Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Analysing the impact of different types of commodities on freight transport growth using a decomposition approach. Ole Kveiborg, Senior Researcher, Danish Transport Research Institute, Knuth-Winterfeldts Allé, Bygning 116 vest, DK-2800 Kgs. Lyngby, Denmark Abstract A large number of papers have investigated the relation between economic development and growth in freight transport. There is consensus in this literature that the traditional oneto-one linkage between the two is no longer given. This paper analyses the question of coupling/decoupling economic growth and freight traffic focussing on the contribution from specific types of commodities. This approach is used to compare the transport impact induced by individual commodities across factors such as average trip length, average load, the handling or logistical element in the (road) transport of the commodities throughout the supply chain. The results contribute to the understanding of drivers of freight transport, and highlight the linkages between specific economic activities and the resulting demand for freight transport and freight traffic. 1 Introduction The relation between economic development and growth in freight transport is investigated in Tapio (2005), Kveiborg and Fosgerau (2007), NEI et al (1999) and Stead (2001). There is consensus in this literature that the traditional one-to-one linkage between the two is no longer given (e.g. Tapio, 2005). Some find a decoupling and in other cases a negative decoupling (higher growth in traffic compared to economic growth) is found. Some papers investigate the causes of the observed trends in coupling/decoupling using different methods (e.g. McKinnon (2006). A few papers contain a really structural and consistent comparison of the magnitudes of the influence different factors have on the development of freight transport (e.g. Kveiborg and Fosgerau, 2007). The published studies are often very aggregate and focus on e.g. the effect of changes in the value density, the handling factors and the average load and length on vehicles (e.g. Mackinnon, 2006). Other studies have focussed on the transport in specific supply chains from which a link to total impacts are difficult to establish (e.g Browne, 2005). This paper combines the aggregate approach with the approach focussing on specific types of commodities through a comparison of the impact driven by NST/R commodity groups. There are large differences in the demand for transport related to different commodities. Heavy, low value gods like sand, dirt, cement and other similar construction materials have a very large volume, but they are not moved very far, whereas other types of goods, also with relatively low value are moved much further (e.g. clothes). It is important for policy makers to know exactly which type of commodities will produced in the future and to know the transport impact of these changes. They need this information as input to decisions on provision of appropriate infrastructure. Heavy construction materials demand one mixture of infrastructure and high valued goods demand another mixture. Our analysis provides insight into the main driving elements on freight transport within each of 25 types of commodities. The historical changes in transport demand for each of these 25 commodities are analysed using a decomposition method that attributes the change in vehicle kilometres on factors such as the composition of industries producing the good, the value to weight ratio of the good, how it is handled, how it is conveyed, and how far it is being moved. The relative impact of each of these factors is analysed and compared across the different commodities. The methodology and data are outlined in Section 2. Section 3 discusses results and finally, some conclusions and discussion of issues to be dealt with in the continuing work on the analysis is given in Section 4. 2 Methodology Our analysis is carried out by decomposing changes in road freight traffic and transport on contributions from independent factors. Decomposition analysis is most often used when data are limited or when data are difficult to compare across different sources, e.g. the combination of economic data and freight transport data (Lakshmanan and Han, 1997). The Divisia index approach is very suited for decomposing time trends into different factors. A good introduction to the method is Boyd et al (1987). We will return to the specific formulation later. The starting point of the analysis is the model outlined in Figure 1. FIGURE 1 about here The boxes correspond to observables: production first by industry, then by commodity and industry, production in tons, tons lifted by truck size and owner, and kilometres driven. Arrows between boxes correspond to transformation factors: commodity mix by industries, value density by commodity, handling factor by commodity, distribution on vehicle size and owner of the vehicle by commodity, and