Michael Baker - Thesis - Problems in Longterm Forecasting and Planning

About | Contents
Previous: Introduction | This: 1. Freight Transport Statistics | Next: 2. Vehicle Refuelling Infrastructure

1. Freight Transport Statistics

1.1 Introduction

In this chapter I give details of the work which I did on gaining an understanding of freight transport. I go on to outline further work which could be done in this field, and end with the lessons I learnt about data, modelling and forecasting.

My aim in doing this work on freight transport was to try and get sufficient disaggregation of data to be able to get reliable explanations of time trends in freight transport.

In this chapter freight transport is taken as the movement of goods by all modes except the movement of fluids by pipe for which, in general, no other method is used. This excludes gas, water and sewage, but includes petroleum movement by pipeline.

The chapter starts with a brief review of some other attempts at understanding or modelling freight transport in Britain. In it I then go on to look at a simple model, which I found to be inadequate to describe the changes which have occurred over the past 25 years. A more detailed model which should help to explain the changes is described and the extra data required for this model is outlined. Associated with this chapter are three appendices. The first gives details of the sources of data which are available, the second describes the different commodity groupings used in the sources and the third describes a method for determining net physical production based on the use of input-output tables.

1.2 Review of some past work

The principal physical measure of freight transport output is tonne kilometres of freight moved. It is roughly comparable between different modes of transport and most costs vary in proportion to it. Consequently several different methods have been developed to project or forecast the total demand for freight transport in Great Britain in terms of tonne-km by all modes. However, earlier work was of a rather superficial nature due to the lack of any theoretical models to explain or clarify the underlying trends.

In one method used for forecasting freight transport demand, the quantity of freight transport is assumed to be proportional to the total quantity of goods produced, GDP is used as a proxy for the total national output, and so demand as measured in tonne-km has been correlated with GDP (Tulpule 1969, Tanner 1974, Sharp 1973). For example Tanner (1974) found a constant ration, between 4.0 and 4.3 tonne-km per pound of GDP, over several years. However he found that there had recently been a fall from the pre 1970 ratio and for his forecasts he substituted an increase in tonne-km of 2/3 of that of GDP in place of a 1 for 1 correspondence.

Another method used for projecting demand has been to look at the number of tonnes lifted and the average distance over which they have been moved. For example in the first appendix to the 1976 Transport Consultation Document (Department of the Environment 1976) the number of tonnes moved was noted to have been roughly constant for the previous 10 years (1965 to 1974, at about 1.9 thousand million tonnes) and all the increase in tonne-km moved could be attributed to an increase in the average distance over which freight was moved (64 to 77 km). This does not seem to be compatible with a causal relationship with GDP which suggests that an increase in freight transport should be explained by an increasing quantity of goods lifted.

An alternative approach was investigated by Brown and Maultby (1974). They developed a model based upon assigning a transport intensity to each element of an input-output table. The model then requires a projection of the absorption table for the year in question to obtain the quantities of each commodity to be transported in that year. However, as official input-output forecasts are unlikely to be made, further development of this method was postponed. Instead they developed a model which explains the volume of freight transport (in tonne-km) in terms of the movement of activity level indicators for the seven industries which account for 90% of freight transport. One criticism which can be made of both of Brown and Maultby's methods is that they do not take account of possible changes in the average distance over which each commodity is moved.

Another possible method of forecasting the volume of freight transport is the construction of an inter-regional commodity flow model, based upon the concepts of urban traffic models. First the volume of each commodity generated by and attracted to each of many zones would be forecast. Then a forecast of the distribution of each commodity between its origins and destinations would be made. O’Sullivan (1972) used details from the 1962 Survey of Road Goods Transport (Ministry of Transport 1964-66) which included details of the movement of 34 commodities between 107 areas in Britain augmented by similar data from a survey carried out by British Rail for 1964. He found that starting from the known volumes of each commodity originating in and destined to each area a linear programming solution describes the distribution of commodity flows well. He also cited other work which has been done on "establishing estimating equations relating tonnages of commodities generated and attracted to employment in various industries and to residential population" (O’Sullivan 1970, Chisholm 1970).

A possible development of the inter-regional commodity flow model would be to construct a multi-region input-output model. The minimum requirement would probably be of the order of 10 regions and 10 industries/ commodities. However apart from the previously mentioned road freight survey in 1962 and rail survey in 1964 there are not sufficient freight statistics to calibrate such a model. Neither are there details of the regions' economies in sufficient detail to build up the individual regional input-output characteristics. One of the problems which would be encountered when using such a model for forecasting purposes would be forecasting the growth or decline of all sectors in all regions. However the work of Chisholm and O'Sullivan is directed to this end. In the conclusion to their book Freight Flows and Spatial Aspects of the British Economy (Chisholm and O’Sullivan 1973) they say that one of the areas which needs further research is the "stability or otherwise of parameters over time".

Although the available historical freight transport data is not sufficiently detailed to give much spatial disaggregation it can yield time series of such parameters as the average length of move for various commodities. It is towards an aspatial model of freight transport that this study was directed.

