How geographic characteristics affect farming practices Workshop on An African Green Revolution Tokyo December 7-8, 2008
Overview Motivation for the research Theory behind the approach Methodology Comparison with an alternative approach Results from a proof (disproof?) of concept exercise
Why the Asian Green Revolution is not relevant for Africa The geographies of Asia and Africa are different Soils Climate Water and hydrology Natural transport systems Proximity of production centers to consumption centers
Consequences Relative prices between output and purchased inputs are lower Transportation costs are higher Farm-gate price for outputs lower Farm-gate price for fertilizer higher Technologies that were successful in Asia require a different geography So need to find a different set of technologies or invent new ones
Endogenous growth theory Farmers assess their own situation Decision environment Act rationally and choose farming methods that suit their needs Technology In practice we observe multiple ways of producing because there is great heterogeneity in farms and farmers
Getting to a production function From a mathematical perspective the problem is to solve for the control variables, conditional on initial conditions and a set of state variables The solution to control variables is always the same for a given set of state variables; they optimize out. Endogenous growth theory, if you have good measures of the state variables, then they define the technology choice If not then observed input choices are a combination of technologies
Geography of crop revenue in Ecuador
Annual precipitation
Fertilizer use
So how can we measure this? Econometric approach Problem finding comparable detailed data to make cross country comparisons Value of REPEAT and the data developed by the World Bank, Yale and China
Spatial-matching approach Place a grid over map of yield for two countries and break the map into thousands of cells Asia and Africa Do the same for maps with geographic information Domino analogy Yield on one side, geography description on other Put the Africa bits in a bag and leave Asia bits out on a table Draw an Africa bit out, find best match best on geography. Flip the pieces over and take the difference in yield and write it down. Put the Asia piece back and start over Calculate a mean difference and standard deviation If the difference is not significant, then geography tells the whole story
Propensity matching Use probit or logit to estimated predicted probability of being in the “treated” group Maize area in Ethiopia Equivalent to a weighted index of treatment group characteristics Other approaches use alternative weighting methods Find closes match outside of “treated” group Maize area NOT in Ethiopia Perfect match is predicted probability based on probit parameters and “non-treated” geography variables Calculated difference yields and standard error Test of difference in yields from moving a maize farm in Ethiopia to best match in Uganda, Tanzania or Kenya
Spatial map of maize yields for Ethiopia, Kenya, Uganda and Tanzania
Temperature map with boundaries
Population density map
Probit results for Ethiopia
Results suggest that yields on farms with similar geography outside of Ethiopia are higher than inside Ethiopia
If this works Some notion about policies that deal directly with geographic hurdles Infrastructure, new technologies And policies that target markets and households Extension, household support programs and market support programs Country by country outcomes likely to differ and regions within countries might differ Side product, a matching of areas for more detailed comparative studies