ACDArab Climate–Debt PortalClimate, trade & sovereign-risk intelligence

From climate and trade shocks to sovereign debt, with explicit model governance.

The portal combines country climate intelligence, Bayesian macro-fiscal climate models, a separate pooled trade-transmission system, economic shock translation and level debt reconstruction across 13 currently published Arab economies, including Kuwait's explicitly labelled reduced-data specification. AI assists the user interface; numerical results remain grounded in analytical engines and stored model outputs.

Open Scenario Lab 13 published debt models · climate + trade scenario families · projection horizon through 2031
01

Analytical framework

ACD uses two linked but distinct analytical systems. The climate system is country-specific and Bayesian. The trade system is a pooled Arab-country transmission layer estimated separately and connected to country-specific debt baselines and debt dynamics. Keeping the two systems separate avoids presenting pooled trade coefficients as if they had been estimated inside each country Dynare model.

Climate intelligence

Historical and physical climate conditions are translated into country-specific standardized climate stress.

Trade & external shocks

Terms of trade, world demand, energy-trade conditions and export-demand innovations are translated into standardized economic shocks.

Sovereign debt reconstruction

Climate or trade responses are applied to institutional debt baselines through a transparent annual debt identity.

Model classification

The climate system is a country-specific linear macro-fiscal transmission model estimated in Dynare using Bayesian methods. The trade system is a semi-structural pooled fixed-effects transmission model with current and lagged shocks. Neither system should be described as a deterministic forecast of sovereign debt.

02

Country model library

The 13-country published model library uses a common debt architecture while retaining country-specific samples, financing blocks, baselines and validation caveats. Production specifications identify the model form and governance treatment; they are not rankings of economic importance.

CountryISOSpecificationEstimation periodNProduction specificationValidation note
EgyptEGYFlexible + spread1999–202426Standard finalOutput-robust; convergence and prior-sensitivity disclosures retained
Saudi ArabiaSAUPeg + global financing1995–202430Standard finalMCMC-clean; wider-prior sensitivity governance note retained
LebanonLBNFlexible + spread2001–201919Standard finalPre-crisis estimation sample; post-2025 baseline is portal status-quo, not IMF forecast
United Arab EmiratesAREPeg + spread2007–202418Standard finalMCMC-clean final specification
QatarQATPeg + spread1995–202430Standard finalOutput-robust; convergence disclosure retained
BahrainBHRPeg + spread1995–202430Standard finalOutput-robust; convergence disclosure retained
KuwaitKWTReduced · peg · no spread1995–202430Reduced-dataSeparately validated reduced-data specification with narrower climate channels
OmanOMNPeg + global financing1995–202430Standard finalMCMC-clean final specification
JordanJORPeg + spread1995–202430Standard finalOutput-robust; convergence disclosure retained
MoroccoMARFlexible + spread1995–202430Standard finalOutput robustness reviewed; convergence disclosure retained
TunisiaTUNFlexible + spread1995–202430Multimodal finalTwo-mode posterior envelope retained; no arbitrary mode selection
AlgeriaDZAFlexible + spread1995–202430Standard finalOutput and prior-sensitivity governance notes retained
IraqIRQPeg + spread2006–202419Standard finalOutput-robust; convergence disclosure retained

Kuwait is available through a reduced-data specification rather than the standard v14.2 Dynare model. It uses Kuwait-specific climate, output and fiscal persistence plus a transparent median calibration from the published Saudi Arabia and Oman peg/no-spread climate models; inflation and sovereign-spread climate channels remain disabled.

03

Climate intelligence

The climate measure combines heat stress, hydrological drought and drought persistence. Monthly observations are assessed against each country's own calendar-month distribution so unusual climate conditions retain a country-specific meaning, including in highly arid environments.

Heat componentThermal pressure

Temperature observations are converted into empirical distribution scores relative to the country and calendar month.

Hydrological componentWater-stress pressure

Precipitation, soil moisture and runoff information are combined into the hydrological stress component.

Persistence componentDuration and accumulation

Dry-month frequency and consecutive dry spells capture persistence rather than only contemporaneous drought intensity.

