A new recession-dating algorithm for South Africa

The SA Reserve Bank (SARB) regularly determines the upper and lower turning points of the South African business cycle, but this is only completed after all the relevant information has been obtained, confirmed and analysed, causing a lengthy time lag between the actual determination and the event. The current research aimed to design a recession-dating algorithm, which could allow the Bureau for Economic Research (BER) to make accurate calls on business cycle turning points substantially sooner after the event than is the case with the official SARB determination, which typically lags actual turning points by 18 to 24 months. The proposed algorithm includes, as a point of departure, the advance signals given by the yield spread (between 3-month and 10-year government bonds), as well as a consideration of the local moments of five high-frequency economic time series. The turning point signals provided by these indicators (and after the application of censoring rules) are integrated by reconciling the differences through the use of the median date in order to derive true business cycle turning points. The algorithm was tested for the five recessions experienced over the 1981 to 2013 period. It was found that the algorithm could be applied successfully in calling the business cycle turning points over this 32-year period avoiding any false positives. A high degree of accuracy was also obtained, i.e. a median two month lag in respect of upper turning points (or peaks) of the SARB-determined business cycle and a one month lead in respect of lower turning points (i.e. troughs). The algorithm will not only allow the BER to make close calls on business cycle turning points, it will be able to do this with a much shorter time delay following actual turning points compared to the SARB’s official determination.

C41, E32

Business cycles, turning points, quantitative analysis of business cycles