Page |1

Traffic Risk Analysis of Tamil Nadu district
Using RADM System and Index Creation for safety estimate

Submitted by
Namit Jain (2011CE10371)

A report of CED 412 - Project Part II submitted in partial fulfillment of the requirements of the degree of
Bachelor of Technology

Department of Civil Engineering
Indian Institute of Technology Delhi
May, 2015

Page |2

Declaration
“I do certify that this report explains the work carried out by me in the Course CED 412 Project Part II under the overall supervision of Dr. Geetam Tiwari and Dr. Dinesh Mohan.
The contents of the report including text, figures, tables, computer programs, etc. have not been reproduced from other sources such as books, journals, reports, manuals, websites, etc.
Wherever limited reproduction from another source had been made, the source had been duly acknowledged at that point and also listed in the References.”

Namit Jain
2011CE10371

Page |3

Certificate

“This is to certify that the report submitted by Namit Jain describes the work carried out by him in the Course CED 412 - Project Part II under my/our overall supervision.”

Dr. Geetam Tiwari
Professor
Department of Civil Engineering
Indian Institute of Technology
New Delhi – 110016
May 2015

Dr. Dinesh Mohan
Professor
Department of Civil Engineering
Indian Institute of Technology
New Delhi – 110016
May 2015

Page |4

Acknowledgement
I take this opportunity to express my profound gratitude and deep regards to my supervisors Dr.
Geetam Tiwari and Dr. Dinesh Mohan for their exemplary guidance, monitoring and constant encouragement throughout the course of this project. The blessing, help and guidance given by them from time to time shall carry me a long way in the journey of life on which I am about to embark.
I also take this opportunity to express a deep sense of gratitude towards my partner Aakash
Budhiraja, for his cordial support, valuable information and guidance, which helped me in completing this task through various stages.

New Delhi
May 2015

Namit Jain
2011CE10371

Page |5

Contents
Figures ........................................................................................................................................................... 7
Tables ............................................................................................................................................................ 8
ABSTRACT...................................................................................................................................................... 9
INTRODUCTION ...................................................................................................................................... 10
Road Safety Improvement ...................................................................................................................... 10
Establishment of Traffic Accident Database ............................................................................................... 11
Designing and Implementing a New System .......................................................................................... 11
Road Accident Data Management System ................................................................................................. 13
Impacts.................................................................................................................................................... 14
Brief Introduction to the System ............................................................................................................ 15
How is Data Recorded ............................................................................................................................. 15
Accessing the Portal ................................................................................................................................ 17
Analysis Options .......................................................................................................................................... 18
Kilometer Analysis................................................................................................................................... 20
The Process Flow................................................................................................................................. 20
The Limitations and Issues associated with the Analysis .................................................................... 21
Grid Analysis............................................................................................................................................ 21
The Process Flow................................................................................................................................. 22
Working of the System........................................................................................................................ 22
Cluster Analysis ....................................................................................................................................... 23
Monitor Site Analysis: ............................................................................................................................. 24
Corridor Analysis: .................................................................................................................................... 24
Safety Benefit Evaluation: ....................................................................................................................... 24
Dynamic Reports ..................................................................................................................................... 24
Query Builder .......................................................................................................................................... 25
DATA EXTRACTION: ..................................................................................................................................... 27
Approach 1: ............................................................................................................................................. 27
Approach 2: ............................................................................................................................................. 27
Approach 3: ......................................................................................................................................... 27
RESULT & ANALYSIS .................................................................................................................................... 27
Traffic fatalities: Total Pedestrian fatality in various districts ................................................................ 28
Traffic fatalities: Pedestrian fatality vis-à-vis literacy rate of he district ................................................ 29

Page |6

Traffic Fatalities: Pedestrian fatality among various age groups in different states .............................. 29
Traffic fatalities: Pedestrian fatality associated with each impact vehicle in different states ............... 32
Field Visit to Tamil Nadu ............................................................................................................................. 35
Primary Recording of the Data................................................................................................................ 40
Filling the ARF ......................................................................................................................................... 42
Role of the Admin Cell ............................................................................................................................ 44
Other Concerns ....................................................................................................................................... 44
Some Observations Pertinent to the System .......................................................................................... 45
Road Safety Index ....................................................................................................................................... 46
Variables ................................................................................................................................................. 48
References .................................................................................................................................................. 56
Appendix ..................................................................................................................................................... 57
Table of Minimum Data Elements .......................................................................................................... 57
Crash Data Elements ........................................................................................................................... 57
Crash Data Elements derived from collected data ............................................................................. 60
Road Data Elements ............................................................................................................................ 60
Vehicle Data Elements ........................................................................................................................ 62
Framework for Index Creation ................................................................................................................ 67
Theoretical framework ....................................................................................................................... 67
Variable selection................................................................................................................................ 68
Imputation of missing data ................................................................................................................. 68
Multivariate analysis ........................................................................................................................... 68
Normalization...................................................................................................................................... 68
Weighing ............................................................................................................................................. 69
Aggregation ......................................................................................................................................... 70
Calculation of RSI ........................................................................................................................................ 70

Page |7

Figures
Figure 1: Percentage distribution of deaths ................................................................................................. 9
Figure 2: Road safety strategy .................................................................................................................... 11
Figure 3: Sample ARF .................................................................................................................................. 15
Figure 4: Sample ARF .................................................................................................................................. 16
Figure 5: Sample ARF .................................................................................................................................. 17
Figure 6: Login Screen ................................................................................................................................. 17
Figure 7: Welcome Screen .......................................................................................................................... 18
Figure 8: Home Screen ................................................................................................................................ 18
Figure 9: Home Screen Map Display Options ............................................................................................. 19
Figure 10: Managing ARFs........................................................................................................................... 19
Figure 11: Kilometer Analysis...................................................................................................................... 20
Figure 12: Grid Analysis............................................................................................................................... 21
Figure 13: Grid Analysis............................................................................................................................... 23
Figure 14: Cluster Analysis .......................................................................................................................... 24
Figure 15: Dynamic Reports ........................................................................................................................ 25
Figure 16: Dynamic Reports ........................................................................................................................ 25
Figure 17: The Query Builder ...................................................................................................................... 26
Figure 18: Results from Query Run ............................................................................................................. 26
Figure 19: Pedestrian fatalities across districts .......................................................................................... 28
Figure 20: Fatalities vis-a-vis literacy rate ................................................................................................... 29
Figure 21: Pedestrian causalities in different age groups vs. districts ........................................................ 30
Figure 22: Pedestrian causalities’ in different age groups vs. Tamil Nadu districts (except Chennai) ....... 31
Figure 23: Percentage contribution of different age groups in pedestrian causality in different states ... 31
Figure 24: Percentage contribution of vehicles in each district ................................................................. 32
Figure 25: Fatality variation among states due to motorcycle ................................................................... 33
Figure 26: Fatality variation among states due to ambulance ................................................................... 34
Figure 27: Fatality variation among states due to LMC .............................................................................. 34
Figure 28: Fatality variation among states due to all vehicles .................................................................... 35
Figure 29: Police headquarter visited during field trip ............................................................................... 36
Figure 30: RADMS data flow ....................................................................................................................... 37
Figure 31: FIR sample .................................................................................................................................. 38
Figure 32: Collision sketch in FIR................................................................................................................. 39
Figure 33: Sample FIR ................................................................................................................................. 40
Figure 34: Proposed FIR form ..................................................................................................................... 41
Figure 35: Proposed FIR form (continued) ................................................................................................. 42
Figure 36: Pros and Cons of the complex index .......................................................................................... 47
Figure 37: Proposed Government Index ..................................................................................................... 50

Page |8

Tables
Table 1: Weightages calculated for each contributory factor .................................................................... 53
Table 2: Ranking of states on three parameters ........................................................................................ 54
Table 3: Variation of ranks under various parameters ............................................................................... 55
Table 4: Index Values for each state .......................................................................................................... 72
Table 5: Data related to causes mentioned ................................................................................................ 73
Table 6: Data related to causes mentioned ............................................................................................... 74

Page |9

ABSTRACT
India’s highways are increasingly becoming killing fields. The country has one of the highest rates of road accidents in the world with about 500,000 accidents a year. Some 130,000 people lose their lives each year, and another 500,000 or so are injured. In fact, more lives are lost in road accidents in India than in epidemics, natural calamities or wars put together. Within India,
Tamil Nadu has recorded one of the fastest increases in the number of vehicles on the roads and the maximum number of fatalities in the country. About 150 accidents take place per day on average, claiming around two lives every hour. Considering the latest statistics, total deaths in
Tamil Nadu on the road for the year 2013-2014 were nearly 15,400. Of this number, 3972 were pedestrian deaths, 4492 were passenger deaths and the remaining were driver deaths.

Deaths in Tamil Nadu (2013-2014)
26%
45%

29%

Pedestrian

Passengers

Drivers

Figure 1: Percentage distribution of deaths

In 1999, about 750000 people were killed globally in road accidents. Of this, a staggering
640,000, which is roughly 85% were from emerging nations. In fact, by 2020 Road traffic accidents will become the 3rd ranked cause in terms of years lost (by death or disability).
(Baguley, 2001) This pushes more pressure on governments around the world to invest more in road safety.
Also, starting from 1980, the number of road fatalities has increase to 65% till 1995 in Asia, whereas it has dipped by roughly 20% in the same time period for Developed countries. Though the traffic in these countries has not risen dramatically, but the number of vehicles on the road have been increasing year on year. The reason for the dip in fatalities can be attributed to the efforts put by the government in road safety.

P a g e | 10

INTRODUCTION
Road accidents is a commonly accepted as “ a rare, random, multi factor event which is always preceded by a situation in which one more road users have failed to cope with the environment”.
First, we make an attempt to interpret this definition piece by piece.
Accidents are rare events in terms of the passage of time, and the composition and flow of traffic on the road network. However, cumulatively, these rare events add up to the nuisance of Road
Hazard, which is an issue of increasing concern for governments globally.
In most of the cases, accidents are found to be multi-factor, e.g. rain, darkness, only 1 headlamp working, partially obscured visibility, driving having consumed intoxicants, etc. may all be factors which could have rendered the driver unable to cope with the situation
Additionally, these are random events, since one cannot really predict when one will occur.
However, they are not entirely random, otherwise no measures could help in curtailing them.
Previous research has indicated that accidents tend to cluster at particular points in the transport network or among particular road user types, implying a non-random component in the occurrence of an accident.

