layout: true --- name: cover class: left, middle background-image: url("data:image/png;base64,#../source-material/img/avocado.png") background-size: cover .pull-left[ ## .white[Ripe for contracts? Avocado contract farming in Kenya] ### .white[Philipp Kollenda] #### .white[May 19, 2022 (last updated: May 18 2022)] .white[Spring Meeting of Young Economist 2022] ] .right[ <img src="data:image/png;base64,#../source-material/img/authors/daniella.jpeg" width="82x"/> <img src="data:image/png;base64,#../source-material/img/authors/jane.jpeg" width="82x"/> <img src="data:image/png;base64,#../source-material/img/authors/philipp.jpeg" width="82x"/> <img src="data:image/png;base64,#../source-material/img/authors/remco.jpeg" width="82x"/> <img src="data:image/png;base64,#../source-material/img/authors/menno.jpeg" width="82x"/> <img src="data:image/png;base64,#../source-material/img/authors/fedes.jpeg" width="82x"/> ] --- name:fact_poverty class: middle .pull-left[ <img class="box" src="data:image/png;base64,#../source-material/img/factoids/Slide4.jpeg" width="1000px"/> ] .pull-right[ ## .white[Fill text] ## .yellow[Two third] .black[of the world population live on] .yellow[less than $10 a day.] ## .red[780 million people] .black[live in extreme poverty on] .red[less than $1.90 a day.] ] --- name:fact_food-security class: middle .pull-left[ <img class="box" src="data:image/png;base64,#../source-material/img/factoids/Slide5.jpeg" width="1000px"/> ] .left.pull-right[ ## .white[Fill text] ## .white[Fill text] ## .brown[26 percent] .black[of the world population was affected by] .brown[moderate or severe food insecurity] .black[in 2019.] ] --- name:fact_smallholder class: middle .pull-left[ <img class="box" src="data:image/png;base64,#../source-material/img/factoids/Slide3.jpeg" width="1000px"/> ] .left.pull-right[ ## .white[Fill text] ## .white[Fill text] ## .darkgreen[86 percent] .black[of agricultural producers are] .darkgreen[smallholder farmers.] ] --- name: motivation_contract class: center, middle, inverse # Can smallholder farmers access global agricultural value chains through contract farming? # Do they receive higher prices, earn more income and improve food security? --- name: preview ### Can smallholder farmers access global agricultural value chains through contract farming? **Research design / Intervention:** Study the adoption of contract farming induced by an external (non-random) intervention that established farmer organisations and matched them with avocado-exporters. Our identifying variation is that some farmers now live closer to a farmer organization - more likely to adopt. **Data:** Panel data (2015, 2017) from 702 Kenian households that own avocado trees and sell to brokers or exporters. Transaction-level data (for avocados) and information on household demographics, consumption, agricultural and non-agricultural assets, production and income. Four categories of outcomes: farmer's behavior (*output*) and production, marketing and welfare (*impact*) **Methodology:** (Non-randomized) Treatment is adoption of contract farming at endline versus no participation. Address selection into treatment with *doubly-robust difference-in-differences* and match based on (quasi-random) change in distance to nearest farmer organization. **Results:** Contract farming increases likelihood that farmers sell to exporters, are trained and certified. Significant and meaningful increase in prices and knowledge, evidence for higher income and *lower* food security. --- name: model ### Contract farming has many benefits over spot market transactions. .pull-left[ .large[Spot market, the status quo] <img src="data:image/png;base64,#../source-material/img/contract-farming-model/Slide6.jpeg" style="width: 400px"/> 👎 No coordination among farmers or with downstream value chain actors. 👎 Take-it or leave-it offers. No traceability, little attention to harvest timing or crop quality. 👎 Need stable relations for certification, training, input provision, no access to credit. 👍 Immediate payment, greater flexibility, lower transport cost, no membership fees. ] -- .pull-right[ .large[Group-based contract farming] <img src="data:image/png;base64,#../source-material/img/contract-farming-model/Slide7.jpeg" style="width: 400px"/> 👍 Contract farming affects welfare through prices, quantities and (opportunity) costs. `\(u_i = f(y_i^{avo}, y_i^{-avo}) \quad y_i = p * q_i - c_i\)` ] --- name: model_evidence ## Empirical studies positive about contract farming - 92% of studies estimate a positive effect on productivity, and - 75% estimate a positive effect on income (Wang, Wang, Delgado 2014; Bellemare, Bloom 2018). - Previous studies also find positive effects on food security, subjective well-being, household assets. ## but usually based on cross-sectional data and challenging identification. - Notable exception: Arouna, Michler, Lokossou (2021) find positive effects on welfare and productivity for rice production in Benin. 💪 ??? Aminou Arouna, Jeffrey Michler and Jourdain Lokossou --- name: overview_intervention class: left, middle, inverse # Outline .large[ 1. Motivation and big picture 2. Our model of contract farming 3. **.red[Intervention and data]** 4. Methodology 5. Results ] --- name: model_whathappened ## A program to strengthen exports from Kenya to EU .pull-left[ 1. .primary[Mid 2015:] Agricultural officers select ~~10~~ 4 villages and establish farmer groups. 