A2.1. Three methods are considered in this appendix:
- the specific adjustment method;
A2.2. These methods are not necessarily mutually exclusive. Each has advantages and disadvantages which are discussed after a brief description of each method.
- the exclusion method; and
A2.3. The specific adjustment method involves removing the estimated effects of specific types or sources of exogenous price pressure from particular components of the index for the period in which they occur. The aim is to separate transaction prices into market prices (known as net prices) and taxes and subsidies, and also remove the effect of seasonal factors and volatility. These are not simple processes. However, the net price indexes which have been developed in Canada, the United Kingdom, Sweden and several other European countries, have generally been accepted as measures of inflation which are useful for commentary and analysis.
A2.4. Once identified, contributions to price change resulting from seasonality, volatility or changes in policy can be removed while retaining any price changes attributable to actual market conditions.
A2.5. Items which are affected by seasonality, volatility or tax change in one or more periods are not simply excluded for the life of the index, which would result in many valid market-determined price signals being excluded. Rather, the specific effect of the tax change/seasonality/volatility is removed, with the ‘underlying’ price of the item being retained in the price index.
A2.6. Specific adjustment results in a more representative price index. For example, the exclusion method adopted for the Treasury underlying inflation measure means that only about 50% of the CPI basket is included. Many food items, petrol and other important items of consumer expenditure are excluded, resulting in an index with low coverage. Under the specific adjustment method, these items would be included in the index, while non-underlying price movements would be specifically removed.
A2.7. Conceptually this is the preferred method when measuring underlying inflation, because of its sound economic basis. However, due to practical difficulties in specifically adjusting for the effects of taxes, seasonality and volatility, the exclusion method is often adopted as a ‘second best’ approach when developing measures of underlying inflation.
A2.7. Commonly accepted methods such as seasonal adjustment and smoothing can be used where appropriate to remove the effects of seasonality and volatility, while retaining the residual market-determined price movement as a relevant contribution to underlying inflation.
A2.8. It is possible to increase the reliability of the index through the application of modelling systems to make specific adjustments. This can also partially overcome any perceived subjectivity in the adjustments made. Such modelling techniques are used effectively in several of the European net price indexes.
A2.9. A large amount of information is usually required regarding the timing and magnitude of shock effects before specific adjustments can be made accurately. This occurs particularly when shocks impact on only one particular price or group of prices, rather than having a general effect on all prices.
A2.10. Specific adjustment can depend partly on judgement to identify supply shocks, in particular when information on policy changes and other supply shocks is inaccurate, unclear or unavailable. This may lead to the resulting index being questioned as to the adjustments made, although this can be partially overcome with improvements in the quantity, quality and timeliness of the information used to make adjustments.
A2.11. The dividing line between price change resulting from government policy changes as opposed to market signals is sometimes blurred. For example, when a government body raises public transport prices, it may be unclear whether this is a result of a new policy initiative, supply and demand forces, or both (and if both, the proportion relating to each).
A2.12. Subsidies have the opposite effect to indirect taxes and should therefore be treated symmetrically in any underlying measure. However, unless the amount of the subsidy is directly relatable to either the quantity or the price of the product, extreme difficulties may arise in assessing the impact of subsidies for each product in each period. For this reason, the United Kingdom net price index is not adjusted for changes in subsidies. An alternative is to exclude goods and services, such as education and medical care, which may be heavily affected by subsidies.
A2.13. If the objective is to produce an index that is ‘tax free’ (by excluding indirect taxes from both base and current period prices), goods which are subject to significant rates of tax will have a greatly reduced weight using the specific adjustment method. The complexity of the Australian taxation system, with indirect taxes being applied at various points in the production chain, presents significant practical obstacles to the calculation of prices net of tax (i.e. the absolute amount of the final transaction price attributable to tax) due to the need to also estimate changing margins. By comparison, the production of an index at a ‘constant rate of tax’ is more straightforward and involves no adjustment of expenditure weights.
A2.14. This is the method currently used to derive the Treasury underlying rate from the CPI. Exclusion involves the outright removal from the original index of subcomponents which have prices deemed to be unduly influenced by policy measures, seasonality or volatility. The Treasury underlying rate excludes 49% of the CPI basket by weight.
A2.15. Exclusion deals unambiguously with subcomponents whose behaviour is judged to differ frequently and significantly from that implied by the concept of underlying inflation.
