Impact of SMS-Based Agricultural ...
Impact of SMS-Based Agricultural Information on Indian Farmers Marcel Fafchamps and Bart Minten This study estimates the benefits that Indian farmers derive from market and weather information delivered to their mobile phones by a commercial service called Reuters Market Light (RML). We conduct a controlled randomized experiment in 100 villages of Maharashtra. Treated farmers associate RML information with a number of deci- sions they have made, and we find some evidence that treatment affected spatial arbi- trage and crop grading. But the magnitude of these effects is small. We find no statistically significant average effect of treatment on the price received by farmers, crop value-added, crop losses resulting from rainstorms, or the likelihood of changing crop varieties and cultivation practices. Although disappointing, these results are in line with the market take-up rate of the RML service in the study districts, which shows small numbers of clients in aggregate and a relative stagnation in take-up over the study period. JEL codes: O13, Q11, Q13. The purpose of this study is to ascertain whether agricultural information dis- tributed through mobile phones generates economic benefits to farmers. We im- plement a randomized controlled trial of a commercial service entitled Reuters Market Light (RML) offered by the largest and best-established private pro- vider of agricultural price information in India at the time of the experiment. Operating in Maharashtra and other Indian states, RML distributes price, weather, and crop advisory information through SMS messages. We offered a one-year free subscription to RML to a random sample of farmers to test whether they obtain higher prices for their agricultural output. Marcel Fafchamps is a professor at Oxford University, United Kingdom his email address is marcel. firstname.lastname@example.org. Bart Minten is a senior research fellow at the International Food Policy Research Institute his email address is email@example.com. The authors thank Dilip Mookherjee, Shawn Cole, Sujata Visaria, three anonymous referees, and the editor for discussion and suggestions. This research would not have been possible without the full cooperation of Thomson Reuters India. The authors are particularly grateful to Raj Bhandari, Rantej Singh, Ranjit Pawar, and Amit Mehra for their constant support, to Alastair Sussock, Sudhansu Behera, Gaurav Puntambekar, and Paresh Kumar for their research assistance in the field, and to Grahame Dixie for his encouragements, suggestions, and comments. Funding for this research was provided by the World Bank and Thomson Reuters. A supplemental appendix to this article is available at http://wber.oxfordjournals.org/. THE WORLD BANK ECONOMIC REVIEW, pp. 1–32 doi:10.1093/wber/lhr056 # The Author 2012. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: firstname.lastname@example.org Page 1 of 32 The World Bank Economic Review Advance Access published February 27, 2012
We also investigate the channels through which the price improvement comes: better arbitrage across space and time ability to bargain with traders and increased awareness about quality premium leading to better agricultural practices and postharvest handling. We present simple models of the first two channels. Both models make testable predictions about farmer behavior in re- sponse to market price information. We test for the presence of these necessary channels. Given that RML circulates weather and crop advisory information, we also examine whether profits increase thanks to better crop management, reduced losses, or improved quality. There is a large interest and takeoff of similar private and public programs in India (Mittal, Gandhi, and Tripathi 2010) and elsewhere (Staatz, Kizito, Weber, and Dembele ´ ´ 2011), indicating that it is important to understand the impact of these interventions.1 Based on this, we expected to find a large and significant effect of RML on treated farmers. These expectations find less support than anticipated in our results. Many treated farmers state that they use the RML information, and they are less likely to sell at the farm-gate and more likely to change the market at which they sell their output. These findings are consistent with the idea that treated farmers seek arbitrage gains from RML information. We do not, however, find any significant difference in price between beneficiaries of the RML service and non-beneficiaries. This result is robust to the choice of estimator and method- ology. There is some evidence of heterogeneous effects: Treated farmers who are young—and presumably less experienced—receive a higher price in some regressions, but the effect is not particularly robust. Why we do not find a stronger effect of RML on the prices received by farmers is unclear. Over the study period there were important changes in prices, which are part of a general phenomenon of food price inflation in India (see, for example, World Bank 2010). Although rapidly changing prices should in principle make market information more valuable, it is conceivable that the magnitude of the change blunted our capacity to identify a significant price dif- ference, given how