Digital footprints of Kashmiri Pandit migration on Twitter

The paper investigates changing levels of online concern about the Kashmiri Pandit migration of the 1990s on Twitter . Although decades old, this movement of people is an ongoing issue in India, with no current resolution. Analysing changing reactions to it on social media may shed light on trends in public attitudes to the event. Tweets were downloaded from Twitter using the academic version of its application programming interface (API) with the aid of the free social media analytics software Mozdeh . A set of 1000 tweets was selected for content analysis with a random number generator in Mozdeh . The results show that the number of tweets about the issue has increased over time, mainly from India, and predominantly driven by the release of films like Shikara and The Kashmir Files . The tweets show apparent universal support for the Pandits but often express strong emotions or criticize the actions of politicians, showing that the migration is an ongoing source of anguish and frustration that needs resolution. The results also show that social media analysis can give insights even into primarily offline political issues that predate the popularity of the web, and can easily incorporate international perspectives necessary to understand complex migration issues.


Introduction
It is often believed that Kashmir has derived the title "paradise on earth" exclusively due to its unmatched natural beauty. In reality, however, there is much more unfolded to it. Out of the many other determinants contributing towards such adulation, Kashmiriyat, an indigenous tradition of communal harmony and religious syncretism, stands out and has definitely been the most indispensable factor to sustain such pursuit. Unfortunately, though, it is this fabric of Kashmiriyat that has been the primary casualty due to the protracted political conflict in the valley of Kashmir. The mass exodus of the Kashmiri Pandit community (followers of Hindu religion) that took place at the very beginning of the armed insurgency built many walls and razed the centuries-old bridges between the majority Muslim population and the minority Pandit population. Such unwanted migration not only dented the secular orientation associated with Kashmiriyat but also led to an unprecedented loss to humanity in general. In the modern history of South Asia, after the post-partition migration of common people between India and Pakistan, the Pandit migration surely has to be the second largest displacement of common people, ironically an exit that nobody would want because it would mean an exit from Paradise.
The Kashmir Valley has been in crisis since the British-administered partition of India in 1947 led to this Muslim-majority area joining India within the Jammu and Kashmir region. A part of the region has been administered by Pakistan (Azad Kashmir, Gilgit-Baltistan) ever since, and China also administers a small part of it, known as Aksai Chin. Such divided regional control amongst three nuclear powers, primarily from a political perspective, has pushed the region into a protracted conflict that, till date, remains unresolved partially, if not completely. Kashmir conflict has many dimensions; however, of the many ongoing controversial issues in the area, this paper focuses on the Kashmiri Pandit migration of the 1990s, when a large Hindu group migrated from the Kashmir Valley due to the onset of political instability and fear for their safety. Whilst this issue is three decades old, forced migration causes a lifelong crisis for the victims and sympathy continues to be expressed, for example through the 2020 and 2022 Bollywood films Shikara and The Kashmir Files. In particular, the focus is on how this issue is discussed on Twitter in the hope of gaining insights into ongoing public opinions about the subject. Social media is suitable for studying political opinions because, unlike traditional media, users play an active part in reporting and sharing online content globally and locally (Odlum;Yoon, 2015). Social media platforms can therefore transform public discourse and assist by adding grassroots contributions to those of established media agencies, governments, and political organizations (Karatzogianni; Nguyen; Serafinelli, 2016). In addition, social media platforms help generate virtual bonds between users who express their opinions and develop relationships through posts, comments, messages, and likes (Öztürk;Ayvaz, 2018). According to Chew and Eysenbach (2010), social media users can share at least 14 different types of information about an issue: news and updated information about an event; individual experiences; personal thoughts and interests; jokes; marketing and products; advertising; irrelevant information; humor and sarcasm; joy; risk reduction; concern, fear, anxiety or sadness; despair and anger; misinformation; and questions related to an event or subject.
Twitter is a popular microblogging platform that is often used in some countries to share news, current status, opinions, and activities during global and local events, making it an attractive source to mine for public opinions about news-related issues (Pak;Paroubek, 2010). For example, sentiment analyses of tweets are sometimes used to study public perspectives about issues on a large scale (Li; Liu; Li, 2020). Twitter has been widely used internationally to discuss refugees and migrants when they are in the news. The different perspectives that it shares (Nerghes; Lee, 2018) may help users to empathize with displaced people. This makes it a suitable platform to analyse recent reactions to the Kashmiri Pandit Migration of the 1990s.

