""" Vision Zero Report Card analysis. Pull truck-involved fatal-crash deaths per city, 2018-2024, from FARS national CSV downloads (NHTSA). Compute year-over-year change and aggregate per Vision Zero city. Truck definition matches federal FMCSA/IIHS standard: BODY_TYP 60-69 = medium/heavy trucks (>10,000 lbs GVWR) 60: Step van 61: Single-unit straight HAZMAT 62-65: Single-unit straight truck (varying weight classes) 66: Medium/heavy pickup 67: Unknown medium/heavy 68: Truck-tractor (semi/combination) 69: Unknown single vs combo Output: derived/report_card.json with per-city series + summary stats. """ import csv import json from pathlib import Path from collections import defaultdict ROOT = Path(__file__).resolve().parent DERIVED = ROOT / 'derived' DERIVED.mkdir(exist_ok=True) # (state_fips, city_fars_code, display_name, vision_zero_year) # Vision Zero adoption years from official city commitments CITIES = [ # state_code, city_fars_code, display, slug, vz_year, population_2020 (36, 4170, 'New York City', 'new-york-ny', 2014, 8_804_190), ( 6, 1980, 'Los Angeles', 'los-angeles-ca', 2015, 3_898_747), (17, 1670, 'Chicago', 'chicago-il', 2012, 2_746_388), (48, 3280, 'Houston', 'houston-tx', None, 2_304_580), ( 4, 370, 'Phoenix', 'phoenix-az', 2017, 1_608_139), (42, 6540, 'Philadelphia', 'philadelphia-pa',2017, 1_603_797), (48, 6090, 'San Antonio', 'san-antonio-tx', 2015, 1_434_625), ( 6, 3260, 'San Diego', 'san-diego-ca', 2015, 1_386_932), (48, 1730, 'Dallas', 'dallas-tx', 2019, 1_304_379), (25, 120, 'Boston', 'boston-ma', 2015, 675_647), (53, 1960, 'Seattle', 'seattle-wa', 2015, 737_015), ( 8, 600, 'Denver', 'denver-co', 2017, 715_522), (11, 10, 'Washington DC', 'washington-dc', 2015, 689_545), (41, 1650, 'Portland', 'portland-or', 2015, 652_503), (48, 330, 'Austin', 'austin-tx', 2016, 961_855), (12, 2010, 'Miami', 'miami-fl', 2017, 442_241), (12, 1510, 'Jacksonville', 'jacksonville-fl',None, 949_611), (47, 1760, 'Nashville', 'nashville-tn', 2017, 689_447), (13, 280, 'Atlanta', 'atlanta-ga', 2020, 498_715), ] # Body type codes for medium/heavy trucks (excludes passenger pickups <10K lbs) TRUCK_BODY_TYPES = set(range(60, 70)) YEARS = [2018, 2019, 2020, 2021, 2022, 2023, 2024] PATHS = { 2018: 'FARS2018', 2019: 'FARS2019', 2020: 'FARS2020/FARS2020NationalCSV', 2021: 'FARS2021/FARS2021NationalCSV', 2022: 'FARS2022/FARS2022NationalCSV', 2023: 'FARS2023/FARS2023NationalCSV', 2024: 'FARS2024/FARS2024NationalCSV', } def _reader(path): """Open CSV with latin-1 (covers all FARS years) and strip BOM from first header. When latin-1 decodes UTF-8 BOM bytes they become '', not ''.""" f = open(path, encoding='latin-1') reader = csv.DictReader(f) # Trigger fieldnames materialization, then mutate in place. if reader.fieldnames: first = reader.fieldnames[0] if first.startswith(''): reader.fieldnames[0] = first[3:] elif first.startswith(''): reader.fieldnames[0] = first[1:] return f, reader def load_year(year): base = ROOT / PATHS[year] accident_path = base / 'accident.csv' vehicle_path = base / 'vehicle.csv' truck_cases = set() f, reader = _reader(vehicle_path) try: for row in reader: try: bt = int(row.get('BODY_TYP', 0) or 0) except ValueError: bt = 0 if bt in TRUCK_BODY_TYPES: key = (int(row['STATE']), int(row['ST_CASE'])) truck_cases.add(key) finally: f.close() by_city = defaultdict(int) by_state_total = defaultdict(int) f, reader = _reader(accident_path) try: for row in reader: state = int(row['STATE']) st_case = int(row['ST_CASE']) fatals = int(row['FATALS']) try: