Technological progress and changes in the occupational structure of the Dutch labour market

Research output: Working paper / PreprintWorking paper


The fear that technological progress will render much of human labour obsolete is not new. However, improved computing power and the decreasing cost of it, together with technological advances such as machine learning and robotics have fuelled the fear about massive job losses. On the one hand, technology does substitute for human labour, and especially those tasks that are routine and can be codified in a sequence of logical ‘ifthen-do’ statements are at risk. On the other hand, for workers in abstract task-intensive occupations, technology is more likely to be a complement that will allow workers to be more performant. We asked experts to provide us with their judgement on which detailed tasks within occupations workers are likely to spend more or less time in the next five years, and used that data to depict the automation risk of occupations. We apply the automation risk data on time series data of employment in the Netherlands for the past 25 years. We find that employment growth in the period 1996-2020 was concentrated in occupations with low automation risk. These are occupations with a small share of automatable tasks, but a relatively large share of tasks on which workers are expected to spend more time. Hence, for those occupations, we expect that the potential loss of tasks that can be performed by machines are not compensated by an increasing demand for human labour in non-automatable tasks. But employment also grew substantially in occupations with a moderate automation risk. A potential explanation for this is that the demand for workers in non-automatable tasks has grown stronger than the substitution of workers in automatable tasks. We confirm that technological change has gone hand in hand with a relative decrease in employment shares of middling jobs. However, our automation risk indicator is highest for low-income occupations and decreases almost linearly with income ranks. One potential explanation is that our automation risk indicator only partly captures the routineness of occupations. Moreover, our automation risk indicator is future oriented, rather than past oriented.

Original languageEnglish
Number of pages8
Publication statusPublished - Jun 2022

Publication series

SeriesROA External Reports
NumberPolicy brief no. 6 Horizon 2020-Technequality

Cite this