Background of Kashmiri Pandit migration
Kashmiri Pandit migration fits within the history of Kashmir as part of a period of unrest starting in 1988unrest starting in or 1989unrest starting in (Bose, 19972009;Burki;United States Institute of Peace, 2007;Hanif;Ullah, 2018;Hussain, 2021;Rajput, 2019). Elements of the unrest were also hostile to the Kashmiri Pandits (Bose, 2021;Evans, 2002) with fears or misunderstandings that the killings were or would target Hindus indiscriminately (Evans, 2002). This then led to a wave of migration of 90,000-100,000 out of 120,000-140,000 Pandits from the Kashmir Valley to escape the threats and violence (Bose, 2021;Sarkaria, 2009, p. 197;Sawhney, 2019;Singh, 2015), despite some Kashmiri Muslim delegations and neighbors asking them to stay (Evans, 2002). The Kashmiri Pandits are officially termed 'migrants,' by the Government of India and that of Jammu and Kashmir (Datta, 2016, p. 53;2017, p. 1099. The migrants initially went to refugee camps in Jammu but most subsequently moved elsewhere in India (Evans, 2002). Thus, it seems clear that the primary reason for the migration was a genuine fear for safety due to the targeted killings, and widespread unrest.
The mass exodus of Kashmiri Pandits was a humanitarian crisis that brought disrepute to the centuries-old socio-cultural tradition of Kashmiriyat, with its communal harmony and religious syncretism (Zutshi, 2019). Since 1990, both Kashmir and Kashmiriyat remain broken and incomplete until the Kashmiri Pandits have a safe return to the Valley.
Few of the Pandits have returned, despite almost all the major parties in Kashmir calling for this, a small government financial incentive to return in 2008, and an apparent attempt to make Kashmir favorable for migration from India (Rai, 2020) and elsewhere. For example, one of the Muslim leaders insinuated, "We want our Kashmiri Pandit mothers, sisters, and brothers to come back. It is their land. They have every right to live in it as we do" (Sharma, 2004).
In this context, on Twitter, it could be expected that Kashmiri Pandits would get widespread support within Kashmir and India.

Twitter as a source of information on crisis and political opinions
Twitter has been widely used for investigating public reactions and information dissemination during natural crises. For instance, Twitter has been used to help trace the spread of discussions about infectious diseases like Zika, Ebola, H1N1, and COVID-19 (Bhat et al., 2020;Guidry et al., 2017;Jain;Kumar, 2015;Medford et al., 2020;Thelwall, 2020;Thelwall, 2021;Vijaykumar et al., 2018), to hold opinions about gender bias (Thelwall, 2018 a, b) and also during natural disasters like floods (Gul et al., 2018;Middleton;Middleton;Modafferi, 2013). In addition, Twitter has also been used to investigate social movement communication during the 2011 UK riots (Gascó et al., 2017) and Wisconsin Labor Protests (Veenstra et al., 2014), for stock market prediction (Mittal;Goel, 2012), to study consumer perceptions (Schivinski;Dabrowsk, 2016), and for election predictions (Ansari et al., 2020;Gul et al., 2016;Wang et al., 2012). All these studies have shown that Twitter can be a useful source of information about public opinion related to political issues, although they do not prove that it is always useful, and all prior research has had substantial limitations. In particular, all have had to ignore the opinions of those that do not tweet. This is a substantial limitation given that tweeters tend to be atypical within a country.

Twitter analyses of migrations to Europe
Twitter data has been used to investigate reactions to migration episodes across the globe, including to contrast the "deserving refugees" and "undeserving migrants" frames of reference. Most of the analyses seem to have focused on Europe. Urchs et al. (2019) analysed 2015 tweets about migrants and refugees to Hungary, Germany, and Austria. The study found that most tweets were sent in the initial stages and were predominantly in English, followed by German. A high proportion were retweets or contained links to information elsewhere.
Similarly, Coletto et al. (2016) examined EU citizens' tweets related to the Mediterranean Refugee Crisis. The analysis showed that the term "refugees" was associated with a positive attitude, but "migrants" was associated with negativity. Most tweets came from the United Kingdom, followed by France and Germany. There was a significant peak in tweets on three news-related days, showing that European users were sensitive to news events.
Several studies have found positivity towards migrants or refugees on Twitter in Europe. (2019) analysed refugee-related tweets referring to the ship Aquarius, which carried refugees to the Spanish port of Valencia after being rejected by Italy. They analysed tweets in Spanish, finding that most of the messages were positive towards the refugees. However, the share of negative messages, including hate speech against the refugees and politicians, grew after the Spanish government welcomed the refugees. This illustrates that Twitter can be used to voice discontent against government policies and is perhaps a more natural medium for protest rather than agreement. Arcila-Calderón et al. (2021) also analysed refugee-related Spanish tweets around high-profile events, finding an overall positive attitude towards the refugees. In this case, the most negative words were mainly reserved for the politicians.

