Step 1
Public source
Agency reports, public datasets, company filings, and academic studies — each with a URL and access date.
45 sources tracked
How we measure market power
How we measure concentration, which comparisons are fair, and where the data runs out. A high score is a reason to investigate, not a legal verdict.
Reusable public method packet
Copy packetEvidence route
Public records pass through an HHI gate.
Sources45Published8Withheld4
Public-interest rule: Concentration is a warning signal, not proof of illegal conduct.
Our measurement framework
A transparent, reproducible process from raw public records to concentration signals, source confidence, and data-gap decisions.
Explore the full methodologyPublic evidence standard
Trust comes from showing the route from public record to public action.
Concentrated markets can shift costs, wages, access, and bargaining power onto ordinary people. The remedy starts with public evidence that readers can inspect, reproduce, and challenge.
Source attached
12/12Every market profile should expose its public source trail before readers are asked to trust a metric.
HHI withheld when unsupported
4 marketsThe site names public data gaps instead of filling missing firm shares with invented precision.
Estimates are labeled
6 marketsEstimated-public rows keep their status visible so readers can separate measured facts from proxy signals.
Public routes stay open
45 sources + dataSources, data exports, methods, reports, and civic action routes are available from the same evidence chain.
Sources45 recordsPublisher, URL, access date, and citation text.Profiles12 marketsProduct, geography, participants, and public limitations.HHI Gate8 publish / 4 holdHHI only appears when public evidence can support it.Data/APIOpen packetsJSON, CSV, observations, participants, and source trails.Reports11 briefsResearch narratives link back to evidence and methods.ActionPublic workConvert evidence into comments, testimony, and oversight.
Reusable method packet
Copy the method, not just the conclusion.
A public-interest competition claim should travel with its source route, HHI gate, open data endpoints, and civic action path. This packet makes the methodology reproducible for analysts, journalists, and public officials.
45
Public sources
8
HHI published
4
HHI withheld
Notebook handoff
Copy the citation, API route, CSV route, JavaScript starter, or Python starter for the public methodology packet.
Method route
Public record to public action
Observatory tutorial
Use the observatory like a research desk
A guided route for turning a fair-competition question into a sourced market profile, reusable data packet, and public action memo.
01
Start with a public question
Search for a market, source, or briefing before choosing a metric.
Browse profiles02
Compare the signal
Use rankings and the comparison lens to find outliers and data gaps.
Open rankings03
Open the market profile
Read market definition, geography, HHI, CR4, top-firm share, and caveats together.
Try housing04
Inspect the data packet
Follow public JSON rows, source records, confidence labels, and notebook snippets.
Open data05
Translate evidence into action
Move from sourced facts to a briefing, disclosure ask, or public-interest memo.
Build actionPractice route
Housing competition data gap
Follow a real incomplete market: compare housing in data-gap rankings, open the local evidence profile, inspect source rows, then read the disclosure ask.
What good use looks like
- Quote the source and access date beside the metric.
- Keep withheld HHI visible when firm-level shares are missing.
- Treat rankings as a triage queue, not a legal finding.
Publication decision lab
Method gateSelect a market and see how source coverage, firm-level shares, residual categories, and confidence labels determine whether Antitrust Radar publishes, estimates, or withholds HHI.
12
markets
4
withheld
6
estimated
Decision queue
Review queue
Withhold HHIincompletehigh confidence
Domestic airlines
The public evidence does not support a reliable HHI publication for this market yet.
Source publisher
Bureau of Transportation Statistics
2026
Concentration class
classification pending
1 visible review flag
CR4 69.0% (>= 60%)
HHI
Withheld
CR4
69.0%
Top firm
17.8%
Residual
1
Publication checks
3/4
The public record can support market context, but not a headline HHI.
Audit focus: Find reproducible firm-level shares before publishing concentration.
Public data request
Turn aggregate residuals into named shares.
The public record has market-share evidence, but an aggregate residual prevents a reproducible headline HHI. The next ask is to split or document the residual category.
SourceBureau of Transportation Statistics
Gap1 aggregate residual row
AskSplit residual shares
Decision memo
Keep HHI withheld until public firm-level shares support a reproducible calculation.
Use source notes, methodology limits, and data-gap rankings to show what evidence is missing.
