HSE researchers examine wellbeing of Russian social media users and rank public holidays by popularity

Researchers of the HSE Graduate School of Business trained a machine-learning (ML) model to infer users’ subjective wellbeing from social media posts. Having processed 10 million tweets, the researchers compiled a rating of holidays celebrated in Russia based on their popularity. The New Year tops the list, but Russian-speaking users of Twitter are also happy to celebrate Defender of the Fatherland Day, International Women’s Day, Victory Day and Halloween. The study findings have been published in PeerJ Computer Science.

As one of the most popular methods for people to communicate and share information and opinions, social media is an important source of data for researchers—particularly because this information can be used to track people’s emotions in real time. 

Knowing how people feel at a given time—also defined as measuring observable subjective well-being (OSWB)—can provide valuable guidance for policymakers, instead of or alongside currently utilised indicators such as gross domestic product. 

Researchers of the HSE Graduate School of Business, prof. Mikhail Komarov and Sergey Smetanin calculated OSWB indices for the Russian-speaking segment of Twitter. Unlike common subjective wellbeing measurements based on survey data collected by research centres such as VCIOM, measuring OSWB via posts in social networks does not require direct contact with users.

‘Social networks became one of the biggest sources of data which should be properly analysed in order to improve processes in the companies and in order to analyse activity of their potential customers’, says Mikhail Komarov, professor of the department of business informatics, HSE Graduate School of Business

The researchers used Twitter Stream Grab—a publicly available historical collection of JSON content grabbed from the general Twitter ‘Spritzer’ API stream—as a data source on tweets in Russian. According to Twitter, this API provides a 1% sample of all complete public tweets and is not tied to specific topics. Thus, the researchers consider it a good and sufficiently representative source of tweets on a wide range of subjects. 

The largest dataset of general-domain tweets in Russian, RuSentiTweet, was selected to train the ML model. This is the largest dataset of tweets with manual annotations for sentiment analysis. RuSentiTweet consists of 13,392 tweets grouped into five classes: Positive, Neutral, Negative, Speech Acts (such as greetings or congratulations), and Skip (which do not express any clear sentiment or attitude).  

The researchers applied the ML model to 10,869,003 tweets posted in Russian by 1,955,827 unique users over 20 months (an average of 5.55 tweets per user).

Based on this data, the study authors compiled a popularity rating of holidays among Russian-speaking Twitter users. As expected, the New Year turned out to be the most popular holiday, with the share of greetings on December 31 being more than triple the annual average and accounting for 12.3% of all tweets for that day. Defender of the Fatherland Day and International Women’s Day rank second and third, respectively. 

Halloween is one of the most popular ‘foreign’ holidays on Russian-speaking Twitter, ranking ninth among all holidays, ahead of Russia Day and International Workers’ Day on May 1st. This finding, however, is different from those reported by VCIOM. According to the researchers, the reason may be that Twitter is dominated by a younger age group that is more inclined to celebrate Halloween, whereas the VCIOM survey provides a representative sample of the Russian population.

Since there is some evidence suggesting gender differences in attitudes towards certain holidays, the holiday rating was first calculated for each gender separately.

‘The share of tweets from women with holiday greetings was higher for all holidays except Cosmonautics Day. Indeed, women are more likely to post greetings and other speech act tweets on ordinary days as well as on holidays,’ explains Sergey Smetanin, doctoral student of the HSE Graduate School of Business. 

The researchers also note that Russian-language tweets from Twitter Stream Grab can only be used in addition to conventional survey-based SWB indicators, not as the main source of information. There are two main reasons for this. First, the analysis for this study included Russian-language tweets from users outside Russia. Their subjective wellbeing may be different, thus affecting the research findings in one way or another. Second, older age groups were underrepresented in the study, as Twitter is mainly popular among a younger audience. 

‘We compared the OSWB findings with the survey-based VCIOM Happiness Index and found a statistically significant correlation. We assume that with access to a larger volume of data, it would be possible to obtain an even stronger correlation and potentially prove that Twitter can be used on its own as a reliable source of data on OSWB,’ says Smetanin.