diff --git a/analyze_cables_concentration.py b/analyze_cables_concentration.py
new file mode 100644
index 0000000..5bc9ad7
--- /dev/null
+++ b/analyze_cables_concentration.py
@@ -0,0 +1,236 @@
+#!/usr/bin/env python3
+"""Quick statistical analysis: are US data centers spatially tied to submarine
+cables, and does the resulting pattern look like concentrated costs / dispersed
+benefits?
+"""
+import math
+import os
+import statistics
+from collections import Counter
+
+import psycopg2
+
+
+def connect():
+ return psycopg2.connect(
+ host=os.environ["PGWEB_HOST"],
+ port=os.environ["PGWEB_PORT"],
+ user=os.environ["PGWEB_USER"],
+ password=os.environ["PGWEB_PASSWORD"],
+ dbname="data_centers",
+ )
+
+
+def quantiles(xs, qs=(0.1, 0.25, 0.5, 0.75, 0.9)):
+ s = sorted(xs)
+ n = len(s)
+ out = {}
+ for q in qs:
+ if n == 0:
+ out[q] = None
+ continue
+ k = (n - 1) * q
+ lo, hi = math.floor(k), math.ceil(k)
+ if lo == hi:
+ out[q] = s[int(k)]
+ else:
+ out[q] = s[lo] + (s[hi] - s[lo]) * (k - lo)
+ return out
+
+
+def gini(values):
+ """Standard Gini coefficient for non-negative values."""
+ v = sorted(x for x in values if x is not None and x >= 0)
+ n = len(v)
+ if n == 0 or sum(v) == 0:
+ return None
+ cum = 0.0
+ for i, x in enumerate(v, 1):
+ cum += i * x
+ return (2 * cum) / (n * sum(v)) - (n + 1) / n
+
+
+def hhi(shares):
+ """Herfindahl-Hirschman Index on shares that sum to ~1. Returns value in [0, 1]."""
+ return sum(s * s for s in shares)
+
+
+def mann_whitney_u_z(xs, ys):
+ """Approximate Mann-Whitney U test z-score (normal approx, large-n).
+ Returns (U, z, p_two_sided). Uses average ranks for ties.
+ """
+ combined = [(v, 0) for v in xs] + [(v, 1) for v in ys]
+ combined.sort(key=lambda t: t[0])
+ ranks = [0.0] * len(combined)
+ i = 0
+ n = len(combined)
+ while i < n:
+ j = i
+ while j + 1 < n and combined[j + 1][0] == combined[i][0]:
+ j += 1
+ avg_rank = (i + j) / 2 + 1
+ for k in range(i, j + 1):
+ ranks[k] = avg_rank
+ i = j + 1
+ r1 = sum(r for r, (_, g) in zip(ranks, combined) if g == 0)
+ n1, n2 = len(xs), len(ys)
+ U1 = r1 - n1 * (n1 + 1) / 2
+ mu = n1 * n2 / 2
+ sigma = math.sqrt(n1 * n2 * (n1 + n2 + 1) / 12)
+ z = (U1 - mu) / sigma if sigma else 0.0
+ # Two-sided p via error function
+ p = math.erfc(abs(z) / math.sqrt(2))
+ return U1, z, p
+
+
+def main():
+ conn = connect()
+ cur = conn.cursor()
+
+ # --- 1. Distance from each US data center to nearest submarine cable ---
+ cur.execute(
+ """
+ with cables_union as (
+ select ST_Union(geom)::geography as g from public.internet_cables
+ )
+ select ST_Distance(
+ ST_SetSRID(ST_MakePoint(dc.longitude, dc.latitude), 4326)::geography,
+ cu.g
+ ) / 1000.0 -- meters -> km
+ from public.us_dc_sample_geocoded dc, cables_union cu
+ where dc.longitude is not null and dc.latitude is not null
+ and (dc.country = 'United States' or dc.country is null)
+ """
+ )
+ dc_km = [float(r[0]) for r in cur.fetchall()]
+
+ # --- 2. Distance from US city-dominance points to nearest cable ---
+ cur.execute(
+ """
+ with cables_union as (
+ select ST_Union(geom)::geography as g from public.internet_cables
+ )
+ select ST_Distance(
+ ST_SetSRID(ST_MakePoint(c.longitude, c.latitude), 4326)::geography,
+ cu.g
+ ) / 1000.0
+ from public.internet_city_dominance c, cables_union cu
+ where c.country = 'US' and c.geom is not null
+ """
+ )
+ city_km = [float(r[0]) for r in cur.fetchall()]
+
+ # --- 3. DC distribution by state (cost concentration) ---
+ cur.execute(
+ """
+ select coalesce(nullif(state_code, ''), 'UNK') as st, count(*)
+ from public.us_dc_sample_geocoded
+ where longitude is not null and latitude is not null
+ group by 1
+ """
+ )
+ state_counts = dict(cur.fetchall())
+ total_dc = sum(state_counts.values())
+ state_shares = {k: v / total_dc for k, v in state_counts.items()}
+
+ # --- 4. IP distribution across US cities (benefit dispersion proxy) ---
+ cur.execute(
+ """
+ select city, coalesce(logical_dominance_ips, 0)
+ from public.internet_city_dominance
+ where country = 'US' and logical_dominance_ips is not null
+ """
+ )
+ city_ips = [(r[0], int(r[1])) for r in cur.fetchall()]
+ total_ips = sum(v for _, v in city_ips)
+ ip_shares = [v / total_ips for _, v in city_ips] if total_ips else []
+
+ # --- 5. Where do the people-with-IPs LIVE relative to the DCs? ---
+ # Top-N US dominance cities, share of national IPs each captures.