Publish market context without inventing missing concentration values.
Publication checklist
Public source is attached1 source in the profile
Concentration basis is reproducible10 firm-level shares
Aggregate residual is disclosed1 aggregate residual row
Confidence label is stronghigh confidence
Method notes
HHI is withheld because Other is an aggregated category rather than firm-level shares.
Relationship method
Pair measures without inventing a composite score
A relationship point exists only when both measures describe the same entity, market universe, and period basis. Missing sides stay missing; no value is carried forward or imputed.
Interpretation boundary
Pearson r summarizes linear association inside the visible matched rows. It does not establish causality, market definition, liability, or welfare effects.
01
Match
Join exact entity ID, year, domain, and calendar or fiscal basis.
02
Retain
Keep each measure's unit, source, status, confidence, and atomic row ID.
03
Summarize
Calculate visible means and Pearson r from matched rows only.
04
Interpret
Read the plane descriptively and return to profiles and source contracts.
Published contract: the API exposes the matched row IDs and both source fields, so every plotted point can be reconstructed from its two atomic observations.
Open relationship atlasState banking screen
Build statewide deposit HHI from a reproducible FDIC ownership rule
Each annual layer from 2015 through 2025 sums positive June 30 branch deposits within each state, groups branches to the FDIC-reported top holding company when available and otherwise to the insured institution certificate, then calculates shares and HHI from those grouped balances.
Screening boundary
Banking markets are often local, deposits are not every banking product, and a statewide balance screen is not a legal market definition.
01
Filter
Keep positive Summary of Deposits branch balances in the 50 states and District of Columbia for each annual vintage.
02
Group
Use the reported top holder ID; fall back to the insured institution certificate when no holder is reported.
03
Measure
Divide each grouped balance by statewide deposits, square the percentage shares, and sum them into HHI.
04
Publish
Retain annual 51-state ranks, state medians, national aggregates, source custody, and the local-market caveat.
- Annual vintages
- 11
- 2015-2025
- State-year rows
- 561
- 51 jurisdictions per year
- National rows
- 11
- Separate U.S. denominator
- Total observations
- 572
- No interpolated values
Individual insuranceKFF / 2024
3,414
median HHI / 36 highly concentrated jurisdictions
The product universe is ACA-regulated individual health insurance and the source publishes state HHI directly.
Branch depositsFDIC / 2025
903
median statewide HHI / 5 highly concentrated screens
The denominator is statewide reported branch deposits after the holder-or-certificate grouping rule above.
No composite: shared state geography permits paired inspection, not averaging, causal interpretation, or an overall state competition rank. The published relationship is Pearson r 0.247, descriptive only.
County banking screen
Recalculate the FDIC deposit denominator within each reported county or equivalent
The local layer applies the same holder-or-certificate ownership rule to positive June 30 branch balances in every annual vintage from 2015 through 2025, then groups and measures them within each FDIC-reported county FIPS.
Interpretation boundary
Branch booking can differ from customer residence or competitive reach. County boundaries can also differ from banking markets, so these rows are screens, not legal market definitions.
01
Locate
Retain the FDIC-reported five-digit county FIPS and county or equivalent label on every positive branch row in each annual vintage.
02
Group
Aggregate balances to the reported top holding company when available and otherwise to the insured institution certificate.
03
Measure
Calculate organization shares, HHI, leader share, organizations above 5%, branch count, and reported deposits within each FIPS.
04
Preserve
Keep every annual rank universe separate, reject duplicate year-FIPS rows, and leave changed or unavailable geographies uninterpolated.
- 2025 screens
- 3,104
- 51 jurisdictions
- Annual observations
- 34,193
- 11 exact vintages
- Median HHI
- 2,826
- Complete county screen universe
- Highly concentrated
- 1,792
- Current HHI at or above 2,500
Annual county rows are independent June 30 branch-deposit screens, not interpolated observations or legal market definitions. Reported ownership structures, branch booking, county names, and FIPS boundaries can change between vintages; deposit values are nominal and should not be read as inflation-adjusted growth.
