Using whole-farm models for policy analysis of Climate Smart Agriculture

Using whole-farm models for policy analysis of Climate Smart Agriculture

• 1.
  Using whole-farm models forpolicy analysis of Climate Smart Agriculture A. Paolantonio1, G. Branca12, R. Cavatassi1, A. Arslan1, L. Lipper1, O. Cacho3 (1FAO, 2Tuscia University, 3University of New England) Montpellier March 16-18, 2015
• 2.
  Outline • Background • Modeloverview & methodology • Malawi case study & data • Results • Conclusions & future model development
• 3.
  Background • The FAO-EPICprogram aims at building evidence-based agricultural development strategies, policies and investment frameworks to achieve the objectives of CSA in Malawi, Zambia and Viet Nam • Why? To create a strong link between research, policy, and investments • How? By providing solid and scientific evidence combining qualitative with quantitative analysis using primary and secondary data at HH and community level + climate and agro-ecological data + institutional data
• 4.
  A model forCSA policy analysis • Econometric models based on HH data are essential tools for policy analysis (but ex-post only) • Mathematical programming (MP) models of farm HHs allow ex-ante analyses to be conducted as well • The key is to calibrate MP optimization models to be consistent with the evidence base (and thus believable)  Positive Mathematical Programming (PMP) [Howitt, 1995] • PMP was developed for a policy analysis that utilizes all the available information, no matter how scarce [especially suitable in agricultural economics]
• 5.
  PMP methodology 1 1.Max 𝜋 = 𝑦𝑝 ′ 𝑥 − 𝑐′ 𝑥 s.t. 𝐴𝑥 ≤ 𝑏 obj. function (LP model) resource constr. 𝑥 ≤ 𝑥 𝑜𝑏𝑠 calibration constr. 𝑥 ≥ 0 2. Use the shadow prices of the calibrating constraint (𝜆 𝐿𝑃) to estimate the implicit cost parameters that calibrate the model to the survey data: 𝑄𝑗𝑗 = (𝜆 𝐿𝑃𝑗+𝑐𝑗) 𝑥 𝑜𝑏𝑠𝑗 3. Max 𝜋 = 𝑦𝑝 ′ 𝑥 − 𝑥′ 𝑄 𝑥/2 s.t. 𝐴𝑥 ≤ 𝑏 𝑥 ≥ 0 obj. function (QP model) resource constr.
• 6.
  PMP methodology 2 •Sensitivity analysis implies parametric change in: - output prices; or - technological coefficients (technical relationships between inputs and outputs); or - resource availability (constraints) that will produce a response on the model’s new solution • Basically, it determines which resource constraint has the most potential impact given the optimal solution • It helps identifying relevant areas of policy intervention based on the observed situation
• 7.
  The MP matrixmodel Technical coefficients Activities Constraints Crops Livestock Off- farm labor… … Max/Min C1 … Cn L1 … Ln X1 Land ac11 … ac1n al11 … al1n …  b1 Labour ac21 … ac2n al21 … al2n …  b2 Capital … … … … … … …  b3 Fertilizer … … … … … … …  b4 Water … … … … … … …  b5 … acm1 … acmn alm1 … almn …  b6 Obj. function
• 8.
  PMP applied tothe case of Malawi • We develop a whole-farm model using PMP with ad hoc collected plot level data on CSA in MW • So the model: - is based on economic theory (optimizing behaviour) - …but has the beauty of utilizing objective data, and therefore - a great potential to provide policy insights through simulations based on observed outcomes
• 9.
  Malawi case study& data • CSA survey carried out in 2013 by FAO-EPIC in collaboration with country FAO office • HH sample and CSA practices selection on the basis of agriculture screening and field visits • Final statistical sample made of 524 HHs cultivating 1,433 fields over 11 Extension Planning Areas (EPA) located in 4 districts (Mzimba, Kasungu, Balaka, Ntcheu) across 4 AEZ • Reference cropping season is 2012-13 • Main evidence found suggests: - Low diffusion of SLM for all crops: 84% tillage systems (conventional), only 16% MSD systems [mainly maize = 61% tillage vs 39% MSD] - No significant difference by AEZ and district/EPA - High heterogeneity of SLM technology packages
• 10.
  Results from theBase Case 1/2 0 1,000 2,000 3,000 Tobacco tillage S-beans tillage G-nuts tillage Maize MSD Maize tillage Yield (kg/ha) 0 200 400 600 Tobacco tillage S-beans tillage G-nuts tillage Maize MSD Maize tillage Capital required ($/ha) 0 50 100 150 200 250 Tobacco tillage S-beans tillage G-nuts tillage Maize MSD Maize tillage Labour required (pd/ha) 0 100 200 300 400 500 Tobacco tillage S-beans tillage G-nuts tillage Maize MSD Maize tillage Fertilizer required (kg/ha)
• 11.
  Results from theBase Case 2/2 0 50 100 150 200 250 300 Tobacco tillage S-beans tillage G-nuts tillage Maize MSD Maize tillage Area planted (ha) How can we increase the adoption of this system?
• 12.
  Sensitivity Analysis • Labourconstraint has almost no effect on crop choice but it significantly matters in the decision to adopt MSD vs tillage 0.40 0.45 0.50 0.55 0.60 0.65 0.70 50 60 70 80 90 100 Maizearea/totcroparea Resource availability (as % of optimal solution) Labour Capital 0.00 0.05 0.10 0.15 0.20 0.25 50 60 70 80 90 100 MSDmaizearea/totmaize area Resource availability (as % of optimal solution) • Capital constraint has strong effect on crop choice with a small change on the proportion that is MSD
• 13.
  Conclusions • PMP modelshave great potential in providing evidence- based insights for CSA policy recommendations • Maize under MSD systems show higher yields, but also higher capital and labour requirements compared to tillage systems in Malawi • Mainly labour constraints the adoption of MSD systems in Malawi, whereas the effects of changes in the availability of capital are limited • Interventions should be primarily targeted to address the labour constraint
• 14.
  Future model development •More simulations on different model parameters • Exploit full sample information: calibrate the model for individual HHs (but need a correct statistical approach) • Multi-period modelling • Extend the analysis to Zambia for cross-country comparison • Add livestock component [Zambia]
• 15.
  Thank you! If interestedin FAO-EPIC CSA evidence-base: www.fao.org/climatechange/epic