Road Safety Improvement
When thinking of methods to enhance road safety, caution has to be maintained appropriate measures are adopted by taking into account the on ground realities. Accident preventive measures that have worked in developed countries might or might not work in developing countries because of differences prevailing between the nations.
3 E’s of road safety, Education, Enforcement and Engineering are well known. A 4th E has been recently added, and it is Encouragement. This is generally regarded as the role of the governments in creating a culture to promote road safety by setting up the best institutional framework and links for coordinating action, and making regulations to enforce efficient safety management, and to back it with adequate funding.
The following figure depicts a framework for a road safety strategy

P a g e | 11

Figure 2: Road safety strategy

The figure above shows 2 ways in which road safety can be improved by the contributions and actions various stakeholders, namely accident prevention {ensuring roads standard are adhered to, drivers are well trained, etc.) or accident reduction {tackling problems on the existing system, spot checks of vehicles/drivers, emergency services, etc.}. It also depicts that fact that the accident database is at the heart of planning for improvement across all sectors and should be used ultimately as the fundamental measure in evaluating how effective various actions that have been taken would be.

Establishment of Traffic Accident Database
The accident database serves as functionality across levels. At macro or national level, it is the tool of the central government to aid them in creating a safety policy. At a regional level, it can be research aid to help develop local police campaigns, say on drunk driving, or child safety education. Next, it can be used at the local level as an aid for engineers for road design and to eliminate blackspots. (Should ideal be placed below the para below it)
(Baguley, 2001)

Designing and Implementing a New System
This section describes the steps to take to create a road crash data system from scratch. The steps are in a sequence, but they are not mutually exclusive and may not be followed in the exact order. Step 1: Establish a Working Group
Step 2: Managing Data Quality Issues
A data system is as robust and reliable as its input is. The data quality problems need to be identified and addressed at the earliest.
Step 3: Select and define minimum data elements

P a g e | 12

A balance needs to be achieved between data that is absolutely necessary, that is desirable and data that is feasible to collect. The following is a list of the minimum data elements with their descriptions Step 4: Defining Data Capture Procedures
Data capture process describes the process used to collate crash data and to then transfer it to the crash database system. Primary data collection is done by the competent authority on site(typically a police officer). Data entry in the system is typically to be managed by another police employee.
Step 5: Identifying System requirements and resources
Before the design is pursued further, it is important to state the requirements of the system and the human and financial resources available for its development and implementation.
Step 6: Choosing the data collection tools
The tools range from simple paper questionnaires, to sophisticated electronic mobile devices which can deliver data in real time. Typically, data collection has been lacking structure and consistency and is proven to subjectivity. Hence, it is essential for these tools to record the parameters that have been suggested previously. Also, these tools should be developed in consultation with the final end users of these tools. They first need to be tested for effectiveness and ease of use. They should be simple, yet be able to deliver precise data in useable form. For this, the minimum data elements can be coded to numerical values against strings (in the cases they are strings) to hasten the data collection process, reduce the scope for errors and obtain data which is easily quantifiable and analyzable.
Step 7: Creating a project timeline
The next step is to develop a time bound action plan for development, testing and implementation of the system. The implementation process should be iterative in nature, where in the system is introduced, then evaluated and then updated as per the observed feedback.
Step 8: Developing a data management plan
The data management plan documents how the system will function, including delegation of responsibilities to various staff and agencies as required.
Step 9: Implementation
The entire process can be summarized as follows

P a g e | 13

I

•Estabish a Working Group of Key Stakeholders
•Develop a long term Strategy and Short Term Plan

2

•Identify Data Quality Issues
•Take Appropriate Steps as necessary

3

•Select and Define Minium Data Elements
•Code them for easier Use
•Define data capture processes

4

5

•Identification of System Requiremens
•Identification of Resources Needed for Development

6

•Choosing Data collection tools. Keep them simple and effective
•Invole Police as stakeholders, since they are end users of the same

7

8

9

•Create Project Deadline

•Create Data Management Plan

•Implementation
•It is an iterative prcocess, where each design is evaluated and then improved upon

Road Accident Data Management System
In April 2007, Tamil Nadu became the first state in the country to announce a Road Safety
Policy. This was followed, in 2009, by a Road Safety Action Plan. As part of the Action Plan, an easy-to-use bilingual software package - known as the Road Accident Data Management System
(RADMS) - was developed, with the help of an international consultant, under the World Banksupported Tamil Nadu Road Sector Project. The GIS-based RADMS software geographically maps all road accidents that take place on Tamil Nadu’s national and state highways, as well as on urban and district roads. The system identifies the most accident-prone spots and displays crash trends and other information at the click of a mouse. The RADMS software, developed

P a g e | 14

after detailed consultations between the police, transport and highways departments, has been helping the authorities analyze the ‘how’, ‘where’ and ‘why’ of road accidents, and enabling them to plan and implement remedial measures. In the two years since the system has been operational, nearly 3000 accident-prone spots have been identified. The implementation of road safety measures based on this analytical data has brought down the number of accident fatalities in Tamil Nadu from 13.39 for every 10,000 vehicles in 2006 to 10.09 in 2010, exceeding the targets set by the state.

Impacts
The RADMS system has been deployed at all the state’s 1,400 police stations. It has helped identify nearly 3,000 accident prone black spots. This has led to a number of interventions that have reduced road accident fatalities in the state from 13.39 for every 10,000 vehicles in 2006 to
10.09 in 2010. It has also resulted in:
Better traffic management
With accurate and real time accident data replacing the voluminous FIR data, the police department has been able to effectively post personnel at critical traffic junctions, deploy radar at appropriate places to detect and prevent speeding, provide speed breakers at relevant spots, upgrade traffic lights, and enforce rules for the use of seat belt and helmets etc.
Better roads
The highways department has been able to improve road conditions where they are most needed such as by creating dividers in areas which have a high frequency of head-on collisions, upgrading unsafe road junctions, providing better street lighting, installing road signs, trimming trees and so on. Increased road safety awareness The transport department has been able to target road safety awareness programs for appropriate groups, such as providing compulsory refresher classes for drivers of heavy vehicles.
Better allocation of budget resources
The availability of accurate information has enabled the authorities to provide financial and economic justification for the allocation of budgetary resources to improve road conditions.
Road safety now high on state agenda
The provision of regular reports to the highest authorities in the state has raised the profile of road safety in Tamil Nadu. Moreover, each department now has ownership of road safety issues, and the people have access to accurate information empowering them to ask questions about the implementation of road safety measures.

Road Ahead
The RADMS software is constantly being improved. It is soon planned to provide a hand-held
GPS devise to each police station, enabling personnel to enter details at the accident spot itself.

P a g e | 15

The device will automatically pick up the date, time, and spot where the accident happened.
Efforts are also on to link the system with medical facilities for quick attention to accident victims. Brief Introduction to the System
How is Data Recorded
The data is collected using ARFs or Accident Report Forms. The system allows us to view all these ARFs which contain all recorded information about the accidents. An ARF has a total of
150 fields, of which the Police Department needs to fill 100, the transport department fills 24 and the highways department needs to fill 16 of them. Under each of these classes, there are both mandatory and optional fields. There are a total 68 mandatory sections that have to be filled by the police department.
Under each ARF, the several broad categories include the General Details about the accident,
Location Details, Collision Details, Vehicle Details, Driver Details, Passenger Details and
Pedestrian Details. To gauge better understanding of the ARFs, a few screenshots have been attached below.

Figure 3: Sample ARF

P a g e | 16

Figure 4: Sample ARF

P a g e | 17

Figure 5: Sample ARF

Accessing the Portal
The access link for the portal is http://117.239.111.243/radms/acegilogin.jsp . IIT Delhi has been provided its unique user name and password. The credentials for the same are: Username: iitdelhi and Password is: rsms#123. A screenshot of the login screen is attached below.

Figure 6: Login Screen

P a g e | 18

Once the credentials are entered, a welcome screen opens which talks about ARFs and their management. The screen shot of the same is attached below

Figure 7: Welcome Screen

Analysis Options
On acceptance of the terms, one can go to the main system. The main system looks like this

Figure 8: Home Screen

By default, the system is open in the Home Tab. Here, basic analysis is allowed. The filters that can be used are date, district, police station and accident severity. On applying the filter, the list

P a g e | 19

of FIRs matching the filters applied are displayed. All this happens in the left half of the screen.
This is the most primitive analysis that can be performed on this system.
Once the results are obtained, there is an option to display these accidents on the map in the center of the page. The display options need to be selected as below:

Figure 9: Home Screen Map Display Options

However, currently this system is not working. This was the first problem we were able to identify in the system right away.
The second tab is for Accident Reporting. It allows for direct access to ARFs and allows the user to search and manage these ARFs. The screenshot for the same is attached below

Figure 10: Managing ARFs

P a g e | 20

The headers under each ARF are as discussed previously.
The third tab is for Accident Analysis. Under this tab, there are several options available for analysis. Each one is discussed below. At the same time, the limitations as well as the inconsistencies for each one of them has been discussed subsequently.

Kilometer Analysis
The screenshot of the Kilometer Analysis is attached below, which will be henceforth used as reference for all explanations about the system.

Figure 11: Kilometer Analysis

The Process Flow

Select the Dates for
Analysis

Select the District for Analysis

Select the Road under that District for Analysis

Flow Chart 1: Kilometer Analysis

Enter the Start and
End Points (in km) on the road for which the Analysis needs to be conducted Enter the resolution for Analysis
(Minimum value is
1).Click on Display to yield results

P a g e | 21

The Limitations and Issues associated with the Analysis
Firstly, the analysis does not work at all. No data has been fed into the backend so that this analysis can be supported. Irrespective of the choice of date, district, road and distances, the answer always remains 0.
Secondly, the minimum resolution to be entered is 1 km, whereas there are several roads of length less than 1 km, where this analysis would prove redundant even if the database had been maintained for the same.

