#41 - David Roodman on incarceration, geomagnetic storms, & becoming a world-class researcher

David Rude's Career Trajectory:

• Studied theoretical mathematics at Harvard (graduated 1990) but transitioned to social science research without a formal plan.
• Attended Cambridge for a one-year mathematics program (Tripos Part 3) but failed exams and left to pursue practical, real-world questions.
• Held roles at the Worldwatch Institute (9 years), Center for Global Development (11 years), Bill & Melinda Gates Foundation (6 months, fired), GiveWell, and Open Philanthropy Project (since 2015).
• Does not hold a PhD, crediting his independence from academic incentives as a key factor in his willingness to challenge established studies.

Research Methodology and Philosophy:

• Specializes in replicating and reconstructing existing empirical studies, often re-running analyses on original data rather than conducting new primary research.
• Approximately 50% of the time, his re-analysis leads to questioning the original conclusions, though fundamental contradictions are rare; issues often arise from methodological robustness or alternative interpretations.
• Believes academic incentives (tenure, publication pressure) discourage challenging senior researchers' work, favoring "new" research over "review" work.
• Uses coding skills to build custom programs when standard software (e.g., Stata) lacks necessary methods for specific replications.
• Adopts a "contrarian" stance, often scrutinizing studies claiming large, statistically significant results from non-experimental data.
• Avoids complex statistical "black boxes," preferring simple, transparent methods to reduce the risk of hidden errors or specification mining.

Criminal Justice and Incarceration Findings:

• Reviewed 20–30 studies on the impact of incarceration on crime; fully reconstructed data/code for 8 studies.
• Deterrence: Found no evidence that harsher sentences (e.g., "Three Strikes" laws) significantly deter crime at policy-relevant margins; concluded deterrence is effectively zero.
• Incapacitation: Confirmed that increasing incarceration reduces crime in the short term (during imprisonment).
• Recidivism/Aftereffects: Concluded that the long-term effects of incarceration are likely harmful, increasing the likelihood of re-offending post-release due to social alienation and lost job prospects.
• Cost-Benefit Analysis: Even assuming the most conservative "devil's advocate" interpretation (that prisons reduce crime), the cost per year of lost liberty ($92,000) roughly equals the value of crime averted ($92,000), resulting in a break-even scenario.
• Parole Bias Discovery: Identified that studies claiming early release increases recidivism were likely measuring "parole bias" (the higher risk of re-incarceration during the parole period itself) rather than the effect of time served.
• Gun Laws: Re-analyzed David Adams' study on gun laws using monthly FBI data (instead of annual) and discovered that the apparent drop in gun crime coinciding with legislation disappeared, revealing a smooth pre-existing decline.

Deworming Intervention Analysis:

• Validated the key 20-year-old study by Miguel and Kremer regarding deworming in Kenya, which showed long-term increases in school attendance and earnings.
• Rejected the study by Hoyt Blakely on deworming in the American South, finding that "sharp jumps" in outcomes were artifacts of data processing and smoothing; results showed long-term convergence instead.
• Current Status: GiveWell's recommendation for deworming now rests almost entirely on the single Miguel/Kremer Kenya study, creating a "weak evidence" scenario.
• Proposed a Bayesian approach to handle weak evidence, acknowledging uncertainty but still favoring the intervention based on priors and the specific benefits of the Kenya study.

Geomagnetic Storm Risk Assessment:

• Evaluated the risk of solar storms disrupting the global electrical grid via induced currents frying large transformers.
• Probability: Concluded the risk of a "Carrington-level" storm (1859) is exaggerated by popular media; estimated the probability at 0–4% per decade based on historical data (1957–present).
• Magnitude: Suggested the 1859 event was likely only twice as large as storms recorded since 1950 (e.g., 1989 Quebec blackout), which were resolved quickly.
• Engineering Mitigation: Noted that modern grids have safety systems to shut down during disturbances, potentially protecting hardware but causing immediate blackouts.
• Tail Risk: Acknowledged uncertainty regarding ancient solar events (700s/900s tree ring data) that may indicate 10–20x larger flares, but the link between flares and geomagnetic storm severity is not well understood.
• Recommendation: Do not panic, but invest in research (e.g., testing transformers) to reduce uncertainty; currently, this is not a primary focus of Open Philanthropy grants compared to AI safety or pandemics.

Career and Organizational Insights:

• Think Tanks vs. Academia: Think tanks offer interdisciplinary exposure and policy relevance but lack the formal progression paths of academia; many lack PhDs at senior levels, though advancement is difficult without one.
• Gates Foundation Experience: Left the Gates Foundation due to a poor fit with the complex internal politics and hierarchy of a 1,200-person organization.
• Open Philanthropy Environment: Values autonomy, small size, and the ability to influence decision-making directly; encourages deep dives into replication and cause prioritization.
• Self-Development: Recommends learning statistics by replicating published papers; suggests that studying technical subjects (math/coding) is more valuable for young researchers than history/literature due to the discipline required to master "slow-mode" thinking.
• Writing Strategy: Advocates for writing press releases or public drafts early to define the bottom line of a project, even if the final output changes.

Book and Public Engagement:

• Wrote a book on microfinance ("Microfinance" series) partly by publishing drafts online (blogging), which generated feedback and interest.
• Microfinance Conclusion: Microfinance is not a "silver bullet" for poverty; over-emphasis on microcredit (loans) created bubbles and risk.
• Recommendation: Advocate for shifting focus toward savings, insurance, and local financing rather than external investor capital.

Recommended Reading:

• The Righteous Mind by Jonathan Haidt (on moral psychology and tribalism).
• Tribal Minds (likely Moral Tribes) by Joshua Greene (on using utilitarianism as a common currency for inter-group cooperation).