6429.0 - Producer and International Trade Price Indexes: Concepts, Sources and Methods, 2014  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 20/08/2014   
   Page tools: Print Print Page Print all pages in this productPrint All

CHAPTER 6 WEIGHTS AND THEIR SOURCES


INTRODUCTION

6.1 This chapter discusses the role of weights to compile the Australian Producer Price Indexes (PPIs) and International Trade Price Indexes (ITPIs) and describes and evaluates the data sources used to generate these weights.


THE ROLE OF WEIGHTS

6.2 The weights used to compile the PPIs and ITPIs determine the impact a particular price change will have on the overall index. The PPIs and ITPIs are calculated from prices of products that are collected from a non-random sample of providers covering in-scope economic activities and products. The collected prices are first combined to compile indexes for each individual product. For example, several prices for different types of transactions for a detailed type of product may be collected from a range of providers. Price relatives are calculated for each specification. These price relatives are combined to produce the price index for the product. Once this has been done, the product price indexes are combined to produce the class, sub-group, group and subdivision indexes, and then the root level index (See Figure 6.1 below). Each product is given a weight determined by its revenue as a proportion of total revenue (or expenditure, for an input index) for all products in the index during the reference period. To arrive at the aggregate index figure the price relatives of the individual products are multiplied by these weights to derive a weighted average aggregate index. Without explicit weights, relative price changes for all products in the price basket would be given equal weight in calculating the index, which may not be a true reflection of the importance of each priced product in terms of its share of total expenditure (for inputs) or market revenue (for outputs).

6.3 The Australian Bureau of Statistics (ABS) periodically updates the weights in PPIs and ITPIs to reflect changes in market structure. The faster the change in an economy’s market structure, the more frequently the weights in the indexes are updated. Recommended international practise is that PPI weights should be updated at a minimum of once every five years. Details on how new weights are introduced into price indexes are discussed in Chapter 12.

6.4 A diagrammatic overview of the typical structure of a price index is provided in Figure 6.1. At the top is the total value of products represented by the index (corresponding to the value aggregate as discussed in Chapter 15). This is progressively divided into finer product groupings, following the structure of the classification until, at the lowest level, there are samples of prices for individual products. These highly detailed price samples are called elementary aggregates. Indexes are only published down to the regimen level as this is the level at which the structure and weights remain fixed between index reviews. See Chapter 11 for a detailed discussion of index reviews.

FIGURE 6.1 EXAMPLE OF PPI AND ITPI STRUCTURES
Diagram: FIGURE 6.1 EXAMPLE OF PPI AND ITPI STRUCTURES


6.5 The division of products into finer product groupings is intended to reflect increasing levels of substitutability of the products in response to changes in relative prices. For an output index, the index structure reflects substitutability in terms of production, reflecting how producers change their outputs in response to the prices they are receiving in the marketplace. For an input index, the index structure reflects substitutability in terms of consumption, reflecting how producers change their inputs in response to the prices they are paying in the marketplace.


COMBINING PRICE SAMPLES AND WEIGHTING PRACTISES

6.6 The root level of a price index (such as the Manufacturing Division for the Output of the manufacturing industries example above) is compiled by weighting price movements (or price relatives) between the reference and current period by their shares of total value in the reference period. This is simply the alternative way of calculating a Lowe index as described in Chapter 4.

6.7 In practise the PPIs and ITPIs are compiled using value aggregates rather than value shares. The concept of a value aggregate is described in Chapter 15, where the value aggregate for a product in period t is calculated by multiplying the reference period value aggregate P0 q0 by the price relative for period t (pt/P0). This is simply the product of the reference period quantity and the period t price. Summing the value aggregates in period t and dividing by the sum of the value aggregates in the reference period yields the Lowe price index.

6.8 Price indexes measure the change over time in the total price of a fixed basket of products when considered in aggregate. For an input index, the aggregate is of all products purchased while for an output index the aggregate is of all products sold. It is important to note that the use of the term ‘fixed’ relates to the quantities underlying the reference period values (or more formally, the quantities in the reference period value aggregate) - it is, after all, the reference period quantities that are fixed in a Lowe index. Weights are expressed in terms of value shares because quantities are not meaningful or consistent across products. Further, value shares will change over time as the rate of price change varies across products.

