Selected witness · Thuringia · AGS 16075
Saale-Orla-Kreis
Included in the predeclared 300-district common sample.
- AfD 2019
- 27.1%
- AfD 2024
- 37.6%
- Change
- +10.6 pp
- Population lens
- −0.12 SD
Data story · Political economy · Updated 3 September 2026
Regional Retreat Reached the Ballot Box.
AfD support did not rise in a straight line. It surged in 2017, receded in 2021, and rose sharply again in 2025. The geography proved more durable than the electoral level; whether demographic retreat produced it—and whether politics now reinforces retreat—remains unidentified.
396 Kreise · 2024 political geographyPopulation retention
vs AfD share
vs AfD share
Case file 01What the record establishes
Median Kreis · AfD share7.03 → 15.18%
2014 to 2024; 9.63% at the 2019 midpoint.
2024 median · geography31.91 / 14.02%
Eastern non-city Länder versus West; Berlin separate.
Population-retention coefficient-2.446 pp
Per standard deviation within Länder; adjusted p=0.0345.
The proposition
The political geography is visible. The mechanism is not yet a verdict.
In August, I argued that regional retreat was becoming political: thinning services, shrinking opportunity, demographic loss, and weaker local capacity could increase demand for parties whose policies might make those conditions worse.
That is a consequential proposition because it joins two different claims. The first is empirical: places under greater regional pressure should show a distinct political pattern. The second is dynamic: electoral success should feed back into capacity through policy. One dataset cannot identify both.
Original proposition · LinkedIn
“Regional retreat is becoming political. The AfD could make it worse.”
Visible fact · The vote moved
The median district shifted by more than five percentage points.
Across the 396 stable Kreise, the median AfD share in the European Parliament election rose from 7.03% in 2014 to 9.63% in 2019 and 15.18% in 2024. The median district-level change between the last two elections was 5.36 percentage points.
Levels remained sharply different. The 2024 median reached 31.91% across the five eastern non-city Länder and 14.02% in the West. Yet that contrast is an orientation, not an explanation. History, party organization, income, migration, settlement structure, and institutional trust all travel with geography.
European-election evidence · District on the witness stand
Choose a Kreis. Keep the claim boundary fixed.
One district travels through the European-election trajectory, capacity field, and coefficient test. The selector changes the witness—not the rules of evidence. Bundestag evidence enters below at the governed 16-Land level; it does not relabel this district estimate.
Exhibit 01 · Descriptive trajectory
AfD support rose. The starting points were already different.
Valid-vote share · European Parliament elections
Current lensWorking-age population retention
Higher values mean stronger retention or growth among residents aged 20–79.
Exhibit 02 · District evidence field
Working-age population retention and 2024 AfD support
Deviation from each Land average
Eastern non-city West Berlin Outside 300-district estimation sample
Exhibit 03 · Predeclared inference
Population retention survives. Aging alone does not.
AfD share percentage points per 1 SD · 300 districts
Selected district · exact values
| Measure | Value | Unit / orientation |
|---|---|---|
| AfD valid-vote share, 2014 | 8.338 | percent |
| AfD valid-vote share, 2019 | 27.057 | percent |
| AfD valid-vote share, 2024 | 37.641 | percent |
| Working-age population change | −0.232 | annualized percent; higher = stronger retention |
| Old-to-young ratio change | +0.285 | annualized percent; lower = less aging pressure |
| Employment per resident change | −1.236 | annualized percent |
Current test · exact estimate
- Coefficient
- −2.446 pp / SD
- Approx. 95% interval
- [−3.687, −1.205]
- Wild-cluster p
- 0.0011
- Family-adjusted p
- 0.0345
The URL preserves this evidence state.
Saale-Orla-Kreis. AfD 27.1 percent in 2019 and 37.6 percent in 2024. Working-age population retention: −0.12 standard deviations.
Exhibit 04 · Federal-election history
Four elections. One durable geography.
AfD support surged in 2017, receded in 2021, and rose sharply again in 2025. The level moved; the regional map persisted.
16 Länder · four final Bundestag results · equal-Land comparisonNational vote pathNot a straight line
- 20134.7%baseline
- 201712.6%+7.9 pp
- 202110.4%-2.2 pp
- 202520.8%+10.4 pp
Stable-Kreis rank persistence
2013 → 2017ρ 0.506
2017 → 2021ρ 0.934
2021 → 2025ρ 0.969
2013 → 2025ρ 0.316
Bundestag election · 2025Young-women deficit and AfD share
valid second votes
Eastern non-city Land City-state Other Land Raw linear gradient
Near-contemporaneous cross-ballot concordance
2025 BTW ↔ 2024 EPPearson 0.995 · Spearman 0.982
2021–25 BTW ↔ 2019–24 EPPearson 0.990 · Spearman 0.974
The matched outcomes are nearly identical across the same 16 Länder. That is ballot robustness and geographic persistence—not an independent replication.
