You are on page 1of 8

Structural

Equation
Modeling
Analysis of APSRTC
Performance

Submitted to
Prof: Sushil Kumar

Submitted by
Brahmananda Rao Peddiboyina
fpm 10007

INDIAN INSTITUTE OF MANAGEMENT


LUCKNOW
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

Contents
INTRODUCTION .....................................................................................................................3

DEFINING THE PARAMETERS ..............................................................................................3

THE GENERAL MODEL..........................................................................................................4

THE LIMITED FACTORS MODEL...........................................................................................5

THE INITIAL PROGRAM .........................................................................................................5

THE FINAL PROGRAM...........................................................................................................6

FINAL OUTPUT ......................................................................................................................7

THE FINAL PATH DIAGRAM ..................................................................................................8

ANALYSIS OF RESULTS........................................................................................................8

2
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

APSRTC PERFORMANCE

INTRODUCTION
I have gathered data on the performance parameters of APSRTC for the last 5 years. APSRTC
has 23 Regions. Each Region caters to the transportation needs of one district. Therefore data
is available region wise. Data is available on each parameter for the last 5 years i.e., from 2004-
05 to 2008-09, both years including. Therefore I could collect 110 (23x5) observations for each
parameter.

DEFINING THE PARAMETERS


Data is gathered on 24 parameters. The parameters and their significance is explained
hereunder:

FLEETAVG: Fleet Average explains the number of buses held by the Region

SCHS: Schedules explains the number of planned buses operated by the Region

FLEETUTN: Fleet Utilization explains how well the Region utilizes the buses without keeping
them off road.

OPTDKMS: This is Number of Kilometers operated by the region

VEH UTN: Vehicle Utilization is given by dividing the Operated Kms by the number of buses.

EARNINGS: Gives the revenue earned in rupees

EPK : Earnings Per Kilometer is obtained by dividing the Earnings by Operated Kms.

EPB : Earnings Per Bus is obtained by dividing the Earnings by No. of Buses

OR : Occupancy Ratio explains how much percentage of seats are occupied by


passengers It is given by formula OR = EPKX100/Seating Capacity X Average
Basic Fare
BDRATE: Break Down Rate explains how many breakdowns are happening for 10000 kms
of operation.
%CANC: Percentage of Cancellation explains what percentage of kms are cancelled out of
planned kms of operation.

%MECHCAN: Percentage of Mechanical Cancellation explains what percentage of kms are


cancelled due to Mechanical Reasons out of total cancelled kilometers.

%CRECANC: Percentage of Crew Cancellation explains what percentage of kms are cancelled
due to absence/late reporting of crew out of total cancelled kilometers.

3
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

HSDKMPL: High Speed Diesel KiloMeters Per Liter explains how many kilometers are run by
the bus by consuming one liter of high speed diesel oil.

LUBKMPL: Lubricating Oil Kilometer Per Liter explains how many kilometers are run by the
bus by consuming one liter of Lubricating Oil.

TYRELIFE: Explains the kilometers run by the tyre before it is scrapped.

%NTS : Percentage of New Tyre Scrap explains how many new tyres are scrapped out of
total new tyres received.

RTFACTOR: Retreadability Factor indicates how many retreads a tyre has undergone before it
is scrapped.

BSR : Bus Staff Ratio explains how many staff are employed per bus operated.

CREWUTN: Crew Utilization explains how many kilometers are operated by one crew i.e. by
one driver or conductor.

EMPPROD: Employee Productivity is obtained by dividing operated kilometers by available


staff in a depot.

GROSSEPK: Gross Earnings Per Kilometer is obtained by dividing the Gross Earnings of the
depot (which includes stall rents besides ticket revenue) by operated kilometers.

GROSSCPK Gross Cost Per Kilometer gives the total cost of operation per kilometer.

PPK: Profit Per Kilometer is the difference between GROSSEPK and GROSSCPK.

THE GENERAL MODEL


Using Data Analysis tools, I have obtained both Correlation and Covariance matrices.

Using Correlation matrix I have tried to develop a model by specifying the following paths:

Observed Variables: FLEETAVG SCHS FLEETUTN OPTDKMS VEH UTN EARNINGS EPK
EPB OR BDRATE %CANC %MECHCAN %CRECANC HSDKMPL LUBKMPL TYRELIFE
%NTS RTFACTOR BSR CREWUTN EMPPROD GROSSEPK GROSSCPK PPK.

