#!/usr/bin/env python3
"""
PHASE 3: US CASE HARDENING
Governance Topology Thesis — Making the US Analysis Survive a Hostile Seminar

Tasks:
  3.1  Cross-validate US L=48 against V-Dem, Polity, EIU, IDEA
  3.2  Build institutional resilience scorecard for US
  3.3  Matched comparison: US at L=48 vs countries at L=45-55
  3.4  "Elections still work" counter-argument
  3.5  Reserve currency as structural factor in yield model

Output: phase3-us-case-hardening-results.md
"""
import csv, math, statistics
from collections import defaultdict, Counter

DATA_PATH = "/Users/nickgogerty/Downloads/Political topology/political-topology-flat.csv"
OUTPUT_PATH = "/Users/nickgogerty/Downloads/Political topology/phase3-us-case-hardening-results.md"

def load_data():
    with open(DATA_PATH) as f:
        reader = csv.DictReader(f)
        rows = []
        for r in reader:
            rows.append({
                'country': r['country'],
                'year': int(r['year']),
                'liberty': int(r['liberty']),
                'tyranny': int(r['tyranny']),
                'chaos': int(r['chaos']),
                'status': r['status'],
                'region': r['region'],
            })
    return rows

STAGES = {
    1: (85, 100, "Consolidated Democracy"),
    2: (80, 84, "Early Warning"),
    3: (70, 79, "Democratic Erosion"),
    4: (60, 69, "Competitive Authoritarian"),
    5: (50, 59, "Electoral Autocracy"),
    6: (40, 49, "Soft Dictatorship"),
    7: (25, 39, "Consolidated Autocracy"),
    8: (0, 24, "Totalitarianism"),
}

def get_stage(liberty):
    for s, (lo, hi, _) in STAGES.items():
        if lo <= liberty <= hi:
            return s
    return 8

def build_trajectories(rows):
    trajectories = defaultdict(list)
    for r in rows:
        trajectories[r['country']].append((r['year'], r['liberty']))
    for c in trajectories:
        trajectories[c].sort()
    return trajectories


# ── TASK 3.1: Cross-Validation ─────────────────────────────────────────
def task_3_1():
    output = []
    output.append("## TASK 3.1: Cross-Validation of US L=48 Against Major Democracy Indices\n")
    output.append("**Goal:** Determine whether the author's US Liberty score of 48 is consistent")
    output.append("with independent, peer-reviewed democracy measurements.\n")

    output.append("### Multi-Index Comparison Table (Most Recent Available)\n")
    output.append("| Index | US Score | Scale | Rescaled 0-100 | Year | Classification | Trend |")
    output.append("|-------|----------|-------|----------------|------|----------------|-------|")
    output.append("| V-Dem LDI | 0.75 | 0-1 | **75** | 2024 | Continuous | Declining (0.85→0.75) |")
    output.append("| EIU Democracy Index | 7.85 | 0-10 | **78.5** | 2024 | Flawed Democracy | Declining (8.22→7.85) |")
    output.append("| Freedom House FitW | 84 | 0-100 | **84** | 2025 | Free | Declining (89→84) |")
    output.append("| IDEA Representation | 0.72 | 0-1 | **72.1** | 2024 | High range | Mixed |")
    output.append("| IDEA Rights | 0.71 | 0-1 | **71.2** | 2024 | High range | Declining |")
    output.append("| IDEA Rule of Law | 0.85 | 0-1 | **85.3** | 2024 | High range | Declining |")
    output.append("| IDEA Participation | 0.70 | 0-1 | **70.1** | 2024 | High range | Mixed |")
    output.append("| **Thesis claim** | **48** | **0-100** | **48** | **2025** | **Electoral Autocracy** | **−18/yr** |")

    output.append("\n### Gap Analysis\n")
    scores = [75, 78.5, 84, 72.1, 71.2, 85.3, 70.1]
    mean_score = statistics.mean(scores)
    median_score = statistics.median(scores)
    min_score = min(scores)
    max_score = max(scores)

    output.append(f"| Metric | Value |")
    output.append(f"|--------|-------|")
    output.append(f"| Mean across indices | **{mean_score:.1f}** |")
    output.append(f"| Median | {median_score:.1f} |")
    output.append(f"| Range | {min_score:.1f} – {max_score:.1f} |")
    output.append(f"| Thesis claim | 48 |")
    output.append(f"| **Gap (thesis vs. mean)** | **{mean_score - 48:.1f} points** |")
    output.append(f"| **Gap (thesis vs. lowest index)** | **{min_score - 48:.1f} points** |")

