Tracking Housing, Rents and Household Moves in Pittsburgh and Allegheny County

Housing decisions often depend on surveys that describe conditions years earlier. We propose updating those baselines as new permits, sales and other records arrive, much as navigation engineers use fast motion measurements to estimate a car’s position between satellite fixes. The system combines each source according to its accuracy and age to estimate neighborhood housing supply, rents and occupancy. As projections extend beyond the survey baseline, their uncertainty ranges widen. When a new American Community Survey release arrives, the system compares averages over the same years, accounts for the survey’s own sampling error, and adjusts future projections for neighborhood types with similar errors. Residents can also contribute rents, which appear only in privacy-protected aggregates. Instead of simply displaying records with different dates, planners get an updated housing map and a companion reliability map: where fast records support dependable estimates, and where gaps in the city’s data leave conditions unclear.

Asking rents tell us what landlords advertise; tenants’ actual payments are harder to verify and dangerous to expose. We build a rent map around the verification process used by iNaturalist’s volunteer naturalists. Tenant reports first pass basic checks against neighborhood rents and county property records. Next, we compare independent reports from the same building with a plausible rental-income range derived from its assessed value, borrowing a property-appraisal calculation. Redacted leases provide the final confirmation. Each stage permits a finer map view, from neighborhoods to buildings, but stronger proof alone never unlocks building detail. Publication also requires at least five independent contributors and checks that no contributor dominates the aggregate. We withhold related figures when someone could otherwise subtract them from totals to recover a hidden value. The result is a map of verified actual-rent aggregates, with building detail only where both evidence and privacy rules permit it.

A rent estimate with one error bar cannot tell a planner whether to survey more tenants, repair mismatched records or obtain newer data. Earthquake analysts separate natural variation in shaking from uncertainty about their models, carrying several plausible models through the calculation. We apply the same principle to housing: keep real differences between rents separate from survey sampling error, uncertain record matches, stale data and figures withheld to protect tenants. When two records might describe the same building, we carry the plausible matches through the calculation and show how much the answer changes. A neighborhood map uses distinct visual marks to show which causes dominate and connects each to an action: expand surveys, repair record links, refresh sources or publish privacy-protected figures for a larger area. Real rent differences remain visible as a condition to plan around. Every displayed number must also carry its source, dates and uncertainty breakdown.