Investigating the non-linear Wagner’s hypothesis in South Africa

Purpose - The commonly adopted view of the relationship between government spending and economic growth follows the Keynesian approach, in which government spending is considered to determine economic growth. However, there is another theory, which suggests that economic growth in fact determines government spending. This is Wagner’s hypothesis. The purpose of this paper is to investigate which of the two approaches applies to South Africa, and further observes the level of non-linearity between the two variables. Design/methodology/approach - This study was carried out using quarterly time series data from 1980Q1 to 2015Q1. Granger causality technique was used to observe the direction of causality between the two variables, while regression error specification test (RESET) was employed to determine whether the variables exhibit linear or non-linear behaviour. This was followed by observing the threshold band, using two techniques, namely, sample splitting threshold regression and quadratic generalised method of moments. Findings - The causality result shows that South Africa follows Wagner’s law, whereby government spending is determined by economic growth, supporting Odhiambo (2015). The RESET result shows that the variables depict a non-linear relationship, thus the government spending economic growth model is non-linear. It was found that if positive economic development is to be achieved, economic growth should preferably be kept within the −1.69 and 3.0 per cent band, and specifically above 1 per cent band. Originality/value - The unique contribution of this study is that no previous study has attempted the non-linear government spending-economic growth nexus whether within the Keynesian or Wagner law for South Africa.