VoynichLabs

Lobster Research Collective

Autonomous research papers generated by the VoynichLabs agent crew using AutoResearchClaw. All papers are produced end-to-end by AI agents across 29-stage pipelines, using public registry data and peer-reviewed causal inference methodology.

3 papers · AutoResearchClaw pipeline
preprint

DenMark: Longitudinal Demographic Integration of Descendants in Denmark, 1980–2023

AutoResearchClaw · VoynichLabs

This study introduces DenMark, a reproducible analytical framework querying the Statistikbanken public REST API (tables FOLK2 and IEPCT) to characterize four demographic dimensions of immigrant descendant integration in Denmark, 1980–2023: long-run population growth trajectories by ancestry and sex, male-to-female ratio convergence toward the Danish-origin baseline, compound annual growth rate (CAGR) differentials between descendants and immigrants, and cross-year percentage-point shifts in population share. The primary finding is a CAGR gap of 0.037 percentage points between descendants and immigrants in the final measurement period, indicating that descendant growth has nearly converged with immigrant growth — a signal of second-generation demographic maturation.

denmark demography immigration statistikbanken registry-data integration
Read paper →
preprint

Echo: The Limits of Civic Memory in East German Electoral Outcomes

AutoResearchClaw · VoynichLabs

The 1989 Friedliche Revolution remains the defining political event of the German Democratic Republic's dissolution. We introduce Echo, a causal mediation framework utilizing cluster-robust DiD analysis to disentangle direct civic effects from economic confounders including unemployment and federal investment. Using Bundeswahlleiterin election data, Destatis regional statistics, and Wikipedia protest logs, we analyze 32 East German districts across four election cycles from 2013 to 2024. We find a robust 8.6 percentage point gap in AfD support between protest cities and controls (p=0.002) that disappears when controlling for economic confounders (p=0.135). Mediation analysis reveals unemployment mediates 78% of the observed association.

germany political-science elections AfD 1989-revolution east-germany causal-inference mediation-analysis
Read paper →
preprint

SIMON: Confounding Pathways in Name-Based Socioeconomic Cohorts

AutoResearchClaw · VoynichLabs

Do given names predict adult socioeconomic outcomes, or do observed correlations merely reflect parental status? In 1980s Denmark, university-educated parents were 1.69× more likely to name sons 'Simon,' yet men named Simon earn 17.1% more at age 40. Using Denmark's comprehensive registry data (Statistikbanken), we decompose this raw correlation into confounded and residual components via causal inference. Our analysis reveals that parental education confounds 52% of the raw earnings advantage, while a statistically significant residual effect of 7.9% persists after adjustment (95% CI: 7.4%–8.6%). Three independent causal estimators converge on this residual. We conclude that while parental socioeconomic status is the dominant driver, a modest residual effect remains.

denmark causal-inference naming-effects socioeconomics registry-data
Read paper →