| description abstract | The accuracy of cost estimates is crucial to the success of construction projects with knowledge of project location being fundamental to the outcome of cost estimating. Location factors are commonly used to adjust construction cost estimates by project location, but these factors are not available for all locations. To date, the nearest neighbor (NN) method, based on simple, proximity-based interpolation has been used when a city lacks a sampled location factor even if this method was not empirically validated. This paper provides several contributions. First, the validity of the proximity-based method was substantially supported through the analysis of the global spatial autocorrelation of the changes in location factors from 2005 to 2009. Using two established cost index databases, it was found that there is spatial correlation between cost indexes for specific cities. Therefore, cost estimators can spatially interpolate values to understand cost indexes at places that are not listed in the database. Second, comparing different spatial interpolation techniques, including NN for how well they can interpolate cost indexes across space revealed that two alternative methods, the conditional nearest neighbor (CNN) and the inverse distance weighted (IDW) methods, lead to more accurate estimates. These findings are relevant to industry practitioners because they justify the use of interpolation as well as identify how to obtain improvements in preliminary estimates by simply replacing an interpolation method with another. Last, this study also provided preliminary evidence supporting the development of spatial prediction models for construction costs, which is an additional point of departure from the existing body of knowledge. | |
