What the French Evidence Actually Says

The French evidence gives occupational polarization a more precise meaning than a chart of “good” and “bad” jobs can provide. This checkpoint connects the stock pattern, the flow accounting, and the entry-side result into one judgment: after 1994, the routine middle shrank mainly because its pipeline from non-employment stopped replenishing it, not because a new wave of routine-to-abstract upgrading or a dominant routine-to-manual downgrade suddenly appeared. That is a strong description of the measured French mechanism. It is not yet a universal theory of polarization or a settled account of its cause.

The narrow answer to the title

The title's claim is best understood as a statement about how an occupational distribution can change. Employment polarization can begin at the entry margin when people arriving or returning to employment are sorted away from routine work, even while workers already in routine occupations continue to move upward at roughly the previous rate. The middle then contracts through a changing pipeline rather than through mass downward movement of its incumbents.

Several facts make that interpretation more than a verbal possibility. The French routine share is broadly stable before 1994 and declines much faster afterward. The routine-to-abstract flow is large and nearly unchanged across the break. The manual-to-routine buffer weakens, but it does not explain the whole change. Most importantly, the routine excess flow with non-employment moves from a positive replenishing channel to a negative one, and the entry deviation changes while the exit deviation remains essentially stable.

This chain matters because it preserves the difference between a transition and a stock. Routine workers can keep using an upward ladder while fewer people enter routine behind them. Abstract employment can grow partly because the ladder remains open, while routine employment shrinks because the entry point narrows. Manual employment can grow because new or returning workers are absorbed there, without that growth being evidence that routine incumbents were systematically pushed downward.

flowchart LR A["Routine share falls after 1994"] --> B["Flow decomposition"] B --> C["Routine-to-abstract upgrading stays broadly stable"] B --> D["Manual-to-routine buffer weakens"] B --> E["Non-employment-to-routine entry collapses"] C --> F["Entry-side selection is the primary measured mechanism"] D --> F E --> F G["ICT-price timing and occupational specificity"] -.-> H["Candidate explanation, not causal proof"]

The structure of the evidence is as important as the headline. The flow decomposition establishes what changed in the observed accounts. The routine-cognitive concentration makes a technology-related explanation plausible. The timing of ICT prices supports that explanation, but the paper also identifies a concurrent low-wage contribution-relief policy and does not claim to disentangle the two. A measured mechanism and a candidate cause should not be put in the same evidentiary box.

What is observed, inferred, and still open

The most secure conclusion is an accounting conclusion. Given the observed occupational stocks, direct job-to-job movements, exits, and the construction of entries, the post-1994 acceleration of routine decline is primarily associated with a collapse in the non-employment-to-routine entry channel.