Statistics Netherlands (CBS) has previously published articles about the development of the social tensions indicator based on social media posts. This indicator has been developed by Statistics Netherlands’ (CBS) Center for Big Data Statistics (CBDS) and is intended to measure tensions in society. What sets the social tensions indicator apart from more general measurements of positive or negative sentiments on social media is its focus on social unrest. This indicator uses a list of words specifically related to feelings of tension and agitation, with this list helping to establish the daily percentage of tweets about these topics.
The indicator was updated every day from 1 January 2010 to mid-December 2018. A visible increase in social tensions began in December 2017. Further analysis has shown that this period saw a relative increase in tweets reporting negative changes in society. Some specific examples include events relating to migration policy (in particular reactions to the possible deportation of two Armenian teenagers) as well as the debate about the controversial cultural figure Zwarte Piet and the associated riots during Sinterklaas celebrations in 2018. Another example relates to the potentially negative consequences of expanding the powers of the intelligence services, in combination with tighter privacy legislation.
The analysis included all public tweets and retweets posted in Dutch by Dutch users located in the Netherlands. Hence, tweets from that proportion of the population of the Netherlands that is active on Twitter were used as the basic principle to reflect social tensions among the Dutch population. Although it is well known that users of social networks, such as Twitter, are not representative of the population at large, and that young people are still overrepresented on these networks (CBS, 2018), previous CBS research has shown that social media can be used to measure indicators that are based on sentiment. It has also become clear that posts on social media are a good indication of sentiments among society (Daas and Puts, 2014; Van den Brakel et al., 2017).
Tweets representing feelings of tension and unrest were offset against the total number of tweets per day. CBS used in-depth interviews to determine which words people use to describe insecurity. Including synonyms, the list contained around 350 words, which were checked by a group of experts from CBS. The words from this list that are used most often on social media were then included in the definitive list of words. Tweets relating to sports events and politics were filtered out of the selection, as they had a distorting effect. Finally, the peaks in the indicator were validated; once CBS had checked which events took place on or immediately before the date in question, the researchers investigated whether the selected tweets did in fact refer to these events. A median filter over a rolling period of 50 days was used to determine the trend. The median – the middle value – was selected because it is less sensitive to extreme values (the peaks).
The visualisation shows variations in the social tensions indicator over time. The indicator also displays fairly regular emphatic peaks. These peaks refer to specific moments when a relatively large number of tweets were posted that expressed feelings of tension and unrest, such as the significant peak in 2010 associated with the disruption of the National Remembrance Day commemorations by the ‘Dam Screamer’ on 4 May in Amsterdam. Other peaks were linked to attacks, such as the MH17 air disaster (17 July 2014) and the terrorist attacks in Paris (13 November 2015) and Brussels (22 March 2016). A clear change in the social tensions indicator is visible starting at the end of 2017, when the long-term trend (the baseline) began to rise significantly.