James Heckman and Daniel McFadden
Citation: For their development of theory and methods for analyzing selective samples (Heckman) and the theory and methods for analyzing discrete choice (McFadden).
The key idea
Heckman: selection bias has structure and can be corrected. McFadden: the multinomial logit and conditional logit make discrete-choice data (yes/no, which-brand) econometrically tractable.
The explanation
Heckman's two-step estimator (1979) corrects for sample-selection bias — e.g., observing wages only for people who chose to work. McFadden's logit framework (1974) gave consistent estimators for choice probabilities and won by transforming travel-demand analysis. Together they made microeconometrics rigorous.
Why Africa should care
Heckman correction is essential for any African study of wages, education returns, or labour-market participation — observations are non-random because of who chooses (or is allowed) to work. McFadden's discrete-choice framework underlies every mobile-money adoption study, voter-behaviour analysis, and brand-choice market research conducted in African contexts.
How to use it
Before running an OLS regression on a non-random sample (e.g., wages of women who worked), implement Heckman correction. The selection equation is often as informative as the outcome equation.
Canonical works
- James J. Heckman (1979) "Sample Selection Bias as a Specification Error" Econometrica
- Daniel L. McFadden (1974) "Conditional Logit Analysis of Qualitative Choice Behavior" Frontiers in Econometrics (Zarembka, ed.)
More from Behavioural, empirical, institutional · 2000-2009
- 2001George Akerlof, Michael Spence, and Joseph Stiglitz
Akerlof: in markets with information asymmetry, bad goods drive out good ones (the lemons problem). Spence: education can signal innate ability even if it produces nothing. Stiglitz: insurance markets unravel under asymmetric information; screening contracts emerge in equilibrium.
- 2002Daniel Kahneman and Vernon Smith
Kahneman: prospect theory — people are loss-averse, weight probabilities non-linearly, frame matters. Smith: experimental markets reach competitive equilibrium remarkably fast even with few traders.
- 2003Robert Engle and Clive Granger
Engle: ARCH/GARCH models — volatility clusters, time-varies, can be forecast. Granger: cointegration — non-stationary series can share a stationary long-run relationship; pairs trading and error-correction modelling follow.
- 2004Finn Kydland and Edward Prescott
Time inconsistency: optimal policy announced today becomes sub-optimal tomorrow once private agents have reacted. Real Business Cycle theory: business cycles can arise purely from technology shocks under flexible prices.
- 2005Robert Aumann and Thomas Schelling
Aumann: repeated games support cooperation via the Folk Theorem; common knowledge is more subtle than it seems. Schelling: focal points solve coordination problems; tipping models explain segregation; deterrence requires credible commitment.