Paper: arXiv 2504.13189

Authors: Sohom Ghosh, Sudip Kumar Naskar

Abstract

Government fiscal policies, particularly annual union budgets, exert significant influence on financial markets. However, real-time analysis of budgetary impacts on sector-specific equity performance remains methodologically challenging and largely unexplored. This study proposes a framework to systematically identify and rank sectors poised to benefit from India’s Union Budget announcements. The framework addresses two core tasks: (1) multi-label classification of excerpts from budget transcripts into 81 predefined economic sectors, and (2) performance ranking of these sectors. Leveraging a comprehensive corpus of Indian Union Budget transcripts from 1947 to 2025, we introduce BASIR (Budget-Assisted Sectoral Impact Ranking), an annotated dataset mapping excerpts from budgetary transcripts to sectoral impacts. Our architecture incorporates fine-tuned embeddings for sector identification, coupled with language models that rank sectors based on their predicted performances. Our results demonstrate 0.605 F1-score in sector classification, and 0.997 NDCG score in predicting ranks of sectors based on post-budget performances. The methodology enables investors and policymakers to quantify fiscal policy impacts through structured, data-driven insights, addressing critical gaps in manual analysis. The annotated dataset has been released under CC-BY-NC-SA-4.0 license to advance computational economics research.

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper uses advanced NLP techniques and statistical metrics like F1-score and NDCG, but the math is mostly applied ML rather than dense theoretical derivations. It demonstrates high empirical rigor with a specific dataset (1,671 entries), concrete backtesting-ready metrics (F1=0.605, NDCG=0.997), and clear data sources, making it practical for implementation.

Research Flowchart

  flowchart TD
  A["Research Goal: Systematic<br>Fiscal Policy Impact Analysis"] --> B["Input: BASIR Dataset<br>Budget Transcripts 1947-2025"]
  B --> C["Core Methodology"]
  subgraph C ["Two-Tier Pipeline"]
      C1["Fine-tuned Embeddings<br>Multi-label Sector Classification"] --> C2["Language Models<br>Performance Ranking NDCG"]
  end
  C --> D["Key Outcomes"]
  subgraph D ["Quantified Results"]
      D1["Class. F1-Score: 0.605"]
      D2["Rank. NDCG: 0.997"]
      D3["Dataset Released<br>CC-BY-NC-SA-4.0"]
  end

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a framework that identifies and ranks sectors likely to benefit from India’s Union Budget announcements, and introduces the BASIR annotated dataset linking budget excerpts to sector impacts.

Source sentence
This study proposes a framework to systematically identify and rank sectors poised to benefit from India's Union Budget announcements.

Assumes 5

  • Sectors are ranked based on expected next‑day post‑announcement returns computed from opening price changes of constituent companies. data
    Source sentence
    For identified sector set ˆS = {sj | P(sj|t) > τ}, develop a model f : ˆS →R+ that ranks sectors by expected next day post‑announcement returns rs using text excerpts t related to the sector s_j.
  • A probabilistic threshold τ is applied to sector relevance probabilities P(sj|t) to select sectors for ranking. model
    Source sentence
    For identified sector set ˆS = {sj | P(sj|t) > τ}, develop a model f :
  • Training uses data up to 2019, validation uses 2020‑2023, and testing uses 2024 budget data. data
    Source sentence
    Data till the year 2019 was used for training, data spanning 2020 to 2023 was allocated for validation, and 2024 data was reserved for testing.
  • Sector‑company mapping from Screener.in accurately reflects sector constituents. data
    Source sentence
    Sector-Company Mapping: We systematically collected a list of sectors and their constituent companies from Screener.in.
Show 1 more
  • Manual validation of DeepSeek‑extracted sector tags ensures correct labeling. other
    Source sentence
    Validation: We manually validated all the outputs.

Data

  • BASIR (Budget‑Assisted Sectoral Impact Ranking) dataset · Authors (compiled from Indian Union Budget transcripts) text-news public
    1947‑2025 · 81 predefined economic sectors
    Source sentence
    We introduce BASIR (Budget‑Assisted Sectoral Impact Ranking), an annotated dataset mapping excerpts from budgetary transcripts to sectoral impacts.
  • Union Budget transcripts · India’s Ministry of Finance portal text-news public
    1947‑2025
    Source sentence
    Aggregated 97 Union Budget documents (1947–2025) from India’s Ministry of Finance portal
  • Sector‑Company Mapping · Screener.in fundamentals public
    Source sentence
    Sector-Company Mapping: We systematically collected a list of sectors and their constituent companies from Screener.in.
  • Historical price data · Yahoo Finance market-prices public
    companies constituting the 81 sectors
    Source sentence
    historical data sourced from yahoo finance.

Recognised sources: Yahoo Finance

Costs

Not applicable The paper reports no trading or execution result.

Finds 2

  • The sector classification model achieves a 0.605 F1‑score. predictive-accuracy
    Source sentence
    Our results demonstrate 0.605 F1-score in sector classification
  • The sector ranking model attains a 0.997 NDCG score on post‑budget performance prediction. predictive-accuracy
    Source sentence
    and 0.997 NDCG score in predicting ranks of sectors based on post‑budget performances

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 14 related papers). Coverage is partial, so this is neither replication nor refutation.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (27)

  1. The Theory of Parsing, Translation and Compiling (1972)
  2. Publications Manual (1983)
  3. Alternation (1981) — link
  4. Scalable training of L1 -regularized log-linear models (2007)
  5. Algorithms on Strings, Trees and Sequences (1997)
  6. Yara Parser: A Fast and Accurate Dependency Parser (2015) — link
  7. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (2025) — link
  8. Impact of budget and GDP announcements on Indian stock market (2019)
and 19 more
  1. Stock Price Reactions to Budget Announcement in Indian Capital Market (2021)
  2. Effects of budget announcement on stock prices in the Indian context (2023)
  3. ANALYZING THE IMPACT OF THE UNION BUDGET ON SECTORAL INDICES IN THE NATIONAL STOCK EXCHANGE (NSE). (2024)
  4. Impact of Union Budget on Stock Market (2018)
  5. A Textual Data Analysis of the Union Budget of India (2024)
  6. Measuring economic policy uncertainty (2016)
  7. Social Listening on Budget—A Study of Sentimental Analysis and Prediction of Sentiments Using Text Analytics & Predictive Algorithms (2022)
  8. Stock market trend analysis on Indian financial news headlines with natural language processing (2021)
  9. Words that Move Markets-Quantifying the Impact of RBI’s Monetary Policy Communications on Indian Financial Market (2024) — link
  10. How do fiscal policy decisions influence business confidence? Case study: the manufacturing sector (2015)
  11. BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding (2019) — link
  12. Nomic Embed: Training a Reproducible Long Context Text Embedder (2024) — link
  13. Deberta: Decoding-enhanced bert with disentangled attention (2020) — link
  14. XGBoost : A Scalable Tree Boosting System (2016) — link
  15. RoBERTa: A Robustly Optimized BERT Pretraining Approach (2019)
  16. DeepSeek-V3 Technical Report (2025) — link
  17. LLaMA: Open and Efficient Foundation Language Models (2023) — link
  18. BERTopic: Neural topic modeling with a class-based TF-IDF procedure (2022) — link
  19. Gemma 3 (2025) — link

0 of 27 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.