Fragile, Resilient, Asymmetric: An Agent-Based Stress Test of European Cooperation under Geopolitical Shocks (2024–2040) (2.0.0)
Agent-based model of cooperation among fourteen EU member states under geopolitical shocks, 2024-2040. States are adaptive agents on the observed intra-EU trade network, updating a cooperation propensity through inertia, network influence, and a repeated-game payoff structure built from verified Eurostat, Eurobarometer and IMF data. An anchoring device makes the observed 2024 configuration exactly stationary absent shocks, so results are read as deviations from an empirical baseline. Five scenarios of increasing severity are simulated with Monte Carlo replication, up to an extreme counterfactual crisis in the Taiwan Strait. Includes six robustness batteries and a full ODD description.
Release Notes
Version 2.0.0
New in this release:
-
Added robustezza_percentile.py: a percentile (rank-based) re-normalisation of all four payoff indicators, rebuilt from the raw 2021-2023 series. Ranks replace min-max scaling, discarding cardinal spacing and retaining only order; the country-specific anchor is recomputed accordingly and all five scenarios are re-run. The two initial configurations correlate at Spearman rho = 0.947. All three qualitative results survive: the strict ordering of scenarios is preserved (0.500 / 0.467 / 0.439 / 0.426 / 0.410, all adjacent gaps positive), fragmentation still compresses under stress (0.0478 to 0.0400, -16.4%), and the core-periphery gap still re-opens (-0.015 to +0.009). Levels shift upward by roughly 2.5 points, as they must when the measurement scale changes, which is why the model reads deviations from the anchored baseline rather than levels.
-
Robust data-path resolution across all scripts: the code now locates the data directory by searching upward, so it runs unchanged both in the CoMSES layout (code and data as sibling directories) and in a local clone.
-
Rewritten README with a script-to-result mapping table and a quick-start sequence.
-
Narrative documentation extended: the ODD description now includes a section documenting all six robustness batteries with their results and the script that produces each.
-
Initial conditions and indicator tables now reported to three decimals, so that the derived quantity c_i(2024) = (R_i + delta_i)/2 reproduces exactly from the published values.
Unchanged: the model itself, all input data, and every headline result. The five scenario outcomes at 2040 (0.476 / 0.448 / 0.425 / 0.414 / 0.400) and all distributional findings
Associated Publications
Fragile, Resilient, Asymmetric: An Agent-Based Stress Test of European Cooperation under Geopolitical Shocks (2024–2040) 2.0.0
Agent-based model of cooperation among fourteen EU member states under geopolitical shocks, 2024-2040. States are adaptive agents on the observed intra-EU trade network, updating a cooperation propensity through inertia, network influence, and a repeated-game payoff structure built from verified Eurostat, Eurobarometer and IMF data. An anchoring device makes the observed 2024 configuration exactly stationary absent shocks, so results are read as deviations from an empirical baseline. Five scenarios of increasing severity are simulated with Monte Carlo replication, up to an extreme counterfactual crisis in the Taiwan Strait. Includes six robustness batteries and a full ODD description.
Release Notes
Version 2.0.0
New in this release:
-
Added robustezza_percentile.py: a percentile (rank-based) re-normalisation of all four payoff indicators, rebuilt from the raw 2021-2023 series. Ranks replace min-max scaling, discarding cardinal spacing and retaining only order; the country-specific anchor is recomputed accordingly and all five scenarios are re-run. The two initial configurations correlate at Spearman rho = 0.947. All three qualitative results survive: the strict ordering of scenarios is preserved (0.500 / 0.467 / 0.439 / 0.426 / 0.410, all adjacent gaps positive), fragmentation still compresses under stress (0.0478 to 0.0400, -16.4%), and the core-periphery gap still re-opens (-0.015 to +0.009). Levels shift upward by roughly 2.5 points, as they must when the measurement scale changes, which is why the model reads deviations from the anchored baseline rather than levels.
-
Robust data-path resolution across all scripts: the code now locates the data directory by searching upward, so it runs unchanged both in the CoMSES layout (code and data as sibling directories) and in a local clone.
-
Rewritten README with a script-to-result mapping table and a quick-start sequence.
-
Narrative documentation extended: the ODD description now includes a section documenting all six robustness batteries with their results and the script that produces each.
-
Initial conditions and indicator tables now reported to three decimals, so that the derived quantity c_i(2024) = (R_i + delta_i)/2 reproduces exactly from the published values.
Unchanged: the model itself, all input data, and every headline result. The five scenario outcomes at 2040 (0.476 / 0.448 / 0.425 / 0.414 / 0.400) and all distributional findings