Grid Analysis
The grid analysis allows users to select a rectangular area on the map of Tamil Nadu and depending on the parameters applied, the grid is further subdivided into cells which are color coded to reflect the desired accident severity type. The screenshot of the same is attached herein

Figure 12: Grid Analysis

P a g e | 22

The Process Flow

Select the dates for which the
Analysis needs to be conducted

Provide a weight to each accident severity type

Select a range of colors (Gives the maximum amount of different colours to be used)

Press Ctrl and select area on the map

Select
Appropriate
Display Option and Grid
Size(Area)

Flow Chart 2: Grid Analysis

Working of the System
1. Area of the Sub-Cell: It governs the total number of sub-cells which will be formed.
Here, a simple mathematical rule is applied where total number of sub cells is calculated by dividing the grid area by the size of each cell.
2. Weightage: The weightage factor for each severity type can be modified as to try and assess the magnitude of accidents of each category in the sub-cells. For instance, if you give the fatal accidents a weightage of 5 and the others 0, then the gird analysis will reveal the magnitude of the Fatal accidents across the grid by showing different colors for the various sub cells.
3. Calculation of Values for of a Sub-Cell: If the various weightage values are x1, x2, x3, x4 and x5 and the number of incidents of each type are n1, n2, n3, n4 and n5, then the value allotted to each cell would be Σ /Σ
4. Allocation of Colors: This is governed by the range entered in the range cell. Also, the color for the lowest value and highest value can also be picked from the color selector.
For allocation of colors, there can be 2 cases possible
a. The number of distinct cell values are less than or equal to the number of colors:
In this case, each value is given a unique color, the one with the lowest weightage given the color selected for the lowest value and similarly for the highest value cell. All sub-cells with same value are given the same color. A sample for the same is attached below:

P a g e | 23

Figure 13: Grid Analysis

Here, there are 6 distinct colors used since under the filters applied, only 6 distinct values cold be obtained. Here, if we change the range to 7, the colors will not change given the other conditions remain constant.
b. The total colors are less than the distinct cell values obtained: Here, say the maximum value obtained was vmax and the minimum value was vmin. The range between these two values is divided by the total number of colors to create buckets (ranges) for cell values with same colors. To further illustrate, say the minimum values was 0, and the maximum was 1, and total colors = 4. Then, 4 buckets of 0-0.25, 0.25-0.5, 0.5-0.75 and 0.75-1 are created. And as appropriate, all sub-cell values are put in the corresponding bucket. Here, if we change the color range and re-run the analysis, the colors for each grid will change.

Cluster Analysis
It is used to analyze clusters of accidents. The process flow is the same for grid analysis. The resolution is governed by the distance between the accident sites. Say there are 4 points of accidents A, B, C, and D and the resolution is 400m. Let’s say the points are arranged such that the distance between A and B is less than 400 m and that between A and others is greater than
400m. Similarly for B, the distance from C and D is greater than 400m. And the separation between C and D is less than 400m. In such a case, 2 clusters will be formed, one comprising of only A and B, and the other of only C and D. And accordingly, each cluster will have a color.
The color will be decided as it is done for the Grid Analysis. However, the Cluster analysis is not operational as of now. Here is a screenshot of the Cluster Analysis Screen

P a g e | 24

Figure 14: Cluster Analysis

Monitor Site Analysis: It is a grid analysis for pre-decided locations. The locations can be selected from a wide range of pre-decide locations. Additionally, new locations can be added and saved to the system.

Corridor Analysis: It is like the gird analysis for a particular corridor only. It has only pre-decide corridors and new additions cannot be made.

Safety Benefit Evaluation: The analysis is currently not operational and hence it uses and working could not be identified.

Dynamic Reports
Here, the system provides the flexibility to select certain attributes of the accident and its stakeholders as filters and obtain statistics on road accidents. Screenshots of the same are attached below

P a g e | 25

Figure 15: Dynamic Reports

Figure 16: Dynamic Reports

Here, as we can see from the second figure, direct graphs can be generated from the dynamic reports. However, the issue with this system is that here only 2 filters can be applied at one time and a very thorough analysis is not possible. Also, there is a problem of over counting which emerges here, since the definitions of the terms used are not very clear. We discussed the above issues with Prof. Kavi Bhalla from John Hopkins university while he was visiting IIT Delhi. Post the discussion, we could arrive at the final data extraction technique which is as discussed below.

Query Builder
Works like an SQL system, where queries can be defined and the database yields results as asked for. Here, the Spatial Query Builder does not work currently and hence no uses or limitations of the same could be assessed. For all the further analysis, I used the General Query Builder, a screenshot for which is attached below. The query builder is self-explanatory in use. It covers a wide range of aspects of the accidents. Once a query is fed in and run, we obtain the FIRs of the associated accidents as well as the values we are directly seeking for. The FIRs can be exported to a pdf report. There is an option to export the data to a csv format, but that does not work as of now, only making the further analysis tougher. A screenshot for the Query Builder and a Sample result it attached below

P a g e | 26

Figure 17: The Query Builder

Figure 18: Results from Query Run

P a g e | 27

DATA EXTRACTION:
3 approaches were taken to extract data :

Approach 1:
Data was extracted using filters given in dynamic report section. It was later found that it involved over counting as mentioned in error section and thus all analysis had to be discarded

Approach 2:
In consultation with Project supervisors and Dr. Kavi Bhalla, it was figured that FIR generated through pdf’s are the only correct source of data but then analyzing whole FIR to get just one data set was not feasible when large data was required for further analysis

Approach 3:
Finally, PDF’s were studied and their results were compared with various query builder results and as found that data here was offset by maximum 10% from the correct FIR data and hence this approach of extracting data was decided. Though this approach too was laborious and time consuming but was better than than initial two in terms of speed and accuracy.
The required data was extracted through RADM System by manually entering each query which gave a single datapoint and thus a whole lot of repeated queries in range of few thousands had to be done and because of which the whole process became quite manual and laborious. Thus, it limited the scope of data that could be analyzed.

RESULT & ANALYSIS
Data was extracted pertaining to fatal injury of pedestrian under following categories:
-

Traffic fatalities: Total Pedestrian fatality in various districts

-

Traffic fatalities: Pedestrian fatality vis-à-vis literacy rate of he district

-

Traffic fatalities: Pedestrian fatality among various age groups in different districts

-

Traffic fatalities: Pedestrian fatality associated with each impact vehicle in different states P a g e | 28

Traffic fatalities: Total Pedestrian fatality in various districts
There’s a large variation among different districts of Tamil Nadu as far as pedestrian fatality is concerned. Chennai district records maximum number of fatality in the range og 600-700 while
Nilgiris has least number of fatality. Total pedestrian fatality which took place in the observation period is 4413 which accounts to 140 fatalities on average per district.

No. of Pedestrian Casualties
700
600
500
400
300
200
100
0
Viluppuram

Tiruvannamalai

Tirunelveli

Thoothukkudi

Thiruvallur

The Nilgiris

Sivaganga

Ramanathapuram

Perambalur

Nagapattinam

Krishnagiri

Kanniyakumari

Erode

Dharmapuri

Coimbatore

Ariyalur

No. of Pedestrian Casualties

Figure 19: Pedestrian fatalities across districts

P a g e | 29

Fatality Rate per 1 million
160
140
120
100
80
60
40
20
0
0

10

20

30

40

50

60

70

80

90

100

Figure 20: Fatalities vis-a-vis literacy rate

Traffic fatalities: Pedestrian fatality vis-à-vis literacy rate of he district
Death rate corresponds to pedestrian fatality per 1 million populations. One cannot see any correlation between literacy rate and the fatality rate which in turn affirms that being more knowledgeable or informed has little or no effect on the extent of exposure to road accident.

Traffic Fatalities: Pedestrian fatality among various age groups in different states
Pedestrian fatality has been divided into various age groups: 0-20, 20-40, 40-60 and 60 above to investigate which age group is more prone to such fatalities. First category in general corresponds to children and school going population, second to college going and working professionals, third to senior working professional and last to old age people.

P a g e | 30

Pedestran Deaths
600
500
400
300
200
100
Ariyalur
Chennai
Coimbatore
Cuddalore
Dharmapuri
Dindigul
Erode
Kancheepuram
Kanniyakumari
Karur
Krishnagiri
Madurai
Nagapattinam
Namakkal
Perambalur
Pudukkottai
Ramanathapuram
Salem
Sivaganga
Thanjavur
The Nilgiris
Theni
Thiruvallur
Thiruvarur
Thoothukkudi
Tiruchirappalli
Tirunelveli
Tiruppur
Tiruvannamalai
Vellore
Viluppuram
Virudhunagar

0

0-20

20-40

40-60

60-Above

Figure 21: Pedestrian causalities in different age groups vs. districts

Fig.21 shows how pedestrian deaths vary over different age groups in different districts of Tamil
Nadu. Chennai sees a huge spike among 20-40 age group because of large college going population and large working professional that commute large distances daily because of it.
Because of high numbers in Chennai, other district data couldn’t be easily seen and thus another fig was analyzed without the data of Chennai and one could see that age group of 40-60 face maximum exposure of fatalities while first group face the minimum probably because of necessary precaution taken by parents as well as by drivers when they see any children crossing the road.

P0

study on the possible reasons for the same.
P20
P40
Virudhunagar

Viluppuram

Vellore

Tiruvannamalai

Tiruppur

Tirunelveli

Tiruchirappalli

Thoothukkudi

Thiruvarur

Thiruvallur

40-60

Theni

The Nilgiris

Thanjavur

Sivaganga

20-40

Salem

Ramanathapuram

Pudukkottai

Perambalur

0-20

Namakkal

Nagapattinam

Madurai

Krishnagiri

Karur

Kanniyakumari

Kancheepuram

Erode

Dindigul

Dharmapuri

Cuddalore

Coimbatore

Ariyalur
Ariyalur
Coimbatore
Cuddalore
Dharmapuri
Dindigul
Erode
Kancheepuram
Kanniyakumari
Karur
Krishnagiri
Madurai
Nagapattinam
Namakkal
Perambalur
Pudukkottai
Ramanathapuram
Salem
Sivaganga
Thanjavur
The Nilgiris
Theni
Thiruvallur
Thiruvarur
Thoothukkudi
Tiruchirappalli
Tirunelveli
Tiruppur
Tiruvannamalai
Vellore
Viluppuram
Virudhunagar

P a g e | 31

140

120

100

80

60

40

20

0

60-Above

Figure 22: Pedestrian causalities’ in different age groups vs. Tamil Nadu districts (except Chennai)

70

60

50

40

30

20

10

0

P60

Figure 23: Percentage contribution of different age groups in pedestrian causality in different states

One could observe inverse relation among age groups 20-40 and 40-60 which requires further

P a g e | 32

Traffic fatalities: Pedestrian fatality associated with each impact vehicle in different states
Various vehicles have been divided into 6 groups: Motorcycle, Ambulance, Animal Drawn, Bus,
LMV and others. All cases involving pedestrian fatality due to some impacting vehicle in all districts of Tamil Nadu are recorded. For eg: whenever a bus hits another vehicle or a pedestrian leading to a pedestrian fatality, the incident gets recorded in the given record and percentage of total such incidents associated with each vehicle has been shown in the fig 24 below

Ariyalur
Chennai
Coimbatore
Cuddalore
Dharmapuri
Dindigul
Erode
Kancheepuram
Kanniyakumari
Karur
Krishnagiri
Madurai
Nagapattinam
Namakkal
Perambalur
Pudukkottai
Ramanathapuram
Salem
Sivaganga
Thanjavur
The Nilgiris
Theni
Thiruvallur
Thiruvarur
Thoothukkudi
Tiruchirappalli
Tirunelveli
Tiruppur
Tiruvannamalai
Vellore
Viluppuram
Virudhunagar

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

Motorcycle

Ambulance

Animal drawn

Bus

Car Lmv

Figure 24: Percentage contribution of vehicles in each district

Others

P a g e | 33

MotorCycle

Ariyalur
Chennai
Coimbatore
Cuddalore
Dharmapuri
Dindigul
Erode
Kancheepuram
Kanniyakumari
Karur
Krishnagiri
Madurai
Nagapattinam
Namakkal
Perambalur
Pudukkottai
Ramanathapuram
Salem
Sivaganga
Thanjavur
The Nilgiris
Theni
Thiruvallur
Thiruvarur
Thoothukkudi
Tiruchirappalli
Tirunelveli
Tiruppur
Tiruvannamalai
Vellore
Viluppuram
Virudhunagar

50
45
40
35
30
25
20
15
10
5
0

Figure 25: Fatality variation among states due to motorcycle

On average 20-30% of such incidents are because of motorcycle which affirms that it is a major contributor to accidents among other vehicles and is one area which needs to be catered. Nilgiris saw almost zero such incidents which can be attributed to hilly region in nilgiris leading to more of buses and cars in the areas than the motorcycles. On the other side, Chennai being the metropolitan records maximum percentage of 44% which can be attributed to more usage of two wheelers by citizens of the area for daily commuting.

