CONSTRUCTION OF CPIX DATA FOR FORECASTING AND MODELLING IN SOUTH AFRICA*

Abstract

South Africa adopted inflation targeting in 2000, targeting the consumer price index excluding mortgage interest cost (or CPIX), for metropolitan and urban areas. Yet there is no clear technical account of the methodology of construction of the consumer price index (CPI) and CPIX by Statistics South Africa, as published by reputable government statistical agencies in other countries. This paper has two main goals. First, we aim to enhance transparency by explaining the CPI and CPIX methodology (as we understand it), and to encourage publication of an official technical handbook. We also raise various technical issues concerning CPI construction. Second, while the CPIX (metropolitan and urban areas) measure only became relevant to monetary policy setting and wage contracts from 2000, and is published monthly only from 1997, a far longer time series is required for the forecasting and modelling exercises of the South African Reserve Bank (SARB), National Treasury and others. We produce estimates of CPIX (metropolitan areas) back to 1970, on a consistent methodology, using monthly price indices, the appropriate weights, and linking correctly when rebasing. The corresponding measure published by Statistics South Africa goes back only to 1994, and shows some troubling discrepancies with our measure. The SARB uses a quarterly approximation to historical CPIX in its forecast models. Our consistent monthly measure should lead to improved inflation modelling.