International Journal of Injury Control and Safety Promotion

Vol. 19, No. 4, December 2012, 331–339

Under-reporting of road traffic crash data in Ghana

Mohammed Salifua* and Williams Ackaahb
a
Department of Civil Engineering, Kwame Nkrumah University of Science and Technology, University Post Office, Kumasi, Ghana;
b
Traffic and Transportation Division, CSIR-Building and Road Research Institute, Kumasi, UPO Box 40, Ghana
(Received 26 February 2011; final version received 19 August 2011)

Having reliable estimates of the shortfalls in road traffic crash data is an important prerequisite for setting more
realistic targets for crash/casualty reduction programmes and for a better appreciation of the socio-economic
significance of road traffic crashes. This study was carried out to establish realistic estimates of the overall shortfall
(under-reporting) in the official crash statistics in Ghana over an eight-year period (1997–2004). Surveys were
conducted at hospitals and among drivers to generate relevant alternative data which were then matched against
records in police crash data files and the official database. Overall shortfalls came from two sources, namely, ‘non-
reporting’ and ‘under-recording’. The results show that the level of non-reporting varied significantly with the
severity of the crash from about 57% for property damage crashes through 8% for serious injury crashes to 0% for
fatal crashes. Crashes involving cyclists and motorcyclists were also substantially non-reported. Under-recording on
the other hand declined significantly over the period from an average of 37% in 1997–1998 to 27% in 2003–2004.
Thus, the official statistics of road traffic crashes in Ghana are subject to significant shortfalls that need to be
accounted for. Correction factors have therefore been suggested for adjusting the official data.
Keywords: road traffic crashes; data shortfalls; under-reporting; adjustment factors; Ghana

1. Introduction
1.1. Background
Completeness, in terms of reporting of road traffic
crashes, is a crucial and desirable attribute of any crash
database. It is required in order that the full extent and
socio-economic implication of the crash problem in
any jurisdiction may be quantified. A complete crash
database also provides an excellent basis for setting
and monitoring the progress of implementation of
realistic targets for safety interventions.
In Ghana, the Motor Traffic and Transport Unit
(MTTU) of the police service is the primary source of
detailed crash data. The relevant road traffic regula-
tions (LI 953 of 1974) make it mandatory for every
motorist to report all road traffic crashes, in which
their vehicle is involved, to the police. It is on these
data that road safety policy formulation, monitoring,
evaluation and research on road traffic injuries, as well
as road safety interventions, are ultimately based.
However, there is a problem when it comes to having a
complete picture of the magnitude of the traffic crash
problem. Notwithstanding the legal obligation on
them to do so, it is generally known that not all
motorists report crashes in which they are involved.
The general problem of ‘under-reporting’ includes
both recording deficiencies (i.e. under-recording) and
non-reporting (Jacobs, Aeron-Thomas, & Astrop,
2000). The national road traffic crash database in
Ghana is hosted by the Building and Road Research
Institute (BRRI). At the beginning of each year, staff
of the institute, on commission by the National Road
Safety Commission (NRSC), visit all police stations to
retrieve and transcribe data on all reported crashes for
the previous year. Under-recording, therefore, repre-
sents the shortfall in recovery (under-recovery) of data
on the number of crashes from police files. Non-
reporting, on the other hand, is where the policemen
are not notified at all of the occurrence of a road crash,
in which case reference is made to an estimate of the
number or proportion of crashes not reported at all.
Significant under-reporting of road traffic crashes
has been documented, even in developed countries
(Nakahara & Wakai, 2001). Thus the idea is not to
establish an ideal database without any reporting
shortfalls, but to be able to estimate the amount of
shortfall so that the necessary adjustment factors may
be derived to make the data reflect the actual level of
occurrence.

*Corresponding author. Email: salifum@hotmail.com

ISSN 1745-7300 print/ISSN 1745-7319 online

Ó 2012 Taylor & Francis
http://dx.doi.org/10.1080/17457300.2011.628752
http://www.tandfonline.com
332 M. Salifu and W. Ackaah

