FARS Report

1. About

In this report we will analyze the analyzing Fatality Analysis Reporting System from National Highway Traffic Safety Administration, to determine if New York's vision zero, that were partly initiated by Amy Cohen and Lizi Rahman, had reduced casualties.

The data used is from 2005 to 2023.

2. Vision zero actions

Lowering the speed limit from 30 to 25
The installation of over 3000 speed cameras that lead to 70% decrease in speeding.
Nearly 1000 intersections are safer to cross because of:
- Curb extension
- Leading pedestrians to intervals
- Narrow vehicle lanes
- Pedestrian Plazas
- Pedestrian safety island
- Raised crosswalk
- Wider sidewalk

3. Variables

3.1. Existing ones

PEDS: People involved in a crash who were not inside a motor vehicle.
- PERNOTMVIT: People involved in a crash who were not inside a motor vehicle, plus, those who were in a motor vehicle but not In-Transport/moving.
PERSONS: Number of people inside a motor vehicle
- PERMVIT: Number of people inside a moving motor vehicle
FATALS: Number of Fatalities
VE_TOTAL: Number of vehicles in crash
- VE_FORMS: Number of moving vehicles in the crash
- PVH_INVL: Number of parked/moving vehicles in the crash

3.2. New variables

ID: Merge the ST_CASE with the year of the accident.
PERNOTMV: The number of people in the accident who were in a vehicle, but they weren't moving; derived from the subtraction of PERSONS and PERMVIT
FT_PED: Fatality count among pedestrians (Amy Cohen case)
FT_CYC: Fatality count among cyclists (Lizi Rahman case)
Pop: The population count at a giving year.
This variables was using on a EDA for both USA overview and NY state overview; Data was gathered from https://www.census.gov/data/tables.html
Date time related
- DATE_ACCIDENT: The accident date in full.
- DATETIME_ACCIDENT: The accident date time in full.
- DATETIME_MEDC_ARR: The arrival of medics on scene date time in full.
- DATETIME_HOSP_ARR: The arrival to the hospital date time in full.
- Delta_acc_medc: The time difference between DATETIME_ACCIDENT and DATETIME_MEDC_ARR
- Delta_acc_hosp: The time difference between DATETIME_ACCIDENT and DATETIME_HOSP_ARR
- Delta_medc_hosp: The time difference between DATETIME_MEDC_ARR and DATETIME_HOSP_ARR
- year_month: Using the dates by only defaulting every date to the first date of the month using DATE_ACCIDENT; example: 2005-03-19 → 2005-03-01

4. EDA

4.1. Pedestrian Fatality trend In the USA

Yearly Pedestrians Fatalities In The United States

The fatalities among pedestrians is going upstarting 2014, that this mean that the New York state effort failed? Let's decompose the trend line into states.
Yearly Pedestrians Fatalities In The Top States With Highest Yearly Casualties

Despite the increase in fatality rates in other states, New York initiatives decreases it.

Yearly Bicyclists Fatalities In The Top States With Highest Yearly Casualties

New York plans did stabilizers/reduce the fatality rate among bicyclists, unlike in other states.

Bonus: Yearly Pedestrians Fatalities vs Fatalities per Population In The United States

I divided the fatality count of a giving year by the population count of the same year.
Starting from 2015, we can notice a huge divergence, meaning that the increase in fatalities isn't only attributed to an increase of populations.

4.2. Pedestrian Fatalities in New York

Density Map Of Pedestrian Casualties In New York Year By Year
(You can zoom, play/select a year, and hover for more details)

4.3. More explorations

Density Map Of Pedestrian Casualties In New York By Hour And Month For Both Pre- & Post-Vision Zero
- There is a weird tilted pyramid pattern, where starting from:
  - 17H (5PM), Months [1,11,12] have most fatalities.
  - 18H (6PM), Months [1,2,10,11,12] have most fatalities.
  - 19H (7PM), Months [1,2,3,9,10,11,12] have most fatalities.
  - …
- Fatalities Post Vision Zero For Month 1, 11, And 12 At 5pm
Density Map Of Pedestrian Casualties In New York By Hour And Age For Both Pre- & Post-Vision Zero
- We can notice a shit in count that start at 17H
- The lack of fatalities of pedestrians aged 10 to 50 can be attributed to them being at school or workplace
- Pre-vision zero, there was a spike of casualties among your people in the early morning
- Post-vision zero, we can see a spike in fatalities after 17H, especially among the elderly.
Box plot describing the distribution of the time it took the medic to arrive to the scene of the accident in all recording weather conditions
How did the medic responsiveness change pre- and post-project zero? One of the variables that might hinder the medics from arriving in time is the weather.
The variable Delta_acc_medc_int represent the time it took the medic to arrive in minutes; the x-axis is in log-scale.

Medic arrival time improved in smoky, sleety, and snowy weather; but not so in cloudy and rainy weather.