2. .primary[End 2015:] Each group matches with one avocado exporter and 113 farmers who join groups included in our sample. 3. .primary[End 2015:] We sample [two comparison groups](#farmer_map): 244 farmers with existing contracts and 345 farmers in comparable (non-targeted) villages. 4. .primary[2015 - 2017:] Farmers [do what they do](#compliance). Groups (may) receive training and certification. 5. .primary[Mid 2017:] Endline data, including avocado transactions from 3 harvest seasons. ] .pull-right[ <img src="data:image/png;base64,#../source-material/img/contract-farming-model/Slide8.jpeg" style="width: 400px"/> ] --- name: data class: middle, center # Lots of data! Household head answered modules on: Household composition | Productive and non-productive assets | Agricultural production and marketing And specifically on: **Avocado transactions** | Farmer organization (membership and quality) | Training and certification .footnote.left[[List of control variables.](#selection)] --- name: overview_methodology class: left, middle, inverse # Outline .large[ 1. Motivation and big picture 2. Our model of contract farming 3. Intervention and data 4. **.red[Methodology]** 5. Results ] --- name: methodology ## Estimating average treatment effects on the treated via doubly-robust difference-in-differences We compare 124 farmers who adopt contract farming between the baseline and the endline with 292 farmers that sell avocado, but never under contract. - We thus exclude farmers that have a contract at baseline (*always-treated and disadopters*). - How does adoption and never-treated relate to the initial classifications? [Take a look!](#compliance) ####Farmers select into contract farming, but the intervention [makes it more likely for some farmers to join](#distance). -- Zhao, Sant'Anna 2020 propose an identification framework that combines [.blue[difference-in-differences]](#did) with [.red[inverse probability weighting]](#matching) to get doubly-robust<sup>1</sup> estimates. .center[ `\(\begin{equation}\widehat{ATT} = \mathbb{E}_n \big[\color{red}{(\hat{w}_1(D) - \hat{w}_0(D, X; \hat{\gamma}^{ipt}))}\color{blue}{\big(\Delta Y - X' \hat{\beta}^{wls}_{0,\Delta} \big)}\big]\end{equation}\)` ] .footnote[ 1: doubly-robust = consistent if one of the two models is correctly specified. ] --- name: tree_trends ### Parallel trends? #### No significant difference in the number of planted trees in the years before the treatment. <!-- .center[<img src="data:image/png;base64,#../../../Output/Figures/treatment_event_plot.png" style="width: 50%" />] --> .center[<img src="data:image/png;base64,#../../../Output/Figures/treatment_event_plot_contractonly.png" style="width: 50%" />] --- name: balance ### Balance on observables? #### Re-weighted sample is significantly more balanced. .center[ <img src="data:image/png;base64,#../../../Output/Figures/descriptivetable2_contract.png" style="width: 45%" /> ] --- name: overview_results class: left, middle, inverse # Outline .large[ 1. Motivation and big picture 2. Our model of contract farming 3. Intervention and data 4. Methodology 5. **.red[Results]** ] --- name: recall_model_intervention class: left ## Recall our main questions .pull-left[ <img src="data:image/png;base64,#../source-material/img/contract-farming-model/Slide8.jpeg"/> ] .pull-right[ ### Part 1: #### Does the .primary[adoption of] contract farming lead to selling to exporters, training and certification? ### Part 2: ####How does the adoption of contract farming change farmers' .primary[production, marketing and welfare outcomes]? ] --- name: results1 class: inverse ### Part 1: YES! [(Table)](#results1_table) ### Contract farming significantly increases the likelihood that farmers sell to exporters, are recently trained and certified. <img src="data:image/png;base64,#index_files/figure-html/results_intervention-1.svg" width="100%" /> ??? The intervention works, in the sense that it **delivers the three immediate changes** to farmer behaviour and outcomes: selling to exporters, training and certfification. - Adopters are **50 percentage points** more likely to sell any hass avocado to a company. We excluded farmers with contracts at baseline, so at baseline almost no-one in this sample sells avocados to a company (2.5 percent). - Farmers don't sell everything to the company, but go from basically 0 to selling more or less half of their avocado to the exporters. - The effect is largest for Hass avocado, which is the variety that exporters mostly demand. - They receive the GAP certification and training, although if we take this as *compliance*, the fact that 50 percent of the farmers are trained between baseline and endline is not incredibly high. - At baseline, training rate is 18.6 percent in treatment group, so it almost triples. - No significant shift in the type of avocado they sell. But we would not expect that (yet). --- name: results2 class: inverse ### Part 2: yes ### Contract farming improves prices and knowledge on agricultural practices. Some evidence for more trees, shift to hired labor, higher income(s) and increased food insecurity.
