salb23/ab-sfc-italy-energy: v0.1.0 — Conference-ready production model

First public release of the AB-SFC Italy framework, packaged as a fully reproducible, conference-ready production model. This release contains the complete codebase, the production driver notebook, and the reference outputs that back every numerical claim in the accompanying paper. What this release contains

model_full/ — the AB-SFC Python package (~3,000 lines), implementing households, firms, banks, energy producers and government within a fully stock-flow consistent architecture. Includes endogenous merit-order electricity dispatch, household heterogeneity calibrated on Italian microdata, bank credit creation with capital-adequacy constraints, Wright's-Law learning curves on renewable capacity, and a four-dimensional Hosseini–type Resilience Index integrated into the simulation environment. notebooks/SIE2026_production.ipynb — production driver: 14-cell scenario grid × 100 seeds under common random numbers, parallel execution on 4 cores, ~3 minutes runtime end-to-end. outputs_production/ — reference CSVs, figures (PNG) and LaTeX tables generated by the production run. Re-running the notebook reproduces these byte-for-byte under the fixed seed (SEED_BASE = 20260101). model_full/tests/ — SFC closure unit tests verifying that the Transaction Flow Matrix and Balance Sheet Matrix balance to floating-point tolerance at every step of every phase of the model.

Headline results reproduced by this release

Diversified generation stack + moderate corrective carbon price (τ = 85 €/tCO₂) dominates the no-policy gas-shock baseline jointly on emissions (−5,150 tCO₂, p < 10⁻⁴) and on the composite Resilience Index (+0.067 points, p < 10⁻⁴). Weitzman price-vs-quantity asymmetry is endogenously reproduced: tight emissions caps drive the wholesale electricity price to 940 €/MWh and collapse the Resilience composite to 0.39, even though they deliver the largest emissions reductions in the grid (−57.5%). High-efficiency appliance adoption is the primary household-side adjustment margin: H-class adoption rises from 3.3% (no-shock baseline) to 43.2% under the gas shock alone, to 63.7% under τ = 85, and saturates at 95% under the tightest cap. Policy instruments amplify, rather than create, this adaptive response.

How to reproduce bashgit clone https://github.com/salb23/ab-sfc-italy-energy.git cd ab-sfc-italy-energy pip install -r requirements.txt jupyter nbconvert --to notebook --execute notebooks/SIE2026_production.ipynb
--output /tmp/executed.ipynb This regenerates every CSV, figure and table in outputs_production/. Expected runtime ≈ 3 minutes on 4 cores. Methodological notes

Calibration: HFCS 2017 (quintile deposits), EU-SILC 2019 (energy expenditure shares, credit access), Istat ICP, ENTSO-E, GME MGP, TTF, Eurostat NRG_T1_11 and SBS. No proprietary or subscription-gated data are used. Inference: paired Wilcoxon signed-rank tests on common random numbers across scenarios; 90% bootstrap percentile CIs (2,000 resamples) on every aggregate KPI. Robustness: one-at-a-time sensitivity sweep over the four principal auxiliary behavioural parameters (χ, φ_pm, κ_invest, ρ_peer); the policy ranking established in the production grid is invariant across the sweep.

License & citation MIT-licensed. Cite via CITATION.cff (rendered automatically by GitHub's "Cite this repository" button) or via the BibTeX entry in the README.