South Africa’s Short and Long Term Interest Rates: A Threshold Cointegration Analysis

Abstract

This paper presents a two-regime vector error-correction model (VECM) with a single cointegrating vector and a threshold effect in the error-correction term. We use a Hansen-Seo (2002) algorithm to extract maximum likelihood estimates in eight threshold cointegration model s that relate short-term to long-term interest rates in South Africa for the period 1990M1-2010M7. We employ a SupLM test to test for the presence of threshold. The Hansen-Seo algorithm yields both linear and non-linear estimates plus critical values used to test threshold effects. The method is applied by relating the South Africa Reserve Bank policy rate, the repo (short-term) to intermediate (TB rate, money market rate) and long-term rates (the 10-year government bond, the loan and deposit rates). In all cases, linear cointegration is rejected in favor of a threshold effect.

On the term structure of South African interest rates: cointegration and threshold adjustment

“…There is, however, no momentum equilibrium adjustment asymmetry, indicating that the nonlinear error correction model may not fit the term structure of interest rates in South Africa well. This finding is in stark contrast to the finding presented in Dube and Zhou (2013). On this account, we fit a linear error correction model and find short-and long-run causal flow from the yield on long-term government bonds to the SARB policy rate.…”

“…The key studies on this issue are Arize et al (2002) and Dube and Zhou (2013). Arize et al (2002) employ linear cointegration techniques to explore the long-run movement of short-and long-term interest rates in 19 countries including South Africa, whereas Dube and Zhou (2013) utilize a two-regime threshold cointegration technique and the Hansen-Seo algorithm to examine whether short-and long-term interest rates co-move in South Africa.…”

“…Dube and Zhou (2013) resolve this problem by fitting a two-regime vector error correction model for the term structure of interest rates in South Africa using the Hansen-Seo algorithm. Our paper adds to the literature by using nonlinear models which are different from the ones utilized in Dube and Zhou (2013) to examine the dynamic correlations between short-term and long-term interest rates in South Africa. The importance of dynamic interest rate correlations is summed up by how such correlations act as leading indicators for business cycle predictions.…”

“…They find the yield curve to be a better predictor than the other indexes at longer horizons. In a recent paper, Dube and Zhou (2013) find evidence of threshold cointegration between short-and long-term interest rates in South Africa.…”

“…The evidence shows that Arize et al (2002), and Dube and Zhou (2013) are the only studies which examine the co-movements of interest rate series for South Africa, to the best of our knowledge. The first authors assume that the cointegrating adjustment mechanism is symmetric.…”