+ top_ip_cities = sorted(city_ips, key=lambda t: -t[1])[:10]
+
+ cur.close()
+ conn.close()
+
+ # ======= report =======
+ print("=" * 70)
+ print("ARE US DATA CENTERS TIED TO SUBMARINE CABLE LANDINGS?")
+ print("=" * 70)
+ print(f"\nN data centers analyzed: {len(dc_km):,}")
+ print(f"N US city-dominance pts: {len(city_km):,}")
+
+ def fmt_q(label, xs):
+ q = quantiles(xs)
+ print(f"\n{label}:")
+ print(f" mean = {statistics.mean(xs):,.1f} km")
+ print(f" median (p50) = {q[0.5]:,.1f} km")
+ print(f" p10 / p25 / p75 / p90 = "
+ f"{q[0.1]:,.1f} / {q[0.25]:,.1f} / {q[0.75]:,.1f} / {q[0.9]:,.1f} km")
+ for thr in (10, 50, 100, 250):
+ frac = sum(1 for x in xs if x <= thr) / len(xs)
+ print(f" share within {thr:>3} km of a cable: {frac*100:5.1f}%")
+
+ fmt_q("Distance: US DATA CENTERS -> nearest submarine cable", dc_km)
+ fmt_q("Distance: US POPULATION CITIES -> nearest submarine cable", city_km)
+
+ U, z, p = mann_whitney_u_z(dc_km, city_km)
+ print(f"\nMann-Whitney U (DCs vs. cities, two-sided): U={U:,.0f}, z={z:.2f}, "
+ f"p≈{p:.2e}")
+ if statistics.median(dc_km) < statistics.median(city_km):
+ diff = statistics.median(city_km) - statistics.median(dc_km)
+ print(f" -> DC median is {diff:,.1f} km CLOSER to cables than city median.")
+ else:
+ print(" -> DCs are not closer to cables than cities.")
+
+ print("\n" + "=" * 70)
+ print("CONCENTRATION OF COSTS (data centers by state)")
+ print("=" * 70)
+ g_dc = gini(list(state_counts.values()))
+ h_dc = hhi(list(state_shares.values()))
+ print(f"States covered: {len(state_counts)}")
+ print(f"Gini of DC counts across states: {g_dc:.3f} (0=even, 1=one state takes all)")
+ print(f"HHI of state shares: {h_dc:.3f} (0.18+ = highly concentrated)")
+ top_states = sorted(state_shares.items(), key=lambda t: -t[1])[:8]
+ cum = 0.0
+ print(f"\nTop states by share of US data centers:")
+ for st, s in top_states:
+ cum += s
+ print(f" {st}: {s*100:5.1f}% ({state_counts[st]:>4} DCs) cum={cum*100:5.1f}%")
+
+ print("\n" + "=" * 70)
+ print("DISPERSION OF BENEFITS (US IPs across cities)")
+ print("=" * 70)
+ g_ip = gini([v for _, v in city_ips])
+ h_ip = hhi(ip_shares)
+ print(f"US cities with IP data: {len(city_ips):,}")
+ print(f"Gini of IPs across cities: {g_ip:.3f}")
+ print(f"HHI of IP shares: {h_ip:.3f}")
+ cum = 0.0
+ print(f"\nTop US cities by share of national IPs:")
+ for city, ips in top_ip_cities:
+ s = ips / total_ips
+ cum += s
+ print(f" {city:<30} {s*100:5.2f}% ({ips:>11,} IPs) cum={cum*100:5.2f}%")
+
+ print("\n" + "=" * 70)
+ print("INTERPRETATION")
+ print("=" * 70)
+ print(f"""
+Cost concentration (DCs across states): Gini={g_dc:.3f} HHI={h_dc:.3f}
+Benefit dispersion (IPs across cities): Gini={g_ip:.3f} HHI={h_ip:.3f}
+
+A "concentrated costs / dispersed benefits" pattern requires:
+ (1) DCs cluster in a few places (high state-level Gini/HHI).
