Presentation on theme: "Solomon Asfaw (Co-authors: Nancy McCarthy, Leslie Lipper, Aslihan Arslan and Andrea Cattaneo) Food and Agricultural Organization (FAO) Agricultural Development."— Presentation transcript:
Solomon Asfaw (Co-authors: Nancy McCarthy, Leslie Lipper, Aslihan Arslan and Andrea Cattaneo) Food and Agricultural Organization (FAO) Agricultural Development Economics Division (ESA) Rome, Italy ICABR Conference June 18-21, Ravello, Italy Climate Variability, Adaptation Strategy and Food Security in Malawi
Background Research questions Why do we do this? Methodology and Data What we find so far (results)? Conclusions Outline
► Malawi is ranked as one of the twelve most vulnerable countries to the adverse effects of climate change - subsistence farmers are most vulnerable to climate related stressors Background ► Adaptation in the agricultural sector to climate change is imperative – requiring modification of farmer behaviour and practices
► At micro (farmer) level, potential adaptation measures include a wide range of activities; most appropriate will be context specific and considered climate-smart agriculture (CSA) option. ► In Malawi, measures with high priority in national agricultural plans and high CSA potential include: o maize-legume intercropping o soil and water conservation (SWC), o tree planting, o conservation agriculture o organic fertilizer o Improved varieties and inorganic fertilizers ► Despite increasing policy prioritization and committed resources, adoption rates are quite low and knowledge gaps exist as to the reasons for this limited adoption Background
Research Questions ► What are the binding constraints of adoption of potential adaptation/risk mitigation measures? ► To what degree is there interdependence between adoption of different practices at plot level? ► What is the effect of adoption on maize productivity? ► What is the distributional impact, particularly where households are heterogeneous on key dimensions such as land holding, gender and geographical location?
Why do we do this? ► Limited research on adoption of multiple practices and little understanding of complementarities and substitution across alternative options; yet these are likely to be increasingly important under climate change. ► The effect of bio-physical and climatic factors in governing farmers’ adaptation decisions & how they are moderated through local institutions/govt interventions is poorly understood. Thus, we need analysis that incorporates: o Role of climate change - rainfall and temperature o Role of institutions o Government interventions o Bio-physical characteristics ► Limited understanding on the synergies and tradeoffs between CSA options and food security
Estimation strategy (1) ► We use multiple maize plot observations to jointly analyze the factors that govern the likelihood of adoption of adaptation measures in Malawi ► A Multivariate Probit (MVP) model: o There exist household and field level inter-relationships between adoption decisions involving various adaptation measures o The choice of technologies adopted more recently by farmers may be partly depend on earlier technology choices --- path dependence o Farm households face technology decision alternatives that may be adopted simultaneously and/or sequentially as complements, substitutes, or supplements ► Unlike the univarite probit model, MVP captures this inter- relationship and path dependence of adoption ► Assumes that the unobserved heterogeneity that affects the adoption of one of the practices may also affect the choice of other practices ► Error terms from binary adoption decisions can be correlated
Data ► World Bank Living Standard Measurement Survey (LSMS-IHS) in 2010/ ,288 households and about 64% maize producers ► Household level questionnaire and community level survey – location are recorded with GPS – link to GIS databases ► Historical rainfall and temperature estimates (NOAA-CPC) ( ) ► Soil Nutrient Availability (Harmonized World Soil Database) ► Malawi 2009 election results at EA level ► Institutional surveys at district level - supply side constraints o Credit; extension and other information sources; agricultural input and output markets; public safety nets and micro-insurance programs; property rights; and donor/NGO programs and projects.
Variables North province (N=1897) Central province (N= 3697) Southern province (N=5614) Total (N=11208) Long term inputs Maize-legume intercropping (1=yes) Planting tree (1=yes) Organic fertilizer (1=yes) SWC measures (1=yes) Short term inputs Improved maize seed (1=yes) Inorganic fertilizer (1=yes) All five None Adaptation measures – in proportion Descriptive statistics (1) NB: No data available on conservation agriculture
VariablesMeanStd. Dev. Household demographics and wealth Age of household head (years) Gender of household head (1=male) Household head highest level of education (years) Livestock ownership (tropical livestock unit (TLU)) Wealth index Agricultural machinery index Plot level characteristics Land tenure (1= own, 0= rented) Nutrient availability constraints (1-5 scale) Land size (acre) Slop of the plot (0=flat, 1=steep) Climatic ariables Coefficient of variation of precipitation ( ) Precipitation in the rainy season (mm) Annual mean temperature (deg C) Drought is a top three shock in the past year (1=yes) Institutions and transaction cost indicators Fertilizers distributed in MT by district per household Distance to major district centre (Km) Seed or fertilizer vender available in the community (1=yes) Village development committees in the community (number) Percentage of plots received extension advice at EA level Collective action index DPP vote as a share of total vote cast Descriptive statistics (2) Some explanatory variables
