Geopolitical Risk Analysis

GPR: overall geopolitical risk
GPRT: geopolitical threats (forward-looking)
GPRA: geopolitical acts (events already realized)
GPR_country: country-level sub-index

import pandas as pd
import requests

def load_gpr_index():
    """Load the official GPR Index data.

    Returns:
        pd.DataFrame: Monthly GPR data with columns such as GPR, GPRT, and GPRA.
    """
    url = "https://www.matteoiacoviello.com/gpr_files/data_gpr_export.xls"
    df = pd.read_excel(url, index_col=0, parse_dates=True)
    return df

def gpr_signal(df, window=12, threshold=1.5):
    """Generate abnormal GPR signals.

    Args:
        df: DataFrame containing GPR data
        window: Rolling mean window in months
        threshold: Z-score trigger threshold in standard deviations

    Returns:
        pd.Series: Boolean signal where True means high-risk state
    """
    gpr = df["GPR"]
    rolling_mean = gpr.rolling(window).mean()
    rolling_std = gpr.rolling(window).std()
    z_score = (gpr - rolling_mean) / rolling_std
    return z_score > threshold

def oil_war_premium(spot_price, mean_5y_price, supply_disruption_prob,
                    disruption_magnitude_pct):
    """Estimate the war-risk premium embedded in crude oil.

    Method:
        A simplified model based on expected supply-disruption value.

    Args:
        spot_price: Current spot price in USD/bbl
        mean_5y_price: Five-year average price as the "no-risk" baseline
        supply_disruption_prob: Probability of supply disruption in [0, 1]
        disruption_magnitude_pct: Price impact of disruption in [0, 1]

    Returns:
        float: Estimated war premium in USD/bbl
    """
    expected_disruption_premium = (
        mean_5y_price * disruption_magnitude_pct * supply_disruption_prob
    )
    observed_premium = spot_price - mean_5y_price
    return max(0, min(observed_premium, expected_disruption_premium))

def gold_geopolitical_premium(gold_price, real_yield_10y, usd_index):
    """Decompose the geopolitical premium component in gold prices.

    Args:
        gold_price: Spot gold price in USD/oz
        real_yield_10y: 10-year real yield in percent
        usd_index: DXY index

    Returns:
        float: Geopolitical premium as the residual component
    """
    import numpy as np
    # Gold fundamentals: real rates (negative) + USD (negative)
    # Linear approximation:
    # Gold ≈ α - β1*RealYield - β2*DXY + ε (geopolitical premium)
    # β1 ≈ 800, β2 ≈ 15 are rough historical estimates that should be updated
    fundamental_value = 2000 - 800 * real_yield_10y - 15 * (usd_index - 100)
    return gold_price - fundamental_value

def update_disruption_probability(prior_prob, new_event_severity, base_rate=0.05):
    """Update supply-chain disruption probability using a new event.

    This is a simplified Bayesian update that adjusts the prior
    using the severity of the new event.

    Args:
        prior_prob: Prior disruption probability
        new_event_severity: Event severity in [0, 1]
            0.0 = diplomatic friction
            0.3 = military standoff
            0.6 = local conflict
            1.0 = full-scale war
        base_rate: Historical annualized baseline disruption rate

    Returns:
        float: Updated disruption probability
    """
    # Likelihood ratio: how much more likely the event is before a real disruption
    # than in a non-disruption state
    likelihood_ratio = 1 + 9 * new_event_severity  # 1x ~ 10x
    posterior = (prior_prob * likelihood_ratio) / (
        prior_prob * likelihood_ratio + (1 - prior_prob)
    )
    return posterior

Gold (GLD/GC): geopolitical shock → immediate rally, but persistence depends on real-rate direction
Silver (SLV/SI): industrial exposure dilutes safe-haven behavior and raises volatility
Palladium / platinum: Russia is a major producer, so sanctions hit supply directly

Estimated impact under a Taiwan Strait crisis:
- Mild military tension (drills): SOX -5% to -10%
- Blockade drill (1 month): SOX -15% to -25%
- Actual military conflict: SOX -40% to -60% (no true historical analogue)

Beneficiaries through substitution:
- Intel (INTC): IDM model with U.S.-based capacity
- GlobalFoundries (GFS): U.S. / Europe / Singapore capacity
- Samsung, though Korea itself is also a geopolitical risk zone

Key ETFs and stocks:
- BDRY: bulk-shipping freight ETF tracking BDI, highly sensitive to Red Sea / Hormuz shocks
- ZIM: Israeli container shipper, directly exposed to Red Sea risk
- FRO (Frontline): tanker beneficiary of Hormuz risk
- STNG (Scorpio Tankers): benefits from rerouting around the Red Sea
- MAERSK.B: container-shipping leader that benefits from freight spikes during crises