size was based on proportion to crash/casualty

1.2. Objective and scope of study population and it was designed to provide a precision
The objective of this study was to establish realistic level of at least +5% on the estimates at 95%
estimates of the amount of shortfall between data on confidence level and 0.5 degree of variability (Israel,
the actual occurrence of crashes on the national 1992).
network of roads and the data available on the official The entire nation was divided into three regional
database at the BRRI. belts as shown in Table 1 below.
The target population of this study comprised of all With these geographical divisions, focal locations
road traffic crashes and the casualties thereof, which (i.e. towns, cities) were selected for the field data to be
occurred in Ghana in the period 1997–2004 and 2003– collected. Tamale was selected to represent the north-
2004 for under-recording and non-reporting, ern belt, Kumasi represented the middle belt while
respectively. Accra, Tema, Sekondi/Takoradi, Nkawkaw and Ko-
foridua represented the southern belt.
The following hospitals were visited:
2. Overview of methodology
(1) Komfo Anokyed Teaching Hospital (KATH) –
2.1. Data collection for estimation of ‘non-reporting’
Kumasi
2.1.1. Sources of data (2) Kwame Nkrumah University of Science and
The data gathered for the purpose of estimating the Technology (KNUST) Hospital – Kumasi
overall level of shortfall in crash reporting consisted of (3) Effia-Nkwanta Regional Hospital – Sekondi/
the most basic information that helped in constructing Takoradi
incidence levels of crashes and casualties. Thus, one of (4) Koforidua Regional Hospital – Koforidua
the main sources of information was driver-interviews, (5) Holy Family Hospital – Nkawkaw
through which information about the occurrence of (6) Tamale Teaching Hospital – Tamale
crashes and casualty outcomes were obtained. Drivers (7) Tema General Hospital – Tema
were chosen at random at selected transport terminals (8) Korle-Bu Teaching Hospital – Accra
and car parks and interviewed. The main purpose was
to obtain alternative information about road traffic The numbers shown in Table 2 represent a pro
crashes in the years 2003 and 2004. They were asked rata breakdown of the overall minimum sample size
about the type of vehicle, collision type, location and for each regional belt and the number of records
time of crash and whether the crash was reported to which was collected. The minimum sample sizes
the police, among others. Although all crashes would were based on the numbers and proportion of
be reportable by the provisions of the road traffic recorded crashes/casualties in the study period i.e.
regulations, drivers generally are less inclined to report 2003–2004.
damage-only crashes in particular, except where there
was an intention to file insurance claims.
As the driver interviews were expected to provide Table 1. Geographical coverage of regional belts.
information essentially on non-fatal crashes, data for Regional belt Spatial extent (geographical coverage)
equivalent periods were also collected from or about
casualties attending hospital, including minor injuries Northern Upper West, Upper East and
(outpatients) and severe injuries and fatalities. Thus, Northern regions
Middle Brong Ahafo, and Ashanti regions
admissions and out-patient records of selected hospi- Southern Western, Eastern, Greater Accra,
tals for the study period were examined for the relevant Central, and Volta
information. The hospital-based records also gener-
ated alternative data for assessing the numbers of
crashes/injuries that actually occurred during the study
period. Table 2. Minimum required sample size and actual samples
Questionnaire sheets were completed for each collected by regional belts.
driver/casualty or in respect of each crash where Crashes/drivers Casualties
applicable.
Regional belt Minimum Collected Minimum Collected
Northern 18 51 24 44
2.1.2. Sampling procedure Middle 89 135 97 108
For all sources of data used in this article, a minimum Southern 293 464 279 319
Total 400 650 400 471
sample of 400 was covered nationally. This sample
 International Journal of Injury Control and Safety Promotion 333