6.9 Weights should be updated regularly to ensure the index remains representative of the market structure.

6.10 If held constant on a permanent basis, the weights would become less representative of the relative importance of products produced (or purchased) by producers the further the series moved away from the reference period. There would also be the problem of products that cease to exist and the entry of new products. Furthermore, the finer the level of detail, the less information that exists about the relative importance of products in the basket, which makes it more complicated to calculate weights at lower levels of the index.

6.11 To reduce these problems, weighting practises vary by the level of aggregation. Three distinct practises arise:
  • Weights for the regimen level and above (also known as upper level weights) in which the implicit quantity weights are fixed between index reviews (see Chapter 11)
  • Weights for the index structure between the regimen level and the elementary aggregate level (also known as lower level weights) which are subject to change, dependent on the outcome of a formal review process
  • Weights for the individual specifications within an elementary aggregate (also known as micro-index weights) which are updated as required to ensure the specifications remain representative.


UPPER LEVEL WEIGHTS

6.12 Chapters 4 and 15 discuss the value aggregate from the Australian National Accounts framework. For an input price index, the value aggregate is the value of products used by an industry for the production of their outputs. In a supply-use framework, this is the value of intermediate inputs at purchasers’ prices. For an output price index, the value aggregate is the value of production at basic prices.


GROSS INDUSTRY WEIGHTING FOR THE MANUFACTURING DIVISION

6.13 The output of one activity can often be considered an input to another activity within the same industry, as discussed in Chapter 2. For example, consider the food, beverage and tobacco manufacturing industry. One of the key outputs of this industry is refined sugar. This product is also an input into many other food and beverage products. Two different approaches to weighting can be used to deal with these intra-industry transactions. Compilation of PPIs on a gross industry basis means that the scope of the indexes, and hence the weights of the indexes, cover all transactions occurring within an industry and between that industry and other industries. However, if compiled on a net industry basis, the PPI for a specific industry is restricted to transactions with parties outside that industry, so their weights exclude intra-industry transactions. Australian PPIs are compiled on a gross industry basis as this is the conceptually preferred method for the compilation of the Australian National Accounts.
ROLE OF CLASSIFICATION

6.14 Classifications play a vital role in determining the weights for price indexes. A classification not only helps determine the appropriate scope of the price index (and hence inclusions and exclusions from the value to be covered), but plays a critical role in defining a common language. That is, the classification is the common language that is used to relate the price index structure to its underlying value data.

6.15 Continuing the example of the manufacturing price indexes from above, the Australian and New Zealand Standard Industrial Classification (ANZSIC) [provides the underlying structure of the price index as well as allowing a direct correspondence between the price index, its value aggregate and the source data upon which it is constructed].


INDUSTRY FOCUS

6.16 PPIs are structured and compiled with an industry focus. Input price indexes measure the change in the prices of all products used by an industry for the production of their outputs. Output price indexes refer to the prices of all products produced by an industry. The output of an industry includes not just the products primary to that industry but it also includes secondary production. For example, some businesses classified to the transport industry may also offer travel agency services. For output indexes, weights for secondary production rather than just primary production have been added to the index. The weights of input indexes cover all products used by the industry. Where possible, components which represent secondary production will source the movements of indexes to which those products are primary. If this is not possible, then components representing secondary production will be treated as empty node components (see 6.18).


CHAIN OF REPRESENTATIVENESS

6.17 Price indexes are constructed using a sample of transactions for a range of product types to represent a broad range of economic activity. This “chain of representativeness” is discussed in more detail in Chapter 7. One outcome of using sampling is that the selected products represent not only themselves but also other related products not included in the selected sample.

6.18 Some industries and products will have very small relative importance in terms of their share of total production. It may not be feasible to maintain a sample for these products, however their weight is still included in the overall index structure. This is achieved through the creation of empty node components. When empty node components are included in the index structure, they are weighted according to the value of the output (or input if it is an input index) of the products represented by the component. When the index is compiled, the price movement of an empty node is derived by using the weighted average price movements of the sampled components within the product group to which the empty node belongs. This approach has the advantage of simplifying the inclusion of a component if it becomes feasible to collect a sample for that component.

