# An implementation plan to annually re-weight the Australian CPI

The ABS maintains a program of periodic reviews of the Consumer Price Index (CPI) which examines CPI concepts, methodologies and data sources

Release date and time

## Preface

The Australian Bureau of Statistics (ABS) maintains a program of periodic reviews of the Consumer Price Index (CPI) to ensure it continues to meet community needs. The 16th Series CPI Review (ABS 2010) was the last comprehensive review which examined CPI concepts, methodologies and data sources. A number of topics and strategies were considered to maintain the relevance of the CPI in a dynamic environment, particularly given recent developments in methods and data sources. These developments present new opportunities to enhance the CPI.

The ABS has primarily used the Household Expenditure Survey (HES) to derive CPI weights at the published level (expenditure class level and above) and have updated these weights in line with the release of the HES data, currently every six years. The International Labour Organization (ILO) Resolution on CPIs recommends published level CPI weights are updated at least every five years. Conducting a more frequent HES would deliver benefits to the measurement of household inflation and to other users. However, a more frequent HES requires additional funding. The use of alternative data sources relating to household expenditure has been investigated by the ABS to enable the more frequent update of CPI expenditure weights, irrespective of the frequency of the HES.

In July 2016, the ABS released the information paper Increasing the Frequency of CPI Expenditure Class Weight Updates (cat. no. 6401.0.60.002), which investigated the feasibility of using Household Final Consumption Expenditure (HFCE) data from the National Accounts to more frequently update Australia's CPI expenditure class (EC) weights. This paper discusses the consultation undertaken by the ABS as part of this investigation, and outlines an implementation plan to annually re-weight the CPI.

The authors of this paper are staff from the Consumer Price Index Section of the ABS.

## Executive summary

In 2016, the ABS proposed annually re-weighting the Australian Consumer Price Index (CPI) using Household Final Consumption Expenditure (HFCE) data from the Australian National Accounts. This proposal was described in the information paper Increasing the Frequency of CPI Expenditure Class Weight Updates (cat. no. 6401.0.60.002). The research presented in this paper found empirical support for the theory that higher frequency re-weighting of the CPI at the expenditure class (EC) level better captures consumers’ substitution effects. Annually re-weighting the CPI using HFCE data also ensures greater coherence across macroeconomic statistics; and improved alignment with international standards.

The ABS consulted broadly in the development of the proposal and methods outlined in the 2016 paper. Development consultation included an external review by Mr Paul McCarthy, an international Price Statistics and National Accounts expert. Following the release of the 2016 paper, the ABS undertook an extensive stakeholder engagement program, including a call for public submissions. The ABS also conducted numerous bilateral and multilateral consultations with key stakeholders, including: the Reserve Bank of Australia; the Treasury; Department of Social Services; Department of Finance; and State Treasuries.

Stakeholders and public submissions were supportive of the ABS proposal and methodology to annually re-weight the Australian CPI. Noting this support, the ABS will:

• update EC level weights for the CPI and Selected Living Cost Indexes (SLCIs) in the December quarter 2017. The principal data source for updating the weights in the December quarter 2017 will be the 2015-16 Household Expenditure Survey (HES) data; and
• update EC level weights in the CPI annually each December quarter. The principal data source for updating the weights for the inter-HES years will be HFCE data from the National Accounts. HES data will continue to be used in the years that it is conducted.

Consultation with key stakeholders raised additional topics for further ABS research. These include:

• methods and data sources to more frequently re-weight the SLCIs; and
• documenting methods to calculate contributions to percentage change with annually re-weighted indexes.

Methods to calculate contributions to change for annually re-weighted indexes can be found in the Appendix.

Additionally, in the coming months the ABS will:

• publish an information paper in early November 2017 containing the new weights that will be implemented in the CPI and SLCIs in the December quarter 2017. This paper will also estimate the amount of upper level substitution bias present in the Australian CPI for the period 2011 to 2017; and
• investigate methods for more frequently updating the weights of the SLCIs. This work will be published following the December quarter 2017 re-weight of the CPI and SLCIs.

