Content uploaded by Chris Grundy
Author content
All content in this area was uploaded by Chris Grundy
Content may be subject to copyright.
Road Safety of London’s
Black and Asian Minority
Ethnic Groups
A report to the London Road Safety Unit
Rebecca Steinbach, Phil Edwards
Judith Green, Chris Grundy
London School of Hygiene and Tropical Medicine
1
Table of Contents
Acknowledgements 2
Summary 3
Part A: Relationships and Risks 10
1. Introduction 11
2. Methods 14
3. Results 20
3.1 Person 20
3.2 Place 24
3.3 Time 29
3.4 Multivariable analysis 32
3.5 Exposure to risk 38
4. Discussion 43
5. Recommendations 48
Appendices 49
Part B: Policy and Practice 60
1. Aims 61
2. Introduction 61
3. Methods 63
4. Findings 64
4.1 How important is the issue to BAME communities? 64
4.2 The boroughs’ perspective 65
4.3 Accounting for ethnic inequalities 70
4.4 Young people’s transport choices 73
4.5 Addressing inequalities 77
5. Discussion 79
6. Conclusion 82
References 84
2
Acknowledgements
Relationships & Risks – Traffic flow and speed data were supplied by Martin Obee at
Road Network Monitoring, Transport for London. The road network used was OS
ITN layer supplied by Transport for London under licence and is copyright Ordnance
Survey. 2001 census data were supplied with the support of ESRC and is crown
copyright. Digital boundaries are Crown and OS copyright. Dale Campbell at
Transport for London provided access to LATS 2001. Athanasios Nikolentzos helped
with data extraction for Part B of the report. We thank those who agreed to talk
with us about their views and experiences.
This work was undertaken by the London School of Hygiene & Tropical Medicine
who received funding from Transport for London. The views expressed are those of
the authors and not necessarily those of Transport for London.
Steinbach R, Edwards P, Green J, and Grundy C (2007) Road Safety of London’s
Black and Asian Minority Ethnic Groups: A report to the London Road Safety Unit.
London: LSHTM.
Further copies of this report are available from
http://www.tfl.gov.uk/streets/roadsafety-reports.shtml
3
Summary
Aims
Our previous study (Edwards et al. 2006) demonstrated a relationship between
deprivation and risk of road traffic injury in London, with pedestrians in particular,
at higher risk of injury in more deprived areas. This study builds on this work to
examine the relationship between ethnicity, deprivation and risk of road traffic
injury in London.
This study addressed four specific questions –
1) Are there differences in the risk of road traffic injury between different ethnic
groups in London?
2) How far can any differences identified between ethnic groups be accounted for
by: measurement errors; different levels of exposure; or different levels of
deprivation across areas of London?
3) Within ethnic groups, how far does deprivation affect the risk of road traffic
injury?
4) Taking into account what we know about differences in risk, possible
explanations for differences, what works to reduce risk, and the policy context
in London – what are the implications for policy and practice?
To do this, we analysed injuries recorded in STATS19 data between 1996 and 2006.
We used census data and GLA population projections to estimate injury rates across
ethnic groupings, and the Index of Multiple Deprivation to rank census Super
Output Areas in terms of deprivation. Ethnicity was coded by mapping STATS19
categories onto census categories, and deriving three broad groupings called
‘White’ , ‘Black’ and ‘Asian’. Interviews with policy makers, practitioners, young
people and parents were used to provide an overview for the policy context.
Background
There has been limited research on ethnic inequalities in road traffic injury risks in
the UK. Although previous studies have identified ‘differences’, these do not provide
any national pattern of which particular communities are at higher risk, and there is
little understanding of ‘what’ about ethnicity might lead to any differences
identified.
Summary
4
In London, research on this issue faces similar problems to elsewhere in the
country:
• London has many diverse ethnic communities, but data available only allow
us to aggregate figures for ‘Black’, ‘Asian’ or ‘White’ which obscure
differences between communities;
• It is difficult to calculate accurate rates for each grouping, as the ethnicity of
injured road users is classified by the police (through STATS19) using
different categories from those used (in the census) to estimate population
numbers. If there are large or systematic errors in how individuals are
classified by STATS19 or census data, we could under- or over-estimate
rates by ethnicity. Further, it is difficult to accurately estimate the size of
populations by ethnicity in small areas.
However, there is some evidence that there are ethnic inequalities in injury risks,
so it is important that we identify these inequalities as robustly as we can, and that
we suggest some possible explanations, in order to inform policy around road
safety which might address inequalities where possible.
Are some ethnic groups at higher risk of injury?
Between 1996 and 2006, there were 428,008 casualties recorded in road traffic
collisions occurring in London. Of those with ethnicity coded, we classified 262,310
(61.3%) as ‘White’, 54,348 (12.7%) as ‘Black’, and 38,858 (9.1%) as ‘Asian’.
Ethnicity was not coded for 64,233 (15.0%) casualties. Road traffic injury rates per
100,000 population differed by ethnicity. In children and adults, road traffic injury
rates were higher in ‘Black’ groups (305 per 100,000 population in children; 617 in
adults) and lower in ‘Asian’ groups (175 in children and 421 in adults), compared
with rates in ‘White’ groups (234 in children and 479 in adults). ‘Black’ Londoners
have been on average 1.3 times more likely to be injured on the roads than ‘White’
Londoners (appendix 1).
Between 2001 and 2006, rates of injury for children and adults in all ethnic groups
declined for all modes of travel. The rate of decline was similar across the ethnic
groups, with one exception: for adult car occupants, ‘White’ rates declined faster
than other groups.
Summary
5
How can we explain these differences between ethnic groups?
Measurement error – Some of the differences may be due to measurement errors.
These might include: systematic bias in under-reporting ethnicity of some groups in
STATS19, or inaccuracies in mapping STATS19 ethnic categories to census ethnic
categories. These could not account for all the differences between ‘Black’ and
other groups, but may explain some of the difference between ‘Asian’ and ‘White’
groups.
Exposure – If, on average, road users in different ethnic groups tend to live in more
dangerous traffic environments, or have different patterns of transport or leisure
activity, they will be more exposed to injury risk. Data on exposure to traffic are
limited and we did not identify significant differences in the average amounts of
walking across ethnic groups. However, more research could be done to examine,
for instance, differences in leisure-related exposure to traffic.
Deprivation – In London, there is a link between ethnicity and area level
deprivation: in least deprived deciles of census super output areas, an average
1.5% of the population is ‘Black’ and 6.6% is ‘Asian’, compared with an average
23.2% ‘Black’ and 15.6% ‘Asian’ in the most deprived deciles. Given that area
deprivation is linked to risk of injury, and more ‘Black’ people, on average, live in
the most deprived areas, we would expect more ‘Black’ people to be injured.
However, these area level effects do not explain all the difference.
How far does deprivation affect the risk of road traffic injury
within ethnic groupings?
For ‘White’ and ‘Asian’ groups, the risk of pedestrian injury was higher for each
decile of deprivation (measured by Index of Multiple Deprivation at census super
output area level). ‘White’ children in the most deprived areas were 2.5 times more
likely to be injured as pedestrians than those in the least deprived. For ‘Asian’
children, the injury rates in the most deprived areas were over 4 times higher than
for ‘Asian’ children in least deprived areas. However, for ‘Black’ children there did
not appear to be any relationship between deprivation and risk – the relative risk of
being injured was the same across deciles of deprived areas.
This suggests that deprivation does not account for all the differences in injury
rates between ethnic groups. It also suggests that deprivation may have different
effects in different ethnic groups. For instance, it is possible that lifestyle (and thus
exposure to traffic) differs between ‘White’ or ‘Asian’ children – depending on where
Summary
6
they live – but that the effect of lifestyle in ‘Black’ children is independent of area.
However, when we examined these relationships by ethnic group for adults injured
as pedestrians, we found similar relationships to those in children. That is, for
‘Black’ adults, the relative risk of injury is also the same across the deciles of
deprivation. This would tend to suggest that any explanation for ethnic differences
in how deprivation relates to injury risk, such as lifestyle or behavioural differences,
would also apply to adults.
It is important to note that the measure of deprivation used in our analysis includes
a number of domains that might be better at discriminating levels of deprivation to
some ethnic groups than others. It may be possible, then, that it is our measure of
deprivation (IMD) that has artificially “flattened out” a real underlying relationship
between deprivation and casualty rates for ‘Black’ children and adults. However, the
two domains of IMD which comprise nearly half of the IMD score are ‘income’ and
‘employment’ deprivation, neither which are likely to discriminate differentially
between ‘White’, ‘Black’ or ‘Asian’ Londoners.
What are the implications for policy and practice?
We have suggested, then, that ‘Black’ groups in London appear to be at higher risk
of road traffic injury, and that at least some of this excess risk is ‘real’ rather than
an artefact of inadequacies in the data available. ‘Asian’ groups appear to be at
lower risk than ‘Black’ or ‘White’ groups. We have also suggested that although
deprivation levels of a neighbourhood are an important influence on risk, they do
not account for all of this risk. In the two most deprived deciles of the population,
there are no differences in the injury rates between ‘White’ and ‘Black’ Londoners,
but in more affluent areas, ‘Black’ rates are higher, suggesting that increasing area
affluence protects ‘White’, but not ‘Black’ road users. There are grounds for
predicting that exposure to traffic may account for some of the risk differential, but
data available have not been able to identify how much.
There are a number of challenges in implementing road safety initiatives in ways
that are likely to reduce the observed ethnic inequalities in injury rates:
• Available data are at a crude aggregated level (e.g. ‘White’, ‘Black’, ‘Asian’)
that both obscures important differences between groups, and bears little
relationship to local communities’ own identification of ethnicity;
Summary
7
• Available data are not sufficient to tell us why there appears to be an
increased rate in those groups identified as ‘Black’, and a possibly lower rate
in those identified as ‘Asian’.
Discussions with key stakeholders in London (including local authority road safety
staff, community organisations, regional policy makers, young people, parents)
raised a number of issues that need to be taken into account:
• Some Black community groups and parents reported a lack of awareness of
road danger as an issue that affects them, and there are opportunities of
raising interest in the issue;
• Young Black people were concerned about the potential for further stigma –
this is another issue where their behaviour is seen as “a problem”;
• Given the uncertainties about both why there are ethnic differences, and
what would work to reduce them, programmes should be broad enough to
meet other goals (e.g. Community engagement) rather than narrowly
directed at ‘Road Safety’;
• Policy should be broadly directed at making London’s roads safer to travel
around, and neighbourhoods safer to play in, rather than in problematising
the behaviour of particular groups.
In general, interventions directed at making the environment safer (e.g. reducing
the speed and volume of traffic) will reduce injury risk for the whole population in
the longer term, as well as reducing the differences across ethnic groups. However,
in the short term, it will be necessary to work with local communities to look at
ways of managing existing risks.
Recommendations
The first three recommendations relate to needs for more robust information:
1) Analysis based on STATS19 data and area-level measures has provided a
‘broad brush’ picture of the relationship between deprivation, ethnicity and
road traffic injury, but further research is needed to:
Summary
8
• Understand in detail different patterns of exposure to risk of road
traffic injury, particularly for children, and how these relate to
deprivation;
• Look at the impact of existing interventions (e.g. 20mph zones) on
ethnic inequalities.
2) To monitor trends in the relationship between road traffic injury and
ethnicity, the most useful outcome measures are rates of child pedestrian
and adult pedestrian casualties
3) Work on improving the completeness of STATS19 data should continue, with
monitoring under-reporting and recording of road traffic injuries.
The final two recommendations relate to potential policy implications:
4) The headline findings on ethnic differences in road traffic injury rates could
be used to raise awareness of the issue of road safety. There is considerable
potential for Local authority road safety teams and Transport for London to
work with both statutory partners (e.g. Equality or Diversity teams) and 3rd
Sector partners representing BAME communities, to include road safety
issues as part of a broader community safety agenda.
5) Although similar rates of decline in road traffic injury rates across ethnic
groups suggest that current strategies are, in general, addressing needs
across the population, to reduce observed inequalities it will be necessary to
reduce injury rates faster in groups identified as ‘Black’. However, given the
limited knowledge we have of how exposure to risk and other variables
interact to put people at higher risk, interventions designed to address
ethnic inequalities need to be carefully designed in consultation with local
communities in order to:
• Avoid ‘victim blaming’;
• Ensure that Road Safety teams understand the precise risks faced
from the perspective of those affected;
• Ensure that programmes are appropriate and tailored to community
needs.
‘Local communities’ in this context will include neighbourhood communities, but
also groups which identify themselves in terms of faith, ethnicity or other
communalities (e.g. young people).
Summary
9
10
Part A: Relationships and Risks
Part A: Relationships and Risks
11
1. Introduction
The London School of Hygiene & Tropical Medicine (LSHTM) recently completed a
research project for the London Road Safety Unit (LRSU) entitled Deprivation and
Road Safety in London that investigated the relationship between road traffic injury
and deprivation in London. The results suggest that there are persisting socio-
economic inequalities in casualty rates for different road user groups in London, and
that differentials are in part associated with minority ethnic status. This raises
issues for equality and inclusion in London’s road safety strategy.
Internationally, studies have found large disparities in road traffic injury rates by
ethnic group (Savitsky et al., 2007; Stirbu et al., 2006; Braver, 2003; Campos-
Outcalt et al., 2002; Stevens and Dellinger, 2002; Schiff and Becker, 1996).
Evidence in the U.K. is limited, but suggests that injury rates are disproportionately
high for some Black and Minority Ethnic (BAME) road user groups (e.g. Lawson and
Edwards, 1991; Christie, 1995). While these international and British studies concur
that ethnic minorities are at greater risk of road traffic injury, they provide
conflicting evidence of who is at risk. In the international studies cited above, ethnic
minorities described as ‘Hispanic’, ‘American Indian’, ‘non-Jewish’, and of Turkish,
Moroccan, Surinamese, or Antillean/Aruban origin have been found to have higher
road traffic injury rates than the native population. Within the UK, both ‘Asian’ and
‘non-White’ groups have been found to be at increased risk of injury, depending on
the timing and location of the study. This suggests that there is nothing
fundamental about belonging to a particular minority ethnic group that causes
traffic injury. Rather, perhaps there is something context-specific about belonging
to a particular ethnic minority within a particular environment that is associated
with high road traffic injury rates.
Why do risks appear to vary between ethnic groups?
The reasons for ethnic differences in road traffic injury are unclear, but are likely to
be at least partially explained by the strong association between ethnicity and
socio-economic status, particularly in London (Edwards et al., 2006; Grayling et al.,
2002). Other explanations offered for the observed differences in road traffic injury
rates by ethnic group include:
Exposure to risk of injury from traffic – There are two components to exposure
risk, the time spent on roads and the relative danger (i.e. due to the volume
Part A: Relationships and Risks
12
and speed of traffic) of the roads that are used. Ethnic differences in road traffic
injury may be explained by exposure differences if people from minority ethnic
groups spend relatively more time as road users, or use roads with higher traffic
volumes and traffic speeds.
Risk perception and behaviour – Cultural factors may play a role in ethnic
differences in risk perception and risk behaviour (DfT, 2002). For example, it
has been suggested that different methods of parental supervision and teaching
of road safety skills may contribute to ethnic differences in child road traffic
injury risk.
Measurement error – Ethnic differences in road traffic injury may be an artefact
of the data due to the inconsistent and differential measurement of ethnicity in
different data sources.
It is also important to remember that any associations found between belonging to
a particular ethnic group and road traffic injury are merely associations. Although
we can assess how far differences are accounted for by socio-economic factors
(and, for instance, suggest that these do not account for all of observed
differences) we cannot control for all other differences between groups defined
through ethnicity. This means that it is not possible to make firm claims about the
relationships between being a particular ethnicity and risk of injury, as we do not
know whether the variable ‘ethnicity’ is simply a proxy for some other unmeasured
variable. Also, even if it were, it is not possible to know what it is ‘about’ ethnicity
that leads to increased risk.
Despite steady casualty reductions for most road users across London (Tfl, 2004)
concerns remain that they have not been shared equally, particularly by minority
ethnic groups. In 2007, Transport for London (TfL) commissioned LSHTM to conduct
a study of ethnicity and road traffic injury risk, in order to provide an evidence base
for recommendations that are applicable specifically to London.
The aims of the study were to: describe the relationship between ethnicity and road
traffic injury risk; identify possible mechanisms that may link ethnicity and road
traffic injury risk; offer recommendations on monitoring ethnic inequality in road
traffic injury risk across London, and on policies to address ethnic inequalities in
road traffic injury risk.
Part A: Relationships and Risks
13
In this report we examine the strength of the association between ethnicity and
road traffic injury risk for different road user groups in London. Our analysis covers
children and adults injured on London’s roads as pedestrians, pedal cyclists,
powered 2-wheeler riders, and as car occupants.
Using STATS19 data collected by the Metropolitan Police and City Police, we
compare the relative risks of road traffic injury to groups of Londoners, using
groups constructed using data on age, sex, ethnicity, borough, year and season.
Then by linking casualties to the areas in which the collisions occurred, we use data
describing area-level deprivation and features of the road network to examine the
relationship between ethnicity, deprivation and the road environment.
