It has become a commonplace to talk about the carbon ‘footprints’ of different economic activities. Whenever a new development is proposed nowadays, it is not just its cost that is estimated but also its carbon content. It is widely believed that electric cars result in lower emissions of carbon than cars with internal combustion engines. But is this factually correct? That will depend partly on the source of the electricity in the car batteries: if the electricity that powers the electric car is generated by burning coal, then that car may leave a greater carbon footprint than one with a petrol engine.
It is not possible to find the correct answers to questions like these without the use of input-output analysis, a method of calculation devised at Harvard University in the 1930s by the Russian economist Wassily Leontief. At that time, concerns about global warming did not yet exist. Input-output analysis was intended to be used for research and planning, for example to estimate the labour content of different commodities. The essential feature of the method is that it recognises explicitly the interdependence of different industries in a modern economy. For example, the production of a car requires inputs of steel, rubber and glass. These commodities in their turn require the use of other commodities in their own production. Steel requires iron ore and electricity amongst other things. The production of any single commodity not only requires directly a number of different inputs, but indirectly requires the use of other inputs as well. Input-output analysis provides the technique by which the amounts of such indirect inputs can be calculated. The data on which any such calculation is to be based come from an input-output table, an account showing the flows of commodities between industries in one country in a given year.
The first ever input-output table, for the United States economy in the years 1919 and 1929, was published in 1941. Today government statistical agencies in advanced countries produce such tables annually. Indeed input-output tables are now produced for individual regions within a country, although these estimated regional tables are almost invariably derived by disaggregating national tables. It is quite rare to find an input-output table for a region that has been built from the bottom up, using only original data from that region.
One of those rare cases happened in Scotland; the year was 1973. The primary purpose then was to provide a statistical foundation for such national accounting entities as GDP, as well as estimates of Scotland’s balances of trade with the rest of the UK and the rest of the World. Almost half a century has passed since this pioneering venture; before it is forgotten, it might be useful to put on record how it was done.
The project to construct a set of I/O tables for Scotland for the year 1973 began in January 1975 under the direction of the late Professor Jim McGilvray. Jim had spent the year 1969/70 at the Harvard Economic Research Project (HERP), the de facto world centre for input-output research, as a Ford Foundation post-doctoral scholar. His visit helped to strengthen ties between HERP and the Fraser of Allander Institute at Strathclyde University. As well as the Fraser Institute, two other organisations co-operated in the project, The Scottish Council Research Institute and IBM’s UK Scientific Centre. The project ended with the publication in 1978 of a set of eight tables each representing 78 industries.[i] These were the first survey-based regional input-output tables to be constructed in the UK.
A basic principle agreed at the outset was that the tables would be derived as far as possible from independently acquired data, and would not use any of the mechanical methods that were then in vogue for estimating regional data by disaggregating national tables. The principal source of data was an extensive sample survey of nearly all industries in Scotland. In principle, the basic unit of the survey was the business establishment, although industries like agriculture and fishing were not surveyed in this way. Estimates of the purchases and sales of these two industries were derived from the accounts routinely collated by the Government’s Agriculture and Fisheries Department. All other data were collected by means of a questionnaire, preceded in almost every case by an interview at which the purpose of the survey and the contents of the questionnaire were discussed with the managers of the business concerned. In many cases several visits took place to clarify points of detail.
After punching and validation, the data were passed through a series of processing and adjustment programmes, the results of which were checked manually. The end result of these adjustments was a column vector of commodity purchases for each establishment. After further manual checking, the column of the sample of absorptions were scaled up to obtain an estimated total Absorption column for each sector. By a similar process sales data were scaled up to create a Make matrix. The Absorption and Make matrices formed the basic elements for the construction of the final input-output tables.