sources is captured in the police records and what

2.2. Data collection for estimation of ‘under-recording’
For the purpose of estimating the shortfalls in crash
data due to under-recording (under-recovery), reports
of crash data collection teams of the BRRI were
interrogated. It is standard practice that such reports
will include information on both the actual number of
crashes reported to the police for any given year and
the number for which information is available and
retrieved. The police data files are manual and stored
in two forms; first there is a register of traffic incidents
which contains only an entry of the fact that a crash
occurred somewhere alongside the name of the person
reporting. The second form is the actual crash
investigation file, which contains all the information
(data) that would normally be required for crash
analysis. For a variety of reasons, it is not always the
case that a crash recorded is subsequently covered by
an investigation. In addition, even for those cases
where files have been opened and investigations carried
out, some files may not be available to the BRRI team
at the time of compilation and retrieval of data. For
these two reasons a discrepancy arises between the
number of crashes in the register and the number as
reflected in files available. The former is always greater.
The reports of the BRRI teams, which typically
capture the numbers from both sources are therefore
sufficient for extracting the level of under-recording of
road traffic crashes.
2.3. Data matching and analysis
The data gathered from the questionnaire surveys were
matched with the police records on which the official
statistics are based. A manual method of data linkage
was used based on name, age, address, date and time of
crash and place of occurrence since electronic records
were not available with both the police and hospital
data. Armed with a selected sample of crash cases,
investigators visited each police station in the study
area and manually searched the police crash registers/
files for approximate matches. As a result, it was
possible to identify what proportion of the casualty
and crash data obtained from drivers and hospital
proportion is not.
The data for the study was analysed and presented
in tabular and graphical format using Microsoft Excel
Spreadsheet.
3. Results
3.1. Non-reporting
3.1.1. Non-reporting of traffic crashes and casualties
Table 3 shows the distribution of non-reporting of
road traffic crashes and casualties for the period 2003
and 2004. It shows that the non-reporting of crashes/
casualties is significant and depends very much on the
degree of severity. Fatal crashes and fatalities were
fully (100%) reported to the police whilst a high
majority (56.9%) of damage, only crashes and slight
injuries (54.2%), went un-reported. It is also observed
that serious crashes with corresponding serious injuries
were more often reported to the police than slight
crashes. On the average, 38.6% of all crashes and
44.6% of all casualties were not known to the police.
Generally, the less severe the crash consequence, the
higher the level of non-reporting to the police. The
accuracy of crash reporting tends to decline with
decreasing injury severity.
This obviously underscores the importance of
estimating the level of non-reporting in order to get
an accurate picture of the overall crash/casualty
problem in the country.
3.1.2. Collision and road user type
The non-reporting rates for the different crash types
were also examined. Single vehicle crashes without
pedestrian (such as vehicles ran-off to ditch, hit object,
etc.) registered the highest of 47.0% followed closely
by multiple vehicle crashes (38.6%) and then single
vehicle crashes with pedestrian (17.1%).
Generally, two wheelers namely cyclists and
motorcyclists were the worst offenders of not reporting
their involvement in road traffic crashes to the police.
The level of non-reporting was very high among

Table 3. Distribution of the level of non-reporting by crash/casualty severity.

Crashes Casualties
Reporting Fatal Serious Slight Damage Total Death Serious Slight Total
Reported 28 122 99 150 399 3 165 93 261
Not reported 0 11 42 198 251 0 100 110 210
Total 28 133 141 348 650 3 265 203 471
Non-reporting (%) 0.0 8.3 29.8 56.9 38.6 0.0 37.7 54.2 44.6
95% CI (%) 0.0–10.4 3.6–13.0 22.2–37.3 51.7–62.1 34.9–42.4 0.0–86.7 31.9–43.6 47.3–61.0 40.1–49.1
334 M. Salifu and W. Ackaah
cyclists (88.2%, n ¼ 17) followed by motorcyclists
(62.2%, n ¼ 37), pedestrians (46.8%, n ¼ 126), pas-
sengers (39.6%, n ¼ 255) and drivers (33.3%, n ¼ 36)
as shown in Figure 1. In comparison, cyclists and
motorcyclists stand the chance of 2.6 and 1.9 times not
to report an injury as a result of road traffic crash to
the police than drivers of other vehicles, respectively.
Similarly, the proportion of non-reporting by pedes-
trians and passengers compared with drivers was 1.4
and 1.2, respectively.
3.1.3. Non-reporting by age groups
Table 4 shows the matched data of the police and the
hospitals for the various age groups and different
casualty injuries. The level of non-reporting of children
up to the age of 15 years was the highest, registering a
little over one half (50.6%) of the total. In the
remaining three age groups, though the levels were
less than 50%, non-reporting was still high with age
group 26–60 years having 45.9%; 16–25 years with
39.8% and the elderly (460 years old) having the
lowest of 29.4% level of non-reporting. The study
demonstrated that when police data was compared
with hospital data, higher non-reporting rates were
found among children and the working class and the
least among the elderly people.
3.1.4. Level of non-reporting by gender
Almost half (49.7%, n ¼ 179) of female road traffic
casualties were not known to the police. Male
casualties, however, have a slightly higher tendency
of reporting a crash to the police than their female
counterpart. The level of non-reporting for male
casualties was 41.4%, n ¼ 292.
Figure 1. Non-reporting by road user type from hospital
data.
3.1.5. Non-reporting by vehicle ownership
Figure 2 also presents the levels of non-reporting by
vehicle ownership of the five main ownership cate-
gories namely, commercial use, company use, hire,
private use and taxi. This is informed by data
generated from both hospital and driver surveys. Taxis
had the highest non-reporting rate of 58.3%. This was
followed by private vehicles (39.1%), commercial
(29.3%) and then company vehicles (25.0%). All
crashes involving hired vehicles were reported to the
police.
3.1.6. Factors influencing non-reporting
3.1.6.1. Time of the day. The time of day was
grouped into night (6:00 p.m.–5:59 a.m.) and day
(6:00 a.m.–5:59 p.m.). This included data from both
hospital and driver surveys. The results shown in
Table 5 indicated that night-time is generally
associated with higher non-reporting rate (47.3%)
than the daytime (35.1%).
3.1.6.2. Effect of rural vs. urban setting. The results,
collated from both hospital and driver surveys,
indicated a higher level of non-reporting of
crashes in the urban environment (40.9%,
n ¼ 369) compared to the rural environment
(35.6%, n ¼ 281). This situation may be due to the
fact that most of the crashes in the urban
environment resulted in damage-only crashes and
involved mostly taxis and private cars which were
together associated with high incidence of non-
reporting compared with most non-urban crashes
which were severe and were most likely to be
reported.
3.1.7. Other reasons for non-reporting
The study also found out from drivers the possible
reasons for not reporting a crash to the police (Table
6). The most common reason given by drivers involved
in single-vehicle crashes was that the crash did not
involve any other vehicle (67.2%). For crashes invol-
ving two or more vehicles, the drivers preferred
reaching a mutual agreement to repair their damaged
vehicles (88.0%) than to reporting the crash to the
police.
3.2. Under-recording
Data relating to under-recording have been retrieved
from crash data collection reports at the BRRI.
Analysis of the data has been carried out and presented
in Figure 3, as well as in Table 7.
 International Journal of Injury Control and Safety Promotion 335