Exact evidence · all 16 Länder
| Land | Women / 1,000 men | BTW 2013 | BTW 2017 | BTW 2021 | BTW 2025 | Δ BTW 2013–17 | Δ BTW 2017–21 | Δ BTW 2021–25 | Δ BTW 2013–25 | EP 2019 | EP 2024 | Δ EP |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 909.6 | 4.57% | 8.18% | 6.78% | 16.29% | +3.62 | -1.40 | +9.50 | +11.72 | 7.45% | 12.18% | +4.73 | |
| 998.2 | 4.17% | 7.82% | 5.03% | 10.87% | +3.65 | -2.80 | +5.84 | +6.70 | 6.49% | 8.01% | +1.53 | |
| 909.9 | 3.73% | 9.09% | 7.44% | 17.84% | +5.36 | -1.65 | +10.40 | +14.10 | 7.95% | 13.24% | +5.29 | |
| 899.2 | 3.74% | 10.00% | 6.88% | 15.07% | +6.27 | -3.12 | +8.19 | +11.34 | 7.68% | 10.23% | +2.55 | |
| 922.2 | 3.92% | 9.42% | 7.26% | 16.82% | +5.50 | -2.17 | +9.56 | +12.90 | 8.52% | 12.62% | +4.10 | |
| 913.4 | 5.60% | 11.91% | 8.82% | 17.78% | +6.30 | -3.09 | +8.97 | +12.18 | 9.90% | 13.61% | +3.70 | |
| 904.1 | 4.81% | 11.25% | 9.22% | 20.09% | +6.44 | -2.03 | +10.87 | +15.28 | 9.85% | 14.67% | +4.82 | |
| 906.5 | 5.25% | 12.19% | 9.60% | 19.79% | +6.94 | -2.59 | +10.18 | +14.54 | 10.00% | 14.73% | +4.73 | |
| 899.7 | 4.31% | 12.39% | 8.98% | 19.01% | +8.08 | -3.41 | +10.03 | +14.70 | 8.51% | 12.64% | +4.12 | |
| 901.1 | 5.21% | 10.07% | 10.05% | 21.56% | +4.86 | -0.02 | +11.51 | +16.35 | 9.64% | 15.69% | +6.05 | |
| 970.4 | 4.93% | 12.05% | 9.45% | 15.23% | +7.12 | -2.60 | +5.78 | +10.31 | 9.90% | 11.60% | +1.71 | |
| 861.6 | 5.98% | 20.20% | 18.14% | 32.49% | +14.21 | -2.06 | +14.36 | +26.51 | 19.91% | 27.52% | +7.61 | |
| 913.0 | 5.63% | 18.59% | 17.99% | 34.99% | +12.95 | -0.59 | +17.00 | +29.36 | 17.68% | 28.34% | +10.66 | |
| 905.0 | 6.76% | 27.02% | 24.65% | 37.30% | +20.26 | -2.37 | +12.65 | +30.53 | 25.28% | 31.77% | +6.49 | |
| 856.3 | 4.21% | 19.63% | 19.59% | 37.10% | +15.42 | -0.04 | +17.50 | +32.88 | 20.37% | 30.52% | +10.16 | |
| 866.7 | 6.17% | 22.72% | 23.97% | 38.55% | +16.55 | +1.25 | +14.58 | +32.38 | 22.53% | 30.74% | +8.21 |
Ten governed descriptive association views
| Election outcome | Spearman ρ | Raw p | Holm p | Raw slope | East + city slope |
|---|---|---|---|---|---|
| Bundestag · 2013 | 0.1853 | 0.4868 | 0.4868 | 0.70 | -0.88 |
| Bundestag · 2017 | 0.5412 | 0.0318 | 0.1590 | 8.61 | 0.02 |
| Bundestag · 2021 | 0.5294 | 0.0389 | 0.1590 | 10.36 | 0.51 |
| Bundestag · 2025 | 0.5971 | 0.0158 | 0.0948 | 17.77 | 2.85 |
| Bundestag · 2013 → 2017 | 0.4941 | 0.0527 | 0.1590 | 7.91 | 0.90 |
| Bundestag · 2017 → 2021 | 0.3118 | 0.2358 | 0.4716 | 1.75 | 0.49 |
| Bundestag · 2021 → 2025 | 0.6147 | 0.0115 | 0.0805 | 7.41 | 2.34 |
| Bundestag · 2013 → 2025 | 0.6647 | 0.0067 | 0.0536 | 17.07 | 3.73 |
| European · 2024 | 0.5294 | 0.0366 | 0.0528 | 15.11 | 1.07 |
| European · 2019 → 2024 | 0.5676 | 0.0224 | 0.0528 | 5.34 | 0.85 |
Thüringen: 866.7 women per 1,000 men; Bundestag 2025, election result dated 2025-02-23, 38.55 percent; demographic snapshot 2025-12-31.