U.S. defense ETFs: ITA (iShares), XAR (SPDR)

Single-stock beneficiaries of geopolitical risk:
- LMT (Lockheed Martin): F-35, missile systems
- RTX (Raytheon): air-defense systems such as Patriot
- NOC (Northrop Grumman): B-21 bomber, nuclear systems
- BA (Boeing): military exposure, though commercial aviation can be hurt by geopolitics

Historical pattern:
Higher geopolitical risk → faster defense budget approvals → effect shows up with a 6-12 month lag

Capital flows during crises:
Risk currencies (AUD/NZD/MXN/KRW/BRL) → outflows
Safe-haven currencies (JPY/CHF/USD) ← inflows

JPY:
- Net-creditor-nation status + repatriation effect
- Historical crisis moves: +1% to +3% vs USD

CHF:
- Neutral country + European financial center
- Major crises: +2% to +5% vs EUR

USD:
- Global reserve currency and final safe haven during crises
- But if the U.S. homeland is directly attacked, USD can weaken instead

Note: High-carry funding currencies such as TRY and ARS tend to suffer the most when global risk aversion rises

SIGNAL_LEVELS = {
    "GREEN": {
        "desc": "Normal geopolitical risk level",
        "gpr_percentile": (0, 50),
        "action": "Standard allocation, no special hedge required"
    },
    "YELLOW": {
        "desc": "Risk rising, watch for escalation",
        "gpr_percentile": (50, 75),
        "action": "Small long-gold position, reduce high-risk asset exposure by 10%"
    },
    "ORANGE": {
        "desc": "High-risk state, potential shock approaching",
        "gpr_percentile": (75, 90),
        "action": "Add safe-haven assets, buy OTM protective options, bullish on oil"
    },
    "RED": {
        "desc": "Extreme risk, crisis may break out",
        "gpr_percentile": (90, 100),
        "action": "Maximize defensive positioning, hold cash / gold / Treasuries, short high-risk assets"
    }
}

Diplomatic:
  [ ] Embassy closures / downgrades
  [ ] Diplomat expulsions
  [ ] UN emergency meeting called
  [ ] Escalation in joint statements by multiple countries

Military:
  [ ] Large-scale exercises (>50,000 personnel)
  [ ] Carrier strike group forward deployment
  [ ] Higher readiness announcements
  [ ] Missile / nuclear system release orders

Financial:
  [ ] Target-country CDS spread breaks historical highs
  [ ] Exchange rate devaluation >3% in one week
  [ ] Sharp decline in FX reserves
  [ ] Accelerating capital flight

def crisis_vol_strategy(underlying, option_chain):
    """Volatility trading framework during crises.

    Crisis outbreaks usually cause:
    1. A short-term VIX spike (long VIX futures / options)
    2. Inversion in the IV term structure (front month > back month)
    3. Steeper put skew

    Args:
        underlying: Underlying asset ticker
        option_chain: Option chain data

    Returns:
        dict: Recommended strategies and sizing guidance
    """
    strategies = {
        "long_vix_futures": {
            "instrument": "Front-month VX futures",
            "trigger": "VIX < 20 and GPR > 75th percentile",
            "target": "VIX spikes to 35-50",
            "stop": "VIX falls 15% below entry"
        },
        "backspread": {
            "instrument": f"Buy OTM Put + sell ATM Put on {underlying}",
            "trigger": "Implied volatility is at a historical low",
            "profit_zone": "Large drop > 10%"
        },
        "calendar_spread": {
            "instrument": "Sell near-month ATM + buy far-month ATM",
            "trigger": "Exit when term-structure inversion becomes excessive",
            "profit_zone": "Volatility mean reversion"
        }
    }
    return strategies

Crisis type        | Gold | Oil | Defense | JPY | Treasuries | EM
Energy conflict    | ++   | +++ | ++      | +   | +          | ---
Nuclear escalation | +++  | +   | +       | +++ | +++        | ---
Sanctions / trade  | +    | +   | +       | +   | +          | --
Food crisis        | +    | 0   | 0       | 0   | +          | -- (importers)
Sea blockade       | +    | ++  | +       | +   | +          | -

1. The initial equity shock from geopolitical events usually recovers within 30 days unless recession hits simultaneously
2. Energy / commodities effects last longer because supply-side changes are structural
3. Go long the most damaged assets once the crisis de-escalates and mean reversion starts
4. Sell safe-haven assets that exploded during the crisis, especially gold after tension fades

