Can information on the distribution of ZAR returns be used to improve SARB’s ZAR forecasts
Matthew Greenwood-Nimmo, Daan Steenkamp, Rossouw van Jaarsveld
Last Modified Date:
2022-10-12, 09:52 AM
Publications > Working Papers | What's New
We characterise the distribution of the rand/US dollar exchange rate returns in two ways: (i) using realised moments estimated from intra-day data; and (ii) using implied market expecta-tions of the rand exchange rate extracted from options prices. We begin by showing that these distributions can be used to contextualise rand movements and summarise the extent of rand uncertainty and tail risk. We then conduct a simple exercise to evaluate whether information on the variance of the rand returns distributions can be used to improve the South African Reserve Bank’s rand forecasts. We find that, while these rand forecasts show little bias, the magnitude of the forecast errors can be reduced and the directional accuracy improved sub-stantially by incorporating historical and implied estimates of rand variance in the forecasting model. This effect is particularly pronounced at longer forecast horizons.