Data & Methodology

Methodology

Difference-in-Differences Design

The analysis compares building permits inside the MIN overlay to permits outside MIN but within the same council districts (CD3 and CD7). This within-district design controls for neighborhood-level trends, political attention, demographics, and market cycles. The only systematic difference between the two groups is MIN exposure.

Because the pre-MIN period (3.54 years) and post-MIN period (3.45 years) differ slightly in length, annualized rates provide a fairer comparison than raw counts. Only permits with confirmed unit counts of 10 or more are included in the primary causal analysis — these are the projects MIN legally regulates.

Deterrent Estimation (units, matched control)

Housing supply is counted in units, and against a control matched on market strength — not just geography. MIN covers the stronger-market submarkets of CD3 and CD7, so the parcels left outside MIN within those districts are the weaker-market leftovers. Comparing the two (a naive within-district difference) books the whole strong-versus-weak gap as “MIN” and overstates the effect.

The headline restricts the non-MIN control to citywide parcels whose land value per square foot falls inside the inside-MIN band. Against that matched control, new construction of 10+ unit buildings was deterred by ~450 units (a DiD of +16 to +23 percentage points), stable across the 10-, 20-, and 50-unit cutoffs and across outlier trimming. Counting conversions — which MIN also regulates — raises the total to a ~1,000–1,400 unit range, our headline for all housing deterred; that conversion channel is directionally clear but spikier (dominated by a couple of mega-conversions), so we cite it as a range.

The older within-district permit×size method — 64 lost permits times average project size — runs to ~1,900–3,900 units. We retain it only for transparency, as an upper bound: its control is the weaker-market parcels left outside MIN, so it overstates.

MIN Spatial Classification

Permits are classified as inside or outside MIN using a spatial join against the Philadelphia City Planning Commission’s Zoning_Overlays shapefile (downloaded from OpenDataPhilly). Each permit’s geocoded location is tested for containment within the /MIN overlay polygon.

TOC Yield Analysis

The forward-looking TOC impact uses a hybrid methodology:

• FAR-bonus zones (CMX-3/4/5/RMX-3) in MIN: 30% FAR bonus blocked by law per §14-513(5)(a)(.2). Zero additional units.
• Density-bonus zones (RM-1/CMX-1/CMX-2/CMX-2.5) in MIN: 50% density bonus is legally available, but a behavioral deterrent is applied to projects with 10+ unit capacity. The rate used here (0.44) derives from the within-district comparison (expected 28.7/yr, actual 16.0/yr) and is therefore an upper bound: the matched-control deterrent is smaller (~450 units citywide). This component is under revision against the matched control, so the behavioral share of the TOC estimate below should be read as provisional.

Parcel-level capacity uses the BPN zoning rules engine with Philadelphia Zoning Code dimensional standards and Eriksen & Orlando (2022) construction cost thresholds for the cost cliff adjustment.

MIHB Revenue Comparison

Lost Housing Trust Fund revenue takes only the conservative new-construction deterrent (~450 units) — not the full conversion-inclusive total — applies the ~77% fee-in-lieu share seen in citywide MIHB behavior (~347 units), and multiplies by the MIN fee rate ($10,900/DU for density zones), giving ~$3.8M. This is conservative — at the MIHB Low Income rate ($17,700/DU) it is ~$6.1M. The Basic Systems Repair Program comparison uses an average grant of ~$15,000 per household (program maximum: $20,000).

Limitations

• Causal identification: The DiD design assumes parallel trends would have continued absent MIN. The 0% growth counterfactual is conservative but may understate losses if the inside-MIN areas would have grown faster than the control group.
• Unit count coverage: 39% of permits have confirmed unit counts. The primary analysis uses only permits with ≥10 confirmed units. Permits without unit counts are excluded from the causal analysis but included in the broader context.
• OPA match rate: 98.9% of permit addresses matched to OPA records. The 1.1% unmatched are excluded from parcel-level analyses (zoning classification, lot size, market value).
• Confounding with tax abatement changes: The 10-year tax abatement was reduced effective January 1, 2022, and a 1% construction tax was imposed — six months before MIN took effect. The DiD design controls for this because both treatment and control groups face the same abatement rules. However, the interaction between abatement reduction and MIN is not separately identified.
• Small control group: The outside-MIN control within CD3+CD7 is smaller than the treatment group (37 vs 125 large permits pre-MIN), which increases variance in the DiD estimate.