For any questions or feedback relating to this paper, please contact prices.statistics@abs.gov.au. The ABS Privacy Policy outlines how the ABS will handle any personal information that you provide to us.

## Implementation plan

### 1 Introduction

1.1 The Australian Consumer Price Index (CPI) is a robust indicator of household inflation that has served Australia well for many decades. While the CPI is a well-respected ABS output, there are particular aspects of the CPI that can be enhanced, particularly with recent developments in methods and the availability of new data sources. In light of this, the ABS commenced a research program in August 2015 aimed at enhancing the Australian CPI through re-examining current approaches to collecting data and the methods used to compile the CPI.

1.2 The Australian CPI measures the change over time in the prices paid by households for a basket of goods and services. The basket reflects the composition of household consumption preferences; is compiled according to international standards; and is based on robust data collection and compilation methodologies. In compiling aggregate measures of price change, the role of expenditure weights is to reflect the economic importance of each item to the total expenditure of Australian households. Therefore, in practice, National Statistical Offices (NSOs) periodically update expenditure weights to accurately reflect the changing purchasing patterns of households.

1.3 The information paper Enhancing the Australian CPI: A roadmap (cat. no. 6401.0.60.001) outlined a research program to enhance the Australian CPI. The research program examined approaches to data collection and methods used to compile the CPI. The research focused on (i) annually re-weighting the CPI; and (ii) maximising the use of transactions data in the CPI.

1.4 The information paper Increasing the Frequency of CPI Expenditure Class Weight Updates (cat. no. 6401.0.60.002) discussed the feasibility of using Household Final Consumption Expenditure (HFCE) data from the National Accounts to more frequently update Australia’s CPI expenditure class (EC) weights. The research identified and addressed the challenges of using HFCE data and conducted an empirical assessment using experimental series. The paper concluded the empirical results supported the theory that higher frequency re-weighting at the EC level more accurately captures consumer substitution effects.

1.5 The ABS has consulted widely on the proposal to more frequently update the CPI EC weights. Public submissions were also sought with the ABS receiving several submissions.

1.6 This paper presents the results of the consultation process, and provides an implementation plan to annually re-weight the CPI. Also discussed is additional research to be conducted by the ABS on topics raised through consultation with the Australian community and key stakeholders.

### 2 Using HFCE data to re-weight the CPI

2.1 HFCE measures expenditure by resident households on goods and services, whether the expenditure is made within the domestic territory or by Australian residents abroad, and expenditure by Non-Profit Institutions Serving Households (NPISH). Conceptually, the measurement of HFCE closely aligns with the HES. HFCE data captures household expenditure, including the Goods and Services Tax (GST). Both HFCE and HES data cover expenditure by Australian households only and exclude expenditure by non-residents in Australia.

2.2 The HES is used as a major benchmark in the compilation of the HFCE series. Benchmarking is a technique used in the National Accounts, where less frequent (e.g. annual) data sources are used to validate more frequent (e.g. quarterly) estimates of HFCE. The other major benchmark used for HFCE is the Retail and Wholesale Industries (cat. no. 8622.0).

2.3 The use of HFCE data for CPI weights has many potential benefits for both internal and external users of inflation statistics. The primary benefit is that more representative weights enhance the CPI in its principal purpose as a macro-economic indicator of household inflation. There are, however, several challenges with using HFCE data for CPI weighting purposes. These challenges can be grouped together as:

• Overarching challenges - encompassing classification, scope, coverage, and revision challenges; and
• Specific EC challenges - encompassing a case-by-case assessment of using HFCE data for specific CPI ECs.

2.4 Challenges arise because the CPI and HFCE estimates are produced for different purposes and use different data sources. Proposed treatments to each of these challenges were addressed in the chapter Methods and practical considerations of the information paper (cat. no. 6401.0.60.002).

