A Task Based Approach to Assessing Skills

Why do some jobs tend to pay much more than others? Sociological approaches to stratification argue that differences in occupational pay reflect workers ability to achieve “closure” in their roles, restricting entry to their positions in order to reduce competition and raise wages. Research in this tradition has examined a variety of political and institutional mechanisms for achieving such closure. I explore how the nature of the skills underlying different kinds of jobs might also facilitate closure, shaping how those jobs are paid.

Much prior work has examined skill demands in terms of the inputs required, such as education and experience. A problem with those measures, though, is that they can equally well be seen as means by which privileged insiders ration opportunities or outsiders signal value. They tell us little about what skills are actually needed to do the job.

I seek to develop a more detailed understanding of how skills generate closure by measuring skills in terms of the underlying tasks that they allow workers to carry out. I use a unique database of online job postings in the US over three months of 2013. The data lists the industry, occupation, skills requirements, education requirements and experience requirements for over 2 million vacancies that were posted online. I use this data to explore how differences in the skills requirements of jobs (defined at the occupation and industry level) affect the average amount that those jobs pay, as recorded in government data.

I using a network analysis to map out the relationships between the different tasks required by jobs. Tasks are connected when they are frequently demanded by the same jobs. Analyzing this network allows me to assess how specialized tasks are, and to what extent they are tightly clustered, indicating that they draw on a coherent, cumulative body of knowledge. A particular advantage of this approach is that it treats every kind of task demand, from “Communication Skills” to “Neurosurgery” as equivalent. No ex ante scheme is used to classify tasks as more or less skilled. Instead, their characteristics are induced purely from how frequently tasks are required alongside other tasks.

Using measures drawn from this network analysis, I show that pay is higher when jobs require workers to carry out a broader variety of skills, and to perform tasks that draw on a coherent, cumulative body of knowledge rather than fragmented, unrelated skills. Indeed, just using these simple characteristics of the skills involved in a job explains 25% of the variation in pay across jobs.

These findings provide a more nuanced explanation of what “highly skilled” means, and why some kinds of skills may be harder to acquire than others. They also shed interesting light on a range of topics of interest to scholars of the labor market, including the challenges created by the decline of manufacturing, the relative wages of male and female dominated jobs, and the different effects of education and experience as means of acquiring human capital.