??? Next, we look at the actual production, marketing and welfare outcomes. - For **production choices** we find that contract farmers - grow more hass trees (up 214%), - shift to hired labor and - know more about good agricultural practices (up 13%). - For **marketing outcomes** we find that contract farmers - receive 1.18 KSh higher prices (up 37%). Recall that they do not sell all avocados to the companies. The results fits our qualitative knowledge that prices by companies are about 3 KSh higher. - higher income, no change in quality or quantity. - For **welfare outcomes** we find - no significant effects, but higher income overall (69%!) and increased food insecurity are notable (and in the case of food insecurity worrying). --- name: conclusion ## In conclusion: #### Our results mainly confirm earlier empirical results on benefits of contract farming, but, arguably, with better identification through panel data. - The intervention delivered on the three main outputs: selling to exporters, training and certification. - Contract farmers get better prices, know more about good agricultural practices and plant more hass avocado trees. -- - [Sales of the local variety decrease](#results_byavo) with no increase in sales of hass variety (yet): important to support farmers through the transition to higher yielding varieties. - Using panel data matters. <!-- - Using panel data [matters](#results_crosssection). --> - Today: contract farming. But, [which part of contract farming matters?](#results_activities) --- name: end class: inverse, middle, center # Thank you very much! ![Avo](data:image/png;base64,#https://media.giphy.com/media/Efm0VUV1Llm07BmltF/source.gif) ### Happy to receive your comments at p.kollenda@vu.nl --- name: end2 class: inverse, middle, center # Thank you very much! <img src="data:image/png;base64,#../source-material/img/avocado-gif.png"/> ### Happy to receive your comments at p.kollenda@vu.nl <!-- From now on the slides should not be played when we go through it, only if linked to on request. --> --- name: farmer_map class: inverse #The intervention (2015 - 2017) <img src="data:image/png;base64,#index_files/figure-html/farmermap-1.svg" width="100%" /> .footnote[ [Back to main text](#model_whathappened) ] --- name: compliance ### Some targeted farmers do not adopt contract farming while some untargeted farmers do. .center[ <img src="data:image/png;base64,#../../../Output/Figures/contract_farming_arms.png" style="width: 40%" /> ] .footnote[ [Back to study design](#model_whathappened) or [back to methodology](#methodology) ] --- name: selection ### Farmers who adopted contracts own more trees, are older, more educated and have higher income at baseline. .pull-left[ #### The outcome and selection model control for the baseline values of: - Household composition - Assets and Land Ownership - Agricultural Production - Access to Banking and Finance ] .pull-right[ <img src="data:image/png;base64,#../../../Output/Figures/descriptivetable2_contract.png" style="width: 90%" /> ] .footnote[ [Back to main text](#data) ] --- name: distance class: top, left .pull-left[ ### Farmers close to the four newly established farmer organizations are significantly more likely to adopt contract farming. <table class="texreg" style="margin: 10px auto;border-collapse: collapse;border-spacing: 0px;caption-side: bottom;color: #000000;border-top: 2px solid #000000;"> <caption>Statistical models</caption> <thead> <tr> <th style="padding-left: 5px;padding-right: 5px;"> </th> <th style="padding-left: 5px;padding-right: 5px;">Large sample</th> <th style="padding-left: 5px;padding-right: 5px;">Small sample</th> </tr> </thead> <tbody> <tr style="border-top: 1px solid #000000;"> <td style="padding-left: 5px;padding-right: 5px;">Intercept</td> <td style="padding-left: 5px;padding-right: 5px;">-2.31<sup>*</sup></td> <td style="padding-left: 5px;padding-right: 