2.5 The ABS has conducted an empirical assessment of the proposed change to annually re-weight the CPI EC level weights. Experimental HFCE price indexes were produced between 2005 and 2015 to compare the annually re-weighted series with the existing CPI series. The results revealed that the experimental HFCE series reported lower average annual household inflation measures relative to the CPI. This supported the theory that higher frequency (annual) re-weighting at the EC level captures a greater amount of consumer substitution when compared to the current CPI (six yearly) re-weighting process.

### 3 Consultation

3.1 The ABS has undertaken broad consultation regarding the proposal and methods to annually re-weight the CPI. The method to use HFCE data for CPI weighting purposes underwent an external review by Mr Paul McCarthy, an international Price Statistics and National Accounts expert. Following the release of the paper in 2016, the ABS sought the views of the Australian community through a call for public submissions, which occurred between July and November 2016. Several submissions were received during the consultation period.

3.2 In addition to the call for public submissions, the ABS conducted numerous bilateral and multilateral consultations with key stakeholders, including: the Reserve Bank of Australia; the Treasury; Department of Social Services; Department of Finance; and State Treasuries.

3.3 Stakeholders and public submissions were supportive of the proposal and methods outlined to annually re-weight the Australian CPI.

3.4 Consultation with key stakeholders raised additional topics of further research for the ABS to consider. These topics include: (i) the possibility of more frequently updating the weights of the Selected Living Cost Indexes (SLCIs)(footnote 1) ; and (ii) the ABS to document methods to calculate contributions to percentage change with annually re-weighted indexes.

3.5 As a result of the feedback received, the ABS has conducted research into methods to calculate contributions to change with annually re-weighted indexes. This information can be found in the Appendix.

3.6 In coming months, the ABS will also conduct further research into methods allowing for more frequent re-weighting of the SLCIs. Both theoretical and empirical research will be undertaken with the results to be published following the December quarter 2017 re-weight of the CPI and SLCIs.

### 4 Implementation

4.1 Updated expenditure class level weights will be implemented into the CPI and SLCIs in the December quarter 2017, due for release on 31 January 2018 and 7 February 2018 respectively. The principal source for the updated weights will be the 2015-16 HES data.

4.2 An information paper will be published in early November 2017 containing the new weights. In addition to this, the ABS will construct a retrospective superlative Fisher-type(footnote 2) index between 2011 and 2017 to estimate the amount of upper level substitution bias in the CPI.

4.3 Following this, the EC level weights in the CPI will be updated annually each December quarter. The primary data source for updating the weights for the inter-HES years will be HFCE data from the National Accounts. The December quarter 2018 will be the first instance where HFCE data is used as the principal data source for the CPI re-weight. The ABS will continue to use HES as the primary data source to re-weight the CPI for the years where it is available.

### Footnotes

1. HFCE data is not available for different household groups. Therefore, the information paper Increasing the Frequency of CPI Expenditure Class Weight Updates (cat. no. 6401.0.60.002) proposed to continue to update the SLCI weights using the six yearly HES data.
2. See detailed information on index theory and aggregation in chapter 4 Price index theory of the CPI Concepts, Sources and Methods (cat. no. 6461.0).

## Appendix - calculating contributions to percentage change

### Show all

#### Introduction

1 As of the December quarter 2017, the Australian Bureau of Statistics (ABS) will move to annual re-weighting of the Consumer Price Index (CPI) expenditure class (EC) weights. The Household Expenditure Survey (HES) will continue to be used to update the weights in the years that it is available. For the inter-HES years, Household Final Consumption Expenditure (HFCE) data from the National Accounts will be used to re-weight the CPI ECs.

2 These short-term CPI series will be chain-linked together to form longer term series. In consultations with users about this change, a number of stakeholders sought advice from the ABS to document methods to calculate contributions to percentage change with annually re-weighted indexes.

3 This appendix explains how the contributions of ECs to quarterly and annual CPI percentage changes are currently derived, then outlines how these contributions can be calculated under an annually re-weighted CPI. An approximation is provided for contributions to the annual inflation rate with an annually re-weighted CPI, which distinguishes between the effect of price changes of the ECs, and changes in the expenditure shares (footnote 1) .