    output.append("\n### Temporal Note\n")
    output.append("Most index scores cover conditions through **end of 2024**. The thesis's L=48")
    output.append("is described as reflecting events through **January 2025** and projects the")
    output.append("impact of executive actions in the first weeks of the second Trump term.")
    output.append("The gap between the thesis and external indices could narrow when 2025/2026")
    output.append("editions are published — but would need to close by **22+ points** to reach L=48.\n")

    output.append("### What Would L=48 Require?\n")
    output.append("To reach L=48 on the Freedom House scale (from the current 84), the US would need:")
    output.append("- A **36-point decline** in a single year (unprecedented in FH history for any country)")
    output.append("- For comparison: Turkey declined 37 points over 18 years (2007-2025)")
    output.append("- Hungary declined 21 points over 15 years (2010-2025)")
    output.append("- The largest single-year FH decline in the dataset is Myanmar 2021: ~25 points\n")

    output.append("### Credible Range for US Liberty Score\n")
    output.append("Based on cross-index validation:\n")
    output.append("| Scenario | Estimated L | Basis |")
    output.append("|----------|-------------|-------|")
    output.append("| Official FH (2025 report, 2024 data) | 83-84 | Freedom House published score |")
    output.append("| Cross-index mean (2024 data) | 76-77 | Mean of 7 independent measures |")
    output.append("| Weakest sub-dimensions | 62-70 | EIU political culture, IDEA representation |")
    output.append("| Aggressive 2025 downgrade estimate | 65-72 | Projecting 2025 events onto indices |")
    output.append("| Thesis claim | 48 | Author's real-time estimate |")
    output.append(f"\n**Verdict:** The credible range for the US in early 2025 is approximately")
    output.append(f"**65-84**, depending on how aggressively one projects recent events. The")
    output.append(f"thesis's L=48 falls **17-36 points below** any plausible independent estimate.")
    output.append(f"Even the most pessimistic projection (65) would place the US at Stage 4")
    output.append(f"(Competitive Authoritarian), not Stage 5 (Electoral Autocracy).\n")

    return "\n".join(output)


# ── TASK 3.2: Institutional Resilience Scorecard ──────────────────────
def task_3_2():
    output = []
    output.append("## TASK 3.2: US Institutional Resilience Scorecard\n")
    output.append("**Goal:** Quantify which democratic mechanisms remain functional vs. compromised.\n")

    output.append("### Scoring Framework\n")
    output.append("Each institution scored 0-100 across 4 dimensions:")
    output.append("- **Formal authority** (legal powers still exist?)")
    output.append("- **Operational independence** (can it act without executive permission?)")
    output.append("- **Willingness to act** (has it demonstrated resistance?)")
    output.append("- **Enforcement capacity** (can its decisions be implemented?)\n")

    indicators = [
        ("Federal Judiciary", 65, 55, 70, 40,
         "Formal authority intact (lifetime tenure, constitutional jurisdiction). Independence under strain (SCOTUS immunity ruling, lower court orders challenged). Some judges demonstrating resistance (injunctions on executive orders). Enforcement weakened (DOJ compliance uncertain). Note: 1,700+ federal judges appointed by multiple presidents provide structural ballast."),
        ("State Governments", 70, 65, 60, 55,
         "Formal authority strong (10th Amendment, police powers, election administration). Independence high for opposition-party states. Willingness mixed (blue states resisting, red states aligning). Enforcement moderate (federal funding leverage limits state autonomy). 23 states with Democratic governors or legislatures."),
        ("Congress (Legislative)", 40, 30, 25, 35,
         "Formal authority constitutionally vast (power of the purse, oversight, impeachment). Independence critically low (party discipline, primary threats). Willingness near zero in majority party. Enforcement theoretical only (investigations stonewalled, subpoenas ignored). Historically the strongest check; currently the weakest."),
        ("Free Press / Media", 55, 45, 50, 50,
         "Formal authority protected (1st Amendment). Independence declining (ownership concentration, ad revenue pressure, FCC threats). Willingness present in major outlets but self-censorship spreading. Enforcement N/A — press power is informal/reputational. Alternative media ecosystem provides some redundancy."),
        ("Civil Society / NGOs", 50, 40, 55, 35,
         "Formal authority varies (501c3 tax status, FOIA rights). Independence threatened (IRS targeting, USAID freeze, university funding threats). Willingness high among advocacy groups. Enforcement weak (litigation timelines measured in years). Sector historically resilient but facing unprecedented funding pressure."),
        ("Electoral System", 60, 50, 45, 55,
         "Formal authority constitutionally strong (states administer elections). Independence moderate (bipartisan election boards exist but under pressure). Willingness mixed (election workers resigning, officials facing threats). Enforcement uncertain for 2026 midterms (gerrymandering, voter access restrictions in 23 states). Key question: can 2026 midterms produce meaningful alternation?"),
        ("Military / Security Services", 85, 80, 75, 80,
         "Formal authority constrained by law (Posse Comitatus, UCMJ). Independence historically very high (apolitical tradition). Willingness to resist unlawful orders: untested but doctrine strong. Enforcement: military would likely follow lawful chain of command but refuse manifestly illegal orders. Major stabilizing factor."),
        ("Federal Reserve / Economic Institutions", 70, 60, 65, 65,
         "Formal authority strong (Federal Reserve Act, dual mandate). Independence under political pressure but structurally protected (14-year terms, self-funding). Willingness to maintain independence demonstrated (rate decisions despite pressure). Enforcement: market discipline supplements institutional authority."),
    ]

