Dr. Michael SchymuraEconomics of AI & Innovation
Pre-submission working paper16 Länder · 544 Land-years · 396 Kreise

Regional demography / convergence / growth

Composition Is Not Adjustment

The coefficient survived. Its meaning did not.

Older German regions grew more slowly. Yet eastern catch-up, falling hours, and within-Land estimates show why that pattern is a warning signal rather than a mechanical forecast of what aging will do.

Open the evidence Download 37-page paper
Revision23 August 2026
Long paperv0.9.3 · 37 pages
Snapshotf22d95bd…

The empirical result

The number is real. The portable interpretation is not.

Across all 16 Länder from 1991 to 2024, a 10% higher period-average ratio of residents aged 50–79 to residents aged 20–49 is associated with 0.317 percentage points less annual growth in real workplace output per broad-adult resident.

That coefficient survives exhaustive small-cluster inference, spatial covariance checks, and deleting each Land in turn. However, it weakens after 2000, approaches zero when the eastern group is removed, and points in the opposite direction when the same Land is followed over time.

A negative comparison across places is not the same estimand as adjustment within a place. Neither is automatically causal.
01 / Benchmark association−0.317 pp

Older Länder grew more slowly.

Annual growth per 10% higher period-average old-to-young ratio. WCR11 p=.010; CRV3 p=.070.

02 / Convergence fact42% → 72%

Eastern output caught up.

East as a share of the western real workplace-output level per broad-adult resident, 1991–2024.

03 / Estimand conflict3 of 5

Composition and adjustment differ.

Familywise within-versus-between tests reject equality for three accounting outcomes.

Länder
16
Land-years
544
Observed Kreise
396
Causal claims
0

Interactive evidence lab

Inspect the geography, the convergence path, and the estimand before accepting the story.

All views use the frozen 0.9.3 outputs. Hover, focus, drag, or use the explicit controls. Every 3D encoding has a complete 2D and textual counterpart.

Evidence 01 / animated aging atlas

Every Land aged. Not every Land aged alike.

Move the annual clock across all 544 observations. Fill records the ratio of residents aged 50–79 to residents aged 20–49; borders retain East, West, and city-state geography without turning it into a treatment.

German Länder old-to-young ratios in 2024An animated choropleth of all 16 Länder. Select a Land for its exact value and 1991 comparison.
Year2024

Selected Land

Mecklenburg–Western Pomerania

132.4

residents 50–79 per 100 residents 20–49

In 1991: 64.7.

Official Länder population panel, 1991–2024. The aging measure is not the conventional old-age dependency ratio. Census-basis changes in 2011 and 2022 remain visible in the data contract.

Spatial instrument / Babylon.js

Aging has height, but not causal altitude.

Each official BKG polygon is extruded by its observed ratio. Drag to rotate; the 2D atlas above remains the authoritative accessible view.

SelectedMecklenburg–Western Pomerania
Year2024
Ratio1.324

Evidence 02 / East–West convergence

Eastern output caught up from a much lower base.

The lines compare real workplace output per resident aged 20–79 for five eastern and eight western territorial Länder. City states are excluded; the shaded distance is a level gap, not an aging effect.

East and West real workplace output per broad-adult resident, 1991–2024Eastern output rises from 18,834 euros to 42,741 euros while western output rises from 44,477 euros to 59,534 euros. The eastern level moves from 42 to 72 percent of the western level.€20k€30k€40k€50k€60k19912001201120212024CENSUS BASISCENSUS BASISMove to inspect a year
East / West level42% → 72%
Faster eastern real output+1.10 pp
Smaller eastern adult base+0.61 pp

Ratio of sums in 2020-euro chain-volume data. Workplace output uses a resident denominator, so commuting remains relevant. The 2022–2024 observations are a descriptive extension beyond the registered windows.

Evidence 03 / composition versus adjustment

A photograph of places is not a film of change.

Each outcome carries three estimands on one axis: persistent differences across Länder, movement within a Land, and year-to-year change. The disagreement is the result.

Within, between, and first-difference aging coefficients for five economic outcomesAcross-Land coefficients are strongly negative for output outcomes while within-Land and first-difference coefficients are positive.Within LandAcross different LänderFirst difference-3-2-10Output / broad-adult resident+0.245−1.863+0.174DIFFERENT · p=0.030Output / actual hour+0.354−1.166+0.327DIFFERENT · p=0.021Actual hours / workplace worker−0.065+0.081−0.066NOT RESOLVED · p=0.527Workplace / resident employment+0.052−0.633−0.022DIFFERENT · p=0.037Resident employment / broad-adult resident−0.095−0.144−0.065NOT RESOLVED · p=0.992

Exact WCR11 intervals. The right-side tests compare within and between coefficients under one direct restriction per outcome with familywise correction across five outcomes. Positive within estimates do not establish a beneficial causal effect.

