Routine-biased technical change, structure of employment, and cross-country income differences

We investigate links between routine-biased technical change, the structure of occupational employment, and cross-country income differences. To implement this, we combine several data sources including national labour force surveys and Penn World Tables. We first document that in our novel dataset spanning 92 countries there is a negative relationship between the employment share of routine occupations and GDP per hour worked. We then conduct a development accounting exercise where we differentiate labour inputs by occupation and allow for occupationspecific technologies. We find a systematic relationship between occupation-specific technologies and GDP per hour worked. More developed economies use technologies that are more routine-biased. The productivity of routine labour is about 11 times higher in the top 25 percent than in the bottom 25 percent of countries ranked by GDP per hour worked. International differences in this routine labour technology by themselves account for about 13 percent of the 90-10 ratio of GDP per hour worked, whereas differences in abstract labour technology do not contribute to the observed GDP dispersion. Eliminating all occupations' and capital's technology differences across the world would compress the GDP distribution by 35 to 41 percent.