REVERSION_SIGNALS = [
    "Ceasefire agreement signed / negotiations announced",
    "Energy / grain exports resume, confirmed by shipping data",
    "Target-country CDS spreads retrace >20% from the peak",
    "GPR Index falls >30% from the peak",
    "VIX drops below 20 after peaking",
    "Safe-haven currencies such as JPY / CHF begin weakening"
]

# Official download, free, monthly updates
GPR_DATA_URL = "https://www.matteoiacoviello.com/gpr_files/data_gpr_export.xls"

# High-frequency daily GPR based on Twitter / news
# Access request required: https://www.policyuncertainty.com/gpr_daily.html

# Related paper:
# Caldara & Iacoviello (2022), "Measuring Geopolitical Risk"
# American Economic Review, 112(4): 1194-1225

# GDELT 2.0 provides global news-event data updated every 15 minutes
# Includes the CAMEO event code system for military / diplomatic / conflict classification

def query_gdelt_events(country_code, event_type, start_date, end_date):
    """Query GDELT geopolitical event data.

    GDELT BigQuery table: gdelt-bq.gdeltv2.events
    CAMEO root codes: 14=protest, 18=assault, 19=fight, 20=mass violence

    Args:
        country_code: FIPS country code, e.g. 'CH' for China, 'RS' for Russia
        event_type: CAMEO root code
        start_date: Start date in YYYY-MM-DD
        end_date: End date in YYYY-MM-DD

    Returns:
        pd.DataFrame: Event records
    """
    from google.cloud import bigquery
    client = bigquery.Client()

    query = f"""
    SELECT SQLDATE, Actor1CountryCode, Actor2CountryCode,
           EventCode, GoldsteinScale, NumMentions, AvgTone
    FROM `gdelt-bq.gdeltv2.events`
    WHERE (Actor1CountryCode = '{country_code}'
           OR Actor2CountryCode = '{country_code}')
      AND EventRootCode = '{event_type}'
      AND SQLDATE BETWEEN '{start_date.replace('-','')}'
                      AND '{end_date.replace('-','')}'
    ORDER BY SQLDATE DESC
    """
    return client.query(query).to_dataframe()

# Armed Conflict Location & Event Data Project
# https://acleddata.com/
# Covers 100+ countries and is free for approved academic access

ACLED_API_BASE = "https://api.acleddata.com/acled/read"

def fetch_acled_events(country, start_date, end_date, api_key):
    """Fetch ACLED armed-conflict event data.

    Args:
        country: Country name in English
        start_date: Start date in YYYY-MM-DD
        end_date: End date in YYYY-MM-DD
        api_key: ACLED API key

    Returns:
        pd.DataFrame: Conflict event data
    """
    import requests
    import pandas as pd

    params = {
        "key": api_key,
        "email": "your@email.com",
        "country": country,
        "event_date": f"{start_date}|{end_date}",
        "event_date_where": "BETWEEN",
        "export_type": "json"
    }
    resp = requests.get(ACLED_API_BASE, params=params)
    return pd.DataFrame(resp.json()["data"])

# Option A: Use the Jina Reader API integrated in the project through read_url
def analyze_geopolitical_news(query: str) -> dict:
    """Read news through Jina and analyze geopolitical-risk sentiment.

    Use together with the agent's read_url tool.

    Args:
        query: Search keywords

    Returns:
        dict: Sentiment-analysis result
    """
    # Recommended news sources:
    news_sources = [
        "https://www.reuters.com/world/",
        "https://www.bloomberg.com/politics",
        "https://www.ft.com/world",
        "https://www.foreignpolicy.com/"
    ]
    # Use read_url to fetch content, then pass it to the LLM to extract risk events

# Option B: Event Registry API (paid, structured news)
# https://eventregistry.org/
# Supports filtering by country / topic / time and returns standardized events

# Option C: VADER / FinBERT sentiment analysis
# Score geopolitical news sentiment and build high-frequency signals

DATA_SOURCES = {
    "oil_tanker_tracking": {
        "desc": "Crude oil / LNG vessel AIS tracking",
        "source": "MarineTraffic API (paid) / VesselFinder (limited free)",
        "use_case": "Real-time monitoring of traffic through Hormuz / the Red Sea"
    },
    "un_vote_data": {
        "desc": "UN General Assembly / Security Council voting records",
        "source": "UN Data API (free)",
        "use_case": "Track changes in great-power alignment"
    },
    "arms_transfer": {
        "desc": "Arms transfers and military aid data",
        "source": "SIPRI Arms Transfers Database (free)",
        "use_case": "Estimate conflict-escalation probability"
    },
    "nuclear_risk": {
        "desc": "Real-time nuclear-risk assessment",
        "source": "Bulletin of the Atomic Scientists Doomsday Clock",
        "use_case": "Tail-risk monitoring"
    },
    "commodity_futures": {
        "desc": "Commodity futures prices, including geopolitical premium",
        "source": "Integrated in this project: Tushare commodity futures / OKX crypto",
        "use_case": "Estimate war premium"
    }
}

