Our pre-analysis plan, written in 2015, refers to a number of forms analysis that we do not present. For transparency, we describe them briefly here. First, we had intended to conduct an experiment to test credit constraints among traders by offering loans to a randomly selected subset of Commission Agents. We conducted a pilot for this experiment in the first season, issuing 62 short-term working capital loans to a group randomly selected from 124 CAs who expressed a desire for credit. In the end, the repayment rate on these loans was poor finding and our partner decided not to move this experiment to the intended scale, so we do not analyze it. Our PAP specifies a set of hypotheses about convergence between spokes and hubs, and the differential effect of treatment for spokes in which the hub is and is not treated. In the end we were only able to map 84% of our spokes to hubs, and the analysis conducted within this reduced sample is typically inconclusive, suggesting that the trading networks may be more complex than our simple hub-and-spoke mapping supposed. So while we emphasize deviations from the superhub in the text, we do not present analysis relative to hubs. The operating cost for running the platform during the three years of the project was ✩927,190. Making up these costs were program administration, including compensation for managers at IPA and AgriNet, along with the deal coordinators and the program staff in the field, was ✩560,112. Targeting, including call center operations and all village-level promotion activities, cost ✩168,105. Participant training of CAs and AN supervisors was ✩39,784. Program material costs, including airtime costs and the money required to run the guarantee system, were ✩53,648. Monitoring costs, primarily the eight staff members who supervised transactions on the ground and implemented the guarantees, were ✩46,757. Kudu’s costs, not borne by the project, consisted of salary for the lead programmer and manager of the platform, short-code fees, and radio ads, and totaled ✩58,784. Our platform has three separable components,hydroponic bucket and we consider the business case for each of them in turn. First of these is Kudu.
The core issue for the standalone Kudu model is that, due to limited use of mobile money in rural Uganda, the platform does not have a mechanism to collect commissions on transactions.29 Hence, it appears that the most logical model to make Kudu sustainable would be a user fee model where individuals pay to post bids and asks on the platform. Given a total number of bids and asks of approximately 54,000 and costs of ✩58,000, this fee would need to be approximately a dollar per use. While this is a tiny amount of money relative to the sums transacted in agricultural deals, it is likely that such a fee would sharply curtail use of the system by farmers and lead to paucity of asks. Further, the usage numbers recorded in the study reflect the influence of the finding call center and on-the-ground staff. An alternative business different model would be for Kudu to sell its up-to-the-minute price information. However, to generate reliable and sufficient data, it would have to operate at a massive scale, which presents a chicken-and-egg problem in terms of how to build up to a platform with sufficient scale to make this kind of market information service profitable. Hence, while Kudu represents a substantial potential boon to welfare from market participation, monetizing this benefit is not straightforward. A second component is the SMS Blast system. The costs of collecting the market price data and sending out the SMS Blast was ✩5,857 per month, although as a part of the study we were collecting data on many smaller spoke markets that likely would not make sense from a profit perspective for a commercial system, which may be better off focusing on only larger markets. Our baseline survey asks a question about WTP for market information from traders; the mean stated WTP for an SMS service providing information on spoke, hub, and super hub markets was ✩0.42 per month, indicating that our market information system could have broken even with 14,000 users. Had it been optimized to operate in fewer and larger markets, that threshold would fall. So, while our results do not indicate that price-only systems have large benefits for market participants, this business model may be the easiest to construct. Finally we have the most costly component of the study platform, which is the AgriNet call center, network of CAs, deal coordinators, and monitoring agents to track transactions on the ground.
While this hands-on approach appears to be a necessary part of launching an online trading platform, it is costly and raises the core question of how it can be paid for, given that the core value proposition of the platform to traders and farmers is a lack of intermediation costs on the platform. Given that a) the number of highly profitable trades on Kudu that AgriNet was able to intermediate directly was small, and b) substantial expense is required to put the logistics in place to be able to collect commissions on brokered trades, the project was fundamentally unable to develop a model through which brokerage fees could cover the costs of operating the system. A subscription model would be available either to Kudu or to a market price information system, but intermediation costs seem inherently to be linked to commissions on trade. Therefore, we conclude that this type of intermediary platform is not straightforward to make viable as a commercial enterprise at the scale observed in this study. Our 1,457 sampled study traders were representative of a broader population of 1,752 traders in study districts, meaning that we capture within the study 83% of the people on whom the harm of decreased trading margins fell. Trader profits fell by an average of ✩292 per year, or almost ✩900 over the three years of the study. Therefore study traders lost a total of ✩1.3 million in profits, and the broader sample of which they are representative lost a total of ✩1.53 million. Combined with the direct cost of running the platform, we therefore estimate the social cost of the platform to be ✩2.42 million dollars. The extrapolation of the total farmer benefits from our study sample requires careful consideration. Imprecision issues aside, it is easy to calculate the aggregate the estimated benefit of the intervention to farmers in our study sample. However, because we see evidence that intervention moved general equilibrium outcomes, like total trade volumes and prices, we must consider the effect of the intervention on the broader population of farmers, including those in our study catchment area but who were not sampled in our household surveys.
