Regional Analyses

1.1 Background

The problem of deaths and injury as a result of road crashes¹ is now acknowledged to be a global phenomenon with authorities in virtually all countries of the world concerned about the growth in the number of people killed and seriously injured on their roads. In recent years there have been two major studies of causes of death worldwide that have been published in the ‘Global Burden of Disease’ (1996, World Health Organisation, World Bank and Harvard University) and in the ‘World Health Report –Making a Difference’ (WHO 1999).

These publications show that in 1990 road crashes as a cause of death or disability were by no means insignificant, lying in ninth place out of a total of over 100 separately identified causes. However, by the year 2020 forecasts suggest that as a cause of death, road crashes will move up to sixth place and in terms of years of life lost (YLL) and ‘disability-adjusted life years’ (DALYs)2 will be in second and third place respectively.

This growing awareness is reflected in the recent establishment of the Global Road Safety Partnership¹ (GRSP). Under the framework of the World Bank’s Business Partners for Development (BPD) Programme, GRSP is a partnership of private sector, civil society and government organisations collaborating to improve the road safety situation around the world. A Steering Committee for GRSP is now in place with the aim of creating a global information network that aims to produce solid evidence of the positive impact of partnerships – both the development impact and the business benefits. Two important aspects of GRSP are the involvement of the private sector in promoting road safety and the promotion of greater awareness of road safety worldwide.

With the setting up of the GRSP it was considered important that a comprehensive summary of the global situation was made available to all involved in the problem of road crashes in developing and transitional nations. For example, using published statistics from countries throughout the world, TRL has in the past attempted to identify the number of people killed in road crashes worldwide and also on a regional basis. The last attempt to do this however used 1990 data and is clearly out of date (Ghee et al 1996). Further, since that time, more information has been collected on under-reporting and a better estimate can be obtained of the global situation. Using different statistics, the World Health Organisation (WHO) investigated deaths from many causes in the two above studies. Differences exist in these early estimates provided by TRL and WHO, and the GRSP Steering Committee proposed that a review should be undertaken to resolve these differences and identify the current situation as accurately as possible. It was agreed that the World Bank, the Department for International Development UK (DFID) and TRL would co-fund a study that would be restricted to the analysis of published material.

The study was carried out by TRL with the following objectives:

• To derive an estimate of road crash fatalities worldwide and on a regional basis for the year 1999 and to derive forecasts of the likely number of deaths in the year 2010and 2020.
• To provide an estimate of crash costs worldwide in relation to GNP.
• To obtain regional analyses of fatality trends, rates and risk (deaths per 10,000 vehicles and per 100,000population respectively) and casualty trends by age, sex and road user type.

The main sources of data used for this study were the International Road Federation (IRF) annual statistical yearbooks, World Health Organisation (WHO) publications, recent regional and country studies, such as the Asian Development Bank (ADB) funded Road Safety in Asia/Pacific, the Inter American Development Bank financed Latin America Study and the European Commission PHARE Multi-Country Transport Programme. Road casualty data also came from TRL publications and country studies.

Previous reviews of global fatalities undertaken by TRL, World Bank and others have produced a wide range of estimates and whilst the problem of data reliability and under-reporting has been regularly acknowledged, traditional reliance has been on the use of officially published statistics based on police reports. In estimating causes of death and disability, the World Health Organisation (WHO) used a different method, based on registered deaths and health sector data that produced higher estimates than those using official police statistics. For example, WHO estimated a million deaths worldwide in 1990 whilst the TRL values were of the order of half this. Using their 1990 figure, WHO then estimated deaths in 1998 to be 1.17 million worldwide.

2.1 Methodology

In keeping with the traditional approach used by transport specialists in compiling road crash statistics, the starting point in this current study was the official fatality figure reported by countries. Using these values to obtain an accurate estimate of the current global fatality situation required several factors to be taken into account as follows:

1. Updating the fatality figure from the latest year (usually1995/96) to 1999.

2. Estimating for those countries where fatality data was not obtained.

3. Under-reporting.

a) recording deficiencies
b) non-reporting to the police

The general problem of under-reporting includes both recording deficiencies, i.e. ‘under recording’ where casualties are reported to the police but are not included in the published statistics, and non-reporting where the police are not notified of road casualties. To highlight the extent of under-reporting, the problems of recording deficiencies and non-reporting have been discussed separately in this study.

