ASSESSING ANNUAL VALUE (OF RENTAL HIGH-RISE RESIDENTIAL APARTMENTS) BY MULTIPLE REGRESSION

Abstract

Practically all the past work done on the application of multiple regression analysis for valuation or assessment purposes, both locally and overseas, have made use of stepwise regression programs in selecting variables to be included in the model. Such an automatic variable selection procedure is fraught with many pitfalls, the most notible of which is the fact that a statistical test (F-value) is the sole criterion used in the selection process. This dissertation is concerned with assessing the Annual Value of high-rise residential apartments- It will propose a method of selecting variables where the findings of consumer research and the valuer's perception of value can be blended with regression theories to arrive at the 'optimum' regression analysis model for mass-assessment. Such an 'optimum' model will be reflective of both the true real estate market and the valuer's perception of value and is thus conceptually superior ; it will also be useful in explanation, apart from prediction, once the order of entry of each variable is determined.