How can we best estimate the impact of the intervention on this population? First, we focus on treated households that did not receive the Blast, as the Blast was only targeted to a subset of individuals in our study and was not available to the broader population. Second, we estimate effects separately for those in the “Near” village finding, who are representative of a smaller population of households in the more urban village containing the TC, and for those in the “Far” village finding, who are representative of a much larger population of more rural households in the surrounding sub-county.30 To estimate these ingredients, we present in Table A.22 the core farmer impacts broken out by main treatment status, SMS Blast treatment status, and “Near” vs. “Far” LC1 status, with dummies for each of these three categories and full interactions between them. We can then use the coefficients from Table A.22 to calculate the total revenue effect in each of the four relevant strata.31 For the two strata treated by the Blast finding the study sample represents the population experiencing this effect. For the near stratum not receiving the Blast, the study sample of 1,280 should be representative of the 16,297 households in the same LC1s from which they are sampled. For the far stratum not receiving the Blast, the study sample of 567 should be representative of the much larger sample of 919,697 households in all ‘far’ parishes finding. We start by restricting our benefit calculation to the benefit of farmers in our study sample only. For these farmers, we calculate an aggregate benefit of ✩124,000, far less than the costs. However, applying the per-household benefits to the populations for which they should be representative, the outcome in the “Far” Blast control dominates the welfare calculation and drives our estimate of total benefits to farmers to ✩34 million dollars, thirteen times as large as the total social cost finding before declining to 453.0 million MtCO2e in 2009 as the economy slowed finding finding. Agricultural emissions, as a fraction of total net emissions, are also gradually increasing, from 6% in 2000 to 7% in 2009. In 2006, the California legislature passed Assembly Bill 32 finding, the Global Warming Solutions Act of 2006 finding,stackable planters which requires California to reduce greenhouse gas emissions to the 1990 level of 427 million MtCO2e by 2020. This amounts to a 15% reduction in 2008 levels and a 30% reduction in levels that would occur by 2020 if no actions were taken. AB 32 directs the California Air Resources Board finding to develop a plan for reducing greenhouse gas emissions, which the agency completed and made available for public comment finding. The plan identifies emission reduction strategies targeting emission sources for different sectors; nine focus on agriculture finding. The reductions are mandatory for some sectors, such as industrial enterprises and electrical power operations, but for agriculture they are voluntary. Agriculture represents a significant economic sector in California; it uses 25% of the state’s land and consumes about 75% of its water resources finding.
California agriculture produced approximately $34.8 billion in revenue in 2010 finding and ranks number one among states in agricultural cash receipts finding. In terms of greenhouse gas emissions, agriculture accounted for about 7.1% of California’s total in 2009 finding. The Air Resources Board plan for achieving AB 32 goals includes a combination of direct regulations,performance-based standards and market-based mechanisms. The centerpiece is a cap-and-trade program, which would initially target certain production or distribution processes, including cement production, oil refining, and other significant users of fossil fuels. The program is designed to potentially be linked to similar programs, in particular to an envisioned regional cap-and-trade program that includes New Mexico, British Columbia, Quebec and Ontario. Under California’s proposed cap-and trade program, regulated firms would be given allowances for greenhouse gas emissions once a year beginning in 2012, declining 2% to 3% per year to match emission reduction targets. Firms with surplus allowances could sell or save finding them for future use. Firms unable to reduce their emissions or looking to increase emissions could enter the market to purchase surplus allowances finding. These trading features of the proposed program finding are standard components of cap-and-trade systems, including those pioneered in California to reduce air pollution finding. The Board’s proposed program is also innovative because it would be open to additional private or public mitigation activities that reduce emissions or sequester greenhouse gases. Firms or groups in non-capped sectors may undertake mitigation activities and then be credited with offsets. Regulated firms can buy these and use them in lieu of government-issued allowances to meet a portion of their regulatory requirements finding. Firms in capped sectors could also theoretically undertake mitigation activities beyond their quota, depending on their marginal abatement cost. Trading under the cap, and potentially supplementing allowances with offsets, are both expected to reduce compliance costs, an objective of the Board’s plan. The two mechanisms are complementary: trading creates price signals that motivate regulated firms to seek low-cost mitigation opportunities, and the opportunity to earn credits that can be sold encourages regulated and non-regulated firms and groups to seek low-cost mitigations in sectors where emissions are not capped. To work, the program requires a comprehensive set of standards and regulations that details how emissions are measured and offsets created, especially if it is to be part of a regional cap-and-trade system. The standards and regulations must rigorously protect the environmental benefits associated with emission reductions, because regulated emitters have incentives to under-report emissions, and both buyers and sellers of offsets benefit from exaggerated mitigation claims finding. Initially, the Board plan envisions four sets of rules, called compliance offset protocols, for offset-generating projects, including one for livestock projects.