Heterogeneous Households and the Limits of Universal Basic Income

Read the paper (PDF)

Sixth paper in the complexity-econ series. Replaces the aggregate household sector from Papers I--V with 100,000 individual agents, each with heterogeneous savings, debt, rent, MPC, skills, and health capital. Reveals that the critical point BDP_c = 500 PLN---previously identified as a policy "sweet spot"---is in fact the zone of maximum microstructural destruction.

Key Findings

• Aggregate metrics mask destruction: BDP_c = 500 shows low aggregate Gini (~0.20) but individual-level analysis reveals peak bankruptcy (17.3%), peak poverty (45%), and peak income Gini (~0.50)
• Scarring creates hysteresis: Skill decay (-2%/month) and health penalties (+2%/month) accumulate during unemployment, producing path-dependent human capital destruction invisible to aggregate models
• Retraining catch-22: Even enhanced retraining (2x probability, half cost, half duration) achieves only ~18% success rate---because success depends on skill × (1 - health penalty), both degraded by the unemployment spell preceding eligibility
• UBI solves income, not human capital: High BDP (≥3,000 PLN) reduces income Gini to 0.15 and poverty below 10%, but produces the highest scarring levels
• MPC heterogeneity is second-order: Three MPC distributions (low/baseline/high variance) produce similar macro outcomes---automation dynamics dominate consumption heterogeneity

Simulation Campaigns

Campaign	Simulations	Description
C1 Validation	60	Aggregate vs individual, 5 BDP levels × 2 modes × 30 seeds
C2 BDP Sweep	630	21 BDP levels (0--5,000 by 250) × 30 seeds
C3 MPC Sensitivity	270	3 MPC distributions × 3 BDP × 30 seeds
C4 Wealth Sensitivity	270	3 initial wealth configs × 3 BDP × 30 seeds
C5 Retraining Policy	270	3 retraining configs × 3 BDP × 30 seeds
Total	1,500

Figures

Distribution Evolution (C2)

Fig 1. Unemployment rate distribution over time by BDP level. Violin widths show cross-seed variation. Higher BDP produces both higher mean unemployment and greater dispersion.

Savings Trajectories (C2)

Fig 2. Terminal savings (mean and median) by BDP. Mean > median at all levels (right-skewed). Dip at BDP=500 reflects selective attrition of low-savings households.

Gini Decomposition (C2)

Fig 3. Income Gini peaks ~0.50 at BDP_c, then falls as UBI provides income floor. Wealth Gini flat at ~0.14---UBI redistributes income but not wealth.

Bankruptcy Cascade (C2)

Fig 4. Bankruptcy rate peaks at BDP=500 (17.3%)---the critical point is the zone of maximum household destruction, not a sweet spot.

Retraining Funnel (C5)

Fig 5. Enhanced retraining generates ~2x more attempts but success rate stays ~18%. The catch-22: scarring degrades the human capital base that retraining operates on.

Welfare Comparison (C2)

Fig 6. Income Gini and poverty rate by BDP. Non-monotonic poverty: peaks ~45% at BDP=500--1,000. Moderate UBI is worse than no UBI on poverty metrics.

MPC Sensitivity (C3)

Fig 7. Consumption distribution across three MPC regimes. Similar P10--P90 bands confirm that macro dynamics dominate micro consumption heterogeneity.

Scarring Dynamics (C2)

Fig 8. Skill decay and health penalty increase monotonically with BDP. More automation → longer unemployment → deeper human capital destruction. Insets show theoretical single-agent scarring curves.

Repository Structure

analysis/              — 8 Python scripts generating figures
figures/               — Generated PNG figures (200 DPI)
paper/                 — Paper source (XeLaTeX + biblatex)
simulations/
  scripts/             — 5 campaign runner scripts + run_all.sh
  results/             — Terminal 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 (~24h with 10K firms + 100K HH)
cd simulations/scripts && bash run_all.sh

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

# Compile paper
cd paper && xelatex paper.tex && bibtex paper && xelatex paper.tex && xelatex paper.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

License

MIT