Reducing revisions in house price indices by use of nowcasting models

Abstract

National Statistical institutes (NSIs) must balance between timeliness and accuracy of the indicators they publish. Because some of the house sales transactions are reported several months after they occur, many countries, including Israel, publish provisional house price indices (HPIs), which are subject to large revisions as further transactions are reported. This happens because the late reported transactions behave differently from the transactions reported on time. In this paper we propose a novel methodology for minimizing the size of the revisions, with illustrations from Israel, but the method can be applied to other countries with appropriate modifications. The proposed methodology consists of nowcasting three types of variables at a sub-district level, and adding them as input data to an extended hedonic model used for the computation of the HPI. (1) the average characteristics of the late-reported transactions such as the average number of rooms and the area size of the sold apartments; (2) the average price of the late-reported transactions; (3) the number of late-reported transactions. The three variables are nowcasted based on models fitted to data from previous months. Evaluation of our methodology shows more than 50% reduction in the magnitude of the revisions.

More information

Identifiers

Catalogue record

Export record

Altmetrics