Almost 200 million chronically undernourished people live in sub-Saharan Africa, where the world’s last large tracts of arable land sit fallow and unfarmed. An African bread basket could help feed an exploding world population, starting with the continent itself–but only one percent of bank loans in East Africa currently go to agriculture, mainly because a lack of data on African food commodities means that lenders can’t assess the risk.
Gro Ventures believes the solution is to create a network of farmers and collect their data, using it to create credit models which allow banks to vet farmers for loans. “Think of it as an operating system for agricultural lending,” says founder Sara Menker. By 2015, Gro Ventures hopes to facilitate loans of $25 million to East African farmers. But this real-life network would need to span millions of people without connectivity, and thousands of miles of land over 54 African countries–collecting everything from population distribution, predominant crop types and soil types, to previous weather patterns and estimates of supply.
Menker was a vice president in Morgan Stanley’s Commodities Group when she predicted that food prices would rocket due to rising population demand. Menker, who is Ethiopian, traveled around commercial farms in her home country and elsewhere in East Africa looking for a good investment. A viable commercial farm needs access to capital, seeds and fertilizers, crop insurance, storage, and distribution. “I quickly realized that farmers didn’t have crop insurance, that they were borrowing at really expensive rates, and that they didn’t have services,” says Menker. “I said this can’t be good business, so it became almost this four-year obsession while I was still at Morgan Stanley. Eventually, I realized that the reason that none of the businesses I looked at worked, or at least the risk profiles didn’t work for me, was because the cost of capital was just too high. That’s simply because there isn’t good data.”
Banks quantify the risk of a loan via a multivariate mathematical model called a credit model which predicts the ability of a borrower to repay and determines the interest rate accordingly. A credit model linked to a commodities market, a highly volatile market dealing in raw materials like agricultural products rather than manufactured goods, will include variables indicating the creditworthiness of the borrower but also factors which will influence crop yields and future prices. “In commodity markets, quantifying the risk means understanding the underlying fundamentals,” explains Menker. “The pieces of the puzzle which make the ecosystem work. It’s understanding weather patterns. It’s understanding the soil type. It’s understanding what are the predominant crops being grown in different regions. What are the populations in these regions? What do storage facilities in the region look like? That starts to give you a very good feel for what supply looks like as well as what demand will look like.”
Gro Ventures tracked down and aggregated data from a subset of the bewildering range of nonprofits, social enterprises, and developmental agencies which work on agriculture in Africa. The company also formed a strategic relationship with the African Risk Capacity Agency in the African Union to use data that organization had previously not made accessible. “From 35 of 54 countries, we have data that is pretty granular. By that I mean on a 10×10 kilometer basis,” says Menker.
One early problem the company encountered was that much of the data available was static or out of date. “If you go to an FAO database (Food and Agriculture Organization of the United Nations), you will find that the last data point that they published was maybe from 2011. Things like prices, especially in East Africa, we are updating on a daily basis. Things like weather we are updating every 10 days. In Africa, it’s unprecedented,” says Menker. A data acquisition layer takes in the data in the form of spreadsheets, XML files, and GIS formats. The data harmonization layer reconciles the incoming data across different geographical and temporal dimensions.
Gro Ventures has also started to collect its own data. “The most underutilized source of information in this whole puzzle is the farmer,” says Menker. “The farmer is the best sensor you can have. He can answer very simple questions like, What are you planting? What have you sold today? Take a picture of a plant because that can indicate pests.” Data collection from farmers is still at the experimental stage. The company plans to partner with enterprises delivering agricultural services in order to bundle products to pay for the airtime of the farmers, or even to give farmers extra airtime in return for their data.
Gro Venture’s credit models take the aggregated data as input and output the optimal interest rate, the duration of a loan, and the frequency of repayments. The same data can be used to build credit models for different types of loans. Gro Ventures’s first product is a credit model for pooled lending. Ninety percent of farms in Africa today are very small, 2.5 hectares or less (the size of two sports fields), so it makes sense to share a tractor or other piece of farm equipment. Banks can use the model to lend to a group of farmers who want to buy a piece of machinery, spreading the risk across the entire portfolio of farmers. “The reason that I like to address mechanization for banks is also that it is also a hard asset which can be collateralized,” says Menker. “We want to be realistic about the types of risks that banks will be willing to take initially. For each farmer you’ll know how much land they have, what kind of crop they are producing, and what they sell. If you have lent successfully to this pool for a hard asset, then you have acquired the individual-level data that you need to lend for other things.”
Mechanization is crucial to increasing yields, which in East Africa currently languish at about a third of the world average. The number of tractors per hundred square kilometers of arable land in Africa today is 13. In Asia it’s 129 and the world average is 200. 3.5 million more tractors are needed in Africa to reach that level of mechanization.
I asked Menker about the similarities between pooled lending, microfinance, and securitization, none of which have enjoyed a stellar reputation of late. “Micro-lending schemes for agriculture are exactly the financial product we are trying to move away from by the creation of new data driven credit products,” she explains. ”We are using the social relationship aspect of micro-lending, but are focused on loans of a larger size and ultimately on lowering the cost of capital. No other outcome is better for farmers than lower costs of capital and increased productivity.” The securitized products created in the sub-prime housing market combined loans of individual homes or commercial real estate into packages which assumed that housing prices would continue to rise. “The pooling in our case happens before the loans are disbursed and are based on a hard asset that depreciates in value over time,” Menker says.
Via pooled lending, Gro Ventures hopes to facilitate loans of $25 million to around 73,000 farmers in co-ops and informal farmer organizations by 2015. When you take into consideration families and others working on farms, Menker estimates that 290,000 people will be indirectly affected. As the amount and variety of data it gathers increases, the company hopes to eventually be able to create models for crop insurance. This week Monsanto spent $930 million on The Climate Corporation, which sells data-driven insurance to farmers.
Ultimately, Gro Ventures could even help lower the sovereign debt costs of African countries. The company gathers estimates of the supply of various crops and those estimates vary between sources. “What that has allowed us to do is to become really smart about where things are being underestimated or overestimated,” says Menker. “A lot of things are underestimated in Africa. Many informal markets don’t get put into official databases. A third of Africa’s GDP comes from agriculture and today we are not estimating that correctly. If we were able to estimate that correctly, then that allows governments to be rated properly in international bond markets so that when they are borrowing money, they are doing it at appropriate rates based on real GDP numbers.”