Are fear and hope of the COVID-19 pandemic responsible for the V-shaped behaviour of global financial markets? A text-mining approach

ABSTRACT

This study found the recent global V-shaped behaviour in major stock and cryptocurrency markets to be attributed to the dramatic variance in public fears and hope regarding the COVID-19 pandemic. Using the Term Frequency-Inverse Document Frequency text-mining technique and a large dataset of tweets, we determined that the public sentiment on joint discourses of COVID-19 and financial topics was a strong driver of recent stock and cryptocurrency markets behaviour. Furthermore, error correction model estimations revealed that the reversal of public sentiment from fear to hope after the initial shock significantly contributed to financial markets’ recovery phase of the V-shaped behaviour, partially explaining their sudden turnaround beginning mid-March 2020.

KEYWORDS:

• Covid-19 pandemic
• stock markets
• cryptocurrency market
• public sentiments
• v-shaped recovery
• text mining

JEL CLASSIFICATION:

• G01
• G15

Acknowledgments

This research is funded by University of Economics Ho Chi Minh City, Vietnam

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Additional summary statistics of the tweet datasets are available from the authors upon request.

² TF-IDF (Term Frequency-Inverse Document Frequency) is a feature extractions technique used to produce a numerical feature vector for a document based on the TF-IDF weight. This weight is a statistical measure on how representative a word could be for a document.

³ Doc2Vec (Doc to Vector) is an unsupervised algorithm to generate numerical feature vectors for a document and is an adaption of the Word2Vec (Word to Vector) algorithm; whereas Word2Vec captures the feature vector for a word in a document, Doc2Vec produces the vector features for a document such that it is sufficient to predict the word in the document.

⁴ The area under the ROC curve (AUC) measures how well algorithms can distinguish between positive and negative sentiment in a sample of tweets.

⁵ We collected the data of ten stock indices: ASX 200 (Australia), FTSE 100 (UK), KOSPI (Korean), CAC 40 (France), DAX (Germany), NIKKEI 225 (Japan), DOWJ (US), S&P 500 (US), and NASDAQ (US). For the cryptocurrency markets, the price data for 14 crypto coins were collected: BTC (Bitcoin), ETH (Ethereum), XRP (Ripple), BCH (Bitcoin Cash), BSV (Bitcoin SV), LTC (Litecoin), ADA (Cardano), LINK (Chainlink), BNB (Binance coin), CRO (Crypto.com Coin), EOS, XTZ (Tezos), XLM (Stellar), and TRX (TRON).

Additional information

Funding

This work was supported by the University of Economics Ho Chi Minh City.