Predictive Accuracy of Logit Regression for Data-Scarce Developing Markets: A Nigeria and South Africa Study

Abstract

This research examines how much forecasting accuracy can be achieved by modelling the relationships between listed real estate and macroeconomic time series variables using the logit regression model. The example data for this analysis included 10-year (2008–2018) transactions. The Statistical Package for Social Sciences (SPSS, version 25) and Microsoft Excel 2016 were used for descriptive and inferential analysis. The data collected on the listed real estate transactions for South Africa and Nigeria represent the largest listed real estate markets in the continent. The study found that 22.2% variance in the Nigerian real estate market was explained by the lending rate, treasure bill rate, and Consumer Price Index, while 9.4% variance in the South African real estate market was explained by changes in the exchange rate and coincident indicators. The strength and similarity of the model capacity in both countries showed that each market signal has a predictive accuracy of 75% (Nigeria) and 80% (South Africa).

Keywords

• economic leading indicators
• real estate
• forecasting
• investment
• market modelling