6.19 An example of the empty node approach can be illustrated by subdivision 11 of Inputs to the manufacturing industries (see Figure 6.2). Not all products under Class 0139 Other Fruit and Tree Nut Growing are included in the sample. Other edible nuts (excluding Peanuts) nec is not sampled. The value of these products was $163.2m (2012-13), or 29.2% of the total value of all Other Fruit and Tree Nut Growing. The value (price movement) of the empty node is derived by using the average price movements of the sampled components within the product group; in this case bananas, orchard fruit and almonds and macadamias.

FIGURE 6.2 THE EMPTY NODE APPROACH
Diagram: Diagram This Flowchart shows the flows from fruit and tree nut growing to it's sub-components.



PERIOD COVERED BY WEIGHTS

6.20 The weighting structure of a price index plays a large part in determining the accuracy and reliability of the index. Key factors in selecting the period used to calculate the weights are:
  • The economic activity over the period should be reasonably normal/stable and representative of likely future activity
  • Close to the link period (the period where the weights are introduced to the index series).

6.21 The weight reference period and the link period used in a price index formula are rarely the same period in practise. For reasons of stability and representativeness, the weight reference period is frequently a year or longer period. New weights are introduced during a specific period, known as the link period. For a Lowe price index, weights are price updated to account for price changes between the weighting period and the link period. For example, the Import Price Index is reweighted each year using the most recent financial year data; however, this re-weighting is undertaken as soon as the data are available, with application in the September quarter at the start of the next financial year. So weights for the Import Price Index from, for example, 2012-13, are price updated to the September quarter 2013. This price updating accounts for changes between the average price over the 2012-13 year and the price observed in the three months ending September 2013. The price updating process is discussed further in Chapter 12.

6.22 In satisfying the stability and representativeness criteria presented above, weights for PPIs and ITPIs are sometimes taken from multiple periods. This practise is followed in those instances where a single year’s data may not be adequate, either because of unusual economic conditions (such as introduction of a new tax system), volatility observed in the marketplace or insufficient sample sizes from survey data. In such cases, an average of several years’ data provides the best weight reference period as it reduces the sampling and seasonal variance of the production or sales data for a given size of the annual sample. For example, whereas the Import Price Index uses weighting data from the most recent financial year, the Export Price Index uses data from the previous two years. Similarly, Input to the House construction industry Price Index uses data over a three-year period.


SOURCES OF DATA FOR UPPER LEVEL WEIGHTS

6.23 Upper level weights are the weights that apply to the components of a price index structure between the root level and the regimen level as illustrated in Figure 6.1. The weights, at the root level including the weight of the regimen level, are fixed in terms of underlying quantities until an index redesign takes place. This process is known as an index review and is detailed in Chapter 11.


Australian National Accounts Input-Output tables

6.24 The key sources of data for upper level weights for the PPIs are the Australian National Accounts Input-Output (I-O) tables. The I-O framework is broken up into two main tables:
  • the Supply table
  • the Use table.

6.25 The supply table shows the output value of product groups by source industry. The product detail tables show a finer disaggregation of supply, showing a comprehensive breakdown of output values by industry at the product level. These data are used as sources for upper level weights for output price indexes. For example, when the supply tables are valued in basic prices the data are used to develop weights for a number of PPIs; namely the Output of the Manufacturing industries Price Index, the price indexes for Services industries, the Output of the Construction industry Price Index, and for the Stage of Production PPIs.

6.26 The use table from the I-O framework shows usage of product groups by industry of consumption. These data are used as sources for upper level weights for input price indexes. For example, when valued at purchasers’ prices these data are used to develop weights for the Input to the manufacturing industries price index. The product detail tables show a finer disaggregation or use, showing a comprehensive breakdown of use values at the product level by industry.

6.27 More information on the I-O framework can be found in:

International merchandise trade data

6.28 The key sources of data for upper level weights for the ITPIs are detailed product level statistics on international merchandise trade. These data are compiled (on a trade basis) from information submitted by exporters and importers or their agents to the Australian Customs Service (ACS).

6.29 The conceptual framework used in compiling Australia's merchandise trade statistics can be found in International Merchandise Trade, Australia: Concepts, Sources and Methods, 2001 (cat. no. 5489.0).