A2.16. The exclusion method is highly visible and the results are readily verifiable through the complete pre-specification of its construction. This pre-specification also provides for ease of analysis, replication and forecasting.
A2.17. The major disadvantage of this method is that relevant information which may be provided by the excluded components is lost. As the components are removed entirely from the index, any true market price signals which may be present in their price movement are excluded.
A2.18. In comparison with the specific adjustment method, the resulting underlying measure is likely to be unrepresentative of market-determined price change for the index as a whole because of low coverage of consumer purchases.
A2.19. The decision as to which components to exclude is arbitrary and may be influenced by events in particular periods. Exclusion offers no formula as to what to systematically exclude. For example, the definition of government charges is considerably broader for the Australian underlying inflation measure than for the New Zealand measure.
A2.20. Exclusion is not well suited to the removal of the effects of general price disturbances, such as indirect tax regime changes, which influence goods and services not excluded from the basket.
A2.21. The method is inflexible in coping with any other price disturbances, in particular price volatility, which come from unanticipated sources and affect goods and services which are included in the underlying index.
A2.22. A price observation with an abnormally large residual (deviation from its estimated expectation, or the sample mean) is termed an outlier. Outlier-based methods rely on the lower weighting, or in some cases removal, of these extreme subcomponent price changes. There are several methods by which this can be done, two of which are described briefly below, namely the trimmed mean and the weighted median.
A2.23. The trimmed mean method excludes (i.e. assigns zero weights to) outlying subcomponent price changes. There will only be a difference between the trimmed mean and the full mean if the distribution of subcomponent price movements is skewed. If this is the case, then compared with the trimmed mean, the full mean inflation rate will be pulled in the direction of the skewness.
A2.24. The weighted median inflation rate is the inflation rate for that item or group of items which is at the centre of the total distribution. In effect, the weighted median places less weight on outlier price movements than does the mean rate of inflation, without excluding them altogether.
A2.25. The main advantage of outlier-based methods is that they are less subjective in nature. Fewer decisions as to what to include in, or exclude from, an underlying inflation measure need to be made as fixed statistical methodology is used.
A2.26. Evidence suggests that periods in which significant deviations of the mean from the median inflation rate are observed, often coincide with periods experiencing price or supply shocks. The weighted median, and other outlier methods, may therefore be very effective in removing the volatility effects of such shocks.
A2.27. There is less economic basis behind outlier methods compared with the exclusion or specific adjustment methods. Different components could be excluded from period to period, depending on the size of their price change relative to the entire index. Whereas the concept of underlying inflation suggests that price changes occurring as a result of government policy should be removed, outlier methods may in some cases include these while excluding (or partially excluding) significant market effects.
A2.28. A related problem is that because components are likely to be included in some periods and excluded in others, it presents considerable problems in interpreting the sources of price change over time.
A2.29. In the case of the trimmed mean, there remains a high degree of judgement in its pre-specification. This involves the decision regarding the point at which to trim the tails in the distribution of price movements in the original mean inflation measure. As any price movements outside this threshold are completely excluded, this decision may have considerable impact on the resulting measure of underlying inflation.
A2.30. On the basis of this assessment, it is concluded that the best method for estimating underlying inflation would be the specific adjustment method, with the possible application of exclusion in some appropriate areas, such as highly subsidised services. Conceptually, specific adjustment would provide the best approach to estimation based on both economic and statistical criteria.
A2.31. Specific adjustment has a significant advantage over exclusion since it retains all market-determined price changes while excluding policy-related and other non-underlying price changes. Both methods are dependent to some degree on elements of subjectivity, which is an unavoidable consequence of there being ‘no uniquely correct measure of underlying inflation’ (Reserve Bank of Australia 1995).
A2.32. Seasonal adjustment and smoothing techniques are also considered to be superior methods for removing seasonality and reducing volatility compared with the alternative of complete exclusion of all price movements of a large proportion of goods and services.
A2.33. Outlier-based methods, while having some benefits in reducing the volatility of the inflation index, are open to questioning due to their lack of economic basis and their inconsistency as to what is included in the measure. Indexes based on outlier methods may, however, remain useful as a supplementary measure for policy analysis.
A2.34. While the specific adjustment method is conceptually preferable, there may be some difficulties in its practical implementation. However, the development of net price indexes by several other countries, and their perceived usefulness, indicate that a valuable and effective measure of underlying inflation could be developed using this approach. The development of net price indexes for household inflation is discussed in section 6.
You have reached the end of publication 6421.0