RDP 2005-05: Underlying Inflation: Concepts, Measurement and Performance 1. Introduction

In principle, underlying inflation, also known as ‘core’ inflation, can be regarded as the medium-term trend in inflation. That is, it is the inflation rate that would be recorded if one were to abstract from (or down-weight) sharp, quickly reversed, movements in prices or one-off shocks that create short-term volatility in measured inflation.^[1] It is particularly important to have an accurate measure of current underlying inflation in the context of an inflation-targeting regime. In addition to providing a less noisy indicator of inflationary pressure than targeted inflation, it is possible that an accurate measure of underlying inflation could enable better forecasts of targeted inflation than would be achieved using the targeted measure alone.

Since 1993, monetary policy in Australia has targeted inflation of between 2 and 3 per cent, on average, over the course of the business cycle. Because of the flexible nature of this target, monetary policy is not required to respond to all movements in the published consumer price index (CPI), and consequently has focused on medium-term inflationary trends. Accordingly, various measures of underlying inflation have been monitored by the Reserve Bank of Australia (RBA) and other economists in Australia in recent years. Indeed, from 1993 to late 1998 the targeted inflation rate was a measure of underlying inflation which excluded a fixed set of items from the CPI.^[2] This series removed the influence of volatile components of the CPI, such as fresh fruit and vegetables, and items whose prices were partly determined outside the market, such as tobacco products. More importantly, the series removed interest rate charges (included in the CPI over that period), thereby precluding a mechanical relationship between changes in monetary policy and targeted inflation. After interest charges were removed from the published CPI in 1998, it became the targeted measure of inflation (Reserve Bank of Australia 1998).

The Australian CPI is a transparent, publicly recognised gauge of price inflation, published quarterly. However, like consumer price indices in other countries, it is subject to volatility, which can make it difficult to distinguish between noise and the inflationary trend. For this reason, since the mid 1990s the RBA has incorporated a selection of underlying inflation measures into its analysis of the economy. These include both simple exclusion-based measures (such as the CPI excluding fruit, vegetables and automotive fuel), which are commonly used in many countries, and statistical measures (such as the weighted median) that exclude various items from the CPI on a time-varying basis. Table 1 lists the statistical measures published by five countries and some relevant research.

Heath et al (2004) calculated a range of statistical underlying inflation measures for Australia and found that after 1993 none of these were better predictors of CPI inflation than a constant forecast of around 0.6 per cent per quarter (2½ per cent annualised). This suggests that underlying inflation measures have become less necessary for forecasting inflationary trends as inflation has become lower and more stable. The aim of this paper is to extend this work by providing a more detailed discussion of underlying inflation concepts and the statistical problems that can hamper accurate measurement, and by subjecting the data to a simple econometric test to determine whether or not CPI inflation adjusts towards its underlying trend.

The rest of the paper proceeds as follows. Section 2 outlines the theoretical underpinnings of several underlying inflation measures, and proposes a test to gauge the extent to which these measures capture the underlying trend in published inflation. In Section 3, the properties of a selection of underlying inflation measures are considered. Seasonal adjustment at the disaggregated component level is shown to eliminate the average bias (with respect to CPI inflation) observed for many of these measures. Calculating underlying inflation measures based on the distribution of four-quarter-ended price changes is shown to achieve a similar result. Evidence is then presented in favour of the proposition that CPI inflation tends to adjust, at least partially, towards statistical measures of underlying inflation. Section 4 concludes.

Footnotes

Although widely used in the literature on underlying inflation, this idea has not been without critics. Keynes (1930, p 78) argued that ‘the hypothetical change in the price level, which would have occurred if there had been no change in relative prices, is no longer relevant if relative prices have in fact changed – for the change in relative prices has in itself affected the price level’ (cited in Diewert 1995). [1]

This series was referred to as ‘Treasury underlying’ inflation. It is no longer published by the Australian Bureau of Statistics. [2]