Grasping Reality on TypePad, by Brad DeLong

Nathan Nunn (2008): The Long-Term Effects of Africa’s Slave Trades, Quarterly Journal of Economics 123 (February): 139–176 <http://www.jstor.org/stable/pdfplus/25098896.pdf>

• From 8 to 6.5 in the natural log…
• exp(1.5) = 4.5: This is enormous! Is this credible? More than 80% of potential wealth today destroyed by something that ended 200 years ago!?
  • Do we need controls?
    • Or, rather, what controls do we need?
• What else might be going on here?
• Do we need instruments?
  • What are our instruments?
  • What is our first stage?
  • Is this a “weak instrument”?
• Under what circumstances is a strong instrument “weak”?
  • Do we need channels?
  • What are our channels?
  • What would a Bayesian say about the publication filter and the file-drawer problem?