The Crop Connection: Impact of Cell Phone Access on Crop Choice in Rural Pakistan Saher Asad Lahore University of Management Sciences saher.asad@lums.edu.pk June 20, 2016 ABCDE 2016 Saher Asad (LUMS) Crop Connection June 20, 2016 1 / 47
Motivation Cash crops can have positive impacts on development (World Bank, 2008; IFPRI,2013; FAO,2014) Cash crops are often highly perishable Profitability In developing countries - storage is limited, market linkages are weak Risk of high post-harvest losses deters production Can the cell phones mitigate this risk? Saher Asad (LUMS) Crop Connection June 20, 2016 2 / 47
This Paper Provides estimates for impact of cell phone access on crop choice along the perishability dimension Mechanisms Farmer-Trader Coordination (FTC) Price Information Effect (PIE) - (e.g. Jensen, 2007; Aker, 2010; Goyal, 2010; Muto and Yamano, 2009; Nakasone, 2013; Mitra et al., 2013) Weather Information Effect (WIE) - (e.g. Cole and Fernando, 2012; Fafchamps and Minten, 2012) Distinguish effects Income and Consumption Saher Asad (LUMS) Crop Connection June 20, 2016 3 / 47
Context: Pakistan Agriculture constitutes 21% of GDP in Pakistan Punjab(study region) is the hub of agriculture Farmers grow 4-6 crops in a year Farmers are constrained by storage and timely sale Traders are constrained by limited capacity and power outages Saher Asad (LUMS) Crop Connection June 20, 2016 4 / 47
Evidence from Field Before I had a cell phone I harvested my crop and then had to wait for a trader to buy my crops; now I talk to the trader and harvest my crops when he will buy it. - Farmer in Rural Pakistan I cannot buy an unlimited amount of crop in one day. The power keeps going out and I cannot store these crops for very long period. The retailers dont like to buy rotten crops. - Trader in a Mandi Farmers can adjust the watering schedule to adjust the harvest date up to 10 days. - Agricultural Expert Saher Asad (LUMS) Crop Connection June 20, 2016 5 / 47
Model Agriculture Household Model (De Janvry and Sadoulet, 2006; Dercon, 1996) highlights the role of coordination in farmer production choices Assumptions Baseline CellPhone Solving the model gives the following results Result 1: CellPhone ProbabilityofSale c SharetoPerishable c Result 2: CellPhone ProbabilityofSale nc SharetoPerishable nc Saher Asad (LUMS) Crop Connection June 20, 2016 6 / 47
Identification Strategy Cell Phone coverage is endogenous - heavily provided to rich areas Coverage Restricted from villages in 10 km of border with India - Dead Zone Enforced with: (a) Strategic Tower Placement; (b) Jammer Placement Does not coincide with the No Man s Land (500m) Spatial Regression Discontinuity India Legend International Border District Boundary Dead Zone Boundary Villages Without Cell Phone Coverage Villages With Cell Phone Coverage 1:1,250,000 Saher Asad (LUMS) Crop Connection June 20, 2016 7 / 47
Village Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 8 / 47
Village Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 9 / 47
Village Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 10 / 47
Village Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 11 / 47
Village Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 12 / 47
Household Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 13 / 47
Household Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 14 / 47
Household Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 15 / 47
Household Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 16 / 47
Household Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 17 / 47
Household Level Data Saher Asad (LUMS) Crop Connection June 20, 2016 18 / 47