Barrier to adoption - Multivariate Probit model Improved Seed Inorganic fertilizer Organic fertilizer Legume intercrop Tree plantingSWC Coefficient of variation of precipitation ( )(+++)(---)(++)(--) Precipitation in the 08/09 season (mm)(+++) (---) Annual mean temperature in 08/09 year(deg C)(---) Drought is a top three shock in survey year (yes=1)(---)(--)(+++) Plot size (acre)(---)(+++) Land tenure (1= own, 0= rented)(---) (+++)(++)(+++) Slop of the plot (1=steep/hilly)(---)(+++) Nutrient availability constraint (1-4 scale, 5= non-soil)(+++)(---) Wealth index(+++) (---) Agricultural machinery index(+++) (---)(+++) Livestock in TLU(+++)(---)(+++) Seed and/or fertilizer vendor in EA (1=yes)(+)(++)(---) (--) Percentage of plots received extension advice at EA level(+++)(---) (+++) Distance to major centre (km)(---) Number of village development committees (+++)(++)(+++)(++)(+++) Collective action index(+++) (++)(+) DPP votes as a hare of total votes case(+++)(---)(--)(+++) Price of maize (MKW/kg)(+++) (-)(++) Fertilizer distributed in MT by district per hh(+++) (---)(--) Proportion of land covered by forest by district(+++) Microfinance & donor agri projects operating in district(+++)(+)(---) (+++) MASAF wages paid out in district in 08/09 season(---)(+++) (---)(+++) Northern Province (Ref: Southern province)(---) (-)(+++)(---) Central province(+)(---) (+)
► Adopting a specific practice is conditioned by whether another practice has been adopted or not –interdependency between adoption decision - complimentarity or substitutability ► Climate risk: Favorable rainfall increases probability of adopting practices with short-term return; unfavorable rainfall increases likelihood of adopting measures with longer term benefits. ► Land tenure: increases the likelihood to adopt strategies that will capture the returns in the long run and reduces the demand for short-term inputs. ► Social capital and supply side constraints: Collective action and informal institutions matter in governing farmers adoption decisions to adopt ► Plot characteristics and household wealth: are important determinants of adoption of adaptation measures Summary of Findings: Adoption
Variables North province (N=1897) Central province (N= 3697) Southern province (N=5614) Total (N=11208) Maize-legume intercrop No Yes Difference (%) 93.7(12.5)***32.4(4.3)***39.8(8.1)***62.7(16.7)*** Tree planting No Yes Difference (%) -9.4(2.2)**5.7(1.2)1.3(0.3)8.7(2.9)*** SWC measures No Yes Difference (%) 5.3(1.1)-13.9(3.8)***25.6(5.5)***10.1(3.4)*** Improved seed No Yes Difference (%) 3.8(0.8)39.1(8.3)***38.6(7.9)***28.6(9.0)*** Inorganic fertilizer No Yes Difference (%) 30.8(5.0)***42.3(6.7)***134.7(15.0)***80.5(15.9)*** Maize productivity by adoption status (kg/acre) Note: Number of observations refers to the number of maize plots. *** p<0.01, ** p<0.05, * p<0.1. t-stat in parenthesis. Impact of adoption on maize yield
Identification strategy (2) ► Random assignment of treatment and control not possible ► No panel data available o Difference-in-Difference (DD) estimator Address time invariant unobservables ► Cross-sectional data o PSM combined with inverse propensity weights (IPW) – Address only observable bias o Instrumental variable (IV) strategy Address observable and unobservable bias
OLSInstrumental Variable (IV) strategy Seed (1)Fertilizer (2) Legume (3)Trees (4)SWC (5) Improved maize seed (1=yes) 0.135***0.611*** Log of inorganic fertilizer (kg/ha) 0.365***1.596* Maize-legume intercropping (1=yes) 0.720***2.128* Perennial trees (1=yes) 0.282***0.917 Soil and water conservation (1=yes) Precipitation in the last rainy season (mm)(+++)(++)(+)(+++) Annual mean temperature (deg C)(---) (--)(---) Drought is a top three shock in survey year (yes=1)(---) (--_)(---)(--) Plot size (acre)(---)(--)(---) Slop of the plot (1=steep/hilly)(--)(-) Nutrient availability constraint (1-4 scale, 5= non-soil)(---) (--) Wealth index(+++) Education of the head (years)(+)(++)(+) Education of the spouse (1=yes)(+++) Age of the head (years)(--)(---)(--) Gender of the head (1=male)(--)(-) Excluded instruments Coefficient of variation of precipitation ( )XXXXX Seed and/or fertilizer vendor in EA (1=yes)XX Percentage of plots received extension advice at EA levelXX Fertilizer distributed in MT by district per householdX Proportion of land covered by forest by districtX District agriculture extension officer per household X Weak identification test (Wald F-stat)43.34***21.55***11.92***19.80***24.92*** Over identification test (Henson J- stat) Impact of adoption on maize yield (log kg/acre) – IV estimator
Improved seed Inorganic fertilizer Legume intercropping Tree planting SWC measures Province North * Central ***-5.23** South ***0.63** Gender of head Male ***2.29* Female2.09** 0.55*** **1.05 Median land size Small 1.24* ** Large2.27*1.27*** Heterogeneous impact of adoption (ATT) – IV estimator Note: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered at EA.
► On average adoption of three of the five farm management practices (short term) have a positive and statistically significant impact on maize yield. ► Average precipitation is positively correlated with maize yield whereas drought and high temperature are negatively correlated ► Plot characteristics, household wealth and human capital are positively correlated with maize productivity ► Heterogeneous impact in key dimensions such as land holding, gender and geographical location Summary of Findings: Impact
► Place matters (and CC makes it even more important)! Plot characteristics, agro-ecology, local institutions and climate regime key factors affecting adoption of practices with adaptation potential ► Given importance of adopting a package of practices for adaptation (e.g. SLM); need to get better understanding of complementarities/substitution- this method is one approach ► Given importance of climate on adoption of practices with short (seeds/fertilizer) vs. long (trees, SWC, legume) term returns; need to improve access to reliable climate forecast information is key to facilitating adaptation - farmers to new sources of information on climate variability will be important; ► Heterogeneity in yield benefits from adoption of different practices across farm size, gender and agro-ecology – suggests possible heterogeneity in synergies/tradeoffs between food security/adaptation. ► Not surprising that fertilizer/seeds gives maize yield effect, but we need to know more about implications for yield variance. We have not estimated the impact on reducing yield variability in the face of variable climate conditions Conclusions and Implications