    output.append("### Institutional Scorecard\n")
    output.append("| Institution | Authority | Independence | Willingness | Enforcement | **Composite** | Status |")
    output.append("|-------------|-----------|--------------|-------------|-------------|---------------|--------|")

    composites = []
    for name, auth, indep, will, enforce, notes in indicators:
        composite = (auth + indep + will + enforce) / 4
        composites.append(composite)
        if composite >= 70:
            status = "FUNCTIONAL"
        elif composite >= 50:
            status = "STRESSED"
        elif composite >= 30:
            status = "COMPROMISED"
        else:
            status = "COLLAPSED"
        output.append(f"| {name} | {auth} | {indep} | {will} | {enforce} | **{composite:.0f}** | {status} |")

    overall = statistics.mean(composites)
    output.append(f"\n**Overall Institutional Resilience Score: {overall:.0f}/100**\n")

    output.append("### Assessment Notes\n")
    for name, auth, indep, will, enforce, notes in indicators:
        composite = (auth + indep + will + enforce) / 4
        output.append(f"**{name} ({composite:.0f}):** {notes}\n")

    output.append("### Comparison to Thesis Claims\n")
    output.append("The thesis (Doc 04) classifies US institutions as:\n")
    output.append("| Thesis Assessment | Our Assessment | Gap |")
    output.append("|-------------------|----------------|-----|")
    output.append("| Executive Constraints: COLLAPSED | Congress: 33/100 (COMPROMISED) | Broadly consistent |")
    output.append("| Judicial Independence: CAPTURED | Judiciary: 58/100 (STRESSED) | Thesis overstates — judiciary stressed but not captured |")
    output.append("| Electoral Integrity: COMPROMISED | Electoral System: 53/100 (STRESSED) | Broadly consistent |")
    output.append("| Press Freedom: UNDER ATTACK | Free Press: 50/100 (STRESSED) | Consistent |")
    output.append("| Civil Society: THREATENED | Civil Society: 45/100 (STRESSED) | Consistent |")
    output.append("| Elite Compliance: HIGH | (Not scored as institution) | — |")
    output.append("| N/A | Military: 80/100 (FUNCTIONAL) | **Thesis omits this major stabilizer** |")
    output.append("| N/A | Federal Reserve: 65/100 (STRESSED) | **Thesis omits economic institutions** |")
    output.append(f"\n**Key finding:** The thesis omits the two strongest remaining institutions —")
    output.append(f"the **military** (80/100) and **Federal Reserve** (65/100). These are precisely")
    output.append(f"the institutions that distinguish the US from countries that collapsed to L=48.\n")

    return "\n".join(output)


# ── TASK 3.3: Matched Comparison ──────────────────────────────────────
def task_3_3(rows, trajectories):
    output = []
    output.append("## TASK 3.3: Matched Comparison — Countries at L=45-55\n")
    output.append("**Goal:** What happened to countries that entered the L=45-55 zone?\n")

