5.1 There are a number of strategies available to keep the CPI contemporary in a dynamic economy. As part of the review the ABS has considered the frequency of weight updates, quality adjustments in the CPI, the use of scanner data, re-referencing the CPI and related series, and the classification used to compile and publish the CPI.
Frequency of weight updates
5.2 The Australian CPI measures the change over time in the price of a fixed basket of goods and services acquired by household consumers. In compiling the CPI, price movements for the different items are combined using weights which represent the relative importance of each of the items to the total expenditure of the CPI population group. The quantities underlying the base period expenditures remain fixed over the life of the basket, with the expenditures being updated by the observed price changes in each period.
5.3 Over time, the expenditure patterns of consumers change in response to factors such as changes in relative prices, changes in tastes, changes in disposable income and the introduction of new products. In order to maintain the relevance of the CPI it is important to update the item weights to reflect current expenditure patterns.
5.4 The ABS currently updates the fixed items of the CPI basket every six years. The periodicity of the Household Expenditure Survey (HES) and the updating of the CPI weights in Australia do not meet the minimum standards in the ILO Resolution on CPIs, which recommends weights be updated at least once every five years.
5.5 During the consultation process of the review there was also considerable demand from users for the ABS to increase the frequency at which the CPI weights are updated to ensure the weights respond to changes in household spending patterns.
5.6 As consumer expenditure patterns change over time, a fixed set of weights used in the CPI runs the risk of becoming unrepresentative, and can lead to potential item (or upper level) substitution bias. Item substitution occurs when households react to changes in relative prices by choosing to reduce purchases of goods and services showing higher relative price change, and instead buy more of those showing lower relative price change.
5.7 Like most CPIs, the Australian CPI is calculated using a base-weighted Laspeyres-type index formula (a modified Laspeyres index) which keeps quantities fixed between major revisions but allows prices to vary. The restriction imposed on the CPI by keeping the quantities fixed and not allowing substitution among goods in response to relative price change results in item substitution bias. This is essentially a divergence between the measured CPI and an 'ideal' index.
5.8 Quantitative estimates of item substitution bias in CPIs are therefore typically undertaken by comparing an index using the fixed basket approach to a superlative index, such as the Fisher index. Superlative indexes make use of both beginning-of-period and end-of-period information on both prices and quantities (expenditures), thereby accounting for substitution across items (the requirement for end-of-period information in real time is the reason this type of index is an impractical option for statistical offices for the compilation of the CPI).
5.9 The ABS constructed a retrospective superlative index to provide an estimation of item substitution bias in the fixed weight Australian CPI for the five year period between June 2000 and June 2005.
5.10 Whilst the economic literature identifies five main sources of bias in CPIs, this analysis focuses on one type only - item substitution bias - in the context of the frequency of weight updates in a fixed-base index. The results in the analysis in this paper should not be taken to equate to the total bias in the CPI, which will be the net sum of all sources of bias. For a brief discussion of the other sources of bias and steps that the ABS takes to minimise such biases in the Australian CPI, see Appendix 2 of the ABS Information Paper: Issues to be considered: 13th series Consumer Price Index Review 1997 (cat. no. 6451.0).
5.11 The analysis found that the All groups CPI was upwardly biased (as measured by the difference between the Laspeyres index and the Fisher index) by 1.2 percentage points at the end of the five year period - 0.2 of a percentage point per year on average - due to the inability of the fixed-base index to take account of the item substitution effect. For more details see Appendix 7.
5.12 These results should be considered with caution, as the period chosen (2000 to 2005) may not be typical. For example the introduction of the GST in 2000 may have altered consumer behaviour, however it is also worth noting that similar analysis undertaken by Statistics New Zealand (SNZ) to help inform the discussion on the frequency of CPI weight updates supports the ABS findings.
5.13 SNZ re-weights its CPI every three years, and demonstrated how the index might have tracked over a six year period in which weights were fixed. The results showed that the simulation in holding weights constant for a further three years (i.e. mirroring the ABS practice of six-yearly weight updates) resulted in an additional (average) annual 0.2 percentage point growth in the index.
