Gamma Blog: Perilfinder - Accident Risk Models

Spatial Risk Models: Analyzing the risk of accident on Irish roads

Drivers, Roads and Road Accidents

In 2011, there were 1,648,409 households in Ireland. These households between them owned 2,219,211 cars, an average of 1.35 cars per household. In Ireland in 2011, 2,704,404 people commuted to their place of work, study or occupation. 1,246,387 (41.6%) did so as the driver of a car or van. (Source: Census 2011).

In 2012, 9,183 cars were involved in 5,603 road accidents leading to minor, serious or fatal injury. That is to say, 0.004 accidents per car. (Source: Road Safety Authority).

This article focuses on two questions. The first is how this risk of road accident is distributed across the country, and the second is how – and where – the new information available from Census 2016 informs and updates our judgement of risk.

Cars and drivers are not evenly spread across the country. Meath has the largest number of cars per household at 1.57 on average, and almost half of Meath commuters travel as the driver of a car or van. On the other end of the scale, only 36% of Dublin commuters drive, and the average Dublin household has only 1.16 cars.

Dublin and Louth have low rates of car ownership and the smallest proportion of commuters who drive. Driving as a commute mode is most popular in Leitrim and Roscommon, while car ownership is highest in Meath and Kildare.

Like drivers, road accidents are not uniformly distributed across the country. Obviously, there is an effect of exposure; we expect to see more accidents where there are more cars. For this reason, we correct for the number of cars owned in each county.

Let’s take a closer look at the area around the border, where we seem to have the highest rate of road accidents. Longford, Louth, Donegal and Cavan have particularly high rates.

What might be contributing to this risk? One possibility is road sinuosity, defined as the ratio of a road segment’s curved length to the straight-line distance between its endpoints. Roads with high sinuosity values are twisty, bendy and potentially dangerous. The Road Safety Authority has made available a dataset of 5,520 1-km segments of the national road network with their sinuosity values.

Ireland’s most sinuous road segment can be found in Rathborney in County Clare.

Gamma Blog: Perilfinder - Accident Risk Models

That is indeed a pretty twisty road. Its speed limit is 100km/h, which one hopes not many drivers actually reach.

In fact, we don’t have to conjecture about the relationship between sinuosity and risk. The Road Safety Authority has also made available a dataset of 1km road segments with the collision rate, corrected for the exposure (millions of kilometres of travel) and road type (urban two-lane roads are the most dangerous with 18.94 accidents per 100 million kilometres of travel; motorways are the safest with only 2.41 collisions per 100 million kilometres of travel).

What is surprising here is that there is no relationship between sinuosity and collision rate! If anything, the most sinuous road segments seem to be safer than expected. The correlation between sinuosity and collision rate is only 0.12 – too small to have any significance. Perhaps very sinuous roads demand care and attention from the driver and so are actually safer than broad, straight roads that encourage recklessness and speeding?

The distribution of sinuosity is so similar for the four classes of collision rate that we can hardly distinguish the lines.

If the structure of the roads is not responsible for the rate at which road accidents are happening, perhaps the culture around cars and driving is a factor. The beauty of the census of population is that it allows us to dive deeply into the spatial element of the data. We can examine the relationship between car ownership and popularity of driving as a commute mode not only at county level but at ED level or indeed small area level.

Looking forward, what changes in driving behaviour can we expect to see described in the 2016 census results? Well, so far we only know how car ownership has changed at a county level, and it’s encouraging to see that although population and car ownership have both increased across the country, the average number of cars owned per household has actually decreased in every county. The smallest change was in County Wicklow where the average number of cars per household went from 1.47 in 2011 to 1.46 in 2016. The biggest change was in County Carlow, where the average number of cars per household went from 1.41 to 2011 to 1.36 in 2016.

It remains to be seen whether the popularity of driving as a commute mode will decline in proportion to car ownership, and to what extent the reduced car usage may translate to reduced road accident risk.

@ 2017 by Charlotte Cuffe


About Perilfinder

Perilfinder™ is Gamma’s flagship SaaS product for the insurance sector and is used by several of the world’s leading insurance companies. Perilfinder™ is a complete risk assessment and reporting solution for underwriters, incorporating address matching, rooftop-level geocoding, risk and accumulation scoring, and spatial catastrophe models. Technically advanced yet user-friendly, Perilfinder™ helps its users control their risk portfolios by better managing policy accumulations and through identifying geographical perils in real time. Perilfinder™ is a world-class, no-fuss, intuitive and feature-rich location intelligence system.

Book a demo at!