• fr
School project · IMT Atlantique · 2025

Road accidents in France

Power BI analysis of road traffic injury accidents in France, using open data and population data.

Power BI scatter plot of French departments: total number of accidents on the x-axis, accident rate on the y-axis, coloured by region. Paris and the other Île-de-France departments stand out for the number of accidents, French Guiana for the highest rate; Corsica, Martinique and Guadeloupe also have high rates.
Number of accidents and accident rate by department (charts in French)

Download the full presentation (PDF, in French)

At a glance

Context

  • School project at IMT Atlantique, in 2025, in a team of 3.
  • A project run by three people: each of us explored the data and built our own analyses, which we then pooled and consolidated into a final presentation.
  • Open data on road traffic injury accidents (data.gouv.fr), combined with population data.

Actions

  • Data preparation and merging with Power Query.
  • Indicators built in DAX.
  • Analysis along four lines: geography, men / women, accident-prone conditions, and an insurance case.

Results

  • A reading of accidents that separates volume, risk relative to population, and severity.
  • A minimum-risk profile identified for the insurance case.

Tools

  • Power BI: visualisations and dashboards.
  • Power Query: data preparation and merging.
  • DAX: indicators.

Context

The goal was to understand where, when and under what conditions road traffic injury accidents happen in France, using open data from data.gouv.fr combined with population data.

Part of the project takes the point of view of an insurance company: identifying the population least exposed to serious accidents, in order to target an offer.

Approach

Data preparation

Accident and population data were prepared and merged with Power Query.

Indicators

Several indicators were built in DAX:

  • accident rate per 100,000 inhabitants;
  • severity rate;
  • uninjured rate;
  • calculated age groups.

Four lines of analysis

  1. Geographical distribution: regions and departments.
  2. Men / women comparison: involvement, severity and safety equipment.
  3. Accident-prone conditions: weather, lighting, road type, time of day and seasons, separating the number of accidents from their severity.
  4. Insurance case: identifying a minimum-risk profile with a decomposition tree.

Results

These results describe correlations observed in the data, not cause-and-effect relationships.

Geographical distribution

In absolute numbers, Île-de-France has the most accidents. Relative to population, French Guiana has the highest rate; other overseas territories and Corsica also stand out.

Men and women

Power BI bar chart of the accident rate per 100,000 inhabitants by age group and sex: the rate for men is higher than for women in every age group, peaking at 20 to 24 for both sexes.
Accident rate by age and sex, per 100,000 inhabitants

Men are more involved at every age, with more serious accidents on average. Women wear safety equipment more often, which may partly explain this gap.

Accident-prone conditions

Power BI table of the number of accidents by day of the week and two-hour time slot, coloured from blue to red: the highest values are on weekdays between 4 pm and 8 pm, then between 8 am and 10 am.
Number of accidents by day and time slot
Power BI charts of the severity rate by road profile, speed limit, surface condition, built-up area and road layout: severity is higher on icy roads, outside built-up areas and on bends, especially S-bends, than on straight roads; a pie chart shows the distribution of accidents by traffic regime.
Severity rate by road characteristics

Accidents are most frequent in towns, on weekdays and at rush hour, but most serious outside built-up areas, at night without lighting, on bends and on icy roads.

Wearing a seat belt is associated with the highest uninjured rate, whatever the speed and the road.

Insurance case

Power BI decomposition tree of the uninjured rate, split first by built-up area, then by age group, then by sex: the rate is higher in built-up areas and reaches 0.57 for people aged 45 to 49 and 55 to 59, for women and men alike.
Decomposition tree of the uninjured rate

The profile with the highest uninjured rate is people aged 45 to 59 driving in built-up areas, men and women alike, with an uninjured rate of 0.57.

What I learned

  • Relating figures to population changes the picture: the area with the most accidents is not the one where the risk is highest.
  • Volume and severity tell two different stories, and they need to be analysed separately.
  • Correlation is not causation: some results hide biases. For example, airbags seem less protective because they mostly deploy in the most violent crashes.
  • An analysis is most valuable when it answers a concrete business question, such as targeting an insurance offer.