1.3 Preliminary analysis

Over the period 1952 to 1975 there was a considerable growth in both the tonnage of freight lifted per year and the total of tonne kilometres of freight moved per year in Great Britain. The quantities lifted and moved are illustrated in Figure 1.1 and Figure 1.2 respectively.

Figure 1.1 Tonnes lifted per year in GB by all modes

Figure 1.2 Tonne-km moved per year in GB by all modes

(The sources of data used for these and all other figures and tables are given in Appendix 1). Although there was considerable growth in the tonne-km of freight moved per year over the whole period it was due to different causes. Over the period 1952 to 1965 the growth was largely due to an increase in the number of tonnes lifted per year. However in the period 1965 to 1975 the growth was mainly due to an increase in the average distance over which freight was moved. The average distance can be found by dividing the tonne-km moved per year by the tonnes lifted per year, for each year. The results of doing this are shown in Figure 1.3.

Figure 1.3 Average distance over which Goods are moved in GB by all modes

To get a further understanding of the number of tonnes lifted per year it is useful to consider the amount of material which is flowing around the country. A simple representation of the system within which the vast majority of freight transport occurs is shown in Figure 1.4.

Figure 1.4 Industrial and Distribution system within which freight transport operates

The system boundary for Imports and Exports are the docks (for convenience the point at which the goods pass through Customs as this is well documented). The system boundary for primary inputs (such as minerals and crops produced in Great Britain, and water and gasses taken from the air which are retained in products) and waste is at the point at which they are removed from or returned to the natural environment. (In the case of primary inputs it is more convenient to take the point at which the production of the respective industry is documented).

In a system in equilibrium, the mass flow over the lines XX and YY would be equal, and we could consider this to be the mass flow through the Industrial and Distribution system. As the mass flow over XX is fairly well documented we can take this as the mass flow through the system. Since the rate of growth in this flow has been of the order of 2.5% per year (for the period 1961-75) the mass flow at XX will be only slightly different from that at YY. Consequently using the flow at XX rather than any other measure of the flow through the system will have a negligible effect on the analysis.

The cumulation of all imports and primary inputs (but excluding retained air and water) over the period 1961 to 1973 is shown in Figure 1.5.

Figure 1.5 Mass flow into the UK industrial and distribution system

It is interesting to note that apart from a switch from coal to oil all other categories have remained essentially static with the exception of UK mined minerals. In this preliminary analysis old scrap was not documented due to lack of readily available data. Also the quantities in Figure 1.5 refer to the UK rather than GB because these figures are more readily available [1] .

The average number of moves for all commodities between entering and leaving the Industrial and Distribution system can be found by dividing the number of tonnes lifted per year by the mass flow through the system. The approximate average number of moves over the period 1961 to 1973 are shown in Figure 1.6.

Figure 1.6 Average number of moves

The reason for this numbers being approximate were that the mass flow used for the UK did not include any packaging and was not complete, whereas the freight lifted per year was for GB and included the weight of packaging.

The relationship between tonnes per year, tonne-km per year and distance moved can be expressed mathematically.

Let

M = sum of all tonnes lifted per year

T = sum of all tonne-km moved per year

Then the average distance moved

D = T/M

This can also be expressed as

Let

m = mass flow through the Industrial and Distribution system (equal to flow across XX or YY in Figure 1.4)

n = average number of moves

Then

n = M/m

1.4 Further analysis

Unfortunately the preceding analysis obscures much of the detail of what was happening. More detail can be obtained by considering the movement of individual commodities. This involves a shift from looking at the inputs to the industrial system to looking at the output of individual industries.

Figure 1.7 and Figure 1.8 show the cumulative number of tonne-km moved per year, and of tonnes lifted per year respectively, of eight commodities for the years 1962, 1967/68 and 1974.

Figure 1.7 Tonne-km moved per year in GB by all modes: by commodity

Figure 1.8 Tonnes lifted per year in GB by all modes: by commodity

As in the previous analysis the average distance over which each commodity was moved was found by dividing the tonne-km moved per year by the tonnes lifted per year. The results of doing this are shown in Figure 1.9.

Figure 1.9 Average distance over which commodities are moved in GB by all modes

To find the average number of moves made by each commodity it is necessary to know the quantity of commodities produced each year. However the total production of a commodity is not the measure required. For example in the chemical industry the total production of all chemicals is larger than that which is used by all users other than the chemical industry itself. This is because many chemicals are used in the production of other chemicals. What is required is the net production of each commodity. That is the quantity which goes to all users of the commodity other than the industry which produces it. I developed a method for determining net production which I explain in Appendix 3. In the appendix I use 1968 as an example and describe how I made an estimate of the net production of the eight commodities. The net production of five of these commodities is very similar to the total production since very little of them are used in their own production.