4 The ABS will conduct empirical testing of the method outlined and publish the results in an information paper, which will also introduce the new CPI weights, to be released in November 2017.

#### Current practice

5 The current Australian CPI is a Lowe-type price index. This is an index where at the upper level of aggregation, the price indexes of the ECs are aggregated using fixed quantity, price-updated weights.

6 Points contributions allow users to understand how much an EC contributes to the overall All groups CPI. Points contributions for each EC are published in Tables 7 and 12 of the Time Series Spreadsheets released with the CPI each quarter (cat. no. 6401.0). The weight of EC n to the All groups CPI in quarter t can be estimated using points contributions:

$$w_{n}^{t}=\frac{P C_{n}^{t}}{P C_{A g}^{t}} \times 100$$ (1)

where $$w_{n}^{t}$$ is the weight (%) of EC n to All groups CPI, $$P C_{n}^{t}$$ is the points contribution of EC n and $$P C_{A g}^{t}$$ is the points contribution (or price index) of the All groups CPI, all in quarter t.

7 These weights can be used in the calculation of contributions to the percentage change (footnote 2) . Under a Lowe-type price index, the contribution of individual EC n to the quarterly percentage change in the All groups CPI in quarter t is given by:

$$\operatorname{Cont}_{n q}^{t}=w_{n}^{t-1}\left(\frac{P_{n}^{t}}{P_{n}^{t-1}}-1\right)$$ (2)

where $$P_{n}^{t}$$ represents the price index of EC n in quarter t. Note that $$\frac{P_{n}^{t}}{P_{n}^{t-1}}-1$$is the percentage change of EC n between quarter t-1 and t.

8 The sum of the quarterly contributions of all ECs from (2) is equal to the quarterly percentage change in the All groups CPI.

9 Similarly, the contribution of EC n to the annual percentage change in the All groups CPI is given by the following:

$$\operatorname{Cont}_{n a}^{t}=w_{n}^{t-4}\left(\frac{P_{n}^{t}}{P_{n}^{t-4}}-1\right)$$ (3)

where $$\left(\frac{P_{n}^{t}}{P_{n}^{t-4}}-1\right)$$ is the annual percentage change of EC n.

#### Annual re-weighting and chaining

10 Under annual re-weighting, short-term CPI series (five quarters in length) will be constructed, each with their own set of weights. These series will then be chain-linked together to form a longer continuous time series. For an explanation and examples of chaining, see paragraphs 12.9 to 12.13 of the CPI Concepts, Sources and Methods publication (cat. no. 6461.0).

11 Under annual re-weighting, each quarterly percentage change of the chained CPI utilises a single set of weights. Therefore, calculating contributions to quarterly percentage change of the chained CPI does not pose an issue and can be derived in the same way as is currently done. For example, the contribution of EC n to the All groups CPI quarterly percentage movement in December quarter 2018 would be given by the following:

$$\large\operatorname{cont}_{n q}^{2018 Q_{4}}=w_{n}^{2018 Q 3}\times\left(\frac{p_{n}^{2018 Q 4}}{P_{n}^{2018 Q 3}}-1\right)$$ (4)

where 2018Q3 and 2018Q4 refer to the third and fourth quarters of 2018 respectively (i.e. the September and December quarters). This is the same expression as (2). Again, summing these contributions for all ECs will equal the quarterly percentage change in the All groups CPI in the December quarter 2018.

#### Decomposing annual percentage changes

12 Annual re-weighting does however pose a challenge when calculating contributions to the annual percentage change in the CPI. This is because the contribution will depend not only on changes in prices, but also changes in the weights.