5.14 The findings of both sets of analysis show that the longer the period between re-weights, the larger the impact of item substitution bias on the CPI. Table 1 and Graph 1 in Appendix 7 illustrate the cumulative nature of this bias.
5.15 To ensure changes in household spending patterns are captured and represented in a fixed weight index, and to minimise the impact of the substitution effect, it is important that weights and baskets are updated regularly. Re-weighting the Australian CPI more frequently would improve the relevance and accuracy of the index, and ensure better alignment/adherence to international standards. While the results suggest a low annual level of item substitution bias, in line with international estimates, the bias is cumulative. More frequent updating of the fixed-level weights would result in the CPI being more responsive to shifting spending patterns of households, and more representative of current household expenditures.
5.16 As the measured impact of item substitution bias is upward, its effect on the CPI is to overstate the rate of inflation. Currently some $172 billion in annual Australian Government expenses alone are tied to the CPI (DoFD, 2010).
5.17 Assuming a measured 0.2 of a percentage point annual upward bias, the potential monetary impact on annual Commonwealth government expenses would be several hundreds of millions of dollars (based on indexation to a CPI calculated under a four-yearly re-weight as compared to the current six-yearly re-weight schedule). This compares with the additional annual average cost of around $3.5 million to implement a four yearly re-weighting of the CPI.
5.18 HES data is the most reliable source for deriving CPI weights. Alternative data sources were examined but they do not provide the socio-demographic information or the financial and employment characteristics about households which are essential for re-weighting the population subgroup indexes.
5.19 A four-yearly HES is preferred for several reasons. The upward bias appears more marked after four years; the HES is a complex survey and has a three and a half year development and processing cycle; and a four-yearly HES enables integration with the Survey of Income and Housing (SIH), which is conducted every two years. If a four-yearly HES were to occur, the next survey could be run in 2013-14 (the next HES is currently scheduled to be run in 2015-16), provided funding was available from 2011-12.
5.20 A more frequent HES would also deliver considerable benefits to a range of users including National Accounts, researchers and policy makers.
5.21 The AG supported more frequent weight updates, and all members supported the use of HES data for this purpose.
5.22 The ABS is also exploring other strategies to further minimise the item substitution effect on the CPI. One of these strategies involves the use of scanner data to improve the quality of lower level CPI weights unavailable from the HES.
5.23 Subject to availability of the additional funding required, the frequency of CPI weight updates will be increased from six-yearly to four-yearly, via a more frequent Household Expenditure Survey (HES).
5.24 To ensure a robust measure of the CPI, strategies to minimise possible sources of bias in the CPI will be continued (e.g. the use of scanner data will be explored).
5.25 The concept of quality adjustment as used in the Australian CPI is based on the notion of consumer utility. The main purpose of applying quality adjustments is to preserve the concept of pricing equivalent items for a fixed basket of goods and services over time such that the price index compares 'like with like'. As products change, their components or ingredients may change resulting in an improvement or degradation in quality. It is important to note that the quality adjustment process may result either in increases or decreases in prices as a result of its application.
5.26 During the consultation phase, user comments were invited on the quality adjustment methodology used by the ABS. Some users suggested that the quality adjustment process used in the CPI is responsible for underestimating price increases (and thus underestimating the cost of maintaining a constant standard of living).
5.27 Most of the criticism directed at pricing to 'constant quality' concerns the impact on income utility and the erosion of purchasing power over time. For instance, some users claim that the CPI relates to a theoretical cost of a good (or service) in such cases irrespective of whether that item can still be obtained. They argue that in many cases no real utility is gained from the apparent change in quality and that the ABS should not be adjusting the prices. These users argue that they cannot buy the same basket of goods today as they could in the past, and that the cost of living is rising quicker than the CPI is reporting.
5.28 AG members supported the concept of quality adjustment in the CPI, however recommended an increased level of transparency for the quality adjustment process used by the ABS. It was suggested that the ABS should be more transparent in explaining quality adjustments on higher expenditure items such as motor vehicles, audio visual equipment and other large durable items.
5.29 Quality adjustment is a conceptual requirement of a CPI. The international CPI Manual (ILO, 2004) advises that failure to pay proper attention to quality changes can introduce serious biases into the CPI. Quality adjustment bias arises from the statistician’s inability to perfectly account for changes in the quality of items over time.