Consequently it was possible to estimate the net quantities produced in 1962 and 1974. I did this by assuming that the ratios of net to total production were constant in the 3 years. These, along with the tonnes of each commodity lifted per year, and average number of moves, are shown in Table 1.1, Table 1.2 and Table 1.3 for 1962, 1968 and 1974 respectively.

Table 1.1 Average Number of Moves in 1962

Commodity	Supply	Goods lifted	Average number of moves
	million tonnes	million tonnes
Food	90	283	3.1
Minerals	244	279	1.1
Coal	202	322	1.6
Petroleum	64	101	1.6
Iron and Steel	30	78	2.6

Table 1.2 Average Number of Moves in 1968

Commodity	Supply	Goods lifted	Average number of moves
	million tonnes	million tonnes
Food	90.5	314	3.5
Minerals	337.1	465	1.4
Coal	171	285	1.7
Petroleum	97.8	147	1.5
Chemicals	18.7	75	4.0
Building materials	42.4	235	5.5
Iron and Steel	36.9	103	2.8
Other	53.9	389	7.2

Table 1.3 Average Number of Moves in 1974

Commodity	Supply	Goods lifted	Average number of moves
	million tonnes	million tonnes
Food	90	290	3.2
Minerals	371	357	1.0
Coal	111	179	1.6
Petroleum	117	174	1.5
Iron and Steel	37	107	2.9

The average number of moves are also shown in Figure 1.10 Average number of moves [2] .

Figure 1.10 Average number of moves

The total supply of petroleum in the UK, tonnes lifted per year and tonne-km moved per year are shown in the Digest of United Kingdom Energy Statistics (Department of Energy annual) for the years 1965 to 1975. From these it is possible to find the number of moves and average distance moved for petroleum. These along with the supply of petroleum are shown in Figure 1.11.

The resulting tonne-km per year of petroleum moved are shown in Figure 1.12.

Figure 1.11 Supply, average number and length of moves of Petroleum and products in the UK

Figure 1.12 Movement of Petroleum in the UK

It should be noted that this is for the United Kingdom rather than for Great Britain, so Figures 1.7, 1.9 and 1.10, are not compatible with Figures 1.11 and 1.12. The average number of moves in 1.10 and 1.11 are very similar. However it is very unlikely that the differences in average length of haul between Figures 1.9 and 1.11 could be accounted for by Northern Ireland. I could find no explanation for this difference.

As in the preliminary analysis the relationships can be expressed mathematically.

Let

n_i = number of times commodity i is moved

m_i = net production of commodity i per year

d_i = average distance over which commodity i is moved

M_i = tonnes of commodity i lifted per year

T_i = tonne-km of commodity i moved per year

where there are p commodities

Then

Also, using the same symbols as in the preliminary analysis:

and

However, because m is primary inputs and imports, whereas m_i is the output of commodity i, some of which goes into the commodities, note that:

1.5 Further work

The next stage in this work would be the compilation of tonnages lifted per year and tonne-km moved per year of different commodities on each mode of transport for as many years as possible (Approximately 1950-78). With suitable adjustments to the commodity classifications these figures could then be combined to get totals for each commodity by all modes of transport. I have listed the sources of data which I have identified, and the problems with these sources, in Appendix 1.

As well as finding the quantities lifted and moved per year for each commodity the net physical volume of each commodity produced could be found. Further details of how this could be done and some of the likely problems are given in Appendix 3. Having obtained the time series of net supply, average number of moves, and average distance moved for each commodity, it would be possible to look for explanations for the values and time trends found.

Having put this data together, it would also be possible to do a disaggregation by mode. If subscript j represents mode, using the same symbols as before, the total tonne-km moved per year is

where

n_(ij) = number of times commodity i is moved by mode j

d_(ij) = average distance over which commodity i is moved by mode j, and there are q modes.

1.6 Conclusions

From this work I learnt several things:

1. There is insufficient data to even approach an "ideal" model (multi-region input-output). However if such a model could be constructed it would be very difficult to use as a basis for forecasting because of the problems of making large numbers of projections to drive the model (i.e. of final demand for each region in the model).
2. There are problems with data which does not exist. For example published statistics are often incomplete without adequate warning. An example of this is the above mentioned problem of the incomplete list of imports given in the Annual Abstract of Statistics. Also the commodity classifications between the different modes are inconsistent.
3. Greater disaggregation of available data does not lead to any understanding of the causes for the underlying trends. For example a constant number of moves for a commodity does not explain why that ratio has remained constant and so is a poor basis for making a forecast.

[1] Since completing this preliminary analysis I have found that the imports listed in the Annual Abstract of Statistics (Central Statistical Office annual a) are not complete. See Appendix 1 for further details.

[2] The same note of caution should be raised about the incompatibility of the tonnes of supply per year (UK basis) and the tonnes lifted per year (GB basis, with packaging) as in the preliminary analysis.

About | Contents
Previous: Introduction | This: 1. Freight Transport Statistics | Next: 2. Vehicle Refuelling Infrastructure
Copyright © Michael Baker 1981,2005. All Rights Reserved.