2.1.1 Regional classifications

There is no standard approach to regional groupings used by the many different international organisations concerned with road safety. However, in order to aid interpretation of data, a total of 192 countries were assigned to six major regional groups as follows:

• Africa.
• Asia/Pacific.
• Central and Eastern Europe (CEE).
• Latin/Central America and the Caribbean (LAC).
• Middle East and North Africa (MENA).
• HMCs, i.e. North America, Australia, New Zealand, Japan and Western Europe.

Less Motorised Countries (LMC) is the collective term used to describe the first five regions where motorisation is typically much lower than in the industrialised HMCs.

2.2 1998 updates

Most countries had published road fatality data available for 1996 and so the first step involved updating to the current year of 1999. Figure 1 shows the recent fatality trends from the different regions and the global level. While the CEE region reported fatalities peaking in 1990before dropping, the reverse was reported in the LAC region with fatalities increasing in recent years. After investigating the series data available for the larger countries within each region (see Table 1), it became apparent that the use of a regional average growth rate could produce misleading predictions because some countries would bias the figures.

Table 1  - Countries with time series data

For example it was found within the MENA region, that fatality statistics from Morocco and Saudi Arabia were three times those elsewhere in the region causing statistics from other countries to be ‘lost in their noise’. However, examination of data from the other countries indicated significantly different trends from those in Morocco and Saudi Arabia. Similar, though less extreme results, were found in data from other regions. For example, the rapid growth in crashes observed in India and China, which already have large number of fatalities, highly influenced any attempt to form a regression over all countries in their region.

A more accurate model was developed by sub-dividing each of the regions into two or three groups (see Table 2). Group 1 contained those countries with large number of road fatalities that dominated the region, all of which tended to have the same trends. Another group was formed from the other countries with fewer road fatalities. In some cases a third group was necessary, when trends between countries were significantly different. The details of the groups are shown above in Table 2.

Regression equations were fitted to the total number of road fatalities in each region, and separate regressions were fitted for each group in each region. Linear regression was used unless a strongly non-linear relationship was in evidence. In other cases LOGIT and quadratic equations were fitted.

The estimates shown in Table 3 are the values produced by the various equations for those countries with trend data. Where trend data was unavailable, the most conservative equation was applied within the region save for LAC and CEE where this would produce unrealistic fatality decreases (based on practical experience).

2.3 Non-reporting by countries

Despite a review that included many sources, road crash data could not be found for many countries in Africa and MENA (see Table 4). Africa had the lowest rate of representation and where data was available, it was often several years out of date.

While representation by motor vehicles might have been more suitable, many countries did not report the size of their motor vehicle fleet so this option was not possible. Accordingly, the number of reported road crash fatalities for the region was adjusted upwards according to the ratio of regional population as follows:

Table 2  - Groupings used in the regions

 Table 3 - 1999 minimum estimates for countries with fatality data

*Higher regional growth rate applied

Table 4  - Regional groups and crash data (1996)

The population adjustment formula shown above assumes a consistent fatality risk, i.e. road fatalities per 100,000 population, within the region, yet as seen in the regional summaries in Chapter 4, this is not always the case. As those countries not reporting road crash data tend to be the less developed, an adjustment based on population may over inflate to a small extent the regional fatality estimate.

As shown in Table 5, Africa was the only region where a large adjustment had to be made to adjust for those countries where published fatality figures could not be found. Overall, the global reported figure had to be adjusted by approximately 6 per cent to accommodate those countries where fatality data was not obtained.

Table 5 - 1999 global Reported fatality estimate

2.4 Under-recording of fatalities

The problem of under-recording occurs when fatalities are reported to the police but are not included in the official database. There are three main causes of under-recording:

• Fatality definition.
• Road crash definition.
• Database updating procedures.

2.4.1 Fatality definition

As defined by the Convention of Road Traffic (Vienna, 1968), a road death is deemed to have occurred when a person injured dies within 30 days of the crash (and as a result of the crash). As shown in the Appendix, however, not all countries use a 30-day definition with some countries using ‘on the spot’, within 24 hours, 3 days, etc. Adjustment factors have been developed by various organisations to bring these countries not using the 30-daydefinition `into line’. If this is not done, then a significant level of ‘under-recording’ will occur.

Adjustment factors identified included those recommended by the Economic Commission for Europe and the European Conference of Ministers of Transport (ECMT) and the United Nations. The more recent ECMT values are shown in the Table 6.