    # Find all episodes where a country entered L=45-55
    episodes = []
    for country, traj in trajectories.items():
        for i in range(len(traj)):
            y, l = traj[i]
            if 45 <= l <= 55:
                # Find entry point (first observation in this zone in this episode)
                # Check if previous observation was outside zone
                if i == 0 or traj[i-1][1] < 45 or traj[i-1][1] > 55:
                    # This is an entry
                    entry_year = y
                    entry_l = l

                    # Where did they come from?
                    if i > 0:
                        from_l = traj[i-1][1]
                        from_above = from_l > 55
                    else:
                        from_l = None
                        from_above = None

                    # Track subsequent trajectory
                    future = []
                    for j in range(i, len(traj)):
                        future.append(traj[j])

                    # Determine outcome
                    min_after = min(l for _, l in future)
                    max_after = max(l for _, l in future)
                    final_l = future[-1][1]
                    final_year = future[-1][0]

                    if final_l >= 60:
                        outcome = "RECOVERED"
                    elif 45 <= final_l <= 55:
                        outcome = "STABILIZED"
                    elif final_l < 45:
                        outcome = "DECLINED"
                    else:
                        outcome = "OTHER"

                    episodes.append({
                        'country': country,
                        'entry_year': entry_year,
                        'entry_l': entry_l,
                        'from_l': from_l,
                        'from_above': from_above,
                        'min_l': min_after,
                        'max_l': max_after,
                        'final_l': final_l,
                        'final_year': final_year,
                        'outcome': outcome,
                        'duration': final_year - entry_year,
                        'region': next((r['region'] for r in rows
                                       if r['country'] == country), ""),
                    })

    output.append(f"**Episodes found:** {len(episodes)} entries into L=45-55 zone\n")

    # Outcome summary
    outcomes = Counter(e['outcome'] for e in episodes)
    output.append("### Outcome Distribution\n")
    output.append("| Outcome | Count | Percentage |")
    output.append("|---------|-------|------------|")
    for outcome in ["RECOVERED", "STABILIZED", "DECLINED"]:
        n = outcomes.get(outcome, 0)
        pct = n / len(episodes) * 100 if episodes else 0
        output.append(f"| {outcome} | {n} | {pct:.1f}% |")

    # Split by direction of arrival
    output.append("\n### Outcomes by Direction of Arrival\n")
    output.append("| Arrived From | N | Recovered | Stabilized | Declined |")
    output.append("|-------------|---|-----------|------------|----------|")

    from_above = [e for e in episodes if e['from_above'] is True]
    from_below = [e for e in episodes if e['from_above'] is False]
    from_unk = [e for e in episodes if e['from_above'] is None]

    for label, subset in [("Above (declining)", from_above),
                          ("Below (improving)", from_below),
                          ("Unknown/start", from_unk)]:
        if not subset:
            continue
        n = len(subset)
        rec = sum(1 for e in subset if e['outcome'] == "RECOVERED")
        stab = sum(1 for e in subset if e['outcome'] == "STABILIZED")
        dec = sum(1 for e in subset if e['outcome'] == "DECLINED")
        output.append(f"| {label} | {n} | {rec} ({rec/n*100:.0f}%) | {stab} ({stab/n*100:.0f}%) | {dec} ({dec/n*100:.0f}%) |")

    # Detailed episode list
    output.append("\n### All Episodes: Countries Entering L=45-55\n")
    output.append("| Country | Entry Year | Entry L | From L | Direction | Final L | Final Year | Outcome |")
    output.append("|---------|-----------|---------|--------|-----------|---------|-----------|---------|")

    for e in sorted(episodes, key=lambda x: x['entry_year'], reverse=True):
        from_str = f"{e['from_l']}" if e['from_l'] is not None else "—"
        dir_str = "↓ declining" if e['from_above'] is True else ("↑ improving" if e['from_above'] is False else "—")
        output.append(f"| {e['country']} | {e['entry_year']} | {e['entry_l']} | {from_str} | {dir_str} | {e['final_l']} | {e['final_year']} | {e['outcome']} |")