Table 4. Age distribution of the level of non-reporting of road traffic casualties based on hospital data.*

Fatal Serious Slight All Casualties

Not Not Not Not Non- 95%
Age Reported reported Reported reported Reported reported Reported reported reporting (%) CI (%)

0–15 1 0 29 18 8 21 38 39 50.6 39.5–61.8

16–25 0 0 38 27 33 20 71 47 39.8 31.0–48.7
26–60 1 0 89 51 50 68 140 119 45.9 39.9–52.0
460 1 0 9 4 2 1 12 5 29.4 7.8–51.1
Total 3 0 165 100 93 110 261 210 44.6 40.1–49.1

Note: *Driver surveys generated data essentially on damage-only crashes and are therefore not reflected in this table.

Figure 2. Non-reporting of road traffic crashes by vehicle ownership.

Table 5. Non-reporting by time of day.

Night Day
Under- Under-
Not All reporting 95% Not All reporting 95%
Crash severity reported Reported crashes (%) CI (%) reported Reported crashes (%) CI (%)
Fatal 0 11 11 0.0 0.0–26.1 0 17 17 0.0 0.0–17.1
Hospitalised 6 43 49 12.2 3.1–21.4 5 79 84 6.0 0.9–11.0
Not hospitalised 18 15 33 54.5 37.6–71.5 24 84 108 22.2 14.4–30.1
Damage only 63 28 91 69.2 59.7–78.7 134 123 257 52.1 46.0–58.2
Total 87 97 184 47.3 40.1–54.5 163 303 466 35.0 30.6–39.3

over the period 1997–2004. The blue bars which

3.2.1. Under-recording by year
Figure 3 shows the pattern of road traffic crashes
recorded nationally every year, which in turn clearly
underscores the importance of quantifying the amount
of shortfalls (in this case due to under-recording or
under-recovery from police files) in official statistics. A
look at the number of expected number of crashes
represented by the total height of the stacked bars from
Figure 3 reveals a pronounced rollercoaster pattern
represent the actual number of crashes captured on
the national database indicate a relatively flatter
trend. Clearly, without accounting for the level
of under-recording, the interpretation of the year-
on-year changes could lead to the wrong
impression that crash numbers have been stable
over the period.
It is important to note, however, that the under-
recording shortfall has changed significantly over the
336 M. Salifu and W. Ackaah

Table 6. Reasons for non-reporting of road traffic crashes.

