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 that include Israel, publish provisional house price indices (HPIs) that 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 to minimize 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 subdistrict 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; and (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.

国家统计机构 (NSI) 必须在其发布指标的及时性和准确性之间取得平衡。由于一些房屋销售交易是在发生数月后报告的，因此包括以色列在内的许多国家都会发布临时房屋价格指数 (HPI)，这些指数会随着进一步交易的报告进行大幅修订。发生这种情况是因为延迟报告的交易行为与按时报告的交易行为不同。在本文中，我们提出了一种新颖的方法来最小化修订的大小，并附有以色列的插图，但该方法可以通过适当的修改应用于其他国家。提议的方法包括在街道级别对三种类型的变量进行临近预测，并将它们作为输入数据添加到用于计算 HPI 的扩展特征模型中：(1) 延迟报告交易的平均特征，例如平均数量房间数量和出售公寓的面积；（二）迟报交易的均价；(3) 逾期报告的交易数量。这三个变量是基于拟合前几个月数据的模型进行即时预测的。对我们方法的评估表明，修订的幅度减少了 50% 以上。(3) 逾期报告的交易数量。这三个变量是基于拟合前几个月数据的模型进行即时预测的。对我们方法的评估表明，修订的幅度减少了 50% 以上。(3) 逾期报告的交易数量。这三个变量是基于拟合前几个月数据的模型进行即时预测的。对我们方法的评估表明，修订的幅度减少了 50% 以上。