+ (2) Users they serve span many places (low city-level Gini/HHI, ideally).
+ (3) That clustering is plausibly tied to fixed infrastructure (cables).
+
+Check signs above:
+ - DC location vs cable proximity: see Mann-Whitney result.
+ - Cost concentration: compare DC HHI to a "competitive" benchmark (~0.10).
+ - Benefit dispersion: IP-share Gini >> DC-state Gini would corroborate
+ the asymmetry (benefits more evenly distributed than costs).
+""")
+
+
+if __name__ == "__main__":
+ main()
diff --git a/analyze_dc_tract_concentration.py b/analyze_dc_tract_concentration.py
new file mode 100644
index 0000000..7fc7a79
--- /dev/null
+++ b/analyze_dc_tract_concentration.py
@@ -0,0 +1,320 @@
+#!/usr/bin/env python3
+"""Tract-level analysis of the 'concentrated costs / dispersed benefits' frame
+for US data-center siting.
+
+Cost-bearing universe = tracts that host at least one DC
+ (public.data_center_census_tracts_2024)
+Comparison universe = ACS 2024 5-yr tracts in the selected states
+ (census_tract_acs_2024_selected_states.csv)
+"""
+import csv
+import math
+import os
+import statistics
+from collections import Counter
+
+import psycopg2
+
+
+CSV_PATH = "census_tract_acs_2024_selected_states.csv"
+
+
+def connect():
+ return psycopg2.connect(
+ host=os.environ["PGWEB_HOST"],
+ port=os.environ["PGWEB_PORT"],
+ user=os.environ["PGWEB_USER"],
+ password=os.environ["PGWEB_PASSWORD"],
+ dbname="data_centers",
+ )
+
+
+def gini(values):
+ v = sorted(x for x in values if x is not None and x >= 0)
+ n = len(v)
+ if n == 0 or sum(v) == 0:
+ return None
+ cum = sum(i * x for i, x in enumerate(v, 1))
+ return (2 * cum) / (n * sum(v)) - (n + 1) / n
+
+
+def hhi(shares):
+ return sum(s * s for s in shares)
+
+
+def median(xs):
+ xs = [x for x in xs if x is not None]
+ return statistics.median(xs) if xs else None
+
+
+def mean(xs):
+ xs = [x for x in xs if x is not None]
+ return statistics.mean(xs) if xs else None
+
+
+def wmean(xs, ws):
+ pairs = [(x, w) for x, w in zip(xs, ws) if x is not None and w is not None and w > 0]
+ if not pairs:
+ return None
+ total = sum(w for _, w in pairs)
+ return sum(x * w for x, w in pairs) / total
+
+
+def to_float(s):
+ try:
+ return float(s)
+ except (TypeError, ValueError):
+ return None
+
+
+def to_int(s):
+ try:
+ return int(float(s))
+ except (TypeError, ValueError):
+ return None
+
+
+def main():
+ conn = connect()
+ cur = conn.cursor()
+
+ # DC-hosting tracts (the cost-bearing universe) ----------------------
+ cur.execute(
+ """
+ select
+ geoid,
+ statefp,
+ data_center_count,
+ population,
+ households,
+ broadband_subscription_pct,
+ median_household_income,
+ per_capita_income,
+ poverty_rate,
+ non_hispanic_white_pct,
+ non_hispanic_black_pct,
+ hispanic_latino_pct,
+ non_hispanic_asian_pct,
+ primary_industry,
+ land_area_sqm,
+ industry_information_workers,
+ industry_total_workers
+ from public.data_center_census_tracts_2024
+ """
+ )
+ dc_tracts = []
+ for r in cur.fetchall():
+ dc_tracts.append(
+ {
+ "geoid": r[0],
+ "statefp": r[1],
+ "dc_count": r[2] or 0,
+ "pop": r[3],
+ "hh": r[4],
+ "broadband_pct": float(r[5]) if r[5] is not None else None,
+ "mhi": r[6],
+ "pci": r[7],
+ "poverty": float(r[8]) if r[8] is not None else None,
+ "white_pct": float(r[9]) if r[9] is not None else None,
+ "black_pct": float(r[10]) if r[10] is not None else None,