A model of the links between ethnicity and road traffic injury
The causal pathways linking ethnicity to road traffic injury risk are likely to be
complex. In principle, the relative risk of injury on the roads is determined by three
variables: the road environment (e.g. speed and volume of traffic; number of
junctions, etc.); an individual’s exposure to that environment (e.g. how often they
are on, or near, roads), and their behaviour when on, or near, the roads (e.g. risk-
taking, crossing at controlled crossings, etc.). These three variables are inter-
related: behaviour and levels of exposure are, to some extent, a consequence of
perceived dangerousness of the road environment.
Ethnicity may impact on injury risk because it is associated with other factors, such
as area-level or individual-level deprivation, that are known to be related to road
traffic injury risk (Edwards et al. 2006). Individual-level deprivation may impact on
exposure to risk (e.g. more likely to use public transport, or to make long journeys
to work or school). Area-level deprivation is associated with more dangerous road
environments. There may also be separate pathways by which ethnicity impacts on
road traffic injury risk, if there is something about being in a particular ethnic group
that influences exposure, road environment or behaviour directly. For example, if
‘White’ or ‘Black’ people are more, or less likely, to use particular modes of
transport, or to socialise in public rather than private space, exposure may vary
between these groups. These indirect and direct effects of ethnicity are also inter-
related, given that cultural differences are shaped by material circumstances.
In this report, we have looked separately at those parts of these causal pathways
for which we have empirical data.
Part A: Relationships and Risks
14
2. Methods
Two types of analyses were carried out to investigate the association between
ethnicity and road traffic injury risk in London: univariable and multivariable. The
univariable analysis describes the distribution of injuries and of injury rates in
groups by age, sex and ethnicity (‘person’); by borough and Inner/Outer London
(‘place’); and by year and season (‘time’). The multivariable analysis describes the
strength of the relationship between injury rates and the individual and area-level
factors (e.g. deprivation; speed and volume of traffic).
The multivariable analysis was carried out using small geographical areas known as
‘census Lower Super Output Areas’ – referred to throughout this report as ‘SOAs’.
London has 4,765 SOAs contained within 33 boroughs. Each SOA includes an
average of 1,500 people and were created by the Office for National Statistics
(ONS) using measures of population size, mutual proximity and social homogeneity
(similarity), to provide robust small-area statistics for use in analyses that seek to
compare areas.
Measures of injury
We were provided with a data file from the London Road Safety Unit containing
STATS19 data for all road traffic injury collisions in London between 1996 and
2006. Casualties were identified according to road user group (pedestrian, pedal
cyclist, powered 2-wheeler and car occupant). Each casualty was assigned to an
SOA based on the Ordnance Survey Grid reference of the location where the
collision occurred. Casualties with home address postcodes outside of London were
removed from the data set.
Population estimates
More than one source of population estimates were used for our analyses. In the
univariable analysis, we used population data from the 2001 census on the number
of children (ages 0–14 years) and adults living in London who classed themselves
as ‘White’, ‘Black’, or ‘Asian’. Additionally, for the analysis of the relationship
between ethnicity and road traffic injury risk over time, we used the Greater
London Authority (GLA) 2006 Round Ethnic Group Population Projections (EGPP) to
estimate injury rates in each year from 2001 to 2006.
For multivariable analysis, we estimated the numbers of ‘White’, ‘Black’ and ‘Asian’
children (ages 0–15 years) and adults living in each SOA, so that analysis of
Part A: Relationships and Risks
15
variation in road traffic injury rates could be conducted at the SOA level. These
were estimated by multiplying the total numbers of children (ages 0–15 years) and
adults living in each SOA (from 2001 census), by the percentages of residents of all
ages that are ‘White’, ‘Black’, or ‘Asian’ (also from 2001 census). These estimates
of SOA-level ethnic group populations were then scaled to ensure that our
estimates of ethnic-specific borough populations were equal to those in the census.
Measures of exposure
We used the London Area Transport Survey (LATS 2001) to estimate the amounts
of walking, cycling, travel by car or by powered 2-wheelers, for each of the ethnic
and age groups. We assessed the evidence for differences in the average amounts
of time spent by each ethnic group as a pedestrian, cyclist, etc., and in the total
distances of trips made by each mode. The LATS 2001 data were collected using
daily travel diaries, kept by the survey participants. Access to the data was
provided through Transport for London.
Measures of ethnicity
STATS19 – In London, police officers assign an ethnicity category to each casualty
and to drivers or riders. This coding of ethnicity in the STATS19 data is unique to
London and police have been including it since 1995. Ethnicity is assigned to one of
seven categories: White-skinned European, Dark-skinned European, Afro-
Caribbean, Asian, Oriental, Arab, and Unknown.
2001 Census – respondents were asked to classify their ethnicity by selecting from
16 categories.
White:
• British • Irish • Other
Black or Black British:
• Caribbean • African • Other
Asian or Asian British:
• Indian • Pakistani • Bangladeshi • Other
Chinese or Other:
• Chinese • Other
Mixed: White and Black
• Caribbean • African
Mixed: White and Asian Mixed: Other
Part A: Relationships and Risks
16
LATS 2001 – ethnicity was self-reported in the survey. Respondents were offered a
choice of nine categories: White, Black-Caribbean, Black-African, Black-Other,
Indian, Pakistani, Bangladeshi, Chinese, and Other.
To incorporate data from each source (LATS 2001, STATS19, and 2001 Census) in
our analysis, the categories of ethnicity used in each data set were mapped
(Table 1). The mapping chosen for the analysis for TfL resulted in four broad ethnic
groupings: ‘White’, ‘Black’, ‘Asian’ or ‘Other’. Census respondents reporting mixed
‘White’ and ‘Black’ identities were included in the ‘Black’ category because it was
assumed that police would more likely identify such identities as Afro-Caribbean.
‘Mixed Asian and White’ census respondents were assigned to the ‘Asian’ category.
Table 1: Ethnicity mapping between data sources
TfL study STATS19 Census 2001 LATS 2001
White White-skinned European British White
Dark-skinned European Irish
Other White
Black Afro-Caribbean Caribbean Black-Caribbean
African Black-African
Other Black Black-Other
Mixed-White & Black
Caribbean
Mixed-White & Black
African
Asian Asian Indian Indian
Pakistani Pakistani
Bangladeshi Bangladeshi
Other Asian
Mixed-White & Asian
Other Arab Other Other
Oriental Chinese Chinese
Mixed-other
For this report we have focused only on the first three of these ethnic groupings:
‘White’, ‘Black’ and ‘Asian’. Analyses for individuals in the ‘Other minority ethnic
group’ have not been reported, as sample sizes for this group were relatively small,
and interpretation of comparisons of injury rates relative to the larger ethnic groups
would be far less reliable. It is also important to note that ethnicity was self-
reported in the 2001 Census, but is defined by police officers in STATS19.
Therefore, in injury rate calculations the numerator (casualties) and the
denominator (population size) may describe two somewhat different populations.
This likely reduces the accuracy of ethnic-specific injury rate estimates, and must
be borne in mind when comparing road traffic injury rates by ethnic group (see
Discussion).
Part A: Relationships and Risks
17
Measures of deprivation
Since deprivation has been found to be highly associated with road traffic injury
rates (Edwards et al., 2006; Grayling et al., 2002), and ethnicity is also associated
with deprivation, we included measures of deprivation in the multivariable analysis,
to investigate this complex relationship.
Index of Multiple Deprivation
The Index of Multiple Deprivation (IMD) brings together 36 indicators across seven
domains of deprivation into an overall score and rank for a geographical area. The
index was designed to provide a robust small-area measure of deprivation which
encompasses the many different dimensions in which deprivation can be recognized
and measured. The seven domains of deprivation are: Income; Employment;
Health and disability; Education, skills and training; Barriers to housing and
services; Crime; and Environment. The IMD score is an ordered scale where higher
IMD scores indicate relatively more deprived areas. For a full description of the IMD
domains, see Deprivation and Road Safety in London (Edwards et al., 2006).
One potential problem of using IMD for this analysis is that an indicator of road
traffic accidents involving injury to pedestrians and cyclists (2000-2002) is included
in the Environment domain of the IMD. Therefore the IMD score might be partially
correlated with pedestrian injury risk, and this could distort the observed
relationship between deprivation, ethnicity, and road traffic injury risk. However,
the road traffic accident indicator only contributes a total of 2.5% to the overall
IMD score. In Deprivation and Road Safety in London (Edwards et al., 2006), we
found no evidence that using the full IMD (including the Environment domain)
biased the observed association between deprivation and child pedestrian injury
risk. We have therefore used the complete IMD score for the analysis presented in
this report.
Deprivation deciles
The values of IMD were obtained for all 4,765 census SOAs in London and were
then used to rank SOAs into deciles (tenths) from 1 (least deprived SOAs) to 10
(most deprived SOAs). These tenths of London’s SOAs are referred to as
‘Deprivation Deciles’ throughout this report.
Part A: Relationships and Risks
18
Road network variables
Road network variables were incorporated into the multivariable analysis to take
into account variations in the complexity of the road traffic environment between
areas. The variables considered included SOA level estimates of the number and
density of road junctions (where two or more roads meet), as well as the length
and density of A, B, and minor roads and motorways in each SOA, and in adjacent
SOAs. At borough level, variables considered were: the number of A, B, and minor
roads and motorways; morning traffic speeds of A, B, and minor roads, and
motorways; the difference between morning and evening traffic speeds and free-
flowing traffic speed; and average traffic flows in 2001. Road densities were
calculated by summing the length of roads within each SOA and dividing by the
area of the SOA. Similarly, the density of road junctions within each SOA was
calculated by summing the number of junctions and dividing by the area of the
SOA.
Statistical analysis
To estimate injury rates for the univariable analysis by age, sex, ethnicity and
borough, we used the number of casualties in each group as the numerator and the
2001 census population of each group (extrapolated to represent the entire 1996-
2006 period) as the denominator. In the analysis of injury rates over time we used
population projections from 2001 to 2006 as the denominator.
In multivariable analysis, we investigated the relationship between injury rates and
deprivation separately for each of the three ethnic groups (‘White’, ‘Black’ and
‘Asian’). We used multivariable regression analysis to calculate injury rate ratios,
with 95% confidence intervals, comparing ethnic group-specific injury rates in the
least deprived areas of London to injury rates in relatively more deprived areas,
adjusting for the road environment. The Poisson distribution was used unless there
was evidence for ‘over-dispersion’, when the Negative Binomial distribution was
used.1 Standard errors of estimates from multivariable analysis were adjusted to
allow for within-borough correlations in SOA injury rates (so called ‘intra-cluster
correlation’).
1 It is common to use the statistical distribution called the ‘Poisson’ distribution when modeling rates in a
population. However, this assumes that the average rate (mean rate) over all SOAs is equal to the variance
(spread) of rates. If the mean is not equal to the variance in the sample, it is common to use the negative
binomial distribution instead.
Part A: Relationships and Risks
19
Road environment variables were selected using both backward and forward
stepwise regression to model injury rates in the three ethnic groups.2 Road
environment variables were then chosen to be included in a final adjusted model if
there was good evidence (i.e. p<0.05) for an association with injury rates in at
least two of the models. All analyses were conducted using the Stata Statistical
Software (StataCorp, 2005).
2 In forward stepwise regression, a model is fitted to the data by adding variables one at a time. A variable will
remain in the model if its p-value (testing no association with injury rates), is less than 0.2. In backward
stepwise regression, a model is initially fitted to all variables, and then variables are excluded if their p-values
are greater than 0.2.
Part A: Relationships and Risks
20
3. Results
The STATS19 file contained data on 450,153 casualties from a total of 374,356
road traffic collisions in London between 1996 and 2006. Of these casualties,
22,135 were excluded from the analysis because they reported their home address
postcodes were outside London. Of the remaining 428,018 injuries, 363,775 (85%)
had been assigned an ethnicity code.
3.1 Person groups
Our initial analysis of the relationship between injury rates and ethnicity was based
on ethnicity-specific injury rates for each road user, age, and sex group.
Pedestrians
There was a total of 78,716 people injured as pedestrians in London between 1996
and 2006. Annual pedestrian injury rates within age-sex groups ranged from 29 to
313 per 100,000 people. Pedestrian injury rates appeared highest in ‘Black’ children
and adults of all ages, males and females.
Table 2: Average annual pedestrian injury rates
per 100,000 people, 1996-2006
Ethnic group
Age group Sex ‘White’ Black ‘Asian’
0-4 M 45 95 68
F 29 52 41
5-9 M 125 235 141
F 72 135 69
10-14 M 254 313 136
F 179 255 97
15-24 M 144 164 84
F 122 148 69
25-34 M 84 124 61
F 63 84 44
35-44 M 75 97 56
F 46 62 38
45-54 M 68 106 61
F 43 69 46
55-64 M 68 102 78
F 49 82 49
65+ M 85 127 109
F 68 101 58
Part A: Relationships and Risks
21
Pedal cyclists
A total of 35,925 pedal cycle injuries occurred in London between 1996 and 2006.
Annual pedal cycle injury rates within age-sex groups ranged from 0 to 127 per
100,000 people. Pedal cycle injury rates appeared higher in ‘White’ children and
adults than in the other two ethnic groups. The exception was in ‘Asian’ boys aged
5–9 years, where the cycling injury rate was higher than in ‘White’ boys.
Table 3: Average annual pedal cycle injury rates
per 100,000 people, 1996-2006
Ethnic group
Age group Sex ‘White’ ‘Black’ ‘Asian’
0-4 M 2 2 1
F 1 0 0
5-9 M
41 9 56
F 11 9 3
10-14 M
127 118 51
F 22 16 4
15-24 M
108 102 32
F 31 11 4
25-34 M
102 86 19
F 43 10 3
35-44 M
80 58 14
F 21 5 1
45-54 M
48 27 8
F 13 2 1
55-64 M
32 17 8
F 8 1 0
65+ M
12 9 3
F 2 0 0
Part A: Relationships and Risks
22
Powered 2-wheeler
There was a total of 63,597 people injured whilst riding powered 2-wheelers in
London between 1996 and 2006. Annual powered 2-wheeler injury rates within
age-sex groups ranged from 0 to 335 per 100,000 people. Powered 2-wheeler
injury rates were highest in ‘White’ adults, male and female, from age 25 years and
older. In ‘White’ and ‘Black’ adolescents and young men, the powered 2-wheeler
injury rates appeared more similar.
Table 4: Average annual powered 2-wheeler
injury rates per 100,000 people, 1996-2006
Ethnic group
Age group Sex ‘White’ ‘Black’ ‘Asian’
0-4 M 0 0 0
F 0 0 0
5-9 M 1 0 0
F 0 0 0
10-14 M
12 14 2
F 2 1 0
15-24 M
321 335 89
F 40 13 5
25-34 M
297 171 83
F 45 10 6
35-44 M
222 106 45
F 20 4 2
45-54 M
105 32 17
F 9 2 1
55-64 M
47 4 7
F 3 1 0
65+ M 8 0 2
F 1 0 0
Part A: Relationships and Risks
23
Car occupants
A total of 187,398 people were injured as car occupants in London between 1996
and 2006. Annual car injury rates within age-sex groups ranged from 44 to 510 per
100,000 people. Over the age of 25 years, car occupant injury rates appeared
highest in ‘Black’ adults and lowest in ‘White’ adults.
Table 5: Average annual car injury rates per
100,000 people, 1996-2006
Ethnic group
Age group Sex ‘White’ ‘Black’ ‘Asian’
0-4 M
48 58 44
F 47 56 47
5-9 M
69 85 72
F 86 103 68
10-14 M
69 61 68
F 94 79 77
15-24 M
423 418 471
F 401 300 258
25-34 M
244 510 407
F 272 373 258
35-44 M
207 426 350
F 219 286 261
45-54 M
160 347 279
F 184 275 238
55-64 M
134 246 207
F 134 156 157
65+ M
98 123 112
F 74 68 75
Part A: Relationships and Risks
24
3.2 Place
Next, our analysis considered the relationship between ethnicity and injury rates
across different London boroughs. The City of London was excluded from the
analysis as this borough tends to have a large day-time population (tourists and
workers) and a small resident population. In this analysis we have linked casualties
to boroughs using the SOA of collision location.3
Pedestrians
‘White’ child pedestrian injury rates ranged from 65 per 100,000 in Richmond upon
Thames to 197 per 100,000 in Newham. ‘Black’ child pedestrian injury rates ranged
from 99 per 100,000 in Sutton to 227 per 100,000 in Wandsworth. ‘Asian’ child
pedestrian injury rates ranged from 37 per 100,000 in Sutton to 159 per 100,000 in
Waltham Forest. Injury rates among ‘White’, ‘Black’ and ‘Asian’ children were
higher in Inner London compared to Outer London. The injury rate ratio comparing
‘White’ children in Inner London to ‘White’ children in Outer London was 139/105 =
1.32, with a 95% confidence interval4 of 1.27-1.38. Among ‘Black’ children the
Inner/Outer London rate ratio was 1.19 (1.12-1.26) and among ‘Asian’ children the
rate ratio was 1.21 (1.12-1.31).
