A study on house price index performance: Mix adjustment and hierarchical linear growth repeat-sales models

In this study, we examined the differences between three house price indexes constructed using hedonic price, mix adjustment, and hierarchical linear growth repeat-sales modeling. The data consisted of housing sales across 13 administrative districts in Kaohsiung City from the third quarter of 2013 to 2022. The predictions were compared using the mean standard error, mean absolute percentage error, mean absolute error, and root-mean-square error. The results revealed that the hedonic price index performed the best; its prediction scores, as reflected by the four aforementioned metrics were 0.072, 1.176, 0.181, and 0.181, respectively. The index with the second best performance was the mix adjustment model, with scores of 0.154, 1.905, 0.293, and 0.293. The worst-performing index was the repeat-sales model, with scores of 0.309, 2.804, 0.439, and 0.439. After comparing the annual prediction errors of the three models, it became apparent that the hedonic price index had the best performance, followed by the mix adjustment index, and then the hierarchical linear growth repeat-sales index.