+ "hisp_pct": float(r[11]) if r[11] is not None else None,
+ "asian_pct": float(r[12]) if r[12] is not None else None,
+ "primary_industry": r[13],
+ "land_sqm": r[14],
+ "info_workers": r[15],
+ "total_workers": r[16],
+ }
+ )
+
+ # Distance from each DC tract to nearest cable (km) ----------------
+ cur.execute(
+ """
+ with cables as (select ST_Union(geom)::geography g from public.internet_cables)
+ select t.geoid,
+ ST_Distance(ST_Centroid(t.geom)::geography, c.g) / 1000.0
+ from public.data_center_census_tracts_2024 t, cables c
+ """
+ )
+ dist_by_geoid = {r[0]: float(r[1]) for r in cur.fetchall()}
+ for t in dc_tracts:
+ t["dist_km"] = dist_by_geoid.get(t["geoid"])
+
+ cur.close()
+ conn.close()
+
+ # Comparison universe from the wider ACS CSV ------------------------
+ universe = []
+ with open(CSV_PATH, newline="", encoding="utf-8") as f:
+ for row in csv.DictReader(f):
+ universe.append(
+ {
+ "geoid": row["geoid"],
+ "statefp": row["statefp"],
+ "pop": to_int(row["population"]),
+ "broadband_pct": to_float(row["broadband_subscription_pct"]),
+ "mhi": to_int(row["median_household_income"]),
+ "pci": to_int(row["per_capita_income"]),
+ "poverty": to_float(row["poverty_rate"]),
+ "white_pct": to_float(row["non_hispanic_white_pct"]),
+ "black_pct": to_float(row["non_hispanic_black_pct"]),
+ "hisp_pct": to_float(row["hispanic_latino_pct"]),
+ "asian_pct": to_float(row["non_hispanic_asian_pct"]),
+ }
+ )
+
+ dc_geoids = {t["geoid"] for t in dc_tracts}
+ non_dc = [u for u in universe if u["geoid"] not in dc_geoids]
+
+ # Restrict comparison to states actually represented in the DC sample
+ dc_states = {t["statefp"] for t in dc_tracts}
+ universe_in_dc_states = [u for u in universe if u["statefp"] in dc_states]
+ non_dc_in_dc_states = [u for u in non_dc if u["statefp"] in dc_states]
+
+ # ============== report ==============
+ print("=" * 72)
+ print("TRACT-LEVEL CONCENTRATED COSTS / DISPERSED BENEFITS ANALYSIS")
+ print("=" * 72)
+
+ total_dc = sum(t["dc_count"] for t in dc_tracts)
+ print(f"\nDC-hosting tracts: {len(dc_tracts):,}")
+ print(f"Data centers in those tracts: {total_dc:,}")
+ print(f"ACS universe (selected states): {len(universe):,} tracts")
+ print(f"States represented in DC sample: {len(dc_states)}")
+ print(f"Universe restricted to DC states: {len(universe_in_dc_states):,} tracts")
+
+ # --- Cost concentration at the tract level ---
+ print("\n" + "-" * 72)
+ print("1. COST CONCENTRATION (DCs across tracts)")
+ print("-" * 72)
+ counts = [t["dc_count"] for t in dc_tracts]
+ shares = [c / total_dc for c in counts]
+ g_dc = gini(counts)
+ h_dc = hhi(shares)
+ print(f"Gini of DC counts across DC-hosting tracts: {g_dc:.3f}")
+ print(f"HHI of DC shares across DC-hosting tracts: {h_dc:.4f}")
+ # Top 1% / 5% of tracts share
+ top1 = max(1, len(counts) // 100)
+ top5 = max(1, len(counts) // 20)
+ s = sorted(counts, reverse=True)
+ print(f"Top 1% of DC-hosting tracts ({top1:>3} tracts) hold "
+ f"{sum(s[:top1])/total_dc*100:5.1f}% of all DCs")
+ print(f"Top 5% of DC-hosting tracts ({top5:>3} tracts) hold "
+ f"{sum(s[:top5])/total_dc*100:5.1f}% of all DCs")
+ print(f"Top 20% of DC-hosting tracts ({len(counts)//5:>3} tracts) hold "
+ f"{sum(s[:len(counts)//5])/total_dc*100:5.1f}% of all DCs")
+
+ # How small a fraction of population lives in a DC tract?