Annual components are standardized over the estimation-era climate history and combined into the country climate-stress state. Historical reconstruction tests for the physical translator reproduce the annual climate index to numerical precision across the currently modelled countries.

04

Physical climate pathway translation

Physical mode accepts annual temperature anomalies and precipitation changes. The same empirical transformations used in the historical climate index translate those assumptions into incremental standardized climate stress.

01 · Temperature

Apply the anomaly to the calendar-month temperature climatology and translate it through the historical empirical distribution.

02 · Precipitation

Apply the percentage change to monthly climatological precipitation and convert the resulting level into a country-specific wetness/drought score.

03 · Hydrology

Map the physical precipitation disturbance through the historical soil-moisture and runoff relationships used in the climate index.

04 · Persistence

Reconstruct dry-month frequency and consecutive drought duration under the physical pathway.

Ht=z(ht¯)(9a)
Dt=z(dt¯)(9b)
Pt=z(pt)(9c)
ct=z[Ht+Dt+Pt3](10)
Historical support and extrapolation

Physical scenarios are evaluated against the country's historical climate support. Results can therefore be labelled within historical support, moderate extrapolation or high extrapolation rather than treating every physical assumption as equally well supported by the observed record.

05

Climate econometric system

The country climate models are linear systems around macroeconomic gaps and standardized external states. Terms enter according to the relevant exchange-rate and financing specification.

Climate, energy and global-activity states
ct=ρcct1+εc,t(1)
qt=ρqqt1+εq,t(2)
wt=ρwwt1+εw,t(3)

Within the country Bayesian climate model, climate, energy and global activity are represented as country-relevant external states.

Output transmission
xt=ρxxt1σr(rt+st1)+χΔeΔet+χcct+χqqt+χwwt+εx,t(4)

Output responds to persistence, financing conditions, exchange-rate movements where relevant, climate stress, energy prices and world activity.

Inflation transmission
πt=ρππt1+κxxt1+κcct+κqqt+κΔeΔet+επ,t(5)

Inflation combines persistence, lagged activity and external cost pressures.

Primary-balance transmission
pbt=ρpbpbt1+ψddt1+ψxxt1+ψcct+ψqqt+εpb,t(6)

The fiscal block captures debt feedback, cyclical conditions and climate-related pressure.

Financing-spread transmission
st=ρsst1+ηddt1+ηpbpbt1+ηΔeΔet+ηcct+ηqqt+εs,t(7)

Where the country specification contains a spread block, financing conditions respond to debt, fiscal performance and external pressures.

Debt-state transmission
dt=ρddt1+αrrtαxxt1αππt1αpbpbt1+αsst1+εd,t(8)

The estimated debt state is a structural consistency measure. The public portal separately reconstructs the level debt path around the institutional baseline.

x output gap · π inflation gap · r global financing benchmark · pb primary balance · s financing spread · d debt state · Δe exchange-rate change · c climate stress · q energy-price state · w world-growth state

Important separation from the trade engine

The q and w states above are controls inside the country Bayesian climate models. The operational trade-to-debt system described below is estimated separately and does not overwrite or re-estimate the v14.2 country Dynare models.

06

Bayesian estimation

Country climate systems are estimated offline in Dynare using Bayesian methods. Preliminary country regressions inform signs, persistence, initialization and economically defensible parameter ranges. The climate-output coefficient χc, climate-fiscal coefficient ψc and, where applicable, climate-financing coefficient ηc are estimated jointly with domestic shock scales.

Climate-output coefficientχc
Climate-fiscal coefficientψc
Climate-financing coefficientηc where applicable
Posterior uncertaintyMeans, modes and 90% HPD intervals
Posterior simulationTwo-chain Metropolis–Hastings
Current portal freezeFinal analytical freeze · 50,000 saved MH draws per chain · two chains for the 12 standard/multimodal systems

The final analytical freeze is complete. Original MCMC convergence diagnostics remain visible rather than being silently converted to passes. Production governance therefore separates chain diagnostics, policy-output robustness, local prior sensitivity and, for Tunisia, explicit multimodal posterior uncertainty. Kuwait is governed separately as a reduced-data specification.

