Mapping Traffic Accidents in Metro Manila
Part 2: A Closer Look
Some weeks ago, I made a post on my introductory project to data science: a Python program to clean and geocode a dataset containing traffic accidents in Metro Manila during 2015. In this post, I dive deeper into the geocoded dataset and reference other articles on traffic statistics in Metro Manila.
A Quick Background
At the time of my previous post, my program geocoded ~15,000 records from the 2015 dataset of ~96,000. I’ve been running my program since, and the current total is ~46,000 geocoded records.
In this post, I visualize and analyze the statistics of the geocoded dataset. I used Carto to generate all geographic visualizations, and Google Sheets to generate all statistical visualizations. For more information on the 2015 dataset, my methodologies, and sources of potential error, please refer to the previous post.
Visualization
First, a visualization of the geocoded dataset.
The dataset contains traffic accidents in the Metro Manila area, and the resulting maps reflect that. Many of the traffic accidents seem to happen along major highways and roads; numerous geocoded points form visible traces of the roads. These main roads are the arteries and veins of Metro Manila, accommodating hundreds of thousands of vehicles every day.
For example, in Southern Metro Manila, many of the reported accidents were concentrated around the main highways, such as Alabang-Zapote Road (lower horizontal road), Sucat Road (upper horizontal road), or Service Road (right side, vertical road).
However, the geocoded traffic accidents aren’t exclusive to the main highways and avenues. In many areas in Northern Metro Manila, the geocoded points aren’t concentrated to lines, and appear instead as patches.
For example, business districts Makati (left half, center) and The Fort (right half, center) are scattered with accidents outside main highways like EDSA. The lesser roads that permeate these business, commercial, and residential districts can often be narrow and crowded.
Carto has a feature that visualizes accident density by regions or cities. Based on the visualization, Quezon City contains the most accidents, followed by Manila and Makati. Unfortunately, I didn’t find any visualizations for further granularity (at the barangay level).
Analysis
Now for some simple analysis on the non location-related fields.
Most reported accidents happened during the morning to early evening (7AM-8PM). Schools and businesses typically operate from the morning to late afternoon. However, office hours can extend into the evening to compensate for arriving late to work due to traffic in morning commutes. In addition, certain industries can require employees to work night or graveyard shifts.
The majority of reported accidents happened during weekdays and Saturday. Sunday, however, sits well below the average of the other days of the week. According to Philippine labor policies, normal working hours are 8 hours a day, 6 days a week.
Each accident record in the original dataset was classified into one of three types: Damage to Property, Non Fatal Injury, and Fatal. Damage to Property consisted the vast majority of reported accident types (81.0%), followed by Non Fatal Injuries (18.5%), and finally Fatal (0.5%).
Cars dominate the vehicle type most involved in all accidents by a large margin, followed distantly by motorcycles, trucks, and vans.
There’s several factors that could explain why cars are most involved in accidents. Some notable points include…
- A steady growth in automobile sales (particularly passenger cars), meaning more cars are being sold and used. This is fueled by…
- More affordable cars and more affordable payment plans that lower the barrier of entry for many Filipinos.
- Outsourcing jobs in the Philippines, which makes car ownership a possibility for a rising middle class.
Males consist of the majority involved in traffic accidents, at 67.2%. Females follow at 27.9%. Finally, there were people whose genders were unknown, at 4.9%.
Most people involved in the reported accidents were between 16 and 40.
There is a sharp increase in accident involvement from people aged 16–20, to 21–25. Between 16–20, people are likely graduating high school and entering college. Between 21–25, people are likely finishing college and entering the workforce.
Working age millennials comprise the three age groups most involved in accidents (21–25, 26–30, and 31–35). In 2015, Millennials (aged 15–34) consisted 47.1% of the working-age Philippine population. Indeed, millennials consist of 1/3rd of the Philippine’s population. They constitute a great fraction of the general population and the working population, and it’s reflected in their involvement in traffic accidents.
Conclusion
I performed some visualizations and analysis on the geocoded dataset. Some interesting observations include…
- Quezon City was the city that contained the most accidents. Manila and Makati follow. There wasn’t any feature to identify granularity by barangay, however.
- The majority of accidents were recorded in the morning to evening (7AM — 8PM). Most businesses and schools operate around these hours, but oftentimes people do have to work later hours to compensate for morning commutes. Finally, there are industries that require employees to work night or graveyard shifts.
- The majority of accidents were recorded on the weekdays and Saturday. Less accidents were reported on a Sunday compared to the other days. This is likely a combination of labor laws and Sunday being a day of rest for many Roman Catholic Filipinos.
- Cars were the vehicle type most involved in all recorded accidents by a huge margin. A combination of a growing percentage of middle class Filipinos, and cheaper and more affordable car plans could point to why cars are the vehicle type most involved in accidents.
- A considerable percentage of people involved in traffic accidents were working-age millennials (between 20 and 35). Millennials comprise a third of the Philippine population, and a majority of the working age Filipino population. It makes sense that many traffic accidents involved at least one millennial.
I don’t expect that geocoding the remaining dataset will change the observed trends very much. The next time I’ll post on this topic will be when I’ve processed at least two (ideally three) years’ worth of traffic accidents. I think observing the point map and statistics as they change over a number of years would reveal additional insights.
Thank you for reading! Please feel free to contact me on LinkedIn.