city = int(row.get('CITY', 0) or 0) except ValueError: city = 0 if (state, st_case) in truck_cases: by_city[(state, city)] += fatals by_state_total[state] += fatals finally: f.close() return by_city, by_state_total def main(): print('Loading FARS years...') series = {(s, c): {} for s, c, *_ in CITIES} state_totals = {} for yr in YEARS: print(f' {yr}...') by_city, by_state = load_year(yr) for (state, city, *_rest) in CITIES: series[(state, city)][yr] = by_city.get((state, city), 0) state_totals[yr] = dict(by_state) # Build output report = [] for state, city, name, slug, vz_year, pop in CITIES: s = series[(state, city)] first = s.get(YEARS[0], 0) last = s.get(YEARS[-1], 0) total = sum(s.values()) baseline_avg = (s[2018] + s[2019]) / 2 if s.get(2018) is not None and s.get(2019) is not None else None recent_avg = (s[2023] + s[2024]) / 2 if s.get(2023) is not None and s.get(2024) is not None else None pct_change = None if baseline_avg and baseline_avg > 0: pct_change = ((recent_avg - baseline_avg) / baseline_avg) * 100 per_100k_baseline = (baseline_avg / pop) * 100_000 if baseline_avg and pop else None per_100k_recent = (recent_avg / pop) * 100_000 if recent_avg and pop else None report.append({ 'name': name, 'slug': slug, 'state_fips': state, 'city_fars_code': city, 'population_2020': pop, 'vision_zero_year': vz_year, 'series': {str(y): s.get(y, 0) for y in YEARS}, 'total_truck_fatalities_7yr': total, 'baseline_avg_2018_2019': baseline_avg, 'recent_avg_2023_2024': recent_avg, 'pct_change_2yr_vs_2yr': round(pct_change, 1) if pct_change is not None else None, 'fatalities_per_100k_baseline': round(per_100k_baseline, 2) if per_100k_baseline else None, 'fatalities_per_100k_recent': round(per_100k_recent, 2) if per_100k_recent else None, }) # Rank by pct_change (winners ascending) ranked = sorted(report, key=lambda x: x['pct_change_2yr_vs_2yr'] if x['pct_change_2yr_vs_2yr'] is not None else 0) out = { 'methodology': { 'source': 'NHTSA FARS National CSV downloads, 2018-2024', 'truck_definition': 'Medium/heavy trucks: BODY_TYP codes 60-69 (excludes passenger pickups <10,000 lbs GVWR)', 'metric': 'Truck-involved fatalities (FATALS column on crashes with at least one truck-class vehicle)', 'baseline_period': '2018-2019 average', 'recent_period': '2023-2024 average', 'change_metric': 'Percent change baseline average vs recent average', 'per_100k_note': 'Per 100,000 residents using 2020 Decennial Census population', 'fars_release_note': 'FARS 2024 Annual Report File (preliminary), released April 2026', }, 'cities': ranked, 'national_totals_all_trucks': { 'note': 'All-states truck-involved fatalities per year (not city-filtered)', 'series': {str(y): sum(state_totals[y].values()) for y in YEARS}, }, } out_path = DERIVED / 'report_card.json' with open(out_path, 'w') as f: json.dump(out, f, indent=2) print(f'\nWrote {out_path}') # Console summary print('\n=== Vision Zero Report Card 2026 ===') print('Truck-involved fatalities, 2018-2019 avg vs 2023-2024 avg\n') print(f'{"City":<20} {"VZ":>4} {"2018":>5} {"2019":>5} {"2020":>5} {"2021":>5} {"2022":>5} {"2023":>5} {"2024":>5} {"Change":>8}') print('-' * 110) for r in ranked: vz = r['vision_zero_year'] or '-' s = r['series'] ch = r['pct_change_2yr_vs_2yr'] ch_str = f'{ch:+.1f}%' if ch is not None else '-' print(f'{r["name"]:<20} {str(vz):>4} ' f'{s["2018"]:>5} {s["2019"]:>5} {s["2020"]:>5} {s["2021"]:>5} ' f'{s["2022"]:>5} {s["2023"]:>5} {s["2024"]:>5} {ch_str:>8}') if __name__ == '__main__': main()