Evidence 04 / inference and robustness

The sign survives deletion. The estimate moves with the design.

No single Land creates the benchmark. Historical and geographic restrictions move it sharply toward zero, while small-cluster methods disagree about certainty.

Headline association across five robustness specificationsThe benchmark is negative and precisely below zero under WCR11. Post-2000, east-deleted, western-only, and east-window specifications are close to zero with wide intervals.-1.0-0.50.00.51.0All Länder, three windowsAll Länder, post-2000Eastern group deletedWestern Länder onlyEast-by-window controlsLEAVE-ONE-LAND RANGE
Leave-one-Land−0.457 to −0.265
Conley 300 kmSE 0.103
Moran windows0 of 3 significant

All estimates remain noncausal associations. Deletion stability addresses single-Land leverage; it does not identify the mechanism or make the full-period magnitude portable.

Evidence 05 / the local economy

The East–West divide is real, but too coarse for local policy.

Every point is one of 396 stable Kreise observed from 2012 to 2021. No density, geographic surface, or missing district is simulated.

Distribution of annualized output-per-resident growth across 396 German KreiseEastern and western distributions overlap substantially. Most variation occurs between districts within the same Land.0%1%2%3%4%5%6%7%East · 72West · 324
Output-growth variation87.2% within Länder
Output/hour variation76.3% within Länder
East–West overlap62.7%

Current-price descriptive layer, 2012–2021. Standard hours are not actual hours. The Kreis evidence changes the geographic scale of the diagnosis; it is not a second aging estimator.

Productivity is not capacity

Output per hour rose because hours fell faster than output.

A productivity ratio can improve while the economy’s total capacity contracts. The exact same-system identity keeps numerator and denominator visible.

output per actual hour+0.388

measured efficiency rose

[+0.136, +0.640]

actual hours−0.464

labor input fell

[−0.759, −0.169]

raw output implied−0.076

raw output was roughly flat

exact accounting sum

Claim ceiling

What survives the analysis

  • The paper establishes a persistent negative benchmark association. It does not establish a causal aging penalty.
  • Output per hour rises while total hours fall. The accounting identifies the margin, not an automation mechanism.
  • East–West differences matter, but 87.2% of local output-growth variation lies within Länder.
  • Spatial corrections and leave-one-Land tests preserve the benchmark sign; historical and geographic restrictions move its magnitude.

Long research edition

The 37-page paper is the primary record.

Version 0.9.3 incorporates the within-versus-between family, East–West and period sensitivities, official BKG spatial inputs, Moran diagnostics, panel-Conley covariance, and the output-hours accounting interpretation. It is prepared for submission and has not yet been submitted.

Pre-submission working paper · v0.9.3

Composition Is Not Adjustment: Regional Aging, Convergence, and Growth in Germany

Pages
37
Revision
23 August 2026
File
PDF · 0.71 MB
Download long paperOpen in browser

SHA-256 · aae16fa8532df4ff068d0ab1481ea2f98cfffbb8afd8da64618b9a02397c99b1

Reproducibility

The visuals inherit the paper’s claim discipline.

A deterministic generator validates row counts, geographic keys, benchmark values, and source hashes before producing the public JSON used by every chart.

Immutable research snapshot Download public evidence JSON

Official geometry

BKG VG250 Länder, 1 January 2026; 16 keyed MultiPolygons. © BKG (2026) dl-de/by-2-0; derived geometry modified by the author.

Bounded inference

Exhaustive WCR11, CRV3, familywise direct tests, Conley, and Moran diagnostics.

Frozen inputs

Research snapshot f22d95bd1de0…; no synthetic observations.

Long-paper PDF
aae16fa8532df4ff068d0ab1481ea2f98cfffbb8afd8da64618b9a02397c99b1
ageing_state_panel_1991_2024.csv
05455ae4dc28b0a08967c63de251863e45b281ff5a5531f0e5536162d4f51809
east_west_real_output_per_adult_1991_2024_audit.csv
2e7fd4f5a5f0d037036e6ef060ae757089f05429c561e65426f6e32ee25c7fc2
germany_lander_simplified.geojson
d720d4fc3e68f32c095fdd944b29d006ba8749b30a7841bca6070612420ca4d4
kreis_output_growth_2012_2021.csv
04c39d73e1e4a3b82d525a1f7c44e660f4b80e879ed7379345c9bb1e7cc1de6a
paper_093_headline_robustness.csv
161c6083a7b7400349fc0f7891028db8bb48f2bcd7dbbd4d40f9b87ea50e6fff
paper_093_annual_coefficients.csv
ddbc3491640e6e36cd2eedd2b4c58360ca24df27302d312a3aa3105043419e46
Historical 0.8.1 publication artifacts
Open archived 0.8.1 HTML report Download archived 0.8.1 source package

Bibliography

References

The long paper contains the complete literature argument and citation context.