6.30 The majority of the PPIs and ITPIs use the Australian National Accounts or international merchandise trade as sources for upper level weights. For some PPIs, the scope of activity covered by the price index does not align directly with the Australian National Accounts. This is particularly the case for the price index of Input to the house construction industry. The Input to the House construction industry Price Index (as further described in Chapter 13) is an input price index concerned with measuring changes in the prices builders pay for materials used in building houses (as opposed to townhouses, apartments and non-residential buildings). This index is weighted using a bill of quantities approach.
SOURCES OF DATA FOR LOWER LEVEL WEIGHTS

6.31 Lower level weights are the weights that apply to the components of a price index structure below the regimen level down to the elementary aggregate. The weights, including the weight of the elementary aggregate (but not the price sample within the elementary aggregate) may be adjusted to reflect changes in either producer or purchaser behaviour in the market and hence changes in the relative importance of products in the basket. Furthermore, the effects of discontinued and new types of products can also be accommodated. The process of updating lower level weights is known as a sample review and is detailed further in Chapter 11.


Australian National Accounts data

6.32 In addition to broad data from I-O tables, other data sources used to construct the Australian National Accounts aggregates are frequently used in the estimation of lower level weights for PPIs. Whilst these data typically come from other ABS economic surveys (see below), the Australian National Accounts component data are generally more complete in terms of consistent coverage and valuation bases. The data used for lower level weighting are at a more detailed level than published Australian National Accounts data.


ABS economic surveys

6.33 ABS economic surveys are also used in the production of lower level weights for the output PPIs. These data typically provide information on type and characteristics of producers, as well as some detailed information on revenue. In addition, these ABS surveys frequently provide information regarding industry outputs in terms of quantity measures. Deriving lower level weights from these quantity data requires combination with measures of average prices.

6.34 Examples of the use of ABS survey data for lower level weights include:
  • A range of revenue estimates from the program of Annual Integrated Collections are adopted to weight lower level components of the Services industry PPIs
  • A range of quantity, expenditure and revenue estimates from the Building and Construction Statistics program are adopted to weight the lower level components of a number of different PPIs, including the price index for the Output of the general construction industry
  • Revenue estimates from the International Trade in Goods and Services (cat. no. 5368.0) are used in weighting the Services industry PPIs, particularly regarding transport (freight) activities
  • Quantity estimates (kilometre-tonnes) from the Survey of Motor Vehicle Use are used together with measures of average prices in weighting the lower levels of the road freight service industry PPI.


Australian Taxation Office data

6.35 The Business Activity Statement (BAS) is a tax return lodged with the Australian Taxation Office (ATO) in respect of:
  • Goods and Services Tax (GST)
  • Pay as you go (PAYG) withholding and instalments
  • Fringe benefits tax (FBT) instalments.

6.36 The BAS must be lodged by all registered businesses, including government entities for each tax period. Since GST is levied on revenue from sales, the BAS data provide information on output (revenue) by size and type of business. These aggregate data can be incorporated into lower level weights for a range of output PPIs.


The bills of quantities approach

6.37 The Input to the House construction industry Price Index uses bills of quantity to select a basket of products to be priced each period and to derive weights for those products. Bills of quantity are documents produced in the building industry by professionals such as quantity surveyors. These documents break construction projects down into elements or products, and then quantify the inputs required for each product. In this way the types and quantities of materials required can be established. The price of the quantity of each specific material is determined so that the materials can be represented in value terms rather than as quantities. The ABS employs quantity surveyors to undertake this exercise for a selection of typical or representative house designs (for example, brick veneer house 100m2, double brick house 180m2 etc.). Product values for different house designs are then weighted together based on the relative shares of the different house designs in total house construction. Products are then aggregated into broad product types. The fixed basket (and hence the upper level weights) is then determined using the values of the broad product types.

6.38 The bill of quantities approach is refined further in practise by allowing the share of construction of different house types to vary by capital city.

6.39 In addition to the use for regimen products in the price index of Input to House construction industry Price Index, the bill of quantities approach is used as part of the process of determining lower level weights for other construction PPIs. Here the representative designs are not of houses but of other building types. Rather than restrict the bill of quantities to just building materials, the approach adopted in parts of the general construction price index includes work in place, which covers labour, plant and materials, plus margins. Again, this approach requires data from quantity surveyors.


