Financial Econometric Project
In the following you can find the detailed description of the files in the repository.
Report.pdf:
- detailed report of what we did in our project.
Folder ResultsPolarity:
- sentiment_evolution19_22: excel file with the evolution of the ESG-related average compound score using polarity approach.
- sentiment_evolution2022: excel file with the ESG-related average compound score of 2022 using polarity approach.
- evolution_bank/car/energy/food: plot of the evolution of companies' average compound score industry by industry
- words_/bank/car/energy/food: word clouds of positive and negative words that are both in the Loughran-McDonald list and in the ESG-related reports.
- change_per_company: percentage change from 2019 to 2022 of companies' average compound score
- change_per_industry: table with the average change within every industry of the percentage change from 2019 to 2022 of companies' average compound score
Folder ResultsVADER:
- sentiment_evolution19_22: excel file with the evolution of the ESG-related average compound score using VADER approach.
- sentiment_evolution2022: excel file with the ESG-related average compound score of 2022 using VADER approach.
- evolution_bank/car/energy/food: plot of the evolution of companies' average compound score industry by industry
- change_per_company: percentage change from 2019 to 2022 of companies' average compound score
- change_per_industry: table with the average change within every industry of the percentage change from 2019 to 2022 of companies' average compound score
- evolution_all_companies: plot of the evolution of all companies' average compound score
- sentiment_per_document2022: companies' 2022 compound score resulting from applying VADER to the entire ESG-reports instead of doing in sentence by sentence
- sentiment_statistics2022: mean and standard deviation (both per industry and in general) of companies' 2022 average compound score
Folder ResultsBERT:
- Sentimental_Bert_on_10K: csv file with the evolution of the ESG-related score using BERT approach.
- evolution_bank/car/energy/food: plot of the evolution of companies' average compound score industry by industry
- change_per_company: percentage change from 2019 to 2022 of companies' average compound score
- change_per_industry: table with the average change within every industry of the percentage change from 2019 to 2022 of companies' average compound score
- evolution_all_companies: plot of the evolution of all companies' average compound score
- sentiment_per_document2022: companies' 2022 compound score resulting from applying VADER to the entire ESG-reports instead of doing in sentence by sentence
Folder 10-K_time_2019-2022:
- collection of 10K-reports from 2019 to 2022 of all the companies selected except Rivian and Royale Energy Inc.
Folder 10-KLast:
- collection of 2022 10K-reports of all the companies selected.
Folder Lists:
- economy/solcial/environmental: lists provided by the outhors of the article "Sustainability and Corporate Social Responsibility in the Text of Annual Reports—The Case of the IT Services Industry"
- LM_MasterDictionary : Loughran-McDonald list
Sentiment_vader_polarity.ipynb:
- complete collection of data and analysis for Vader and polarity approaches
- complete analysis for Bert approach
Folder To_run_Sentiment_vader_polarity:
-everything you need to run Sentiment_vader_polarity.ipynb
Sentiment_ecological_Bert.ipynb:
- to replicate our 3rd approach (Bert to recognise Ecological sentiment) you need this code plus the file ecological_bert (support code);it is possible to change the document in which the analysis is run in ecological_bert/src/predict_historical_10K. Results are in RESULTS folder.
Sentiment_FinBert.ipynb:
- to replicate our 4th approach (FinBert for sentiment analysis) you need this code. Results are in RESULTS folder.
Select_ESG_sentences_via_Bert.ipynb:
- to replicate the selection of ESG sentences via BERT you need this code. One may need adaptation regarding the path of the 10_k reports within the notebooks. Results are in RESULTS folder.
Folder Data_analysis_summary_code:
- code of Data_analysis
- merged by industry reports(pre-processed and not pre-processed) (8 .txt files to run the code)
Folder Regression:
- folder 1: code for regression and data necessary to run the code
- folder 2: second code for regression (necessary data to run this code are all in Folder To_run_Sentiment_vader_polarity)
- Msci scores: necessary to run the two above codes for regressions
