Road safety as a scientific discipline is based on the experience of many scientific fields, such as urban planning, mathematics, and road traffic management. All this makes it an interdisciplinary field. This paper is built on the logic of the theory-empirics-model. (Jamroz, 2008) points out that there are four types of theories in road safety (stochastic, causal, systemic and behavioural). This article uses the stochastic theory, namely _random events theory_, with Bortkiewicz, who discovered that the distribution of killed in road accidents in a year is almost perfectly random and the Poisson model gives a good description of the random process (Bortkiewicz, 1898; Jamroz, 2008) as the authors investigate the accidents as random events. Bortkiewicz investigated that accidents are purely unexpected and that we do not have control over them (Elvik et al., 2009). By causal theories, it is claimed that human factors are major research problem, estimating that 90% of road accidents are connected with human factors (alcohol, training, age) (Jamroz, 2008). However, these factors were not explored in this research.

Most os the previous models predict the number of road accidents taking into account the severity of accidents, casualty count and the number of vehicles involved (Gatarić et al., 2023; Pourroostaei Ardakani et al., 2023). Previously, logistic regression and other data mining techniques and machine learning has been used as the statistical method for prediction (Ait-Mlouk et al., 2017; Mohanta et al., 2022). However, the authors of this research focus on the model predicting the number of fatal accidents on the road across the countries based on Poisson distribution, the topic which has not been studied in detail.

The literature underlines RTS as a multifaceted concept influenced by temporal, seasonal, and demographic factors. Vision Zero by the EU and UN aims to eliminate road fatalities by 2050. Factor analysis and Poisson regression are commonly applied to model and predict accident trends and severities.

It has been demonstrated () that the accident frequency with fatalities—rather than the number of fatalities alone—is a more accurate indicator of road traffic safety (RTS), as high-fatality events (e.g., bus crashes) can distort the overall picture (; ). Relying solely on a single indicator such as the fatality rate may not capture the full complexity of RTS, especially when such incidents can disproportionately influence the statistics. A more comprehensive assessment of RTS can be achieved by considering multiple indicators simultaneously—including accident rates involving injuries, fatalities, and accidents with fatalities—in a unified model.

In this study, we begin by analyzing Estonian traffic data using heatmaps and Poisson regression to determine whether RTS has improved over time. The variables included in the analysis are weekday, month, and year. Poisson regression is a widely used method for examining trends and risk factors in road traffic accidents (; ; ). Although the fatality rate (i.e., the number of traffic fatalities per year) has traditionally been the most commonly used indicator of RTS, it often fails to provide a complete and stable representation of the situation. This is due to its sensitivity to rare but severe events, such as a single accident resulting in multiple fatalities.

There is an ongoing debate regarding how best to evaluate road traffic safety (RTS). This complexity arises from the multifaceted nature of RTS, which is influenced by various factors such as seasonality (e.g., summer versus winter), day of the week (e.g., workday versus weekend), and broader socioeconomic dynamics. Additionally, RTS assessments are complicated by the distinction between short-term and long-term effects, population changes (e.g., migration), and per capita calculations, all of which can introduce bias into safety metrics. For example, recent immigration to Estonia from Ukraine has increased the country’s population by approximately 50,000, affecting per capita indicators.

This paper provides a contextual background for RTS, followed by a review of the relevant literature and theoretical framework.

Road traffic accidents impose a substantial economic burden, costing countries approximately 3% of their gross domestic product (). Commonly used indicators to assess RTS include the total number of accidents, accident severity (injuries and fatalities), and temporal variables such as time of day, day of week, or season (; ; ; ).

RTS has emerged as a dedicated field of scientific inquiry (; ; ; ). Understanding the role of different indicators is essential for the accurate interpretation of safety trends (; ).

The implications of RTS are not only methodological but also economic and societal. In the European Union, over 20,000 people die each year due to road traffic accidents (). The EU Road Safety Policy Strategy sets a target of halving road fatalities by 2030 and achieving near-zero fatalities by 2050—an initiative known as Vision Zero (; ; ).

According to the World Health Organization (), the world is facing a road traffic safety crisis, with 1.3 million deaths annually due to traffic accidents. The global road fatality rate ranges from 3 to 40 per 100,000 population (; ). In the EU, nearly 19,000 people died in traffic accidents in 2020, equating to 42 fatalities per million inhabitants—a decrease compared to 2019, partly reflecting reduced road use during the COVID-19 pandemic ().

Innovations in road construction technologies and sustainable urban planning initiatives are helping to improve RTS (; ). In 2021, the European Parliament launched a new Road Safety Report supporting the EU’s long-term goal of eliminating traffic deaths by 2050 (; ).