+ pop_dc = sum(t["pop"] or 0 for t in dc_tracts)
+ pop_universe = sum(u["pop"] or 0 for u in universe_in_dc_states)
+ print(f"\nPopulation living in a DC-hosting tract: {pop_dc:>11,}")
+ print(f"Total population (DC-states ACS universe): {pop_universe:>11,}")
+ if pop_universe:
+ print(f" -> {pop_dc/pop_universe*100:.2f}% of people in DC-hosting states "
+ f"live in a DC-hosting tract")
+ # Per-capita DC density
+ if pop_dc:
+ print(f" -> 1 DC per {pop_dc/total_dc:,.0f} residents in DC-hosting tracts")
+ if pop_universe and total_dc:
+ print(f" vs. 1 DC per {pop_universe/total_dc:,.0f} residents "
+ f"averaged across DC-state population")
+
+ # --- Profile of cost-bearing communities ---
+ print("\n" + "-" * 72)
+ print("2. WHO BEARS THE COSTS? (ACS profile of DC tracts vs. peer tracts)")
+ print("-" * 72)
+ fields = [
+ ("Median household income ($)", "mhi", "{:>10,.0f}"),
+ ("Per-capita income ($)", "pci", "{:>10,.0f}"),
+ ("Broadband subscription (%)", "broadband_pct", "{:>10,.1f}"),
+ ("Poverty rate (%)", "poverty", "{:>10,.1f}"),
+ ("Non-Hispanic White (%)", "white_pct", "{:>10,.1f}"),
+ ("Non-Hispanic Black (%)", "black_pct", "{:>10,.1f}"),
+ ("Hispanic/Latino (%)", "hisp_pct", "{:>10,.1f}"),
+ ("Non-Hispanic Asian (%)", "asian_pct", "{:>10,.1f}"),
+ ]
+ label_w = max(len(lbl) for lbl, *_ in fields)
+ print(f"{'Field':<{label_w}} {'DC tracts':>12} {'Non-DC peers':>14} "
+ f"{'Δ (DC − peer)':>15}")
+ for label, key, fmt in fields:
+ dc_med = median([t[key] for t in dc_tracts])
+ peer_med = median([u[key] for u in non_dc_in_dc_states])
+ if dc_med is None or peer_med is None:
+ continue
+ delta = dc_med - peer_med
+ cell_dc = fmt.format(dc_med)
+ cell_pe = fmt.format(peer_med)
+ cell_dl = fmt.format(delta)
+ print(f"{label:<{label_w}} {cell_dc} {cell_pe} {cell_dl}")
+
+ print("\nPopulation-weighted means (DC tracts):")
+ pops = [t["pop"] for t in dc_tracts]
+ for label, key, _ in fields:
+ wm = wmean([t[key] for t in dc_tracts], pops)
+ if wm is not None:
+ print(f" {label:<{label_w}} {wm:>12,.1f}")
+
+ print("\nPrimary-industry mix of DC-hosting tracts (count of tracts):")
+ for industry, n in Counter(t["primary_industry"] for t in dc_tracts).most_common(10):
+ print(f" {n:>4} {industry}")
+
+ # --- Cable vs. inland subgroups ---
+ print("\n" + "-" * 72)
+ print("3. CABLE-ADJACENT vs. INLAND DC TRACTS")
+ print("-" * 72)
+ near = [t for t in dc_tracts if t.get("dist_km") is not None and t["dist_km"] <= 100]
+ far = [t for t in dc_tracts if t.get("dist_km") is not None and t["dist_km"] > 100]
+ print(f"≤100 km from a submarine cable: {len(near):>3} tracts, "
+ f"{sum(t['dc_count'] for t in near):>4} DCs")
+ print(f">100 km from a submarine cable: {len(far):>3} tracts, "
+ f"{sum(t['dc_count'] for t in far):>4} DCs")
+ if near and far:
+ print(f"{' Median MHI':<28} near={median([t['mhi'] for t in near]):>10,.0f} "
+ f"far={median([t['mhi'] for t in far]):>10,.0f}")
+ print(f"{' Median broadband %':<28} near={median([t['broadband_pct'] for t in near]):>10,.1f} "
+ f"far={median([t['broadband_pct'] for t in far]):>10,.1f}")
+ print(f"{' Median DC count':<28} near={median([t['dc_count'] for t in near]):>10,.0f} "
+ f"far={median([t['dc_count'] for t in far]):>10,.0f}")
+
+ # --- Benefit-side proxy ---
+ print("\n" + "-" * 72)
+ print("4. BENEFIT DISPERSION (broadband subscribers across all tracts)")
+ print("-" * 72)
+ # Total broadband subscribers approx = households * broadband_pct
+ subs = []
