This paper re-examines the statistical properties of five different price index methods with the objective of identifying one that is most accurate and robust when estimated at frequent time intervals and/or for highly localized markets. This paper adopts the split sample technique to allow for a consistent basis upon which the price index methods are compared. The results show that: 1) when estimating price indexes, the arbitrary pooling of data across time and geography is not warranted, 2) the hedonic imputation model outperforms the alternative models on all measures of accuracy and robustness, 3) the differences in the levels of accuracy among the models are statistically significant and 4) there are significant economic disadvantages associated with using a sub-optimal price index method in decision-making.