RDP 2014-07: International Trade Costs, Global Supply Chains and Value-added Trade in Australia 4. The Structure of Australia's Domestic Supply Chain

4.1 Measuring the Domestic Supply Chain

The fragmentation of production across firms and industries within the domestic economy has also been an important feature of structural change in Australia. To the best of our knowledge, there has been no research into the structure and evolution of the domestic supply chain in Australia.

Ideally, to measure changes in the structure of the Australian supply chain we would have information on transactions at the plant-level between buyers and suppliers. Unfortunately, these data are not available for Australia. Nonetheless, we can extract useful information on the length of supply chains and an industry's position along a supply chain from industry-level input-output tables.

The most conventional measure of inter-industry linkages and supply chains is the ratio of intermediate consumption to gross output. As mentioned earlier, gross output is the total market value of goods and services produced in an economy, which can be divided into value-added and the cost of intermediate inputs (or intermediate consumption). Value-added reflects the returns to labour and capital used by the industry. The higher the share of output that is accounted for by intermediate consumption, the more of the industry's value is added outside of the industry. A high share of intermediate consumption indicates that production in the industry is ‘vertically fragmented’.

While the ratio of intermediate consumption to gross output is easily estimated, it does not account for the full complexity of inter-industry linkages involved in production, nor the length of the supply chain between a good's production and its consumption. More sophisticated measures have been developed to describe the relative position of an industry in the value-added chain – ‘fragmentation’ and ‘upstreamness’ (Fally 2012).

The ‘fragmentation’ statistic measures the number of stages involved in the production of a good or service and how the overall value-added of the product is distributed along these stages. Fragmentation is calculated using a good or service's inputs. Fragmentation is defined as one plus a weighted sum of the number of stages involved in the production of good i's intermediate inputs, where the weight corresponds to the value added by each input. The index takes the value of one if there is a single production stage in the final industry and increases with the length of the production chain.

For example, if half of the value of industry A's gross output is accounted for by intermediate inputs from industry B, and the inputs from industry B do not require any inputs themselves, then the ‘fragmentation’ measure of industry A will be 1 + 0.5 = 1.5. If, however, half of the value of industry B's output is spent on intermediate inputs from industry C (which themselves do not require any intermediate inputs) then the fragmentation measure of industry A will be 1 + 0.5 × (1 + 0.5) = 1.75.

An industry's supply chain will, therefore, be more (or less) fragmented depending on the extent to which the production of its final output depends on intermediate goods which are themselves more (or less) fragmented. Low fragmentation does not necessarily mean a ‘short’ supply chain, but could indicate that the bulk of value-added is concentrated at only one or two stages of a long supply-chain, rather than being dispersed across the length of the chain.

The fragmentation measure is mathematically comparable to the measure of total backward linkages of a sector in traditional input-output theory, first proposed in the late 1950s and equivalent to measures of sector-to-economy ‘output multipliers’ (Miller and Blair 2009). The interpretation of fragmentation as a sector's weighted average number of production stages clarifies the definition of backward linkages, while avoiding the shortcomings of the ‘multiplier’ interpretation.^[13]

The ‘upstreamness’ statistic measures the average number of stages occurring between production and final demand of a good or service. Upstreamness is calculated using the good or service's outputs. It is defined as one plus a weighted sum of the number of stages between production of the goods that take output from industry i as an input and these goods' own final demand, where the weight corresponds to the fraction of industry i's total production going to each use.

For example, if half of the gross output of industry A is used for final consumption and half is used as intermediate inputs by industry B, which produces entirely for final consumption, then the measured upstreamness of industry A will be 1 + 0.5 = 1.5. If, however, only half of the value of industry B's output is used for final consumption, with the other half used as intermediate inputs by industry C (which produces entirely for final consumption), then the upstreamness measure of industry A will be 1 + 0.5 × (1 + 0.5) = 1.75.

Industries with low measured upstreamness produce largely for final consumption. An industry that mainly produces for intermediate use will be more upstream, particularly if it produces for other industries that are also upstream.^[14] The measurement of fragmentation and upstreamness requires detailed input-output data, giving the relative values of the intermediate inputs that each industry requires for production. We use a combination of ABS input-output tables and the WIOD to estimate the supply chain statistics. The calculations are explained in Appendix B.

4.2 The Domestic Supply Chain in Australia

The two supply chain measures indicate that the Australian domestic supply chain involves about two stages of production, on average, and most of this production occurs two stages away from final demand. However, there is significant variation in the degree of vertical fragmentation and upstreamness across different sectors of the economy (Table 3). The manufacturing, construction and utilities sectors tend to be the most fragmented, while the resource sector is the most upstream. In contrast, the services sector tends to be the least fragmented and most downstream.^[15]

This can be seen even more clearly if we decompose these sectoral estimates and examine the variation in fragmentation and upstreamness across individual industries (Table 4). The most fragmented industries tend to have long supply chains along which little value-added occurs at each stage. The most fragmented industries are typically in the manufacturing sector, including meat, dairy and basic metals manufacturing. In contrast, the least fragmented industries are typically services industries, such as education, finance and insurance, and health and community services.