Estimation Covariates are balanced Household Fuzzy RD for village data Village Sharp RD for household data Due to the sample size limitations I adopt a semi-parametric estimation method based on Dell(2010) Present results for quadratic, cubic and quartic polynomials in Euclidean distance and lat-long India Legend International Border District Boundary Dead Zone Boundary Villages Without Cell Phone Coverage Villages With Cell Phone Coverage 1:1,250,000 Saher Asad (LUMS) Crop Connection June 20, 2016 19 / 47
Discontinuity in Probability of Producing Perishable Crop Probability of Growing Extremely and Highly Perishable Crops 0.0 0.2 0.5 0.7 0.9 10 5 0 5 10 Distance from Restriction Density Smoothness Saher Asad (LUMS) Crop Connection June 20, 2016 20 / 47
Discontinuity in Area Allocated to Growing Perishable Crops Area Under Extremely and Highly Perishable Crops 0.0 0.3 0.5 0.8 1.0 5 0 5 Distance from Restriction Density Smoothness Saher Asad (LUMS) Crop Connection June 20, 2016 21 / 47
Is this a Major Crop in the Village? (Degree of Perishability of Crop) (Extremely & Highly) (Least) (Extremely & Highly) ( Least) (Probit) Cell Phone 0.31*** 0.09 0.29*** 0.12 (0.06) (0.08) (0.11) (0.16) (Quadratic in Distance from Restriction) Cell Phone 0.23*** 0.032 0.26*** 0.02 (0.03) (0.04) (0.03) (0.04) (Quadratic in Lat-Long) Cell Phone 0.24*** 0.012 0.25*** 0.09 (0.02) (0.09) (0.03) (0.07) Controls Y Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Y Number of Observations 1785 1785 1094 1094 Size of Band Around Restriction 10km 10km 5km 5km Mean of Control Group 0.15 0.89 0.18 0.92 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 22 / 47
Percentage of Area Under Crop (Degree of Perishability of Crop) (Extremely) (Highly) (Least) (OLS) Cell Phone 0.28*** 0.19*** -0.48*** (0.08) (0.06) (0.18) (Quadratic in Distance from Restriction) Cell Phone 0.245*** 0.16*** -0.411*** (0.11) (0.02) (0.09) (Quadratic in Lat-Long) Cell Phone 0.265*** 0.18*** -0.37*** (0.08) (0.03) (0.06) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 Mean of Control Group 0.11 0.13 0.76 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 23 / 47
Farmer Trader Coordination: Can the Farmer Sell Crops on First Trip to Market? (Degree of Perishability of Crop) (Extremely) (Highly) (Least) (Probit) Cell Phone 0.58*** 0.42*** 0.03 (0.16) (0.19) (0.07) (Quadratic in Distance from Restriction) Cell Phone 0.40*** 0.32*** 0.06 (0.07) (0.04) (0.19) (Quadratic in Lat-Long) Cell Phone 0.49*** 0.31*** 0.07 (0.02) (0.065) (0.17) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 Mean of Control Group 0.35 0.39 0.86 p < 0.01, p < 0.05, p < 0.1 Number of Days Saher Asad (LUMS) Crop Connection June 20, 2016 24 / 47
Farmer Trader Coordination: Percentage of Output Lost to Post-Harvest Losses? (Degree of Perishability of Crop) (Extremely) (Highly) (Least) (OLS) Cell Phone -0.37*** -0.21*** -0.11 (0.13) (0.09) (0.14)) (Quadratic in Distance from Restriction) Cell Phone -0.21*** -0.15*** 0.04 (0.09) (0.07) (0.11) (Quadratic in Lat-Long) Cell Phone -0.25*** -0.13*** 0.05 (0.03) (0.045) (0.15) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 Mean of Control Group 0.63 0.51 0.12 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 25 / 47
Weather Information Effect: Difference Between Actual and Optimal Plantation Date? (Degree of Perishability of Crop) (Extremely) (Highly) (Least) (OLS) Cell Phone -8.5*** -4.9*** 1.3 (1.85) (0.75) (1.14) (Quadratic in Distance from Restriction) Cell Phone -6.43*** -4.28*** -1.12 (0.97) (0.84) (0.91) (Quadratic in Lat-Long) Cell Phone -4.73*** -3.18*** -0.98 (1.15) (1.12) (0.97) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 Mean of Control Group 12.91 10.85 9.98 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 26 / 47