In the most recent UK Casualty Report (1998), the ECMT values were applied to Turkey (30% increase for a 1 day reporting definition) and Korea (15% increase for a 3 day reporting period) (DETR, 1999). Despite this, it was assumed that the ECMT values would not apply for all LMCs; a larger percentage of road crash fatalities can, unfortunately, be expected to die within the first day with the lack of advanced medical facilities as well as the higher percentage of vulnerable road user fatalities.

Table 6 - ECMT standardised 30 day road crash fatality adjustment factors

Given the lack of information on the timing of LMC road crash deaths, the assumption was made to use half the ECMT values for LMCs. Thus those LMCs reporting road crash fatalities occurring only within the first day of occurrence would have their figures increased by 15 percent rather than 30 per cent. This approach avoids any overestimate of deaths.

A second assumption was made regarding the standardised death definitions in LMCs; while many countries state the use of a 30-day definition, this could be interpreted to apply at the local level and for prosecution purposes. Road crash statistics are based on report forms that are often to be completed as soon as possible, i.e. ‘24hour report form’. It can be very difficult to modify previously reported or submitted figures, especially where manual reporting systems are used which is the case in most LMC at the local level where casualty reporting occurs. Accordingly, it was decided that regardless of the official definition, a one-day reporting time period would be assumed to apply for all of LMC and thus a 15% inflation factor was logically applied to all LMC reported fatalities.

2.4.2 Road crash definition

According to the Vienna Convention, the standard international definition of an injury road crash involves a collision of a moving vehicle on a public road in which a road user (human or animal), is injured (IRTAD, 1992). Some countries, however, require the involvement of a motor vehicle and others exclude certain types of crashes. For instance, China’s reporting policy excludes road crashes occurring at road works or rail crossings, a restriction that is believed to exclude thousands of deaths that are reported by the health authorities as road fatalities.

Likewise, Mexico’s official statistics refer only to deaths on the Federal road network and only account for 30 per cent of those reported by the national health authorities (ADB, 1998, WHO 1996). Other countries reported incomplete crash data sets with fatalities limited to urban areas (Madagascar) or for less than a full year (Benin).

2.4.3 Updating procedures

Updating and transferring records is a process prone to errors, especially when manual systems are involved, as so often is the case in LMCs. An early under-reporting study in Colombo, Sri Lanka found that while just over a half (53%)of the adult fatalities had been recorded by the hospital police post, when records were matched with the official police database, only 41 per cent were found to be included (Sayer and Hitchcock, 1984). After the introduction of a nationwide computerised crash reporting system, road fatalities in Bangladesh increased by 55 per cent. Until recently, Zimbabwe used to lose one month’s data each year when the official annual database was compiled.

Even computerised crash database systems in HMCs can exclude fatalities if required details are not provided within the reporting deadline. In the UK, a crash must be reported within 30 days of its occurrence with all the required information for it to be entered onto the STATS 19database. National centralisation policy will also affect the completeness of the database. For example, while the UK allows 6 months for the local police forces to provide their data to the national database, conversely France requires all road crashes of the preceding year to be reported by the end of January, a decision which is believed to result in some reported fatalities being excluded from the official national accident statistics.

The low priority given to accurate reporting of road crashes is also a contributory factor, with data collection often seen as only a paperwork exercise and with no real practical value. In many countries, road crashes are reported by the general and not the Traffic Police who could be expected to give the data collection more priority. This is partly because the general police, unlike most Traffic Police, work around the clock seven days a week.

Large inconsistencies in the LMC crash databases have been highlighted in previous studies (ADB 1996, WB 1998). Police may also be put under pressure to avoid documenting a worsening safety record. Such factors may explain how Indonesia could experience such a high rate of growth in motor vehicles in the past decade, i.e. almost 80 percent between 1986-95 while only reporting a 3 per cent rise inroad fatalities.

It has not been possible to develop standard adjustment factors for the problems mentioned above of updating and crash definitions but these data weaknesses will be considered in the assumptions made for adjusting for under-reporting.

2.5 Non-reporting by general public

Almost all countries require road crashes to be reported to the police. The UK is unusual in that it does not have a unilateral requirement to report road crashes, even fatal ones, to the police; the parties involved are only legally required to exchange names and addresses. If this is not done, then the crash should be reported to the police. Despite this, police reported statistics are still used to assess the road safety situation in the UK. The other extreme in reporting is found in Bahrain, where vehicle damage is not permitted to be undertaken without proof that the crash has been reported to the police.

