A carregar...
Publicação
Corporate reports and the changing sentiment of economic power: Text mining your way into the soul of capitalism
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.56 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This thesis explores the potential of financial sentiment extracted from corporate disclosures as
an indicator of industry-level economic performance. Annual filings (10-K form) submitted in
2003-2021, as mandated by the U.S. Securities and Exchange Commission (SEC), were
downloaded from the EDGAR database and sentiment analysis was run on them using
FinBERT. These sentiment scores were subsequently aggregated at the industry level and
aligned with macroeconomic and sector-specific indicators from BLS, FRED and BEA
databases. The resulting panel dataset was utilized for conducting fixed-effects regression
analysis across multiple time lags to test predictive power of sentiment scores. The findings
indicate that sentiment scores are associated with BLS-derived sector performance indicators,
but no significant positive association was found with BEA-sourced industry performance data.
The thesis provides empirical evidence supporting the value of text mining in broader
macroeconomic analysis and early-warning frameworks.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Driven Marketing, specialization in Marketing Research and CRM
Palavras-chave
Text mining sentiment analysis FinBERT 10-K filings real sectoral output SDG 8 - Decent work and economic growth SDG 16 - Peace, justice and strong institutions