Developing a construction cost function for residential buildings in Dhaka city- an econometric approach

Abstract

The preliminary cost estimate of a new building project is very significant which provides the basis for the clients' budgeting, funding and controlling the project costs. It is also the starting point on which the stakeholders decide whether to accept or reject the project in question. A cost model should represent the significant cost items in a form which will allow analysis and prediction of cost according to changes in the design variables and price of cost elements. Only then it can be utilized in the decision-making process. Considering the above fact the main objectives of this research is to identify the possible cost elements and developing a general cost function for residential building at Dhaka city. The developed model is then validated to be useful for future study in this regards. For the above study primary and secondary data were collected from the developers of Dhaka city and Bangladesh Bureau of Statistics respectively. Total three models were formulated using multiple linear regression (MLR) on 85 data with 26 independent variables (IV) and construction cost per sq. ft as dependent variable (DV). Model-1 integrated construction materials' cost and labour wage and explained 91.4% of variability with standard error 65.461. Model-2 incorporated only design variables as IV and explained only 32.9% variability with standard error 185.938. Model-3 took account of all the variables explaining an increase of 0.5% of variability only. Sand and Paint cost with Mason wage could describe the construction cost, whereas design related variables displayed little influence on the DV. Models and variables were statistically significant below 5% level. All models met MLR assumptions and found suitable after cross validation and sensitivity analysis. Hence it is concluded that model with materials' cost and wage explain the DV better. However, a discrepancy is observed here, as steel and cement were not found statistically significant whereas, these cause the maximum cost in reality. Conversely, sand and paint being smaller in cost is contributing in these models. Foundation and Structural systems is more important cost contributor but these are not statically significant in our case. Concrete Strength, Steel Grade did not show desired indication as in reality. This discrepancy might be because of data being collected from developers of various standards and in different time frames, when sudden rise and fall of the materials' cost took place. At the end, it could be noted that an estimated project cost is not directly calculated from project components rather an approximate indication of the cost derived from minimum possible number of variables.