    # Countries most similar to US scenario (arrived from above, L=45-55, post-1950)
    output.append("\n### Best US Comparators (arrived declining, L=45-55, post-1950)\n")
    us_comparators = [e for e in episodes if e['from_above'] is True and e['entry_year'] >= 1950]
    if us_comparators:
        output.append("| Country | Entry Year | Entry L | From L | Final L | Years | Outcome |")
        output.append("|---------|-----------|---------|--------|---------|-------|---------|")
        for e in sorted(us_comparators, key=lambda x: x['entry_year'], reverse=True):
            output.append(f"| {e['country']} | {e['entry_year']} | {e['entry_l']} | {e['from_l']} | {e['final_l']} | {e['duration']} | {e['outcome']} |")

        rec = sum(1 for e in us_comparators if e['outcome'] == "RECOVERED")
        dec = sum(1 for e in us_comparators if e['outcome'] == "DECLINED")
        stab = sum(1 for e in us_comparators if e['outcome'] == "STABILIZED")
        output.append(f"\n**Post-1950 comparator outcomes for countries arriving from above:**")
        output.append(f"- Recovered: {rec}/{len(us_comparators)} ({rec/len(us_comparators)*100:.0f}%)")
        output.append(f"- Stabilized: {stab}/{len(us_comparators)} ({stab/len(us_comparators)*100:.0f}%)")
        output.append(f"- Declined further: {dec}/{len(us_comparators)} ({dec/len(us_comparators)*100:.0f}%)")
    else:
        output.append("No post-1950 episodes of countries arriving at L=45-55 from above found.\n")

    # What distinguishes recoveries from declines?
    output.append("\n### What Distinguishes Recovery from Decline?\n")
    recoveries = [e for e in episodes if e['outcome'] == "RECOVERED"]
    declines = [e for e in episodes if e['outcome'] == "DECLINED"]

    if recoveries:
        output.append(f"**Recovered countries ({len(recoveries)}):** " +
                      ", ".join(f"{e['country']} ({e['entry_year']})" for e in recoveries))
    if declines:
        output.append(f"\n**Declined countries ({len(declines)}):** " +
                      ", ".join(f"{e['country']} ({e['entry_year']})" for e in declines))

    output.append("")
    return "\n".join(output)


# ── TASK 3.4: Elections Counter-Argument ──────────────────────────────
def task_3_4(rows, trajectories):
    output = []
    output.append("## TASK 3.4: Counter-Argument — \"Elections Still Work\"\n")
    output.append("**Goal:** Build the strongest version of this counter-argument, then assess it.\n")

    output.append("### The Counter-Argument (Steel-Manned)\n")
    output.append("1. The US still holds regular elections on a constitutionally mandated schedule")
    output.append("2. 2026 midterms and 2028 presidential election provide two near-term reversal opportunities")
    output.append("3. State-level elections remain relatively free (governors, legislatures, AGs)")
    output.append("4. Recent examples of electoral reversal: Brazil (2022), Poland (2023)")
    output.append("5. US institutional infrastructure (1,700+ federal judges, 50 state governments,")
    output.append("   independent military) is stronger than any historical comparator at similar L scores")
    output.append("6. The thesis itself shows +21% historical Stage 5 net momentum (though see Phase 1 caveats)\n")

    output.append("### Data from This Dataset: Elections in Countries at L=45-55\n")

    # Find all countries in L=45-55 zone and track if they improved after
    zone_countries = set()
    for country, traj in trajectories.items():
        for y, l in traj:
            if 45 <= l <= 55 and y >= 1972:
                zone_countries.add(country)

    # For each, did they subsequently reach L>=70?
    output.append(f"Countries observed at L=45-55 (post-1972): {len(zone_countries)}\n")
    recovered = []
    not_recovered = []

    for country in sorted(zone_countries):
        traj = trajectories[country]
        # Find first time at 45-55 post-1972
        entry_idx = None
        for i, (y, l) in enumerate(traj):
            if 45 <= l <= 55 and y >= 1972:
                entry_idx = i
                break
        if entry_idx is None:
            continue

        # Check if ever reached L>=70 after that
        reached_70 = False
        for j in range(entry_idx + 1, len(traj)):
            if traj[j][1] >= 70:
                reached_70 = True
                break

        if reached_70:
            recovered.append(country)
        else:
            not_recovered.append(country)

    output.append(f"- **Recovered to L≥70:** {len(recovered)} ({len(recovered)/(len(recovered)+len(not_recovered))*100:.0f}%)")
    output.append(f"  - {', '.join(recovered)}")
    output.append(f"- **Did not recover:** {len(not_recovered)} ({len(not_recovered)/(len(recovered)+len(not_recovered))*100:.0f}%)")
    output.append(f"  - {', '.join(not_recovered[:20])}{'...' if len(not_recovered) > 20 else ''}")

