不動產投資信託分析師盈餘預測與房價指數變動之前後關聯性探討

Abstract

本研究討論不動產投資信託證券分析師的盈餘預測與房價指數變動之領先落後關係，探討：(1)分析師盈餘預測資訊對未來房價報酬的預測能力；(2)分析師所作盈餘預測受到前期房價增減率的影響程度，以及(3)除其受前期房價增減率解釋部分，分析師預測殘差對未來房價增減率關聯顯著性。本文除了使用過去盈餘預測研究者常檢視之盈餘預測修正值作為估計變數，基於對長期房價預估值應屬於多期累積分析所得，故亦納入當期預測之水準值，觀察分析師預測力以減少偏誤。就分析師1~7年期盈餘預測預測力的議題，本研究以最小平方法回歸進行探討；而針對1~3年期預測受影響程度議題，本文進一步採單根檢定檢視序列資料狀態，依結果分成平穩與非平穩序列，分別進行因果關係檢定及向量自我回歸檢定流程；4~7年期預測則採最小平方法回歸檢驗。研究發現：(1)盈餘預測水準值顯示分析師具預測房價指數變動能力；(2)非住宅型的偏門REITs分析師具顯著預測力；(3)分析師預測也會受到過去房價變化影響，尤其接近住宅型不動產投資信託的預測更為顯著。但隨預測年期變長，預測力和受影響程度趨於不顯著。

This paper investigates the lead-lag relationship between real estate investment trust earnings forecasts issued by security analysts and change of house price index, showing (1) the extent to which these forecasts add to predicting future U.S. house prices, (2) the extent that these analysts’ forecasts are affected by previously reported house price changes, and (3) the significance of analysts’ forecast revisions as a predictor for future change of house prices. To enhance the power of tests in identifying analysts’ forecast performance, this study not only adopts forecast revisions, which typically serve as a predictor in prior studies, but also incorporates the level of forecast into the analysis. Specifically, for the topic on forecast power for 1 to 7 years, we analyze them with Ordinary Least Square regressions. Relatively, as for the topic on the extent of that analyst one- to three-year ahead forecasts are affected by prior house price changes, we further use Unit Root Test to see whether the time series are stationary or nonstationary, and then put both results into Pairwise Granger Causality Test and Vector Autoregression Test respectively; as for four- to seven-year ahead forecasts, we conduct Ordinary Least Square regression. Interestingly, we find that: (1) the level of analyst forecasts have forecast power for future house prices; (2) analysts for non-residential sector REIT have statistically significant forecast power; (3) REIT Analyst forecasts appear to be affected by previous house price changes, especially in the property-sector of apartment. However, the results of both forecast power and the extent of analysts’ being reactive to previous house price changes become less significant statistically for longer horizons.