‘White’ adult pedestrian injury rates ranged from 37 per 100,000 in Bexley to 311
per 100,000 in Westminster. ‘Black’ adult pedestrian injury rates ranged from 37
per 100,000 in Bexley to 314 per 100,000 in Westminster. ‘Asian’ adult pedestrian
injury rates ranged from 27 per 100,000 in Sutton to 179 per 100,000 in
Westminster. Similar to the pattern in child pedestrians, adult pedestrian injury
rates were higher in Inner London compared to Outer London in all ethnic groups.
The pedestrian injury rate ratio comparing ‘White’ adults in Inner London to ‘White’
adults in Outer London was 2.21 (2.17-2.26). Among ‘Black’ adults the ratio was
1.53 (1.46-1.60) and among ‘Asian’ adults it was 1.58 (1.49-1.67).
3 This allows the entire STATS19 data set to be used, but assumes that casualties are injured in the borough in
which they live. This assumption is valid for pedestrians and cyclists, but less so for other road user groups.
However, an analysis of distance from home to collision location (Appendix 3), provides evidence that
casualties of all age, ethnicity, and road user groups tend to be injured on average less than 6Km from home,
with median distances from home below 4.5Km.
4 The confidence interval represents the range of values that are likely to contain the true injury rate ratio.
Part A: Relationships and Risks
25
Table 6: Average annual pedestrian injury rates per 100,000 people, 1996-2006
Children (0–14) Adults
Borough ‘White’ ‘Black’ ‘Asian’ ‘White’ ‘Black’ ‘Asian’
Inner London 139 188 103 119 128 84
Camden 134 187 80 159 176 112
Hackney 139 203 95 99 122 55
Hammersmith and
Fulham 109 181 66 96 132 107
Haringey 136 199 94 97 118 71
Islington 153 191 99 129 150 107
Kensington and Chelsea 82 115 41 133 150 102
Lambeth 134 223 89 104 137 104
Lewisham 153 176 75 81 92 70
Newham 197 147 141 85 80 64
Southwark 151 174 76 89 119 69
Tower Hamlets 187 182 89 98 150 70
Wandsworth 106 227 126 64 132 87
Westminster 154 219 96 311† 314† 179†
Outer London 105 159 85 54 84 53
Barking and Dagenham 141 140 105 45 51 38
Barnet 88 165 54 57 72 42
Bexley 100 153 77 37 37 32
Brent 124 199 87 89 97 64
Bromley 89 172 58 41 85 28
Croydon 124 154 67 63 90 43
Ealing 109 182 99 74 119 88
Enfield 113 140 72 59 67 45
Greenwich 161 171 64 63 77 54
Harrow 78 134 67 58 68 37
Havering 96 139 90 40 73 47
Hillingdon 102 136 81 47 94 45
Hounslow 125 155 86 58 100 53
Kingston upon Thames 69 101 47 49 119 38
Merton 101 159 60 51 87 44
Redbridge 85 114 107 51 53 47
Richmond upon Thames 65 144 39 48 99 43
Sutton 84 99 37 41 85 27
Waltham Forest 149 160 159 62 79 73
Greater London 115 175 92 78 109 64
†Adult pedestrian injury rates in Westminster reflect the particularly high number of visitors
to this borough.
Part A: Relationships and Risks
26
Pedal cyclists
Pedal cyclist injury rates for ‘White’ and ‘Asian’ children were lower in Inner London
compared with Outer London (rate ratios were 0.77 (0.71-0.83) for ‘White’ children
and 0.77 (0.62-0.97) for ‘Asian’ children). There was less evidence for a difference
between Inner and Outer London rates for ‘Black’ children injured as cyclists (rate
ratio 0.95, 95%CI 0.83-1.09). Among ‘White’, ‘Black’ and ‘Asian’ adults, pedal
cycling injury rates were higher in Inner London compared with Outer London. Rate
ratios were 2.88 (2.80-2.96) for ‘White’ adults, 1.84 (1.68-2.01) for ‘Black’ adults,
and 2.13 (1.86-2.44) for ‘Asian’ adults.
Table 7: Average annual cycle injury rates per 100,000 people, 1996-2006
Children (0–14) Adults
Borough ‘White’ ‘Black’ ‘Asian’ ‘White’ ‘Black’ ‘Asian’
Inner London 27 30 11 76 42 16
Camden 21 14 15 97 62 18
Hackney 27 35 10 75 42 17
Hammersmith and Fulham 32 52 4 76 63 28
Haringey 20 24 7 30 21 13
Islington 28 26 12 110 63 35
Kensington and Chelsea 11 47 0 81 75 33
Lambeth 23 31 6 86 45 24
Lewisham 34 27 14 36 23 7
Newham 40 21 16 47 21 11
Southwark 36 26 11 82 36 18
Tower Hamlets 34 28 8 63 67 12
Wandsworth 32 52 11 61 55 16
Westminster 17 31 16 141 99 24
Outer London 36 31 14 27 23 8
Barking and Dagenham 52 26 31 20 8 5
Barnet 25 27 9 18 14 3
Bexley 33 8 19 12 9 7
Brent 27 37 9 38 21 6
Bromley 29 30 4 19 12 4
Croydon 30 31 8 25 21 5
Ealing 32 31 18 40 42 13
Enfield 35 31 12 19 14 5
Greenwich 32 23 11 26 16 11
Harrow 32 56 13 19 19 4
Havering 39 46 24 14 13 10
Hillingdon 55 29 24 28 13 13
Hounslow 51 41 17 54 46 10
Kingston upon Thames 33 30 13 44 33 10
Merton 36 38 14 35 39 9
Redbridge 39 31 15 21 16 6
Richmond upon Thames 32 11 13 47 76 10
Sutton 35 22 4 22 20 5
Waltham Forest 34 26 20 29 30 9
Greater London 33 30 13 45 34 10
Part A: Relationships and Risks
27
Powered 2-wheeler
Powered 2-wheeler injury rates appeared higher for ‘White’ adults compared with
‘Black’ or ‘Asian’ adults in almost all London boroughs. Powered 2-wheeler injury
rates in Inner London appeared higher than in Outer London in all ethnic groups:
rate ratios comparing Inner with Outer London were 2.20 (2.16-2.24) for ‘White’
adults, 1.42 (1.34-1.51) for ‘Black’ adults, and 1.97 (1.82-2.13) for ‘Asian’ adults.
Table 8: Average annual powered 2-wheeler injury rates per 100,000 people,
1996-2006 Children (0–14) Adults
Borough ‘White’ ‘Black’ ‘Asian’ ‘White’ ‘Black’ ‘Asian’
Inner London 4 3 1 156 83 45
Camden 4 3 1 176 150 51
Hackney 3 3 0 122 67 46
Hammersmith and Fulham 10 4 0 142 107 90
Haringey 4 4 0 77 55 30
Islington 6 5 0 162 122 72
Kensington and Chelsea 4 2 0 170 142 95
Lambeth 3 3 0 187 88 85
Lewisham 3 1 0 117 61 52
Newham 6 0 1 88 27 20
Southwark 6 3 0 161 66 66
Tower Hamlets 4 7 1 189 122 23
Wandsworth 1 1 0 132 110 72
Westminster 3 2 0 285 220 80
Outer London 2 2 0 71 58 23
Barking and Dagenham 2 0 0 57 27 36
Barnet 2 0 1 70 64 17
Bexley 2 0 0 46 25 15
Brent 1 1 0 104 71 20
Bromley 0 2 0 59 83 38
Croydon 2 2 0 84 70 21
Ealing 3 4 1 93 70 30
Enfield 1 4 0 53 49 23
Greenwich 4 3 0 97 36 17
Harrow 2 2 0 52 48 12
Havering 1 0 0 46 35 25
Hillingdon 3 3 1 61 42 23
Hounslow 3 5 0 110 87 23
Kingston upon Thames 2 10 0 75 74 33
Merton 3 7 0 87 67 30
Redbridge 1 0 1 61 36 16
Richmond upon Thames 2 11 4 85 198 73
Sutton 1 0 0 66 65 28
Waltham Forest 5 1 0 71 44 30
Greater London 3 2 0 103 73 30
Part A: Relationships and Risks
28
Car occupants
Car occupant injury rates for all age and ethnic groups were lower in Inner London
compared to Outer London. The injury rate ratio comparing ‘White’ children in Inner
London with ‘White’ children in Outer London was 0.79 (0.75-0.84); in ‘Black’
children the ratio was 0.92 (0.84-0.99), and in ‘Asian’ children it was 0.53 (0.47-
0.59). For adult car occupants, ratios comparing injury rates in Inner and Outer
London were 0.69 (0.68-0.70) in ‘White’ adults, 0.88 (0.86-0.90) in ‘Black’ adults,
and 0.74 (0.72-0.77) in ‘Asian’ adults.
Table 9: Average annual car occupant injury rates per 100,000 people, 1996-2006
Children (0–14) Adults
Borough ‘White’ ‘Black’ ‘Asian’ ‘White’ ‘Black’ ‘Asian’
Inner London 57 70 41 172 331 249
Camden 50 39 20 159 349 204
Hackney 57 67 37 185 300 222
Hammersmith and Fulham 23 58 53 113 273 317
Haringey 61 83 61 197 308 256
Islington 39 44 16 135 330 301
Kensington and Chelsea 36 37 51 126 342 310
Lambeth 51 79 47 169 364 328
Lewisham 100 69 75 225 304 285
Newham 94 59 48 279 253 253
Southwark 56 76 40 185 337 266
Tower Hamlets 80 100 31 204 541 194
Wandsworth 37 81 64 118 333 249
Westminster 66 108 31 194 516 291
Outer London 73 77 76 251 378 335
Barking and Dagenham 84 72 105 290 347 567
Barnet 66 79 70 263 497 325
Bexley 65 77 82 199 273 260
Brent 75 84 84 228 354 307
Bromley 68 62 27 233 385 193
Croydon 89 95 57 262 342 207
Ealing 75 52 89 240 452 412
Enfield 78 86 69 319 385 349
Greenwich 94 83 53 256 353 327
Harrow 67 50 67 221 302 212
Havering 89 119 138 311 666 598
Hillingdon 71 84 85 335 562 545
Hounslow 82 96 95 314 542 419
Kingston upon Thames 52 71 44 199 356 225
Merton 59 59 51 171 253 208
Redbridge 88 88 90 303 448 346
Richmond upon Thames 33 67 43 153 434 383
Sutton 62 27 78 207 329 246
Waltham Forest 72 62 65 211 300 348
Greater London 68 73 63 222 351 306
Part A: Relationships and Risks
29
3.3 Time
There were two components to our investigation of injury rates by ethnic group
over time: annual changes and seasonal changes.
Injury rates by year
First, we examined the relationship between ethnicity and injury rates among age
and road user group from 2001 to 2006. Population data on the number of people
in each age and ethnic group were from the GLA 2006 Round Ethnic Group
Population Projections. Injury rates shown are casualties per 100,000 people.
Figure 1: Pedestrians
Child pedestrians
W
B
A
0
50
100
150
200
2001 2002 2003 2004 2005 2006
Injury
rate
Adult pedestrians
W
B
A
0
50
100
150
200
2001 2002 2003 2004 2005 2006
Injury
rate
Figure 2: Pedal cyclists
Child pedal cyclists
W
B
A
0
10
20
30
40
2001 2002 2003 2004 2005 2006
Injury
rate
Adult pedal cyclists
W
B
A
0
10
20
30
40
2001 2002 2003 2004 2005 2006
Injury
rate
Pedestrian injury rates among children and adults in all three ethnic groups appear
to have decreased over time. A formal statistical test (not shown) indicated no
evidence for ethnic differences in the rates of decline in either children or adults.
For pedal cyclists, injury rates among ‘Black’ children steadily declined. In ‘White’
children, cycling injury rates mostly declined over the time period except for an
increase seen in 2004. Similarly, cycling injury rates in ‘Asian’ children declined
over the time period except for a small increase in 2003. Generally, cycling injury
Part A: Relationships and Risks
30
rates appeared to decline among ‘White’ adults and ‘Black’ adults over the period.
However, cycling injury rates among ‘Asian’ adults appeared to remain relatively
constant. A formal statistical test (not shown) indicated no evidence for ethnic
differences in the rates of decline in pedal cycling injury rates for either children or
adults.
Figure 3: Powered 2-wheeler
Child powered 2-wheeler
(Figure not shown as sample
sizes are too small)
Adult powered 2-wheeler
W
B
A
0
50
100
150
2001 2002 2003 2004 2005 2006
Injury
rate
Figure 4: Car occupants
Child car occupants
W
BA
0
100
200
300
400
2001 2002 2003 2004 2005 2006
Injury
rate
Adult car occupants
W
B
A
0
100
200
300
400
2001 2002 2003 2004 2005 2006
Injury
rate
Powered 2-wheeler injury rates have steadily declined among ‘White’, ‘Black’ and
‘Asian’ adults since 2001.
Car occupant injury rates were considerably higher in adults, yet rates have
decreased considerably over time in all ethnic groups. A formal statistical test (not
shown) indicated no evidence for ethnic differences in the rates of decline in car
occupant injury rates in children. However, among adults there was strong
evidence that the decline over time in car occupant injury rates was higher in
‘White’ adults than in either ‘Black’ or ‘Asian’ adults (annual reduction for ‘White’
adults 14.4% versus 10.9% in ‘Asian’ adults (p=0.001) and 13.1% in ‘Black’ adults
(p=0.013)).
Part A: Relationships and Risks
31
Seasonal analysis
The second component of our analysis of variation in injury rates over time
considered seasonal differences. A table showing injury rates by age, ethnic group,
road user group and season is included in Appendix 2. The figures below present
the average age/ethnicity-specific injury rates per 100,000 people for each season
(combining data for the period 1996-2006), for pedestrians and pedal cyclists.
Among children injured as pedestrians, the highest rates occurred in the Summer
months (June to August) among ‘Black’ children and ‘Asian’ children. However in
‘White’ children the pedestrian injury rates in the summer months remained similar
to that in Spring. A formal statistical test (not shown) provided some evidence that
the relationship between season and child pedestrian injury rates differed by
ethnicity (p-value testing no association between the interaction of ethnicity and
summer months on injury rates was p=0.085 for ‘Black’ children and p<0.001 for
‘Asian’ children).
Among adults, pedestrian injury rates in all three ethnic groups appeared higher in
the autumn months, and lower in the summer months.
Figure 5: Pedestrians
Child pedestrians
W
B
A
0
10
20
30
40
50
Winter Spring Summer Autumn
Injury
rate
Adult pedestrians
W
B
A
0
10
20
30
40
50
Winter Spring Summer Autumn
Injury
rate
Pedal cyclist injury rates increased in the summer months in ‘White’, ‘Black’ and
‘Asian’ children. Cycling injury rates appeared higher in ‘White’ children than in
‘Black’ children in every season apart for the summer, when they appeared slightly
lower. The size of difference in rates in ‘Asian’ and ‘White’ children was higher in
the summer compared to other seasons. A formal statistical test (not shown)
provided some evidence that the relationship between season and child cycling
injury rates differed by ethnicity (p-value testing no association between ethnicity
and injury rates in the summer was p=0.002 for ‘Black’ children and p=0.014 for
‘Asian’ children).
Part A: Relationships and Risks
32
Figure 6: Pedal cyclists
Child pedal cyclists
W
B
A
0
5
10
15
20
Winter Spring Summer Autumn
Injury
rate
Adult pedal cyclists
W
B
A
0
5
10
15
20
Winter Spring Summer Autumn
Injury
rate
Among adults, pedal cycling injury rates appeared highest in the summer months in
all three ethnic groups. Cycling injury rates in ‘White’ and ‘Black’ adults were more
similar over the summer months, and were substantially higher than in ‘Asian’
adults. A formal statistical test (not shown) provided strong evidence that the
relationship between season and adult cycling injury rates differed by ethnicity
(p-value testing no association between the interaction of ethnicity and summer on
injury rates was p=0.001 for ‘Black’ adults and p<0.001 for ‘Asian’ adults).
3.4 Multivariable analysis
The second part of our analysis explored the relationship between ethnicity and
road traffic injury, taking account of other variables known to influence risk, namely
levels of deprivation and the road environment in areas where people live. In a
“multi-variable” analysis, we are able to measure the effect of each explanatory
variable (e.g. level of deprivation) on injury rates in each ethnic group, whilst
controlling for the effects of other variables (e.g. number of road junctions). For our
analysis, each casualty record in STATS19 was linked to the SOA in which the
collision occurred.5
Child pedestrians
As recommended in Deprivation and Road Safety in London, use of SOA of collision
instead of SOA of residence of casualties is acceptable for analyses of child
pedestrian casualties. This is due to the close proximity of child pedestrian (and
cyclist) casualties to their home addresses. We examined whether this assumption
5 Although the SOA of residence of casualties was available, and arguably preferable, the home address
postcodes in the STATS19 data are incomplete, with levels of completeness as low as 11% in some boroughs
(see Edwards et al., 2006, Table A1, p62). Linking casualties to their SOA of residence using home address
postcode would therefore exclude a large proportion of the total casualties in the STATS19 data.
