A modified storey enclosure model

Abstract

James’ Storey Enclosure Method (JSEM), developed in 1954, is considered by many to be the most sophisticated single-rate method ever devised for early-design-stage tender price forecasts. However, the method is seldom used in practice partly because it has been superseded by multi-rate methods (such as the elemental method) and partly due to the arbitrary nature of the weightings prescribed for its use. This paper describes the further development of the approach, and in which empirical values of the weightings are derived by multivariate regression analysis.

A set of 50 completed Hong Kong private housing projects is used to demonstrate the use of the technique. This involves, firstly, the modification of the variables used in the original JSEM to incorporate the special characteristics of Hong Kong multi-storey residential buildings. This results in what is termed here as a Modified James’ Storey Enclosure Model (MJSEM). Next, the optimal number of variables for inclusion in the model is identified by means of a dual stepwise cross validation regression procedure - resulting in a Regressed Modified Model for James’ Storey Enclosure Method (RMJSEM). Also, using an amended version of MJSEM, the dual stepwise cross validation regression is used to produce a Regressed Modified Model for Amended Storey Enclosure Method (RMASEM).

The forecasting accuracy of RMJSEM and RMASEM is then compared with that of MJSEM together with the floor area and cube method to provide an indication of the improvement achieved. It is shown that the RMASEM provides significantly more consistent forecasts than the MJSEM and floor area models, leading to the conclusion that RMASEM may be the best model.