Data & Methodology

All analysis is based on publicly available City of Philadelphia data. The underlying datasets are available for download below so anyone can verify our findings.

Download the Data

What's in the workbook:

Data Sources

Methodology

Study Area

The study area is the intersection of (a) the CTR Old City Residential Area overlay polygon (Phila. Code § 14-502, the union of five sub-areas in the city's Zoning_Overlays dataset) and (b) the OCD BID polygon. This means a property must fall inside both definitions to be included — "where the zoning code AND the BID definition agree it's Old City." All business licenses and appeals are matched via spatial join (point-in-polygon) using GIS coordinates, not street-name text matching.

Time Window

January 2007 – February 2026 (19 years). Pre-2007 business licenses are excluded because the appeals dataset does not extend back that far — including them would silently classify older licenses as "by-right" for lack of matchable appeal data.

Appeal Scope

Zoning Board of Adjustment (ZBA) appeals only. The L&I Review Board (LIRB) and Board of Building Standards (BBS) handle code-enforcement and building-code questions, not zoning variances, and are excluded from this analysis.

F&B Classification

Business licenses are classified as F&B based on their revenue code (matching standard food, beverage, and restaurant codes). Appeals are classified by a single canonical regex classifier on appeal grounds text into one primary use type: Residential, Sit-down restaurant, Take-out restaurant, Prepared food shop, Other F&B, Smoking lounge, Body art / tattoo, Assembly / entertainment, Hotel / visitor accommodation, Retail / commercial, Roof deck, Signage, Parking / lot, or Other. Procedural appeals against L&I permit decisions are excluded.

Variance Feature Flags

A single appeal can touch multiple variance and special-exception elements (e.g., a residential project with a roof deck and parking variance). Each appeal is also tagged with boolean flags — has_roof_deck, has_height_variance, has_use_variance, has_parking, has_sign_element, has_lot_variance — so we can count cross-cutting elements (e.g., "appeals touching a roof deck") without double-counting appeals.

Approval Rate

Approval rate = approved ÷ (approved + denied). Withdrawn and dismissed cases are reported separately and not in the denominator. This rule is applied uniformly across every chart, table, and figure on this site.

License Deduplication and Opening Events

Philadelphia issues separate licenses for "Food Establishment, Retail" (revenue 3118) and "Food Preparing and Serving" (3120) — many businesses have both, often issued the same day. License renewals are also issued as separate records. We deduplicate to one row per unique (address × business name), keeping the earliest license date as the "opening event."

Two-Path Timeline

Each opening event is classified as either ZBA path or by-right, and a timeline is computed:

• ZBA path (n=30): the closest approved F&B-use appeal at the same address within 3 years before opening → business license issuance date. The 3-year window matches the realistic ZBA-path timeline (file appeal → hearing → permit → build → open).
• By-right path (n=97): most recent fit-out permit issuance date → business license issuance date. 2-year lookback window (non-binding — by-right opens are typically much faster).

New-building construction permits are excluded from the by-right start-of-clock to avoid counting building-construction time as restaurant opening delay. Medians are reported to minimize the impact of outliers.

Strict ZBA-Path Classification

A business is on the ZBA path only if its opening can be directly attributed to a recent (≤3 years) approved F&B-use appeal at the same address. Addresses where the only F&B-use appeals were denied, withdrawn, dismissed, or filed too long ago to plausibly cause the current business are classified as by-right. This is a tighter rule than the older approach (which counted any address with any historical F&B-use appeal as ZBA), but it more honestly reflects which openings actually required the ZBA process.

RCO Voting Data

76 voting records from Old City District covering RCO meetings 2012–present. Each record includes the project address, ZBA application/calendar/permit number, meeting date, board and public vote tallies, and final Support/Oppose outcome. 74 of 76 records were matched to a ZBA appeal in the open-data appeals file via permit number, calendar number, or address-and-date proximity matching.

The published download reflects three OCD-confirmed corrections (April 2026 review): RCO has no "No opinion" outcome — the single stray record was reclassified to Support; 222 Market St is shown as OCD-Opposed → ZBA-Approved (the smoking-lounge case); and "105 Market Street" is the legacy address typo for 104 Market Street. Final tally: 74 Support, 2 Opposed.

Gave-Up Analysis

"Gave-up candidates" are F&B use appeals with no matching subsequent F&B business license at the same address. Appeals filed after March 2025 are excluded as still in process. Range addresses are expanded for matching.

Known Limitations

• Sample sizes for some use types are small (e.g., smoking lounges: n=3, body art: n=3).
• The "gave up" analysis cannot distinguish between applicants who abandoned projects due to zoning costs vs. other business reasons.
• Some addresses may have had use approvals under prior zoning frameworks that carry forward.
• Appeals without GIS coordinates are matched via text-based street classification (minimal impact; <5% of records).

Questions about the data or methodology? Contact jon@buildphillynow.org