average load and trip length by commodity, truck size and owner of the truck, which we have calculated. We analyse the change in these factors as the causes for the observable growth in freight traffic and transport. The quantities and factors are described in more detail below. We use the industry production values instead of GDP as our economic measure. The GDP of an industry is the value of the production value minus the value of intermediate inputs other than capital and labour. Thus GDP is a poor measure of the volume of goods transported, since the whole product is transported, not just the part that is added by the industry. To measure the volume of goods traded we therefore use the production value, which includes the value of all inputs. The two economic measures are closely related and follow similar patterns throughout the analysed period as illustrated in Figure 2. GDP growth was 2.3 % and production similarly values grew by 2.3 % p.a. from 1981 to 2004. The only difference in growth rates occurred in 1987 and 1988. This was probably the result of a very restrictive fiscal policy in Denmark with a huge cutback in government spending. The analysis here will thus reveal the same patterns irrespective of which of the two economic activity measures is used. FIGURE 2 about here The data we use in the model and illustrated by the boxes in Figure 1 are: Xijt : The production value in fixed prices in industry i of commodity j in year t. Mjt : The weight of the production of commodity j in year t. Ljt : The tons lifted of commodity j in year t. Tjt : Trips by commodity j in year t.1 Kjt : Vehicle km by commodity j in year t. The data in Xijt give production values in fixed prices for 19 industries and 26 groups of commodities in the Danish economy. Lists of the industries and the commodity groups are provided in the appendix. This information is given for a period covering 1981 to 1992 for commodity details, but until 2004 for the industry disaggregation. All values are total accounts based on national make tables. The make tables classify production according to the producing industry, i, and the commodity, j. Similarly use tables classify inputs according to commodity and the using industry. The combination of make and use tables leads to the input-output tables. The make and use tables comprise 2900 commodities at the fundamental level; in our data set this has been aggregated according to the standard goods classification for transport statistics (NST/R-25), which is used in many European transport statistics. The commodity groups are linked through the make table to 19 aggregate industries, corresponding to those used in the ADAM model of the Danish economy (Dam, 1995), which among other things, forecasts production values by industry. The national accounts data are given in monetary values. We have supplemented with information about the production in industries and of commodities measured in tons. These data have been produced by Statistics Denmark (Pedersen, 1999) by looking at each of the 2900 commodities and using available information about the weight of each of these commodities. Often this information is directly available from the national trade statistics, 1 Used in the calculation of the average trip length and load. and not shown in the figure but sometimes assessments have been made, for example of the weight of 100 square meters of carpet etc. This is very time-consuming work, which is why these data were only constructed for six years (1981, 1983, 1985, 1987, 1990 and 1992). We have constructed data for intermediate years by linear interpolation. The data also contain transport volumes (tons lifted, trips, tonne kilometres and vehicle kilometres including empty running) by the same types of commodities. These data are recorded in an annual survey of a sample of Danish trucks. The recorded transport includes both transport by own-account and hire-and-reward transport, but only domestic trips within Danish borders. This means that trips starting or ending outside Danish borders and cabotage trips by foreign trucks in Denmark are not accounted for in the data. These data are available from 1981 to 2005. The transport statistics have been used for official freight transport statistics in Denmark since 1981. Participation in the survey is mandatory for those sampled by Statistics Denmark. The odometer of all Danish motor vehicles has been registered at regular intervals since 1998. This information is used to scale the survey data has thus improved the already rather high quality of the transport data we have used. We use the data to construct the following variables, which are represented in Figure 1 as the arrows between the boxes: jt jt jt jt K jt : T jt L jt : T jt L jt : the average trip length by commodity type j, the average load per trip by commodity type j, the handling factor by commodity type j, M jt M jt X jt : the weight to value ratio by commodity type j , (the inverse value