13 As chaining is a multiplicative operation, and due to the arithmetic form of the short-term price index, it is not possible to calculate exact contributions of the ECs to the annual percentage change of the chained CPI. However, an approximation can be derived which distinguishes between the effects of annual percentage price change, and changes in the weights of the ECs. For the contribution of EC n to the annual percentage change between March quarter 2018 and March quarter 2019, the approximation is given by:

$$\large\operatorname{Cont}_{n a}^{2019 \mathrm{Q} 1} \approx w_{n}^{2018 \mathrm{Q} 1}\left(\frac{P_{n}^{2019 \mathrm{Q} 1}}{P_{n}^{2018 \mathrm{Q} 1}}-1\right)+\frac{P_{n}^{2019 \mathrm{Q} 1} / P_{n}^{2018 \mathrm{Q} 3}}{\hat{P}^{2019 \mathrm{Q} 1} / \hat{P}^{2018 \mathrm{Q} 3}}\left(w_{n}^{2018 \mathrm{Q} 3}-\hat{w}_{n}^{2018 \mathrm{Q} 3}\right)$$ (5)

where $$\hat{P}^{2019 \mathrm{Q} 1}$$ is the price index for the All groups CPI in March quarter 2019 using the weights of the previous short-term series and $$\hat{w}_{n}^{2018 \mathrm{Q} 3}$$is the weight of EC n in September quarter 2018 under the previous short-term series.

14 The first term in the expression represents the contribution due to the annual percentage price change, while the second term measures the contribution due to relative quantity (or weight) changes. The latter term is calculated with reference to the September quarter, which is the link quarter that connects the short-term series.

15 The ABS will provide additional information on the derivation of the approximation to interested users upon request. Requests can be sent to prices.statistics@abs.gov.au.

16 Further work will see the ABS conduct empirical testing of the decomposition shown in (5). The results of this will be published in an information paper to be released in November 2017.

#### Conclusion

17 As of the December quarter 2017, the ABS will move to annually updating the CPI EC weights. The primary data source to update the weights for the inter-HES years will be National Accounts HFCE data.

18 Updating the EC weights and chain-linking the short-term (Lowe-type) CPI series in this way does not change the method of calculating EC contributions to quarterly percentage changes. This is because each quarterly percentage change of the chained CPI utilises a single set of weights. The contributions are calculated in the same way as has traditionally been done for the Lowe-type CPI with fixed quantity weights.

19 However, with annually updated weights, the contribution of an EC to the annual percentage change in the All groups CPI has to be approximated. A decomposition has been provided in this appendix which distinguishes between the contribution due to price change, and that due to changes in the weights. The ABS will provide additional information on the derivation of the approximation to interested users upon request.

20 The ABS will conduct empirical testing of the decomposition and publish the results in a November 2017 information paper.

#### Footnotes

1 The ABS acknowledges the input and advice provided by Professor Jan de Haan of Statistics Netherlands and Delft University of Technology in the development of this method.

2 This contribution is different to the contribution to change published in Table 6 of the Time Series Spreadsheets released with the CPI each quarter. The values reported in this table measure an EC's contribution to the change in index points of the All groups CPI, rather than the percentage change.

## Bibliography

### Show all

ABS (Australian Bureau of Statistics) 2010, Outcome of the 16th Series Australian Consumer Price Index Review, cat. no. 6469.0, ABS, Canberra.

ABS 2015, Enhancing the Australian CPI: A roadmap, cat. no. 6401.0.60.001, ABS, Canberra.

ABS 2016, Increasing the Frequency of CPI Expenditure Class Weight Updates, cat. no, 6401.0.60.002, ABS, Canberra

ABS 2017, Consumer Price Index: Concepts, Sources and Methods, 2016, cat. no. 6461.0, ABS, Canberra.

Balk, B.M. 2017. Mixed-Form Indices: A Study of Their Properties. Paper presented at the 15th Meeting of the Ottawa Group. Altville am Rhein, 10-12 May.

International Labour Office 2004, Consumer price index manual : theory and practice, International Labour Office, Geneva.

Ribe, M 1999. Effects of Subcomponents on Chained Price Indices Like the HICPs and the MUICP. Statistics Sweden, Stockholm.

Walschots, J 2016. Contributions to and Impacts on Inflation. Statistics Netherlands, The Hague.

### Previous catalogue number

This release previously used catalogue number 6401.0.60.005.