Sovereign Debt Crisis MCP

A Model Context Protocol server implementing an endogenous political economy sovereign default simulator. Fuses macroeconomic, trade, FX, regulatory, and geopolitical data from 15 Apify actors into a sovereign debt network, then applies game theory, econometrics, and mechanism design to simulate defaults, contagion, and political transitions.

Tools (8)

#	Tool	Method	Price
1	simulate_sovereign_default	Eaton-Gersovitz (1981) value function iteration on (b,y) state space	$0.045
2	compute_clearing_contagion	Eisenberg-Noe (2001) + CDS circularity (Scarf-type damped iteration)	$0.045
3	estimate_fiscal_sustainability	Hansen (1999) panel threshold regression with bootstrap inference	$0.035
4	design_optimal_restructuring	Myerson mechanism design + Nash-Rubinstein alternating offers	$0.040
5	model_speculative_attack	Morris-Shin (2003) global games with unique switching equilibrium	$0.040
6	detect_crisis_jumps	Lee-Mykland (2008) non-parametric jump detection via bipower variation	$0.035
7	forecast_multi_frequency	MIDAS regression (Ghysels 2004) with Almon exponential lag polynomials	$0.035
8	assess_political_transition_risk	Acemoglu-Robinson (2006) inequality-driven regime transitions	$0.035

Data Sources (15 actors across 5 categories)

Macro (6): IMF, World Bank, FRED, BLS, OECD, Eurostat Trade (2): UN COMTRADE, World Bank Projects FX (3): Exchange Rates, Historical Exchange Rates, ECB Exchange Rates Regulatory (2): Federal Register, Congress Bills Geopolitical (2): OpenSanctions, REST Countries

Mathematical Foundations

Tool 1: Sovereign Default — Eaton-Gersovitz Model

The canonical model of strategic sovereign default. A small open economy with income y and debt b chooses between repayment and default:

• Continuation value: V_c(b,y) = max_{b'} u(c) + β·E[max(V_c(b',y'), V_d(y'))] where c = y - b + q(b',y)·b'
• Default value: V_d(y) = u(y^def) + β·E[θ·V_c(0,y') + (1-θ)·V_d(y')]
• Bond price: q(b',y) = E[1{no default}]/(1+r_f) — endogenous risk pricing

The default threshold, borrowing policy, and equilibrium spreads are all simultaneously determined as a fixed point.

Tool 2: Clearing Contagion — Eisenberg-Noe with CDS

Standard Eisenberg-Noe finds clearing payments via monotone fixed-point iteration (Tarski's theorem). CDS circularity breaks monotonicity: bank A's solvency depends on bank B paying CDS protection, but B's ability to pay depends on C, which depends on A. The non-monotone mapping requires Scarf-type damped iteration with a convergence rate of O(1/k) instead of the geometric convergence of standard methods.

Tool 3: Fiscal Sustainability — Hansen Threshold Regression

Tests whether fiscal behavior changes regime at a debt-to-GDP threshold γ. The model y_it = α_i + β₁·x·I(q≤γ) + β₂·x·I(q>γ) + ε classifies countries into sustainable vs. unsustainable fiscal regimes. The threshold is identified by minimizing concentrated SSR over a grid, with significance assessed via Hansen bootstrap (asymptotic distribution is non-standard — mixture of chi-squareds).

Tool 4: Optimal Restructuring — Myerson + Rubinstein

Combines mechanism design with bargaining theory:

• Myerson revelation principle: Debtor type θ (capacity to pay) is private info. IC constraints require monotone transfer schedules.
• Nash bargaining: max (u_d - d_d)^α · (u_c - d_c)^(1-α) gives fair surplus split.
• Rubinstein alternating offers: First-mover share = (1-δ_c)/(1-δ_d·δ_c) where δ are discount factors. More patient player captures more surplus.

Tool 5: Speculative Attack — Morris-Shin Global Games

Resolves the Obstfeld (1996) indeterminacy (multiple equilibria in currency crises) via incomplete information. Speculators receive private signals x_i = θ + σ·ε_i about fundamentals θ. With finite signal precision (σ > 0), there exists a unique switching equilibrium θ*: speculators attack iff x_i < θ*, and the peg collapses iff θ < θ*. This is perhaps the most elegant application of global games to macroeconomics.

Tool 6: Crisis Jumps — Lee-Mykland Test

Non-parametric detection of jumps in CDS/spread data:

• Bipower variation σ̂² estimates continuous volatility while being robust to jumps
• Test statistic L_i = |Δlog(X_i)| / σ̂_i
• Under H₀: (L_i - C_n)/S_n converges to Gumbel distribution
• Centering: C_n = √(2·log(n)) - [log(π)+log(log(n))] / [2√(2·log(n))]

Tool 7: Multi-Frequency Forecast — MIDAS Regression

Bridges the frequency gap between daily financial data and quarterly macro data:

• y_t^(Q) = β₀ + β₁ · Σ_k B(k;θ)·x_{t-k}^(D) + ε_t
• Almon exponential weights: B(k;θ₁,θ₂) = exp(θ₁k+θ₂k²) / Σexp(...)
• Parsimonious: 2 parameters govern the entire high-frequency lag structure
• Nowcasting: current-quarter prediction using within-quarter high-freq data

Tool 8: Political Transition — Acemoglu-Robinson

Models political transitions as inequality-driven Markov dynamics:

• Elite regime: Concedes redistribution when revolution threat θ exceeds threshold
• Democratic regime: Median voter taxation at τ_m
• Coup risk: Elite military capacity × (1 - cost of coup)
• Gini threshold: Inequality crossing critical level triggers transition
• Markov chain over {democratic, hybrid, authoritarian} with endogenous transition probabilities

Architecture

Client ──► Express (port 3019) ──► McpServer factory
                                       │
                          ┌─────────────┼─────────────┐
                          ▼             ▼             ▼
                    SSE transport  Streamable HTTP  Health endpoint
                          │             │
                          ▼             ▼
                   getOrBuildNetwork() (5-min TTL cache)
                          │
              ┌───────────┼───────────┐
              ▼           ▼           ▼
        15 Apify actors   Graph build   8 analysis tools
        (parallel fetch)  (fusion)      (per-request)

Usage

npm install
npm run build
npm start

Environment variables:

• APIFY_TOKEN — Apify API token for actor execution
• PORT — Server port (default: 3019)

Endpoints

• GET /sse — SSE transport for MCP
• POST /messages?sessionId=... — SSE message handler
• POST /mcp — Streamable HTTP transport
• GET /health — Health check with network stats