Improved method for heritage property valuation

Heritage property value can be used as part of an economic indicator. This underlines the importance of having an accurate value of heritage property for better-informed decision-making. The appropriate approach to valuation should produce an accurate value which is reliable and practical. There are many methods that can be used to estimate the value of heritage property including sale comparison method, cost method, contingent valuation method (CVM) and regression models. Recent literature has shown evidence of increasing use of the last two methods in advanced countries. However, there has been no conclusive evidence as to what is the appropriate method of valuation for heritage property. This study aims to determine the appropriate approach for heritage property valuation by testing the applicability of multiple regression analysis (MRA), rank transformation regression (RTR) and contingent valuation method (CVM) in assessing the value of private heritage property using pre-war shophouses in Kota Bharu, Kelantan, Malaysia as a case study. The methods are selected for this study due to their relative acceptance and potential as being indicated in the literature. Empirically, data on 25 pre-war shophouses in the study area transacted from 2004 to 2014 were used to construct the regression models of MRA and RTR. In doing so, 10 independent variables and three functional forms of linear, semi-log and log-log were tested in order to select the best-fit model for each method. Secondly, self- administered questionnaire survey was conducted among 100 owners and tenants of pre-war shophouses in the case study to assess the applicability of CVM. The survey has introduced a new section of monetary valuation that has not been evident in previous studies of CVM. In order to determine the best method, the statistical performance of the models was compared by focusing on Mean Absolute Percentage Error (MAPE) values. The results indicate that the linear functional form suits the data best for MRA method. Meanwhile, the log-log functional form suits the data best for the RTR method. The CVM and MRA models produced lower MAPE than RTR (6.08%, 7.55% and 73.84% respectively). Except for RTR, the MAPE values for CVM and MRA fall within the range, ±10% differences from the baseline value of market price. Unfavourably, the RTR’s MAPE value exceeds the parameter. However, the idea of comparing between MRA, RTR and CVM needs to be viewed more thoroughly in future studies because these methods differ in functional forms and explanatory variables. Notwithstanding, after considering the reliability and practicality aspects, it is fair to conclude that using these three methods simultaneously for cross-checking constitutes appropriate approach for private heritage property valuation.