Skip to content

complexity-econ/paper-07-io-coupling

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

I-O Sectoral Coupling and the Leontief Cascade Threshold

Read the paper (PDF)

Seventh paper in the complexity-econ series. Introduces input-output sectoral coupling into the SFC-ABM engine via a 6×6 technical coefficients matrix calibrated from GUS symmetric I-O tables (2015/2019). Discovers a sharp first-order phase transition at critical coupling strength κ_c ≈ 0.93, universal across BDP levels.

Key Findings

  • Critical coupling strength κ_c ≈ 0.93: sharp 1st-order phase transition in I-O pass-through, universal across BDP levels. Below κ_c, I-O is a mild perturbation of baseline (±3 pp adoption). Above κ_c, Leontief cascade dominates---deflation spiral, mass unemployment (87%), NPL explosion.
  • Transition is BDP-independent: same κ_c at BDP=500 and BDP=2000. At κ=1.0, cascade triggers even at BDP=0 (natural AI adoption seeds the supply chain feedback loop).
  • Reentrant inverted-U survives: the adoption peak at BDP≈500 from Paper-04 is preserved for all I-O variants below κ_c. At κ=1.0, the reentrant peak shifts to BDP≈2250.
  • Manufacturing is the hub sector: backward linkage BL=2.04, forward linkage FL=2.69---highest in both directions, consistent with Acemoglu et al. (2012) cascade theory.
  • Empirically grounded range κ ∈ [0.75, 0.90]: produces +10% unemployment amplification, consistent with Baqaee & Farhi (2020) estimates. The cascade at κ ≥ 0.95 is an artifact of fixed A, instantaneous pass-through, and capacity-proportional revenue.
  • Three propagation channels: demand (Keynesian), supply chain (Leontief), and credit (SFC/Minsky) interact nonlinearly through the intermediate market.

Simulation Campaigns

Campaign Simulations Description
C1 Validation 300 No-IO vs IO baseline, 5 BDP levels × 2 modes × 30 seeds
C2 Scale Sweep 660 11 κ levels (0.0--1.0) × 2 BDP (500, 2000) × 30 seeds
C3 BDP Phase 2,520 21 BDP levels (0--5,000) × 4 I-O variants × 30 seeds
C4 Sectoral 150 5 BDP levels × κ=0.85 × 30 seeds
Total 3,630

Figures

Scale Sweep: Phase Transition (C2)

fig01 Fig 1. Adoption and unemployment vs κ at BDP=500 and BDP=2000 (left: terminal, right: time evolution). Green band marks empirical range κ ∈ [0.75, 0.90]. The κ=1.00 trajectory diverges catastrophically after M30.

I-O Flow Volume (C2)

fig02 Fig 2. I-O flows vs coupling strength (left: terminal, right: timeseries at BDP=2000). Flows peak at κ≈0.85 then collapse---at κ=1.00, firms bankrupt cascadingly through the supply chain and flows drop to zero.

Adoption Phase Diagram (C3)

fig03 Fig 3. Adoption vs BDP for four I-O variants (left: terminal, right: timeseries for no-IO baseline). The reentrant inverted-U is preserved across all sub-critical I-O strengths.

Unemployment Phase Diagram (C3)

fig04 Fig 4. Unemployment vs BDP (left: terminal, right: timeseries). Clear post-shock bifurcation: BDP=0 rises fastest (no compensation), BDP=5000 also high (inflation-induced tightening).

Inflation Phase Diagram (C3)

fig05 Fig 5. Inflation vs BDP (left: terminal, right: timeseries). All BDP paths converge to deflation post-shock. BDP=5000 delays onset by ~20 months but ends at -20%.

Intermediate Market Volume by BDP (C4)

fig06 Fig 6. I-O flows at κ=0.85 (left: terminal bar chart, right: timeseries). Higher BDP reduces I-O volume---UBI-sustained demand dampens adoption, hence fewer automated firms purchasing intermediate inputs.

Repository Structure

analysis/              — 6 Python scripts generating figures
figures/               — Generated PNG figures (200 DPI)
latex/                 — Paper source (XeLaTeX + biblatex)
simulations/
  scripts/             — 4 campaign runner scripts + run_all.sh
  results/             — Terminal + timeseries CSV files (European format: ; separator, , decimal)

Dependencies

  • Engine: complexity-econ/core (Scala 3.5.2, sbt)
  • Analysis: Python 3 (matplotlib, numpy, pandas)
  • Paper: XeLaTeX + biblatex

Running

# Build the engine JAR first
cd ../core && sbt assembly

# Run all simulation campaigns
cd simulations/scripts && bash run_all.sh

# Generate all figures
for f in analysis/python/fig*.py; do python3 "$f"; done

# Compile paper
cd latex && xelatex paper_en.tex && biber paper_en && xelatex paper_en.tex && xelatex paper_en.tex

Series

# Paper DOI
01 The Acceleration Paradox 10.5281/zenodo.18727928
02 Monetary Regime & Automation 10.5281/zenodo.18740933
03 Empirical σ Estimation 10.5281/zenodo.18743780
04 Phase Diagram & Universality 10.5281/zenodo.18751083
05 Endogenous Technology & Networks 10.5281/zenodo.18758365
06 Heterogeneous Households pending
07 I-O Sectoral Coupling pending

License

MIT

Releases

No releases published

Packages

 
 
 

Contributors