density) ijt X ijt : the share of production of commodity j in industry i, X it X it j X ijt : production of commodity j by industry i. We can decompose the growth in traffic on these variables using an approach similar to Jorgensen et al (1987). The method is known as the Divisia index decomposition (Liu et al, 1992, and Lakshmanan and Han, 1997). This method gives accurate estimates for infinitesimal changes and is a very good approximation for short-term changes (Lakshmanan and Han, 1997). The following expression for vehicle kilometres (Kt) holds by definition: K jt jt jt jt i ijt X it . jt (1) Denoting the growth rate in a variable as Y ln Y t we can write K t as K ijt K ijt K jt jt jt jt jt i ijt i X it K jt K jt (2) The kilometres Kijt per industry i and commodity type j are found by distributing the kilometres per commodity type by the production of this commodity on industries measured in ton. This means that we can decompose the growth in traffic into the following factors starting from the left: jkt growth in average trip length, jt growth in average load, jt growth in the handling factor, jt weight to value ratio (growth in the inverse value density), i i K ijt K jt K ijt K jt ijt Changes in the production of commodity j in the industries, (X it X t ) growth in the industries’ production value, and finally X t growth in overall production value. To link the commodity specific growth rates to aggregate growth rates we need to make the following adjustment: K t j K jt Kt K jt (3) The model is formulated in continuous time. Our data are discrete, hence we approximate by Yt Yt Yt 1 Yt 1 for most of the series and by Yt 2(Yt Yt1 ) (Yt Yt1 ) when the series change to and from zero; the weights Kijt/Kit are furthermore replaced by two-year averages in the latter type of series. Most of the series include a continuous positive (or negative) trend. Twoyear averages in these cases lead to growth rates below (above) the true averages. When series differ between positive and negative changes the over- and under estimations cancel out. We do not include interaction terms in our decomposition (2). The interaction terms are the joint contribution to growth from the included factors. This can be illustrated by the following example. Assume that the model can be described as x=yz. Then the following decompositions are equivalent: x x1 x0 y1 z1 y 0 z 0 yz 0 y1 z x0 x0 y0 z0 y0 z0 y z yz y0 z0 y0 z0 From this we can se that the final interaction term is not included in the discrete approximation we use. Oosterhaven and Hoen (1998) discuss the differences between different decomposition methods including and excluding the interaction terms. They argue that the interaction terms will only be important for long time intervals between subsequent observations of five years and more. This is supported by Lakshmanan and Han (1997). Since we use a time interval of one year we do not include interaction terms. 3 Results We apply the model for each separate type of commodity to give us an insight in which of the included factors matter for the overall development generally, and the commodity specific changes in particular. In Figure 3 we have shown the average growth in vehicle kilometres in each of the 25 types of commodities over the period 1981 to 2004. Fig. 3 about here The overall growth during this period is 0.8 % per year on average, but the commodity specific average annual growth varies very much more and ranges from -2.83 to 4.15 % This result illustrates the importance of going into more details in order to understand why and how the development in freight traffic is composed. Transport with coal (10) and crude oil (11) and related products are the groups with the largest decrease. A reason for this result is that a larger and larger share of these products are conveyed though pipelines, which are not part of the statistics used. It means that we can only observe the ‘modal shift’ towards pipelines from one side and see that the number of kilometres driven is declining. The commodities with the highest growth are paper waste (21) and furniture etc. (25), but neither of these groups or the groups with the largest decline in vehicle kilometres are among the groups which are conveyed the most. This can also be seen from Figure 3. The last information shown in Figure 3 is the weighed impact on total growth in the vehicle kilometres that each commodity group has. This shows that even though some groups have very high growth rates they do not contribute by much, because the total number of vehicle kilometres with this type of commodity is limited (the values range from -0.04 to 0.20). The detailed commodity results are shown in Figure 4, where it must be remembered that the growth rates are weighed by the relative share of vehicle kilometres that can be attributed to the specific type of commodity (see equation 3). This weighing procedure secures that the small commodity groups with large growth rates only contribute to the