For best experience please turn on javascript and use a modern browser!

Measuring the Routine and Non-Routine Task Content of 427 Four-Digit ISCO-08 Occupations

AIAS Working Paper Series - Emil Mihaylov & Kea Tijdens

E. Mihaylov & K. Tijdens (2019) Measuring the Routine and Non-Routine Task Content of 427 Four-Digit ISCO-08 Occupations, AIAS Working Paper 8

This paper develops new measures of the task content of occupations that are based on the International Standard Classification of Occupations 2008 (ISCO-08). Using a detailed set of 3,264 occupation-specific tasks, we construct five measures of non-routine analytic, non-routine interactive, routine cognitive, routine manual and non-routine manual tasks for 427 four-digit occupations. To generate these measures, first we assign each of the 3,264 tasks to one or more of the five task categories. The decision to classify tasks as routine or non-routine, and as cognitive or manual, depends on whether the tasks can be replaced by computer-controlled technology and whether the performance of the tasks requires cognitive or manual skills. We judge the automation potential of tasks on a case-by-case basis and classify tasks to one or more of the five task categories. Because the classification of 3,264 tasks can be prone to errors, we devote substantial attention to the possibility of misclassifying tasks. We discuss three particular types of task misclassifications and provide examples of tasks that could be potentially misclassified.   
In line with the previous literature, we find that non-routine analytic and interactive tasks are most prevalent in the work of Managers and Professionals, routine cognitive tasks are mainly concentrated in the work of Clerical Support Workers, and routine and non-routine manual tasks are most common in the work of Plant and Machine Operators and Assemblers and Elementary Occupations, respectively. We compare the newly developed task measures with three previous studies (Acemoglu and Autor, 2011; Dengler, Matthes and Paulus, 2014; Frey and Osborne, 2017) and demonstrate that our measures are moderately to strongly positively correlated with the previous papers’ indexes. Based on our task content measures, we provide a back of the envelop estimation of the number of occupations that might be at risk of automation. We find that approximately 16 percent of the 427 ISCO-08 occupations fall into the so-called high risk of automation category – they contain 70 percent or more routine tasks. The 16 percent of automatable occupations correspond roughly to 11 percent of total employment in the Netherlands.