variable prices are. Point estimates of the treatment effect on price received are, however, extremely small and sometimes negative. We also find no significant evidence of an effect on transaction costs, net price, rev- enues, and value-added. In other channels by which RML may affect farmers, we find a statistically significant but small increase of the likelihood of grading or sorting the crop. This is especially true for younger farmers, possibly explaining why they achieve better prices. RML shows prices by grade to farmers and might have helped to inform them about the benefit from grading. With respect to other 1. In India, apart from RML, other initiatives include the Indian Farmers Fertilisers Cooperative Limited (IFFCO) Kisan Sanchar Limited (IKSL), a partnership between Bharti Airtel and IFFCO, as well as the fisher friend program by Qualcomm and Tata Teleservices, in partnership with the MS Swaminathan Research Foundation. In western Africa, Manobi and Esoko, private ICT providers, have developed a number of SMS applications to facilitate agricultural marketing there. Page 2 of 32 T H E WO R L D B A N K E CO N O M I C R E V I E W
RML information, we find no systematic change in behavior due to weather in- formation and no change on the crop varieties grown or cultivation practices. Farmers who changed variety or cultivation practice do, however, list RML as a significant source of information in their decision. To summarize, RML bene- fits appear minimal for most farmers, even though RML is associated with sig- nificant changes in where farmers sell their crops. There is some evidence that young farmers benefit more from the service, but the effect is not robust. I. D E S C R I P T I O N O F T H E I N T E R V E N T I O N A N D E X P E R I M E N T A L D E S I G N The Indian branch of Thomson-Reuters distributes agricultural information to farmers through a service called Reuters Market Light (RML). Subscribers are provided with SMS messages in English or the local language of their choice— in total 75 to 100 SMS per month. Subscribers are offered a menu of three agricultural markets and three crops to choose from. This menu is based on market research of farmers’ information needs and a pilot marketing program carried out in 2006-07. Prior to the experiment, Reuters gathered encouraging anecdotal information from farmers regarding the usefulness of RML. At the time of the experiment, Reuters had about 25,000 RML subscribers in Maharashtra. The RML content included market information, weather fore- cast, crop advisory tips, and commodity news. At the time of the study the price for the service was about $1.50 per month. Thomson Reuters is a trans- national corporation specializing in market information services, its core busi- ness. The corporation has a long experience collecting and selling market information and a reputation for accuracy. Although Thomson Reuters did not disclose how it obtains market price information, we have no reason to suspect that the provided information is inaccurate. Doubts about data accuracy would be detrimental to the firm’s reputation and would hurt its effort to market RML across India. A randomized controlled trial (RCT) was organized by the authors to test the effect of RML on the price received by farmers. The RCT was conducted in close collaboration with Thomson Reuters. Five crops were selected as the focus for the study: tomatoes, pomegranates, onions, wheat, and soybeans. In Maharashtra these crops are all grown by smallholders primarily for sale. Whereas wheat and soybeans are storable, the other three crops are not and should be subject to greater price uncertainty due to short-term fluctuations in supply and demand (Aker and Fafchamps 2011). We expect market informa- tion to be more relevant for perishable crops. Soybeans have long been grown commercially in Maharashtra, but tomatoes, onions, and pomegranates are more recent commercial crops. We expect farmers to be more knowledgeable about well-established market crops—such as wheat and soybeans—than about crops whose commercial exploitation is more recent—such as tomatoes, onions, and pomegranates. Finally, pomegranate is a tree crop that is sensitive to unusual weather and requires pesticide application. We expect the benefit Fafchamps and Minten Page 3 of 32
from weather information and crop advisories to be more beneficial to pom- egranate farmers. For each of the five crops, we selected one district where the crop is widely grown by small farmers for sale: Pune for tomatoes, Nashik for pomegranates, Ahmadnagar for onions, Dhule for wheat, and Latur for soya. All five districts are located in the central region of Maharashtra we avoided the eastern part of the state where sporadic Maoist activity has been reported, and the western part of the state which is less suitable for commercial agriculture. A total of 100 villages were selected for the study, 20 in each of the 5 dis- tricts. The villages were chosen in consultation with Thomson Reuters to ensure they were located in areas not previously targeted by RML marketing campaigns. Ten farmers were then selected from each village, yielding a total intended sample size of 1,000 farmers.2 Two treatment