aggregate growth with its relative share. This also explains why the pattern in the growth rate for all commodities in total cannot be directly compared to sum of the individual commodity growth rates. Another very important precaution is the inaccuracy in the model. All calculations are averages over a long period, and two-year approximations have been used as well as other approximations. This means that the model in several of the commodity groups cannot replicate the growth in vehicle kilometres exactly. To deal with this we have added an adjustment value for each of the groups. The figure indicates both variables that have a negative impact on the growth in vehicle kilometres (less than 0 in the figure) and variables that lead to higher growth in vehicle kilometres (those above the horizontal 0-axis in the figure). The sum of the positive and negative contributions equals the growth rate in the vehicle kilometres shown in the figure as the dotted line. Figure 4 about here The most important result to be obtained from figure 4 is the huge variety each individual explanatory factor has on the specific commodity. The handling factor plays a very significant role for some commodities changes (5, 7, 16 and 17) and the average trip length seems more important for other groups (7, 8, 16 and 25). The most important factor for all commodities is the general growth in production (including the growth in industry differentiated production). This factor has the same value in each of the commodity groups because production is not split on commodities, and al industry sectors can (in principle) produce all types of goods. The growth in this factor is 2.2 % pa. This factor could have been included, but here it is left out as it would merely lead to a reduction in the industry specific growth rate with the same value. It is also interesting to note the relatively large importance of the commodity mix, which is where production is split on the specific commodities. It is especially in the food commodity group (7) this becomes visible. The food commodity group is further interesting because there are many factors with a large influence on the overall development in traffic. The aggregate growth in vehicle kilometres for this very large group measured in vehicle kilometres (see figure 3) is not very high (0.38 % per year), but truck size, commodity mix, average load, handling factor and average trip length all influence the development in opposite directions. Figure 5 about here To further illustrate these differences Figure 5 show the different in annual growth rates in the average trip length and average load of the different commodity types. We can see very large differences, but we should also notice that the average load and average trip length seem to neutralise each other for most of the commodities (meaning that the growth rates are of equal size, but with opposite sign). We find the same pattern for all the different influential elements included in the model. 4 Discussion In this paper we have applied a decomposition analysis to the growth in freight traffic for different types of commodities. The analysis has been carried out in a joint model with identical elements used as explanatory variables for the growth in the commodity individual growth rates. This procedure has resulted in some difficulties with respect to the possibility to derive the exact size of individual explanatory variables. The reason for using this method nevertheless, is that the changes are comparable across commodity groups and that the contribution to the overall growth pattern is maintained. In the paper we have only focussed on the differences in the explanatory factors and highlighted the focal points. There are naturally many things to be obtained from the analysis that we have not taken into consideration here. One important lesson from the analysis and the results is that there is a huge variety in the causes for changes in freight traffic. Modellers should therefore be careful when trying to generalise too much in models and use the results of too aggregated measures to explain possible outcomes of e.g. specific commodity groups. There are still many unanswered questions in relation to the linkage between economic growth and freight transport. This paper has highlighted a few elements that are important for the understanding of the relationship. However, there is still much work to be done with the analysis described here. One main point to be clarified is that the sum of the commodity specific elements must add up to the aggregate growth rate. This is not yet shown in the figures in this paper. However, the relative size of the individual commodity types and the varying factors will remain more or les unaffected by these changes. The changes in traffic have a close relationship with the environmental effects through energy and emission factors. The traffic is thus an indicator of the energy and emission impacts that have evolved