Access is not only a park boundary.
Recreational access shapes opportunity for physical activity, but a public-park-only map can miss resident-only facilities that many households actually use. This project measures how much that omission changes the estimated geography of access inequality in one town before drawing any causal or health conclusion.
This measures geographic opportunity, not physical activity or health effects.
What changes when private provision counts?
How does including resident-only HOA recreation facilities change estimates of recreational-access inequality in Cary, North Carolina, and does private provision compensate for limited public access or compound existing advantage?
Four signals, one routed network.
Entitlement is constructed from four signals — parcel geometry, name-text matching, plat records, and nesting — then kept three-valued when the evidence cannot support a binary answer. The walking network combines OpenStreetMap streets with Cary's official greenways and uses network distance, not a straight line.
- Entitlement
- Geometry, name text, plat records, and association nesting are reconciled before an access state is assigned.
- Typology
- Dual access, public-reliant, private-reliant, and dual deficit follow Liu et al. (2026); they are not presented as an original classification.
- Uncertainty
- When entitlement remains undetermined, each share is reported as a band rather than collapsed to a single estimate.
- Walking speed
- 1.31 m/s, the pooled usual outdoor speed from Murtagh et al. (2021); thresholds are 10, 15, and 20 minutes.
Derived access point is a candidate greenway door where the sourced network meets an official greenway — not an automatic claim of public entry. Candidates are recorded as unreviewed, confirmed, ambiguous, or rejected. The five featured routes terminate at registered public entrances; derived candidates remain in the pipeline sensitivity analysis until destination evidence is reviewed.
Upchurch Farms is the worked association example: a geometric-proximity resolution, followed by a PROPDESC name-matching correction, with its amenity evidence retained at community scale.
Because an HOA is a community rather than an individual, this page names associations directly — no output below that level (for example, individual households) is shown.
Definition choice changes the deficit.
At the primary 15-minute threshold, the same classification behaves differently depending on whether the recreation universe includes trails alongside parks. Values are housing-unit-weighted bands and remain preliminary pending methods review.
| Access class | Parks + trails | Parks only |
|---|---|---|
| Dual access | 41.5–50.1% | 16.3–19.8% |
| Public-reliant | 32.3–40.9% | 15.5–19.0% |
| Private-reliant | 7.5–9.1% | 32.7–39.4% |
| Dual deficit | 8.5–10.1% | 25.3–32.0% |
Changing what counts as recreation moves the dual-deficit estimate by roughly 3.5× — larger than any other single modeling choice made so far.
Failures are part of the method.
D-50 caught a silent source defect before it reached a result: a raw 100,453 TOTUNITS count included storage units and hotel rooms, while the defensible housing count was 76,411 — a 33% inflation corrected with a whitelist.
Automated unittest cases plus three independent script checks run against the current pipeline.
Every fallback path is logged and reviewed, and each run is reproducible from the fixed inputs and documented decisions.
What this page does not claim.
This is not a causal or health-effects study. The distinction between bundled and opt-in HOA fees remains unresolved because no source data exists yet. Demographic cross-referencing (Stage 8) is intentionally held until the methods above have been reviewed by someone with relevant expertise.
References
- Liu et al. (2026). Four-class recreational-access typology: dual access, public-reliant, private-reliant, and dual deficit. Classification framework used for comparability.
- Murtagh, E. M., et al. (2021). Outdoor Walking Speeds of Apparently Healthy Adults: A Systematic Review and Meta-analysis. Sports Medicine, 51(1), 125–141. Pooled usual outdoor walking speed: 1.31 m/s (95% CI 1.27–1.35).