GOVERNMENT RESIDENTIAL LAND SALES (YEAR 1989-1996) : DEVELOPING A BID PRICE MODEL

Abstract

This study formulates a bid price model and a subset model for the residential land bids. In order to determine the basis for the selection of variables, a detailed review from theoretical studies is done. A set of eight macro variables is then selected for the construction of a multiple regression model. Using two variable selection methods, some variables are identified as being the most significant contributors to the variation of residential land bids under the subset model. The findings of this study indicate that economic conditions and price movement of private residential properties constitute 41.44% of the total variability of the land bids during the period between the fourth quarter of 1989 and second quarter of 1996. Finally, the Gross Domestic Product has a negative statistical relationship with the bid price which theoretically may not be logical. This is due to the strong correlational structure among the explanatory variables. Therefore, this may be a limitation of the subset model used in the study. Despite the fact that large number of variables have been excluded from the final model, there is only a slight reduction of the overall mean error value from 41% to 40%. Due to the existence of multicollinearity, the future development of a bid price model will be the use of ridge regression.