07

Trade data and shock construction

The trade layer begins with a 13-country analytical panel. It combines the portal macro-fiscal dataset with internationally comparable trade indicators and the external-shock series already used by the project. Kuwait now enters the debt-scenario layer through its separately governed reduced-data specification.

Terms of tradeRelative export/import prices

Net barter terms of trade are transformed into annual log changes and standardized using each country's historical volatility.

World demandGlobal real activity

World real-GDP growth deviations provide a common external demand shock across the Arab-country panel.

Energy tradeCountry-oriented energy shock

Oil and LNG price movements are oriented by the importer/exporter role so a positive shock is favourable and a negative shock is adverse for the country.

Export demandExport-volume innovation

Export-demand stress is based on innovations in export-volume growth rather than an arbitrary percentage change in trade value.

Energy-role construction

Energy importers receive the opposite orientation from energy exporters. Qatar is treated as a gas exporter using a 35-percent oil and 65-percent LNG price mix before re-standardization. The construction is designed so the sign has a consistent economic interpretation across countries.

GVC structural exposure and production validation

ACD now reports genuine UNCTAD–Eora GVC indicators as 2015–2018 structural averages for backward participation, forward participation, total GVC participation and GVC position. These measures are treated as slow-moving production-network characteristics rather than as 2024 observations. They are available for descriptive country and regional intelligence but are not used as scenario multipliers.

Trade × GVC interactions were evaluated in 03G and then subjected to the 03H production gate: explicit hydrocarbon- exporter shock heterogeneity, restricted wild-cluster bootstrap-t inference, leave-one-country-out sign stability and Benjamini–Hochberg multiple-testing adjustment. Eleven interactions entered the robustness stage and zero satisfied the full production-eligibility criteria. The validated 03C debt-scenario coefficients therefore remain unchanged.

08

Trade-to-macro transmission estimation

The operational trade layer is estimated using pooled country fixed effects with standard errors clustered by country. Each equation contains the contemporaneous shock and one annual lag. Country-specific and exporter/importer-group regressions are retained as diagnostics; they are not used as the operational coefficients in the scenario engine.

yi,t=αi+β0zi,t+β1zi,t1+γXi,t+ui,t(11)

Operational channels pass a reliability gate based on the cumulative current-plus-lag effect, statistical support at the 10-percent level and the expected economic sign where the sign is theoretically unambiguous. Channels that fail the gate are exported with zero operational coefficients while their raw diagnostic estimates remain preserved.

ShockResponseCumulative 1-SD effectOperational status
Terms of tradeReal GDP growth+0.695 ppOperational
Terms of tradePrimary balance+1.803 pp GDPOperational
Terms of tradeCurrent account+2.617 pp GDPOperational
World growthReal GDP growth+1.894 ppOperational
World growthPrimary balance+1.890 pp GDPOperational
World growthSovereign spread−66.15 bpOperational
World growthCurrent account+0.860 pp GDPOperational
Favourable energy tradePrimary balance+2.427 pp GDPOperational
Favourable energy tradeCurrent account+2.737 pp GDPOperational
Export-demand innovationReal GDP growth+1.071 ppAugmented
Export-demand innovationCurrent account+1.778 pp GDPAugmented

No direct inflation channel passes the current operational reliability gate. Energy-trade shocks therefore affect debt primarily through the primary balance and external-account response in the present specification, while world-demand shocks additionally transmit through growth and, where available, the sovereign spread.

Interpretation of pooled coefficients

Cross-country differences in trade scenario debt results should not be interpreted as 12 separately estimated structural trade elasticities. The operational transmission coefficients are pooled. Country differences arise from the economic-to-SD translation, country debt baselines, financing structure and debt dynamics.

09

Economic trade-shock translation

The user enters an economically interpretable assumption. A dedicated translation layer converts that assumption into the exact standardized shock units used by the trade-transmission estimates. Exposure variables are descriptive and are not used as arbitrary multipliers on pooled coefficients.