No. of vehicles Reasons for non-reporting No. of crashes %

1 Fear of being engaged in a lawsuit or prosecution 4 3.0
Do not possess a valid driving license at the time of crash 3 2.2
Do not know the legal obligation to report a crash 5 3.7
Non-proximity of police station to the place of crash 9 6.7
Did not involve other vehicle 90 67.2
Other 23 17.2
Total 134 100.0
2 or more Fear of being engaged in a lawsuit or prosecution 2 1.7
Do not possess a valid driving license at the time of crash 2 1.7
Do not know the legal obligation to report a crash 2 1.8
Non-proximity of police station to the place of crash 3 2.6
Mutual agreement to repair damaged vehicle(s) 103 88.0
Other 5 4.3
Total 117 100.0

Figure 3. National crash figures by status of reporting. Received – in official statistics; not received – not captured in official
statistics.

period from a high value of 32.5–42.0% in 1997–1998 3.3. Under-reporting

to 28.1–25.9% in 2003–2004. This improvement is
attributable to a variety of measures taken over time
between the police, BRRI and the National Road
Safety Commission (NRSC) to improve on crash
record keeping and early data retrieval.
3.2.2. Under-recording by region
Table 7 shows the level of shortfalls by regions. The
overall weighted average shortfalls due to under-
recording for the period 1997–2004 cover the range
14.6% for Upper East region to 47.5% for Ashanti
region. The data for the year 2004, however, show
that Upper West (29.6%), Western (28.2%), Ashanti
(28.1%) and Greater-Accra (27.8%) regions
recorded the highest shortfalls in marginally decreas-
ing order.
As stated in the objectives of this article, the under-
reporting of road traffic crashes/casualties shall be
established from the component elements of non-
reporting and under-recording.
The average national recorded cases for the study
period 2003–2004 is 73.0% (i.e. 27.0% under-record-
ing). This value was used to represent the different
severity levels of crashes and casualties. The assump-
tion here was that since the crash files were randomly
made available to the staffs who visit the police
stations, the ones which were not made available
should contain the same data.
Let
x ¼ total number of crashes or casualties
y ¼ number of crashes or casualties reported to the
police
 International Journal of Injury Control and Safety Promotion 337

Table 7. Levels of shortfall by region.

Level of shortfall (%)

Region 1997 1998 1999 2000 2001 2002 2003 2004 W. average
Ashanti 39.8 51.4 70.4 52.3 61.3 25.3 24.3 28.1 47.5
Brong-Ahafo 44.2 50.0 46.7 30.8 46.8 17.4 28.5 23.4 37.2
Central 43.1 64.8 15.6 45.8 29.3 11.6 25.0 25.9 36.1
Eastern 0.0 4.8 31.1 33.2 15.3 40.1 26.0 24.1 22.7
GT.-Accra 30.8 37.3 33.0 32.9 35.6 19.0 32.7 27.8 31.7
Northern 45.0 31.5 24.5 30.9 31.3 24.0 27.0 25.5 29.5
Upper east 16.1 26.1 10.3 11.8 15.4 10.3 13.1 17.3 14.6
Upper west 60.0 64.4 34.5 4.2 7.9 13.2 19.8 29.6 32.9
Volta 34.8 63.2 60.0 2.5 41.1 15.9 21.5 9.7 34.8
Western 22.4 13.2 11.2 28.0 27.0 20.0 26.4 28.2 22.3
Total 32.5 42.0 42.8 36.5 39.3 20.7 28.1 25.9 34.2

