The question
Is far-right activism a sequence of isolated events—or a local ecosystem?
A concert, a march, an assault: the public record often presents them as separate incidents. Geography asks a harder question. Do they keep returning to the same places, and does one channel reliably precede another?
Co-locationThe maps can show whether channels share a local geography.
SequencingThe county tape and lag model can show whether activity tends to precede activity.
CausationNeither design can prove that concerts produce protests, violence, or votes.
Opening the county files…
07 · The verdict
A shared local geography is not a causal machine.
Local ecosystem
Music and protest recur in many of the same counties. The overlap remains above a benchmark that preserves each Land’s own totals, especially across the full 2013–2024 period. Adjacent covered county-years are also highly persistent.
Conveyor belt
Once persistent county differences, common year shocks, population exposure, and prior protest activity enter the model, lagged music is estimated imprecisely. The data do not resolve a predictive dose–response.
Causal mechanism
Selection, reporting, organisation, opportunity, policing, and omitted local conditions can shape every channel. Voting correlations are ecological. No individual pathway is identified.
Germany’s far right has three maps. They overlap often enough to describe an ecosystem—and too ambiguously to call it a pipeline.
08 · Methods and data
The audit trail travels with the story.
Every public value is generated from frozen intermediate files. Violence non-coverage remains distinct from zero from the raw panel to the maps, state typology, transitions, Bundesland aggregates, and vote pairing.
Source correction
The deposited panel is preserved unchanged. In the derived analysis only, Hamburg’s 2015–2021 violence cells are treated as coverage unavailable, matching the published coverage table rather than replication-code-created zeros.
Model boundary
The negative-binomial panel model is descriptive. County and year effects, a population offset, and lagged protest activity improve the comparison; they do not turn the estimate into a causal effect.
Geographic bridge
All 400 FARAC counties map without loss to 396 stable election regions. The 2025 source begins with 401 official Kreis rows and reaches that same frame through five additive pairs, including Berlin East and West.
CSV extracts are offered for inspection and documented reuse under their source-specific terms. The complete claim ledger, source register, story contract, and chart plan remain machine-readable alongside them.
- County-year activity4,800 rows · counts, rates, typology, and violence coverage
- Bundesland aggregates192 Land-years · totals, rates, and coverage
- Stable election regionsSeven elections · six prior-year activity pairings · 396 regions
- Bundestag election archiveSeven elections · 2002–2025 · outcome and pairing status
- Voting correlationsSix elections · raw, rank, weighted, and within-Land correlations
- Voting regression results18 election × channel Land-FE OLS summaries · HC3 intervals
- Lag-model resultsPrimary NB2, PPML, and sensitivity estimates