0

Ariyalur
Chennai
Coimbatore
Cuddalore
Dharmapuri
Dindigul
Erode
Kancheepura…
Kanniyakum…
Karur
Krishnagiri
Madurai
Nagapattinam
Namakkal
Perambalur
Pudukkottai
Ramanathap…
Salem
Sivaganga
Thanjavur
The Nilgiris
Theni
Thiruvallur
Thiruvarur
Thoothukkudi
Tiruchirappalli
Tirunelveli
Tiruppur
Tiruvannama…
Vellore
Viluppuram
Virudhunagar

Ariyalur
Chennai
Coimbatore
Cuddalore
Dharmapuri
Dindigul
Erode
Kancheepuram
Kanniyakumari
Karur
Krishnagiri
Madurai
Nagapattinam
Namakkal
Perambalur
Pudukkottai
Ramanathapuram
Salem
Sivaganga
Thanjavur
The Nilgiris
Theni
Thiruvallur
Thiruvarur
Thoothukkudi
Tiruchirappalli
Tirunelveli
Tiruppur
Tiruvannamalai
Vellore
Viluppuram
Virudhunagar

P a g e | 34

Ambulance

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Figure 26: Fatality variation among states due to ambulance

Very less incidents of ambulance being the impacted vehicle were seen. Its contribution was

maximum in Viluppuram and that too 0.8% which may be because of extra care people take in

giving way to ambulances or such emergency cases. Moreover, distance travelled by ambulances

is generally quite less in comparison to other modes like motorcycles and cars.

LMC

40

35

30

25

20

15

10

5

Figure 27: Fatality variation among states due to LMC

P a g e | 35

Large variation in percentages could be seen in the case of cars in different districts of Tamil
Nadu. On average it contribution is around 15%. This variation can be attributed to different adoption rates of cars in different states and the kind of road and safety infrastructure available.

All
50
40
30
20

Animal Drawn

Viluppuram

Virudhunagar

Vellore

Tiruppur

Tiruvannamalai

Tirunelveli

Tiruchirappalli

Thiruvarur

Bus

Thoothukkudi

Thiruvallur

Theni

The Nilgiris

Thanjavur

Salem

Sivaganga

Pudukkottai

Namakkal

Ambulance

Perambalur

Nagapattinam

Madurai

Karur

MotorCycle

Krishnagiri

Kanniyakumari

Kancheepuram

Erode

Dindigul

Dharmapuri

Cuddalore

Coimbatore

Ariyalur

Chennai

0

Ramanathapur…

10

Lmc

Figure 28: Fatality variation among states due to all vehicles

Overall, Motorcycle dominates as far as impacting vehicle is concerned. Also, one could see an inverse relationship between LMV and motorcycle which is something that needs to be further investigated to find reasons for it. Also, animal drawn and ambulances are negligible contributors to pedestrian fatality.

Field Visit to Tamil Nadu
From 28th April to 2015, to 29th April 2015, we visited the Police Headquarters of the state of
Tamil Nadu and interacted with various officials starting from the head constable level to ADGP,
Technical Services, Tamil Nadu. The aim of the visit was to better understand our working of the system and gain firsthand experience on using the application for extraction of data for conducting further analysis. The experiences of all the officials was recorded, like their feedback on the utility of the system and what initiatives is the Tamil Nadu government taking to further curtail road accidents.

P a g e | 36

Figure 29: Police headquarter visited during field trip

RADMS took birth after the confluence of 3 stakeholders within Tamil Nadu, namely, the Police
Department, Highway Department and the Transport Department. Its design was headed by the
Transport Department, and the operations have always been managed by the Police. Thus, for any changes to the system (i.e any technical or functionality changes), the competent authority to approach transport department and not the Police.
At the police headquarters, RADMS is managed by the State Transport Planning Cell, which is the central body which regulates the use of the system. We have discussed below the operational procedure of how the system is operated and the role of each of the staff members involved as each level.

P a g e | 37

Admin

Figure 30: RADMS data flow

When an accident first happens, the nearest police officer will rush to the site to analyze the accident and record data for the FIR. The firs record is made on a blank paper, and then on the
FIR. The pictures of the same are attached below

P a g e | 38

Figure 31: FIR sample

P a g e | 39

Figure 32: Collision sketch in FIR

P a g e | 40

Figure 33: Sample FIR

Primary Recording of the Data
If the accident is Fatal (i.e. there is at least 1 fatality, or a highly grievous injury which later gets converted to a fatality), the accident FIR has to be uploaded to the RADMS within 24 hours of the incident. However, if the accident is non-fatal, a time period of 3 days is provided to fill up the ARF for the particular accident.
The ARF is filled not by the police officer who visited the site, but by those who are well versed with computers and their operations. Thus, the person who observes the accident and the one who reports it online may not be the same. Thus, there is a scope of an error in interpretation of the FIR data, even though the form is well coded and the staff is well trained, but differences in the way humans interpret things cannot be ruled out. To reduce this, the following form is suggested P a g e | 41

Figure 34: Proposed FIR form

P a g e | 42

Figure 35: Proposed FIR form (continued)

Filling the ARF
There are 7 possible headers which have to be filled, namely, General Details, Location Details,
Collision Details, Vehicle Details, Driver Details, Passenger Details and Vehicle Details. The computer operator first begins by filling the General Details. There are 150 fields to be filled, of which 68 are marked as mandatory.
First, a unique Crime Number is generated by the police department for the accident which is fed into the ARF. For the purpose of this system, it is nothing but the FIR number. The police officials are also required to analyse the event before filling the form to try to identify the major causes for the accident and propose counter remedial measures, which have to be filled in this

P a g e | 43

section. From the feedback we got, in most cases, the causes are mentioned as Rash & Negligent driving, which is true for over 90% of the cases. In the general details, they also have to mention the number of vehicles, drivers, passengers and pedestrians involved in the incident. If any one of them is 0, the form for that particular category is disabled. This not only makes the process of uploading the data simpler, but prevents fake details from being recorded.
Then, there are certain other parameters that have to be filled, but the most important of them is the location of the accident, which is fed in terms of latitude and longitude. When the system was started, it was filled using google maps. Later, GPS were introduced and provided at all police stations to record the exact location of the accident. This means that in the older ARFs, the locations provided may not be as accurate as for the accidents that have occurred since the introduction of GPS.
Next, the Collision details are to be filled, where there is a provision for a collision sketch/photograph to be added, which is done only in the case of very severe accidents since those graphics can then be used for investigatory purposes. There is also a voice recording feature which allows for the views of police officer on site to be recorded.
In the collision details itself, a field by the name of contributory actor is to be filled, where there are 4 options available, namely Engineering, Human, Natural and Others. These are filled on the basis of the judgment of the police. However, since accidents are multi-factor in nature, and human judgment may be biased towards a certain cause, a possibility of an error might creep in.
This is a very critical field, since on the basis of the parameter recorded, the issue is escalated to the Transport Department or the Highway Department for further remedial action.
In the 3 form sections that have been discussed till now, the system has been provided with the functionality to ensure that the quality of the data recorded is good and fake reporting is prevented. The Admin cell has been empowered by a set of queries to ensure quality control and robustness of the data. For instance, if the time of an accident is entered as 9 pm, and the lighting conditions are mentioned as Daylight conditions, the Admin will be able to identify this discrepancy and in this case, they will immediately notify the competent authority of the error, thus pushing for transparency and increasing accountability of the system.
Next, the vehicle details are to be filled. On the top of the form, the operator has to select whether the vehicle is of the victim or the accused. Once this selection is done, they feed the registration number of the car and other details. They are soon planning to automate the process,
i.e. on entering the car registration number, the details of the vehicle and driver fill be filled automatically by using an application called the Wahan Saarthi, which has an interface the
Transport Department, all vehicle particulars and associated driver particulars get filled automatically. The driver parameters can also be entered manually, since it may not be necessary that the vehicle is driven by a single person only. The idea of this automation is to develop a system in the future, where the police have all accident records of a driver at the click of the mouse and it automatically recommends appropriate action against the driver (eg. Imposing of fine, compounding of vehicle, cancel of licence, etc.)

P a g e | 44

Once the ARF is submitted into the system by the police department, a mail is automatically delivered to the concerned authority in the Transport and Highway Department since certain form parameters have to be filled by them as well. As per the SOP, they have a variable length of allowed time to do their job. If however, they do not do so, the issue is escalated to senior authorities. Thus, it is ensured that work is done in time and order.

