EU-harmonised and comparative measurement of occupations and skills

This report is a deliverable of Work Package 21 ‘Innovative tools and protocols for working condi­tions & vulnerability research.’ Section 2 provides a review of the measurement of occupations in surveys in Europe. It specifies how occupations are measured in web surveys, and outlines the meth­odology currently used to test the comparability of the job content and skill requirements in occupa­tional titles. The results of the validation efforts are presented, including the design of a project to measure occupations on a global scale. Occupation is a key variable in socio-economic research, used in a wide variety of studies. Where such studies use quantitative approaches, they usually rely on survey data. In this paper an inventory of 33 surveys is analysed with respect to the phrasing of the question. The vast majority uses an open text format for the occupation question, but the phrasing of the question is different across almost all surveys. In an additional question, half of the surveys ask for a job description, and again the phrasing varies largely across the surveys. Coding of the open format question is usually (semi-) automatic, survey agencies applying dictionary approaches for automatic occupational coding. In web surveys closed survey questions can be asked using text string matching and search trees for navigating. Recently, machine learning algorithms appear to be a prom­ising development, requiring a substantial amount of manually coded occupations to be used as train­ing data for the automatic classification. A huge training set is required for an auto-coder to apply machine learning algorithms. This Section details a design to develop such a training set in a multi-country approach. Occupation is a key variable in socio-economic research, used in a wide variety of studies, but the measurement of occupations is a major challenge. In case of open-ended survey questions, web sur­veys pose extra challenges, as respondents are more likely to key in odd text compared to other survey modes, particularly those with an interviewer. Web surveys however offer new opportunities for closed survey questions with self-identification of occupations, particularly when using semantic matching and a database with large numbers of coded occupational titles. Section 3 details the design requirements for such a database to facilitate surveys in multiple countries. First, this Section describes how the WISCO Database of Occupations has developed since its first use in 2000, when the database was applied in the WageIndicator web survey on work and wages. By the end of 2015 the web survey and the database had expanded to 85 countries. In these fifteen years, the number of occupations in the database had increased from 55 to 1,896 and the languages covered from 1 to 43. The section then presents an overview of the ISCO occupational classification with four hierarchical levels and their coding. It also reviews how occupations can be measured in web surveys in open-ended or closed survey questions. For the closed question three approaches are detailed, notably scrolling, search trees and semantic matching. Given that any national labour market easily cover 10,000 or more occupational titles, the semantic matching is considered the best approach provided that the look-up tables include 5,000 or more occupational titles for each country. Finally, the Section details the design principles in the database used for semantic matching. It discusses the size of the source list in relation to the ISCO coding. It details the issue of male and female titles and the use of start or end years for occupations It specifies how the database copes with highly aggregated occu­pational titles such as clerk or manager, with abbreviations and organisation specific job titles, with synonyms, with skill levels within and across occupations, with occupations in the corporate hierar­chy, with composite jobs, with handicraft workers, and with subsistence farmers and hunters. Next, it details how to cope with respondents indicating that their job title is not in the database. The Section ends with a future outlook for a program to test the validity and reliability of occupational measurement. Section 4 addresses the question whether occupations refer to the same work activities, as assumed in occupational classifications such as ISCO-08? Up to now, no large-scale empirical testing of this assumption has been conducted, whereas occupations are a core variable in socio-economic research. Using the task descriptions provided for all ISCO-08 4-digit occupations, the frequency of task implementation was tested using respondents in the multi-country, multilingual WageIndicator web survey on work and wages in 13 countries. The web survey targets individuals in the labour force. Depending on their self-selected occupation, the relevant task list was shown and respondents were asked to tick on a 5-point scale how often they performed each task. For 427 occupations (ISCO-08 4-digits) in total 3,237 occupation-specific tasks were available. Between November 2013 and August 2015 33,678 respondents had completed the tasks questions for their respective occupations. The results show that task measurement is feasible because it can generate sufficient observations to allow for analysis for a range of detailed, 4-digit occupations. Moreover, given that the WageIndicator web survey also holds data on wages, the median and average hourly wages (in euro) could be computed for each task separately, showing that the average wages of tasks performed on a daily or weekly basis ranged between 5 and 34 euro. The data collection challenges future empirical testing of hypo- theses concerning the variation in task frequencies and their related wage premiums within and across coun­tries, across occupations’ skill levels, across firm sizes, across regions and alike