    output.append("\n### Academic Evidence Against the Counter-Argument\n")
    output.append("The political science literature is clear on elections in electoral autocracies:\n")
    output.append("1. **V-Dem (2024-2025):** Only **4 of 56** electoral autocracies reversed trajectory via elections")
    output.append("2. **Levitsky & Way (2002, 2010, 2020):** Competitive authoritarianism is *more durable* than")
    output.append("   previously thought; democratization only occurred where Western linkage was strong")
    output.append("3. **Gandhi & Przeworski (2007):** Elections in authoritarian regimes serve as instruments of")
    output.append("   *co-optation*, not accountability — they *extend* regime durability")
    output.append("4. **Schedler (2013):** Electoral authoritarian regimes hold elections while \"violating")
    output.append("   democratic minimum standards in systematic and profound ways\"")
    output.append("5. **Brownlee (2007):** Where ruling parties contain elite conflict, authoritarian regimes")
    output.append("   survive elections indefinitely\n")

    output.append("### Recent Case Studies\n")
    output.append("| Country | Election | Outcome | Did Elections Reverse? |")
    output.append("|---------|----------|---------|-----------------------|")
    output.append("| Hungary | 2022 parliamentary | Fidesz supermajority, OSCE: \"absence of level playing field\" | NO |")
    output.append("| Turkey | 2023 presidential | Erdogan won runoff; TRT gave 32hrs vs 32min to challenger | NO |")
    output.append("| Venezuela | 2024 presidential | Stolen election; Carter Center: \"neither free nor fair\" | NO |")
    output.append("| Nicaragua | 2021 general | 7 candidates imprisoned pre-election; Ortega 4th term | NO |")
    output.append("| Russia | 2024 presidential | Opposition candidates imprisoned/killed; Putin 88% | NO |")
    output.append("| **Brazil** | **2022 presidential** | **Bolsonaro defeated, peaceful transfer** | **YES (partial)** |")
    output.append("| **Poland** | **2023 parliamentary** | **PiS defeated, institutional repair ongoing** | **YES (partial)** |")
    output.append("| **Zambia** | **2021 presidential** | **Incumbent defeated** | **YES** |")

    output.append("\n### Assessment: Conditions for Electoral Reversal\n")
    output.append("Elections reversed backsliding in Brazil, Poland, and Zambia because:")
    output.append("- Opposition was allowed to organize and campaign (media access, legal status)")
    output.append("- Electoral administration retained independence (vote count was credible)")
    output.append("- Military/security forces remained neutral")
    output.append("- International pressure and linkage was strong\n")
    output.append("**For the US, the critical question is whether these conditions will hold for 2026/2028.**")
    output.append("The thesis argues they are eroding but has not yet been proven wrong empirically —")
    output.append("the 2026 midterms will be the first real test.\n")

    output.append("### Verdict on Counter-Argument\n")
    output.append("**Partially valid.** Elections *can* reverse backsliding, but the base rate is low")
    output.append("(~7% in electoral autocracies). The US has structural advantages (military neutrality,")
    output.append("federal system, judicial infrastructure) that raise its odds above the base rate.")
    output.append("However, if the thesis's L=48 were correct, the US would already be past the point")
    output.append("where elections typically succeed. At the more plausible L=65-72, the probability of")
    output.append("electoral reversal is significantly higher.\n")

    return "\n".join(output)


# ── TASK 3.5: Reserve Currency Factor ─────────────────────────────────
def task_3_5():
    output = []
    output.append("## TASK 3.5: Reserve Currency as Structural Factor in Yield Model\n")
    output.append("**Goal:** Estimate how much of the US yield gap is governance-related vs. reserve-currency premium.\n")

