Labeled firm links from New York Times articles

Time span: January 1981 through December 2023.

Frequency: Monthly.

Link types: Credit, financing, peer, supply chain, and other links.

Network Data: Google Drive folder. 

Training Data: HS_chatgpt_sentences.csv. 

Related papers:

• "The Different Networks of Firms Implied by the News" (with V. Hilt). Online Appendix. Codes.

• "The Network of Firms Implied by the News" (with H. Zheng). Online Appendix. Codes. 

These data contain links between firms as reported by The New York Times. The links are categorized as credit, non-credit financing, peer, supply chain, and other links. The data set contains the following monthly files (where YYYYMM stands for the corresponding month and TYPE stands for the type of network):

• nytimes_labeled_links_YYYYMM.csv (Labeled links between two firms)

• nytimes_network_plot_TYPE_YYYMM.png (Plot for network of a given type)

For details, see the Online Appendix. 

Firm links from New York Times articles

Time span: January 1981 through December 2023

Frequency: Monthly

Link types: Unlabelled.

Data: Google Drive folder. 

Related papers:

• "The Network of Firms Implied by the News" (with H. Zheng). Online Appendix. Codes. 

• "The Different Networks of Firms Implied by the News" (with V. Hilt). Online Appendix. Codes.

These data contain links between firms as reported by The New York Times. The links are not categorized (i.e., they do not distinguish whether a link between two firm is competitive or credit in nature). The data set contains the following monthly files (where YYYYMM stands for the corresponding year and month):

• nytimes_entities_YYYYMM.csv (Total number of mentions for each identified firm)

• nytimes_firm_mentions_YYYYMM.csv (Sentences in which a firm was identified) 

• nytimes_sentences_with_2links_YYYYMM.csv (Sentences with exactly two firm mentions)

• nytimes_sentences_with_links_YYYYMM.csv (Sentences with at least two firm mentions)

• nytimes_network_YYYYMM.csv (Network implied by sentences with exactly two firm mentions)

• nytimes_network_plot_YYYYMM.png (Network plots)

For details, see the Online Appendix. 

Firm links from Reuters articles

Time span: Oct 2006 through Nov 2013

Frequency: Monthly

Link types: Unlabelled.

Data: Google Drive folder. 

Related papers:

• "The Network of Firms Implied by the News" (with H. Zheng). Online Appendix. Codes. 

These data contain links between firms as reported by Reuters. The links are not categorized (i.e., they do not distinguish whether a link between two firm is competitive or credit in nature). The data set contains the following monthly files (where YYYYMM stands for the corresponding year and month):

• reuters_entities_YYYYMM.csv (Total number of mentions for each identified firm)

• reuters_firm_mentions_YYYYMM.csv (Sentences in which a firm was identified) 

• reuters_sentences_with_2links_YYYYMM.csv (Sentences with exactly two firm mentions)

• reuters_sentences_with_links_YYYYMM.csv (Sentences with at least two firm mentions)

• reuters_network_YYYYMM.csv (Network implied by sentences with exactly two firm mentions)

• reuters_network_plot_YYYYMM.png (Network plots)

For details, see the Online Appendix. 

Crypto peer links from Cryptocompare articles

Time span: October 1, 2017, through November 30, 2020.

Frequency: Weekly. 

Link types: Peer.

Data: Google Drive folder. 

Related papers:

• "News-Driven Peer Co-Movement in Crypto Markets" (with H. Zheng). Online Appendix.

These data contain competitive links between cryptocurrencies as reported by Cryptocompare. The data set contains the following files:

• crypto_peers_training_sentences.csv: training data for deep learning classification model. 

• crypto_peers_all_link_sentences.csv: output of classification model containing all sentences that describe peer relationships between cryptocurrencies. 

• crypto_peers_weekly_peer_weekly_network.csv: weekly adjecency matrix for crypto peer network. 

• crypto_peers_exogenous_endogenous.xlsx: file characterizing shock events as endogenous or exogenous.

In addition, the folder also contains codes that were used to: 1) scrape online news articles, 2) extract crypto mentions from online news articles, and 3) label sentences mentioning two cryptos as describing a competitive or non-competitive relationship.