5.30 While the quality adjustment issue raises important conceptual and practical challenges, the important role assigned by macroeconomic policy makers to price statistics underlines the priority that should be attached to ensuring that price statistics are not distorted by inappropriate quality adjustment procedures. The ABS has a number of strategies in place to minimise quality adjustment bias.
5.31 The ABS publication Consumer Price Index: Concepts, Sources and Methods, 2009 (cat. no. 6461.0) provides information on the quality adjustment methods used by the ABS and the circumstances in which they are generally applied. The ABS will improve the transparency of the documentation in future releases of that publication. The ABS will not, however, produce a CPI index excluding quality adjustment, as this would be a misleading indicator of price change.
5.32 The ABS will provide more information on the quality adjustment processes used in the CPI to better inform the public on circumstances where quality adjustments take place.
5.33 The ABS uses the expression administrative by-product data to describe information which is not obtained from conventional statistical surveys. The ABS uses such administrative data in the compilation of various CPI components such as automotive fuel, property rates, and hospital and medical services. The ABS is investigating the potential uses of scanner data (the electronic capture of product information at the point of sale) to assist in the compilation of the CPI and ALCIs.
5.34 The use of scanner data in the CPI should be seen as experimental. Several national statistical agencies have had success using scanner data in their CPIs, as outlined in Ivancic (2010). Statistics Norway has been using scanner data since August 2005 to compute its index for food and non-alcoholic beverages. Statistics Netherlands introduced supermarket scanner data into its CPI in June 2002. In both Norway and The Netherlands both prices and expenditure weights for a large sample of grocery items are derived from scanner data. The Swiss Federal Statistical Office also uses scanner data in its CPI.
5.35 Scanner data is collected and aggregated by retail chains and may be used to estimate unit values, changes in unit values, and household expenditure on the items offered for sale by the stores.
5.36 During the consultation phase of the review, most users supported efforts by the ABS to improve the accuracy of prices used for the CPI. The AG also supported the ABS in its endeavour to explore the uses of scanner data.
5.37 The ABS sees scanner data as a rich source of information which has potential benefits across many areas of the prices program. Access to electronic data files will help reduce provider burden and the associated costs of physically obtaining prices from many stores. The use of scanner data will also enable the ABS to price a wider spread of products and from more geographical locations. Furthermore, as both unit values and quantities are available for all transactions, it should be possible to use aggregated sales data to update lower level item expenditure weights between HES cycles.
5.38 Scanner data may also enable the ABS to compare price movement differences between rural and urban locations which may provide important information on whether the CPI needs to cover a wider geographic coverage than capital cities.
5.39 Utilising scanner data for the CPI will not replace field collection operations entirely. The ABS estimates that there are potential scanner data sources for use in components that contribute around 50% of the weight of the CPI basket (as at June 2010).
5.40 It is clear that undertaking this work will require significant effort. Statistical organisations that have commenced using scanner data for their CPIs have advised that any potential savings from reducing the current price collection operations are at least partly offset by the added cost of managing and quality assuring the large volumes of scanner data.
5.41 The ABS sees potential with the use of scanner data in the CPI, both to improve reliability and reduce data collection costs, and will continue to explore the use of scanner data for CPI and related purposes, leveraging international experience with the use of such data.
5.42 The ABS changes the index reference period (a process known as re-referencing) of the CPI from time to time, but not frequently. Re-referencing of indexes is undertaken to keep the level of the index manageable and usable to clients.
5.43 The index reference period is the period in which all index numbers in the CPI have a value of 100.0. The current index reference period for most of the components in the Australian CPI is 1989-90. The index reference period for the ALCIs is currently June quarter 1998 = 100.0, and for the PBLCI is June quarter 2007 = 100.0.
5.44 Re-referencing should not be confused with re-weighting. Re-referencing does not change the relative movements between periods - it merely resets the index reference period of the CPI basket to a period in which all index numbers in the CPI have a value of 100.0. Re-weighting involves introducing new weights and recalculating the aggregate index for each period which will affect the relative movements between periods.
5.45 Although the issue of re-referencing was not raised by the ABS during the public consultation phase of the review, a small number of users have requested it be considered in the context of the review.
