In the national accounts labour inputs are collected by industry. Homogenising means transforming labour inputs by industry into labour inputs by product. This homogenisation is done using mathematical techniques. The paper compares the results for two wellknown techniques (product technology and industry technology) and discusses the effects of homogenisation on Belgian data for the years 2000 and 2005. Labour inputs are detailed by gender and education level. An additional distinction is made between employees and self-employed. The paper proposes a solution for the negatives problem that arises when applying the product technology model in the case of self-employed workers. It also assesses the plausibility of results by showing the effects of homogenising on wage costs and value added per head as well as on the ranking of industries by education level. The product and the industry technology model yield significantly different results, most particularly for the employment use of wholesale and retail trade. The results of the product technology model are judged to be most plausible.
In the national accounts (NA) a firm is allocated to the industry that corresponds to its main activity. Besides their main activity many firms perform secondary activities like wholesale, software development, R&D, real estate & rental services or restaurant services. These activities have their own industry, but are also performed outside it. Because firms often perform more than one activity, the industries in the NA are not homogeneous.
Homogenising means transforming a variable by industry into one by activity. Each activity (or product) is either a single good or a single service. All the key variables in the NA, including value added, wage costs and labour inputs per industry can be homogenised. This paper focuses on the homogenisation of labour inputs in terms of number of workers. It also reports on the effects of homogenisation on wage costs and value added per head. The provision of a homogenised series for labour inputs is a part of the obligatory programme for the transmission of the NA to Eurostat. The homogenised employment series appears as supplementary information below the Input-output table.
Besides the labour input data by industry, the only data requirement to generate homogenised labour inputs is a Make table. A Make table specifies each industry’s output by product. Starting with this Make table, the homogenisation is done using mathematical techniques. The paper compares the results for two well-known techniques: product technology and industry technology. It discusses the effects of homogenisation on Belgian data for the years 2000 and 2005.
Workers have been subdivided by gender and education level and into self-employed workers and employees. These subdivisions lead to a multiplication of cells to be homogenised. In theory this can worsen the negatives problem that goes with the theoretically superior product technology model. The negatives problem has been reduced by isolating two special groups of workers: self-employed company administrators and temporary workers. Both groups are exclusively used in only one activity. For this reason they are better left out of the homogenisation process.
Both the product and the industry technology model lead to a fairly similar and stable ranking of industries with respect to the use of high skilled labour and value added and output per head. The product technology model tends to increase the differences between activities, while the industry technology model tends to reduce them. Thus activities that employ many (fewer) highly educated workers or more (fewer) female workers, do so even more (less) after being homogenised by the product technology model.
A similar result is obtained for the ratio of value added per worker and that of wage costs per employee. Thus, the activities with the highest value added per worker (e.g. Electrical energy, gas steam and water; Real estate and rental services and Refineries, pharmaceutical & chemicals) have an even higher ratio after value added and workers have been homogenised using the product technology model.
When compared to their non homogenised industry some activities “gain” workers, other “lose” workers. The activities that “gain” most workers are Wholesale trade, Computer & related activities and R&D, Machinery, Electrical & equipment as well as Other community, social and personal services. Activities that “lose” workers are retail trade and public administration. This means that, in 2000 and 2005, the latter industries used workers to perform secondary activities.
When workers are detailed only by gender and education level, applying product technology leads to almost no negatives problem. Still, its results in terms of employment per activity are almost as far from the results for industry technology as from the original employment per industry data. The industry technology model results are judged implausible, mainly because they draw too many workers away from retail trade towards wholesale trade.
If a distinction between employees and the self-employed is introduced, using product technology
leads to a negatives problem in the group of the self-employed, while industry technology
results are implausible. Some of these negatives are caused by the presence of secondary market
activities in non market industries. When performing these market activities, these industries
do not use self-employed workers; however the product technology model does not recognise
To solve this problem, self-employed workers and employees with the same levels of education
and same gender have been treated as perfect substitutes. When replacing negative values for
self-employed workers with appropriate positive ones or zeroes, the results for employees are
obtained as the difference between the homogenised series for all workers and that for selfemployed
workers. This approach yields plausible employment figures and plausible wages
per head for employees.