Presentation on theme: "(joint with Juan Pablo Rud, Royal Holloway)"— Presentation transcript:
1 (joint with Juan Pablo Rud, Royal Holloway) Modern Industries, Pollution and Agricultural Productivity: Evidence from Mining in GhanaFernando Aragon (SFU)(joint with Juan Pablo Rud, Royal Holloway)CEA ConferenceMay 2013
2 Main issue Negative spillovers of modern industries What is the effect of modern industries on agricultural production?In this paper:Case of gold mining in GhanaModern, capital-intensive industrySevere concerns of pollutionNear fertile rural area (cacao)This paper explores empirically the effect of modern mining on agricultural productivity.
3 Why is this relevant? Effect of pollution on agriculture not explored Literature focuses on effects on human healthBiological evidence that pollution affects cropsSpillovers of modern industriesThought in terms of input competitionOther non-input negative spillovers (e.g. pollution) neglectedEconomic policyPrivate and social costsCompensation and mitigationAnd through that channle may affect livelihoods as well….
4 What do we do? Main idea: pollution may affect crop yields Non-input channels residual productivity
5 What do we do? Estimate an agricultural production function Effect of mine activities on total factor productivityEmpirical strategyRepeated cross sections of HH surveysD-i-D: expansion of mining x exposure to mines (distance)Endog. inputs: IV and imperfect IV (partial identification)
6 Main findings Reduction of agricultural productivity 40 % decrease between , near minesBut, no change in input use nor prices.Results consistent with pollution channelSatellite imagery: increase in air pollutants (NO2)Increase in rural poverty
7 Unsolved issues Cannot measure pollution directly (not available) Effect on residual productivity does pollution affect quality of inputs (land, water) or crops’ health?Large scale and artisanal mines overlap cannot separate source
9 Background – Gold mining in Ghana Important industry in Ghana97% of mineral revenue, 45% of total exports, 12% of fiscal revenue.Modern, large scale, capital intensive96% large scale, rest artisanal/galamseyMostly foreign owned, exports all production as raw material.
10 Background – Gold mining in Ghana Located in fertile agricultural landAshanti gold belt: Western, Ashanti and Central regions.Cocoa producing regionsNegative spilloversPopulation displacementEnvironmental pollution: anecdotal and scientific evidenceSignificant increase in late 1990sWe exploit this source of variation.
12 Why would mining affect agriculture? Input competition channelDemand-Supply Increase in price of local inputs
13 Why would mining affect agriculture? Input competition channelDemand-Supply Increase in price of local inputsMining has potential to pollute: air, water and soilIndustry-specific pollutants cyanide, acid drainages, heavy metalsSimilar to small city or power plant emissions from heavy machinery (air pollutants)Biological evidenceExposure to air pollutants from burning fossil fuels reduction in yields 30-60%, more susceptible to diseases.Heavy metals in water and soil vegetation stunted or dwarfed (Environment Canada 2009)
14 Analytical framework Production function: F(A, Labor, Land) Consumer-producer household choose inputs to maximize HH utility.We will discuss validity later…
15 Analytical framework With perfect input markets: input demand is function of: prices and A endogeneity of inputs problemIf farmers cannot buy/sell inputsInput demand constrained by HH endowmentsExtreme case: Input demand = input endowmentUse endowments as instruments for input use.We will discuss validity later…
16 Analytical frameworkTwo possible channels for mining to affect agricultural output (and HH income)Change on input prices change in input usePollution change on AWe can isolate effect on A, by estimating the production function i.e. conditioning output on input use
17 Methods – empirical implementation Assume Cobb – Douglas,y, m, l : log of agricultural output, land and laborA is function of:Svt = measure of exposure to mine activityFarmer characteristics Zi: age and literacy, land ownership, place of birthDistrict, year fixed effects , dummy prox. each mine
18 Methods - Data Household data Ghana Living Standard Surveys (GLSS): GLSS 4 ( ) and GLSS5 ( ), repeated cross sectionsInput and output (farming households)Real output calculated using local agric prices.Poverty (all HHs)Geographical coordinates of Enumeration AreasDistance to mining areas (GIS)
19 Methods – empirical implementation Two issues:Endogeneity of mining activity: mining areas may be systematically different.Endogeneity of input choice
20 Methods - solutions Difference in difference: Treated and control group defined by proximity to mine“mining” area = within 20 km of an active mineTreatment (continuous) : cumulative gold productionSvt = cumulative gold production within 20 km
23 Results – Mining and Agricultural Productivity Go to: Crop yieldsGo to: First Stage
24 Methods - solution Use of instrumental variables: Endowments as instruments of input use (with imperfect input markets)But input endowments may be correlated to error termUse an imperfect IV strategy (Nevo and Rosen, 2012)If correlation between instrument and error is weaker than for the instrumented variable andThe sign of that correlation is the same Bounds of parameter values, i.e., partial identification
25 Results – Mining and Agricultural Productivity Go to: Crop yieldsGo to: First Stage
26 Results – Mining and Agricultural Productivity Go to: Crop yieldsGo to: First Stage
27 Increase of one S.D in gold production reduction of 30% in residual productivity Between 1998/99 and 2005 40% decrease.Too large? Consistent with biological evidence: 30-60% decrease in yields of crops exposed to polluted urban air.
29 Robustness checks No evidence of compositional change Farmer’s observablesAgricultural practicesRobust to alternative specificationsParsimonious vs saturatedSimilar for locals and immigrantsPlacebo test (future mines)CES production fct
30 Is this pollution?No ground measures of pollution satellite imagery (cross section only, 2005)Detect NO2 air pollutant linked to fuel combustion , toxic & precursor of tropospheric ozone
37 Increase in rural poverty (both farmers and non-farmers) But nothing on urban poverty
38 Final remarks Expansion of mining associated to Significant reduction in agricultural productivityDeterioration of living standards for rural populationSeems to be driven by pollution instead of competition for inputsImportant crowding out effect of modern industriesSignificant spillovers and re-distributive effectsLocal farmers lose, rest of country may gainDisregard for these spillovers over-estimate net benefits of sector