Run at the start of each month:
1. Download the latest GPR Index data
2. Calculate CDS spread changes for each hotspot country
3. Analyze tanker insurance rates
4. Summarize counts of high-intensity GDELT conflict events
5. Output a composite risk score (0-100) plus allocation guidance

Trigger:
Major geopolitical event breaks out, such as a missile strike or sanctions announcement

Execution flow:
1. Identify the event type and intensity (0-10)
2. Map the affected asset classes
3. Estimate the short-term price shock range
4. Identify hedging instruments (options / futures / ETFs)
5. Set stop-loss rules and position size

# Geopolitical scenario stress tests for a portfolio
SCENARIOS = {
    "hormuz_blockade_30d": {
        "oil_price_shock": +40,
        "gold_shock": +8,
        "equity_shock": -12,
        "usd_shock": +3,
        "description": "30-day Strait of Hormuz blockade scenario"
    },
    "taiwan_conflict_mild": {
        "semioconductor_shock": -25,
        "gold_shock": +5,
        "equity_shock": -15,
        "jpy_shock": +8,
        "description": "Mild Taiwan Strait military conflict scenario"
    },
    "russia_gas_cutoff": {
        "eu_natgas_shock": +80,
        "eu_equity_shock": -20,
        "eur_shock": -8,
        "gold_shock": +6,
        "description": "Russia fully cuts off gas to Europe"
    }
}

def portfolio_stress_test(portfolio_weights, scenarios=SCENARIOS):
    """Run geopolitical scenario stress tests on a portfolio.

    Args:
        portfolio_weights: dict mapping asset ticker to weight
        scenarios: Scenario-definition dictionary

    Returns:
        pd.DataFrame: Expected portfolio PnL under each scenario
    """
    results = {}
    for scenario_name, shocks in scenarios.items():
        portfolio_pnl = sum(
            portfolio_weights.get(asset, 0) * shock / 100
            for asset, shock in shocks.items()
            if asset != "description"
        )
        results[scenario_name] = {
            "portfolio_return": portfolio_pnl,
            "description": shocks["description"]
        }
    return results

# Strategy logic:
# When GPR > 75th percentile, hold 5% gold + 5% oil calls
# When GPR < 25th percentile, revert to standard allocation
# Historical backtests suggest a roughly 30-40% reduction in tail losses
# across major crises from 2001-2023

def gpr_dynamic_hedge_backtest(returns_data, gpr_data,
                                hedge_assets=["GLD", "USO"],
                                hedge_weight=0.05):
    """Backtest a GPR-driven dynamic hedge strategy.

    Args:
        returns_data: pd.DataFrame of daily asset returns
        gpr_data: pd.Series of monthly GPR Index values
        hedge_assets: List of hedge assets
        hedge_weight: Allocation weight per hedge asset

    Returns:
        pd.DataFrame: Return comparison before and after hedging
    """
    import pandas as pd

    # Map monthly GPR to daily frequency.
    gpr_daily = gpr_data.resample("D").ffill()
    gpr_threshold = gpr_daily.quantile(0.75)

    hedge_signal = gpr_daily > gpr_threshold

    base_return = returns_data.drop(columns=hedge_assets, errors="ignore").mean(axis=1)
    hedge_return = returns_data[hedge_assets].mean(axis=1) if hedge_assets else 0

    hedged_return = base_return.copy()
    hedged_return[hedge_signal] = (
        base_return[hedge_signal] * (1 - len(hedge_assets) * hedge_weight) +
        hedge_return[hedge_signal] * len(hedge_assets) * hedge_weight
    )

    return pd.DataFrame({
        "base": base_return,
        "hedged": hedged_return,
        "hedge_active": hedge_signal.astype(int)
    })

Academic papers:
- Caldara & Iacoviello (2022), "Measuring Geopolitical Risk", AER
- Apergis et al. (2021), "Geopolitical Risks and Asset Prices"
- Mueller & Rauh (2018), "The Hard Problem of Prediction for Conflict Prevention"

Data resources:
- GPR Index: https://www.matteoiacoviello.com/gpr.htm
- GDELT: https://www.gdeltproject.org/
- ACLED: https://acleddata.com/
- SIPRI: https://www.sipri.org/databases

Market-analysis tools:
- BDI (Baltic Dry Index): https://www.balticexchange.com/
- CDS spread data: Bloomberg / Refinitiv (paid) / FRED (partially free)
- Vessel AIS tracking: MarineTraffic.com