Weather Information Effect: Log of Crop Yield (Degree of Perishability of Crop) (Extremely) (Highly) (Least) (OLS) Cell Phone 0.06*** 0.03*** 0.02 (0.03) (0.01) (0.08) (Quadratic in Distance from Restriction) Cell Phone 0.05*** 0.04*** 0.01 (0.01) (0.009) (0.02) (Quadratic in Lat-Long) Cell Phone 0.03*** 0.02*** 0.009 (0.01) (0.006) (0.02) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 27 / 47
Price Information Effect: Percentage Difference Between Farmers and Market Price (Degree of Perishability of Crop) (Extremely) (Highly) (Least) (OLS) Cell Phone -0.34*** -0.38*** -0.29*** (0.14) (0.17) (0.05) (Quadratic in Distance from Restriction) Cell Phone -0.31*** -0.34*** -0.36*** (0.15) (0.16) (0.09) (Quadratic in Lat-Long) Cell Phone -0.29*** -0.35*** -0.26*** (0.11) (0.12) (0.07) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 Mean of Control Group 0.48 0.43 0.45 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 28 / 47
Income and Consumption (Agricultural Income) (Consumption) (OLS) Cell Phone 0.17*** 0.14*** (0.05) (0.04) (Quadratic in Distance from Restriction) Cell Phone 0.10*** 0.08*** (0.03) (0.02) (Quadratic in Lat-Long) Cell Phone 0.13*** 0.10*** (0.05) (0.03) Controls Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Number of Observations 450 450 p < 0.01, p < 0.05, p < 0.1 Saher Asad (LUMS) Crop Connection June 20, 2016 29 / 47
Income By Crop Category and Mechanism Saher Asad (LUMS) Crop Connection June 20, 2016 30 / 47
Robustness Checks Use of higher order polynomials Alternative identification strategy and external validity Full Country Radio test radio Pre-coverage placebo test Placebo Saher Asad (LUMS) Crop Connection June 20, 2016 31 / 47
Policy Implications and Conclusion Improving extension services - when to provide critical information? Post-harvest risk mitigation - coordination, post-harvest Insurance? Cell phone coverage makes farmers shift towards producing perishable cash crops Analysis of mechanisms suggest Farmer-Trader Coordination Effect dominates in economic importance Future work: Impacts on Traders, Impacts on Non-Agricultural Households Saher Asad (LUMS) Crop Connection June 20, 2016 32 / 47
THANK YOU Saher Asad (LUMS) Crop Connection June 20, 2016 33 / 47
Crop Profitability By Degree of Perishability Return Saher Asad (LUMS) Crop Connection June 20, 2016 34 / 47
Assumptions Return A1: 2 crops A2: Land normalized to 1 and all land cultivated A3: Extreme Perishability Assumption(EPA) A4: Participation Constraint (PC) holds A5: Fair Trader Assumption(FTA) A6: One Market A7: Utility Function is: U(c) = c ρ and ρ < 1 Saher Asad (LUMS) Crop Connection June 20, 2016 35 / 47
Baseline Case Return Based on FTA probability of sale when no one has cell phone is 1 θ = k Budget Constraint c = σ b Y p + (1 σ b )Y np Maximization gives σ b σ b = Ynp Y np+ Yp D 1 where D = [( k Yp Ynp 1 k )( Y np )] 1 1 ρ Saher Asad (LUMS) Crop Connection June 20, 2016 36 / 47
Case with Cell Phones Return Some fraction α gets cell phone access Those with cell phone; k = 1 Those without cell phone; k = 1 αθ θ αθ ; k < k σ c = 1 ; σ nc = Ynp Y np+ Yp D 1 where D = [( k 1 k )( Yp Ynp Y np )] 1 1 ρ Saher Asad (LUMS) Crop Connection June 20, 2016 37 / 47
Village Level Diff of Means Return Saher Asad (LUMS) Crop Connection June 20, 2016 38 / 47
Household Level Diff of Means Return Saher Asad (LUMS) Crop Connection June 20, 2016 39 / 47
