Ownify Research · Boulder, Colorado · September 2026
Do real-estate agents price their own homes differently? Boulder, 2021–2026
A Levitt–Syverson replication on 7,066 recorded Boulder sales, with 76 homes sold by the listing agent who owned them and 73 agent-owned homes listed by a colleague. Companion to the San Francisco study (September 2026). By Frank Rohde, .
Key finding: Agents who sell their own Boulder home get about 3.3% more relative to market value than the civilians they advise — the same premium, to the decimal, that the San Francisco study found. What differs is how they get it. In San Francisco agents under-price and let an auction run; in Boulder they ask more, hold the number, and wait.
+3.3 pts
Self-listed premium on sale ÷ market value, controlled (p = 0.002, n = 76)
103.1% / 102.2%
Median asking ÷ market value, self-listed vs civilian
38% / 21%
Sold exactly at asking, self-listed vs civilian
–4.5 pts
What agents pay as buyers, relative to value (p = 0.065, n = 92)
1. The question
Levitt and Syverson (2008) showed that Chicago agents selling their own homes got about 3.7% more and took ten days longer than the clients they advised, and argued the gap measures the agent's incentive to close fast rather than to close high. We asked the same question of Boulder. It matters for the Ownify pricing report because the answer tells a seller what the expert does when the expert's own money is at stake — which is a better guide than what the expert recommends.
2. Identification
The seller of each recorded sale is the grantee on the prior deed (the previous buyer). We matched those names — first and last, trusts and entities excluded — against the roster of 2,066 listing agents who appear in the Boulder listing data itself. Three tiers: any match (149 sales, 2.1% of civilian-seller sales), strict (surname rare in the deed population and unique in the roster; 81 sales, 43 self-listed), and self-listed — the agent's own name on the deed and on the listing that produced the sale (76 sales). Agent-owned homes listed by a colleague are the agent-other group (73). Entity sellers (LLCs, trusts, 1,000 sales) are excluded from all comparisons.
| Group | Sales | Houses | Median sq ft | Median sale | Median market value | Cash buyer |
|---|---|---|---|---|---|---|
| Civilian sellers | 6,917 | 67% | 1,756 | $918,900 | $936,300 | 20% |
| Self-listed agents | 76 | 68% | 1,651 | $980,000 | $918,000 | 20% |
| Agent-owned, listed by another agent | 73 | 67% | 1,908 | $1,003,000 | $1,012,600 | 12% |
Unlike San Francisco, where agents owned more houses than civilians, Boulder's three groups have the same house share and similar size and vintage; composition does not drive the result. Agent-sellers are spread evenly across market areas (1.3–3.3% of sales in each).
3. Results
Every regression is OLS on the log ratio with HC1 standard errors and controls for property type, size, market-value level, assessor sub-market and month of sale. "Market value" is Ownify's fair-value estimate for the home; "asking" is the verified original ask recovered from listing history (3,111 of the 7,066 sales).
| Outcome | Civilian median | Self-listed median | Self-listed effect, controlled | Agent-other effect, controlled |
|---|---|---|---|---|
| Sale ÷ market value | 100.7% | 101.8% | +3.3 pts (p = 0.002) | +2.2 pts (p = 0.18) |
| Sale ÷ hedonic value (value-model-free) | 102.1% | 103.9% | +2.3 pts (p = 0.09) | +3.6 pts (p = 0.13) |
| Asking ÷ market value (the anchor) | 102.2% | 103.1% | +3.0 pts (p = 0.11) | +3.9 pts (p = 0.015) |
| Sale ÷ asking | 98.9% | 100.0% | -0.4 pts (p = 0.62) | +0.9 pts (p = 0.33) |
| Sold exactly at asking | 21% | 38% | agent-other 28% | |
| Sold over asking | 26% | 18% | agent-other 24% | |
| Days, listing to sale | 41 | 46.5 | +12% (p = 0.26) | -8% (p = 0.48) |
| Listings over 60 days | 26% | 35% | agent-other 21% | |
| Failed to sell in the listing | 17.2% | agent-owned 12.9% (12 failures; indicative) |
Robustness. The self-listed premium is +4.3 pts in 2021–22 (p = 0.04) and +2.7 pts in 2023–26 (p = 0.03); +2.4 pts on the strict-name tier (p = 0.06); +4.9 pts on houses (p < 0.001) and zero on condos (+0.1, n = 24), which is where you would expect skill and patience to matter. The raw-median gap is small (+1.1 pts) and its bootstrap interval spans zero (–1.9 to +3.5); the effect lives in the controlled means, i.e. agents sell better for what they own and where. Agents as buyers pay 4.5 pts less than value (p = 0.065, n = 92) — the mirror image.
