Phylogenetic
Methods

Phylogenetics
Main Page

Input-output
Data

Data Conversion

Algorithms

Software Use

An Alternative
Approach

Bibliography

One part of the STAN project is the OECD input-output database, which is fully available in electronic form (as a large set of electronic worksheets) and partially in paper form. This subproject has, however, been suspended since the startup period. So although input-output tables that include more recent years and that are based on up-to-date classifications (ISIC Rev. 3 and SNA93) are currently being developed, we at the moment have to stick to the originally published input-output tables. These tables cover 10 OECD economies from about 1970 to about 1990. For the purposes of the present paper input-output tables have been selected from the first and the last available year for a few countries (US, UK and Denmark). For each of these countries here are for each year several available tables. The most interesting are the tables that include intermediate flows in current prices from both domestic and foreign suppliers. This kind of table is assumed to allow the best comparison of industrial characteristics across countries, while the tables that only contain domestic flows are assumed to be less comparable due to differences in import intensities across countries. The use of current prices rather than constant prices means that we avoid the related distortions of the data. The capital goods flows are covered by other OECD tables, but the industrial characteristics contained in these tables have not been explored in the present paper.

The presently available OECD data are aggregated according to the International Standard Industrial Classification (ISIC Rev. 2), which otherwise has been abandoned. The industries selected by the STAN project for its input-output tables are:

1:Agriculture
2:Mining
31:Food
32:Textiles
33:Wood
34:Paper
351/2-:IndChem
3522:Drugs
353/4:Petro
355/6:Plastics
36:Nonmetallic
371:IronSteel
372:Nonferrous
381:MetalProd
382-:Machinery
3825:Computers
383-:ElectricApp
3832:Communic
3841:Ships
3842/4/9:Transp
3843:Automobile
3845:Aircraft
385:ProfGoods

This list uses shortened ISIC codes and industry names, but full information is found at the OECD/STAN/IO project page. The list reflects the already mentioned purposes of the project. Thus we see that agriculture, mining, electricity and construction are aggregated to the one-digit level amd several other industries and services are available at the two-digit level. However, important parts of manufacturing are separated out at the tree-digit and even the four-digit level. This highest level of disaggregation is reserved to high-tech industies and otherwise strategically important industries. Thus there are four transport equipment industries. Furthermore, drugs, computers and communication equipment are separated out of the larger industries in which they are contained. These classificatory decisions are, of course, related to the STAN project relation to OECD's DSTI: this directorate has special interest in these disaggregate industries. However, the mix of different levels of aggregation is also relevant for the exploration of the present approach to the reconstruction of industrial trees. For instance, the highly aggregate industries may be supposed to be much more heterogeneous than the disaggregate industries.

The OECD classification is constructed to allow the linking of disparate forms of industrial statistics that are available at different levels of aggregation. This means that a relatively high level of aggregation has been necessary. Thus there are only 35 real industries in each table. This means that for each industry we can only compute 35 input characteristics and 35 output characteristics. This is insufficient to reconstruct a reliable tree that covers all the 35 industries. The exact number of characteristics needed for such an exercise is difficult to determine. But a guess (based on the experience of biological systematists) is that we really need a few hundred independent characteristics. So from the very beginning there is a need for defining subsets of the industries for which we shall try to reconstruct phenograms and/or phylograms. The large group contains the 23 first industries of the classification.

The OECD classification reflects the general approach of industrial systematics of placing industries within a hierarchy. This hierarchy is immediately obvious from the related numbering system. This hierarchy and numbering system is used as a reference point in the discussion of the reconstructed trees. Therefore, we shall name the individual industries both by their number codes and by shortened versions of their names. 382-:Machinery is non-electrical machinery except computers, whole 3825:Computers is office and computing machinery. In all the trees we as far as possible order the industries so that lower-numbered industries are placed above/before higher-numbered industries. This ordering (which is a useful feature of MEGA2) does not change the topography of the tree but it makes a comparison with the OECD hierarchy as easy as possible.

If the OECD hierarchy reflects the similarity patters of the input-output characteristics and if the input-output data are adequate, then the reconstructed tree will more or less look like the tree version of the OECD hierarchy in the figure below (constructed with MacClade). A full similarity cannot be obtained, however. The reason is that industrial classifications are made as multifurcating trees while reconstructed trees are always of the bifurcating type. Thus we have to split up the internal nodes with multiple descendants that are depicted in the figure. But this procedure cannot change the basic structure of the tree. When we observe that a reconstructed tree has a different basic topology, we have found a contradiction that has to be explained.