Forecasting commercial property rentals in South Africa

Abstract

ABSTRACT The commercial property sector is of considerable importance to the South African economy, yet despite this, there appears to be relatively little published literature on the local market. This study identifies and tests a model of aggregate commercial property rentals in South Africa. Drawing on literature from studies conducted around the world, this paper identifies the most relevant predictors of commercial property rentals and attempts to utilise them in producing a forecast. Predictors are sorted according to Archer and Lings’ (1997) three-market framework of space, property and capital. Variables identified for the space market include, GDP and unemployment. Property market variables include, construction costs, plans passed and vacancy rates. The capital market was represented by prime interest rates. Secondary time series data were obtained from various well established sources for a seventeen year period. A single equation regression model was employed through the PROC Autoreg function in SAS. Allowing for lags of between zero and three years, all permutations are considered through the running of 2 731 regressions. A further 12 607 regressions were run to evaluate which variables to include in the model. The final model selected included GDP and plans passed, both with a lag of three years, building costs with a lag of two years and vacancy rates lagged for one year. The unemployment and prime lending rate variables were excluded from the model. The final equation was able to account for 98 percent of the variation in gross rent for the period of 1996 to 2012. Of the explanatory variables, building costs exhibited the largest effect on rentals and showed a positive relationship. Building plans passed and vacancy rates exhibited the second and third largest effect on rentals respectively and both showed a negative relationship. The GDP variable had the smallest relative impact on rentals and showed a positive relationship.

When forecasting, the model offers reasonable accuracy and was able to capture the general trend in gross rentals, with only minor deviations from the actual price levels while prices were trending. In all but two instances, the model was able to predict the turning point in rentals. While not without its limitations, the model could serve as a useful tool in the structuring of leases or for better time portfolio allocations. More importantly, the model could be built on to enhance the body of knowledge relating to the South African commercial property industry.