For agricultural finance, you need agricultural data
I was recently speaking with someone at a large international development bank about Harvesting’s agricultural credit scoring tool. He asked me about the role of mobile phone data and so-called Call Detail Records (CDRs) in our models. This is something we’ve been thinking about for a while at Harvesting, and in a way, it’s an obvious question.
The emerging market fintech space is awash with players building lending models based on CDRs, social media activity and other mobile-driven data. Ever since Safaricom started playing around with the idea of M-Shwari (and probably before) it has been clear that there are huge opportunities here. Most digital lending models are premised on the idea that new streams of data will dramatically bring down the cost and improve the accuracy and scalability of small-ticket lending.
But Harvesting is about financing farmers. We help banks and MFIs close the financing gap that inhibits farmers from investing in agricultural inputs, machinery or anything else that could help boost productivity and improve their livelihoods. Where do CDRs and other mobile-driven data fit into this?
When a bank or MFI lends, it wants to know that the borrower can make the repayments. An obvious way to do this is to help finance a purchase of an income-generating asset that will directly connect to repayments. That could be a mortgage on a shop, for which regular sales will cover repayments. Or it could be helping a smallholder purchase a cow, with repayments based on sales of milk. If a bank is lending for farming, it needs to know that the borrower’s farming income will be enough to cover loan repayments (and also when that income will materialize — usually at harvest time). A mis-match between what the bank thinks the loan is for and what the borrower wants to use the money for is major driver of non-performance.
So why would Call Detail Records help us lend to a farmer? To facilitate a loan to a farmer, the purpose of which is agricultural, we need data about their agriculture. To predict their ability to repay, we need to predict what will be their agricultural cashflows and when. That means what are they growing, what are the agricultural conditions, what are the historical yields, what inputs have they used, and a whole load of other data points which we use to build an overall picture of how agriculture will link to financial activities.
In the rural villages that I see in India, people do use mobile phones and they do use social media. But the density of data is still sparse and while it might tell us something about their lives, it won’t tell us much about their agricultural lives.
I think the world of alternative credit scoring will grow and grow. The growing demand for short term loans in emerging markets — particularly among younger, urban populations — will mean more opportunities to finance smart phones, scooters, refrigerators and a million other items that might no longer seem like luxuries. For these markets, CDR data, social media data, utility bill payments and payment records on online retailers are already valuable data streams — but these are still essentially urban data, describing urban lives. As mobile phones and the internet penetrate deeper into rural areas, and we see more take up among farmers, we expect that they will start to tell us more about agricultural lives. But this will take some time.
To design financial services for farmers, you need data that tell you something about their agricultural activities. This is why we’re using data sources that get us closer to remote, rural areas. Data from remote sensing satellites can tell us about what is happening on farms miles from any bank branch. Water and soil mapping can tell us about soil fertility. Value chain data can tell us about historical and future transactions. But I’m yet to see firm evidence that the phone records of a smallholder farmer will tell us much that can support greater financial inclusion.
— Howard Miller, Director Client Strategy, Harvesting Inc
