Julia Berazneva and Leah Bevis are PhD candidates at Cornell’s Dyson School.
Soils are a fundamental resource for smallholder farmers: they support production of cultivated and managed vegetation used for food, animal feed, fiber, fuel, and medicinal products. They also store and filter water, support biodiversity, and play a key role in the planet’s carbon cycle (more on this in a later post).
So it comes as no surprise that higher food production is achieved on healthier soils. The health or fertility of soils can be measured by many metrics; one of the most common ones, however, is the level of soil organic carbon or soil organic matter. Increases in the soil organic carbon pool, for example, enhance yields through increasing available water capacity, improving supply of nutrients, and enhancing soil structure and other physical properties. The available data show that with every 1 ton/ha increase in the soil organic carbon pool, crop yields can increase by 30-300 kg/ha for maize, 20-70 kg/ha for wheat, and 10-50 kg/ha for rice (Lal 2006).
Not only do healthy soils directly translate to higher yields, but soil health may also influence the use, productivity, and agronomic use efficiency of conventional inputs. For example, Marenya and Barrett (2009a) show that in western Kenya maize yield response to nitrogen fertilizer is higher for soils with initially high soil organic matter levels. Based on an experiment providing fertilizer grants to female rice farmers in Mali, Beaman et al. (2013) find that women who receive fertilizer increase not only fertilizer application, but also the application of complementary inputs such as hired labor and herbicides, resulting in significantly higher yields.
Since smallholder income and welfare are often highly correlated with agricultural yields, soil degradation may also contribute to households falling into poverty. A correlation between soil fertility and poverty has been empirically observed in Uganda (Krishna et al. 2006) and suggested by the simulation exercises in the context of Kenya (Stephens et al. 2012) and Peru (Antle et al. 2006). As expected, the numerous constraints of poverty may also decrease investment in soils and lead to higher soil degradation (as demonstrated by Shively (2001) in the Philippines).
Soils, however, can be improved—land use and management practices that focus on increasing and maintaining soil fertility are numerous. Integrated soil fertility management (ISFM), for example, is defined as a set of practices that include the use of purchased fertilizer, organic inputs, and improved seeds in combination with the adaptation of these practices to local conditions (Vanlauwe et al. 2010). Other practices include no-till farming, residue retention, terracing, applications of animal manure and biochar, incorporation of grass and legumes in the rotation cycle, and use of agroforestry systems, among many others. They are often context-specific. Many of them also fall under the umbrella of “climate-smart” or “low emissions” agriculture.
Not surprisingly, the initial soil health of a plot influences a farmer’s decision to engage in soil conservation and/or restoration, as well as the agricultural practices the farmer employs to do so. Marenya and Barrett (2009b), for example, find that Kenyan farmers apply less fertilizer to plots with less fertile soils, where the marginal yield response to fertilizer is low. Conversely, application of amendments such as mulch or compost, which increase soil organic matter, may elicit higher yield gains on poor soils than on fertile soils (Bayala et al. 2012). The same may hold true for conservation structures such as terraces. For example, Amsalu and de Graff (2007) show that Ethiopian farmers are more likely to invest in stone terracing on plots with steep (i.e., easily eroded) slopes and poor soil fertility, where the payoff for such investment is high. Hagos and Holden (2006), on the other hand, find similar though weaker results in the same setting, while Yirga and Hassan (2006) find that perception of soil degradation is a significant, positive predictor of stone/soil bund adoption in central Ethiopia. At the same time, there is descriptive evidence that farmers may not alter fertilizer application rates according to perceived soil fertility (Sheahan and Barrett (2014) as well as this blog post).
Estimation challenges
Many challenges are associated with understanding and measuring the contribution of soil fertility—and the inputs/technologies/practices aimed at achieving it—to agricultural productivity and household poverty, both from farmers’ and researchers’ point of view. The vast literature on agricultural technology adoption deals with many of these challenges and provides nuanced answers as to why farmers may adopt (and dis-adopt) agricultural practices. See, for example, Moser and Barrett (2003), Conley and Udry (2010), or Suri (2011).
One particular challenge lies in the researchers’ ability to accurately measure soil quality. Soil sampling and analysis can be expensive and time consuming, so in many agricultural surveys soil quality goes unmeasured and becomes an “unobserved input” when estimating crop production. In other cases soil health is inferred from farmers’ answers to pre-coded questions about perceived soil quality (good, fair, poor), slope (flat, steep), or type (sandy, clay, loam), as in the World Bank’s detailed Living Standards Measurement Study – Integrated Surveys on Agriculture (LSMS-ISA). Such crude measurements, however, can be misleading and econometrically difficult to analyze. Time-invariant soil characteristics can be controlled for when panel data are available (Chamberlain 1984), but in many contexts soil fertility varies across time.
Another econometric challenge lies in dealing with reverse causality and self-selection. Correlation between soil fertility and wealth, for example, may reflect the income effects on soil fertility, but also the capacity of wealthy farmers to invest in soils (reverse causality). Consistent estimates of the effect of agricultural practices are difficult to uncover when farmers select into agricultural practices based on perceived soil quality (self-selection). Moreover, most econometric models used to date have not incorporated dynamic nature of soil resource management and the endogenous decisions farmers make in response to changes in soil health.
Ultimately, integrating biophysical and social sciences is a difficult task. New and exciting data (e.g., from Africa Soil Information System or detailed global maps of soil moisture from NASA) and collaborations across disciplines, however, can help make this integration easier. See, for example, a recent paper by Berazneva et al. (2014), which combines the efforts of economists and soil scientists and uses detailed data in Kenya to show that combined applications of mineral fertilizer and organic resources can double maize yields and simultaneously sequester carbon. More research aimed at understanding farmers’ perceptions of soil health and how soil health influences agricultural practices, as well as understanding adoption/dis-adoption of soil fertility management practices and natural resource-based poverty traps is sorely needed.
