The Reserve Bank's forecasting performance

For most of the period since 1994, the target range for annual Consumers Price Index (CPI) inflation, established by the Policy Targets Agreements (PTA), was 0 to 3 per cent. Over this period, actual CPI inflation has averaged 2 per cent. As one might expect, analysis shows that our medium-term CPI inflation forecasts since 1994 have been biased towards under-prediction, which is the subject of this article. In any particular period, inflation is unlikely to be exactly as forecast, given that the economy is affected by unforeseeable events and inflation is far from perfectly controllable. However, it is important to have a good understanding of why inflation has evolved as it has, and not as predicted. We need to know whether particular events in the period under consideration have dominated inflation outcomes, or whether there is a fundamental problem with the policy process - such as a fundamental misunderstanding of the workings of the economy - that would systematically affect future monetary policy outcomes unless corrected. In this article we focus particularly on our CPI inflation forecasting performance, but also examine our forecasts of other key macroeconomic variables, given their relevance for explaining our CPI forecasts. We conclude that, in the mid-1990s, underestimation of growth, and overestimation of the economy's capacity to grow without generating inflation pressures, were the source of most of our under-prediction of medium-term CPI inflation. From 1998 until recently, the major factor explaining the under-prediction of inflation appears to have been sizeable and persistent differences between the assumptions we used for the path of the exchange rate and its actual evolution. We also conclude that contributions to forecast inaccuracies have at times been made by our understanding of the noninflationary output growth rate, the equilibrium exchange rate and exchange rate pass-through into CPI inflation. However, these factors do not appear to be systematic sources of inflation forecast bias.