摘要 : Abstract This study aims to examine whether US consumer home-buying assessments can potentially help reduce the random walk prediction errors of mortgage rates. Forecast evaluations under flexible loss reveal that the random walk ... 展开 Abstract This study aims to examine whether US consumer home-buying assessments can potentially help reduce the random walk prediction errors of mortgage rates. Forecast evaluations under flexible loss reveal that the random walk predictions for 1992–2020 imply asymmetric loss, meaning that they are of value to a user who assigns more (less) cost to over-predictions (under-predictions). Further results indicate that such survey measures as consumer home-buying attitudes and consumer opinion about interest rates can help improve the accuracy of the random walk predictions of mortgage rates. As such, we recommend that forecasters consider using such survey measures in predicting mortgage rates. 收起