4. Two markets, one premium, opposite playbooks
The San Francisco and Boulder self-listed premiums are the same size (+3.3 pts, p = 0.002 in both), which is itself the finding: the agent's information advantage is worth about three percent of the price wherever the agent is. But the mechanism is inverted, and it tracks each market's pass-through of asking price into sale price — 0.44 in San Francisco, 0.93 in Boulder.
| Lever | San Francisco (auction market) | Boulder (negotiation market) |
|---|---|---|
| Listing anchor, self-listed vs civilian | 87.3% vs 91.2% of value — agents ask less | 103.1% vs 102.2% — agents ask more (agent-other 107.4%, p = 0.015) |
| Sale vs asking | +14.8% over ask, 75% clear over — agents run the auction harder | 100.0% of ask, 38% close exactly at ask — agents hold the number |
| Time on market | 15.5 vs 15 days — no difference | 46.5 vs 41 days; 35% vs 26% past 60 days — agents wait |
| Agent-owned, listed by a colleague | +40% days (p = 0.03) — patience shows | Asks highest (107.4% of value), sells fastest (41 d, 21% past 60 d) |
| Failure rate | No difference (10.9 / 11.8 / 12.1%) | Agent-owned lower (12.9% vs 17.2%; 12 failures) |
| Measure | Civilian | Self-listed | Agent-other |
|---|---|---|---|
| Anchor (asking ÷ value, median) | 102.2% | 103.1% | 107.4% |
| Sale ÷ value, median | 100.7% | 101.8% | 103.8% |
| Sale ÷ value, mean | 98.3% | 101.6% | 101.2% |
Read together with the Boulder pricing model, the agents are doing what the model says a Boulder seller should do. Because 93% of the ask passes through to the sale, under-pricing is not recovered by bidding — only 26% of civilian sales close over ask, and the discount is given away for nothing. The agents anchor a point above the civilian convention, refuse the two-percent give-back that civilians accept (their median sale is exactly the ask), and pay for it in time: a week longer at the median, a third of their listings past 60 days. The probability-of-sale curve is flat inside ±10% of value, so the extra ask costs them almost nothing in odds — and their failure rate is, if anything, lower.
5. What it means for a Boulder seller
The civilian Boulder convention — ask 4–5% above value, accept 2% under ask after six weeks — is close to what agents do with their own homes, but not identical. The three-point gap is the difference between accepting the first reasonable offer and holding for the ask. The Ownify pricing report therefore recommends, for houses in particular: anchor at or slightly above the sub-market convention, budget eight weeks rather than six, and treat an under-ask offer in the first month as the opening of a negotiation, not its end. For condos the study finds no agent premium; the convention is the right price.
6. Method notes
Data: 8,417 Boulder recorded residential sales, January 2021 – September 2026, restricted to 7,066 clean, non-entity-seller sales with a market value; 3,111 carry a verified original asking price recovered from listing history. Market value = Ownify's fair-value estimate (automated valuation × 0.971 trailing-four-quarter sale-to-valuation factor × property adjustment). Hedonic value = leak-free ridge on characteristics, sub-market and month FE, five-fold cross-fitted (R² 0.83, residual sd 0.29). Seller identification: prior-deed grantee names parsed LAST FIRST, matched exactly on first and last name to parsed listing-agent names (2,066 of 2,131 roster names parsed); entity patterns (LLC, trust, inc., etc.) excluded. Self-listed requires the listing agent of the listing dated within 400 days before the sale to match a seller name. Failure analysis uses the current owner's post-sale expired or withdrawn listings (2023-07 onward) as failures and verified-ask sales weighted to all sales as successes. All effects are OLS coefficients on log ratios × 100 with HC1 standard errors; controls: condo indicator, log square feet, log market value, 46 assessor sub-market fixed effects, 68 month fixed effects. Bootstrap: 2,000 resamples of the group medians. Buyer statistics are aggregated only; no individual is identified in this report or its underlying files.
Reference: Levitt, S. and Syverson, C. (2008), "Market Distortions When Agents Are Better Informed," Review of Economics and Statistics 90(4).
Frequently asked questions
- Do real-estate agents get more when they sell their own home in Boulder?
- Yes. Across 7,066 Boulder sales from 2021 to 2026, agents who listed and sold their own home got about 3.3 percentage points more relative to market value than civilian sellers, after controlling for property type, size, value, sub-market and month of sale (p = 0.002, n = 76).
- How do Boulder agents price their own homes?
- They ask slightly more (a median 103.1% of market value versus 102.2% for civilians), hold the number (38% sell exactly at asking versus 21% of civilian sales), and wait longer (46.5 days at the median versus 41, with 35% of listings past 60 days versus 26%).
- Is the agent premium the same for condos?
- No. The premium is 4.9 points on houses (p < 0.001) and essentially zero on condos (+0.1 points, n = 24). For condos, the market convention is the right price.
- How does Boulder compare with San Francisco?
- The premium is the same size in both markets, +3.3 points. In San Francisco agents under-price and let bidding run; in Boulder, where 93% of the asking price passes through to the sale price, agents ask more and hold the number.
- What should a Boulder home seller do with this?
- For houses, anchor at or slightly above the sub-market convention, budget eight weeks rather than six, and treat an under-ask offer in the first month as the opening of a negotiation, not its end.
Related reading
- Owen Boulder pricing intelligence
- Boulder neighborhood guides
- How Ownify prices homes
- Buying a home in Boulder with 2% down
Ownify Research · Frank Rohde · [email protected] · Not an appraisal.