Concentration uncertainty desk
HHI residual sensitivity lab
Separate observed firm shares from aggregate residual categories such as “Other” or “Unknown” before treating one arithmetic choice as the market.Evidence stateResidual range2026 exact release
Named share91%10 source-labeled participants
Residual share9%1 residual category
Sensitivity range1,285–1,366fragmented floor to consolidated ceiling
Selected scenario1,305Moderately Concentrated
HHI range / 0–10,000
Domestic airlines · 2026
Every residual allocation in the displayed 1,285–1,366 range stays moderately concentrated under this threshold reference.
Contribution ledger
Each share stays inspectable.
Delta Air LinesNamed source row316.8417.8% share
American AirlinesNamed source row306.2517.5% share
Southwest AirlinesNamed source row289.0017% share
United AirlinesNamed source row278.8916.7% share
Alaska AirlinesNamed source row39.696.3% share
JetBlue AirwaysNamed source row20.254.5% share
Other airlines4 equal firms per residual category20.259% share
Frontier AirlinesNamed source row13.693.7% share
Spirit AirlinesNamed source row10.893.3% share
SkyWest AirlinesNamed source row6.762.6% share
Hawaiian AirlinesNamed source row2.561.6% share
- Named contribution
- 1,284.82
- Residual contribution
- 20.25
- Scenario total
- 1,305
HHI threshold bands
This observatory classifies with the 2010 Horizontal Merger Guidelines bands (1,500 / 2,500). The 2023 Merger Guidelines lowered the highly-concentrated line to 1,800, so every market flagged here clears both standards.
HHI < 1,500
Unconcentrated
Markets are less likely to raise structural concern on concentration alone.
1,500 <= HHI < 2,500
Moderately concentrated
Markets may deserve closer scrutiny depending on evidence and harms.
HHI >= 2,500
Highly concentrated
Markets are likely concentrated enough to warrant serious policy attention.
HHI calculator
Enter market shares in percentage points to compute descriptive concentration metrics.
FirmShare (%)Squared
Firm A1,225
Firm B625
Firm C225
Firm D100
Firm E25
Other firms100
Total100.0%2,300
HHI
2,300
Moderately concentrated
CR4
85.0%
Top four firms
Effective firms
4.35
10,000 / HHI
01,5002,50010,000
HHI is the sum of squared market shares. A firm with 30% share contributes 30² = 900 points.
Source and data limitations
Coverage gaps
Not all firms, transactions, or geographies are public.
Reporting lags
Agency and market data are released on different schedules.
Definition variance
Markets and firm definitions can vary by source.
Proxy measures
Usage shares, passenger shares, and prescription shares may not equal revenue shares.
Credibility rules
We follow these rules before publishing concentration metrics.
Public evidence
At least one public source is required for every key input.
Market test
Product and geography definitions must be visible before metrics get persuasive.
Reproducible
Methods, source rows, and calculations are documented for replication.
Uncertainty disclosed
Assumptions, aggregate residuals, and missing shares stay visible.
Current enough
Vintage and access dates travel with the metric.
Human review
Analysts review outputs and flag concerns before publication.
When we withhold HHI
We do not publish or score HHI when any of these conditions apply.
Less than three firms have observable public shares.
A public table aggregates Other in a way that prevents firm-level HHI.
Market definition cannot be defended from public evidence.
Source vintage, geography, or proxy measure is too limited for a concentration claim.
Disclosure would risk exposing confidential company data.
How to cite
Antitrust Radar is public research infrastructure — cite it freely. The observatory’s compilation (curated rows, computed metrics, and API outputs) is licensed CC BY 4.0; upstream facts keep their publishers’ terms. Every market profile also carries a copyable field citation with its metrics, data vintage, and source basis.
The observatory
Antitrust Radar, A Public Interest Observatory: U.S. market concentration profiles, sources, and methodology. https://antitrustradar.org (accessed [date]).
A market profile
Antitrust Radar, “U.S. search engine usage” (market profile), HHI 7,359 (2026); source basis: StatCounter Global Stats. https://antitrustradar.org/markets/us-search-engine-usage (accessed [date]).
Methodology and resources
Explore how we measure markets and where the data comes from.
Data sources registryPublic datasets and agency records used by the observatory.HHI primerA public guide to reading concentration metrics.Data gaps reportWhy the observatory withholds HHI when public coverage is incomplete.Data-gap rankingsMarkets where public evidence is incomplete or confidence-limited.