Role of the Admin Cell
The Admin cell, or the STPC which is based out of Chennai is from where the system is being monitored. Also, they constantly perform R&D to identify areas for improvement in the functionality of the system and operational procedures. The responsibilities of the STPC are as listed below







Monitor monthly tally of accidents and tally with the data collected by State Crime
Review Board (SCRB) to ensure correctness of data on every 4th day of the month or first
Tuesday, whichever is later.
On the lines of reports by the IRC, they have to send quarterly and annual reports to the government to discuss the situation of road safety and propose further action
Maintaining the quality of data by conducting checks on ARFs from time to time. For this, there is a 2 prong strategy that is in practice. First, an ARF is selected at random and a query is run to check if all the mandatory 68 fields have been recorded. If it emerges that not all fields has been entered, the issue is highlighted, and the concerned police station and officer have to immediately take appropriate corrective measures. If however the 68 fields are filled, then as suggested previously, the Admin can run certain queries on the system to identify fake reporting. Thus, the system has been designed to overcome the two foremost biggest challenges of data collection, under reporting and false reporting. Conduct training sessions for the staff every 3 months on using RADMS

Other Concerns
From our past work, there were certain queries which had emerged in regards to the interpretation of the data provided by the system. We discussed them with the concerned individuals here and the following describes the current status of the concerns.
1. How are the total number of deaths caused in road accidents calculated, and how is the distribution amongst drivers, passengers and pedestrians arrived at ? The solution to this lies in the use of dynamic reports which allow the user to generate the reports as per their choice of parameters. The process for the same is suggested as below
a. While selecting the required output, if the number of fatal accidents are to be determined, chose the output as Number of Accidents. If pedestrian deaths, then pedestrian casualty and so on and so forth for passengers, drivers and total (for a cumulative value)
b. For determining X casualty (X = driver/passengers/pedestrian), one of the report parameters (i.e. row or column) has to be X injury severity or accident injury severity. The other parameter can be decided on need basis

P a g e | 45

c. For determining total casualty or total number of fatal accidents, one of the selected parameters should be accident injury severity, and the second parameter is again need based.

2. Does the system allow us to extract data on impacting vehicle in collision?
a. No, it does not with its current functionality

3. Are the Kilometer Analysis and Safety Benefit Evaluation operational?
a. We have been told they are, though when the analysis was performed here for several examples, no values were obtained. They have told us that they will talk about this to the necessary authority
4. Why there are so many vehicle categories included which unnecessarily burden the system? a. As per Mr. Ravi Kumar, Head Constable at STPC, who manages the entire operations and explained the system to us told us that it has been done to make work easier for the data collector, and they will not feel lazy while reporting the data and filling the entire ARF manually. Rather, they can just pick the correct option as per the situation. When asked whether the police officer at site, who does not have this list can identify which vehicle it is, he replied that they can because of the training they have been provided.

Some Observations Pertinent to the System
Post the visit and our interactions, on the outset, it is a highly advanced system when it comes to functionality and its level of service is comparable to most international systems in place. A better remark can be made on the state of the system only post an interaction with the people managing the technicalities of the system. However, as a researcher with access to the system, the system appears to lag behind since there is not much clarity on how to use the system. Apart from this, on the operational front, there is reliability on human interpretation and judgment which may lead to incorrect data collation. Though the system has provisions for maintaining consistency in terms of quality of data, yet the scope for error remains. And this problem can be overcome by further standardization and reduction in unnecessary form parameters.
The system has proved really valuable and has helped improve the situation of road safety in
Tamil Nadu. However, equally important has been the commitment of the state government and its officials to the cause, which is visible only in one on one interactions. This can be understood from the kind of remedial measures they have taken to counter the menace of road accidents. For eg: On highways, especially at night, police officials have been stationed to stop cars and interact with the driver, and even serve them tea to ensure they are not fatigued, since as told to us, on highways, driver fatigue is a cause for a lot of accidents, especially at night. Also, slip roads have been developed in certain stretches which can be used as a resting bay by the drivers.

P a g e | 46

Road Safety Index
In recent years, the interest in and the use of indicators and indexes are rapidly increasing. In general terms, an indicator is a quantitative or a qualitative measure derived from a series of observed facts that can reveal relative positions (e.g. of a country) in a given area. Their benefit for policy support and communication is a key advantage. Possible applications include trend identification, problem prediction, target and priority setting and impact assessment of measures.
The most important advantage of one index over an accumulation of individual indicators is that all relevant information is aggregated in one final score which can be used for ranking countries, tracking changes over time, etc. However, road accident data are characterized by various inconsistencies, like the lack of uniformity in definitions and the problem of under-registration.
The most important drawback is that knowledge of the number of accidents and casualties in a country is insufficient to understand the processes that lead to traffic accidents and casualties.
To rank a set of countries according to their road safety level, the underlying risk factors and the road safety measures that determine the road safety performance should therefore also be taken into account. Road safety indicators are useful in this respect, as they offer a means to include these dimensions in the road safety ranking. The main virtue of composite indicators is their usefulness for policy analysis in that they can summarize complex and sometimes elusive issues in wide ranging fields, e.g., environment, economy, society or technological development. Composites often seem easier to interpret than finding a common trend in many separate indicators and have proven useful in benchmarking country performance.

Following diagram shows the pros and cons of going for composite indicators in detail and what are the considerations while going for one

P a g e | 47

Figure 36: Pros and Cons of the complex index

Before starting the analysis of the various steps in constructing a composite indicator we clarify some basic definitions
Dimension: is the highest hierarchical level of analysis and indicates the scope of objectives, individual indicators and variables. For example, a composite indicator on sustainability can include economic, social and environmental dimensions; a composite indicator on active citizenship can include participation to elections and to voluntary associations, but also values and attitudes of the citizens. The development of attitudinal indicators requires economic, social and psychological dimensions.
Objective: an objective indicates the direction of change desired. For example, within the economic dimension GDP has to be maximized; within the social dimension social exclusion has to be minimized; within the environmental dimension CO2 emissions have to be minimized. Steps in the construction of a composite indicator
We followed following sequence of steps to develop an ideal composite indicator to evaluate safety of various districts of Tamil Nadu using tremendous data generated by RADMS.
1. Theoretical framework

P a g e | 48

2.
3.
4.
5.
6.
7.

Variable selection
Imputation of missing data
Multivariate analysis
Normalization
Weighing
Aggregation

Variables
From our discussions and literature review, the stakeholders/parameters involved in road safety are Road Characteristics, People, Police, Government and Vehicles. Within each of one these, there are further sub components which influence road safety of a given area and hence need to be evaluated. We have selected the following variables. The following section discusses the subcomponents of each of the above mentioned attributes, explanations for their inclusion and the measure used to evaluate them.
Road Characteristics


Road Pavement Status - Typically, when the road is paved, vehicular movement is smoother. Hence, this enhances mobility and reduces risk associated with road travel.
It is calculated by normalizing the percentage of paved roads of the area on a scale of 0 to
1. For this, first, the percentage of paved roads is calculated in an area and arranged in ascending order. Then, the value assigned to each area is given by the formula. =

−
( − )

The accepted minimum value will be slightly lesser than the absolute minimum value.
This has been done to obtain non-zero values for coefficients.


Lighting Conditions – This factor plays a pivotal role during early morning and late night movements, where a lot of accidents happen.
For this, first, we determine the number of street lights per km as Total Number of street lights in the area/ Total length of roads in the area. The value will then be normalize between 0 and 1 as per the above suggested normalization.



Convenience of Travel – This is a subjective measure, which includes the following subcomponents o Sharpness of curves o Road surface nature (ditches, potholes), o Risk from Construction Activities, o Presence of undesirable objects o Intersection design characteristics

P a g e | 49

Each of the parameters will be rated from 1 to 5, 5 being Highly safe and 1 being the lowest when these parameters are analyzed from a perspective of road safety. For 1 given street, first, all 5 values will be determined. Then, they will be averaged (simple arithmetic mean) to obtain the convenience parameter for 1 road. The above process will be repeated for X number of roads per district, where X will be decided as per convenience and feasibility.
For deciding X, a suggested sampling technique is to first sort all streets by the number of accidents on them. From them, a certain percentage of roads will be selected (mostly from the top of the list) to develop a worst case scenario.
Another suggested sample could be on the basis of major roads as classified by total flow volume and peak time flow characteristics.


Junctions Characteristics – To evaluate this, the roads from the sampling of the
Convenience factor will be considered. More precisely, the junctions on these roads.
Then, the percentage of junctions which have traffic controls will be identified for each area. Once we have the percentages, the values will be normalized to obtain a value between 0 and 1 as per the previously suggested normalization technique.

People


Directly Measureable o GDP per person – Typically, higher GDP has been associated with higher levels of technology adoption and higher safety. The log10 will be evaluated for the GDP per capita, post which it will be normalized as above. o Education Levels – Measured in terms of literacy rate. The values are then normalized before being used for evaluating the index.



Indirectly Measurable – These are measure of people’s attitude towards road safety, in terms of their awareness of the rules and how much they adhere to them. They are measured as a product of exposure and risk. Exposure is defined as the chances of an occurrence of an incident whereas risk is defined as the risk of death on the occurrence of the accident. Thus, it is measured as IM = E x R. E = Number of fines recorded in 1 year/ 1 Lakh Population. R = Number of fatalities due to the violation/ Total number of fatalities o Alcohol Violations o Over Speeding o Rash Driving o Safety Belts and Helmet
The values for IM will be calculated in the units as suggested. Then, these will be normalized to a scale of 0 to 1. However, at this point of time, higher values indicate higher risk of accident. Thus, we will invert these values to obtain a measure of safety, and then repeat the normalization process once again.

P a g e | 50



Sensitivity to Victim – Measured as a percentage of Hit and Run cases, its contribution is calculated the same way as IM is.

Police
It has two components. The first is the total size of the police force. The second is the efficiency of the police force with which it works. These 2 are captured as follows:



Size of Police force - Traffic policemen per 1,00,000 people will be first evaluated. Then, it will be normalized using the technique being used till now.
Efficiency - 4 parameters have been identified to measure the efficiency of the police force, namely o Activity – Measured in terms of attendance of police force. It will be normalized between 0 and 1 as per the used technique. o Ethics - Measured through psychometric testing. The results will be on the scale of 5 to 1, which will then be scale down to the range of 0 to 1, 5 being high standards of ethics and workmanship, 1 being the lowest. o Knowledge – Measured through awareness of traffic rules by conducting a test, the marks for which will be normalized. o Past Performance – Challans issued per unit population, the values for which will again be normalized.
These parameters will be first evaluated for 1 policeman. Then, for each district/ area,
X% of policeman will be evaluated, and the final value of efficiency will be obtained by averaging over all of them.

Government
The contributions of the government are of measured in terms of the following indicators

Figure 37: Proposed Government Index

For these indicators, an (a) is equivalent to 3 on a scale of 3 in terms of better road safety, and (c) is equivalent to 1, 1 being the lowest. To evaluate the final value of G, first A1 to A5 are

P a g e | 51

determined for the area. Then, we take an arithmetic mean of these values and obtain a values between 1 and 3, which is then divided by 3 to scale it down to a value less than 1.

Vehicle


Measurement of Vehicular default – It is measured as ratio of number of accidents where contributory factor is Vehicle default to Total number of Accidents. The ratio is normalized, then inverted, and then normalized again as we did for IM.