regimes were implemented. In the first regime—treatment 1—all 10 farmers in the selected village were offered free RML. In the second regime—treatment 2—three farmers randomly selected among the village sample were offered RML. The purpose of treatment 2 is to test whether the treatment of some farmers benefit others as well. Farmers who are not signed up are used to evaluate the externalities generated if farmers share RML information. Bruhn and McKenzie (2009) show that, in RCTs, stratification improves effi- ciency. Randomization of treatment across villages was thus implemented by constructing, in each district, triplets of villages that are as similar as possible along a number of dimensions that are likely to affect the impact of the treat- ment.3 A description of the process is provided in the supplemental appendix 1, available at http://wber.oxfordjournals.org/. We also offered RML to randomly selected extension agents covering half of the treatment 1 and treatment 2 villages. In principle, extension agents could disseminate the relevant information they receive, making it unnecessary to dis- tribute it to individual farmers. Whether they do so in practice is unclear, given that extension agents visit villages infrequently. The object of the RCT is to estimate the impact of RML on farmers who voluntarily sign up for the service because they benefit from it. In villages tar- geted by RML, only a small proportion of farmers sign up. They tend to be larger farmers with a strong commercial orientation in RML crops. There is no 2. This sample size was determined as follows. The primary channel through which we expect SMS information to affect welfare is the price received by producers. We therefore want a sample size large enough to test whether SMS information raises the price received by farmers. Goyal (2010) presents results suggesting that price information raises the price received by Indian farmers by 1.6 percent on average. Based on this estimate and its standard error, a simple power calculation indicates that a total sample size of 500 farmers should be sufficient to identify a 1.6 percent effect at a 5 percent significance level. To protect against loss of power due to clustering, we double the sample size to 1,000. We did not have sufficient information to do a proper correction of our power calculations for clustering. 3. More precisely, 6 triplets and one pair. Page 4 of 32 T H E WO R L D B A N K E CO N O M I C R E V I E W
point estimating the effect of RML on people unlikely to benefit from it. For this reason, participants are limited to farmers growing the district-specific selected crop for sale. These farmers have a large enough marketed surplus to amortize the cost of information gathering and enough experience with the crop to benefit from agricultural information. It remains that the farmers to whom we offered RML did not express an initial interest in it and may have dismissed the usefulness of something they did not pay for. Since a mobile phone is required to receive RML messages, we limit partici- pants to farmers with a mobile phone at the time of the baseline survey. We omit farmers who already were RML customers prior to baseline because they could not be used for impact analysis. Because we carefully avoided areas in which RML had already been actively promoted, few farmers were eliminated due to this condition. Further details on the sample and survey design are found in supplemental appendix 2. II. C O N C E P T U A L F R A M E W O R K Our primary objective is to test whether farmers benefit from the SMS-based market information and, if so, how. Information gathered by Thomson Reuters and conversations with farmers and RML customers in villages of Maharashtra suggest that farmers benefit in several ways. According to this information, timely access to market price information at harvest helps farmers decide where to sell, as in Jensen (2007). It also enables them to negotiate a better price with traders. To illustrate how informed farmers may obtain a higher price through better arbitrage, consider a farmer selecting where to sell his output. There are two possible markets, i and j, with transport costs ti and tj. We assume tj . ti, that is, market j is the more distant market. To focus on arbitrage, we assume that the distribution of producer prices F(p) is the same in both markets. In particu- lar, E[pi] ¼ E[pj] ¼ m. Given this, it is optimal for an uninformed farmer to always ship his output to market i since EU (pi - tj) , EU (pj – ti) for any utility function U(.).4 The average price received by farmers is thus m. A rele- vant special case is when i is selling at the farm-gate to an itinerant buyer and j is selling at the nearest market. In that case the farmer incurs no transaction cost, receives price pi, and does not learn the market price pj. Now suppose that the farmer is given information on prices in i and j. Shipping to i remains optimal if pi – ti pj – tj otherwise, the farmer ships to j. The average farmer price now is: E pi pi ti pj tj Pr pi ti pj tj þ E pj pi ti , pj tj Pr pi ti , pj tj m ð1Þ 4. Since p — ti . p — tj point wise. There is no role for risk aversion in this model. Fafchamps and Minten Page 5 of 32