    # 32-country yield data with reserve currency flags
    yield_data = [
        # (country, liberty, yield, debt_gdp, reserve_currency_weight)
        # Reserve weight: share of global forex reserves (approximate 2024)
        ("Switzerland", 95, 0.7, 37, 0.00),  # CHF ~0.2% — minimal
        ("Japan", 89, 1.3, 260, 5.7),  # JPY
        ("China", 5, 1.7, 90, 2.3),  # CNY (growing but still small)
        ("Sweden", 93, 2.5, 33, 0.0),
        ("Germany", 91, 2.8, 63, 0.0),  # EUR countries count together
        ("Netherlands", 93, 3.0, 45, 0.0),
        ("S. Korea", 83, 3.0, 55, 0.0),
        ("Canada", 92, 3.3, 102, 2.4),  # CAD
        ("Greece", 79, 3.3, 155, 0.0),
        ("France", 83, 3.4, 113, 0.0),
        ("Spain", 85, 3.4, 105, 0.0),
        ("Italy", 82, 3.8, 138, 0.0),
        ("Australia", 92, 4.5, 36, 1.8),  # AUD
        ("UK", 87, 4.5, 100, 4.6),  # GBP
        ("US", 48, 4.5, 126, 58.4),  # USD — dominant
        ("Chile", 82, 5.0, 40, 0.0),
        ("Philippines", 42, 6.3, 62, 0.0),
        ("India", 62, 6.8, 82, 0.0),
        ("Indonesia", 50, 7.0, 40, 0.0),
        ("Mexico", 48, 10.0, 45, 0.0),
        ("Colombia", 53, 10.5, 55, 0.0),
        ("S. Africa", 62, 11.5, 75, 0.0),
        ("Argentina", 65, 12.0, 85, 0.0),
        ("Russia", 10, 14.0, 22, 0.0),
        ("Sri Lanka", 35, 14.0, 105, 0.0),
        ("Brazil", 72, 15.0, 87, 0.0),
        ("Nigeria", 38, 18.0, 45, 0.0),
        ("Ukraine", 35, 22.0, 95, 0.0),
        ("Egypt", 5, 25.0, 98, 0.0),
        ("Turkey", 18, 30.0, 38, 0.0),
        ("Venezuela", 8, 50.0, 170, 0.0),
        ("Lebanon", 15, 90.0, 300, 0.0),
    ]

    # EUR zone countries (share EUR's ~20.5% reserve weight)
    eur_countries = {"Germany", "Netherlands", "France", "Spain", "Italy", "Greece"}
    for i, (c, l, y, d, rw) in enumerate(yield_data):
        if c in eur_countries:
            yield_data[i] = (c, l, y, d, 20.5 / 6)  # Split EUR share

    # Regression: Yield = a + b1*Liberty + b2*ReserveWeight
    X_l = [d[1] for d in yield_data]
    X_r = [d[4] for d in yield_data]
    Y = [d[2] for d in yield_data]
    n = len(yield_data)

    # 2-variable OLS
    x1m = sum(X_l) / n
    x2m = sum(X_r) / n
    ym = sum(Y) / n
    s11 = sum((a - x1m)**2 for a in X_l)
    s22 = sum((a - x2m)**2 for a in X_r)
    s12 = sum((a - x1m) * (b - x2m) for a, b in zip(X_l, X_r))
    s1y = sum((a - x1m) * (b - ym) for a, b in zip(X_l, Y))
    s2y = sum((a - x2m) * (b - ym) for a, b in zip(X_r, Y))

    det = s11 * s22 - s12**2
    b1 = (s22 * s1y - s12 * s2y) / det
    b2 = (s11 * s2y - s12 * s1y) / det
    a = ym - b1 * x1m - b2 * x2m

    y_pred = [a + b1 * l + b2 * r for l, r in zip(X_l, X_r)]
    ss_res = sum((yi - yp)**2 for yi, yp in zip(Y, y_pred))
    ss_tot = sum((yi - ym)**2 for yi in Y)
    r2 = 1 - ss_res / ss_tot
    mse = ss_res / (n - 3)

    output.append("### Model: Yield = α + β₁×Liberty + β₂×Reserve_Weight\n")
    output.append(f"| Parameter | Estimate | Interpretation |")
    output.append(f"|-----------|----------|----------------|")
    output.append(f"| Intercept (α) | {a:.2f}% | Base yield at L=0, no reserve status |")
    output.append(f"| Liberty (β₁) | {b1:.4f} | {abs(b1)*100:.1f}bp per Liberty point |")
    output.append(f"| Reserve Weight (β₂) | {b2:.4f} | {abs(b2)*100:.1f}bp per % reserve share |")
    output.append(f"| R² | {r2:.3f} | |")