Evaluating the Contributions to the Index
We have divided road safety index into five stakeholder indices which have been given equal weightages i.e. 0.2 each for getting the aggregated safety index.
Thus, the final value of the traffic safety index of a region can be calculated as
0.2 * Road Index(R) + 0.2 * People Index(P) + 0.2* Government Index(G) + 0.2 * Police(O)
Index + 0.2 * Vehicle Index(V), where the values of each of these indices lies between 0 and
1.
Calculations for individual indices are shown below
Road Characteristics
Road characteristics component is further divided into 4 sub components where in again we have given equal weightages to each considering each contributes equally to safety to begin with.
In Convenience of travel component, all 5 further divisions are ranked out of 5 and their average taken to get a normalized value from the convenience factor.
The value is calculated using the following formula
R = 0.25 * Road Pavement Status + 0.25 * Lighting Condition Status + 0.25*Convenience
Status + 0.25*Junction Traffic control status
People’s Attitude
We have divided this components into 3 sub components and again equal weightage of 1/3 to each of them. Further division into these sub components have been treated equally leading to
1/6 weightage to individual direct subcomponents factor and 1/12 for indirect subcomponent factor and 1/3 for hit and run case factor.
P=0.33*Direct component + 0.33*Indirect component + 0.33*Hit and run case component
Government

P a g e | 52

Here, we have 5 subparts which are graded out of 3 and each part has equal weightage of 1/5 i.e.
0.2
G= 0.2*A1 + 0.2*A2 + 0.2*A3 + 0.2*A4 + 0.2*A5 where Ai denotes value of 5 sub parameters
Vehicle index
This index involves only single component of vehicular defect with all weightage contributed by it V= 1.0* Vehicular Defect Status
Police Index
Police index comprise of 2 sub components, size of police force and efficiency of police force which further has 4 components which are again given equal weightage. Thus size gets a weightage of 0.2 and each subcomponent under efficiency gets a weightage of 0.2
O = 0.2* Size status + 0.8*Efficiency Status
0.8 *Efficiency Status = 0.2* Attendance + 0.2*Ethics + 0.2*Knowledge + 0.2*Past performance The following graphic illustrates a sample calculation that was performed by using randomized data, and the values obtained at each step of calculation. (it is attached as an excel to this document) Alternate Approach:
The index we suggested in the above section involves lot of subjective measures which at present were not feasible to get or accounted for in our index calculation like rating government initiatives. Thus, we have designed an alternate index calculation method which incorporates available data attributes and calculates index values for each state and union territories of the country We have identified various factors which ultimately lead to accidents and these factors are in a way crucial for both people inside the car and people outside the car i.e. pedestrians.
Various factors identified are:





Fault of driver
Fault of cyclists
Fault of pedestrians
Defect in condition of motorcycles

P a g e | 53





Defect in Road condition
Weather conditions
Other factors which includes poor lightening , Stray animals and negligence of civic bodies We have used the total accidents data for the country under various factors to assess how much each individual factor contribute to the accidents in the country in general and thus have been used to calculate the weightages.
The weightages we obtained were:
Weightage
0.732044 0.017713 0.034262 0.019811 0.016968 0.013742 0.165459
Table 1: Weightages calculated for each contributory factor

respectively for seven factors chosen and were used for state wise data we obtained under given categories. Now we also extracted data on the lines of population, No of registered vehicles and Total road length for each state in order to get Accident estimates per 1000 of each attribute which helps in a way to normalize accident values as per the these parameters.
Further, an ascending order of states was made under three parameters i.e. Per 1000 population, per 1000 vehicles, per 1000 km of road length in terms of person killed in accidents giving an idea how various states fair against each other.
Following table shows ranking of states under three parameters:

State

per 1k per 1k
Population vehicle

Lakshadweep 1

1

per 1k road length
1

Manipur
Nagaland
Meghalaya

2
3
4

3
2
4

3
2
6

Delhi**
Sikkim
West Bengal

5
6
7

6
5
22

13
5
10

Bihar

8

31

20

Uttar Pradesh 9

33

15

Arunachal

35

4

10

P a g e | 54

Pradesh
Uttarakhand

11

18

11

Tripura
Jharkhand

12
13

16
7

7
33

Andaman and Nicobar
Islands
Jammu and
Kashmir
Assam
Mizoram

14

11

14

15

17

23

16
17

23
12

9
8

Odisha

18

27

12

Chhattisgarh

19

30

18

Daman and
Diu
Madhya
Pradesh
Maharashtra

20

8

35

21

34

24

22

19

21

Gujarat

23

15

27

Punjab

24

29

25

Kerala
Chandigarh
Karnataka
Rajasthan

25
26
27
28

9
14
25
32

16
29
22
26

Haryana
Andhra
Pradesh
Puducherry

29
30

20
28

34
28

31

21

30

Himachal
Pradesh
Dadra and
Nagar Haveli
Goa

32

24

17

33

13

31

34

10

19

Tamil Nadu

35

26

32

Table 2: Ranking of states on three parameters

P a g e | 55

Following figure shows data about certain states and how they fair in rankings on different parameters like Sikkim has no effect with underlying parameter and ranking remains constant throughout while one can see huge differences in ranking in Bihar and Uttar Pradesh across three parameters showing how these parameters play a role in road accident fatalities.
Here, lower the index, more safer is the state and higher the rank, more accident prone is the state. Meghalaya 4

4

6

Delhi
Sikkim
West
Bengal
Bihar

5
6
7

6
5
22

13
5
10

8

31

20

Uttar
Pradesh
Arunachal
Pradesh

9

33

15

10

35

4

Table 3: Variation of ranks under various parameters

P a g e | 56

References
A.S. Hakkert, M. V. (2007). Road Safety Performance Indicators .
Akaateba, M. A. (n.d.). Comparing Road Safety Performance of Selective EU and African Countries using a composite Road Safety Performance Index.
Baguley, C. (2001). The importnace of a road accident data system and its utilisation. International
Symposium on Traffic Safety Strengthening and Accident Prevention. Nanjing.
C bax, P. W. (2012). Developing Road Safety ndex. Belgian Road Safety Institute.
Data Systems: A Road Safety Manual. (n.d.).
Elke Hermans, D. R. (2009). Road safety risk evaluatio by means of ordered weights averaging operators and expert knowledge. Elsevier, 48-52.
F. Pirotti, A. G. (n.d.). Road Safety Analysis Using Web Based Collaborative GIS. Legnaro.
Fred Wegman, S. O. (2010). Benchmarking road safety performances of countries. Safety Science 48,
1203-1211.
Haji, G. A. (n.d.). Towards a Road Safety Development Index.
India Stats. (n.d.). Retrieved from http://indiastat.com/
Jaehoon SUL, L. O. (n.d.). Road Accident Data Collection and Management System in Korea. The Korea
Institute of Transport.
K.Evangelidis, S. B. (2006). A GIS web-based traffic accident information system. WIT Transactions on
Information and Communication Technolgies, Vol 36, 363-371.
Paul de Laur, T. S. (n.d.). Development of a Road Safety Risk Index.
Peter Hollo, V. E. (2010). Road Safety performance indeicators and their explanatory value: A critical view based on Central European Countries. Saftey Science 48, 1142-1150.
Ramasaamy, N. (n.d.). GIS based information system on Urban Road Safety - A Case of Pune City.
(2013). Reports on Road Accidents in India. NCRB.
Road Crash and road crash injury data for setting and monitoring targets. (2009).
T.H. Law, r. U. (n.d.). The Development of Road Accident Database Management System for Road Safety
Analysis and Improvement. Road Safety Research Center, Unviersity Putra Malaysia.
T.H.Law, R. U. (n.d.). The Development of Road Accident Databse Management System for Road Safety
Analysis and Improvement.
Victoria Gitelman, E. D. (2010). Desiging a composite indicator for road saftey. Safety Science 48, 12121224.

P a g e | 57

Appendix
Table of Minimum Data Elements
Crash Data Elements
The crash data elements describe the overall nature of the crash
Crash Data
Element
Crash Identifier

Definition

Obligation

Data Type

Comments

A unique (eg. X digit number) within a given year that identifies the crash Mandatory

Numeric or
Character String

Crash Date

The date (day, month, year) of crash Mandatory

Numeric

The time at
Mandatory
which the crash occurred in 24 hr format Crash
The municipality Mandatory
Municipality and and country or
Region
equivalent entity

Numeric

This value is usually assigned by the police as they are responsible at the crash scene.
Other systems may reference the incident using this number. If a part of the crash date us unknown, the respective places should be filled with a 99(for day and month).
Absence if a year should result in an edit check. Important for season comparisons, time series analysis, management/ administration, evaluation and linkage Allows for analysis at various times of the day
Important for analysis of local and regional

Crash Time

Character String

P a g e | 58

the crash occurred in

Crash Location

Crash Type

The exact
Mandatory
location at which the crash occurred. Optimum definition is route name and
GPS/GIS
coordinates that can relate geographic coordinate to specific locations in road inventory and other files. The minimum requirement for documentation is street name, reference point, distance and direction from reference point
The crash type is Mandatory characterized by the first injury or damage producing event of the crash

programs and critical for linkage of crash file to local data files. Also helps in interregional comparisons Character String Critical for to support problem latitude/longitude identification, coordinates, prevention linear programs, referencing engineering method, or link evaluations, node system mapping and linkage purposes

Numeric

Possible Data
Values:
Crash with pedestrian, Crash with parked vehicle, crash with fixed obstacle, Crash with Non-Fixed
Obstacle,
Animal, Single
Vehicle
Crash/Non
Collisi0on,
Crash with

P a g e | 59

Impact Type

Indicates the manner in which the road motor vehicle involved initially collided with each other.
The variable refers to the firs impact of the crash, if that impact was between tow road motor vehicles Mandatory

Numeric

Weather
Conditions

Prevailing atmospheric conditions at crash location

Mandatory

String

Light Condition

The level of natural and artificial light at the crash location at the time of crash

Mandatory

String

multiple
Vehicles and
Other Crashes
Data Values:
No Impact between motor vehicles, Rear end Impact,
Head on Impact,
Angle Impact – same direction:
Front of first vehicle collides with side of second, Angle impact – opposite direction, Angle impact-right angle, Angle
Impact-direction
not specified,
Side by Side impact- Same
Direction, Side by Side impactOpposite
Direction, Rear to side impact
Clear,
Rain(heavy or light), Snow,
Fog/mist/smoke,
Sleet/hail, Other,
Unknown
Possible Values:
Daylight,
Twilight,
Darkness,
Darkness with street lights unlit, Dark with street light unlit

P a g e | 60

Crash Data Elements derived from collected data
Crash Severity

Described the severity of the road crash, based on the most severe injury of the person involved

Mandatory

Road Data Elements
Type of
Describes the
Mandatory
Roadway type of road, whether it has two directions of travel, and whether the carriageway is physically divided. For crashes occurring at junctions where the crash cannot be clearly allocation in ne road, the road where the vehicle with priority was moving is indicated Road
Describes the
Mandatory
Functional character of
Class
service or function of the road where the first harmful event took place. For crashes occurring at junctions, where the crash cannot be clearly allocated in one road, the road where the vehicle

