The recent finding of job polarization −the decline in the share of employment in middle occupations relative to high- and low-skilled occupations− has raised concerns about its causes, being automation one of the main candidates. Surprisingly, existing studies have been mostly descriptive. The lack of a formal index has prevented comparisons of the degree of job polarization across countries and time. To overcome these problems, we propose an index of job polarization which, among other things, allows us to distinguish between different labor processes: job upgrading, job downgrading, job polarization and job concentration. In addition, the use of our index of job polarization permits testing the Routine Biased Technological Change hypothesis which claims that middling jobs in the center of the distribution tend to consist mostly of routine and repetitive tasks that can be easily computerized and thus eliminated.