Complexity Insights
VOL. I  ·  APR 2026
§ III — The Portfolio

Applied research,
working models.

Four projects that turn complexity-economics theory into code you can run. Each is a question, the data brought to bear, and the result — reproducible and open.

↓ SCROLL FOR THE INDEX
FIG. 01 · CORR-MAT Σ · 11 SECTORS STREAMING
ρ ∈ [-1, +1] WINDOW · 60-DAY ROLL
Projects shipped
0
Across 3 methodological domains
Observations analyzed
M
County · ticker · agent-step
Lines of research code
K
Python · R · JavaScript
Public GitHub repos
Fully reproducible
III.AThe Toolkit

Methods, languages, taste.

The same stack across every project — no black boxes, everything on GitHub.
Financial analysis01
Financial modeling Risk assessment Valuation & synergy modeling
Data science02
Python · Pandas · NumPy SQL & time-series Reproducible notebooks
Research methods03
Econometrics Network science Agent-based modeling
Specialty04
Non-linear dynamics Correlation structure Complexity economics
§ III.B — The Portfolio 04 ENTRIES · 2 LIVE · 1 IN PROGRESS · 1 PLANNED
№ 001 Geospatial analysis ● LIVE BNK · COMPLETED 2025
~3,200 U.S. COUNTIES
BIVARIATE CHOROPLETH

Banking deserts in America.

Where does the formal financial system end — and what does it correlate with when it does?

Interactive bivariate choropleth mapping FDIC branch density against median household income across ~3,200 U.S. counties. Surfaces the geography of financial exclusion and the demographic patterns that accompany it.

3,243
Counties mapped
81K+
FDIC branches
4.2%
Unbanked desert
Leaflet.jsFDIC dataCensus ACSGeospatial
View the analysis → [ github ↗ ]
FIG · BIVARIATE3×3 LEGEND
DATA · FDIC + ACSN = 3,243
№ 002 Network analysis ● LIVE CORR · 5-YEAR WINDOW
N = 500 TICKERS
1.2M DATAPOINTS

S&P 500 systemic-risk analysis.

When diversification stops working, what does the correlation graph look like just before it fails?

Five years of daily returns, rolled into a weighted correlation graph and tracked through regime changes. Surfaces the sector clusters that tighten before volatility spikes and the ones that come apart.

0.71
Mean ρ, crisis regime
11
Sector communities
+38%
Edge density shift
PythonNetworkXCorrelation structureMicrostructure
View the analysis → [ github ↗ ]
FIG · FORCE-DIRECTEDG(500, ρ>0.4)
COMMUNITIES · LOUVAIN|V|=500
№ 003 Agent-based model ● IN PROGRESS LIQ · Q2 2026
N ≈ 200 AGENTS
3 POPULATIONS

Algorithmic liquidity-spiral model.

Under what population mix do fast strategies stop providing liquidity and start consuming it?

Agent-based simulation of price formation with heterogeneous trading rules — fundamentalist, chartist, and noise populations — calibrated against flash-crash microstructure. An extension of the front-page Lab demo toward a publishable model.

200
Heterogeneous agents
10K
Monte-Carlo runs
σ · 4.1×
Vol amplification
Python · MesaMonte CarloMicrostructureABM
In progress — Q2 2026 [ preview repo ↗ ]
FIG · PRICE PATHT = 500 STEPS
FUND · CHART · NOISEμ=100 σ=σ(t)
№ 004 Economic research ● PLANNED ECI · 2026
~120 COUNTRIES
HIDALGO–HAUSMANN

A global economic-complexity index.

What does the diversity and sophistication of a country's export basket forecast about its growth?

Comparative analysis of export diversification across developing economies. Reproduces and extends the Hidalgo–Hausmann framework, then tests its predictive power on GDP trajectories since the original publication.

~120
Countries
24yrs
Panel horizon
R² · ?
To be determined
RRegression analysisMacroeconomicsTrade data
Planned — 2026 [ read the working notes → ]
FIG · ECI vs GDPSCATTER · LOG-LOG
n ≈ 120HARVARD ATLAS
Commission a project

Have a problem that deserves a proper model?

Contract engagements for firms, research groups, and hiring pipelines — scoped to the question, priced to the scope.

Start a conversation →
© 2026 Complexity Insights · Nicholas Thomas · Washington, D.C. ← Return home
Tweaks ×
Animated correlation heatmap
Show count-up stats
Show methods · languages
Show data-viz previews
Changes persist to the file.