Vázquez and Pérez
Aydemir and Akyol (2020) explored the negotiation of pro-migrant online activism campaigns and evaluated the response of individuals in France and India during the COVID-19 pandemic. The tweets tended to express support, solidarity, and help for migrants.
In contrast, other studies have found negativity towards migrants or refugees. Backfried and Shalunts (2016) analysed the refugee crisis in German, finding that the temporal sentiment in traditional media was neutral, whereas social media was dominated by negative sentiment.
Yantseva (2020) analysed migration attitudes in Sweden with newspapers, an online forum, and Twitter. The study concluded that messages shared on social media sound negative, and some media frames contributed to a migration-hostile discourse. Ozduzen, Korkut and Ozduzen (2020) studied the evolutionary modes of online anti-immigrant and racist discourse related to Syrian Refugees in Turkey. The qualitative content analysis and sentiment analysis of tweets revealed that most of the tweets express a clear racist attitude towards Syrians. The negativity in Turkey, Germany, and Sweden may reflect negative media frames or more popular or better-organized anti-immigrant right-wing groups.
The mainstream news media, politicians, and influential tweeters can all impact the way in which migrants are discussed.
Ferra and Nguyen (2017) analysed the public discourse and online platforms related to the migrant crisis in 2016. They deduced that many international media outlets, media-related individuals, and NGOs play an active role in transmitting information and shaping public discourse. Also, the wide geographical distribution of users reflected by their Twitter accounts shows that the discourse on the topic has a transnational-global reach.
Vilella et al. (2020) found a strongly separated network of tweets about migration in Italy that showed clear interaction with the Italian political and social landscape that seemed to be disconnected from the actual geographical distribution and relocation of migrants.
Gökçe and Hatipoglu (2021) analysed four months of tweets about the Syrian refugee crisis, finding that half of the tweets were associated with at least one political party in Turkey.
Nerghes and Lee (2018) analysed tweets posted two days after the death of the three-year Syrian refugee toddler Alan Kurdi, who drowned in the Mediterranean Sea. They found that more influential users made less dramatic statements.

Analysis of migrations in other countries through Twitter
Twitter has also been used to study migrations in other parts of the world.
In relatively mono-ethnic South Korea, 500 Yemeni refugees isolated offshore caused substantial debates about gender, nationalism, and identity (Kim et al., 2020).
In Cameroon, tweets discuss diaspora communities, with social media giving activists a platform to engage with government narratives (Nganji; Cockburn, 2020).
In the USA, Walker and Boamah (2020) analysed the tweets related to the "Central American Migrant Caravan" travelling from the Guatemala-Mexico border to the Mexico-United States border in 2018. Whilst tweets that were positive about the migrants expressed emotions like trust, joy, and anticipation, negative tweets mostly expressed fear and anger.
In India, internal migration issues have been mainly tweeted in relation to . Analyses of tweets about the internal migrations of workers returning from cities to their home villages due to Covid-19 economic closures have shown widespread trust in, and reliance upon, government intervention to support those affected by the disruption (Misra; Gupta, 2021). In partial contrast, a smaller-scale analysis focusing on the deaths of migrant families in a train accident led to Twitter pressure on the government for greater support to poor workers with Covid-19 unemployment (Sharma, 2022). This issue faded from Twitter relatively quickly. This is unsurprising since Twitter is news driven to some extent.