z ¼ number of crashes or casualties recorded reported fully. Again, there was a considerable non-
(found in official statistics) reporting for slightly (54.2%) and seriously (37.8%)
z injured casualties. This report is consistent with
Proportion recorded ¼ ð1Þ previous studies elsewhere. Harris (1990) found out
y
that the police data were not representative; the
y completeness declined according to severity of the
Proportion reported ¼ ð2Þ injuries: inpatients, about 70%; outpatients 26%.
x
Lopez, Rosman, Jelinek, Wilkes, and Sprivulis (2000)
z
y¼ ð3Þ showed that only 82% of road-related injuries in a
z=y trauma registry had matching police record. Even
though the level of shortfall between the hospital and
y
x¼ ð4Þ police data is quite substantial, the number might be
y=x even greater because some casualties do not visit the
hospital but resort to traditional treatment.
Substitute Equation (3) into Equation (4). The data collected did show that the collision type
z= with the highest level of non-reporting (47.0%) is
z=y 1 1 single vehicle collision which did not involve a
x¼ ¼z  ¼ z  k;
y=x z=y y=x pedestrian. Nakahara and Wakai (2001) found that
  out-of-court settlements are more common in vehicle-
1 1 to-vehicle collisions in which drivers’ responsibilities
where k is  ð5Þ
z=y y=x are almost equal when compared to pedestrian-vehicle
collisions in which drivers’ responsibilities are one
From Equation (5), x can be calculated since both z sided. This finding explains why the non-reporting rate
and k are known. The aggregated levels of shortfall are for multiple vehicle collision was higher than single
presented in Table 6. If k ¼ 1, then the level of vehicle collision with pedestrian (38.6% vs. 17.1%).
shortfall from the official statistics is 0. From the survey, 88% of respondents for crashes
From Table 8, 37.0% of fatal, 49.4% of serious, involving two or more vehicles indicated that they did
95.1% of slight and 217.8% of damage only crashes not report because there was mutual agreement
are under-reported. Similarly, 37.0% of fatal, 119.9% between the drivers to repair the damaged vehicles.
of serious and 199.1% of slightly injured casualties are The results also showed a high level (88.2%) of
not included in the official statistics. On the average, non-reporting among cyclists. This was consistent with
123.1% of all crashes and 147.3% of all casualties are a previous study by Hvoslef (1994), which established
under-reported. that the level of reporting of single bicycle crashes is
almost negligible. Of the 11.8% of cyclist who
reported, the collision type was either bicycle–pedes-
4. Discussion
trian or bicycle–vehicular collision. It is believed
4.1. Non-reporting strongly that the cyclists may think that whatever
There were considerable non-reporting for damage injury and damage occurs to themselves and to their
only (56.9%), slight (29.8%) and serious (8.3%) bicycles are their own trouble and do not see any
crashes. Fatal crashes and fatalities were, however, reason to report. The same reason can be assigned to
338 M. Salifu and W. Ackaah

Table 8. Aggregated level of shortfall (under-reporting) of road traffic crashes.

Crash/casualty severity % recorded % reported k-values Under-reporting (%) 95% CI (%)

Crashes
Fatal 73.0 100.0 1.3699 37.0 37.0–52.9
Serious 73.0 91.7 1.4939 49.4 42.1–57.5
Slight 73.0 70.2 1.9514 95.1 76.1–118.5
Damage only 73.0 43.1 3.1783 217.8 183.6–261.4
All crashes 73.0 61.4 2.2310 123.1 110.4–137.8
Casualties
Death 73.0 100.0 1.3699 37.0 37.0–930.0
Hospitalized 73.0 62.3 2.1988 119.9 101.2–142.9
Not-hospitalized 73.0 45.8 2.9910 199.1 159.9–251.2
All casualties 73.0 55.4 2.4727 147.3 128.7–169.1

