Advanced

Twitter Sentiment and Stock Returns

Wolf, Fredrik LU and Bergdorf, Oskar LU (2019) NEKN02 20191
Department of Economics
Abstract
This study aims to investigate if the sentiment expressed on Twitter has an effect on individual stock returns. We use a uniquely large data-set consisting of129 million tweets related to 31 companies over a 15-month period. The sentiment is derived on a daily basis with Loughran and McDonald’s established finance-focused method, as well as with Vader, a sentiment analysis tool developed specially for social media. Using a panel data regression, we empirically test Twitter sentiment effectual forecasting power on individual stock returns. Our main findings are: 1) the sentiment derived from Twitter can help predict individual stock returns up to two days ahead 2) the individual stock returns are more sensitive to ’cashtag’-tweets than... (More)
This study aims to investigate if the sentiment expressed on Twitter has an effect on individual stock returns. We use a uniquely large data-set consisting of129 million tweets related to 31 companies over a 15-month period. The sentiment is derived on a daily basis with Loughran and McDonald’s established finance-focused method, as well as with Vader, a sentiment analysis tool developed specially for social media. Using a panel data regression, we empirically test Twitter sentiment effectual forecasting power on individual stock returns. Our main findings are: 1) the sentiment derived from Twitter can help predict individual stock returns up to two days ahead 2) the individual stock returns are more sensitive to ’cashtag’-tweets than general-company tweets 3) Twitter data can be used to explain the stock return volatility 4) Loughran and McDonald Lexicon-based method outperforms Vader. The results indicate that Twitter data is a suitable data source to understand and forecast stock market movements. (Less)
Please use this url to cite or link to this publication:
author
Wolf, Fredrik LU and Bergdorf, Oskar LU
supervisor
organization
course
NEKN02 20191
year
type
H1 - Master's Degree (One Year)
subject
keywords
sentiment analysis, Twitter, stock return, volatility, prediction, forecasting
language
English
id
8984394
date added to LUP
2019-08-08 10:28:30
date last changed
2019-08-08 10:28:30
@misc{8984394,
  abstract     = {This study aims to investigate if the sentiment expressed on Twitter has an effect on individual stock returns. We use a uniquely large data-set consisting of129 million tweets related to 31 companies over a 15-month period. The sentiment is derived on a daily basis with Loughran and McDonald’s established finance-focused method, as well as with Vader, a sentiment analysis tool developed specially for social media. Using a panel data regression, we empirically test Twitter sentiment effectual forecasting power on individual stock returns. Our main findings are: 1) the sentiment derived from Twitter can help predict individual stock returns up to two days ahead 2) the individual stock returns are more sensitive to ’cashtag’-tweets than general-company tweets 3) Twitter data can be used to explain the stock return volatility 4) Loughran and McDonald Lexicon-based method outperforms Vader. The results indicate that Twitter data is a suitable data source to understand and forecast stock market movements.},
  author       = {Wolf, Fredrik and Bergdorf, Oskar},
  keyword      = {sentiment analysis,Twitter,stock return,volatility,prediction,forecasting},
  language     = {eng},
  note         = {Student Paper},
  title        = {Twitter Sentiment and Stock Returns},
  year         = {2019},
}