Is there a Phillips curve relationship present in South Africa and if so, what form does it take? Traditionally the way to estimate the Phillips curve is merely to regress the change in the price level on a measure of the output gap (or the deviation of actual unemployment from the NAIRU). However, Gordon (1990:481-5) has argued that estimating the Phillips curve in this manner biases the estimated results. Instead, Gordon (1997; 1989) puts forward his so-called triangular model that controls for inertia effects, output level effects and rates-of-change (in output) effects. He applies the model to several European countries, the US and Japan and finds meaningful results. The question this paper poses is whether or not the triangular model also applies to South Africa. In estimating the Phillips curve for South Africa the paper also experiments with four versions of the output gap, based on four different methods to estimate long run output, including the standard Hodrick-Prescott (HP) filter and the production function approach. There are several variants of the Phillips curve. The first, as estimated by Phillips (1958) himself, measures the relationship between wage inflation and unemployment. However, other versions consider the relationship between price inflation and unemployment or price inflation and output. This paper focuses on the latter, given the absence of quarterly unemployment data in South Africa, as well as the lack of a reliable and sufficiently long unemployment time series. The paper first presents an overview of literature on the Phillips curve and its estimation for South Africa and other countries. This is followed by the second section that considers the model to be estimated, the data as well as the discussion of the alternative measures of the output gap. The third section presents the estimated results followed by section four that contains the conclusion and a discussion of the policy implications.
by 