Methods
The research design was to gather a large sample of tweets mentioning Kashmiri Pandits and then conduct an exploratory content analysis of them and report quantitative information about the tweets. This is a common approach in Twitter research (e.g., Chew; Eysenbach, 2010), although data mining (Misra; Gupta, 2021), network analysis (Nerghes; Lee, 2018), and sentiment analysis (Bhat et al., 2020) are also used. We did not use data mining or sentiment analysis because the issue is fundamentally multilingual so these would not be very effective. We did not use network analysis because interactions between tweeters seemed too rare for this topic to give relevant results. As described below, the content analysis was inductive, with the coders designing relevant categories after reading samples of tweets. The inductive approach is appropriate because this is the first analysis of Twitter for the topic and there are no guides about the issues that are likely to be discussed.
A keyword search method was developed to identify many tweets that were clearly relevant to Kashmiri Pandit migration.
It is impossible to retrieve Twitter for all tweets relating to Kashmiri Pandit migration because some relevant tweets contain only images and others may allude to the issue without mentioning it directly (e.g., "I agree!", "19 Jan is a day of sad memories for me.", "I do not feel safe here anymore"). The stages of the methods, including the content analysis part, were as follows. 1. Eight relevant hashtags were found by exploring Twitter: #KashmiriPanditMigration, #KPExodusDay, #HumWapasAayenge, #kashmirexodus1990, #iheartkashmir, #KashmiriHindus, #Kashmiri_Pandit_targetted, #kashmiri_pandit. These were selected from a much larger initial set of potentially relevant tags after rejecting those matching a substantial proportion of irrelevant tweets. 2. All tweets matching these eight hashtags from the start of Twitter (2006) to August 2021 were downloaded from Twitter using the academic version of its application programming interface (API) with the aid of the free social media analytics software Mozdeh (mozdeh.wlv.ac.uk). 3. The top 1000 hashtags were extracted from the matching tweets. 4. The 1000 hashtags were manually checked to identify those that were directly about Kashmiri Pandits, finding 85. 5. All tweets matching these 85 hashtags from the start of Twitter (2006) to August 2021 were downloaded from Twitter using the academic version of its API. Accordingly, a total of 2,768,695 tweets were downloaded. 6. The tweets were filtered to remove duplicates (e.g., multiple copies of the same retweet, otherwise identical tweets). Before the duplicate check, all hashtags and @usernames were removed from the tweets to eliminate spam tweets that add random hashtags or usernames to marketing information. 7. To avoid domination by individual tweeters, a maximum of one tweet per tweeter per month was allowed, chosen at random by Mozdeh with a random number generator. This is a compromise rule that allows multiple tweets from prolific tweeters but restricts them to one per month so that they do not overwhelm the data. This gave 111,788 tweets. 8. The countries of the tweeters were identified by checking their Twitter biographies for country names (e.g., India, Pakistan) or major cities (e.g., Islamabad, Delhi). The countries of users with biographies not containing these indicators were left undefined. 9. A set of 1000 tweets was selected for the content analysis with a random number generator in Mozdeh. Two coders (Library and Information Science researchers) based in Kashmir identified a relevant set of categories for these tweets, coded all the tweets, and designed descriptions for the categories (i.e., a content analysis codebook). Categories were chosen based on the main points of the tweets when several tweets made similar points. Tweets not in a language that the coders could read were translated using Google Translate. 10. A further 100 tweets were randomly selected and coded by the two coders for a consistency check with Cohen's kappa (Cohen, 1960). The inter-coder consistency check had a Cohen's kappa score of 0.84 on the 99 non-missing tweets, which is high enough to validate the categories and coders.

Results
The first tweet collected by the hashtag queries was from 2019 and there was at least one tweet every month from January 2020. The number of tweets tended to increase each year, led by increases in January on the anniversary day (Graph 1). This suggests that the issue is growing in online attention, but the results may be distorted by increases in the proportion of the Indian population using Twitter, by old tweets being deleted by users when they leave the platform or clean their timeline, or by Twitter banning of controversial tweets. Perhaps more importantly, major news events like related film releases may have increased the volume of tweets for recent years.

Geographical distribution of tweets
The 111,788 tweets (one per user per month) matching the 85 Kashmiri Pandit migration hashtag queries were mainly from India (89%), with low representation from other countries ( Table 1). The dominance by India is expected because the area of concern is within India. Tweets from other countries, such as the USA, might also be partly from Indians that have moved abroad, whether temporarily or permanently.

Content analysis of tweets
A content analysis of a random sample of 1000 tweets matching the 85 Kashmiri Pandit migration hashtags produced ten categories of tweets ( Table 2), nine of which reflect different ways of tweeting about the topic. The inter-coder consistency check had a Cohen's kappa score of 0.84 on the 99 non-missing tweets, which is high enough to validate the categories and coders.
Almost a third of the tweets were from three ostensibly neutral topics but often expressed implicit emotions or criticism (Graph 2). These three topics were: Sharing news and information (16.2%), Providing suggestions (10.7%), and Asking questions (6.4%). Thus, given that the remaining categories were explicitly emotional or critical, this shows that the issue of Kashmiri Pandit migration is emotionally charged on Twitter.
Over a fifth of the tweets about Kashmiri Pandit migration criticized others (16.8%) or the government of Jammu and Kashmir (6.2%), representing widespread dissatisfaction with the situation. A fifth of the tweets primarily expressed emotions: Anguish (8.2%) or others (12.0%). The remaining 16% of tweets were implicitly critical of the current situation by Extending support (to Kashmiri Pandits) (10.8%) or Raising protests (4.8%).
It is perhaps unsurprising that the tweets were overall very critical of the current situation of the Kashmiri Pandit migration because those who are not critical of the situation would not need to tweet about it. Therefore, it is unclear whether the Graph 1. Number of tweets per month from January 2010, after filtering out duplicate tweets situation on Twitter reflects the general opinion of people who know about the issue. However, some of the tweets were not explicitly critical of the situation but emphasized other aspects. For example, many tweets included links to news stories about an old Kashmiri Pandit who had remained in his ancestral village in Kashmir and had no co-religionists to cremate him when he passed away. The news stories celebrated that his Muslim neighbors had carefully performed his last rites according to Hindu traditions, emphasizing this small act of kindness.
Some news stories announced the release of movies relating to Kashmiri Pandits. These films raised the profile of the issue, and therefore tweets about them would do the same. For example, the most retweeted tweet (20,178 retweets) was from filmmaker Vivek Ranjan Agnihotri announcing that he was making a film, The Kashmir Files, about the genocide. Nevertheless, some people may have tweeted about the movies for their own sakes. Another highly retweeted movie was Vidhu Vinod Chopra's romance/drama Shikara (6,164 retweets for the top tweet), although this also attracted criticism from some Kashmiri Pandits for its marketing strategy.