+ for u in universe_in_dc_states:
+ hh = None # households not in CSV; use population/2.5 fallback
+ if u["pop"] and u["broadband_pct"] is not None:
+ est_hh = u["pop"] / 2.5
+ subs.append(est_hh * u["broadband_pct"] / 100.0)
+ total_subs = sum(subs)
+ sg = gini(subs)
+ sh = hhi([s / total_subs for s in subs]) if total_subs else None
+ print(f"Estimated total broadband subscribers (DC states): {total_subs:>14,.0f}")
+ print(f"Gini of subscribers across {len(subs):,} tracts: {sg:.3f}")
+ print(f"HHI of subscribers across tracts: {sh:.5f}")
+ # Compare to DC HHI
+ print(f"\nSide-by-side concentration (lower = more dispersed):")
+ print(f" HHI of DCs across DC-hosting tracts: {h_dc:.4f}")
+ print(f" HHI of broadband subs across DC-state tracts: {sh:.5f} "
+ f"({h_dc/sh:.0f}x more concentrated for DCs)")
+
+ print("\n" + "=" * 72)
+ print("BOTTOM LINE")
+ print("=" * 72)
+ n_above_500 = sum(1 for t in dc_tracts if t["dc_count"] >= 5)
+ print(f"""
+- DCs are extremely concentrated at the tract level: top 1% of host tracts
+ hold {sum(s[:top1])/total_dc*100:.0f}% of all DCs; top 5% hold {sum(s[:top5])/total_dc*100:.0f}%.
+- Only {pop_dc/pop_universe*100:.2f}% of residents of the DC-hosting states actually
+ live in a DC-hosting tract — costs (land use, power draw, water, traffic,
+ noise) fall on a tiny minority of communities.
+- DC-hosting tracts skew {'wealthier' if median([t['mhi'] for t in dc_tracts]) > median([u['mhi'] for u in non_dc_in_dc_states]) else 'poorer'} and {'higher' if median([t['broadband_pct'] for t in dc_tracts]) > median([u['broadband_pct'] for u in non_dc_in_dc_states]) else 'lower'} broadband than peer
+ tracts. See deltas above for the demographic profile.
+- Broadband subscribers (proxy for who consumes cloud services) are far more
+ evenly distributed than DCs — HHI roughly {h_dc/sh:.0f}× lower.
+ That asymmetry IS the classic concentrated-cost / dispersed-benefit shape.
+""")
+
+
+if __name__ == "__main__":
+ main()
diff --git a/cables_concentration_report.docx b/cables_concentration_report.docx
new file mode 100644
index 0000000..2d61508
Binary files /dev/null and b/cables_concentration_report.docx differ
diff --git a/cables_concentration_report.md b/cables_concentration_report.md
new file mode 100644
index 0000000..c6bfb49
--- /dev/null
+++ b/cables_concentration_report.md
@@ -0,0 +1,194 @@
+# Data Centers, Submarine Cables, and the Concentrated-Costs / Dispersed-Benefits Frame
+
+**Author:** David Adams · **Date:** 2026-05-17
+**Data:** PostGIS `data_centers` DB — `us_dc_sample_geocoded` (1,489 DCs),
+`data_center_census_tracts_2024` (611 tracts, ACS 2024 5-yr enriched),
+`internet_cables` (693 cables), `internet_city_dominance` (4,552 cities),
+`census_tract_acs_2024_selected_states.csv` (83,811 tracts, 46 states).
+
+---
+
+## 1. Are US data centers spatially tied to submarine cables?
+
+Distance from each point to the nearest submarine cable line (km):
+
+| Group | n | Mean | p10 | p25 | **p50** | p75 | p90 | ≤10 km | ≤50 km | ≤100 km | ≤250 km |
+|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
+| US data centers | 1,489 | 358.7 | 21.6 | 163.1 | **276.1** | 477.4 | 867.4 | 5.2% | 16.8% | 21.4% | 32.2% |
+| US population cities | 1,291 | 339.7 | 18.7 | 61.2 | **256.1** | 528.0 | 811.0 | 6.8% | 22.5% | 31.8% | 49.5% |
+
+Mann-Whitney U two-sided: **z = 2.66, p ≈ 0.008** — significant, but in the
+*opposite* direction. DCs are **not** systematically closer to cables than
+ordinary US cities.
