A Pipeline-and-Place Diagnosis for Structural Change
Occupational polarization can start at the entry margin, but the meaning of that claim is precise. It does not say that incumbent workers are never displaced, that technology has one universal effect, or that every person excluded from a shrinking occupation can be moved into a green one. It says that an occupational stock can change because the flow of people entering and re-entering it changes, even while established workers continue along existing ladders. This final synthesis combines that French finding with the OECD's skills-and-place evidence to answer a practical question: how can a region tell whether a transition is a problem of entry, skills, vacancies, job design, or geography?
The answer is not a single index. It is a sequence of distinctions. First identify the stock change and the flow that is moving. Then identify which workers face the altered access. Map skill-related destinations, check whether vacancies exist locally, and choose a response that addresses the actual bottleneck. The result is a diagnosis that can guide employers, workers, public employment services, educators, and local authorities without pretending that a correlation is a causal estimate or that a vacancy is a guaranteed job.
The title's precise answer
The French evidence gives the strongest support for the entry-margin thesis in a specific setting. After 1994, France's routine employment share declines more quickly, but the routine-to-abstract upgrading flow remains large and broadly stable. The manual-to-routine buffer weakens. The non-employment channel stops replenishing routine, especially routine-cognitive work, and the entry deviation changes while the exit deviation remains essentially stable. The aggregate middle therefore contracts primarily because fewer people enter or re-enter it relative to the share-neutral benchmark, not because a new wave of routine incumbents is pushed into manual jobs.
That mechanism changes how polarization should be understood. A worker can be affected by the disappearance of an occupational opportunity without ever holding the job that disappeared. Someone who would once have entered routine work may instead enter manual work, remain outside observed employment, or seek a different occupation. Meanwhile, an incumbent routine worker may still upgrade into abstract work. The same stock chart combines those different experiences unless the analyst follows flows.
The claim is not universal. The IZA study follows French salaried employment and uses an out-of-employment state that can include forms of work outside the panel. Its discussion of ICT prices is a plausible explanation of the timing and routine-cognitive concentration, not a completed causal identification. The OECD study is a cross-country proof of concept for Italy, the United Kingdom, and the United States, not a test that every labour market follows the French pattern. The defensible conclusion is a framework: structural change must be analysed through entry, incumbent movement, local skills, and local demand together.
Keep the evidence layers connected but distinct
A useful diagnosis has several layers that should inform one another without being collapsed into one measure.
The stock layer describes where employment is observed. It can show that routine work is shrinking while abstract and manual work expand, or that vulnerable work is concentrated in a region. It cannot show whether the change comes from entry, exit, direct job-to-job movement, or a changing denominator.
The flow layer explains how the stock changes. The French decomposition separates direct movements among manual, routine, and abstract work from entries and exits through non-employment. It reveals that a stable upgrading ladder can coexist with a narrowing entry pipeline. This layer identifies the affected margin, but it does not by itself establish the ultimate cause.
The worker layer asks who can use each route. Education changes per-worker transition probabilities, yet non-college workers account for most aggregate routine-to-abstract flows because they are a larger group. Tenure is associated with greater stability, and gender differences appear across upgrading, exit, and re-entry. These patterns show unequal access, but the regressions are partial associations, not levers with guaranteed effects.
The skill layer maps possible destinations. The OECD's occupational-proximity measure uses current job-posting requirements, weights specific skills more heavily than ubiquitous ones, and treats overlap directionally. A destination can cover much of an origin worker's skill set without the reverse being true. The score helps identify candidate pathways across sectors and titles, but it does not certify competence, price training, or measure job quality.
The place-and-demand layer determines whether a candidate destination is available. A vacancy must exist in the local market or be reachable through a realistic mobility decision.