2.5.1 HMCs

A 1991 review on under-reporting studies worldwide included studies from the UK, USA and Canada that reported complete coverage of road crash fatalities while in Germany 5-9 per cent of road crash fatalities were not reported to the police (James, 1991). A 1994 International Road Traffic and Crash Database (IRTAD) Special Report on the under-reporting of road traffic crashes quoted studies indicating a 3 percent level of fatality under-reporting in Spain and 2 percent in Switzerland. Research in Western Australia has also found that 5 per cent of road crash deaths were unknown to the police (Giles, 1994). Table 7 shows a comparison of the official police reported fatalities with that reported by WHO based on death certificates. The under reporting adjustment factor is the amount required to be added to the police reported figure in order to arrive at the WHO figure. Whilst the adjustment factor was found to be low in the USA, a surprisingly high 26 per cent needed to be added to official police statistics in Italy.

Table 7  - Road crash fatality comparisons

2.5.2 LMCs

WHO World Health Statistics (1996) included examples of much worse under-reporting of fatalities in LMCs, with for example only one out of every five medically reported road deaths being included in police statistics in the Philippines (see Table 8).

Table 8  - Road crash fatality under-reporting estimates

Similarly, in Indonesia, insurance companies reported 15,080 road fatalities in 1995, some 37 per cent more than the police reported. The Department of Health in Taiwan reported 7,250 road deaths in 1995, some 130 per cent greater than that reported by the police (3,094) although police statistics are limited to those fatalities occurring within 24 hours of a crash (Lu, 1999).

Due partially to the restricted definition of a reportable road crash, the under-reporting in China appears to be very high, unfortunately so for the country with already the largest number of road fatalities in the world. The Beijing Research Institute of Traffic Engineering estimated the actual number of fatalities in China for 1994 to be111,000, 42 per cent greater than the 77,860 reported officially by the police (Liren, 1996).

Despite the progress made by computerising the police reporting system in Bangladesh, road casualty reporting is still incomplete. A review of the road trauma fatalities in the capital Dhaka in 1996-1997 found 30 percent of those road fatalities reported in the newspapers were not included in the police database (IDC, 1997).

In Karachi, a recent study compared the road crash casualties reported by the police with those treated by the main ambulance service. Using the capture-recapture method, the study estimated that only 56 percent of road crash fatalities had been reported by the police in Karachi in 1994 (Razzak, 1998).

Another recent urban review of fatality under-reporting was conducted in Bogotá, Colombia. A comparison of the police reported fatalities with those on the city’s mortuary database found major discrepancies. Only 27 per cent of the mortuary’s road fatalities were found as fatalities on the police system (another 19 per cent were reported as injured). A further concern was that 19 per cent of the total fatalities were not found on the mortuary system so that even the (presumably) most reliable system is missing almost one fifth of all road fatalities (TRL, 1998).

2.5.3 Non-reporting adjustment factors

Evidence that under-reporting of fatalities ranged from 0-26 percent in HMCs and up to as high as 351 per cent in LMCs with fatalities in China estimated to be as high as 42per cent more than officially reported. Accordingly, to adjust for the extent of non-reporting of fatalities, as well as the under-recording weaknesses that could not be easily quantified, the conservative but realistic decision was made to use the following factors:

• HMC 2-5 per cent adjustment
• LMC 25-50 per cent adjustment

The factors indicate a probable range of fatalities. Given all the uncertainty in the estimation, a range is much more appropriate than a supposedly precise figure and estimates will be shown for both factors.

2.6 Current global fatality estimate

Based on the methodology described above, a realistic estimate of global road deaths is between 750,000 and 880,000 for the year 1999. The calculations and regional totals are presented in Table 9.

2.6.1 Comparison with WHO estimates

Whilst the fatality range presented here is lower than the recent WHO estimate, two points should be kept in mind. Firstly, the recent WHO estimates were based on the data synthesis conducted by the Global Burden of Disease (GBD) (1996). The ambitious objective of the project required major assumptions to be made, especially where data were absent. For example, it was reported that estimates for the entire region of sub-Saharan Africa were based only on South Africa, which in turn represented only about 1 percent of the region’s population (Cooper et al, 1997).

