Hypothetical Modeling of Contractor’s Bid Markup Estimation for Road Construction Projects in Ethiopia
No Thumbnail Available
Date
2020-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Markup is a factor that estimators apply to certain work activities or to the total cost
of a bid to cover general overhead and profit. Estimating markup is an important decision for
contractors as its size has to be low enough to win a contract, but high enough to make a
profit. Studies which are done on cost estimation practice in Ethiopian construction industry
agree in the need for a change in the current practice of bid markup estimation. But these
studies have a gap in introducing a systematic tool for solving the problem. This research
focuses on identifying and analyzing factors affecting bid markup in road projects and
developing a model which will support local contractors’ decision in estimating bid markup
size for road construction projects. The research uses integrated review of various literatures
and questionnaire survey as data collection methods.
Twenty-one factors that are considered to affect bid markup have been identified from
literature review. Based on the results of the analysis, it appears that ‘complexity of project’,
‘number of competitors with strong desire to win a project’, ‘project location (region)’,
‘immediate need for work’ and ‘security need of project location’, are the top five ranked
factors in terms of influencing bid markup size.
Multiple linear regression (MLR) and artificial neural network (ANN) were selected
for modeling bid markup estimation. The developed MLR equation contains eleven factors
based on stepwise regression technique. The coefficient of correlation (R) was obtained as
0.882 with adjusted value of coefficient of determination (R
) = 0.745. The overall regression
model was statistically significant, F (11, 73) = 23.297, p< 0.05. For developing the ANN
model, various network structures were generated and tested. The most satisfactory model
was the ANN8, which consists of 8 neurons in the hidden layer with R, R
2
2
, MAPE and RMSE
values of 92.06%, 84.75%, 6.43% and 2.47 respectively. Cross validation for both models
was done and satisfactory result was obtained.
Statistical performance indicators shows that the ANN method of modeling predicts
bid markup better than MLR method. But, the obtained values of the statistical performance
indicators for the two models are closer to each other. Thus, both models can be considered
as a satisfactory prediction tools for bid markup and can provide a starting point for
estimators in a given road project bid markup estimation task.
Description
Keywords
Markup, Multiple linear regression, Artificial neural network