Terms-of-trade percentage change
ziToT=100ln(1+ΔToTi100)si,ToT(12)

A user-specified percentage change is therefore scaled by the country's observed historical terms-of-trade volatility.

World-growth deviation
zW=ΔgWsgW(13)

The same global-growth disturbance is expressed relative to the historical volatility of the common world-growth series.

Broad energy-price assumption
ziE=qi*si,E(14)

Importer/exporter sign orientation and, where relevant, energy composition are applied before standardization.

Export-demand innovation
ziX=νiXsi,X(15)

This is an export-volume innovation, not a direct percentage change in export value.

Country-specific economic interpretation

A common economic shock does not necessarily imply the same standardized shock across countries. For example, the same energy-price movement maps differently when historical volatility and energy-trade role differ.

10

Validation, stability and evidence tiers

Validation is not treated as a single pass/fail statistic. Climate models are reviewed for dynamic stability, MCMC behaviour, posterior identification, parameter-bound sensitivity, persistence, climate reconstruction, debt-baseline reconciliation and economic interpretation. The trade layer is separately validated through data coverage checks, sign and significance gates, exact zero-shock reproduction and end-to-end economic-scenario replication.

Climate-model stability

Published systems passed the required structural model checks after documented country regularization. Posterior convergence is reported separately and is not relabelled as a pass where the original diagnostic thresholds were not met.

Climate reconstruction

The physical translator reproduces the annual climate-stress index to numerical precision across the modelled countries.

Trade zero-shock test

The trade engine reproduces the stored baseline exactly when the standardized trade shock is set to zero.

Economic-scenario replication

Economic assumptions translated by the trade layer reproduce the same debt paths as the corresponding standardized 03C scenario.

Material caveats are retained, not hidden

Examples include Lebanon's pre-2019 structural climate sample, Bahrain's stability regularization, Tunisia's persistent debt dynamics, shorter climate-model samples for the UAE and Iraq, pooled trade-transmission coefficients and the absence of operational GVC amplification.

11

Scenario engines

The Scenario Lab exposes three user-facing analytical modes: standardized climate stress, physical climate pathways and trade or external-sector stress. Climate and trade scenarios are executed by separate engines but converge on a common debt accounting framework.

Climate · standardizedModel shock in SD units

A standardized climate-state disturbance is propagated through the country Bayesian response system.

Climate · physicalTemperature and precipitation

Physical assumptions are translated to the same standardized climate state before macro-fiscal propagation.

Trade · economicEconomic assumption to debt

An economic trade assumption is translated to SD units, transmitted through the operational trade equations and then passed to the sovereign debt identity.

Climate shock propagation

εc,t=ctρcct1(16)
Rj,t=k=0tεc,kIRFj,tk(17)

Trade shock propagation

Rj,t=ρjRj,t1+β0,jzt+β1,jzt1(18)

Standardized trade stress uses positive values for favourable or expansionary shocks and negative values for adverse shocks. The economic translator applies this sign convention before the debt engine is called.

12

Public-debt reconstruction

The portal distinguishes model-state responses from the projected level debt path. Both climate and trade scenarios ultimately operate around the institutional baseline stored in the country bundle.

Dt=1+it1+gtnDt1PBt+SFAt+Ct(19)

For climate scenarios, Cₜ can represent an explicit direct climate fiscal-cost assumption when the user activates that feature. The trade engine does not add a direct trade fiscal cost; its effects enter through estimated macro-fiscal and financing responses. Current-account responses are reported as an external-sector diagnostic and are not mechanically inserted into the debt identity.

Climate debt pathCountry Bayesian responses + baseline

Output, inflation, fiscal and financing responses are applied recursively to the institutional debt baseline.

Trade debt pathPooled trade transmission + country debt dynamics

Trade-induced growth, primary-balance and applicable spread responses are applied to the same country debt accounting structure.

Output gap is not GDP growth

Climate v14.2 treats x as an output-gap level, so the annual real-GDP-growth response used in the climate debt identity is approximated by Δxₜ = xₜ − xₜ₋₁. The trade system instead estimates the real-GDP-growth response directly.

