Estimating cost at the conceptual phase of highway earthwork construction projects is challenging owing to the limited amount of information present at the initial stages. Methods that have been applied in the past to perform such estimates are regression, mathematical equations, artificial intelligence, and others. The objective of this research is to use two spatial statistical methods – indicator kriging and empirical Bayesian kriging – together with exponential, whittle, and k-Bessel semi-variograms to build an appropriate model to predict highway earthwork construction cost at the conceptual stage. Iowa Department of Transportation provided the data for the study. The cross-validation results were used to compare the performance of the kriging algorithms. It was observed that the exponential semi-variogram showed better performance for both types of kriging when they were analyzed separately. However, indicator kriging results were more accurate than empirical Bayesian kriging when they were compared for the same type of semi-variogram.