Data-Driven Road Safety
Evidence turns individual crashes into patterns you can act on.
Synopsis
Road safety decisions are stronger when they are supported by reliable evidence. Data can help identify where crashes occur, who is affected, when risks are highest and which factors may be contributing. This lesson introduces the role of crash data, exposure data, trends, indicators and evidence-based decision-making in road safety. You will also understand why data must be interpreted carefully and why numbers alone do not explain every crash.
Why this matters
Raw crash counts can mislead. Understanding exposure, severity and data quality is what turns numbers into safer roads.
Expected outcome
You will understand how to source, question and interpret road safety data, and how to move from evidence to targeted action and evaluation.
Learning objectives
After completing this lesson learners should be able to:
- Explain what data-driven road safety means
- Identify common sources of road safety data
- Understand the difference between crash frequency and crash risk
- Explain why exposure matters when comparing road safety performance
- Identify useful road safety indicators
- Recognise limitations and biases in crash data
- Understand how data can support targeted interventions
- Explain why interventions should be evaluated using evidence
What is data-driven road safety?
Data-driven road safety means using reliable evidence to understand road safety problems and guide decisions. Instead of asking only where crashes happen, we can ask who is involved, when and where crashes occur, what type of crash occurs, how severe the outcomes are, what road or environmental conditions are present, whether certain road users are disproportionately affected, and whether the situation changed after an intervention. Data helps transform individual events into patterns that can be analysed. The objective is not to collect numbers for their own sake — it is to use evidence to make roads safer.
Sources of road safety data
Road safety information can come from police crash records, hospital and emergency medical records, road authority records, traffic-volume data, speed measurements, vehicle registration information, road condition assessments, road safety audits, traffic-camera information, surveys and observational studies, and public transport data. Each source provides a different perspective, and no single dataset necessarily provides the complete picture.
Crash data
Crash records may contain location, date, time, road-user type, vehicle type, crash type, injury severity, weather conditions, road conditions and contributing factors where recorded. When collected consistently, these records can reveal patterns. For example, a road may show a concentration of pedestrian crashes during evening hours, which may lead investigators to examine lighting, pedestrian movement, vehicle speeds, crossing facilities and traffic volume. The data does not automatically provide the solution — it helps identify where deeper investigation is needed.
Frequency vs severity
Two locations can have very different safety profiles. Location A may experience many minor crashes while Location B experiences fewer crashes but several fatal ones. Simply counting crashes would not tell the complete story. Road safety analysis should therefore consider crash frequency, injury severity, fatalities, serious injuries and crash type. A useful analysis asks not only how many crashes happened, but also how serious their consequences were.
Why exposure matters
Crash numbers need context. Suppose Road A has 100 crashes and Road B has 50 crashes — it may appear that Road A is less safe. But if Road A carries 500,000 vehicles per day and Road B carries 20,000, the amount of traffic using each road is very different. Exposure can refer to the amount of travel or interaction with the road system: vehicle kilometres travelled, number of pedestrians, number of cyclists, traffic volume, population or number of trips. Risk comparisons become more meaningful when crash outcomes are considered alongside relevant exposure.
Identifying patterns
Data can reveal patterns across time (hour, day, week, month, season), location (intersection, road section, school zone, residential area, highway), road user (pedestrian, cyclist, motorcyclist, car occupant, heavy-vehicle occupant) and crash type (rear-end, head-on, side-impact, pedestrian collision, single-vehicle crash). Finding a pattern does not prove a cause. It identifies something that deserves investigation.
Data quality matters
Poor-quality data can produce poor decisions. Potential problems include missing information, incorrect locations, inconsistent reporting, duplicate records, different definitions, under-reporting and changes in reporting practices. For example, a reduction in recorded crashes might appear positive, but if reporting became less complete the apparent improvement may not represent a real reduction. Good analysis therefore asks how reliable the dataset is before asking what it tells us.
Under-reporting
Not every crash is necessarily recorded in the same way. Minor crashes may never reach police records, and some injuries may be treated without being linked to a road crash database. Serious crashes are more likely to generate multiple records across different systems. This can create an incomplete picture. Combining information from different sources — police, healthcare, road authority and others — can provide a more complete understanding of road safety.
Data does not equal cause
A common analytical mistake is assuming that correlation automatically proves causation. Imagine that crashes increase on rainy days. That tells us there is an association, but further analysis is needed to understand why. Possible factors could include reduced visibility, reduced tyre grip, longer stopping distances, reduced traffic speeds, changes in pedestrian behaviour or different traffic volumes. Data identifies patterns; investigation is required to understand causes.
From data to action
A useful data-driven process is: collect reliable information; clean it by checking for missing, inconsistent or duplicate records; analyse it to identify patterns and trends; investigate to understand the underlying factors; intervene with an appropriate safety measure; and evaluate whether the intervention improved outcomes. This creates a continuous improvement cycle: data, analysis, action, evaluation.
Measuring road safety
Road safety performance can be monitored using indicators such as road deaths, serious injuries, pedestrian fatalities, motorcycle fatalities, crash rates, vehicle speeds, seatbelt usage, helmet usage, drink-driving prevalence and emergency response times. Some indicators measure outcomes; others measure behaviours or system performance. Using multiple indicators provides a more complete picture than relying on a single number.
Evaluating interventions
Suppose a dangerous intersection receives a new safety treatment and the work is completed. Does that automatically mean the problem is solved? No — the intervention should be evaluated. Possible evidence might include changes in crash frequency, crash severity, vehicle speeds, pedestrian behaviour, traffic conflicts and user observations. Evaluation helps determine whether the intervention achieved its intended safety outcome.
Real-world scenarios
Crash counts
Two roads are compared. Road A has 80 crashes per year and Road B has 40, but Road A also carries ten times as much traffic.
→ What should the safety team do before concluding that Road A is more dangerous?
Show suggested response
A — Compare crash outcomes with relevant exposure and other safety indicators. Raw crash counts do not provide sufficient context; exposure and other indicators make comparisons more meaningful.
Apparent improvement
A city reports that recorded crashes have fallen by 20% after changing its reporting process.
→ What should analysts consider before declaring a safety improvement?
Show suggested response
B — Check whether the change in reporting practices could have affected the recorded numbers. Changes in data collection or reporting can affect apparent trends, so analysts should understand the quality and consistency of the underlying data.
Rain and crashes
Crash data shows that more crashes occur during heavy rain.
→ What is the most appropriate conclusion?
Show suggested response
C — The pattern should be investigated to identify the factors associated with increased crash risk. The data identifies an association, but it does not by itself establish the complete cause.
Safety intervention
A road authority installs a new pedestrian crossing at a location with repeated pedestrian crashes.
→ What should happen next?
Show suggested response
D — Evaluate whether the intervention produced the intended safety improvement. Evidence can show whether crashes, conflicts, speeds or other relevant indicators changed after implementation.
Key takeaways
- Data helps identify road safety problems and patterns
- Crash counts alone do not provide the complete picture
- Exposure is important when comparing risk
- Crash frequency and crash severity should both be considered
- Data quality and consistency are essential
- Under-reporting can create an incomplete picture of road safety
- A pattern in data does not automatically prove causation
- Multiple data sources can provide complementary information
- Data should lead to investigation and targeted action
- Safety interventions should be evaluated after implementation
- Evidence-based decisions make interventions more targeted and effective
Complete this lesson
Take the short quiz to mark this lesson complete and unlock the next.
Lesson 87 of 89 available · 12 min · India-specific