Cell Phones versus Radios (Probability of Being a Major Crop in the Village) (Degree of Perishability of Crop) (Highly & Extremely) (Less & Least) (Highly & Extremely) (Less & Least) Radio 0.036 0.023 0.034 0.025 (0.226) (0.212) (0.242) (0.016) R-Square 0.24 0.18 0.21 0.14 Controls Y Y Y Y District FE, Crop FE and District-Crop FE Y Y Y Y Number of Observations 52,378 52,378 27,059 27,059 Sample Full Country Full Country Full Punjab Full Punjab Controls include percentage of area cultivated, pre-coverage distance from pucca road network, wealth measures like roof structure of majority of households in the village, elevation, access to agricultural credit and access to irrigation. The regression also includes set of district specific, crop specific and district-crop specific fixed effects. p < 0.01, p < 0.05, p < 0.1 Return Saher Asad (LUMS) Crop Connection June 20, 2016 40 / 47
Pre-Coverage Placebo Test (Probability of Being a Major Crop in the Village) (Degree of Perishability of Crop) (Highly & Extremely) (Less & Least) Cell Phone 0.01 0.007 (0.03) (0.04) R-Square 0.15 0.17 Controls Y Y District FE, Crop FE and District-Crop FE Y Y Number of Observations 498 498 Controls include percentage of area cultivated, pre-coverage distance from pucca road network, wealth measures like roof structure of majority of households in the village, elevation, access to agricultural credit and access to irrigation. The regression also includes set of district specific, crop specific and district-crop specific fixed effects. p < 0.01, p < 0.05, p < 0.1 Return Saher Asad (LUMS) Crop Connection June 20, 2016 41 / 47
Farmers Lined Up at the Market to Sell Crops Saher Asad (LUMS) Crop Connection June 20, 2016 42 / 47
Perishability Ranking of Crops Extremely Perishable : Tomatoes, Oranges, Mangoes, Corn, Sugar Cane Highly Perishable : Onion, Garlic, Potato, Taro, Peas, Fresh Turmeric Least Perishable : Millet, Feed, Sorghum, Rice, Wheat, Cotton Saher Asad (LUMS) Crop Connection June 20, 2016 43 / 47
Density Smooth at Village Level 0.02.04.06.08 20 10 0 10 20 Return Saher Asad (LUMS) Crop Connection June 20, 2016 44 / 47
Density Smooth at Household Level 0.05.1.15 5 0 5 Return Saher Asad (LUMS) Crop Connection June 20, 2016 45 / 47
Conditional Correlation (Probability of Being a Major Crop in the Village) (Degree of Perishability of Crop) (Highly & Extremely) (Least) (Highly & Extremely) (Least) Cellphone 0.29*** 0.01 0.31*** 0.04 (0.004) (0.02) (0.002) (0.1) R-Square 0.38 0.25 0.29 0.17 Controls Y Y Y Y District FE, Crop FE and District-Crop FE Y Y Y Y Number of Observations 52,378 52,378 27,059 27,059 Mean of Control 0.12 0.78 0.17 0.85 Sample Full Country Full Country Full Punjab Full Punjab Controls include percentage of area cultivated, pre-coverage distance from pucca road network, wealth measures like roof structure of majority of households in the village, elevation, access to agricultural credit and access to irrigation. The regression also includes set of district specific, crop specific and district-crop specific fixed effects. p < 0.01, p < 0.05, p < 0.1 Return Saher Asad (LUMS) Crop Connection June 20, 2016 46 / 47
Mechanisms : FTC (Number of Days Between Harvest and Sale) (Degree of Perishability of Crop) (Extremely) (Highly) (Less and Least) (OLS) Cell Phone -8.9*** -6.5*** 2.3* (2.23) (1.56) (1.40) (Quadratic in Distance from Restriction) Cell Phone -6.13*** -5.34*** 0.51 (0.05) (0.12) (0.45) (Quadratic in Lat-Long) Cell Phone -6.96*** -6.15*** 1.23 (1.12) (1.55) (1.85) Controls Y Y Y District Dummies, Crop Dummies, Dist Crop Dummies Y Y Y Number of Observations 450 450 450 Mean of Control Group 10.35 8.68 5.38 p < 0.01, p < 0.05, p < 0.1 Return Saher Asad (LUMS) Crop Connection June 20, 2016 47 / 47