Part A: Relationships and Risks
33
is true for child pedestrians of all ethnicities. Appendix 3 summarises the
distributions of distance from place of residence to site of collision for different age,
ethnic, and road user groups. In this analysis we can confirm that ‘White’, ‘Black’,
and ‘Asian’ child pedestrians and cyclists do tend to be injured very close to home
(median distances from home for ‘White’, ‘Black’ and ‘Asian’ pedestrians are around
600 metres, confirming that they are likely to be injured in the SOA in which they
live).
Since child pedestrians are by far the most vulnerable of road user groups, we have
chosen to focus our analysis of the relationship between ethnicity, deprivation and
road traffic injury using this group. To remain consistent with our report
Deprivation and Road Safety in London we have used the age group 0–15 years in
the analyses by deprivation. A separate analysis of the relationship between
ethnicity, deprivation and child cyclist injury rates may be found in Appendix 5. The
total numbers of children in each deprivation decile injured as pedestrians between
1996 and 2006 is shown below. It may be seen that in all ethnic groups the
numbers of children injured as a pedestrian is higher in more deprived areas of
London.
Table 10: Child (0–15 years) pedestrian casualties by ethnic group and deprivation
decile (1996–2006)
Deprivation decile
1 2 3 4 5 6 7 8 9 10 Total
‘White’
673
885
944
1,097
1,118
1,162
1,260
1,422
1,240
1,414
11,215
‘Black’
62
135
180
293
459
539
636
844
943
1,311
5,402
‘Asian’
52
90
131
200
259
246
327
362
385
465
2,517
Child pedestrian injury rates per 100,000 ‘White’, ‘Black’ and ‘Asian’ children living
in each decile are presented in the figure below. For these rates, we estimated the
child population by multiplying the total numbers of children and adults living in
each SOA (from 2001 census), by the percentages of residents of all ages that are
‘White’, ‘Black’, or ‘Asian’ (also from 2001 census). These estimates of SOA-level
ethnic group populations were then scaled to ensure that our estimates of ethnic-
specific borough populations were equal to those in the census.
For ‘White’ and ‘Asian’ children, the rate of pedestrian injury increased for each
decile of deprivation. However, for ‘Black’ children, there did not appear to be any
Part A: Relationships and Risks
34
relationship between deprivation and injury – the rates of pedestrian injury were
similar across the deprivation deciles.
White
Black
Asian
0
50
100
150
200
250
12345678910
Least deprived
Deprivation deciles
Most deprived
Rate per 100,000 residents
Figure 7: Average annual ‘White’, ‘Black’ and ‘Asian’ child pedestrian
injury rates per 100,000 children (0–15), by deprivation decile of
residence
We ran two sets of multivariable models of child pedestrian injury rates. The first
model describes child pedestrian injury rates within deprivation deciles for each
ethnic group. The second model includes road environment variables to adjust for
their effects on child pedestrian injury rates within each deprivation decile. [Note–
all models are presented in detail in Appendix 4 of this report.]
‘White’ child pedestrians
In the figures below, the graph on the left shows the relationship between ‘White’
child pedestrian injury rates and increasing deprivation. The ‘Black’ diamonds
represent the injury rate ratios, comparing injury rates in each deprivation decile to
those in the least deprived decile. The vertical lines running through the ‘Black’
diamonds represent 95% confidence intervals. Where two confidence intervals
overlap, there is not enough evidence to say whether one injury rate is higher than
the other. The graph on the right shows the relationship between ‘White’ child
Part A: Relationships and Risks
35
pedestrian injury rates and deprivation after adjusting for the road environment
variables.
Figure 8
‘White’ child pedestrians ‘White’ child pedestrians
(adjusted* estimates)
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
*The road environment variables in the adjusted model included the density of junctions, A roads, and B
roads in an SOA; the length of minor roads and motorways in an SOA; and borough level estimates of
traffic speed and traffic flows. A full list of variables in the adjusted models can be found in Appendix 4.
These figures show a strong positive relationship between deprivation and
pedestrian injury rates in the ‘White’ child population. The injury rate for ‘White’
child pedestrians in the most deprived decile was more than 2.5 times the injury
rate in the least deprived decile. Compared to the least deprived areas, ‘White’ child
pedestrians in all other areas experienced significantly higher injury rates.
Furthermore, there appears to be ‘dose-response’ relationship – injury rates
increase with increasing area deprivation. Adjusting for the road environment
increased the width of confidence intervals (indicating somewhat less certainty
about true rate ratios), particularly in the more deprived deciles, however the
underlying relationship between ‘White’ child pedestrian injury and deprivation
remained.
‘Black’ child pedestrians
The figures below show ‘Black’ child pedestrian injury rate ratios by deprivation
decile. ‘Black’ child pedestrian injury rates in the most deprived areas were broadly
similar to rates in the least deprived areas. All confidence intervals include the rate
ratio of 1.0, indicating no real differences in ‘Black’ child pedestrian injury rates for
any deprivation decile compared with the least deprived. Even after adjusting for
the road environment, ‘Black’ child pedestrian injury rates demonstrated no
evidence for a relationship with different levels of deprivation.
Part A: Relationships and Risks
36
Figure 9
‘Black’ child pedestrians ‘Black’ child pedestrians
(adjusted estimates)
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
‘Asian’ child pedestrians
The figures below show ‘Asian’ child pedestrian injury rate ratios by deprivation
decile. The relationship seen between ‘Asian’ child pedestrian injury rates and
deprivation levels is similar to that observed for ‘White’ children, however the
gradient appears steeper.
Figure 10
‘Asian’ child pedestrians ‘Asian’ child pedestrians
(adjusted estimates)
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
6
7
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
Relative to least deprived areas, ‘Asian’ child pedestrian injury rates increased with
increasing deprivation, but then plateau in the most deprived areas. After adjusting
for the road environment, the injury rates in the most deprived areas were over
4 times higher than in the least deprived areas. Compared with the rate in the least
deprived decile, ‘Asian’ child pedestrian injury rates were significantly higher in all
other deprivation deciles.
Part A: Relationships and Risks
37
These analyses suggest that there are real differences in the relationships between
deprivation and child pedestrian injury by ethnic group, and that these remain after
controlling for the road environment. In a formal statistical test for interaction
between pedestrian injury risk, deprivation and ethnicity (not shown), we found
strong statistical evidence that these relationships differ by ethnic group: while the
deprivation relationship is similar for ‘Asian’ and ‘White’ children, it differs for
‘Black’ children.
Adult pedestrians
To understand more about these ethnic differences in the relationships between
pedestrian injury rates and deprivation, we next examined adult pedestrians. Some
caution is required when interpreting these results for adult pedestrian casualties,
as the use of SOA of collision instead of SOA of residence in the analysis will be less
reliable than for child pedestrian casualties (adults are injured further away from
home than are children).
The figures below show the relationship between adult pedestrian injury rates with
increasing area deprivation. Again, rate ratios are presented comparing injury rates
in tenths of London (according to deprivation) with the rate in the least deprived
tenth. The figures presented below show estimates after adjustment for the effects
of road environment variables.
Figure 11 Figure 12
‘White’ adults ‘Asian’ adults
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
A strong positive relationship may be seen between increasing levels of area
deprivation and increasing pedestrian injury rates in both ‘White’ and ‘Asian’ adults.
The injury rate for ‘White’ and ‘Asian’ adult pedestrians in the most deprived decile
was more than 3 times the injury rate in the least deprived decile.
Part A: Relationships and Risks
38
‘Black’ adults
Figure 13
1
2
3
4
5
12345678910
Injury
rate ratio
There was no good evidence for an
equivalent relationship between ‘Black’
adult pedestrian injury rates and
increasing area deprivation.
Only in the tenth most deprived areas of
London was there weak evidence that
‘Black’ adult pedestrian injury rates were
higher than in the least deprived tenth.
Least deprived Most deprived
As with child pedestrians, there is strong evidence for a relationship between
increasing deprivation and increasing pedestrian injury rates in both the ‘White’ and
‘Asian’ adult population. However, there is very little evidence for any relationship
between deprivation and pedestrian injury rates in the ‘Black’ adult population.
3.5 Exposure to risk (LATS 2001)
Ethnic differences in exposure to traffic as a pedestrian, a cyclist, a car occupant,
etc., are a potential explanation for the observed ethnic differences in road traffic
injury rates. In Deprivation and Road Safety in London an analysis of LATS 2001
data suggested that ‘Black’ children tend to take a larger percentage of their trips
as a pedestrians, compared to ‘White’ children, and that ‘Black’ adults take a
greater proportion of trips by bus (which includes walking to and from bus stops)
compared to ‘White’ adults. To add to this evidence, for this report our analysis
considered two measures of exposure: average daily time spent walking (in
minutes) and average daily distance walked (in km). We calculated averages
separately for each age, sex, and ethnic group.
The LATS 2001 data include records of daily travel for 67,252 individuals from
29,973 households in London. There was a total 176,447 trips made, comprising a
total 360,389 interchanges (parts of trips made by different travel modes). A total
of 51,427 trips were made where walking was the only mode of transport used for
the entire trip. A further 163,885 interchanges were made by foot.
To estimate average times spent walking and average distances walked for the
whole of London, LATS data must be weighted to allow for different selection
probabilities between age, sex and ethnic groups. The numerators (total numbers
Part A: Relationships and Risks
39
of minutes/Km walked by a particular age-sex-ethnic group) were weighted by
interchange-level weights. The denominators (total number of persons in each age-
sex-ethnic group) were weighted by person-level weights. All weights used were
provided by TfL.
Daily time spent walking
Among children, both male and female, ‘Black’ and ‘White’ children appear to walk
for a similar amount of time per day. ‘Asian’ males and females appear to walk
slightly less than their ‘Black’ and ‘White’ counterparts. For adults up to age 44,
‘White’ males and females seem to have higher average walking times compared to
either ‘Black’ or ‘Asian’ males and females.
Figure 14: Average daily time spent walking (minutes)
Males
A
B
W
0
5
10
15
20
25
30
5-9 10-
14 15-
24 25-
34 35-
44 45-
54 55-
64 65+
Minutes
Females
A
B
W
0
5
10
15
20
25
30
5-9 10-
14 15-
24 25-
34 35-
44 45-
54 55-
64 65+
Minutes
Age group Age group
Daily distances walked
The figures below show average daily distances walked. ‘White’ male children
appear to walk slightly further than ‘Black’ or ‘Asian’ male children. However, young
‘White’ male adults appear to walk considerably further than their ‘Black’ and
‘Asian’ counterparts. Older ‘Black’ males appear to walk further than ‘White’ or
‘Asian’ males ages 45 to 64 years old.
Among females, ‘Black’ girls appear to walk slightly further on average than their
‘White’ and ‘Asian’ counterparts. Young ‘Black’ females 15-24 and 35-44 seem to
walk considerably further than ‘White’ and ‘Asian’ females of the same ages. Across
all ages, ‘Asian’ females appear to walk shorter distances than either ‘White’ or
‘Black’ females.
Part A: Relationships and Risks
40
Figure 15: Average daily distances walked (Km)
Males
A
BW
0
1
2
3
4
5
5-9 10-
14 15-
24 25-
34 35-
44 45-
54 55-
64 65+
Km
Females
A
B
W
0
1
2
3
4
5
5-9 10-
14 15-
24 25-
34 35-
44 45-
54 55-
64 65+
Km
Age group Age group
Table 11: Average daily time walking Table 12: Average daily distances
(minutes) walked (Km)
Age group Sex ‘White’ ‘Black’ ‘Asian’ Age
group Sex ‘White’ ‘Black’ ‘Asian’
5-9 M 15 17 14
5-9 M 1.0 0.8 0.7
F 16 18 14 F 0.7 0.8 0.5
10-14 M 22 23 20
10-14 M 1.5 1.5 1.0
F 22 21 16 F 1.1 1.3 0.9
15-24 M 23 19 19
15-24 M 3.0 2.0 1.5
F 25 20 20 F 2.3 4.4 1.1
25-34 M 20 18 15
25-34 M 2.6 1.3 1.5
F 25 22 19 F 2.0 2.1 1.0
35-44 M 17 15 13
35-44 M 2.0 1.2 0.8
F 22 19 18 F 2.0 3.0 0.9
45-54 M 16 13 11
45-54 M 1.4 2.2 0.6
F 18 19 10 F 1.5 1.6 0.6
55-64 M 17 19 14
55-64 M 1.4 1.8 0.8
F 20 17 12 F 1.0 0.7 0.4
65+ M 18 20 17
65+ M 0.9 0.8 0.6
F 16 14 8 F 0.7 0.6 0.4
Daily time spent walking by household income level
To further investigate the relationship between ethnicity and exposure to traffic as
pedestrians, we stratified the analysis of time spent walking by household income
level (collected during the LATS household survey). We used three household
income groups: under £15,000 per year, between £15k and £49k, and £50k or
more per year. The minimum income group was based on household incomes below
60% of median income levels in the LATS survey. The figures below show the
average amounts of time spent walking by ‘White’, ‘Black’ and ‘Asian’ children in
households with these levels of income.
Part A: Relationships and Risks
41
Figure 16
White children
<£15k
£15-49k
£50k+
0
5
10
15
20
25
30
35
5-6 7-8 9-10 11-12 13-14
Age group
Black children
<£15k
£15-49k
£50k+
0
5
10
15
20
25
30
5-6 7-8 9-10 11-12 13-14
Age group
Asian children
<£15k
£15-49k
£50k+
0
5
10
15
20
25
5-6 7-8 9-10 11-12 13-14
Age group
Average daily time spent walking (minutes) by ‘White’,
‘Black’ and ‘Asian’ children in households on three
different income levels
Part A: Relationships and Risks
42
It may be seen that for all ethnic groups, the average amount of time spent walking
is highest in children from households on lowest incomes, and lowest in children
from households on highest incomes. [Note– walking times are not shown in some
ages of ‘Black’ and ‘Asian’ children in high income households, due to small sample
sizes in LATS survey.]
Part A: Relationships and Risks
43
4. Discussion
Principal Findings
This study has examined associations between ethnicity and road traffic injury risk
for different road user groups in London. Our analysis has covered children and
adults injured as pedestrians, pedal cyclists, powered 2-wheeler riders, and as car
occupants. Our principal findings are summarised below:
- Person groups: In the 1996-2006 period, ‘Black’ pedestrians of all age-sex
groups appeared to have higher injury rates than their ‘White’ and ‘Asian’
counterparts. ‘Asian’ pedestrians of most age-sex groups appeared to have lower
injury rates than ‘White’ and ‘Black’ pedestrians. In most age-sex groups, pedal
cycle injury rates among the ‘White’ ethnic group appeared higher than rates
among ‘Black’ and ‘Asian’ ethnic groups.
- Place: Compared to Outer London, Inner London had higher pedestrian injury
rates for all age and ethnic groups; higher pedal cycling injury rates among adults
of all ethnic groups; higher powered 2-wheeler injury rates among adults of all
ethnic groups; and lower car occupant injury rates for adults and children of all
ethnic groups.
- Time: Casualty rates among children and adults of all three ethnic groups
appear to have decreased over time for most road user groups. Car occupant injury
rates appear to have decreased more in ‘White’ adults than in either ‘Black’ or
‘Asian’ adults.
- Season: There appeared to be seasonal differences in ‘White’, ‘Black’, and
‘Asian’ child pedestrian injury rates. There was also some evidence that the
relationship between season and child cycling injury rates differ by ethnicity.
- Deprivation: There was strong evidence that the relationship between
deprivation and child pedestrian injury rates differs by ethnic group: ‘White’ and
‘Asian’ child pedestrian injury rates increase with increasing levels of deprivation,
whereas ‘Black’ child pedestrian injury rates are not related to deprivation levels.
Similar results were found for adult pedestrian injury rates.
- Exposure: On average, ‘Black’ and ‘White’ children walk similar amounts of time
each day. ‘White’ adults aged under 44 years walk longer than ‘Black’ or ‘Asian’
adults. Young ‘White’ men walk further than their ‘Black’ and ‘Asian’ counterparts,
and young ‘Black’ women walk further than their ‘White’ and ‘Asian’ counterparts.
In all ethnic groups the average amount of time spent walking was highest in
children from households on lowest incomes, and lowest in children from
households on highest incomes.
Part A: Relationships and Risks
44
Before we consider potential underlying mechanisms for these observed
differences, we will first consider the methodological issues that have a bearing on
our results.
Methodological Issues
Perhaps the most important methodological issue that must be borne in mind when
considering the results of part A of this report is that they are based on an
‘ecological analysis’. We have analysed SOAs which are small geographical
populations of around 1,500 people. Any inferences about the relative risks of road
traffic injury for different ethnic groups using different modes of travel and living in
different levels of deprivation are based on aggregates of individual data and not
the individuals themselves. The results therefore only provide evidence for
relationships between ethnicity, deprivation and road injury risk at an ecological
(i.e. population) level, and do not necessarily for hold true for all individuals living
within those areas.
Are ethnic variations in injury rates an artefact of the data?
Ethnic variations in child pedestrian injury could be an artefact of the data for three
possible reasons: (1) bias in the numerator, (2) bias in the denominator, or (3) the
population described in the numerator differs from that in the denominator.
