San Francisco County presents a valuation-versus-income tension: in Zillow’s June 2026 county record, the $1,396,131 median home value and $4,401 monthly median asking rent produce a reported 3.78% gross yield before costs. This is a market for investigators who can substantiate property-level expenses and tenant durability, not for underwriting that assumes price appreciation resolves thin initial income. Buyers depending on low insurance, low turnover or rapid resale should be cautious.
Zillow’s value increased 9.48% year over year while its asking rent increased 20.25%, a favorable spread in the same Zillow vintage but not proof of net cash flow. The FHFA 2025 annual repeat-transaction HPI gained 0.50%; it is an index, not a home value, and its different timing and method must not be averaged with Zillow’s movement. The effective property-tax rate is 0.71%, a carrying cost against the gross-yield starting point. HUD’s two-bedroom FMR is a payment standard, not a market-rent estimate; only the published asking rent supports gross yield. Insurance, repairs, vacancy, financing and utilities are not published, preventing a net-yield conclusion.
Demand evidence is mixed. Annual QCEW 2025 covered jobs at county workplaces declined 1.48%; this is neither resident employment nor an unemployment measure. Professional and business services is the largest disclosed private supersector, not the whole economy. Tax-return migration is slightly negative, and arriving households had lower average AGI than departing households; this is a headwind signal, not a forecast or a tenant-demand measure. Investor participation was 366 of 3,783 purchases, or 9.67%, so non-occupant competition exists but does not define all purchasing. Realtor.com reports fewer active listings and shorter marketing time; its listing prices are asks, inventory is visible supply, and neither establishes closed-sale pricing nor buyer demand.
Earthquake is the dominant hazard. The modeled climate loss ratio of 0.35% of building value per year is a separate loss model and does not quantify earthquake severity, insurance availability or seismic damage. The record lacks seismic retrofit status, coverage terms and deductibles, condition, lease and occupancy data, closed-sale comparables, and operating expenses. Those omissions prevent a net-operating-income, cap-rate, replacement-risk or exit-liquidity conclusion.