through the analysed period. However, the factors also change over time because the vehicles that are used change to newer vehicles with lower energy consumption and lower emissions (for example the EURO norm requirements for European vehicles). To complete the analysis in this paper for energy and emissions the changes in these factors can be included in a similar way. References Browne, M. (2005). The Validity of Food Miles as an Indicator of Sustainable Development: Final report. Prepared for DEFRA (Department for Food and Rural Affairs), London, July 2005 Jorgenson, D.W., F.M. Gollop and B.M. Fraumeni (1987), Productivity and U.S. Economic Growth, Cambridge, Massachusetts: Harward University Press Kveiborg, O. and M. Fosgerau (2007). Decomposing the decoupling of Danish road freight traffic growth and economic growth. Transport Policy, 14 Liu, X.Q., B.W. Ang and H.L. Ong (1992). The application of the Divisia index to the decomposition of changes in industrial energy consumption. The Energy Journal, 13(4), 161-177 Lakshmanan, T.R. and X. Han (1997). Factors Underlying Transportation CO2 Emissions in the U.S.A.: A Decomposition Analysis. Transportation Research D. 2(1), 1-15 McKinnon, A.C. (2006). Decoupling of Road Freight Transport and Economic Growth Trends in the UK: An Exploratory Analysis. Transport Reviews (Forthcomming) Mckinnon, A.C. and A. Woodburn (1996), Logistical Restructuring and Road Freight Traffic Growth. An Empirical Assessment. Transportation, 23, 141-161 NEI et al. (1999), REDEFINE, Relationship Between Demand for Freight-transport and Industrial Effects. Final Report. Netherlands Economic Institute, The Netherlands Oosterhaven, J. and A.R. Hoen 81998). Preference, Technology, Trade and Real Income in the European Union. – An Intercountry Decomposition Analysis for 1975-1985. the Annals of Regional Science. 32, 505-524 Pedersen, O.G. (1999), Physical Input-Output Tables for Denmark Products and Material 1999 Air Emissions 1990-92. Statistics Denmark. Stead, D. (2001). Transport Intensity in Europe – Indicators and Trends. Transport Policy. 8, 29-46 Tapio, P. (2005). Towards a theory of decoupling: degrees of decoupling in the EU and the case of road traffic in Finland between 1970 and 2001. Transport Policy. 12, 137-151. Appendix List of industries and commodity types Note: Industries 12 through 19 produce non-physical services only Industry Commodity type 1 2 Construction materials Food products 1 2 Live animals Corn 3 Agriculture 3 Potatoes and other vegetables 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Metal industries Chemical products Petroleum refineries Other manufacturing Beverages and tobacco Transportation equipment Energy extraction Other services Electricity, gas and water Construction Wholesale and retail trade Sea transport Other transportation Financial services Housing Government services 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Sugar beets Wood Fur, hides etc. Edible products Feed and fodder Fatty subst. Coal and charcoal Crude oil Gazoline and other mineral oils Iron ore and scrap from iron Ore and scrap of other metals Intermediate products from iron Gravel, dirt, sand and salt and steel Cement, chalk, and other Fertilizers construction materials Tar and asphalt Chemicals Celluloses and paper waste Machinery Metal products Glass, cheramic etc. Furniture, clothes, paper Miscellaneous Empty runnig Illustrations Figure 1. The conceptual model used in the analysis. 180 GDP Production values 160 Tonne km 140 Vehicle km 120 100 80 19 81 19 82 19 83 19 84 19 85 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 60 Figure 2. Growth differences between GDP and production values together with the development in freight transport and traffic in Denmark between 1981 and 2004 Vehicle km per type of commodity (left axis) 4,E+05 5 Vehicle km growth (right axis) 21 4 4,E+05 25 6 13 9 2 3,E+05 3 7 2,E+05 22 17 2 23 20 16 12 Total 1000 km 3 15 1 19 14 24 8 0 18 4 1 2,E+05 % 5 3,E+05 -1 -2 11 1,E+05 10 -3 5,E+04 -4 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 9 8 7 6 5 4 3 2 10 To t 1 -5 al 0,E+00 Commodity type Figure 3. Vehicle kilometres, average annual growth in vehicle kilometres, and.the weighed impact on growth in total vehicle kilometres per commodity. 6 5 4 Total production Production by industry Commodity mix Value density Handling factor Truck size Average load Average length Vehicle km grow th 3 2 1 0 -1 -2 -3 -4 t To al 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Figure 4. Factors contributing to the growth in freight transport per commodity. 21 22 23 24 25 8,00 Average load 10 Average length 6,00 3 2 20 12 4,00 Total 4 2,00 7 5 15 16 9 8 21 23 17 1 25 22 19 11 24 % 13 18 66 14 -2,00 Total 11 4 5 7 12 9 2 14 15 1 -4,00 17 21 19 25 24 23 22 21 20 18 19 18 17 16 15 14 13 13 12 11 9 10 8 7 6 5 4 3 2 1 To tal 0,00 23 25 22 16 10 24 8 3 20 -6,00 Figure 5. The average growth rates in average trip length and average load for the commoditites.