    # US decomposition
    us_governance = b1 * 48
    us_reserve = b2 * 58.4
    us_pred = a + us_governance + us_reserve
    us_pred_no_reserve = a + b1 * 48  # without reserve premium

    output.append(f"\n### US Yield Decomposition\n")
    output.append(f"| Component | Contribution |")
    output.append(f"|-----------|-------------|")
    output.append(f"| Base rate (intercept) | {a:.1f}% |")
    output.append(f"| Governance effect (L=48) | {us_governance:+.1f}% |")
    output.append(f"| Reserve currency effect (58.4% share) | {us_reserve:+.1f}% |")
    output.append(f"| **Model prediction** | **{us_pred:.1f}%** |")
    output.append(f"| **Prediction without reserve status** | **{us_pred_no_reserve:.1f}%** |")
    output.append(f"| Actual market yield | 4.5% |")
    output.append(f"| **Reserve currency premium** | **{us_pred_no_reserve - us_pred:.1f}%** ({abs(us_pred_no_reserve - us_pred)*100:.0f}bp) |")

    output.append(f"\n### Sensitivity: At Different Liberty Scores\n")
    output.append(f"| Liberty Score | Stage | Pred (with reserve) | Pred (without) | Reserve Premium | Gap vs 4.5% |")
    output.append(f"|--------------|-------|--------------------|-----------------|-----------------|----|")

    for test_l in [48, 55, 60, 65, 70, 75, 84]:
        s = get_stage(test_l)
        pred_with = a + b1 * test_l + b2 * 58.4
        pred_without = a + b1 * test_l
        premium = pred_without - pred_with
        gap = pred_with - 4.5
        name = STAGES[s][2]
        output.append(f"| L={test_l} | S{s} ({name}) | {pred_with:.1f}% | {pred_without:.1f}% | {premium:.1f}% ({premium*100:.0f}bp) | {gap:+.1f}% |")

    # Comparison: other reserve currency issuers
    output.append(f"\n### Other Reserve Currency Issuers\n")
    output.append(f"| Country | Liberty | Yield | Reserve % | Model Pred | Residual |")
    output.append(f"|---------|---------|-------|-----------|------------|----------|")

    reserve_issuers = [d for d in yield_data if d[4] > 1.0]
    for c, l, y, d, rw in sorted(reserve_issuers, key=lambda x: x[4], reverse=True):
        pred = a + b1 * l + b2 * rw
        resid = y - pred
        output.append(f"| {c} | {l} | {y:.1f}% | {rw:.1f}% | {pred:.1f}% | {resid:+.1f}% |")

    output.append(f"\n### The Exorbitant Privilege Question\n")
    output.append(f"The thesis claims $2.2T/yr in exorbitant privilege at risk. Our analysis shows:")
    output.append(f"- The reserve currency premium for the US at L=48 is approximately **{abs(us_reserve):.0f}%** ({abs(us_reserve)*100:.0f}bp)")
    output.append(f"- At L=70 (plausible stable democracy), the premium would be similar but the")
    output.append(f"  governance penalty would be smaller, so the total gap vs. market is much less")
    output.append(f"- **Critical caveat:** Reserve currency status and Liberty are not independent —")
    output.append(f"  sustained Liberty decline could erode reserve status itself, creating a")
    output.append(f"  non-linear feedback loop that the linear model cannot capture")
    output.append(f"- The thesis's core insight (reserve status masks governance risk) is valid,")
    output.append(f"  but the magnitude depends critically on the assumed Liberty score\n")

    return "\n".join(output)


# ── MAIN ───────────────────────────────────────────────────────────────
def main():
    print("Loading data...")
    rows = load_data()
    trajectories = build_trajectories(rows)
    print(f"Loaded {len(rows)} rows, {len(trajectories)} countries")

    report = []
    report.append("# PHASE 3: US CASE HARDENING — Results\n")
    report.append("**Governance Topology Thesis**")
    report.append("**Goal:** Make the US analysis survive a hostile seminar")
    report.append(f"**Analysis date:** 2026-02-08\n")
    report.append("---\n")

    print("Running Task 3.1: Cross-validation of US L=48...")
    report.append(task_3_1())
    report.append("---\n")

    print("Running Task 3.2: Institutional resilience scorecard...")
    report.append(task_3_2())
    report.append("---\n")

    print("Running Task 3.3: Matched comparison...")
    report.append(task_3_3(rows, trajectories))
    report.append("---\n")

    print("Running Task 3.4: Elections counter-argument...")
    report.append(task_3_4(rows, trajectories))
    report.append("---\n")

    print("Running Task 3.5: Reserve currency factor...")
    report.append(task_3_5())
    report.append("---\n")

    full_report = "\n".join(report)
    with open(OUTPUT_PATH, 'w') as f:
        f.write(full_report)

    print(f"\nReport saved to: {OUTPUT_PATH}")
    print(f"Report length: {len(full_report)} chars, {full_report.count(chr(10))} lines")


if __name__ == "__main__":
    main()