Discussion
A limitation of the results is that they only reflect the opinions of social media users. Since the Twitter users in the dataset seem to be primarily from India and Kashmir is part of India, it is not clear whether they present a non-Kashmiri perspective or a Kashmiri perspective. Another limitation is that the short length of tweets means that they may be misinterpreted in the content analysis. Moreover, the random sample used for the content analysis gives an overall picture of how the issue is discussed but will be dominated by popular time periods and does not help to distinguish between the comments made at different points in time. The sample size (1000) gives only rough estimates of the proportions of tweets matching each category (Kim et al., 2018).
Kashmiri Pandit migration led to the settlement of Kashmiri Pandits mainly in Jammu and other parts of India, which may be the reason why most of the tweets related to the migration are from India. However, some families may have settled abroad, expressing their sentiments related to the event. The results also confirm that even 30 years after the migration, sentiments are still alive, and Twitter has helped channel the feelings from India and other global regions. Whilst, as reviewed, many investigations of tweeting about migrants found prevalent negative attitudes towards migrants (Backfried; Shalunts, 2016; Ozduzen; Korkut; Ozduzen, 2020; Yantseva, 2020), the current study did not find negativity towards the Kashmiri Pandits. Instead, many tweets were strongly supportive of them, echoing the view of almost all political parties in Kashmir and India that they should be welcomed home. The strong negative emotions in the tweets tended to be directed instead against those that were perceived to be blocking this return, and not doing enough to facilitate the homecoming. The latter echoes some previous analyses of tweets about migrants, in the sense of Twitter being used to express frustration against, or disagreement with, politicians (Arcila-Calderón et al., 2021;Nganji;Cockburn, 2020). In addition, there were many news-sharing and question-asking tweets. The time series graph (Graph 1), confirms that the situation is still considered current and evolving, despite the migration occurring well before the debut of Twitter.
One of the critical concerns is the return and rehabilitation of the Kashmiri Pandits. As evident from the tweets, many are dissatisfied with the overall government actions. Some of the tweets also express love and warmth from Kashmiri Pandits to their former Muslim neighbors in the Valley, suggesting that the Kashmiriyat bond has not been completely broken by the migration.

Conclusion
This Twitter analysis has shown that the Kashmiri Pandit migration of the 1990s that ruptured the social, cultural, religious, political, and ethnic heterogeneity of the Valley is still an emotionally charged issue that is commemorated in January each year. In this context, there is still pressure on governments to resolve the issue of their return peacefully, supporting the aspirations of both Muslim and Pandit communities. The local support to welcome back the Kashmiri Pandit community may also restore the reputation of the centuries-old socio-cultural tradition of Kashmiriyat. The syncretistic social fabric of Kashmir for which it was and is known will become more strengthened if the Kashmiri migrant Pandits will find a return to the Valley. An open-door policy from all societal sectors will help the community to find their lost homes and identities.
Recall that the Kashmiri Pandit migration predated Twitter for decades and occurred before the web became popular, unlike the topics of most political Twitter analyses. This study was still possible because Twitter is popular in India (Statista, 2022). Because of this, there were enough tweets to analyse even though the Kashmiri Pandits are not a "social media issue", in the sense of the Arab Spring, #MeToo, or #BlackLivesMatter, even though there are important geographic dimensions. Thus, this study has shown that Twitter can be used to investigate important political issues that are not natural case studies for social media, as long as social media are widely enough used for substantial data gathering. Finally, for those studying issues around migrant communities, the results confirm one key advantage of social media analysis: the ability to easily incorporate global perspectives for an issue that is intrinsically international.