13

Baselines, source treatment and data lineage

Country scenarios are anchored to the baseline stored in each portal model bundle. Where a complete institutional projection is available, the debt identity is reconciled through an annual stock-flow adjustment so the model reproduces the published baseline before stress is introduced.

Climate lineageClimate history → translator → country model → debt path

Physical and standardized climate scenarios preserve the country model version and baseline metadata.

Trade lineageTrade panel → pooled transmission → translator → debt path

Trade scenario outputs preserve the economic assumption, standardized shock, operational channels and baseline source.

14

AI analytical layer

The AI assistant is designed as an orchestration and explanation layer over the analytical platform. It does not replace the climate or trade engines and should not generate sovereign-risk numbers from language-model memory.

Interpret the question

Extract country, shock family, economic magnitude, duration, start year and requested output from natural language.

Call analytical tools

Invoke climate, trade, comparison, data-catalogue and reporting endpoints rather than inventing results.

Explain with provenance

Translate engine output into decision-ready language while retaining assumptions, caveats and source information.

AI governance principle

The assistant may formulate and explain a scenario, but every quantitative result presented as a portal estimate must come from an analytical endpoint or stored validated output.

15

Model governance and versioning

Every published country is associated with a model tier, publication status, model version, data vintage, baseline vintage and explicit caveats in the country registry. Climate and trade engines are versioned separately so additions to the trade layer do not silently alter validated climate-model bundles.

  • Climate engine. Code 02 v14.2 remains the current country climate-debt scenario engine for the published final bundles.
  • Trade data. 03A builds the Arab trade-debt analytical panel without changing the climate models.
  • Trade transmission. 03B v2 estimates the pooled operational current-and-lag shock responses and exports only channels that pass the reliability gate.
  • Trade translation. 03D converts economic assumptions into standardized shock units without arbitrary exposure multipliers.
  • Trade debt engine. 03C v1 applies the operational responses to the standard v14.2 baselines; the Kuwait-compatible v1.1 availability layer uses the same equations and coefficients once the validated reduced-data Kuwait bundle exists. 03E provides an end-to-end economic scenario runner.
  • Kuwait. Published as a reduced-data specification. Climate-output and climate-primary-balance loadings are regional calibrations checked against Kuwait country diagnostics; the model must not be presented as a standard country-specific Dynare estimate.
16

Interpretation, scope and disclaimer

Portal outputs are conditional scenario estimates. Climate results describe model-implied macro-fiscal responses to selected climate stress. Trade results are semi-structural stress-test estimates based on pooled operational transmission coefficients and country-specific debt baselines and dynamics. Neither should be presented as an official forecast.

  • Conditional interpretation. Results depend on the selected baseline, shock magnitude, start year, duration, persistence and the relevant model configuration.
  • Climate interpretation. Temperature and precipitation enter the macro-fiscal system through the composite climate-stress state rather than as independent structural coefficients.
  • Trade interpretation. The current operational trade coefficients are pooled across the Arab-country panel; cross-country rankings are scenario sensitivities rather than structural rankings of exposure.
  • Current-account channel. Trade-induced current account responses are disclosed but are not mechanically added to the public-debt identity.
  • Financing spread. A spread response is only applied where the country debt bundle contains the corresponding spread block.
  • GVC. Structural GVC intelligence is operational as a descriptive layer using genuine UNCTAD–Eora 2015–2018 averages. No synthetic or post-2018 extrapolated GVC series are used. GVC amplification is inactive because zero interaction candidates passed the full 03H small-cluster production gate.
  • Uncertainty reporting. Trade scenarios expose conditional parameter-uncertainty envelopes from the validated bootstrap layer; these are not forecast probabilities. Climate scenarios emphasize central paths and deterministic sensitivity, while original posterior convergence and multimodality disclosures remain visible rather than being converted into probability claims.
17

Author

HM
Portal author and model developer

Hassan Mansour

Research and modelling interests include macro-fiscal analysis, sovereign debt, climate risk, trade transmission, economic modelling and sustainable development.