The mechanism register.
A mechanism is a claimed coupling between conditions: the reason two indicators multiply rather than add. Nothing is described as connected unless it has an entry here.
Evidence statuses: supported (multiple independent lines), correlational (co-movement shown, causation argued), hypothesis (mechanism argued, evidence thin), retired (failed its terms, kept visible with the autopsy).
M1. Conversion and shelter: the household that never forms
Pathway
Without the entry-level job there is no deposit. Without the deposit there is no household. Without the household, family formation is deferred or abandoned. The two pressures are one biography: the graduate in the childhood bedroom.
Micro anchor
The bottom rung is where AI bites first. The Stanford and ADP payroll finding shows a 16 per cent relative employment fall for 22-to-25-year-olds in the most exposed occupations, through hiring cessation. The coupling tightens fastest where shelter pressure is already highest.
Control case
A country with high youth unemployment but cheap housing should show weaker downstream effects than where both are tight. Spain against the UK and Ireland.
What would falsify it
Panel countries where both families sit in the worst tercile but living-with-parents and partnering rates track the panel average for five years or more.
Status
Supported. The co-movement is documented across EU-SILC and the causal chain is standard household economics. The AI acceleration of the first term is new and rated correlational.
M2. The valve: exit substitutes for voice
Pathway
Where the young can leave, grievance becomes a private decision and departs, taking the people with the energy and qualifications to organise. Where they cannot, grievance pools and becomes collective. Anti-immigration politics tends to close the valve at both ends.
Historical anchors
Ireland in the 1950s and 1980s. Portugal and Italy after the war. Argentina in 1989 and 2001. The UK is the live experiment: three centuries of exporting surplus young ended in 2021.
Control case
Ireland today: severe shelter pressure with an open valve. M2 predicts its legitimacy indicators deteriorate more slowly than a closed-valve country under equivalent pressure.
What would falsify it
A closed-valve, high-pressure country whose legitimacy indicators hold stable for a sustained period, or an open-valve country that ruptures while its exit routes are in use.
Status
Supported for the historical claim. Hypothesis for the strong version, that the valve dominates the rate term ahead of all other variables.
M3. Binds and the posted bond
Pathway
A person with dependants and property has posted a bond to the existing order. Childlessness and non-formation release it. The claim is not that the unbound become violent. It is that the state stops holding anything of theirs, which changes what they will risk.
Control case
High-childlessness, high-pressure societies with strong alternative binds, such as religious affiliation or dense civic association, should show weaker mobilisation than bind-poor equivalents. Poland against the UK.
What would falsify it
Mobilisation waves drawing proportionally from the bound: partnered, propertied, parenting, across multiple cases.
Status
Correlational. The recruitment-demography record is old and broad. The modern panel-data version is untested, and this register is where we test it.
M4. Churn and conversion: the closing of the question
Pathway
Rapid leadership turnover under unresolved structural pressure teaches an electorate that the system, not the incumbent, is the variable. Each failed last chance transfers discredit from persons to the order itself.
Control case
High-churn countries where churn reflects coalition arithmetic rather than performance failure should not show the same anti-system conversion. Dutch coalition churn against UK failure churn.
What would falsify it
Sustained high churn and high pressure with flat or falling anti-system vote share across several panel countries.
Status
Hypothesis.
M5. The accelerant: AI as a rate multiplier
Pathway
General-purpose capability arrives at the entry level first, because the tasks juniors are hired to do are the tasks the systems do. It removes the bottom rung rather than the chain, through hiring cessation, which produces no event for the news to report.
Micro anchors
The Stanford and ADP cohort finding of November 2025, with its February 2026 follow-up. An earlier anchor, a claimed rise in entry-level postings requiring prior experience, was withdrawn in August 2026 when no primary source could be found. Strength is rated on field data, never demos.
Control case
Low-diffusion, high-exposure countries should show the acceleration weakly or not at all. Within countries: less-exposed occupations in the same labour markets.
What would falsify it
Exposed-cohort employment recovering to trend while diffusion continues, or acceleration appearing equally in low- and high-diffusion economies.
Status
Correlational, upgraded from hypothesis in November 2025 by the payroll finding. The most important entry to keep honest, because it carries the project's name.
Last updated