A year's worth of education for under $1 and other "best buys" in development | Rachel Glennerster

Current Work & Research Focus:

• Rachel Glenister currently serves as Chief Economist at the UK's Department for International Development (DFID).
• She is evaluating a randomized mass media radio campaign by Development Media International on family planning in Burkina Faso.
• Burkina Faso is selected for evaluation due to diverse radio stations across different languages (reducing spillover) and the ability to randomize radio distribution to women without devices.
• Family planning is prioritized for its potential to drive demographic transitions, benefiting women's economic status, economic development, and child health.

Trajectory of Development & Aid (Past 30 Years):

• The aid sector has shifted from theory-driven "fads" (e.g., physical capital vs. human capital) to evidence-based decision-making, particularly within DFID and the World Bank.
• Academic development research has moved from purely descriptive work and ideology toward rigorous causal analysis via Randomized Control Trials (RCTs).
• RCTs have bridged the gap between academic research and policy implementation, creating a "50-50" split in methodology usage today.
• There is a trend of integrating development economics with global economics, allowing lessons from developing contexts (e.g., behavioral economics) to inform rich-country policy.

Role and Limitations of RCTs:

• RCT proponents argue RCTs are a tool, not the only methodology; descriptive work is essential to identify the correct problem before testing solutions.
• RCTs are most valuable when testing generalizable theories of human behavior rather than just specific program outcomes.
• Example: An immunization study in India revealed "persistence" (dropping out of the schedule) was the key barrier, not distrust or clinic closures; however, the specific intervention (lentils) was not scalable.
• Most RCT funding targets developing country governments (who control the majority of poverty relief spending) rather than foreign aid donors.
• RCTs can identify broad policy shifts, such as the need to lower curriculum levels in India to match student actual grade levels, rather than just testing specific inputs like textbooks.

Economic Growth vs. Specific Interventions:

• Glenister argues that economic transformation (e.g., market reforms in India and China) yields the largest poverty reductions, often dwarfing micro-interventions.
• DFID is shifting focus to supporting economic transformation opportunities, such as Ethiopia's current reforms, where policy changes can have massive impact.
• Health and education remain critical because they provide the human capital necessary for sustained economic growth (e.g., 10% return on education investment).
• Glenister views "aid effectiveness" as a misnomer; the goal is "government effectiveness" and poverty effectiveness at the local level.

Overrated/Underrated Interventions & Concepts:

• Underrated:
  • Cash Transfers: Historically appropriately rated, but recent negative results have led to them being underrated despite long-term benefits.
  • Cracking Illicit Financial Flows: Stopping money laundering out of developing countries to tax havens is high-impact; Western nations share culpability for hosting these flows.
  • Micronutrient Supplementation & Anemia: Underrated solutions, though current delivery mechanisms need improvement; anemia significantly impacts productivity and cognition.
  • Starting Businesses in Developing Worlds: Social entrepreneurship is often overrated; managing existing businesses to improve management practices and productivity offers higher impact.
  • Macroeconomic Policy: Inflation control is "nailed," but further optimization remains under-discussed.
• Overrated:
  • Charter Cities: Viewed as politically unfeasible due to institutional resistance; Glenister prefers focusing on practical, incremental institutional improvements (e.g., women's political quotas in India).
  • Social Entrepreneurship: Often yields small-scale, non-scalable results compared to working within large organizations or improving private sector management.
  • Pre-registration (Pre-analysis plans): Can hinder scientific discovery by preventing researchers from adapting analysis when new patterns emerge in the data; theory-based priors may be superior.
  • Confirmatory News Consumption: Reading only news that confirms existing biases; Glenister advocates for reading diverse, specialized sources (e.g., African politics).

Debate on Generalizability (Eva Vavaut vs. Glenister):

• Glenister disputes Eva Vavaut's finding that RCT results vary by ~100% across contexts, arguing Vavaut conflates different causes for variance.
• Sources of Variation:
  • The underlying problem may not exist in the new context.
  • Implementation quality varies (e.g., 80% vs. 20% take-up).
  • Behavioral responses to incentives are generally consistent across contexts if the program is implemented correctly.
• Glenister argues that most variation stems from bundling different program "packages" or poor implementation, not a lack of generalizable behavioral principles.
• Prediction of Replication:
  • Experts (and even aggregated non-experts) are surprisingly good at predicting which studies will replicate, though this information is often not in the written paper.
  • Glenister advocates for soliciting "priors" (pre-study predictions) from experts to improve meta-analysis and Bayesian statistical applications.
  • Policy insights should focus on cross-cutting principles (e.g., price sensitivity, convenience) rather than specific program mechanics, which reduces the need for new RCTs.

Impact Distribution & Cost-Effectiveness:

• Returns on social interventions are highly skewed: a small fraction of interventions (e.g., specific education nudges, deworming) generate the vast majority of impact.
• Education: The most cost-effective interventions involve rearranging students by ability level or sending information via mobile phones; traditional inputs (textbooks, extra teachers) often show zero impact.
• Scale vs. Depth: Glenister critiques the development community's preference for "perfect" small-scale projects over "okay" large-scale interventions, citing a psychological need for visible, tangible success stories (e.g., a new school building vs. text message reminders).
• Portfolio Approach: A few massive, scalable interventions (e.g., mobile extension, graded classrooms) can achieve massive world-GDP gains despite having low individual barriers.

Career & Policy Advice:

• Policy Influence: Success requires making partners feel ownership of the idea; data should inform decisions quietly rather than being used as a blunt instrument.
• Civil Service Impact: Civil servants should prioritize reading peer-reviewed academic journals over glossy think-tank reports to avoid "bad science" that matches political narratives.
• Skill Investment: Learning advanced data analysis and economics tools significantly increases a researcher's ability to influence complex policy.
• Addressing Myths: There is a need to counter public misconceptions that poverty is increasing globally; evidence shows the majority of the world's poor are better off.
• Illicit Flows: Listeners in the UK/US can impact policy by lobbying against the use of their jurisdictions for laundering illicit funds from developing nations (e.g., the Mozambique loan scandal).

Future of Life & 80,000 Hours Context:

• Glenister agrees with the core Effective Altruism (EA) goal of finding high-impact interventions but emphasizes the difficulty of scaling "invisible" successes like policy nudges.
• She notes that while DFID can commission RCTs, the ecosystem often lacks organizations capable of scaling single, evidence-based interventions to the magnitude required for maximum impact.
• The podcast concludes with a response from Eva Vavaut acknowledging the heterogeneity of RCT results but agreeing with Glenister on the need for better modeling and the use of ex-ante forecasts to improve generalization.