String

Fatal, Severe,
Slight

String

Motorway/freeway,
Express Way,
Urban road- 2 way,
Urban road – 1 way, Road outside built up area,
Restricted road,
Other

String

Principal Arterial,
Secondary Arterial,
Collector, Local

P a g e | 61

Speed Limit

Road Obstacles

Road Surface
Conditions

Junction

with priority was moving is indicated.
The legal speed limit at crash location The presence of any person or object which obstructed the movement of the vehicles on the road. Includes any animal standing or moving (either hit or not), and any object not meant to be on the road.
Does not include vehicles (parked or moving vehicles, pedestrians) or obstacles on the side of the carriageway (e.g. poles, trees).
Road condition at time and place of crash Mandatory

Numeric

Mandatory

String

Yes, No, Unkown

Mandatory

String

Indicates whether Mandatory the crash occurred at a junction (two or more roads
Intersecting) and defines the type of the junction. In at-grade junctions all roads intersect at the same level. In not-at-grade String

Dry,
Snow/Frost/Ice,
Slippery,
Wet/Damp, Flood,
Other, Unknown
At-Grade,
crossroad; At-grade, roundabout; T or staggered junction,
At-grade multiple junction; Not at grade, Not at junction, Unknown

P a g e | 62

Traffic Control at Junction

Road Curve

Road Segment
Guide

junctions roads do not intersect at the same level.
Type of traffic
Mandatory if
String
control at the crash occurred junction where at a junction crash occurred.
Applies
only to crashes that occur at a junction. Indicates whether Mandatory the crash occurred inside a curve, and what type of curve Indicates whether Mandatory the crash occurred on a road segment with a steep gradient String

String

Authorised person,
Stop sign, Give wat sign or markings,
Other traffic sings,
Automatic working traffic signal,
Automatic nonworking Traffic
Signal,
Uncontrolled, Other
Tight curve, Open
Curve, No Curve,
Unknown

Yes, No, Unknown

Vehicle Data Elements
The vehicle data elements describe the characteristics and events of the vehicle(s) involved in the crash. Vehicle Number Unique vehicle
Mandatory
number assigned to identify each vehicle involved in the crash. Vehicle Type
Type of vehicle Mandatory
In crash

Alpha-Numeric

String

Allows the vehicle record to be crossreferenced to the crash record and person records
Bicycle, Other
Non-Motor
Vehicle, 2/3 wheel motor vehicle, passenger car, Bus/coach/trolley,
Light goods vehicle (=3.5t),
Other Motor

P a g e | 63

vehicle, Unknown
Vehicle Male

Indicate the make (distinctive name) assigned by motor vehicle manufacturer

Mandatory if the vehicle is a motorized vehicle. Not applicable to bicycles, tricycles, rickshaws and animal-powered vehicles

Vehicle Model

The code assigned by the manufacturer to denote a family of motor vehicles (within a make) that have a degree of similarity in construction Mandatory if the vehicle is a motorized vehicle. Not applicable to bicycles, tricycles, rickshaws and animal-powered vehicles

Vehicle Model
Year

Year assigned to motor vehicle by manufacturer

Engine Size

The size of the

Mandatory if the vehicle is a motorized vehicle. Not applicable to bicycles, tricycles, rickshaws and animal-powered vehicle
Mandatory, if

Character string.
Alternatively, a list of motor vehicle makes can be
Composed, with a code corresponding to each. Such a list allows for more consistent and reliable recording, as well as for easier interpretation of the data
Character string.
Alternatively, a list of motor vehicle models can be composed, with a code corresponding to each. Such a list allows for more consistent and reliable recording, as well as for easier interpretation of the data
Numeric

Numeric

P a g e | 64

vehicle’s engine is recorded in cubic centimeters

Vehicle Special
Function

Vehicle
Manoeuvre

The type of special function being served by this vehicle regardless of whether the function is marked on the vehicle The controlled manoeuvre for this motor vehicle prior to the crash

vehicle is motorized. Not applicable to bicycles, tricycles, rickshaws and animal-powered vehicles
Mandatory, if vehicle is motorized. Not applicable to bicycles, tricycles, rickshaws and animal-powered vehicle
Mandatory

String

No special function, taxi,
Vehicle used passenger bus,
Police/military,
Emergency,
Other, Unknown

String

Reversing,
Parked, Entering or leaving parking position, Slowing or stopping,
Moving off,
Waiting to turn,
Turning,
Changing Lane,
Avoidance
manoeuvre,
Overtaking
vehicle, Normal,
Other, Unknown

Person data elements
The person data elements describe the characteristics, actions, and consequences relating to the people involved in the crash. These elements are to be completed for every person injured in the crash, and also for the drivers of all vehicles (motorized and non-motorized) involved in the crash. Persona Number

Number
Mandatory
assigned to uniquely identify each person involved in the crash Numeric (2 digit) The persons related to the first (presumed liable) vehicle will be recorded first.
Within a specific vehicle, the

P a g e | 65

Occupant’s
Vehicle Number

Pedestrian’s linked vehicle number Date of Birth

Sex
Type of Road
User

The unique number assigned for this crash to the motor vehicle in which the person was an occupant
The unique number assigned for this crash to the motor vehicle which collided with this person. The vehicle number assigned under to the motor vehicle which collided with this person
Indicates the date of birth of the person involved in the crash Gender of person involved
Indicates role of each person at time of crash

driver will be recorded first, followed by the passengers.
Allows the person record to be crossreferenced to crash, road and vehicle records to establish a unique linkage with the Crash
ID and the
Vehicle
number
Allows crossreference to vehicle records

Mandatory

Numeric

Mandatory

Numeric

Allows crossreference to vehicle records

Mandatory

Numeric

Use for age calculation Mandatory

String

Mandatory

String

Male, Female,
Transgender
Driver,
Passenger,
Pedestrian,

P a g e | 66

Seating Position

Location of
Mandatory for person in vehicle all vehicle occupants String

Injury Severity

The injury severity level for a person involved in the crash Describes the use of occupant restraints, or helmet use by a motorcyclist or bicyclist

Mandatory

String

Mandatory

String

Pedestrian
Manoeuvre

The action of the pedestrian immediately prior to the crash

Mandatory

String

Alcohol Use
Suspected

Law enforcement officer suspects

Mandatory for all drivers of motorized String

Safety
Equipment

Other, Unknown
Subfield Row:
Front, Rear, NA,
Other, Unkown
Subfield Seat:
Left, Middle,
Right, NA,
Other, Unknown
Fatal, Serious
Injury, Slight
Injury

Subfield
Occupant
restraints: Seat belt available and used, seat belt not available, seat belt available and not used,
Child restrain used, Child restrain not present, Child restrain present but not used,
NA, Other restrains, Unknown
Subfield Helmet:
Helmet Worn,
Not Worn, NA,
Unknown
Crossing,
Walking on
Carriageway,
Standing on
Carriageway,
Not on carriageway, Other, Unknown
Yes, No, NA,
Unknown

P a g e | 67

that person involved in the crash has used alcohol

Alcohol Test

Describes alcohol test status, type and result Drug Use

Indication of suspicion or evidence that person involved in the crash has used illicit drug

Driving License
Issue Date

Indicates the date (month and year) of issue of the person’s first driving licence, provisional or full, pertaining to the vehicle they were driving vehicles, recommended for all non-motorists (pedestrians and cyclists) Conditional
(Mandatory if suspected for alcohol use)

Mandatory for all drivers of motorized vehicles, recommended for all non-motorists (pedestrians and cyclists) Mandatory for all drivers of motorized vehicle

String

String

Numeric

Alcohol-related crashes are a major road safety problem.
Information
on alcohol involvement in crashes facilitates evaluation of programs to reduce drink-driving
No evidence,
Suspicion of Us,
Evidence of drug use, NA,
Unknown

Allows calculation of number of years’ driving experience at the time of crash Framework for Index Creation
Theoretical framework
A framework should clearly define the phenomenon to be measured and its sub-components and select individual indicators (and weights) that reflect their relative importance and the

P a g e | 68

dimensions of the overall composite. Ideally, this process would be based on what is desirable to measure and not which indicators are available. And the transparency of the whole exercise is essential in constructing credible indicators.
Variable selection
The strengths and weaknesses of composite indicators largely derive from the quality of the underlying variables. Ideally, variables should be selected on the basis of their relevance, analytical soundness, timeliness, accessibility, etc. While the choice of indicators must be guided by the theoretical framework, the data selection process can be quite subjective as there may be no single definitive set of indicators. Finally, one has to make sure that the type of the selected variables −input, output or process indicators − match the definition of the composite indicator.
Imputation of missing data
Often data sets are not complete. Some countries or some years could lack data on a relevant indicator. Imputation of missing data is the art of filling empty spaces in a data matrix. In general there are three methods for dealing with missing data: I) case deletion, ii) single imputation or iii) multiple imputation. As a rule of thumb, if a variable has more than 5% missing values, cases are not deleted
Multivariate analysis
Unfortunately, individual indicators are sometimes selected in an arbitrary manner with little attention paid to the interrelationships between them. This can lead to indices which overwhelm, confuse and mislead both decision-makers and the general public. The underlying nature and properties of the data need to be carefully analyzed before embarking on the construction of a composite indicator. This preliminary step is helpful in assessing the suitability of the data set and will provide an understanding of the implications of the methodological choices.
Grouping information on individual indicators. The analyst must first decide whether the nested structure of the composite indicator is well-defined and if the set of available individual indicators is sufficient or appropriate to describe the phenomenon. This decision can be based on expert opinion and the statistical structure of the data set. Different analytical approaches, such as principal components analysis, Factor analysis and Cronbach Alpha can be used to explore whether the dimensions of the phenomenon are statistically well- balanced in the composite indicator. If not, a revision of the individual indicators might be needed.
Normalization
Most of the times the individual variables forming the composites have different measurement units. For example GDP can be expressed in Euro, unemployment in number of persons, health in number of diseases, survey data in high, medium, low, or important, not important. In order to avoid summing up apples and oranges then one need to normalize the variables in order to make them comparable. There are many possible normalization ethos

P a g e | 69

and practitioners should take into account data properties, as well as the objectives of the composite indicator in choosing one method rather than another. One has to be aware that different normalization methods will yield different results in composite scores. Therefore, the robustness of the final rankings (scores) should be tested by employing different normalization techniques.
Weighing
Central to the construction of a composite index is the need to combine in a meaningful way different dimensions measured on different scales. This implies a decision on which weighting model will be used and which procedure will be applied to aggregate the information.
Weights should ideally be selected according to a theoretical framework that is established or at least clearly stated
Weighting implies a “subjective” evaluation, which is particularly delicate in case of complex, interrelated and multidimensional phenomena. The menu of weighting methods is rather large and increasing with the creativity of the practitioners. Ideally, weights should reflect the contribution of each indicator to the overall composite. Different weights may be assigned to component series in order to reflect their economic significance (collection costs, coverage, reliability and economic reason), statistical adequacy, cyclical conformity, speed of available data, etc.
There will almost always be some positive correlation between different measures of the same aggregate. Thus, a rule of thumb could be introduced to define a threshold beyond which the correlation is a symptom of double counting.
Weights may also reflect the quality of the data, thus higher weight could be assigned to statistically reliable data (data with low percentages of missing values, large coverage, sound values). In this case the concern is to reward only easy to measure and readily
Available base-indicators, punishing the information that is more problematic to identify and measure. Statistical models such as principal components analysis or factor analysis could be used to group individual indicators. These methods account for the highest variation in the data set, using the smallest possible number of factors that reflect the underlying “statistical” dimension of the data set. Weighting only intervenes to correct for the overlapping information of two or more correlated indicators, and it is not a measure of the theoretical importance of the indicators. Weights, however, cannot be estimated if no correlation exists between indicators. Other statistical methods, such as the benefit of the doubt (BOD) is extremely parsimonious about weighting assumptions
Conjoint analysis derives the worth of the single indicator from the worth of a composite,
i.e. it reverses the process of AHP, with which it shares advantages and disadvantages.

