74 rows → a single metric (annual profit pool, nominal $M + CPI→2020), linked to graph nodes, with independent verification of the figures.
42 normalized cells96 graph nodes tagged by margin5 facts verified
Headline
Migration of the annual profit pool across periods (nominal $M)
high marginmediumlow (commodity)
The profit pool grew ≈ $5M→$49M→$390M→$495M (P2→P5, nominal; ×CPI ≈ $0.2→1.4→5.5→6.4 billion in $2020). Color = margin: the mass of profit clearly shifts toward high-margin nodes (patent-protected lamps, aluminum, patents/advertising, real estate, financial capital), while commodities (copper/coal, in gray) grow by volume but stay thin on margin. P5: Insull's financial collapse of −$750M is one-off and shown separately (not in the stack).
Self-check (what had been missing)
Verification log — 5 load-bearing figures
Fact
What I had
Independent verification
Outcome
GE share of the lamp market, 1921
was 71.9%
69% of $68.3M (US v. GE 1926); ~85% including licensees
corrected
Collapse in the price of aluminum
was ×200
$4.86/lb (1888) → ~25¢ (1900s) ≈ ×20 (Hall); ~×48 from the pre-Hall $12
corrected
Coal's share of generation
was 56%
~50% from 1920, sustained for decades
refined
Insull collapse
~$750–800M / ~600k
confirmed: >500k investors, ~$750–800M, a $4 billion empire
confirmed
GE revenue/margin 1929 (~$390M/~16%)
secondary source
NOT independently confirmed (no public 1929 series)
flagged as unverified
3 corrections made, 1 confirmed, 1 flagged as unverified. Sources: US v. GE (Justia), History of aluminium (Wikipedia/ACS), Coal power US (Wikipedia/VisualizingEnergy), Insull collapse (chicagology/EBSCO).
Data discipline
Normalized model (42 cells)
Metric: annual profit ≈ revenue × margin in a representative year of the period; where no margin is available, a tier estimate by node type; confidence H/M/L. This is an order-of-magnitude estimation model, not accounting.
Segment
Period
$M nominal
$M (2020)
tier
conf
copper(elec)
P2 ~1888
$3M
$87M
lo
L
equipment_mfg
P2 ~1888
$2M
$58M
mid
L
utility/generation
P2 ~1888
$0.5M
$14M
lo
L
equipment_mfg
P3 ~1902
$12M
$336M
mid
M
lamps
P3 ~1902
$8M
$224M
hi
M
copper(elec)
P3 ~1902
$8M
$224M
lo
M
real_estate(uplift)
P3 ~1902
$5M
$140M
hi
L
coal
P3 ~1902
$5M
$140M
lo
L
utility/generation
P3 ~1902
$3M
$84M
lo
M
finance/holding
P3 ~1902
$3M
$84M
hi
L
aluminum/electrochem
P3 ~1902
$2M
$56M
hi
L
appliances
P3 ~1902
$1M
$28M
mid
L
transport
P3 ~1902
$1M
$28M
lo
M
hydro
P3 ~1902
$1M
$28M
mid
L
finance/holding
P4 ~1925
$80M
$1120M
hi
M
equipment_mfg
P4 ~1925
$60M
$840M
mid
H
real_estate(uplift)
P4 ~1925
$60M
$840M
hi
L
utility/generation
P4 ~1925
$45M
$630M
mid
M
copper(elec)
P4 ~1925
$40M
$560M
lo
M
appliances
P4 ~1925
$25M
$350M
mid
M
aluminum/electrochem
P4 ~1925
$20M
$280M
hi
M
lamps
P4 ~1925
$15M
$210M
hi
H
coal
P4 ~1925
$15M
$210M
lo
M
hydro
P4 ~1925
$10M
$140M
mid
L
refrigeration/cold
P4 ~1925
$8M
$112M
mid
L
patents_royalty
P4 ~1925
$7M
$98M
hi
M
broadcasting_ads
P4 ~1925
$5M
$70M
hi
M
transport
P4 ~1925
$-5M
$-70M
neg
M
utility/generation
P5 ~1940
$120M
$1560M
mid
M
equipment_mfg
P5 ~1940
$70M
$910M
mid
M
aluminum/electrochem
P5 ~1940
$60M
$780M
hi
M
real_estate(uplift)
P5 ~1940
$50M
$650M
hi
L
appliances
P5 ~1940
$40M
$520M
mid
M
refrigeration/cold
P5 ~1940
$35M
$455M
mid
M
copper(elec)
P5 ~1940
$30M
$390M
lo
M
broadcasting_ads
P5 ~1940
$30M
$390M
hi
M
lamps
P5 ~1940
$20M
$260M
hi
H
hydro
P5 ~1940
$20M
$260M
mid
L
patents_royalty
P5 ~1940
$10M
$130M
hi
M
coal
P5 ~1940
$10M
$130M
lo
M
transport
P5 ~1940
$-10M
$-130M
neg
M
finance/holding
P5 ~1940
$-750M
$-9750M
neg
H
S3↔S4 linkage
Linking to the graph
96 of 298 graph nodes are tagged with margin_tier (47 hi · 24 mid · 17 lo · 6 neg · 2 labor) — recorded in graph/graph-data-v3.json. The S3 hubs can now be colored by margin: GE/lamps/aluminum/patents — hi; copper/coal/utility throughput — lo/mid; the streetcar — neg. This confirms: centrality ≠ margin.
Honest disclosure. Profit pools are order-of-magnitude estimates (revenue×margin), nominal in the anchor year of each period plus a rough CPI→2020 adjustment; segments are not strictly additive (different bases). Low confidence (L) applies to real estate / aluminum in early periods / hydro. The financial −$750M (P5) is a one-off loss, outside the annual stack. The goal is the shape of margin migration, not precise sums. The row source is profit-pools-raw.json; the model is profit-pools-normalized.json.