+
+**Interpretation.** At the national level the "cables drive DC siting" story
+fails. The largest clusters — Loudoun County VA (Ashburn), central
+Washington, Hillsboro OR, Columbus OH, Iowa — are inland, anchored to
+terrestrial fiber, cheap power, and tax incentives rather than submarine
+landings. Only 21.4% of DCs sit within 100 km of any cable.
+
+---
+
+## 2. Cost concentration at the state level
+
+| Measure | Value |
+|---|---:|
+| States covered | 46 |
+| Gini of DC counts across states | 0.648 |
+| HHI of state shares | 0.080 |
+
+Top states by share of US data centers:
+
+| State | DCs | Share | Cumulative |
+|---|---:|---:|---:|
+| VA | 319 | 21.4% | 21.4% |
+| CA | 129 | 8.7% | 30.1% |
+| TX | 120 | 8.1% | 38.1% |
+| OR | 102 | 6.9% | 45.0% |
+| WA | 90 | 6.0% | 51.0% |
+| OH | 69 | 4.6% | 55.7% |
+| AZ | 60 | 4.0% | 59.7% |
+| IA | 58 | 3.9% | 63.6% |
+
+Five states hold **half** of all US data centers.
+
+---
+
+## 3. Cost concentration at the tract level
+
+Much sharper than state-level:
+
+| Measure | Value |
+|---|---:|
+| DC-hosting tracts | 611 |
+| DCs in those tracts | 1,489 |
+| Gini of DC counts across DC-hosting tracts | 0.499 |
+| HHI of DC shares across DC-hosting tracts | 0.0069 |
+| **Top 1% of host tracts (6 tracts) hold** | **14.6% of all DCs** |
+| Top 5% of host tracts (30 tracts) hold | 33.3% of all DCs |
+| Top 20% of host tracts (122 tracts) hold | 60.6% of all DCs |
+
+Population scaling:
+
+| Metric | Value |
+|---|---:|
+| Population living in a DC-hosting tract | 2,868,863 |
+| Total population (DC-state ACS universe) | 332,343,349 |
+| **% of DC-host-state residents in a DC-hosting tract** | **0.86%** |
+| DCs per resident, DC-hosting tracts | 1 per 1,927 |
+| DCs per resident, DC-state average | 1 per 223,199 |
+| **Per-capita DC burden, host vs. average** | **~115×** |
+
+---
+
+## 4. Who bears the costs? (ACS profile of DC tracts vs. peer tracts in same states)
+
+| Field | DC tracts (median) | Non-DC peers (median) | Δ (DC − peer) |
+|---|---:|---:|---:|
+| Median household income ($) | 91,082 | 76,637 | **+14,446** |
+| Per-capita income ($) | 48,111 | 38,546 | +9,565 |
+| Broadband subscription (%) | 94.2 | 92.0 | +2.2 |
+| Poverty rate (%) | 8.8 | 10.8 | −2.0 |
+| Non-Hispanic White (%) | 52.4 | 64.7 | −12.3 |
+| Non-Hispanic Black (%) | 6.7 | 3.9 | +2.8 |
+| Hispanic/Latino (%) | 11.9 | 9.8 | +2.1 |
+| Non-Hispanic Asian (%) | 5.2 | 1.5 | +3.7 |
+
+Population-weighted means in DC tracts: MHI **$109,145**, broadband **93.2%**,
+poverty 11.1%. The actual residents of host communities are concentrated in
+affluent tech corridors (Loudoun, Silicon Valley, Seattle eastside,
+Hillsboro OR).
+
+Primary-industry mix of host tracts (count of tracts):
+
+| Tracts | Primary industry |
+|---:|---|
+| 351 | Educational services, and health care and social assistance |
+| 133 | Professional, scientific, management, administrative, and waste management services |
+| 35 | Manufacturing |
+| 26 | Arts, entertainment, recreation, accommodation, and food services |
+| 22 | Retail trade |
+| 14 | Agriculture, forestry, fishing and hunting, and mining |
+| 10 | Finance and insurance, and real estate and rental and leasing |
+| 9 | Construction |
+| 4 | Transportation and warehousing, and utilities |
+| 3 | Public administration |
+
+---
+
+## 5. Cable-adjacent vs. inland DC tracts
+
+| | ≤100 km from a cable | >100 km from a cable |
+|---|---:|---:|
+| Tracts | 159 | 452 |
+| Data centers | 319 | 1,170 |
+| Median household income ($) | 106,406 | 86,289 |
+| Median broadband (%) | 95.2 | 93.9 |
+| Median DC count | 1 | 1 |
+
+Inland DCs are roughly **3.7×** the cable-adjacent count. Coastal/cable
+tracts skew even wealthier than inland DC tracts.