Bias in the numerator
In the univariable analysis, the numerator (number of casualties) comes from the
STATS19 data. A number of issues could affect the accuracy of STATS19 data in
estimating numbers of casualties by ethnic group: under-reporting and under-
recording of accidents, reliability of the police-assigned ethnicity codes, and missing
data on ethnicity. Estimates of under-reporting of pedestrian injuries in the
STATS19 vary based on location, but generally fall between 25-40% (Ward et al.,
2006). In London, Ward et al. (2005, cited in Ward et al., 2006) estimate that no
more than 70% of casualties are reported to police. Under-recording occurs when
collisions are reported but do not appear in the STATS19 data, usually due to
clerical errors or latent (i.e. not yet recognised) injuries. Estimates of under-
recording in the STATS19 data are around 20% (Ward et al, 2006). If under-
reporting and under-recording disproportionately affect different ethnic groups,
estimates of ethnic variations in injury rates will be biased.
In the STATS19 data, police are responsible for assigning an ethnicity code to each
casualty. Deciding a person’s ethnic identity based on their appearance is a difficult
Part A: Relationships and Risks
45
task and some casualties may have selected a different ethnic identity, if asked.
These potential classification errors could bias ethnic-specific injury rate estimates.
Finally, 15% of casualties in the STATS19 data did not have any information
recorded on ethnicity. If casualties from one or more ethnic groups are less likely to
have been assigned an ethnicity code in STATS19, then estimates of ethnic-specific
injury rates may be biased. We examined the characteristics of the casualties that
were missing an ethnicity code and found that coding was less likely to be complete
for very young (0 to 5 years) and for older (65 years and over) casualties. Codes
were also less likely to be complete for cyclists than for pedestrian casualties.
Coding was more complete for seriously injured casualties than for casualties with
less severe injuries. Furthermore, the completion of the ethnicity code has fallen
each year from around 93% in 1996 to under 80% in 2006.
Bias in the denominator
The univariable analysis used the 2001 census estimate of the population of London
to estimate the population at risk for the entire 1996-2006 period. Population
estimates from 2001 are likely to underestimate the London population in later
years, and overestimate the population in earlier years. If the rate of population
growth differs by ethnic group, then this too could introduce bias. The multivariable
analysis estimated the numbers of ‘White’, ‘Black’, and ‘Asian’ adults and children in
each SOA by multiplying the percentage of persons in each ethnic group of all ages
to the number of adults and number of children living in each SOA. As the
percentages of persons in each ethnic group is likely to vary by age, this method
may have introduced error in our estimates of injury rates for each age-ethnic
group.
Bias from combining two sources of data
The numerator and denominator in the injury rate calculations may be describing
two different populations. First, ethnicity in the STATS19 is police-assigned, while
ethnicity in the 2001 census was self-reported. Furthermore, ethnic categories in
the STATS19 do not reliably map onto the ethnicity categories used in the
2001 census. For example, not only is it unclear who the STATS19 category ‘Arab’
is meant to represent, it is also unclear to which census category ‘Arab’ would be
mapped. The numerator in the injury rate analysis represents injuries occurring in
London, while the denominator represents the resident population of London.
Efforts were made to exclude injuries occurring to non-London residents, however
postcode of residence of casualties were missing in many cases. Therefore, the
numerator and denominator may depict slightly different populations, again
decreasing the accuracy of the injury rate estimates.
Part A: Relationships and Risks
46
Although it is difficult to prove that ethnic variations in injury rates are entirely an
artefact of the data, the potential for bias, for the reasons outlined above, must be
borne in mind when interpreting the ethnicity-specific estimates of injury rates in
this report.
Mechanisms
Assuming that the ethnic differences observed in road traffic injury rates are not
entirely artifacts of the data sources used, we now consider the mechanisms that
may explain them. In our analysis we considered two mechanisms: exposure and
deprivation.
Exposure
Our analysis of ethnic differences in exposure suggests that ‘Asian’ pedestrians
walk less than their ‘White’ and ‘Black’ counterparts, which may partially explain
the lower pedestrian injury rates in this group. The exposure analysis found little
evidence that ‘Black’ pedestrians walk more than ‘White’ or ‘Asian’ pedestrians,
suggesting that their higher pedestrian injury rates cannot be explained by
differences in the time spent walking, or in distances walked.
However, the data used to measure exposure have some notable limitations: First,
data from LATS 2001 were collected during school term-time only. Our seasonal
analysis of pedestrian injury rates found evidence for ethnic differences in child
pedestrian injury rates in the summer months. Since LATS did not collect data in
the summer, key exposure differences in these months may have been missed.
Secondly, exposure data from 2001 was extrapolated to represent the entire 1996-
2006 period. Any potential changes in walking patterns over this time period will
have been missed. Finally, not all pedestrians are injured while walking. In an
urban American study of injured children, Posner et al. (2002) found that one-third
of all pedestrian injuries occurred while playing in the road environment. Measures
of playing exposure were not available in the LATS data. A UK study examining
police data (Sentinella and Keigan, 2006), also found that ‘most’ fatal injuries to
child pedestrians aged 9-15 years occurred when the child was playing.
Deprivation
The hypothesis that deprivation explains ethnic differences in road traffic injury
rates is based on the assumption that injury rates increase with increasing
deprivation. Deprivation may confound the relationship between ethnicity and
Part A: Relationships and Risks
47
injury found in this report, if children living in more deprived areas have higher
pedestrian injury rates, and if more ‘Black’ children live in deprived areas.
Our multivariable analysis provides strong evidence that the relationship between
deprivation and pedestrian injury is modified by ethnicity: ‘White’ and ‘Asian’ child
pedestrian injury rates increase with deprivation, however, ‘Black’ child pedestrian
injury rates were not found to be related to levels of deprivation. Deprivation
cannot therefore fully explain the observed differences in ‘Black’-’White’ ethnic
group injury rates.
It is important to remember that measures of ethnicity in quantitative data are
merely proxies of a multi-faceted social construct. Furthermore, the definitions of
ethnic groups used in our analysis are imperfect and do not necessarily represent
any real communities in London. But our findings do suggest that there are
differences in road traffic injury risk by ethnicity, posing a critical question: What is
it about the complex construct of ethnicity that could be related to road traffic
injury?
Part A: Relationships and Risks
48
5. Recommendations
Following from the analyses presented within Part A Relationships and Risks, our
recommendations for future research and monitoring the relationship between road
traffic injury and ethnicity in London are as follows:
1) Analysis based on STATS19 data and area-level measures has provided a ‘broad
brush’ picture of the relationship between deprivation, ethnicity and road traffic
injury, but further research is needed to:
• Understand in detail different patterns of exposure to risk, particularly for
children, and how these relate to deprivation;
• Look at the impact of existing interventions (e.g. 20mph zones) on ethnic
inequalities.
2) To monitor trends in the relationship between road traffic injury and ethnicity,
the most useful outcome measures are rates of child pedestrian and adult
pedestrian casualties
3) Work on improving the completeness of STATS19 data should continue, with
monitoring under-reporting and recording of road traffic injury.
Part A: Relationships and Risks
49
Appendices
Part A: Relationships and Risks
50
Appendix 1
Injury rates by severity
Table 13: Average annual injury rates per 100,000 people, all transportation modes
1996-2006
Children (0–14) Adults
All injuries
(95%CI) KSI
(95%CI) All injuries
(95%CI) KSI
(95%CI)
‘White’ 234 (231-237) 41 (40-42) 479 (477-481) 72 (71-73)
‘Black’ 305 (299-312) 53 (50-56) 617 (611-622) 78 (76-80)
‘Asian’ 175 (170-180) 29 (27-31) 421 (417-426) 49 (48-51)
Table 14: Rate ratios comparing ‘Black’ and ‘Asian’ injury rates to ‘White’ injury
rates, 1996-2006
Children (0–14) Adults
All injuries
(95%CI) KSI
(95%CI) All injuries
(95%CI) KSI
(95%CI)
‘White’ - - - -
‘Black’ 1.30 (1.27-1.34) 1.29 (1.22-1.37) 1.29 (1.27-1.30) 1.08 (1.05-1.11)
‘Asian’ 0.75 (0.72-0.77) 0.72 (0.66-0.78) 0.88 (0.87-0.89) 0.69 (0.67-0.71)
Part A: Relationships and Risks
51
Appendix 2
Injury rates by season
Table 15: Average seasonal injury rates per 100,000 children,
1996-2006
Mode of transport Season ‘White’
‘Black’
‘Asian’
Pedestrian Winter 23 35 16
Spring 33 47 24
Summer 30 49 29
Autumn 30 45 23
Pedal cycle Winter 3 2 1
Spring 9 8 3
Summer 13 14 6
Autumn 8 7 3
Car Winter 16 17 16
Spring 17 17 14
Summer 17 21 18
Autumn 18 18 15
Winter 1 1 0
Powered 2-
wheeler Spring 1 1 0
Summer 1 1 0
Autumn 1 0 0
Table 16: Average seasonal injury rates per 100,000 adults,
1996-2006
Mode of transport Season ‘White’
‘Black’
‘Asian’
Pedestrian Winter 20 28 17
Spring 19 27 15
Summer 18 25 15
Autumn 22 31 18
Pedal cycle Winter 8 6 1
Spring 11 7 2
Summer 14 12 4
Autumn 13 10 3
Car Winter 54 84 75
Spring 55 86 74
Summer 53 89 75
Autumn 60 94 83
Winter 22 14 7
Powered 2-
wheeler Spring 25 17 7
Summer 27 22 8
Autumn 30 20 9
Part A: Relationships and Risks
52
Appendix 3
Table 17: Distance from home to site of collision (Km)
Mode of
transport Age
Group Ethnic
group N
Records Mean SE Median 5th
centile 95th
centile
Pedestrian 0-14 ‘White’ 3015 1.38 2.18 0.53 0.04 5.47
‘Black’ 257 2.01 3.29 0.58 0.03 8.61
‘Asian’ 1834 1.82 2.99 0.63 0.04 7.19
16+ ‘White’ 9786 3.38 4.72 1.28 0.07 13.76
‘Black’ 974 3.72 4.49 2.14 0.08 12.51
‘Asian’ 2868 3.56 4.42 1.77 0.10 12.87
Cycle 0-14 ‘White’ 838 0.89 1.47 0.42 0.05 3.17
‘Black’ 48 0.69 1.11 0.18 0.03 2.28
‘Asian’ 282 1.02 1.56 0.44 0.04 4.35
16+ ‘White’ 7543 3.73 3.48 2.71 0.31 10.68
‘Black’ 446 3.53 3.42 2.53 0.22 10.41
‘Asian’ 896 3.27 3.63 1.89 0.24 11.32
Car 0-14 ‘White’ 1981 3.32 4.04 2.05 0.18 10.70
‘Black’ 246 3.96 4.01 2.69 0.14 12.56
‘Asian’ 938 3.97 4.23 2.60 0.22 12.17
16+ ‘White’ 34784 4.38 4.84 2.75 0.22 14.11
‘Black’ 3993 4.84 4.98 3.29 0.25 14.96
‘Asian’ 11394 5.25 5.24 3.58 0.28 15.77
0-14 ‘White’ 83 2.85 3.95 1.08 0.16 11.58
Powered
2-wheeler ‘Black’ 8 4.56 3.27 4.45 0.07 9.24
‘Asian’ 23 1.77 2.10 1.12 0.11 4.87
16+ ‘White’ 18016 5.92 5.45 4.23 0.38 17.08
‘Black’ 1997 5.85 4.90 4.57 0.45 15.64
‘Asian’ 2402 5.26 5.00 3.72 0.31 15.00
All modes 0-14 ‘White’ 6314 2.06 3.14 0.93 0.06 7.97
‘Black’ 616 2.84 3.79 1.41 0.05 10.82
‘Asian’ 3351 2.42 3.45 1.06 0.06 9.29
16+ ‘White’ 77020 4.61 5.02 2.89 0.19 15.09
‘Black’ 7903 4.89 4.92 3.39 0.22 14.87
‘Asian’ 19846 4.84 5.01 3.21 0.21 15.07
Part A: Relationships and Risks
53
Appendix 4
MULTIVARIABLE MODEL: ‘WHITE’ CHILD PEDESTRIANS
Unadjusted Adjusted
Variable Name IRR SE 95% CI p IRR SE 95% CI p
Deprivation
IMD decile 2 1.47 0.10 1.29 - 1.68 <0.001 1.36 0.10 1.19 - 1.57 <0.001
IMD decile 3 1.73 0.10 1.54 - 1.95 <0.001 1.55 0.09 1.38 - 1.74 <0.001
IMD decile 4 2.14 0.18 1.81 - 2.52 <0.001 1.94 0.14 1.68 - 2.23 <0.001
IMD decile 5 2.19 0.16 1.90 - 2.53 <0.001 1.98 0.18 1.66 - 2.37 <0.001
IMD decile 6 2.22 0.17 1.92 - 2.57 <0.001 2.10 0.21 1.73 - 2.54 <0.001
IMD decile 7 2.35 0.15 2.08 - 2.66 <0.001 2.25 0.18 1.92 - 2.64 <0.001
IMD decile 8 2.67 0.17 2.36 - 3.03 <0.001 2.50 0.27 2.03 - 3.08 <0.001
IMD decile 9 2.45 0.19 2.10 - 2.85 <0.001 2.38 0.30 1.85 - 3.05 <0.001
IMD decile 10 2.76 0.22 2.36 - 3.23 <0.001 2.55 0.34 1.96 - 3.33 <0.001
Education domain 1.00 0.00 1.00 - 1.01 0.030
Barriers domain 0.98 0.00 0.97 - 0.98 <0.001
Crime domain 1.21 0.04 1.13 - 1.30 <0.001
Percentage of children
without GCSEs 1.00 0.00 1.00 - 1.00 0.201
Traffic environment
Difference between free
flowing and morning
traffic speeds 1.00 0.00 0.99 - 1.01 0.888
Junction density 0.82 0.02 0.78 - 0.86 <0.001
Density of A roads 1.01 0.00 1.01 - 1.01 <0.001
Density of B roads 1.01 0.00 1.01 - 1.01 <0.001
Density of minor roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.201
Length of minor roads in
SOA 1.00 0.00 1.00 - 1.00 0.002
Length of motorways in
SOA 1.00 0.00 1.00 - 1.00 0.007
Number of A roads 1.00 0.00 1.00 - 1.00 0.015
Number of road junctions 1.01 0.00 1.01 - 1.02 <0.001
Number of A roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.020
Traffic flow 2001 1.00 0.00 1.00 - 1.00 0.816
Number of adjacent SOAs 1.07 0.01 1.06 - 1.09 <0.001
Part A: Relationships and Risks
54
MULTIVARIABLE MODEL: ‘BLACK’ CHILD PEDESTRIANS
Unadjusted Adjusted
Variable Name IRR SE 95% CI p IRR SE 95% CI p
Deprivation
IMD decile 2 1.33 0.18 1.02 - 1.74 0.035 1.19 0.14 0.94 - 1.51 0.148
IMD decile 3 1.09 0.21 0.74 - 1.60 0.668 0.97 0.17 0.69 - 1.36 0.840
IMD decile 4 1.24 0.24 0.85 - 1.82 0.268 1.10 0.21 0.75 - 1.61 0.640
IMD decile 5 1.35 0.21 0.99 - 1.83 0.055 1.20 0.19 0.88 - 1.63 0.246
IMD decile 6 1.33 0.22 0.97 - 1.83 0.078 1.22 0.20 0.88 - 1.69 0.234
IMD decile 7 1.28 0.19 0.95 - 1.72 0.109 1.19 0.19 0.87 - 1.63 0.288
IMD decile 8 1.26 0.19 0.93 - 1.70 0.135 1.16 0.23 0.78 - 1.71 0.460
IMD decile 9 1.06 0.16 0.79 - 1.42 0.697 0.94 0.19 0.63 - 1.41 0.775
IMD decile 10 1.13 0.16 0.85 - 1.50 0.414 1.02 0.23 0.66 - 1.60 0.917
Education domain 0.99 0.00 0.99 - 1.00 0.055
Barriers domain 0.98 0.01 0.97 - 0.99 <0.001
Crime domain 1.38 0.06 1.27 - 1.50 <0.001
Percentage of children
without GCSEs 1.00 0.00 1.00 - 1.00 <0.001
Traffic environment
Difference between free
flowing and morning
traffic speeds 1.01 0.00 1.00 - 1.02 0.002
Junction density 0.86 0.04 0.78 - 0.95 0.003
Density of A roads 1.01 0.00 1.01 - 1.02 <0.001
Density of B roads 1.01 0.00 1.01 - 1.01 <0.001
Density of minor roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.845
Length of minor roads in
SOA 1.00 0.00 1.00 - 1.00 0.670
Length of motorways in
SOA 1.00 0.00 1.00 - 1.00 0.039
Number of A roads 1.00 0.00 1.00 - 1.00 <0.001
Number of road junctions 1.01 0.00 1.00 - 1.02 0.001
Number of A roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 <0.001
Traffic flow 2001 1.00 0.00 1.00 - 1.00 <0.001
Number of adjacent
SOAs 1.12 0.02 1.09 - 1.15 <0.001
Part A: Relationships and Risks
55
MULTIVARIABLE MODEL: ‘ASIAN’ CHILD PEDESTRIANS
Unadjusted Adjusted
Variable Name IRR SE 95% CI p IRR SE 95% CI p
Deprivation
IMD decile 2 1.39 0.13 1.16 - 1.67 <0.001 1.42 0.14 1.16 - 1.73 0.001
IMD decile 3 1.55 0.21 1.18 - 2.03 0.002 1.63 0.21 1.27 - 2.11 <0.001
IMD decile 4 2.23 0.36 1.62 - 3.05 <0.001 2.35 0.36 1.73 - 3.18 <0.001
IMD decile 5 2.75 0.43 2.02 - 3.74 <0.001 3.17 0.48 2.36 - 4.26 <0.001
IMD decile 6 2.54 0.40 1.87 - 3.46 <0.001 3.07 0.47 2.27 - 4.14 <0.001
IMD decile 7 3.28 0.58 2.32 - 4.62 <0.001 4.17 0.69 3.01 - 5.77 <0.001
IMD decile 8 3.26 0.51 2.39 - 4.44 <0.001 4.08 0.63 3.02 - 5.51 <0.001
IMD decile 9 3.17 0.42 2.45 - 4.10 <0.001 4.21 0.82 2.86 - 6.18 <0.001
IMD decile 10 3.02 0.49 2.20 - 4.14 <0.001 4.45 1.15 2.69 - 7.37 <0.001
Education domain 0.98 0.00 0.98 - 0.99 <0.001
Barriers domain 0.98 0.01 0.97 - 0.99 <0.001
Crime domain 1.26 0.07 1.13 - 1.40 <0.001
Percentage of children
without GCSEs 1.00 0.00 1.00 - 1.00 0.012
Traffic environment
Difference between free
flowing and morning
traffic speeds 0.98 0.01 0.97 - 0.99 0.005
Junction density 0.84 0.03 0.77 - 0.91 <0.001
Density of A roads 1.01 0.00 1.01 - 1.01 <0.001
Density of B roads 1.01 0.00 1.00 - 1.01 <0.001
Density of minor roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.317
Length of minor roads in
SOA 1.00 0.00 1.00 - 1.00 0.099
Length of motorways in
SOA 1.00 0.00 1.00 - 1.00 0.024
Number of A roads 1.00 0.00 1.00 - 1.00 0.185
Number of road junctions 1.01 0.00 1.01 - 1.02 <0.001
Number of A roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.004
Traffic flow 2001 1.00 0.00 1.00 - 1.00 0.019
Number of adjacent
SOAs 1.07 0.02 1.04 - 1.11 <0.001
Part A: Relationships and Risks
56
Appendix 5
MULTIVARIABLE MODELS: CHILD PEDAL CYCLISTS
‘White’ ‘White’ (adjusted estimates)
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
‘Black’ ‘Black’ (adjusted estimates)
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
‘Asian’ ‘Asian’ (adjusted estimates)
1
2
3
4
5
12345678910
Injury
rate ratio
1
2
3
4
5
12345678910
Injury
rate ratio
Least deprived Most deprived Least deprived Most deprived
There was some evidence that ‘White’ children living in the moderately deprived areas
experienced higher cycling injury rates compared to children in least deprived. There was no
evidence for a relationship between ‘Black’ or ‘Asian’ child cycling injury rates and
deprivation.
