The validity of the triangular distribution assumption in Monte Carlo simulation of construction costs: empirical evidence from Hong Kong

Abstract

This paper investigates the validity of the triangular distribution assumption which is commonly adopted in Monte Carlo simulations of construction costs. The study begins with an examination of the asymmetric nature of the distribution of construction costs and deduces theoretically that the triangular distribution assumption leads to an upward bias in the probability of exceeding the conventional single figure estimate for the subsystem variables and therefore the system variable. This assumption is also inconsistent with the estimators' subjective perception. An experiment has been performed to generate empirical data that test the above theoretical arguments and assesses the magnitude of the bias, if it indeed exists. Subjective estimates of the construction costs of the ten major subsystems of electrical services contracts of government clinics in Hong Kong are examined. These estimates are extracted from seven experienced estimators. The results of the analysis of the data confirm that the underlying distribution of the subsystem costs is asymmetric with a long thin tail towards the right and that the triangular distribution assumption does in fact lead to bias in the subsystem input variables and therefore the simulated output system costs. An alternative modelling approach which can reduce the bias has also been outlined in this paper.

Keywords:

• Monte Carlo simulation
• probability density function
• risk analysis
• subjective estimates
• triangular distribution