Real Florida transaction evidence · v2

Tampa for development.
Orlando for the test.

Offer-to-Exit reconstructs named iBuyer ownership spells from county deeds, fits only on Hillsborough County, and measures what transports to Orange County without refitting on Orlando outcomes.

Evidence at a glance

Valuation errors give each parcel equal metric weight. Hazard fit and evaluation use the ordinary person-period likelihood. Kaplan-Meier summaries use ownership episodes.

2,657,072

normalized deed rows

Two official county systems

5,060

linked iBuyer ownership spells

2,391 common-window descriptive episodes

$129,039

Orlando repeat-sale MAE

12.8% lower than the rolled-forward prior-sale baseline

0.649

Orlando disposition-hazard AUC

Tampa-only specification; no Orlando refit

Geographic transport

The repeat-sale valuation exercise uses the prior eligible sale price relative to its own county-quarter market median, elapsed time, and sale quarter. The prediction is scaled by the lagged current-quarter median. The duration model uses acquisition-time information and operator indicators.

Orlando recorded prices versus Tampa-model predictions — Tampa-only repeat-sale model applied to the Orlando common-window test without refitting.

Kaplan-Meier inventory exit curves for Tampa and Orlando — Recorded acquisition-to-resale duration, which is not MLS days on market.

Orange qualification sensitivityThe subset with an observed county-qualified description contains 4,124 test transactions. Its model MAE is $144,459 versus $162,374 for the rolled-prior benchmark. The subset spans 2025-01-02 through 2026-08-19, so it changes both qualification observability and calendar period rather than isolating proxy error.

Repeat-sale seasoning sensitivityThe main design already excludes prior deeds from the target sale's quarter. Requiring an additional 30-day gap leaves 10,506 Tampa and 6,021 Orlando test transactions. Model MAE remains 2.9% below the rolled-prior benchmark in Tampa and 12.8% below it in Orlando.

The Opendoor slice

Opendoor remains visible as the motivating operator, while Offerpad enlarges the common-market empirical support for the shared inventory problem. Operator comparisons are descriptive.

Orlando

935

Opendoor acquisition spells

KM exit by 120 days: 50.1%

Tampa

1,190

Opendoor acquisition spells

KM exit by 120 days: 43.9%

Named intermediary episodes

The comparable duration sample requires an improved single-family home, operator support in both counties, and an acquisition in January 2022 or later. Deed-to-deed duration is not MLS days on market. Open spells are censored at each county's observation end: Tampa 2026-08-21 and Orlando 2026-08-31.

Market	Operator	Acquisitions	Observed exits	KM exit by 120d	Median observed hold
Orlando	Offerpad	92	91	71.4%	90 days
Orlando	Opendoor	935	881	50.1%	118 days
Tampa	Offerpad	174	171	58.6%	104 days
Tampa	Opendoor	1,190	1,132	43.9%	134 days

What transported

An external-market test is a hard test of a Tampa-only specification, not proof of national generalizability.

$147,902

Orlando baseline MAE

Prior sale rolled forward with the county market

80.5%

Orlando 90% interval coverage

Coverage under geographic shift

-0.373

Tampa future Brier skill

Against constant training hazard

-0.253

Orlando Brier skill

Same Tampa-only model and time window

Identification boundaryCounty deeds support prediction of recorded prices and descriptive ownership duration. They do not reveal rejected offers, weekly list-price paths, repairs, operating costs, days on market, or causal price response. The separate randomized policy experiment identifies response only inside its known data-generating process. No Florida seller or buyer elasticity is claimed here.