Part A: Relationships and Risks
57
MULTIVARIABLE MODEL: ‘WHITE’ CHILD PEDAL CYCLISTS
Unadjusted Adjusted
Variable Name IRR SE 95% CI p IRR SE 95% CI p
Deprivation
IMD decile 2 1.22 0.13 0.99 - 1.50 0.059 1.14 0.10 0.95 - 1.35 0.155
IMD decile 3 1.19 0.11 0.98 - 1.43 0.073 1.10 0.10 0.93 - 1.31 0.259
IMD decile 4 1.34 0.13 1.11 - 1.62 0.002 1.26 0.11 1.06 - 1.51 0.010
IMD decile 5 1.32 0.12 1.10 - 1.57 0.002 1.21 0.11 1.01 - 1.46 0.039
IMD decile 6 1.38 0.14 1.13 - 1.68 0.002 1.28 0.12 1.06 - 1.55 0.011
IMD decile 7 1.63 0.18 1.31 - 2.03 <0.001 1.49 0.18 1.17 - 1.88 0.001
IMD decile 8 1.38 0.16 1.11 - 1.73 0.005 1.27 0.15 1.01 - 1.60 0.037
IMD decile 9 1.06 0.11 0.86 - 1.30 0.580 1.04 0.15 0.79 - 1.38 0.763
IMD decile 10 1.07 0.09 0.91 - 1.26 0.435 0.99 0.18 0.70 - 1.42 0.967
Education domain 1.01 0.00 1.00 - 1.02 <0.001
Barriers domain 0.99 0.00 0.98 - 0.99 0.002
Crime domain 1.14 0.04 1.06 - 1.23 0.001
Percentage of children
without GCSEs
1.00 0.00 1.00 - 1.00 0.554
Traffic environment
Speed of A roads in the
morning 1.00 0.00 1.00 - 1.01 0.204
Junction density 0.90 0.04 0.82 - 0.99 0.027
Junction density squared 1.01 0.02 0.98 - 1.04 0.553
Density of A roads 1.01 0.00 1.01 - 1.01 <0.001
Number of A roads 1.00 0.00 1.00 - 1.00 0.007
Number of road junctions 1.01 0.00 1.00 - 1.01 <0.001
Number of A roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.001
Number of motorways in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.038
Number of adjacent
SOAs 1.05 0.02 1.02 - 1.08 0.002
Part A: Relationships and Risks
58
MULTIVARIABLE MODEL: ‘BLACK’ CHILD PEDAL CYCLISTS
Unadjusted Adjusted
Variable Name IRR SE 95% CI p IRR SE 95% CI p
Deprivation
IMD decile 2 1.02 0.27 0.60 - 1.72 0.947 1.02 0.28 0.60 - 1.74 0.943
IMD decile 3 1.11 0.36 0.59 - 2.09 0.749 1.17 0.40 0.60 - 2.29 0.636
IMD decile 4 0.95 0.29 0.52 - 1.73 0.864 1.04 0.32 0.57 - 1.90 0.903
IMD decile 5 1.10 0.33 0.61 - 1.96 0.752 1.28 0.40 0.69 - 2.36 0.429
IMD decile 6 0.84 0.29 0.42 - 1.66 0.612 1.07 0.39 0.52 - 2.17 0.858
IMD decile 7 0.92 0.30 0.49 - 1.74 0.802 1.26 0.44 0.63 - 2.51 0.518
IMD decile 8 0.75 0.23 0.41 - 1.36 0.341 1.10 0.38 0.56 - 2.18 0.779
IMD decile 9 0.68 0.22 0.36 - 1.28 0.234 1.09 0.40 0.53 - 2.25 0.818
IMD decile 10 0.62 0.20 0.33 - 1.16 0.131 1.08 0.44 0.48 - 2.40 0.859
Education domain 0.98 0.01 0.97 - 0.99 0.001
Barriers domain 0.97 0.01 0.95 - 0.99 0.001
Crime domain 1.31 0.08 1.16 - 1.47 <0.001
Percentage of children
without GCSEs
1.00 0.00 1.00 - 1.00 0.928
Traffic environment
Speed of A roads in the
morning 1.00 0.00 0.99 - 1.01 0.737
Junction density 1.03 0.06 0.92 - 1.15 0.588
Junction density squared 0.97 0.02 0.94 - 1.00 0.087
Density of A roads 1.01 0.00 1.00 - 1.01 <0.001
Number of A roads 1.00 0.00 1.00 - 1.00 0.365
Number of road junctions 1.01 0.00 1.00 - 1.01 <0.001
Number of A roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.047
Number of motorways in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.659
Number of adjacent
SOAs 1.08 0.02 1.04 - 1.13 <0.001
Part A: Relationships and Risks
59
MULTIVARIABLE MODEL: ‘ASIAN’ CHILD PEDAL CYCLISTS
Unadjusted Adjusted
Variable Name IRR SE 95% CI p IRR SE 95% CI p
Deprivation
IMD decile 2 1.35 0.36 0.80 - 2.29 0.264 1.19 0.31 0.72 - 1.97 0.502
IMD decile 3 1.24 0.26 0.82 - 1.86 0.307 1.00 0.24 0.63 - 1.60 0.987
IMD decile 4 1.40 0.39 0.81 - 2.41 0.226 1.05 0.31 0.59 - 1.89 0.859
IMD decile 5 1.47 0.41 0.86 - 2.53 0.162 1.03 0.31 0.57 - 1.87 0.929
IMD decile 6 1.05 0.30 0.59 - 1.85 0.872 0.67 0.23 0.34 - 1.30 0.235
IMD decile 7 1.08 0.30 0.63 - 1.88 0.773 0.66 0.26 0.31 - 1.41 0.285
IMD decile 8 1.41 0.36 0.85 - 2.34 0.180 0.80 0.31 0.38 - 1.70 0.562
IMD decile 9 1.25 0.33 0.75 - 2.08 0.399 0.68 0.25 0.33 - 1.39 0.287
IMD decile 10 0.89 0.25 0.51 - 1.55 0.681 0.41 0.24 0.13 - 1.29 0.126
Education domain 1.01 0.01 0.99 - 1.03 0.328
Barriers domain 1.00 0.01 0.98 - 1.02 0.951
Crime domain 1.26 0.12 1.04 - 1.54 0.017
Percentage of children
without GCSEs
1.00 0.00 1.00 - 1.00 0.225
Traffic environment
Speed of A roads in the
morning 1.00 0.00 0.99 - 1.01 0.730
Junction density 0.85 0.05 0.75 - 0.96 0.006
Junction density squared 1.04 0.02 0.99 - 1.08 0.139
Density of A roads 1.01 0.00 1.00 - 1.01 0.003
Number of A roads 1.00 0.00 1.00 - 1.01 0.426
Number of road junctions 1.00 0.00 1.00 - 1.01 0.318
Number of A roads in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.115
Number of motorways in
adjacent SOAs 1.00 0.00 1.00 - 1.00 0.926
Number of adjacent
SOAs 1.08 0.03 1.02 - 1.15 0.005
60
Part B: Policy and Practice
Part B: Policy & Practice
61
1. Aims
The analysis of road traffic injury data in Part A has provided a detailed picture of
how risk of injury, ethnicity and deprivation may be related statistically. As we have
noted, though, the identification of statistical relationships cannot explain why or
how ethnicity is related to injury risk, or what the policy implications of addressing
the issue might be. The aims of this part of the report are to put the findings of Part
A in policy context, and to use the knowledge of key stakeholders to suggest
possible mechanisms that could be explored in future research.
Specifically, the aims of this part of the project were to:
• Use existing data on borough professionals’ views, and additional interviews
with key stakeholders, to describe the current context in which policies to
address ethnicity are developed;
• undertake qualitative pilot work to identify potential research questions in
this area, and generate exploratory hypotheses for future studies.
2. Introduction
Given the difficulties in identifying possible mechanisms that link ethnicity and
injury risk, qualitative work is an important element in unpacking exactly how
environments, socio-economic conditions and ethnic identity might interact to
shape exposure to, and behaviour in, road environments. It also has a role to play
in understanding the relationships between ‘ethnicity’ as measured in routine data
sets and ‘ethnicity’ as it is understood as a description of a community or an
individual. Finally, qualitative research can help identify suggestions for further
research.
Assuming that the relationship between ethnicity and injury risk is not purely
artefactual, it could be explained by a number of factors, such as road
environments, relative deprivation, different levels of exposure to risk, or different
behaviours in particular environments. Explaining why ‘Black’ children in particular
seem to be at higher risk in London will require a local understanding of how these
factors inter-relate. One issue, for instance, is the apparently puzzling finding that
in an earlier study in Birmingham (Lawson and Edwards 1991) young ‘Asian’
Part B: Policy & Practice
62
pedestrians were at higher risk of injury compared with ‘White’ and ‘Black’ children,
whereas in London it appears that young ‘Black’ pedestrians are at higher risk than
either ‘Asian’ or ‘White’ children. Clearly there is nothing inherent about being
categorised as ‘White’, ‘Black’ or ‘Asian’ that puts one at higher risk, but there may
be something specific about being categorised in that way in particular places that
is associated with risk.
As Karlsen and Nazroo (2002) argue, ethnic identity (i.e. how we choose to define
ourselves or others) may be less important to health outcomes than ethnicity as
structure (i.e. those social factors that we have less control over at an individual
level, such as experiences of racial discrimination or the ways in which ethnicity
may be related to housing availability or employment opportunities). These are
likely to be variable across the country – what it means to be, say, Afro-Caribbean
in an Inner-London borough might be different to what it means in Birmingham. For
instance, different patterns of housing stock might mean that ‘being Black’ in
Birmingham has very different implications in terms of where you live and your
exposure to road danger compared with ‘being White’ or ‘being Asian’ than it might
in London.
So, although there are probably no direct effects of ethnicity on road traffic injury
risk, the interplay of ethnicity and environment does have a number of implications
for risk exposure. Some relate to structural factors, such as where people from
different ethnic communities are likely to live within London, which affects variables
such as how much traffic there is, whether there are safe places for children to
play, and what choices are available for transport. Some relate to the how ethnic
identity might shape behaviour. Here, identifying yourself within one group might
have implications for kinds of leisure activity undertaken, or transport mode
choices. A systematic investigation of these factors is outside the frame of a small-
scale project. However, discussions with some key stakeholders did generate some
suggestions for further research on potential mechanisms operating in London
which might link particular ethnic identities and experiences with injury risk, and
describe the context within which policies will be developed.
Part B: Policy & Practice
63
3. Methods
Three sources of data were drawn on for this part of the project:
• Analysis of existing data set on key stakeholders in London. For the previous
project on deprivation and road safety (Edwards et al. 2006), we
interviewed 40 borough professionals and other stakeholders in London and
reviewed 32 borough Road Safety Plans. We revisited this data to address
specifically: what boroughs are currently doing to address possible links with
ethnicity; what data they need; and what challenges they see.
• Further interviews with selected stakeholders. Seven community
organisation and London-wide agency representatives were interviewed to
identify how, if at all, the ‘problem’ of ethnicity and road traffic injury has
been framed in London, and canvass their views on potential interventions
and further research needed. The aims of these interviews were to identify
the relative importance of road safety to key groups, and to identify
potential policy and practice implications.
• Pilot work with young people and parents. A small opportunistic sample of
seven young people and three parents (from different ethnic groups) were
interviewed to explore their views on the links between their exposure to
risk, behaviour and ‘ethnicity’ as both a structural factor and identity. The
aim was not to include a representative sample of participants, but rather to
explore the feasibility of using interview data to shed light on possible
differences in exposure and behaviour, and to begin to unpack some of the
problems with the indicators for ethnicity identified above. Interviews had
three aims:
- To generate some pilot data on travel patterns for work, school and
leisure and identify how theses patterns might relate to risk
exposure;
- To get a range of views on why there might be differences between
London’s ethnic groups;
- To gather views on possible strategies for addressing inequalities in
road traffic injury.
Approval for the interview study was granted by the LSHTM Ethics Committee.
Those quoted in this report gave consent to be interviewed and to be quoted
anonymously. Some details and geographical identifiers have been changed or
removed to maintain confidentiality.
Part B: Policy & Practice
64
4. Findings
This section summarises the views of the stakeholders included in the study. These
are an essential context for considering how policies should address inequalities in
injury, as they illustrate current awareness of the issue and the ways in which the
relationships between ethnicity and risk are understood.
4.1 How important is the issue of road safety to Black and
Minority Ethnic (BAME) communities in London?
For most community organisations, parents and young people, there was low
awareness of road safety as a priority issue in general, and little awareness that it
was an issue that might be of specific concern to London’s BAME communities.
Some considered road traffic injury in general as an ‘inevitable’ risk, and therefore
not an issue that would be tied to social inequalities:
“Accidents are just part of life, aren’t they?” (Community organisation)
“I would go as far as to say that in the past you’d see road casualties being just an
acceptable hazard that people would seek to live with and I think that would go
across all communities”. (Policy maker)
Even if road traffic injury was considered to be potentially the result of social
factors, rather than something that ‘just happened’, it was felt to be a relatively low
priority compared with more pressing issues such as gun and knife crime. It was
also not, in general, seen as a specifically ‘Black’ or ethnic minority issue, and not
one that they had experienced community demands about:
“I was quite shocked to be told that it was an issue specific to the black
community” (Community Organisation)
“I wouldn’t say that I’m aware that there was any major drive from the community
around this, partially because I think the community was probably not aware that
this was an issue, it was not aware that there was an inequality”. (Policy maker)
This comment does suggest that if there was more awareness of both the
relationship between ethnicity and road traffic injury risk, and the number of
injuries on London’s roads, it might become an issue for communities to mobilise
Part B: Policy & Practice
65
around, and there was a considerable amount of willingness to raise awareness and
consult with communities through working with bodies such as Transport for
London. However, some did note that ‘working with communities’ would be a
challenge on this issue, given the difficulties of identifying exactly which
communities are at high risk, and the low levels of understanding of how and why
there is an over-representation of some groups in road injuries:
“There are so many different cultures and countries, so identifying the exact
people to work with is a daunting task” (Policy maker)
This reflected the views of road safety professionals, who were also concerned
about using the relatively crude data from STATS19 as a basis for developing
interventions. We turn now to the views of road safety professionals. If road traffic
injury was a low priority for BAME groups in London, ethnicity was a relatively low
priority for borough professionals.