motorcyclists who also registered a high level (62.2%)
of non-reporting. The results also indicated a higher
non-reporting rate of pedestrian (46.8%) compared to
vehicle occupants, i.e. passengers (39.6%) and drivers
(33.3%). More scrutiny into the high non-reporting
rate for pedestrian showed that, 79.4% of slightly
injured casualties did not report as against 35.6% of
seriously injured casualties confirming the decline in
reporting with decreasing injury severity (Nakahara &
Wakai, 2001).
The study established higher non-reporting rates
for police data related to children when compared with
hospital data. One reason why injuries to children are
less reported may be because parents consider them as
minors and so they do not give statements to the police
at all.
Overall, male fatalities from 1991 to 2004 continue
to outnumber their female counterparts by an approx-
imate ratio of 3:1 (Salifu, Mosi, Addae-Bofah, &
Larbi, 2005). Against the background that males
represent only 49% of the national population (Ghana
Statistical Service, 2002), it is clear that males are
involved in more serious crashes and, therefore, report
to the police than their female counterparts.
There are many other reasons for drivers and
casualties not reporting a crash to the police. The most
frequent ones are: did not involve other vehicles,
mutual agreement to repair damaged vehicle/s, do not
know the legal obligation to report a crash and non-
proximity of police station to the place of crash.
4.2. Under-recording
It can be deduced from Figure 3 that without
accounting for the shortfalls due to under-recording
the official statistics suggest a relative stabilisation in
the crash trends over the period. Clearly, this is far
from the real picture and a grossly misleading
impression. This buttresses the need to have realistic
estimates of overall shortfalls since that can
significantly distort the overall time-series trends in
crashes figures and subsequently mislead efforts at
setting and monitoring implementation of realistic
targets for crash intervention.
The national trend of shortfalls in data recording
was generally downwards, although still at substantial
levels, starting from 32.5–42% in 1997–1998 to 28–
26% in 2003–2004. The study indicated a clear and
reassuring downward trend in the level of under-
recording. This may be due in large measure to the
promptness with which crash data is recovered from
police files by the years. Until 1997/1998, most crash
data were retrieved as backlogs of two to three years.
These days, however, data for any particular year are
promptly and regularly retrieved within the first
quarter of the ensuing year. When the data is recovered
relatively quickly, shortfalls due to filing losses in
particular are minimised.
The police also give many reasons for their failure
to release some crash files for recording. The most
frequent ones are: the file has been forwarded to the
attorney’s general department for advice, the officer
handling the case is on transfer, leave or gone on peace
keeping operation. It is suspected that some of the
police officers give these reasons in order to avoid the
extra work of getting into their archive to bring out the
files for recording.
5. Conclusions and recommendations
The levels of under-reporting of road traffic crashes in
Ghana have been estimated and these came from two
sources, namely, ‘non-reporting’ and ‘under-record-
ing’. There is a significant under-reporting in police
crash data compared to hospital data and the level
varied according to the crash/casualty severity. Gen-
erally, the less severe the crash consequences, the
higher the level of non-reporting to the police.
Collision type, road user type, age and sex of casualty,
vehicle ownership, time of crash occurrence, road
 International Journal of Injury Control and Safety Promotion
environment, region, time of retrieval of data from the
police were all identified to have profound influence on
the level of under-reporting. Currently, official road
traffic fatality figures in Ghana require an adjustment
factor of 1.37 to take account of the shortfall in under-
reporting.
Police crash reports still give the most reliable and
detailed information about the crashes/casualties and
the circumstances surrounding their occurrence com-
pared to hospital data. Hospital data can on the other
hand give supplementary information. Improving the
police crash report system and setting up a more
comprehensive trauma registry in the hospitals for
crash victims will help give a true picture of the crash/
casualty situation in the country. This will help
improve the road traffic crash data for evaluation of
preventive measures and for determination of econom-
ic costs of crashes to the nation.
There is the need to educate motorists on the
relevant road traffic regulations (LI 953 of 1974) which
make it mandatory to report all road traffic crashes
involving their vehicles to the police. Such education
will lead to a reduction in the level of under-reporting
of road traffic crashes in the country.
The police should be assisted to improve their crash
filling systems. Additionally, early retrieval of crash
data by the BRRI from the police for computer storage
339
and processing is highly needed to enhance the current
police crash data in Ghana.
References
Ghana Statistical Services. (2002). 2000 population and
housing census (Summary Report of Final Results).
Accra, Ghana: Ghana Statistical Service.
Harris. S. (1990). The real number of road traffic accident
casualties in the Netherlands. A year-long survey.
Accident Analysis and Prevention, 2, 371–378.
Hvoslef, H. (1994). Operational committee of IRTAD and
Norwegian public roads administration (Special). Oslo,
Norway: Operational committee of IRTAD and Norwe-
gian public roads administration.
Israel, G.D. 1992. Determining sample size. Fact Sheet
PEOD-6. Tampa, FL: Florida Cooperative Extension
Service, Institute of Food and Agricultural Sciences,
University of Florida.
Jacobs, G., Aeron-Thomas, A., & Astrop, A. (2000).
Estimating global road fatalities (TRL Report 445).
Crawthorne, Berkshire: Transport Research Laboratory.
Lopez, D.G., Rosman, D.L., Jelinek, G.A., Wilkes, G.J., &
Sprivulis, P.C. (2000). Complementing police road-crash
records with trauma registry data – An initial evaluation.
Accident Analysis and Prevention, 32, 771–777.
Nakahara, S., & Wakai, S. (2001). Under-reporting of traffic
injuries involving children in Japan. Injury prevention, 7,
242–244.
Salifu, M., Mosi, I., Addae-Bofah, K., & Larbi, J. (2005).
Road traffic accidents in Ghana, statistics 2004. Kumasi,
Ghana: Building and Road Research Institute.
Copyright of International Journal of Injury Control & Safety Promotion is the property of Taylor & Francis Ltd
and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for individual use.