily that they cannot retrieve semantically similar Pages and Groups without an exact text match.
Data access: We have not been able to look at individual user profiles because of privacy restrictions: it was not possible to obtain the profile information of all of the key actors and accounts that administered or posted frequently in many of the Pages and Groups that we studied. Having access to this information would have made it possible to attribute certain activities to individual actors instead of broader networks of Pages and Groups.
Admin affiliation: We relied on ground truth and local expertise to confirm if a Page or Group bearing the political party’s name is officially associated with the party. Admin information is outside of the scope of this analysis.

¹https://blog.twitter.com/en_us/topics/company/2021/disclosing-state-linked-information-operations-we-ve-removed
²https://graphika.com/reports/bad-reputation
³https://graphika.com/reports/likes-for-lorestan
⁴Tech Global Institute
⁵SimPPL

Data visualization by Sheikh Waheed Baksh, Nazmul Ahasan and Sabhanaz Rashid Diya. Graphic design and illustration by Shafique Hira and Rishad Hasan Sourov.

InfoLab