4.2 Ethnicity and road traffic injury: the perspective of London
boroughs
Road Safety Plans
One indicator of the relative priority of ethnicity for road safety teams is how far it
is addressed in the borough Road Safety Plans (RSPs). However, as we found for
deprivation in general (Edwards et al. 2006: 90), RSPs focused largely on the
broader targets they had been given for reducing the numbers of casualties, rather
than on issues that might be important, but with no specific targets. A few reports
did discuss ethnicity in terms of the diversity of the population, but of the 32 RSPs
we had available only a minority addressed the implications of ethnicity for road
traffic injury:
• 5 reported specific casualty numbers by STATS19 ethnic groups, with 3 of
these reporting changes over time. One of these made a commitment to
reducing numbers for ethnic minorities.
• 3 did not report figures, but made reference specifically to the added risk for
‘Black’ pedestrians, either in their Borough or as reported by TfL. One of
these recommended more research on the issue.
• 4 reported that they had carried out an Equality Impact Assessment,
although 1 one these boroughs explicitly reported that there were no issues
to do with Race Equality that arose from road safety
Part B: Policy & Practice
66
These low overall rates of coverage do not necessarily indicate a lack of interest in
addressing ethnic inequalities. However, in a field where there are already a
number of policy priorities, they do suggest that boroughs will inevitably
concentrate resources (or at least public statements) on those issues that are the
subject of specific targets.
Views of professionals
Our previous study on deprivation and road safety in London discussed the range of
ways in which road safety teams addressed ethnicity in their work, noting that,
given the comparatively weak evidence at the time that there were ethnic
inequalities in injury risk, ‘ethnicity’ was largely seen as an issue to be taken into
account when delivering road safety interventions, rather than an issue of
inequalities that should be addressed through prioritising resource allocation
(Edwards et al. 2006:100-104). In summary, borough road safety professionals
reported that:
• Given that each borough had a unique mix of settled and more recently-
arrived communities, crude findings on ‘ethnic’ differences might not be
useful at the local level;
• Too little was known about why ethnicity might be linked to injury risk to
‘target’ interventions effectively;
• Work with local ethnic minority communities relied on good links with
community groups.
Here, we revisit the views of road safety professionals to address three particular
areas that have implications for addressing ethnic differences in injury rates: the
problems of insufficient evidence; different perspectives on how to address ethnic
differences and developing good community links.
1) The evidence base.
At the time of the original interviews (early 2006), there was little robust data
about the statistical relationships between measures of ethnicity and road traffic
injury in London available to borough professionals. Some participants in the study
had seen presentations at the Pan-London Road Safety Forum suggesting that
‘Black’ Londoners were over-represented in the injury data, and others were aware
from local studies that there may be some over-representation, but given that
there was no certainty about the relationship, much of the discussion about
potential inequalities was around the needs for data and, more importantly for
practitioners, needs for information about what they could do to address the
problem, if there was one. Further, there was no detailed evidence to suggest
Part B: Policy & Practice
67
whether ‘ethnic’ differences were ‘real’, or an artefact of either data collection
methods (see Part A of this report) or simply reflected the different levels of
deprivation across London’s population.
In the absence of any reliable data, professionals had to draw on personal
knowledge and observation, which was limited, as this practitioner noted:
“It [ethnicity] isn’t given in the normal statistics… we really just don’t know, the
only way we can get it is by feel, when you’re going to places .. but you go to
another area, and it might be completely the opposite, so it’s quite difficult to
establish” (Interview 13)
The key challenge from the perspective of borough level practitioners was the lack
of relevance of London-level data to their locality. At a borough level, there are too
few injuries (particularly serious injuries) to analyse by particular ethnic groups to
identify where there might be over-representation, but at the London level, data
are inevitably aggregated to crude ‘Black’, ‘Asian’ and ‘White’ groups derived from
STATS19 which do not relate to the specific communities defined by ethnicity,
religion, or other communalities. The specific needs, say, of recently-arrived people
from central Africa may well be different from those of the Somali community and
those of well-established Afro-Caribbean groups. Individuals from these groups are
all likely to be defined as ‘Black’ in terms of STATS19, but the risks they face are
likely to be varied, resulting from very different exposure to traffic risks, travel
patterns and risk behaviours. As one officer, who was knowledgeable about the
data suggesting an over-representation of ‘Black’ children in the injury statistics,
noted:
“It starts getting complicated of course because it is different for different ethnic
mixes – you’ve got different groupings” (Interview 11)
A further problem noted with reliance on STATS19 data for our knowledge about
ethnic differences was that, as one road safety officer put it, “that only tells us
about the visible minorities”. This officer was concerned about the potential high
rate of injuries among a large local Jewish community, which was unlikely to be
identified through London-level data derived from STATS19.
Even if professionals were aware of the issue, and considered it one they would
prioritise in their borough, a more pressing lack of evidence was that of what would
Part B: Policy & Practice
68
be effective in addressing inequalities. Similar issues were raised in terms of
addressing deprivation: simply knowing that there are inequalities does not help,
given the problems of ‘targeting’ resources (see Edwards et al 2006) and the lack of
evidence on what works to reduce inequalities. These ‘evidence gaps’ were more
profound for ethnicity than deprivation. For deprivation, which was measured at
area level, and had a step-wise relationship with injury risk, at least particular
geographical areas could be prioritised in terms of their relative deprivation (for
instance in terms of Index of Multiple Deprivation score) for engineering solutions
such as 20mph zones. ‘Ethnicity’ is a rather different measure, given that it is an
attribute of individuals (as measured in STATS19), and the only way of ‘targeting’
areas is to identify which ones have the highest proportion of ‘Black’ residents.
2) Different perspectives on addressing ethnicity
The views of borough professionals ranged from those (largely in Inner London)
who were concerned about the potential risk differences across ethnic groups but
unsure what they could do, to those (largely in Outer London boroughs) who did
not consider this to be an issue of relevance to their borough. Some did not see it
as a productive strategy to frame the issue of one as ‘ethnicity’, given the
weaknesses of the data available (see above) and the lack of knowledge about how
to address it. There was also the issue of priorities: given the multiple policy goals
they were asked to address, it was difficult to focus on ethnicity, especially as there
were few community demands for this to be at the top of the agenda. As one
noted, partner organisations such as the police and local BAME groups focused on
the high priority issues of gun and knife crime, and within their road safety teams,
the focus was on meeting the Mayor’s targets for reducing overall rates of injury
(see Edwards et al. 2006)
For most, an approach of what we described (Edwards et al. 2006) as ‘tailoring’ was
seen as most appropriate. Rather than ‘targeting’ particular communities, this
involved the careful tailoring of interventions such as educational programmes to
the needs of the recipients, and their particular needs. That would include cultural
needs associated with ethnicity, but also needs related to age, disability, or other
differences across local communities. One described this as ‘tweaking’ (Interview 9)
to the needs of particular audiences.
One area of agreement was the move away from translating educational materials
into other languages, because it was seen as not cost-effective with so many local
Part B: Policy & Practice
69
languages, and often unnecessary, as the main beneficiaries of educational
materials (children) had good English skills:
“We have spent thousands doing that, it just isn’t worthwhile” (Interview 3)
Although specific materials might be translated for newly-arrived communities, or
to publicise consultation events, in general, translation of promotional materials
was not seen as a productive method of addressing the diverse needs of local
ethnic minority communities within the borough.
3) Developing community links
Given the importance of detailed local knowledge to ‘tailoring’ interventions, good
and sustainable links with local communities were needed. For long settled
communities, this was relatively unproblematic, as many had local councillors and
officers working for the boroughs who could help identify needs and appropriately
meet them. Many of the examples of work with local communities mentioned in the
study arose from officers who were from local ethnic minority communities taking
the lead on this issue. Well-established communities with their own organisations
also had a route for asking for particular services. Examples included an Islamic
school which has requested help with road safety and a Bengali women’s groups
which had requested road safety officers provide some information for them. Some
borough staff noted that such requests were the only route to providing tailored
information:
“We would do it on request .. we have been asked to do something on Turkish
radio” (Interview 14)
If providing tailored help relied on receiving specific requests, the challenge was
clearly in working with more recently-arrived communities, or those without
organisational resources and knowledge needed to liaise with statutory authorities.
This was recognised as a challenge for involvement in consultations as well, with
the borough often having good links with those organisations it traditionally worked
with (Schools, neighbourhood groups, religious organisations) but having more
difficulty with less visible or more transient communities, or those reluctant to deal
with statutory authorities.
Part B: Policy & Practice
70
4.3 Accounting for ethnic inequalities
Our previous report (Edwards et al. 2006: 103) noted that when road safety
professionals were asked for potential explanations for ethnic differences, they were
offered tentatively. Not surprisingly, given the lack of good published evidence in
the area, those that did suggest possible explanations noted that these were
speculative, and based on common-sense or personal observation, rather than any
robust evidence. Similarly, when the community leaders and other stakeholders
interviewed in this study were asked for their views on why some groups might be
more at risk, they were drawing on personal experience, and often aware that this
experience was filtered through stereotypical assumptions about the behaviour of
both their own (given experience can only ever be partial) and other communities.
These provisional explanations were offered and are reported here, then, as
opinions, which do generate potential avenues to explore in future research, but
should not be read as ‘evidence’ about the possible explanations of the relationships
described in Part A of this report.
Structural accounts
For some, the key reasons for differences probably lay in the structural differences
between communities in London, particularly around deprivation, which was seen to
influence the dangers of the road environment and access to alternatives to playing
on the street for young people:
“in the poorer deprived areas of London you’ll find that a lot of the roads don’t
actually have … this middle part, the island” (community organisation 10)
“in some poor boroughs there isn’t a lot of options and activities for young people.
Most schools have got rid of their parks and sports centres, so many young people
in deprived areas don’t have any social activities to get on with, so most of them
are just, if you like, hanging out on the road sides because they haven’t literally got
anything to do. So that’s another practical problem really because there isn’t a
place for them to go and socialise” (Community organisation 10)
Knowledge
In the previous study, road safety professionals raised lack of knowledge about
road layouts and crossing types as one potential issue that might make some
recently-arrived communities more at risk on London’s roads:
Part B: Policy & Practice
71
“There are so many different kinds of crossings… It’s confusing for anybody, lets
alone someone who has not been used to that … so [we have] a talk which
introduces them to the road network and introduces different types of crossing”
(Interview 12)
This was a view echoed in discussions with community leaders and policy makers in
this study, but not by the young people and parents interviewed, who saw
themselves as knowledgeable about safety. Currently, the data are not detailed
enough to identify how much of the excess injuries in the ‘Black’ group are
accounted for by recently-arrived individuals.
Culture
In general, respondents were circumspect about attributing cultural differences as
explanations of risk differences, given that comments about cultural difference
drawing on ‘common-sense’ knowledge are often based on racist stereotyping
about others’ behaviour. There were a few comments that could be attributed to
these kinds of stereotypes in the data:
“a lot of people who are Gujarati speakers, their whole attitude to life is different,
they undervalue life”
“some of the Afro-Caribbean kids have no self-discipline when they are crossing the
road”
“in that community you get a lot more of children looking after children”
In general, professionals, community leaders and policy makers did not refer to
cultural differences. However, there were a few comments that suggested that
exposure to risk might be a result of differences in transport choices that might be
tied to aspects of an ethnically-defined identity. One example were the comments
reflecting on the relatively low rates of cycling in ethnic minority communities,
which did suggest the different meaning cycling as a mode of transport might have
across different groups:
“It is sometimes that people at the lower end of the economic spectrum sometimes
think that actually things like cycling is indicative of your status. So basically it’s
people can’t afford to drive that actually will cycle … and as it happens, the black
community, broadly speaking, is the poorest section of the community … I can
Part B: Policy & Practice
72
recall even walking, for example, and having people from my community saying
‘Why are you walking?’” (Community organisation)
“Ethnic minorities feel that perhaps that [cycling] may be perceived that that’s their
only method of transport, you see.. people feel they have choices and the ethnic
minorities maybe feel ‘well, people might perceive that’s my only choice”
(Community organisation)
Young people and parents were more likely to draw on cultural differences in
offering explanations for different rates of injury. Parents referred to local
differences in, for instance, whether children were accompanied on their way to
school, or in the different leisure activities that were typical of different ethnically
based peer groups (“the Afro-Caribbean boys are more likely to be on their bikes in
the park” (Parent); “I think the white kids aren’t out on the streets so much”
(Parent)).
Not perhaps surprisingly, few young people discussed their own culturally-specific
risk-taking behaviour, but they did identify cultures that might put others at risk,
which were on occasion linked to particular ethnic ‘styles’. This [‘White’] boy for
instance, discusses how, despite having similar leisure activities, there was a
certain ‘Black’ style at his school may be linked to road traffic injury:
“the style they like to follow is, is about appearing cool and part of that is never
rushing or never kind of moving out of the way for anyone else… it’s something
about the image that means that they don’t, you know, they feel almost the traffic
should stop for them, rather than they should stop for the traffic. (YP15)
One [‘Black’] girl suggests a mix of structural and cultural reasons that might
protect ‘Asian’ young people from injury:
“The Asian kids are more in their houses, because of their religion, and black and
white kids are out in the street more – the Asian kids, ‘cos there aren’t that many
of them, they might be worried about people being racist and not want to go out,
so they stay in” (YP12)
The view that young ‘Asian’ people may be less likely to be exposed on the street to
risk is echoed by this [‘Asian’] girl reflecting on her peers:
Part B: Policy & Practice
73
“It’s because Indian girls have a really strict upbringing. If we do anything wrong
we get punished for it, so it’s like we’re not going to get injured. I’m allowed out by
myself, but I see Indian girls are really less out by themselves unless they’re like
eighteen or something” (YP14)
However, in general, parents, young people and professionals stressed the
‘sameness’ of behaviour, particularly that of young people in road environments,
saying they found it hard to believe that there were ‘ethnic’ differences that might
explain different risk outcomes. One young man suggested that the problem was
more likely to be the difficulty in identifying a suitable denominator: his account of
why ‘Black’ children might have more injuries was that:
“In this area, there are like loads and loads of black kids, so of course there’s going
to be more of us knocked over” (YP11)
4.4 Young people’s transport choices: convenience, safety and
socialising
The seven young people included in this study were of course not representative of
the population of London, but their accounts of experiences travelling and
socialising do suggest some differences across peer groups that might be
productive to explore further in research as potential explanations for differential
exposure to risk.
First, it is important to note that for the young people interviewed in this study,
road safety was not a high priority. All were well aware of road safety advice (such
as advice to wear cycle helmets, to cross roads in a safe way), and could talk
knowledgably about road safety advice they had received. However, their accounts
of travel and socialising suggested that other dangers were more significant, and
other priorities more pressing. Other priorities could include getting to school on
time (a challenge in many parts of London, where many made long journeys to
school, and where bus services might be unreliable):
“ It [bus number] takes longer, then I’d be late for school and get detention”
(YP13)
“Sometimes I have to run across Padstock Road in the mornings ‘cos I see the bus
coming and you have to get on it” (YP 11)
Part B: Policy & Practice
74
“Sometimes I’m in a hurry, and sometime I’m on [local high street known for fast
moving traffic with few crossing places] I’ll just stand in the middle of the road and
make the cars stop” (YP12)
Similarly, following what planners call ‘lines of desire’ could reduce the chance of
road safety advice being followed. These are the favoured routes that we choose to
navigate roadways because they are the most obvious, even if less ‘safe’. As this
young man notes, engineering solutions need to take these ‘lines of desire’ into
account to make streets safer, and this would require a knowledge of how local
people actually move around the road environment:
“there are some, some obvious places where if you looked at a map you might not
think we don’t need to put a crossing there, but when you’re actually there it’s very
obvious that they need a crossing. It’s like that road I was talking about at the end
of the street, if you looked at it doesn’t look a particularly busy road, but a lot of
people who will use it to skip a bit of Rowbridge Hill, and it’s the only way you can
actually get from here in the direction of North End … “ (YP15)
More significant dangers were primarily those of other young people. Postcode,
school and small neighbourhood allegiances were widely referred to even by these
young people, who did not report belonging to specific gangs. Coming across those
from other schools or postcodes when alone was potentially dangerous, with a risk
of being assaulted, or mugged:
“You always see fights like kids from different schools or the same school, the kids
on the top of the bus, there’s always fights breaking out – someone gives a look, or
says something” (YP 12)
“I wouldn’t go to [neighbouring locality] – there’s all different bandanas and gangs,
and they’d know we weren’t from there, so we’d get beaten up probably” (YP11)
Asked about their main concerns in terms of keeping themselves safe, violent crime
was the main issue:
“Guns and knives, there’s been so much shooting. I don’t even know if my friends
is carrying [weapons]. Even the parents don’t know if their kids, like they’re
carrying guns or not” (YP11)
Part B: Policy & Practice
75
“It’s the gun crime and gangs and all the shootings and stabbings” (YP12)
In choosing ‘safe’ routes around London, then, the accounts of all the young people
we talked to suggested that road traffic injury was a less pressing concern than
other dangers, and one that was sometimes traded against other goals (such as
getting to school on time).