Counter-evidence · Three limits that change the claim
The strongest result is narrower than the original story.
Aging by itself does not survive.
The lower-aging-pressure component has a coefficient of -0.374 percentage points and an adjusted p-value of 0.9989. The evidence does not support a portable sentence that older places simply produce more AfD support.
The composite is not the result.
The three-component capacity measure points in the expected direction, but its family-adjusted p-value is 0.2238. Compressing population, aging, and employment into one “retreat score” would hide the fact that the components behave differently.
Four elections are not four mechanism tests.
The Bundestag record reveals a non-monotonic national path and an increasingly stable regional ranking. Pairing any of those elections with a demographic stock measured only at the end of 2025 remains retrospective. It describes alignment, not temporal precedence, prediction, or an independent mechanism test.
Capacity and political support move together in one bounded way.
- Official AfD valid-vote shares for 2014, 2019, and 2024 on one stable district geography.
- Official 2022–2023 population and employment inputs.
- One family-adjusted component result: stronger working-age population retention is associated with lower 2024 AfD support within Länder.
- National AfD second-vote share moved from 4.70% in 2013 to 12.64% in 2017, 10.39% in 2021, and 20.80% in 2025.
- Stable-Kreis rank persistence rose from ρ=0.506 in 2013–2017 to 0.934 in 2017–2021 and 0.969 in 2021–2025.
Identification gap
AfD influence causes services and regional capacity to retreat further.
- A policy exposure or quasi-experimental treatment is required.
- Comparable service, investment, migration, and employment outcomes must follow the treatment.
- Pre-trends, concurrent shocks, and selection into political control must be tested.
Verdict · publish with limits
The electoral level moved. The map proved harder to move.
The longer Bundestag record changes the shape of the story. AfD support did not rise monotonically: it surged in 2017, receded in 2021, and rose sharply again in 2025. What persisted more reliably was the regional geography once it had formed. That evidence supports a durable political map exposed to election-specific waves—not a single demographic law. The only family-adjusted district capacity result remains the 2024 European-election association with working-age population retention. Neither the retrospective demographic overlay nor cross-ballot similarity identifies individual voters, establishes a demographic mechanism, predicts future elections, or shows that AfD policy deepens regional retreat.
Regional retreat is visible in the political geography. The feedback loop remains a hypothesis.
Retain people
For regional policy, the measurable warning is loss of working-age population—not chronological aging alone.
Measure services
Build a harmonized panel of public-service access, local investment, migration, and employment before claiming a retreat mechanism.
Test the loop
Align demography before each election, then measure services and regional capacity after actual policy exposure. Landtag elections remain a later empirical check, once their outcomes and consequences can be observed.
Methods · compact, predeclared, reproducible
One stable geography. Separate components. No causal shortcut.
The release harmonizes official 2014, 2019 and 2024 European-election results to 396 stable Kreise. AfD support is votes divided by valid votes. Capacity is measured through separate 2022–2023 changes in residents aged 20–79, the old-to-young ratio, and employment per resident.
The association results use equal-weight OLS with Land fixed effects, a 300-district common sample, CRV3 standard errors, 9,999 null-imposed wild-cluster replications, and Romano–Wolf adjustment across the predeclared family. The approximate intervals shown in the chart use 14 cluster degrees of freedom.
The Bundestag update is a separate 16-Land descriptive history—not a relabeled district capacity model. It uses final 2013 and 2017 results, the revised final 2021 result, and the final 2025 result; it exposes all four levels, the three adjacent changes, and the 2013–2025 long-run change. Counts are aggregated before shares are calculated. Stable-Kreis rank correlations measure geographic persistence, not individual-voter continuity.
The young-adult sex ratio is measured on 31 December 2025. Comparing it with earlier elections asks whether today’s demographic geography aligns with a persistent political map. It cannot show that the measured imbalance preceded or caused votes in 2013, 2017, 2021 or 2025. Eight Bundestag-history rank views use a separate fixed-seed permutation and Holm family; East and city-state terms are descriptive sensitivity checks. The European comparisons retain their original robustness family.
This story extends the public research record in Composition Is Not Adjustment, which explains why comparison across places, adjustment within places, and causal effects are different estimands.
For the longer demographic baseline, see Germany Aged Everywhere. Growth Started from Different Places. That article explains why similar aging outcomes can emerge from different regional trajectories.
396stable districts in the descriptive panel
300districts in the common estimation sample
15Land clusters in the primary estimates