The most upstream industries are typically in the resource sector, although the property and business services industry is quite upstream too. Manufacturing industries that produce mainly primary commodities, such as basic metals, also occupy very upstream positions. In contrast, the most downstream industries are generally in the service sector, such as education, health and community services, and retail trade. Some manufacturing industries, such as motor vehicles and clothing, are also downstream.

We can construct time-series indicators of the domestic supply chain for the aggregate economy using historical input-output tables at roughly three-year intervals back to the mid 1970s. These supply chain indicators suggest that the Australian economy has become more fragmented and more upstream since the 1970s (Figure 8), while there was a notable increase in both measures over the 1990s.^[16] Methodological changes to the input-output tables by the ABS appear to explain at least some of this ‘jump’.^[17] Given this, for much of the subsequent analysis, we focus on the WIOD data and the period since the mid 1990s. This is also the period for which we have comparable international data.

The WIOD data also indicate that, for the Australian economy in aggregate, the degree of fragmentation has been relatively unchanged over the past decade (Figure 9) while the extent of upstreamness has increased (Figure 10). The increase in aggregate upstreamness reflects two factors: i) an increase in the value of aggregate output accounted for by the resource sector (which is the most upstream sector) and ii) an increase in the level of upstreamness within the resource sector.^[18] The first factor reflects the economy's response to the global commodities boom. The second factor is more subtle and appears to reflect a change in the composition of Australia's supply chain. While, on average, about 85 per cent of Australian production occurs domestically, the upstreamness indicator suggests that the foreign share of the production line has been gradually increasing over the past decade, particularly in the resource sector. This, in turn, reflects two overseas developments highlighted earlier: i) Australian resource production is becoming further removed from the source of final demand (e.g. Europe and North America) as its exported content is increasingly re-directed via intermediate suppliers, such as China; and ii) the Chinese supply chain itself has become more upstream.

International comparisons based on the WIOD data allow us to put the Australian estimates into context. Australia's degree of vertical fragmentation and upstreamness is higher than in the rest of the world, on average (Figure 11). This is particularly true in terms of upstreamness due to Australia's relatively large resource sector. In contrast, the supply chains of the G7 countries tend to be relatively downstream and less fragmented by international standards. Clearly, China stands out in international comparisons as a country that has a particularly long and upstream aggregate supply chain.

In fact, Australia has bucked a global upward trend of rising fragmentation since the early 2000s (left-hand panel, Figure 11). This increase in fragmentation around the world has been particularly pronounced in the emerging economies and especially China. On the other hand, the WIOD estimates imply that the degree of upstreamness has risen in Australia over the past decade at about the same rate as the world average, and much less quickly than in China.

Footnotes

For details on these inherent problems in deriving sector-to-economy output multipliers from input-output tables see Gretton (2013). [13]

The upstreamness measure is mathematically comparable to the measure of total forward linkages of a sector in traditional input-output theory (see Jones (1976)), which have also been interpreted as ‘supply-driven’ or ‘cost-push’ multipliers, as opposed to the ‘demand-driven’ multipliers which are mathematically related to the fragmentation measure (see Miller and Blair (2009)). Antràs et al (2012) show how an equivalent measure of distance to final demand can be reached using an alternative derivation. [14]

The estimated level of fragmentation and upstreamness across sectors is somewhat sensitive to the level of sectoral aggregation used in the calculations. However, the estimated trends for fragmentation and upstreamness are little affected when we calculate each statistic based on different degrees of sectoral aggregation. See Appendix B for more details on the issue of aggregation. [15]

These longer-run trends for the Australian economy are in stark contrast to those of the United States; the US economy has become progressively less fragmented and less upstream over the same period (Antràs et al 2012; Fally 2012). Fally attributes this to a shift towards a more service-oriented, and hence downstream, economy. [16]

The ABS input-output tables are constructed from supply and use (S-U) tables, which detail industries' production and uses of goods and services and are compiled as part of the Australian System of National Accounts (SNA). While past S-U tables are revised for all periods when historical revisions are made (like other national accounts measures such as GDP), previously published input-output tables are not revised, and therefore do not form a consistent time series. Significant methodological changes were undertaken in the 1990s, including those associated with the implementation of SNA93. For details, see Gretton (2005) and Australian Bureau of Statistics (2013). [17]

In Appendix C, we conduct a more detailed ‘shift-share analysis’ to examine the factors driving the aggregate changes in fragmentation and upstreamness. [18]