nyc-ride-market

L10 · Policy studies

Wave 3 · Phase 3 (9 – 29 Nov) · Spec §6 (congestion effect, pay-rise event study), §12 · Decisions D7, D9, D12 · Feeds L11, L15

Goal

The two policy analyses: a year-over-year difference-in-differences for congestion pricing (Q2, Q3) and a weekly event study around the 1 Aug 2025 pay rise (Q6).

Owns

sql/metrics/cp_effect.sql, sql/metrics/pay_event.sql, pipeline/studies.py
config/periods.yaml: the excluded-holiday-weeks key only
reports/congestion_pricing.md, reports/pay_rise.md
tests/test_cp_effect.py, tests/test_pay_event.py

Starts when

L09’s runner exists. Build on factories first; real numbers after G3

Tasks

Holiday weeks: compute in code which US federal holidays fall in different ISO weeks in 2024 and 2025; write them to periods.yaml with each holiday’s name
Congestion effect: per 2025 ISO week (from W02 if D12 is agreed) × mode × metric (trips_all, fare_base / miles, driver_pay / fare_base): growth of treated ÷ growth of control − 1. Report the median and the p10–p90 range across weeks, for both zone lists (D9)
Pay rise: 12 ISO weeks before and after 2025-08-01; HVFHV driver_pay / miles, driver_pay / minute and fare_base / miles; the same weeks of 2024 as the seasonal reference; W31 as a transition week (D12)
pipeline/studies.py writes the report tables. The narrative is written by hand: “estimated effect”, never “caused”; limits from SPEC §12
Sanity sections in each report: the weekly series per group, and which weeks were excluded and why

Tests

A fixture with known weekly sums → effect, median and p10/p90 worked out by hand
Excluded weeks are dropped; week-boundary handling

Done when

Both metrics run on the back-filled aggregates; both reports merged
python -m pipeline.studies reproduces every number in the reports

Not this lane

Site pages (L15); finding sentences for the site (L11)

Status log

Only this lane writes here. Newest first.

Date	State	Note
	not started

Blocked: none

Hand-off

Built:
Checked:
Known limits:
Follow-ups: