The predictability of general election trends on the basis of by‐election results in South Africa

Abstract

This article deals with the possibility of predicting the outcome of general elections on the basis of the results of by‐elections, with specific reference to a technique proposed and used by Professor Kleynhans. Three questions are dealt with. Firstly can by‐elections be used as valid samples of voters’ preferences regarding the different political parties; if so, secondly, can the by‐election prediction technique detect the correct trend in the change in the voters’ mood and lastly, if the trend is predicted correctly, does the prediction fall within acceptable limits of accuracy? The main conclusion reached is that the by‐election prediction technique tends to be general election‐by‐election trend biased in the sense that it cannot detect a change in the voters’ mood between the last by‐election and the forthcoming general election. To this extent it may fail to predict the general trend of the change in the voters’ mood correctly. Even in cases where the trend is correctly predicted, the accuracy of the prediction is, judged by past performance, not reliable enough. On the basis of qualitative reasoning it is argued that by‐elections in South Africa are not comparable to general elections. The by‐election prediction technique therefore does not seem a reliable one.