Although mechanical procedures such as RAS could have been used to balance the tables, most of the reconciliation was accomplished by a lengthy process of checking estimates from the original data sources, often on a cell-by-cell basis. This manual reconciliation involved a team effort in five iterative stages over three months; it entailed changes in the interindustry flows, final demands, imports and commodity outputs. By the fifth round of balancing, almost all deviations had been reduced to 5% or less. The final stage of balancing was then completed by the RAS method.
In order to construct symmetrical tables, i.e. industry to industry and commodity to commodity, it was necessary to make assumptions about the distribution pattern of secondary products. Sometimes it was evident that in particular cases the application of a general assumption would be inappropriate, so in these cases adjustments were made manually.
A special suite of computer programmes was specifically constructed to process most of the data from the questionnaire to the final tables. The work was done on an IBM 370/145 running under VM and CMS. This system provided interactive facilities using visual display terminals that made it easy to edit in small changes to the data bases.
Results and Uses:
The picture of the Scottish economy in 1973 that emerged from the input-output tables was of an open and highly diversified economy, in which only a modest degree of specialisation was evident in production and in trade with the rest of the United Kingdom. Exports to the rest of the world, however, were quite specialised. From the large number of sectors where the proportion of exports to output was matched by a similar proportion of imports to output, it could be inferred that specialisation in external trade took place more often within commodity groups than between them.
From the tables one could read off estimates of Scottish GDP that could be compared with the Government’s estimates derived from UK national accounts data. Although the two estimates of GDP at factor cost turned out to be quite similar, they concealed much larger differences in some of the principal components of GDP. In particular, the I/O estimates of other value added in manufacturing were significantly higher than the official estimates[ii].
For the first time it was possible to estimate Scotland’s balances of trade with the rest of the UK and with the rest of the World. It appeared that in 1973 Scotland had an overall deficit in its external balance of trade in goods and non-factor services of some £253 million, or 4.5% of its GDP, made up of a deficit of £358 million on trade with the rest of the UK and a surplus of £105 million with the rest of the world.
To communicate this picture of the Scottish economy to a wider audience in schools and offices a multi-coloured wallchart was created. Each cell in the chart contained three pieces of data: the value of the intersectoral flow, the value of the direct input coefficient and the value of the inverse coefficient. A copy of this wallchart has been preserved, and can be seen today in The Fraser of Allander Institute at Strathclyde University.
The 1973 Scottish Tables were subsequently used in a range of applications, from CGE modelling to employment and environmental analyses, undertaken over a number of years at the Fraser of Allander Institute[iii]. Perhaps more importantly, the construction of the 1973 table acted as a catalyst for the Government to take responsibility for producing input-out tables for Scotland. Beginning with 1979, official Scottish tables have been published under the auspices of the Scottish Office/Scottish Government, in close collaboration with the UK’s Central Statistics Office/ National Statistics Office. The regular production of official tables has made possible decades of research on the Scottish economy that would not otherwise have happened.
Today the Scottish Government Input-Output team publishes annually a full set of Supply, Use and Analytical Input-Output Tables, along with Leontief Type 1 and Type 2 Output, Employment, Income and GVA Multipliers and Effects. These are time series consistent tables starting at 1998 to the latest year. 1998-2016 Tables were published in July 2019.
These annual
tables are at the core of several linked statistics and economic modelling
programme that are used today by the Scottish Fiscal Commission for official
forecasts of Scotland’s economy. The Scottish IO Tables are also used
extensively for in-house economic impact assessments and have an environmental
extension to allow estimates of greenhouse gas emissions associated with
Government spending plans.
[i] The Fraser of Allander Institute et.al., Input-Output Tables for Scotland 1973, Edinburgh: Scottish Academic Press 1978.
[ii] F.Harrigan, J. McGilvray and I. McNIcoll, “A Comparison of Regional and National Technical Structures”, Economic Journal vol 90 No 360, Dec 1980 pp 795-810
[iii] For example, D. Henderson, P.G. McGregor and I.H. McNicoll, “Measuring the Effects of Changing Structure on Employment Generation Potential”, International Regional Science Review vol 12 No 1, 1989 pp.57-65