Latent Variables: Performance, Operformance, Mperformance, Fperformance. (O=Operational ,


M = Mechanical, F = Financial).

4
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

Relationships:

Perfor = Fperfor Operfor Mperfor

BSR CREWUTN EMPPROD GROSSEPK GROSSCPK PPK = Fperfor

FLEETAVG SCHS FLEETUTN OPTDKMS VEHUTN EARNINGS EPK EPB OR = Operfor

BDRATE %CANC %MECHCAN %CRECANC HSDKMPL LUBKMPL TYRELIFE %NTS


RTFACTOR = Mperfor

However, the model did not converge as the matrix is NOT positive definite.

Many of the variables are inter related. A look at the formulae to calculate these observed
variables show that Operated Kilometers, Earnings are common in many parameters.

Therefore, I have selected the following parameters which are not related to each other , yet has
good correlation with ppk.

THE LIMITED FACTORS MODEL


EPK HSDKMPL TYRELIFE GROSSCPK.

THE INITIAL PROGRAM


The initial program is as follows:

APSRTC PERFORMANCE
Observed Variables
EPK HSDKMPL RTFACTOR GROSSCPK
Correlation Matrix:
1.00
-0.61 1.00
-0.49 0.08 1.00
0.91 -0.62 -0.36 1.00
Means:
1423.86 5.30 3.22 1651.94
Standard Deviations:
164.91 0.17 0.50 165.50
Sample Size = 110
Latent Variable: Perform
Relationships:
EPK HSDKMPL RTFACTOR GROSSCPK = Perform
Path Diagram
End of Problem.

5
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

The goodness of fit statistics are as follows:


Chi Square df p-value RMSEA
20.90 2 0.00003 0.294

To improve the model, the modifications suggested by the LISREL software are looked at and in
the second iteration the variances between RTFACTOR and EPK are set free. The goodness
of fit statistics are as follows:
Chi Square df p-value RMSEA
80.37 3 0.00000 .0486

In the third iteration, the variances between RTFACTOR and HSDKMPL are set free.
Chi Square df p-value RMSEA
89.77 4 0.00000 0.444

In the fourth iteration the variances between GROSSCPK and HSDKMPL are set free.
Chi Square df p-value RMSEA
326.95 5 0.00000 0.769

THE FINAL PROGRAM


APSRTC PERFORMANCE

Observed Variables

EPK HSDKMPL RTFACTOR GROSSCPK

Correlation Matrix:

1.00

-0.61 1.00

-0.49 0.08 1.00

0.91 -0.62 -0.36 1.00

Means:

1423.86 5.30 3.22 1651.94

Standard Deviations:

164.91 0.17 0.50 165.50

6
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

Sample Size = 110

Latent Variable: Perform

Relationships:

EPK HSDKMPL RTFACTOR GROSSCPK = Perform

Set the Variances of HSDKMPL RTFACTOR Free

Set the Variances of RTFACTOR EPK Free

Set the Variances of RTFACTOR GROSSCPK Free

Path Diagram

End of Problem

FINAL OUTPUT
The final output is as follows:
Measurement Equations
EPK = 164.91*Perform,, R² = 1.00
(11.17)
14.76

HSDKMPL = - 0.10*Perform, Errorvar.= 1569.55, R² = 0.00


(3.79) (122.75)
-0.027 12.79

RTFACTOR = - 0.24*Perform, Errorvar.= 1569.55, R² = 0.00


(3.79) (122.75)
-0.065 12.79

GROSSCPK = 150.60*Perform, Errorvar.= 1569.55, R² = 0.94


(10.88) (122.75)
13.84 12.79

7
Brahmananda Rao Peddiboyina fpm 10007 SEM PROJECT

THE FINAL PATH DIAGRAM


The final path diagram is as follows:

ANALYSIS OF RESULTS
From the path diagram and results, it appears only EPK and GROSSCPK explain the variance
in Performance better. Though APSRTC puts maximum stress on HSDKMPL, it is evident that
this parameter does not explain the variance in performance. The R-square is zero. Similar is
the case with RTFACTOR. It does not explain the variance in the performance.
We may conclude that the Earning parameters explain the variance in performance. The
Mechanical Parameters may not be explaining the variance in performance, but they may be
good indicators of the health of the fleet.
If a separate study is conducted taking ‘fleet health’ as the latent variable, the usefulness of
mechanical parameters might be found.

You might also like