Secondly, two different methods of projection were used by WHO. Information on most causes of death (inc. road crashes) were based on complex forecasting techniques (usually from a base year of 1990) whilst others such as malaria, HIV/AIDS, tuberculosis were based on regional information collected as part of detailed and specific studies of these diseases. Even so, considerable variation exists in the forecasts derived for leading causes of death. Thus, for Sub Saharan Africa, deaths in 1998 from HIV/AIDS was stated to be 1.83 million but with a possible range of 1.1 million to 2.4 million. Similarly in Africa, deaths from malaria are estimated to be in the range 758,000 to 1.3 million and global deaths from tuberculosis of 1.49 million actually lies in the range 1.1 to 2.2 million.

The higher estimate derived in this study is a third less than the WHO value of 1.18 million estimated for 1998.However, as previously stated, the WHO study is based on 1990 data, and a number of assumptions about the distribution of fatalities in 1998 are made (World Health Report 1999). This, as shown above, leads to a range of possible fatalities for different diseases. Therefore, presenting a range of possible fatalities is not incongruent with the method employed by the WHO.

2.6.2 Global road crash injury estimate

While the extent of under-reporting is known to be even worse with injuries than with fatalities, a minimum estimate within a likely range has been included here. Based on the IRTAD report mentioned previously and earlier studies that had estimated approximately 50 percent of road injuries were reported, a ratio of 100 injuries for every fatality was assumed to apply in the HMCs. For LMCs, a ratio of between 20 to 30 was accepted as a conservative estimate. These values produce annual road crash injury estimates for 1999 of at least:

• 11 million in HMCs;
• 12 to 23 million in LMCs;
• Global estimate of between 23 and 34 million road crash injuries per annum.

Table 9  - 1999 Estimated road fatalities with under-reporting (UR) adjustments

This estimate is approximately twice other current global road injury figures (GRSP, 1999). An estimate of the number or percent of injuries that are disabling was beyond the scope of this review.

2.7 Fatality forecasts

Forecasting future deaths worldwide is fraught with difficulties. For example, past trends may be thought to give a reasonable picture of what may happen in the future. However some countries, such as Japan experienced rapid deterioration in road safety in the 1960s with an 80 per cent growth in road fatalities but then with massive investment reduced deaths by almost 50 per cent over the next decade. However deaths started to increase once again in the early1980s due in part to a continued increase in vehicle ownership but with a slowing down of investment in life-saving activities. Additionally, trends in many parts of the world are not consistent and there is evidence (see Section 4) that rapid increases of deaths in Africa and Asia/Pacific show signs of slowing down (that said growth rates in Africa and Asia are still high and of concern).

Social, political, and economic changes may also play apart and ideally would be taken into account in any forecasting activity. However, these changes are difficult to predict. For example, in the CEE region, changes in motor crash reporting most likely has changed with the transition to market economies. Whilst the trend in this region is one of fewer fatalities, it is probable that with economic development and rapid motorisation, there is potential for growth in the number of crashes and fatalities.

Forecasting future trends should be approached cautiously for the reasons outlined above. With these caveats in mind, we suggest that for 2010 the likely range of global road deaths will be between 900 thousand and 1.1 million and between 1 million and 1.3 million in 2020.

2.8 Summary

Based on officially reported fatalities, this study has attempted to produce a realistic estimate of global road crash deaths for the year 1999. Different recording definitions and procedures have been considered, as has the impact of under-reporting. One of the main objectives of this report was to document the estimation process and the data reliability. As the data has been shown to have many weaknesses, the estimate produced is considered to be realistic but conservative and the true toll could be higher.

The fatality estimate produced may be lower than the WHO figure but it still indicates that road crashes are a cause for concern. The burden of road fatalities is on the LMCs where 86 per cent of the world’s road fatalities occur, with almost half of all fatalities in Asia. Figure 2shows the regional distribution of 750,000 fatalities, the low end of the range suggested for 1999.

Figure 1 -  Estimated road fatalities – Regional Distribution

It should also be borne in mind that fatalities are only the tip of the casualty iceberg and that road safety, especially road safety engineering, is concerned with the reduction of injury road crashes. Worldwide, at least 30-45people are being injured for every life lost. However, data on injuries and their social and economic impacts is very limited, and it is outside the scope of this study to present trends in injuries. It can reasonably be assumed that serious injuries have a major social and economic impact. Nationwide costs of road crashes (including those for injury accidents) are presented in the following chapter.