International merchandise trade

6.40 The use of International Merchandise Trade data for upper level weighting is described above. Data provided by the ACS is coded to very detailed levels of the Harmonised System (HS) trade classification. This fine level of detail allows use of these data in constructing lower level weights for the ITPIs.


Other sources of weights

6.41 Administrative data: A wide variety of administrative data on production values are available from public agencies charged with regulating or monitoring certain economic activities. Examples of data obtained from other public agencies include: data for agricultural and mining activities, production and consumption of energy and outputs and consumption of transport services. These types of data are used in developing weights below the regimen product for a range of PPIs.

6.42 Industry associations: Another source for weighting data is industry associations. Many associations conduct surveys of their membership that include detailed information on value of sales by product. Alternatively, where production of a type of product is dominated by one or two large firms, the market shares for these firms can be a source of weighting data. Both types of data are adopted for use in PPIs, particularly for the service industry output PPIs.


SOURCES OF DATA FOR MICRO-INDEX WEIGHTS

6.43 The calculation of the broad price indexes starts with the measurement of the relative price change for an elementary aggregate, which represents the first level at which price observations are combined to calculate an index. At this level, weights are needed to combine individual price observations in order to calculate higher-level indexes. The elementary aggregate index covers all prices collected for one detailed product type. Each elementary aggregate is composed of price observations for products that are similar in terms of material composition, end use, and price behaviour.

6.44 It is important that the weight for each price observation covers the value of all products that the individual transaction represents. That is, most price observations will have a weight that represents other products and transactions in addition to the value of the sampled product alone. Micro-index weights are frequently adjusted to account for the introduction of new product varieties within a product type (such as a unit of sale or new flavour). Similarly, they are adjusted to account for the removal of discontinued individual product lines.

6.45 For example, consider an elementary aggregate for bottled beer for an output price index. The specifications within the elementary aggregate typically incorporate information on brewery, brand, bottle size and units of sale (such as 12 X750 ml bottles or 24 X375 ml bottles etc.). Through the process of sample selection (see Chapter 7), sampled products are selected to represent other products. A particular sampled product may be a best-selling brand of beer in a 24 X375 ml bottle carton; but the weight of the specification would include not only the value of sales of such a product but also the value of sales of other brand varieties sold by the particular brewery. The weight may also include the value of sales from other breweries.

6.46 In calculating such a weight, it is necessary to know several critical pieces of information regarding the values of transactions. Continuing the bottled beer example from above, the following data are required:
  • the value of sales of the selected beer product (brewery, brand, size, units of sale)
  • the value of sales of other brands of the same size from the selected brewery
  • the value of sales of 24 X375 ml cartons of beer from other (non-sampled) breweries.

6.47 In general, value data (either revenue or expenditure) are required for the sampled product and also for any other products that are within scope of the elementary aggregate. Determining such information requires the co-operation of sample providers. The ABS generally collects this information via a personal visit during the activity known as a sample review (see Chapter 11).

6.48 The use of International Merchandise Trade data for both upper and lower level weighting is described above. Data provided by the ACS is coded to very detailed levels of the Harmonised System (HS) trade classification. This fine level of detail allows use of these data in constructing lower level weights for the ITPIs, particularly the Import Price Index.

6.49 The other key source of detailed product information used in micro-index weighting are industry associations, as described above for lower level weights.


Difficulties with specification specific weighting

6.50 Determining value data at the product specification level is a difficult process that sometimes proves to be burdensome for sample providers. In addition, it can become an increasingly complex task to ensure that the micro-index weights are correctly maintained over time. Continuing the bottled beer example from above, the introduction of an immediately popular new brand (such as a boutique beer) requires collection of not only the sales revenue of the new product but also a measure of how the sales values of existing products have changed in response to the new competition. Collecting such information in a timely manner frequently proves difficult.


Equal weighted index formulae

6.51 In the case where specification weighting might be difficult to achieve, an alternative is to adopt a micro-index formula where the sampled specifications have the same weight within an elementary aggregate. This process is currently applied to the accommodation price index.