Evaluating information content of earnings calls to predict bankruptcy using machine learnings techniques
Bibliographic record
Abstract
This study investigates the prediction of firms’ health in terms of bankruptcy and non-bankruptcy based on the sentiments extracted from the earnings calls. Bankruptcy prediction has long been a critical topic in the world of accounting and finance. A firm's economic health is the current financial condition of the firm and is crucial to its stakeholders such as creditors, investors, shareholders, partners, and even customers and suppliers. Various methodologies and strategies have been proposed in research domain for predicting company bankruptcy more promptly and accurately. Conventionally, financial risk prediction has solely been based on historic financial data. However, an increasing number of finance papers also analyze textual data during the last few years. Company’s earnings calls are the key source of information to investigate the current financial condition and how the businesses are doing and what the expectations are for the next quarters. During the call, management offers an overview of recent performance and provide a guidance for the next quarter expectations. The earnings calls summary is provided by the management and can extract the CEO’s sentiments using sentiment analysis. In the last decade, Machine Learnings based techniques have been proposed to achieve accurate predictions of firms’ economic health. Even though most of these techniques work well in a limited context, on a broader perspective these techniques are unable to retrieve the true semantic from the earnings calls, which result in the lower accuracy in predicting the actual condition of firms’ economic health. Thus, state-of-the-art Machine Learnings and Deep Learnings techniques have been used in this thesis to improve accuracy in predicting the firms’ health from the earnings calls. Various machine learnings and deep learnings method have been applied on web-scraped earnings calls data-set, and the results show that LONG SHORT-TERM MEMORY (LSTM) is the best machine learnings technique as compared to the comparison set of models.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".