Peers were an important part of travel choices. Not surprisingly, young people
preferred to travel to school with friends, and would alter journeys to meet up or
socialise at the bus stops. Peer opinions were also a factor in risk behaviours such
as wearing cycle helmets. One had stopped cycling because his mother had banned
it unless he wore a helmet. Another said:
“ Like none of your friends wear one, so you’d feel odd, different. Some do, this one
girl, but she’s not my friend. No one does, not our friends” (YP12)
Knowledge is unlikely to be a key factor in explaining differences in risk across
ethnic groups, given that all the young people in this study knew (for instance)
about safe places to cross roads, and that cycle helmets protect you, and there was
no suggestion that this knowledge was differentially distributed across London’s
ethnic groups (although it may apply to those more recently arrived). However,
there was little direct relationship between knowledge and behaviour, and how this
knowledge gets put into practice might be different. Patterns of risk exposure
associated with socialising were also determined by peer group norms. This small
group of three ‘Black’, two ‘White and two ‘Asian’ young people is obviously not
representative of London’s population, and indeed the young people interviewed
said mostly the ethnic groups they were familiar with did ‘the same things’ in their
leisure time. There were, though, indications of differences in patterns of
socialising. For the two ‘White’6 young people included, socialising was focused on
6 Young people were asked to describe their own ethnicity using the census categories. For those who did not
tick one of the ‘White’ choices, this was a difficult task: the categories simply did not reflect the ways in which
they understood their own identities. Several comments made whilst attempting to choose a category
illustrate this:
‘What should I tick? My Dad is like from Africa, but he went there from India, and my mum is Indian and
now we all live in England’
[to friend] ‘Why are you ticking Black, not mixed? Your dad is English, isn’t he?’
‘But he doesn’t live with us, so that doesn’t count, does it?’
‘My mum is [Caribbean] and my dad is from [European background], so I’m Caribbean, but I’m English as
well’
Part B: Policy & Practice
76
visiting each others’ houses, and mainly being indoors, watching videos or playing
computer games:
“You know, we’d, we’d hang out at their houses some of the time, but I’d often
meet up with a few friends and we’d kind of criss-cross between houses, so, you
know, we just, we just kind of go where our mood takes us” (YP15)
[where do you go when you meet up with friends?]”Go round her house, listen to
music, go to the cinema, or we just hang out here [own house] in my bedroom”
(YP16)
Two of the ‘Black’ children talked about being outdoors, on the streets outside
friends’ houses, in the playground or park. When asked specifically why outside
rather than inside, one girl explained:
“‘Cos your house might be messy or something, and you can just be outside, sitting
on the wall and chatting or riding your bike or playing football with your friends”
(YP12)
Clearly this is not ‘evidence’ of different patterns of socialising and exposure, but it
does reflect an element of exposure that travel data does not necessarily pick up,
that of simply ‘hanging out’ rather than travelling, which might be more likely to
put some people at risk of road traffic injury.
All young people reported similar strategies for maximising convenience,
opportunities for socialising and safety when travelling around London. These
included drawing on detailed knowledge of local bus routes, safer places to cross
roads and avoiding known dangers (often adjacent neighbourhoods, which were
perceived as more dangerous in terms of the potential for trouble from other young
people than more distant neighbourhoods). There were a number of specific
strategies reported for minimising risk, including choosing transport modes for
particular times of day, and avoiding the more ‘dangerous’ parts of the bus if alone:
“there are some areas I’d rather walk through in the morning than walk through
kind of mid-afternoon when there’s lots of people about that I might run into.”
(YP15)
Part B: Policy & Practice
77
“If I’m on my own I sit downstairs on the bus, I don’t look, I don’t make no
comments and you don’t involve yourself” (YP12)
There were, then, indicators that transport mode choices might be influenced by
ethnicity, both directly in terms of the structural constraints that arose from where
you lived, but also in terms of ‘identity’ in that choices of transport clearly had
symbolic meanings that might be shaped by one’s own ethnic identity, as well as
how peers and the wider community might react to particular choices.
4.5 Addressing inequalities
The challenges faced by borough professionals in addressing ethnicity in road safety
work were discussed above. For community organisations, if road traffic injury was
to be on their agenda, it was most productively done as part of a broader concern
with ‘community safety’. One issue mentioned by several was the need to provide
more alternatives to the street as a space for young people to socialise, as this
addressed the problems of gang culture that were perceived as a high priority in
many parts of London, but would also possible reduce road injuries:
“We’ve actually been lobbying our current borough here, [borough], and saying
they should be doing more for young people in terms of activity centres, community
centres, youth centres and that’s a big challenge that we actually face, just trying
to get young people off the street into something a bit more” (Community
organisation)
Policy makers felt that if there was solid evidence of ethnic differences in injury
rates, then this was clearly an issue of inequalities, and targeting (in terms of
putting more resources into some communities) would be appropriate:
“we have to continue to educate and to work towards reducing the amount of
casualties that we have across the board, but I do think there is a role for
identifying and taking, where appropriate, target of action to maximise the impact”
(Policy maker)
However, there were concerns that this would have to be sensitively, and without
victim blaming those communities that were at high risk:
Part B: Policy & Practice
78
“So, this has to be done sensitively, not just because other people might actually
think it’s not a good idea, but actually there is a sense that people feel that they
are being scapegoated again.” (Policy maker)
This was a view also given by borough professionals, who were concerned that,
particularly if educational interventions were designed on the basis of stereotypical
assumptions about ‘cultural differences’, they could be merely ‘victim-blaming’ and
counter-productive. Young people also noted the danger of targeting young, ‘Black’,
boys in particular. After hearing an explanation of why the study was happening,
one said:
“Why are you saying Black people? Why is it always us black kids that is the
problem?” (YP11)
Policy makers suggested that one danger was that these sensitivities could become
an excuse not to act:
“Some [practitioners] have said that ‘Well, we don’t want to look like we’ve
stigmatised [some communities]’ It’s rubbish … but that’s what their fear is” (Policy
maker)
Part B: Policy & Practice
79
5. Discussion
This part of the report summarised the views of some key stakeholders in London
(children, road safety professionals, policy makers, community organisations).
Taken with the data presented in Part A, which described the statistical risks
associated with being ‘Black’, ‘Asian’ or ‘White’, they suggest that there are a
number of issues around ethnicity and road traffic injury that could be addressed by
both Transport for London and the London boroughs. However, decisions about
what, if anything, could be done, depend as well on political values, around both
competing priorities and the costs and benefits of focusing on ethnicity. There are
two key questions that arise from this description of the policy context of London’s
road safety programmes:
1) Should road traffic injury be addressed as an ‘ethnic’ issue?
We have confirmed that there are ethnic differences, at a crude level, in the risk of
being injured on London’s roads, which are not accounted for purely by differences
in deprivation across London’s ethnic communities (see Part A of this report). Given
the evidence that for ‘Black’ groups, there is a higher risk, there is clearly an issue
of potential inequality here, but a number of costs and benefits associated with
describing road traffic injury as an ‘ethnic’ issue.
One benefit is the scope, and willingness from community organisations, to raise
awareness across the BAME communities of road safety, given the evidence that
this is an issue that affects them disproportionately. Similarly, flagging road safety
as an issue of ‘ethnicity’ would provide leverage, and possibly further resources, for
road safety teams to work with communities that may be rather marginalised in
planning and consultations.
However, there are also some disadvantages. First, several participants in this
study noted the potential for ‘victim blaming’ those at highest risk (‘Black’ boys),
and framing this as ‘their problem’. Demands for action need to come from
communities, rather than be imposed on them by those with little understanding of
the complex mix of factors that might put them at risk. Second, if road traffic injury
is to be framed as an ‘ethnic’ issue, there is problem of how to address it, given
that, as we have argued, there is unlikely to be any direct link between ethnicity
and risk. Rather, the link is between what being defined as a particular ethnicity
means specifically in London.
Part B: Policy & Practice
80
2) How should ethnicity be addressed?
If road safety is to be addressed as an ethnic issue, policy organisations suggested
that it might be appropriate to ‘target’ resources or services at particular groups.
There are two problems here: First, is knowing which communities to target, given
the widespread recognition that STATS19 categories are not that useful in
identifying the high risk groups. The crude categories of ‘Black’, ‘Asian’ and ‘White’
will each contain a range of groups identified by communalities (for instance) of
religion, ethnic identity or nationality. The data are not detailed enough to identify
which of these groups are at relatively high risk. Data derived from STATS19 are
unlikely to be able to provide this level of detail, and more detailed local research is
unlikely to be drawing on large enough data sets to identify significant differences
between groups. It may be impossible to accurately identify those ‘communities’
that are at high risk.
Second, even if such groups could be identified statistically, and then identified as
‘real’ communities, it is difficult to know what would be targeted at them. More
education, for instance, is unlikely to helpful, given the lack of evidence that there
are any knowledge differences between ethnic groups in London. Targeting
behaviour is also problematic. If it is, for instance, the behaviour of ‘Black’ children
that is different – in that they are found more likely to be out on the street than
other groups – do we really want interventions that reduce the amount of their
active transport, when other policies are encouraging walking, cycling and outdoor
activity? More generally, we need to think carefully about what the goals of policies
are. If they are aimed at removing young people from danger, do we really want to
discourage young people from public space, and from active leisure activities?
In terms engineering solutions, there is good evidence that these reduce injury
rates (see Part B1, Edwards et al. 2006). It is, though, difficult to see how these
would be targeted at particular ethnic groups, except by prioritising those
geographic areas with higher proportions of ‘Black’ residents for traffic calming
measures. Reducing speed and volume makes London’s roads safer for everyone,
whatever their behaviour. In the longer term, this would also even out any
differentials between ethnic groups based on exposure differences. However, as a
relatively ‘upstream’ intervention, which does not obviously ‘target’ ethnic
communities, this may be politically difficult to frame as an intervention designed to
address ethnic inequalities.
Part B: Policy & Practice
81
A possible solution
There are, then, advantages in using the evidence on differences between the
ethnic ‘groups’ derived from STATS19 data for awareness raising. There was
considerable support from those 3rd sector organisations that represent BAME
groups, for using the data that ‘Black’ people are at relatively higher risk, to
mobilise BAME community organisations around road safety, as part of broader
community safety programmes. There was concern about low awareness of road
safety, compared with problems such as crime and gang culture, and the data on
higher rates within the ‘Black’ community are both a useful resource for BAME
groups wanting to advocate for community safety programmes, and a way of
generating interest in road safety among those groups.
The data are also potentially useful for borough professionals, as a route for
engaging with communities, particularly those that have been traditionally
marginalised from consultation and planning processes. However, ‘targeting’
particular communities is problematic. Instead, these data could be seen as an
opportunity to mobilise and engage multiple local communities, and taking into
account their priorities for safety, such that programmes can be carefully tailored to
local needs.
Part B: Policy & Practice
82
6. Conclusion
Policy
There is currently low awareness of road safety as an issue that affects London’s
ethnic minority communities, and some scope for raising awareness in collaboration
with community organisations. There are several challenges to addressing road
safety as an ‘ethnicity’ issue, particularly the difficulties in identifying precisely
which communities are at higher risk, and why. Explanations are likely to be
specific to London’s diverse areas, and relate to the specific mix of environmental,
social and behavioural factors that affect ethnic communities in those areas. The
most productive strategies available for professionals who are concerned about
addressing inequalities, will require sustained links with local community
organisations, both to design appropriate programmes for needs identified, and to
avoid implementing inappropriate programmes based on inadequate understanding
of why some people are at higher risk. To be effective, there is good evidence that
programmes should be designed primarily to make road environments safer (see
Edwards et al. 2006, Part B1).
Further research
The limitations of STATS19 data for deriving anything other than broad brush
pictures of the issue have been noted. However, this broad picture is a useful one
for highlighting differences, and attempts to improve data recording should
continue for monitoring purposes. These are the best data available for examining
ethnic differences, given the low rates of completion of ethnic coding on other data
sources such as hospital admission records. It is unlikely, though, that further
analysis of STATS19 data will generate more useful understanding of the
relationships between ethnicity and road traffic injury risk, and further research on
the particular links between ethnicity and road traffic injury risk will require primary
data generation.
The views of those participating in this study suggest several areas that could be
investigated further, including detailed work with young people on strategies for
keeping themselves safe on transport in London. The young people in this study
were knowledgeable about road safety, and insightful about the potential for
solutions such as better location of road crossings. Research with communities
Part B: Policy & Practice
83
(including the more recently arrived), young people and local communities
identified as potentially at high risk by borough professionals, might be productive
in identifying both road safety issues from the perspective of vulnerable road users,
and possible solutions.
Recommendations
1) The headline findings on ethnic differences in road traffic injury rates could
be used to raise awareness of the issue of road safety. There is considerable
potential for Local authority road safety teams and Transport for London to work
with both statutory partners (e.g. Equality or Diversity teams) and 3rd Sector
partners representing BAME communities to include road safety issues as part
of a broader community safety agenda.
2) Although similar rates of decline in road traffic injury rates across ethnic
groups suggest that current strategies are, in general, addressing needs across
the population, to reduce observed inequalities it will be necessary to reduce
injury rates faster in groups identified as ‘Black’. However, given the limited
knowledge we have of how exposure to risk and other variables interact to put
people at higher risk, interventions designed to address ethnic inequalities need
to be carefully designed in consultation with local communities in order to:
• Avoid ‘victim blaming’;
• Ensure that Road Safety teams understand the precise risks faced
from the perspective of those affected;
• Ensure that programmes are appropriate and tailored to community
needs.
‘Local communities’ in this context will include neighbourhood communities, but
also groups which identify themselves in terms of faith, ethnicity or other
communalities (e.g. young people).
References
84
References
Bradby H. (2003) Describing Ethnicity in Health Research. Ethnicity & Health,
8(1):5-13.
Braver, ER. (2003) Race, Hispanic origin and socioeconomic status in relation to
motor vehicle occupant death rates and risk factors among adults. Accid Anal Prev
35:295–309.
Campos-Outcalt D, Bay C, Dellapena A, Cota MK. (2003) Motor vehicle crash
fatalities by race/ethnicity in Arizona, 1990-96. Inj Prev 9:251-6.
Christie N. (1995) The high risk pedestrian: Socio-economic and environmental
factors in their accidents. Project report 117, Transport Research Laboratory.
Crowthorne: TRL.
Department for Transport (Dft) (2002). Road accident involvement of children from
ethnic minorities. Road Safety Research Report No. 19. Department for Transport:
London.
Edwards P, Green J, Roberts I, Lutchmun S (2006). Deaths from injury in children
and employment status in family: analysis of trends in class specific death rates.
BMJ 333:119-121.
Edwards P, Green J, Roberts I, Grundy C, and Lachowycz K (2006) Deprivation and
Road Safety in London: A report to the London Road Safety Unit. London: LSHTM.
Grayling T, Hallam K, Graham D, Anderson R, Glaister S (2002). Streets Ahead:
safe and liveable streets for children. London: IPPR.
Lawson S, Edwards P. (1991) The involvement of ethnic minorities in road
accidents: data from three studies of young pedestrian casualties. Traffic Eng
Control 32:12-8.
Karlsen S and Nazroo JY. (2002) Agency and structure: the impact of ethnic
identity and racism on the health of ethnic minority people Sociology of Health &
Illness 24(1); 1-20.
References
85
Savitsky B, Aharonson-Daniel L, Giveon A. (2007) Variability in pediatric injury
patterns by age and ethnic groups in Israel, Ethnicity and Health, 12(2).
Schiff M, Becker T. (1996) Trends in motor vehicle traffic fatalities among
Hispanics, non-Hispanic whites and American Indians in New Mexico, 1958-1990.
Ethn Health 1:283-91.
Sentinella J, Keigan M. (2006) Young adolescent pedestrians' and cyclists' road
deaths: analysis of police accident files. TRL Report TRL620, Crowthorne: Transport
Research Laboratory.
Stevens JA, Dellinger AM. (2002) Motor vehicle and fall related deaths among older
Americans 1990-98: sex, race, and ethnic disparities. Inj Prev 8:272-5.
Stirbu I, Kunst AE, Bos V, van Beeck EF. (2006) Injury mortality among ethnic
minority groups in the Netherlands. J Epidemiol Community Health, 60:249-55.
Transport for London (TfL) (2004). Towards the year 2010: monitoring casualties in
Greater London. London: Transport for London.
Ward H, Lyons R, Thoreau R (2006) Under-reporting of Road Casualties – Phase 1.
Road Safety Research Report No. 69. Department for Transport: London.
Ward H, Robertson S, Pedler A, Townley K. (2005) Reporting of Road Traffic
Accidents in London: Matching Police STATS19 with Hospital Accident and
Emergency Department Data. Supplementary Report for St. Thomas’ Hospital
Central London. TRL unpublished project report (Cited in Ward, et al, 2006).
Crowthorne: Transport Research Laboratory
86
