Job Polarisation: The Shrinking Middle

Why middle-skill routine jobs have declined while both high-skill and low-skill jobs have grown, and what this means for workers and inequality.

For decades, economists expected technology to help mainly skilled workers. But research in the 2000s revealed a more complex pattern: job polarisation.

What happened

In the United States and many European countries from the 1980s onwards:

High-skill, high-pay jobs such as managers, professionals and engineers grew.
Low-skill, low-pay jobs such as food service, cleaning and personal care also grew.
Middle-skill jobs such as clerical work, bookkeeping and routine factory work shrank.

The job market was hollowed out in the middle.

The routine task explanation

Economists David Autor, Frank Levy and Richard Murnane explained this in 2003 by looking at tasks:

Routine tasks follow clear rules, such as filing, calculating or assembling. Computers and machines can do these, so jobs made up of routine tasks declined.
Non-routine cognitive tasks, such as problem solving and creativity, were complemented by computers, boosting high-skill jobs.
Non-routine manual tasks, such as cleaning, caregiving or driving in traffic, were hard to automate but did not need much formal education, so these jobs persisted.

Trade also played a role

Offshoring of manufacturing to lower-cost countries, especially China, also reduced middle-skill factory jobs in rich countries.

Effects

Rising inequality: wages grew at the top, while many displaced middle-skill workers moved into lower-paying jobs.
Regional impacts: towns reliant on routine manufacturing suffered.

Developing economies

Evidence for developing countries, including India, is more mixed. Some studies find polarisation, while others find that routine jobs are still growing as economies industrialise.

AI

Generative AI can perform some non-routine cognitive tasks, such as writing and coding, raising questions about whether high-skill jobs will now face pressure too.

The bank office

In the 1980s, a bank branch employed many clerks to process cheques and records. Computers and ATMs automated much of this routine work. Today, the bank employs fewer clerks, but more IT specialists and relationship managers, while security and cleaning staff remain.

Thinking technology only replaces low-skill jobs

Early automation hit middle-skill routine jobs hardest, not the lowest-skill ones. AI may now affect some high-skill tasks.

Key takeaways

Job polarisation means growth in high- and low-skill jobs and decline in the middle.
Routine tasks are easiest to automate.
Offshoring also reduced middle-skill manufacturing jobs.
AI may extend automation to some non-routine cognitive tasks.