5.46 The ABS does not regularly undertake re-referencing as frequently changing the reference base is inconvenient for users, particularly those who use the CPI for contract escalation. Re-referencing may result in a small loss of detail in historic data, especially for long series. Relative movements of any series over time are not generally affected by a reference base change, except for minor differences, due to rounding, between the percentage changes published on the old base and those on the new base.
5.47 Although re-referencing incurs some cost to the ABS, the ABS is planning to re-reference the CPI and related series because some expenditure class index numbers have become either very small or very large and difficult to manage over time. For example, the September 2010 index number for the Audio, visual and computing equipment expenditure class was 14.5, whereas for the Tobacco expenditure class was 594.9.
5.48 The ABS sees the opportunity to re-reference the CPI to 2009-10 (the 16th series CPI weight reference period) as having many advantages, including enabling all CPI and related series (including the living cost indexes) to be presented on the same reference base.
5.49 To minimise the impact on, and inconvenience to users, re-referencing the CPI and related series is expected to be undertaken one year after the implementation of the 16th series weighting pattern.
5.50 The CPI and all related series will be presented on a reference base of 2009-10 = 100.0, commencing with the September quarter 2012 CPI.
5.51 A CPI classification is a framework for the systematic categorisation of all goods and services acquired by the consumer household sector. The CPI classification was raised during the review to determine whether the ABS-designed CPI Commodity Classification (CPICC) required any updates.
5.52 There was limited interest during the public consultation phase of the review and from the AG on the issue of commodity classification and item coverage. There were a number of requests for a correspondence between the CPICC and the Input-Output Product Classification (IOPC) to be published. No users were opposed to the use of the CPICC as the main classification framework for the CPI and there was no interest in changing to the Classification of Individual Consumption according to Purpose (COICOP).
5.53 Although few submissions commented on this issue, the ABS conducted an investigation to determine if there were any realisable gains from adopting the international standard classification (COICOP) instead of the CPICC.
5.54 In assessing the most appropriate classification for use in the CPI the ABS considered numerous criteria. The classification should: align with Australian CPI concepts; represent the economic reality faced by Australian households; and facilitate both international comparisons of CPIs and internal coherence with other ABS statistics.
5.55 The CPICC is a purpose designed classification for use in an acquisitions CPI. The customised classification allows items to be grouped that are substitutable and that generally display similar price behaviour in the Australian market.
5.56 COICOP is the international standard classification. It is based on an economic cost-of-use approach which is more closely aligned to a CPI that measures household standards of welfare.
5.57 The different concepts underpinning the Australian CPI and COICOP lead to significant divergences, especially in the treatment of OOH. These conceptual differences between the Australian CPI and COICOP would make it necessary to alter the classification to accommodate the different scopes. A straight adoption of COICOP is not advisable for the ABS while the ABS adopts an acquisitions approach.
5.58 Despite these conceptual differences the CPICC is comparable to the COICOP at the top level.
5.59 Although the CPICC aligns best with the underlying concepts of the Australian CPI there are various places within ABS macroeconomic statistics (such as the Australian System of National Accounts (ASNA) Household Consumption Final Expenditure (HFCE)) which may benefit from closer classification alignment between COICOP and CPICC. There is also the need for a correspondence between the CPICC and the IOPC which is used to structure the Input-Output (I-O) tables in the ASNA.
5.60 Greater coherence between ABS statistics is the main internal driver for a classification change in the ABS. Appendix 8 contains information and further examples of where greater alignment between classifications (or creating and improving correspondences) may improve statistical coherence.
5.61 Adopting COICOP in the CPI is not expected to produce improvements due to the modifications required to make the structure suitable for the Australian CPI. However, potential gains have been identified in the development of a joint classification for use in the ASNA HFCE and CPI.
5.62 The Australian CPI will continue to use the ABS Consumer Price Index Commodity Classification (CPICC), however the ABS will take action internationally to influence a revision of the Classification of Individual Consumption According to Purpose (COICOP) to address shortcomings which prevent its use as the primary product classification in the CPI.
5.63 A correspondence between CPICC and the Input-Output Product Classification (IOPC) will be compiled and published for the 16th series CPI.