Measuring Corporate Reputation Through Online Social Media: A Case Study of Volkswagen Scandal
Name
Paria Molayemvand
Abstract
This research investigates if there is any relationship between public opinion on social media and corporate reputation. The hypothesis of this study is that public comments on social media influences corporate reputation. Studies have shown that social media channels have an impact on corporate reputation. Reputation is increasingly recognized for its impact on value creation for corporations. This study assumes that the financial performance of a company is the direct indicator of its reputation. Therefore, this study investigates the influence of people's comments on social media on corporation stock market value.
This research is a case study of the Volkswagen scandal and it focuses on data of Twitter posts between 2015 and 2016 and tries to find how sentiments from public opinion can influence corporate reputation and its financial performance in the crisis situation. The process is that the opinion of people on Twitter about Volkswagen is extracted from the tweets in the form of sentiment value. Then, the correlation between stock market price and volume and the sentiment of tweets is calculated.
For the data selection, a semi-manual approach is used to remove commercial, political and unrelated tweets from the tweet data set. This approach shows a high average accuracy of 0.92. Retweets are treated as new tweets and are not removed from the dataset to find out the influence of retweets on the correlation. Then, three different sentiment analysis tools are used and compared to find out which one has more correlation with stock market price and volume of a corporation. These tools are "Microsoft Azure text analysis API", R package "Sentimentr" and R package "SentimentAnalysis". Comparing the resulting sentiments shows that Sentimentr tool has a higher correlation with stock market data.
The correlation results show that there is a correlation between the sentiment of tweets and corporations' stock market data. The average sentiment of tweets per day has the highest negative correlation (-0.84) with the stock market volume of trades the first month after the scandal. As the months pass, the correlation drops dramatically (By the fourth month after the crisis, the correlation has dropped to -0.27). This means that first month after crisis while the average sentiment gets more negative, more stocks are traded. However, this doesn't necessarily indicate that negative opinion of people on Twitter influences the stock volume of trade. The correlation results show that stock price of day D has more correlation with the average sentiment of day D+4. This can indicate that actually, fluctuations in the stock price of the company can influence the sentiment of tweets. This is against our original hypothesis that public opinion on social media influences corporate reputation.
This research is a case study of the Volkswagen scandal and it focuses on data of Twitter posts between 2015 and 2016 and tries to find how sentiments from public opinion can influence corporate reputation and its financial performance in the crisis situation. The process is that the opinion of people on Twitter about Volkswagen is extracted from the tweets in the form of sentiment value. Then, the correlation between stock market price and volume and the sentiment of tweets is calculated.
For the data selection, a semi-manual approach is used to remove commercial, political and unrelated tweets from the tweet data set. This approach shows a high average accuracy of 0.92. Retweets are treated as new tweets and are not removed from the dataset to find out the influence of retweets on the correlation. Then, three different sentiment analysis tools are used and compared to find out which one has more correlation with stock market price and volume of a corporation. These tools are "Microsoft Azure text analysis API", R package "Sentimentr" and R package "SentimentAnalysis". Comparing the resulting sentiments shows that Sentimentr tool has a higher correlation with stock market data.
The correlation results show that there is a correlation between the sentiment of tweets and corporations' stock market data. The average sentiment of tweets per day has the highest negative correlation (-0.84) with the stock market volume of trades the first month after the scandal. As the months pass, the correlation drops dramatically (By the fourth month after the crisis, the correlation has dropped to -0.27). This means that first month after crisis while the average sentiment gets more negative, more stocks are traded. However, this doesn't necessarily indicate that negative opinion of people on Twitter influences the stock volume of trade. The correlation results show that stock price of day D has more correlation with the average sentiment of day D+4. This can indicate that actually, fluctuations in the stock price of the company can influence the sentiment of tweets. This is against our original hypothesis that public opinion on social media influences corporate reputation.
Graduation Thesis language
English
Graduation Thesis type
Master - Software Engineering
Supervisor(s)
Rajesh Sharma, Peter Ormosi
Defence year
2019