  1. 01

    Acemoglu, Autor, Beirne & Scott (2026). Baby Busts and Growth Booms: Demographic Change and the Macroeconomy. NBER Working Paper 35401.

  2. 02

    Acemoglu & Restrepo (2017). Secular Stagnation? The Effect of Aging on Economic Growth in the Age of Automation. American Economic Review 107(5), 174–179.

  3. 03

    Acemoglu & Restrepo (2022). Demographics and Automation. Review of Economic Studies 89(1), 1–44.

  4. 04

    Adão, Kolesár & Morales (2019). Shift-Share Designs: Theory and Inference. Quarterly Journal of Economics 134(4), 1949–2010.

  5. 05

    Bode, Dohse & Stolzenburg (2023). Aging and Regional Productivity Growth in Germany. Review of Regional Research 43, 409–432.

  6. 06

    Bönte, Falck & Heblich (2009). The Impact of Regional Age Structure on Entrepreneurship. Economic Geography 85(3), 269–287.

  7. 07

    Borusyak, Hull & Jaravel (2022). Quasi-Experimental Shift-Share Research Designs. Review of Economic Studies 89(1), 181–213.

  8. 08

    Brunow & Hirte (2006). Age Structure and Regional Economic Growth. Jahrbuch für Regionalwissenschaft 26(1), 3–23.

  9. 09

    Brunow & Hirte (2009). The Age Pattern of Human Capital and Regional Productivity. Papers in Regional Science 88(4), 799–823.

  10. 10

    Canay, Santos & Shaikh (2021). The Wild Bootstrap with a “Small” Number of “Large” Clusters. Review of Economics and Statistics 103(2), 346–363.

  11. 11

    Daniele, Honiden & Lembcke (2019). Aging and Productivity Growth in OECD Regions. OECD Regional Development Working Papers 2019/08.

  12. 12

    Dauth, Findeisen, Suedekum & Woessner (2021). The Adjustment of Labor Markets to Robots. Journal of the European Economic Association 19(6), 3104–3153.

  13. 13

    Feyrer (2007). Demographics and Productivity. Review of Economics and Statistics 89(1), 100–109.

  14. 14

    Goldsmith-Pinkham, Sorkin & Swift (2020). Bartik Instruments: What, When, Why, and How. American Economic Review 110(8), 2586–2624.

  15. 15

    Gregory & Patuelli (2015). Demographic Ageing and the Polarization of Regions. Environment and Planning A 47(5), 1192–1210.

  16. 16

    Hayashi (2026). The Root Cause of Japan’s Reversal of Fortune. SSRN Working Paper 4553699.

  17. 17

    Hoshi (2026). Demographic Challenges and Economic Stagnation in Japan. The Japanese Economic Review 77, 411–433.

  18. 18

    Hoshi, Kodama & Li (2025). The Effect of Aging on Entrepreneurship and Aggregate Productivity. The Journal of the Economics of Ageing 32, 100605.

  19. 19

    Lindh & Malmberg (1999). Age Structure Effects and Growth in the OECD, 1950–1990. Journal of Population Economics 12(3), 431–449.

  20. 20

    MacKinnon, Nielsen & Webb (2023). Cluster-Robust Inference: A Guide to Empirical Practice. Journal of Econometrics 232(2), 272–299.

  21. 21

    Maestas, Mullen & Powell (2023). The Effect of Population Aging on Economic Growth, the Labor Force, and Productivity. American Economic Journal: Macroeconomics 15(2), 306–332.

  22. 22

    Mitze & Reinkowski (2011). Testing the Neoclassical Migration Model. Journal for Labour Market Research 43(4), 277–297.

  23. 23

    Monte, Redding & Rossi-Hansberg (2018). Commuting, Migration, and Local Employment Elasticities. American Economic Review 108(12), 3855–3890.

  24. 24

    Romano & Wolf (2005). Stepwise Multiple Testing as Formalized Data Snooping. Econometrica 73(4), 1237–1282.

Public interpretation

The next design must identify the mechanism, not merely another aggregate coefficient.

The companion essay translates the evidence for a broader audience. The paper remains the authoritative research record.

Read the public argument