5px;">-1.72</td> </tr> <tr> <td style="padding-left: 5px;padding-right: 5px;"> </td> <td style="padding-left: 5px;padding-right: 5px;">(0.97)</td> <td style="padding-left: 5px;padding-right: 5px;">(1.16)</td> </tr> <tr> <td style="padding-left: 5px;padding-right: 5px;">Distance Reduction</td> <td style="padding-left: 5px;padding-right: 5px;">-0.48<sup>**</sup></td> <td style="padding-left: 5px;padding-right: 5px;">-0.38<sup>*</sup></td> </tr> <tr> <td style="padding-left: 5px;padding-right: 5px;"> </td> <td style="padding-left: 5px;padding-right: 5px;">(0.16)</td> <td style="padding-left: 5px;padding-right: 5px;">(0.18)</td> </tr> <tr> <td style="padding-left: 5px;padding-right: 5px;">Distance Dummy</td> <td style="padding-left: 5px;padding-right: 5px;">1.52<sup>***</sup></td> <td style="padding-left: 5px;padding-right: 5px;">1.20<sup>**</sup></td> </tr> <tr> <td style="padding-left: 5px;padding-right: 5px;"> </td> <td style="padding-left: 5px;padding-right: 5px;">(0.31)</td> <td style="padding-left: 5px;padding-right: 5px;">(0.37)</td> </tr> <tr style="border-top: 1px solid #000000;"> <td style="padding-left: 5px;padding-right: 5px;">Covariates</td> <td style="padding-left: 5px;padding-right: 5px;">Yes</td> <td style="padding-left: 5px;padding-right: 5px;">Yes</td> </tr> <tr style="border-bottom: 2px solid #000000;"> <td style="padding-left: 5px;padding-right: 5px;">Num. obs.</td> <td style="padding-left: 5px;padding-right: 5px;">416</td> <td style="padding-left: 5px;padding-right: 5px;">259</td> </tr> </tbody> <tfoot> <tr> <td style="font-size: 0.8em;" colspan="3"><sup>***</sup>p < 0.001; <sup>**</sup>p < 0.01; <sup>*</sup>p < 0.05</td> </tr> </tfoot> </table> ] -- .pull-right[ ### And predicted propensity scores for contract farming adoption have sufficient overlap. <img src="data:image/png;base64,#index_files/figure-html/distance_plot-1.png" width="100%" /> .footnote[ [Back to methodology](#methodology) ] ] --- name: did class:middle, left ### .center[The trend in the control group is the counterfactual for the treatment group.] .pull-left[ <img src="data:image/png;base64,#../source-material/img/identification/did.png" style="width: 100%" /> ] .pull-right[ .large[Identifying assumptions: ] - Parallel trends - Conditional on observables (Heckman, Ichimura, Todd 1997: Outcome regression) .footnote[ [Back to methodology](#methodology) ] ] --- name: matching class:middle, left ### .center[Households in the control group who - based on observables - are as likely to be treated, are the counterfactual for the treatment group.] .pull-left[ <img src="data:image/png;base64,#../source-material/img/identification/matching.png" style="width: 100%" /> ] .pull-right[ .large[Identifying assumptions: ] - Treatment assignment is random conditional on the propensity score. .footnote[ [Back to methodology](#methodology) ] ] --- name: results1_table class: inverse ### Part 1: YES! ### Contract farming significantly increases the likelihood that farmers sell to exporters, are recently trained and certified.
.footnote[[Back to results](#results1)] --- name: results_byavo class: inverse ### Benefits of contract farming are driven by hass variety, farmers shift away from local varieties.
.footnote[[Back to conclusion](#conclusion)] <!-- .footnote[[Back to conclusion](#conclusion) or view as [table](#results_byavo_table)] --> --- name: results_activities ### Which part of contract farming matters? -- Farmers that adopt selling to exporters: - get 2.60 KSh more per Hass unit. - double their income from Hass avocado sales. - But decrease total income 🤔 -- Farmers that adopt certification or training: - plant 2.65 more hass trees (sign. for certification) - have higher knowledge scores (around 10%-15%) - and slightly higher prices (0.90-1.17KSh). .footnote[[Back to conclusion](#conclusion)] <!-- .footnote[[To Table](#table_results_activitites) or [back to conclusion](#conclusion)] -->