P a g e | 70

Participatory methods that incorporate various stakeholders − experts, citizens and politicians
− can be used to assign weights. This approach is feasible when there is a well-defined basis for a national policy.
Unobserved components is similar in spirit to the multiple regression models but it does not need an explicit value for the “dependent variable” as it treats it like another unknown variable to estimate.
Multiple regression models can handle a large number of indicators. This approach can be applied in cases where the model input are indicators related to various policy actions and the model output is the target.
Aggregation
Weighting is strongly related to how the information conveyed by the different dimensions is aggregated into a composite index. Different aggregation rules are possible.
Individual indicators could be summed up, multiplied or aggregated using nonlinear techniques.
Each technique implies different assumptions and has specific consequences.
Linear aggregation method is useful when all individual indicators have the same measurement unit and further ambiguities due to the scale effects have been neutralized while geometric aggregations(in which indicators are multiplied and weights appear as exponents) are appropriate when non-comparable and strictly positive individual indicators are expressed in different ratio-scales.

Calculation of RSI
The tables are attached in the subsequent pages

per 1k
Index Value
Population
Lakshadweep 0

per 1k
Index Value vehicle Lakshadweep 0

per 1k road
Index
length
Value
Lakshadweep 0

Manipur
Nagaland
Meghalaya

0.004977
0.00709
0.012748

Nagaland
Manipur
Meghalaya

0.109063
0.452385
0.630468

0.411224
0.707956
1.844223

Delhi**
Sikkim
West Bengal

0.018429
0.02117
0.023666

Sikkim
Delhi**
Jharkhand

0.899268
1.100821
1.441575

Nagaland
Manipur
Arunachal
Pradesh
Sikkim
Meghalaya
Tripura

Bihar

0.023837

1.783702

Mizoram

5.745915

Uttar

0.028345

Daman and
Diu
Kerala

1.929828

Assam

6.491159

2.778603
3.153076
4.565035

P a g e | 71

Pradesh
Arunachal
Pradesh
Uttarakhand

0.028752

Goa

1.933817

West Bengal

7.22506

0.039532

2.383267

Uttarakhand

8.116056

Tripura
Jharkhand

0.041996
0.042211

Andaman and Nicobar
Islands
Mizoram
Dadra and
Nagar Haveli

2.476818
2.8913

Odisha
Delhi**

9.991885
10.41372

Andaman and Nicobar
Islands
Jammu and
Kashmir
Assam
Mizoram

0.048168

Chandigarh

2.897077

13.20426

0.050068

Gujarat

3.441806

0.050354
0.051665

3.451858
4.053699

Odisha

0.061655

Tripura
Jammu and
Kashmir
Uttarakhand

4.106993

Andaman and Nicobar
Islands
Uttar
Pradesh
Kerala
Himachal
Pradesh
Chhattisgarh

Chhattisgarh

0.062336

Maharashtra

4.192415

Goa

18.3335

Daman and
Diu
Madhya
Pradesh
Maharashtra

0.066403

Haryana

4.416331

Bihar

18.94025

0.068487

Puducherry

4.588166

Maharashtra

20.25899

0.07401

West Bengal

4.642918

Karnataka

20.60098

Gujarat

0.083379

Assam

5.076203

23.28773

Punjab

0.084163

5.417602

Kerala
Chandigarh
Karnataka
Rajasthan

0.093784
0.094396
0.094957
0.09861

Haryana
Andhra
Pradesh
Puducherry

0.101829
0.103644

Himachal
Pradesh
Karnataka
Tamil Nadu
Odisha
Andhra
Pradesh
Punjab
Chhattisgarh

Jammu and
Kashmir
Madhya
Pradesh
Punjab
Rajasthan
Gujarat
Andhra
Pradesh
Chandigarh
Puducherry

0.103969

Bihar

9.077206

Dadra and
Nagar Haveli

48.01777

Himachal
Pradesh
Dadra and
Nagar Haveli

0.110967

Rajasthan

9.16553

Tamil Nadu

52.67326

0.113163

Uttar
Pradesh

9.297927

Jharkhand

58.21586

5.465497
6.555892
6.590154
6.590796
7.351042
8.60961

14.49604
15.56122
15.86323
16.94324

25.20097
27.69452
28.04071
32.23503
36.86997
43.58934
47.22127

P a g e | 72

Goa

0.133654

Tamil Nadu

Madhya
Pradesh
Arunachal
Pradesh

0.140439

10.64193

Haryana

61.86747

-

Daman and
Diu

68.34753

Table 4: Index Values for each state

Fault of Driver
No. of
Persons

Killed

Total No. of Road
Accidents

Killed

Killed

0

0

0

0

0

0

11492

378

144

876

330

504

217

79

46

23

5

10

4

24

19

Assam

6184

2134

85

42

213

95

0

0

Bihar

6225

3082

159

89

143

83

368

194

419

136

0

0

0

0

0

0

Chhattisgarh

9178

1991

158

36

93

26

682

211

Dadra and
Nagar Haveli

85

53

0

0

0

0

0

0

Daman and
Diu

40

20

0

0

0

0

0

0

0

0

0

0

0

0

0

0

3954

265

65

4

116

22

6

0

23980

6811

605

118

2648

717

69

27

Haryana

7890

3364

118

70

306

132

146

74

Himachal
Pradesh

2662

1027

4

3

15

3

15

7

Jammu and
Kashmir

4350

776

6

1

335

0

69

14

Jharkhand

3374

1769

120

45

182

177

266

101

Karnataka

36545

7625

259

77

837

236

477

133

Kerala

36142

4277

7

2

21

7

0

0

Andaman and Nicobar
Islands
Andhra
Pradesh
Arunachal
Pradesh

Chandigarh

Delhi**
Goa
Gujarat

Lakshadweep

Killed

Total No. of Road
Accidents

236

25

33222

Defect in Condition of
Motor Vehicle
No. of
Persons

Fault of Pedestrian
No. of
Persons
Total No. of Road
Accidents

States/UTs

Total No. of Road
Accidents

Fault of Cyclist
No. of
Persons

3

0

0

0

0

0

0

0

Madhya
Pradesh

40975

6532

425

61

847

116

1112

204

Maharashtra

56418

11087

447

171

1991

973

442

54

0

0

0

0

0

0

211

62

Manipur

P a g e | 73

Meghalaya

23

26

0

0

22

23

117

44

110

77

0

0

0

0

0

0

5

11

0

0

0

0

6

10

Odisha

8781

3511

143

48

90

34

33

12

Puducherry

1125

172

0

0

0

0

0

0

Punjab

3804

2839

125

103

82

76

163

106

21939

9181

12

5

21

9

28

7

11

12

0

0

0

0

0

0

57507

13453

735

186

1495

685

107

53

705

201

5

3

42

12

20

5

12759

6164

1803

1147

1313

742

1482

870

Uttarakhand

1013

471

3

3

9

2

21

25

West Bengal

6191

2581

179

86

478

233

669

290

Mizoram
Nagaland

Rajasthan
Sikkim
Tamil Nadu
Tripura
Uttar
Pradesh

Table 5: Data related to causes mentioned

Defect in Road Condition
No. of
Persons

Killed

Killed

0

0

0

0

580

392

196

5813

2005

23

14

15

20

77

30

0

0

0

0

53

20

283

155

457

218

2685

1235

0

0

0

0

0

0

Chhattisgarh

316

74

243

63

2841

766

Dadra and
Nagar Haveli

0

0

0

0

0

0

Daman and
Diu

0

0

0

0

10

9

Delhi**

0

0

0

0

6937

1866

Goa

1

0

4

1

166

0

32

9

37

6

578

129

142

55

131

88

1332

663

34

18

0

0

169

51

Andaman and
Nicobar
Islands
Andhra
Pradesh
Arunachal
Pradesh
Assam
Bihar
Chandigarh

Gujarat
Haryana
Himachal
Pradesh

Killed

Total No. of Road
Accidents

0

0

1339

All Other Causes*
No. of
Persons
Total No. of Road
Accidents

States/UTs

Total No. of Road
Accidents

Weather Condition
No. of
Persons

P a g e | 74

Jammu and
Kashmir

28

10

6

3

1915

361

Jharkhand

239

103

252

113

1278

510

Karnataka

241

57

263

56

5826

1264

Kerala

0

0

0

0

4

0

Lakshadweep

0

0

0

0

0

0

Madhya
Pradesh

818

131

365

54

6668

1077

Maharashtra

315

61

21

8

6682

979

Manipur

102

23

18

1

440

72

32

12

37

13

252

101

Mizoram

0

0

0

0

0

0

Nagaland

0

0

0

0

31

35

19

12

10

2

209

82

0

0

0

0

56

21

92

77

162

147

1913

1472

209

47

30

11

730

268

0

0

16

20

131

23

1072

219

291

65

6550

1514

0

0

37

13

79

38

756

345

824

412

11035

6469

Uttarakhand

43

15

0

0

383

328

West Bengal

660

329

919

390

3194

1488

Meghalaya

Odisha
Puducherry
Punjab
Rajasthan
Sikkim
Tamil Nadu
Tripura
Uttar Pradesh

Table 6: Data related to causes mentioned