During the period 2014-2016, the MRR Innovation Lab’s predecessor, the Assets & Market Assets Innovation Lab funded a research project assessing whether soil testing and fertilizer subsidies would increase fertilizer use and maize yields in Tanzania. The researchers found that farmers who received site-specific recommendations for fertilizer—based on tests of their own plot—and vouchers to buy fertilizer purchased and applied recommended fertilizers during the project period, and experienced significant increases in yields and farm revenues. 

Curious whether the results had been sustained when vouchers were no longer offered, the team— led by Abdulrazzak Tamim, in close partnership with Sokoine University—returned in 2019. They found that the use of fertilizer had dropped off and returned to levels observed at the project’s 2014 baseline. Yield gains made during the intervention were not maintained. The results suggest that a lack of liquidity during planting time continues to constrain fertilizer use. 

In this Q&A, we sit down with Hope Michelson, one of the PIs on the original and follow-up studies. 

Thanks so much for reengaging with the Innovation Lab. To start off, could you tell me about the role of the Innovation Lab in this work? 

We were able to do the original 2014-16 study thanks to a grant from the Innovation Lab. Due to the richness of the data that we collected, we were able to do a lot with a relatively small investment for this follow-on work.

One of the best things about the Innovation Lab funding is the research community. The Innovation Lab makes an effort to get PIs from a cohort of funded projects together for several meetings in which researchers present study plans and preliminary findings to each other. When you get all of those people in the room, all thinking about aspects of agricultural development, dimensions of what the projects are implementing and learning emerge that weren’t always apparent before. 

You found that farmers who applied the recommended fertilizer during the 2014-16 project period saw significant increases in farm profits. Were you surprised that fertilizer use had dropped off among this group when the provision of vouchers ended? 

We were disappointed but maybe not surprised. The project solved the liquidity constraint and provided farmers with information about their plot-specific soil nutrient limitations. In theory, the information would persist when the project left, but the liquidity obviously ended. So an interpretation is that it’s the liquidity constraint that’s really binding. This is consistent with findings during the project period, when the group of farmers that received information only but not the vouchers did not increase their fertilizer usage. 

One possible longer-term effect that we thought we might see when we initiated the follow-up was that farmers had learned and observed the yield gains from using the fertilizers that are specific to their soil constraints and would be able to find the cash for it. Even so, the season under the initial project, in which farmers would have learned about and observed benefits of the fertilizer for their field, was impacted by a drought. Farmers could still observe a benefit of the fertilizer—they had reduced losses compared to those without fertilizer—but this may have changed the learning dynamics from what we would have seen in a non-drought year. Also, because of the drought, farmers may not have experienced large enough profits to facilitate purchases in the following seasons. 

Soil heterogeneity—that is, differences in soil nutrients over some defined spatial scale—is a topic that frequently comes up in this literature, as the optimal type of fertilizer can depend on very locally-specific soil conditions. Are you looking at that topic any differently based on your findings? 

Ten years ago, when we first conceived of this project, we were in a team of economists and soil scientists and agronomists and there was a hypothesis that fertilizer recommendations were too broad, too coarse to be of use to farmers in many locations. 

Since then, we’ve learned from our work and others that understanding the degree of heterogeneity in these contexts is really important. We regularly find meaningful differences in soil conditions even across different plots of a single village area. What is the relevant heterogeneity for farmer technology adoption? For farmer learning? For policy makers?

As we think about how soil heterogeneity matters for production decisions and for farmer learning and investment, we also have to consider how farmers perceive differences in soil. Research has shown that farmers categorize soils in their area according to color and texture, though these categories are only loosely correlated with the measures we take to determine fertilizer recommendations. 

We’ve also observed farmers making decisions that would be consistent with a perception that their soils are quite similar. For instance, in an experiment I worked on with a research team in Malawi, farmers demonstrated significant willingness to pay for soil testing on other people’s plots. 

Coming back to the Tanzania study, in the beginning we identified a broad constraint of sulfur deficiency across almost all the farmers’ fields that was not being addressed by current, regional recommendations from the government. When we gave farmers more granular soil quality information about their specific field, they didn’t really respond to that, but they bought the fertilizer that addressed the broader-scale issue of sulfur deficiency. Now, perhaps it was because the ammonium sulfate was the cheaper fertilizer relative to those recommended at the individual level. Or, maybe they had more confidence in the ammonium sulfate recommendation because others received it as well.  

It’s worth considering if there’s a meso-scale at which we should be reaching farmers that might be “good enough.” So farmers aren’t applying fertilizer according to generalized regional recommendations, or at the hyper-specific plot level, but maybe at an inbetween level of precision at the sub-regional level, addressing broader deficiencies?

That dovetails with my next question around demand for soil testing, and the challenge of farmers being unlikely to get their soil tested if they’re unlikely to use the resulting recommendations for fertilizer. Your Malawi study explored farmer preferences around collective soil testing. What future directions do you see for soil testing? 

Right, so as I mentioned above, in Malawi, we found significant willingness to pay for soil tests among farmers on plots other than their own in their village, again suggesting that farmers perceive that soil heterogeneity is not so considerable over a local spatial scale; they still think they can benefit from results from the test of another farmer’s plot. It was also interesting in that, unlike the Tanzania project, there was no infusion of outside capital or fertilizer. This was just information – about the nutrient constraints on a neighbor’s plot. And farmers were willing to pay for it. 

But in terms of the Tanzania results, one of our takeaways is that information likely has more salience for farmers when the other constraints are solved. In terms of soil, that is, only once farmers have solutions to the constraints of liquidity for purchasing fertilizer, access to testing and the fertilizer they need, and the risk that their investment may be lost in case of a shock to their crop, then the results of soil tests will be relevant to them and we will start to see demand. 

Another important area of work for soil testing is building guidance around how to treat what is a multidimensional measure. Testing for soil quality, you end up with at least six salient measures; you’ve got organic matter, you’ve got macronutrients, you’ve got pH, and so on. There’s not much consensus in the soil science literature on which is the most relevant measure. For example, if one measure shows a value that is extremely low, then do none of the others really matter until that deficiency is addressed? And while we’re trying to sort through this complexity, there are products going full steam ahead on providing digital recommendations to farmers based on pretty broad soil map assessments.

What else is on your mind these days around soil? 

Based on meetings I’ve been to recently, I’m thinking a lot about stagnation and declines in agricultural productivity in Sub-Saharan Africa and what that means in the context of my work, especially our work related to technology adoption and heterogeneity. But also – lots of discussion of late about whether we should be using yields as our primary outcome measure of interest, versus something like total factor productivity. Because you can increase yields while expending way more on labor or finance on inputs, so yields alone may not be the right number to watch, depending on what we care about.

At the end of the day, people are just working with really degraded soils in some contexts. And they don't have the resources to either rotate their fields or make the investments to remediate the soils that they're working on. And that hasn't really changed. And there doesn't seem to be any magic solution.

***
The MRR Innovation Lab recognizes the important contributions of Cheryl Palm to this work.