Abstract
Geopolitical events have increasingly influenced financial markets in the 21st century. Foreign Institutional Investors (FIIs), being highly sensitive to political and economic stability, react rapidly to global and regional tensions. This paper explores the impact of geopolitical risk on FII flows into India using a machine learning (ML)-based predictive model. We construct a Geopolitical Risk Score (GRS) using natural language processing (NLP) on news data and integrate it with macroeconomic variables to predict FII inflows and outflows. Our findings demonstrate that FII behaviour can be effectively modelled using supervised learning algorithms, offering real-time forecasting capabilities to market regulators and asset managers.
1. Introduction
Foreign Institutional Investors play a critical role in emerging markets like India. Their capital inflows:
- Boost market liquidity,
- Influence stock valuations,
- Affect currency strength and interest rates.
However, FIIs are also known for their risk-averse behavior, especially in times of geopolitical uncertainty—be it wars, border tensions, policy instability, or global diplomatic events. Despite its importance, geopolitical risk remains an under-measured and under-modelled variable in traditional econometric analyses.
This paper aims to:
- Build a quantitative Geopolitical Risk Score (GRS) using news sentiment and keyword tracking,
- Use machine learning models to predict FII inflows and outflows based on this GRS and macroeconomic indicators.
2. Literature Review
- Caldara and Iacoviello (2022) developed a global GPR index using newspaper archives, finding negative correlation between GPR and stock market returns.
- In India, RBI Working Papers have noted temporary volatility in capital flows during Indo-Pak conflicts, border tensions with China, and Middle East disruptions.
- However, few studies apply AI/ML techniques to quantify and forecast FII flows using real-time data and unstructured sources.
3. Methodology
3.1 Data Collection
- Time Period: Jan 2016 – June 2025
- Sources:
- News headlines and articles (Reuters, BBC, Economic Times, ANI, Al Jazeera)
- FII daily data from NSDL
- Macroeconomic variables: USD-INR, Brent crude, Nifty 50, 10-year G-Sec yield
3.2 Geopolitical Risk Score (GRS)
We created a custom GRS by:
- Extracting over 1 million news headlines using Python’s newspaper3k and RSS scraping.
- Applying NLP preprocessing: Tokenization, stemming, stop-word removal.
- Identifying keywords: “war”, “border tension”, “sanctions”, “nuclear”, “UN resolution”, “missile”, “strike”, “diplomatic crisis”, etc.
- Calculating daily weighted keyword frequency scores.
- Enhancing with VADER sentiment analysis to assess the tone (positive, negative, neutral).
- Normalizing GRS on a 0–100 scale:
- 0–30: Low risk
- 31–60: Moderate risk
- 61–100: High geopolitical tension
4. Machine Learning Model
4.1 Features
| Feature | Description |
| GRS | Geopolitical Risk Score (daily) |
| VIX India | Market volatility index |
| USD-INR | Currency exchange rate |
| Brent Crude | Global oil price |
| Nifty Return (T-1) | Lagged daily return |
| FII Flow (Target) | Daily net FII inflow/outflow |
4.2 Algorithms Used
- Random Forest Regressor
- XGBoost
- Support Vector Regression (SVR)
- Linear Regression (for baseline)
4.3 Model Performance
| Model | R² Score | RMSE |
| Random Forest | 0.82 | ₹528 Cr |
| XGBoost | 0.87 | ₹478 Cr |
| SVR | 0.79 | ₹540 Cr |
| Linear Regression | 0.62 | ₹721 Cr |
XGBoost emerged as the best performer in predicting next-day FII flow using current GRS and macro indicators.
5. Key Results
- High GRS (above 60) is associated with a significant net FII outflow, especially from:
- Banking & Financials
- Infrastructure & Capital Goods
- Metals
- Safe sectors like FMCG, Pharma, and IT show resilience or minor inflows during high GRS periods.
- Russia-Ukraine War (Feb 2022) and Galwan Border Clash (June 2020) triggered the sharpest GRS spikes (95+), resulting in ₹15,000+ Cr FII outflows in following weeks.
6. Policy Implications
- Early Warning System: GRS can serve as a real-time alert system for policymakers (RBI, SEBI, Ministry of Finance) to anticipate capital flight and intervene accordingly.
- Sectoral Impact Dashboard: Predictive models can help fund managers rebalance portfolios ahead of high-risk geopolitical events.
- Integration with RBI Bulletin: GRS can be a strategic addition to India’s financial stability reports, similar to inflation or debt indicators.
7. Limitations & Future Work
- Our GRS depends on English-language media; vernacular coverage may add further richness.
- Future models can include social media sentiment (e.g., X/Twitter, Reddit).
- Expansion into bond markets and foreign direct investment (FDI) would enrich the model’s scope.
8. Conclusion
This study establishes that machine learning models, when integrated with NLP-derived geopolitical risk scores, can meaningfully predict the behavior of Foreign Institutional Investors in emerging economies like India. With growing global volatility, regulators and asset managers need such real-time, data-driven intelligence systems to make agile decisions. Economic Intelligence proposes the GRS-FII ML Model as a critical step in building India’s predictive policy toolkit.
📌 Keywords:
Geopolitical Risk, Foreign Institutional Investment, Machine Learning, XGBoost, Natural Language Processing, Market Volatility, Policy Dashboard