+
+---
+
+## 6. Benefit dispersion (broadband subscribers as a benefit proxy)
+
+| Measure | Value |
+|---|---:|
+| Estimated broadband subscribers (DC states) | 119,719,313 |
+| Tracts with subscriber data | 81,839 |
+| Gini of subscribers across tracts | 0.253 |
+| HHI of subscribers across tracts | 0.00001 |
+
+Side-by-side concentration:
+
+| Series | HHI |
+|---|---:|
+| DCs across DC-hosting tracts | 0.0069 |
+| Broadband subscribers across DC-state tracts | 0.00001 |
+| **Concentration ratio** | **~464× more concentrated for DCs** |
+
+---
+
+## 7. Verdict
+
+| Element of the frame | Holds? |
+|---|---|
+| Costs concentrated geographically | **Yes** — top 6 tracts carry 15% of DCs; <1% of host-state population lives in a DC tract; per-capita burden ~115× the average. |
+| Driven by submarine cable infrastructure | **No, broadly** — proximity test fails nationally; submarine cables matter for a coastal subset only. Terrestrial fiber, power, water, land, and tax incentives dominate. |
+| Benefits dispersed among users | **Yes** — broadband subscribers ~464× more dispersed (by HHI) than DCs. |
+| Classic political failure mode (weak losers vs. diffuse winners) | **No.** Host tracts skew wealthier, higher-income, higher-broadband than peers. The cost-bearing communities are affluent tech corridors with strong bargaining capacity — they tend to convert concentrated costs into concentrated *rents* (tax base, jobs, infrastructure concessions). |
+
+**Bottom line.** The structural asymmetry that defines "concentrated costs /
+dispersed benefits" is unambiguous in the data — DC siting is hyper-local
+while benefits are continental. But the predicted political dynamic doesn't
+fit cleanly, because the loser side here is not weak. A more targeted test
+would split host tracts into power-stressed exurban tracts (parts of
+Loudoun's edges, central Oregon, Iowa) and urban-suburban tech-corridor
+tracts, and look at whether the *exurban* subset shows the weak-loser
+pattern (lower income, slower broadband, higher poverty than its
+neighbors).
+
+---
+
+## Caveats
+
+- The ACS universe is the 46 DC-host states (already DC-heavy); excludes
+ states with no DCs in the sample.
+- `data_center_census_tracts_2024` only contains tracts that host at least
+ one DC, by construction.
+- Broadband-subscription rate is a coarse benefit proxy; cloud services
+ benefit any internet user globally, not just local subscribers.
+- 45 of 1,489 DCs use city-precision fallback coordinates, so a small share
+ of tract assignments are approximate.
+- The `logical_dominance_ips` field in `internet_city_dominance` measures
+ IP blocks routed/hosted at each city — a supply-side measure that
+ duplicates the DC signal, not a demand-side user-location measure. It
+ was excluded from the benefit-dispersion calculation for that reason.
+
+## Reproducible scripts
+
+- `load_postgis_internet_cables.py` — ingest cables/landings/cities
+- `make_internet_cables_map.py` — render the combined Leaflet map
+- `analyze_cables_concentration.py` — state-level + cable-proximity analysis
+- `analyze_dc_tract_concentration.py` — tract-level analysis used here
diff --git a/data_centers_cables_map.html b/data_centers_cables_map.html
new file mode 100644
index 0000000..3339d11
--- /dev/null
+++ b/data_centers_cables_map.html
@@ -0,0 +1,153 @@
+
+
+
+ US Data Centers + Submarine Cables
+
+
+
Data Centers + Cables
+
Data centers
+
Submarine cables
+
City dominance pts
+
+
Layers
+
+
+
+
+
Data center source
+
+
+
+
+
+
City dominance
+
+
+
+
US Data Centers + Submarine Cables
+
+
+
Data Centers + Cables
+
Data centers
+
Submarine cables
+
City dominance pts
+
+
Layers
+
+
+
+
+
Data center source
+
+
+
+
+
+
City dominance
+
+
+
+
US Data Centers + Submarine Cables
+
+
+
Data Centers + Cables
+
Data centers
+
Submarine cables
+
City dominance pts
+
+
Layers
+
+
+
+
+
Data center source
+
+
+
+
+
+
City dominance
+
+
+
+