One concept · Applied economics lesson
Measuring Local Market Concentration
How do economists measure whether a market is concentrated—and how much can the answer change when the researcher cleans names or redraws the market?
Why this matters
Market shares and the Herfindahl–Hirschman Index (HHI) are common starting points for describing competition. The arithmetic is simple. The consequential work often comes first: deciding which observations belong together, which names identify the same economic organization, and which geography is the market.
Research
Yelowitz and coauthors measured brokerage concentration across 90 local markets and then examined six metropolitan areas in greater detail.
Teaching
This lesson makes the data work visible. Generated spelling variants fragment one firm until you clean them; offices can then be analyzed separately or grouped by franchise.
Policy
HHI thresholds appeared in the 1982 U.S. merger guidelines, remained in the 1992/1997 guidelines, rose in 2010, and returned to earlier levels in 2023. The same synthetic market can receive a different screening label across vintages.
Interpretation boundary: An HHI or structural screen is evidence about market structure, not a legal conclusion. A ZIP comparison is a sensitivity exercise, not an automatic antitrust market definition.
Measure one market
Hold the generated rows fixed and change one analytical choice at a time. Start with Chicago ZIP 60614, a label appearing in the source paper's FSBO appendix. HHI sums squared percentage shares; CR4 is the combined share of the four largest identities. For a hypothetical combination of shares a and b, the HHI rises by 2ab.
Synthetic teaching exercise: The listings, assignments, and shares below are generated. Historical place and brokerage labels do not describe actual market shares or mergers.
Loading the synthetic listings…
Selected market
- Synthetic listings
- —
- Identities
- —
- HHI
- —
- CR4
- —
—
Hypothetical top-two combination
- Identity A
- — ()
- Identity B
- — ()
- ΔHHI = 2ab
- —
- Post-combination HHI
- —
- Vintage HHI screen
- —
This mechanical illustration combines the two largest identities under the selected rule. Separate raw strings or offices can share a franchise brand; this is not an actual transaction or legal conclusion. Exact share ties use alphabetical name order; changing tied names does not change HHI. Official guideline source.
Same rows, three identity rules
Cleaning duplicate spellings reduces artificial fragmentation. Grouping offices under a franchise brand is a separate, contestable economic-unit assumption: independent franchisees may compete. The source study treats separate offices and combined franchises as useful lower and upper concentration measures.
HHI approaches 0 as many firms become tiny and reaches 10,000 with one firm. The bars zoom to 0–1,500 because every scenario in this synthetic release falls between 145.1 and 1,219.0; the table gives exact values.
| Identity rule | Identities | HHI |
|---|
Look inside the calculation
| Rank | Synthetic identity | Listings | Share | HHI contribution |
|---|---|---|---|---|
| All identities | — | 100.0% | — | |
See what cleaning does
These generated variants are deliberately messy. Treating them as separate firms inflates the identity count and biases this exercise's measured HHI downward.
| Generated raw string | Cleaned office | Combined firm |
|---|
Test a synthetic submarket definition
Each generated ZIP is assigned one price tier. This table compares ZIP-based groups within the four-ZIP pool, not price specialization within a ZIP. The generated tier pattern does not reproduce the published paper's higher-price concentration finding.
| Tier | N | Firms | HHI | Largest identity | Share |
|---|
What changed?
Try this: keep Chicago selected and compare raw strings, cleaned offices, and combined firms. Then switch between its four-ZIP pool and individual ZIPs. Explain whether the number moved because the underlying rows changed, the economic unit changed, or the geographic market changed.
The basic idea
Calculate each firm's percentage share, square it, and add the squared shares.
HHI = Σ sᵢ²
Four equal firms each have a 25 percent share, so HHI = 25² + 25² + 25² + 25² = 2,500. A monopoly has HHI = 100² = 10,000. Squaring gives more weight to large shares.
Four equal pieces still produce a concentrated-market arithmetic example.
Squaring makes the largest shares matter most.
The synthetic data
The lab contains 1,620 deterministic synthetic listings: 270 in each of six four-ZIP pools labeled for metropolitan areas studied in the 2013 Cityscape paper. These pools are not full MSAs. One ZIP label (60614) appears in the published paper's appendix; the other 23 were checked against paths in the archived scrape. Period brokerage names come from the source materials. Every row-level assignment and value is generated. Version hhi-synthetic-v1 uses seed 20130411.
Synthetic teaching exercise. No address, agent, listing URL, sale, or property record was copied. The generated shares do not state any organization's historical or current market share.
Yelowitz evidence
The lesson simplifies the workflow. The published studies supply the historical questions, samples, and findings.
Concentration and Market Structure in Local Real Estate Markets
Jason Beck, Frank Scott, and Aaron Yelowitz, Real Estate Economics 40(3), 2012. Across 90 local markets, the authors find brokerage concentration generally low in medium and large markets; smaller markets show more concentration. These historical listing-based estimates depend on the market and firm definitions used.
The Market for Real Estate Brokerage Services in Low- and High-Income Neighborhoods: A Six-City Study
Aaron Yelowitz, Frank Scott, and Jason Beck, Cityscape 15(1), 2013. Across six MSAs and more than 300,000 listings, full-MSA HHIs range from 36–341 with offices separate and 302–678 when franchises are grouped. ZIPs with higher house prices tend to be more concentrated; less advantaged neighborhoods do not show worse access by this structural measure. These are historical associations, not causal effects.