Title Evaluation of consumer confidence indicators using social media and administrative data /
Translation of Title Vartotojų pasitikėjimo rodiklių vertinimas naudojant socialinių tinklų ir administracinius duomenis.
Authors Vitkauskaitė, Akvilė
Full Text Download
Pages 89
Keywords [eng] Vartotojų pasitikėjimo rodiklis, Socialiniai tinklai, Twitter, Sentimentų analizė, SARIMAX, VECM, Random Forest, XGBoost Consumer confidence, Social Media, Twitter, Sentiment Analysis, SARIMAX, VECM, Random Forest, XGBoost
Abstract [eng] The main objective of this study is to nowcast and forecast the Consumer Confidence Index (CCI). The aim is to estimate the current month's CCI values faster than those obtained using the traditional survey methodology, which usually provides results at the end of the month. For instance, while the official CCI for November would typically be available in the last few days of November, this research aims to provide an early estimate at the beginning of November, utilizing data collected at the start of the month. This is achieved by combining key economic indicators with historical CCI values. The research includes examining the relationship between traditional survey-based indicators and consumer sentiment expressed on social media platforms. Social media expressions, particularly from X (Twitter), are analyzed through its official API. The sentiment analysis of tweets has enabled us to create a Social Media Indicator (SMI) that offers a distinct advantage in our predictive models. In addition, the study explores the possibility of integrating key economic indicators from administrative data, such as inflation rate, income statistics, and unemployment. In general, obtaining data for research from popular social platforms such as Facebook and Instagram is not possible due to stringent privacy policies and data protection regulations. Nevertheless, data are easily and legally available from X, but this platform is not so popular in Lithuania. Therefore, the representativeness of X data raises special issues. Taking everything into account, by combining traditional economic indicators with advanced sentiment analysis from X, the study seeks to deliver prompt CCI predictions ahead of standard survey timelines.
Dissertation Institution Vilniaus universitetas.
Type Master thesis
Language English
Publication date 2024