For smallholder farmers in rural areas, learning advanced agricultural technologies and adopting improved agricultural technologies can significantly boost productivity and income, and has been an important policy objective. Social ties can often play a critical role here, as farmers learn not only from formal training but from neighbors, friends, and family who have adopted successful practices.
However, in some circumstances, the same socially tied individuals may also be competitors. Because developing country markets are often highly localized, individuals with social ties may compete directly for the same economic rents. For example, among workers in the same village, the number of employers is largely fixed, and relatively modest changes in the number of skilled workers could affect wages.
In my job market paper, I examine how these dynamics unfold in rural Burundi. I argue that in some cases, individuals may hoard, rather than share knowledge with socially connected individuals when doing so may reduce their economic rents.
A Puzzle: Uneven Diffusion of Profitable Technologies
In rural Burundi, farmers are exposed to some key agricultural techniques: row planting and composting. Some farmers are members of NGOs, which train them directly to use these technologies. Others, instead, rely on transmission from their peers to learn the technologies, absent other sources of training. We first document a very uneven diffusion of these technologies beyond the NGO members in several Burundian villages.
One key difference between these technologies that could explain this result is the extent to which the returns to knowing the technology decline if the technology diffuses more broadly. Row planting---which involves precise planting in rows to maximize yields---is labor-intensive and commands a wage premium in the local labor market. Because Burundian agricultural labor markets are highly localized, row planting is a technology with "rival" returns: even a moderate increase in the number of laborers who master the technology can affect, in principle, equilibrium prices. In contrast, composting, while beneficial, is not valued in the labor market and its diffusion is unlikely to affect the output prices, making it a “non-rival” technology because the adoption by one farmer doesn’t reduce returns for others.
The Experiment: Testing the Existence of Knowledge Hoarding
To test whether rivalry of returns affects sharing, I ran a first experiment where I organized training events where skilled farmers were paired with those who had not yet learned row planting. However, I added a twist: some training pairs came from the same village (labor market), while other (randomly assigned) pairs were from different villages.
The findings were striking. When skilled farmers were paired with trainees from different villages, 38% of the unskilled farmers learned the technique effectively. But when both the trainer and trainee came from the same village, that rate dropped to less than 3% (see Figure 1). This result seems to be explained by the quality of the training provided in the two arms: incumbents were much more likely to provide feedback and corrections to unskilled who were not competitors (i.e., from a different labor market).
Interestingly, we found that farmers in smaller villages, where the labor market is even more constrained, incumbents were 25% more likely to hoard their skills. This suggests that the closer the competition, the more intense hoarding can become.
Figure 1
To show that this effect was driven by a fear of losing returns, rather than something about the individual being shared with, I also introduce training events where training was in the non-rival technology (composting). Here, in contrast, we find very high rates of diffusion regardless of whether the pairs competed in the same labor market or not.
The Winners and Losers when Hoarding is Reduced
To understand the downstream impacts of knowledge hoarding, in 121 additional villages we invite 30% of the labor force who is unskilled to a training event---again randomizing at the village level whether the workers are paired with a skilled worker in their own village or different village. This enables us to compare the status quo (skilled and unskilled workers from the same village) against a counterfactual world where knowledge-hoarding motives are substantially lower (skilled and unskilled workers operating in different markets).
We collect data on labor market outcomes, adoption of row planting, and farm output for over 6,500 farmers across the 121 villages and find that reducing knowledge hoarding creates winners and losers. For unskilled workers who were successfully trained in row planting where hoarding was limited, the benefits were substantial: they saw their earnings increase by 7% thanks to a shift in their days of employment toward the more productive skill. However, this gain came with a cost for skilled workers in the same villages, who experienced an average earning drop of around 6% as competition in the labor market increased. This suggests that the concerns voiced by the incumbents about fear of economic losses are, at least in the short-term, warranted.
The Costs of Hoarding: Lost Productivity
These shifts also impacted the overall market equilibrium. In villages that experienced an increase in labor supply skilled in row planting, the equilibrium wage for row planting fell by about 3%. The effects extended to productivity as well: limiting knowledge hoarding led to a noticeable efficiency gain, with yields increasing by roughly 19% due to a broader adoption of row planting. This implies that hoarding, representing a significant missed opportunity for aggregate productivity in these communities.
Why Overcoming the Knowledge-Sharing Barriers is Hard
While, in principle, the gains to the unskilled workers would suggest that they could pay someone for training, we document that this may be hard in practice. First, the unskilled workers, being the poorest members of the community, may not afford to pay for their training. Second, the incumbents’ expected cost for sharing may be substantial: incumbents mention that they expect that other incumbents would socially sanction (for instance, by denying future work referrals) the individual who is found training someone else.
Implications for Development Programs
In small markets or in areas with inelastic demand, the diffusion of skills can lead to negative externalities or a loss of economic rents for incumbent workers. Policymakers need to take these costs into account when designing agricultural development programs. Structuring diffusion programs to include targeted incentives for knowledge-sharing is essential. This might involve strategies like group-based rewards or cross-community training networks to reduce competitive pressures and encourage skill transfer. By aligning economic incentives with diffusion goals, development programs can create an environment where knowledge-sharing supports both individual livelihoods and broader community growth.
Limitations of the Study
There are limitations to these findings. A key concern is that it can be hard, ex-ante, to define a technology as generating rival returns or not. For example, in the context of our study, it would be unclear that row-planting may have rival returns without knowing in detail of how the labor market functions.
Conclusion
Knowledge-sharing is often touted as a key driver of growth in low-income settings, yet the reality is more complex. In tightly-knit communities with limited economic opportunities, the same social ties that, under certain conditions, can foster learning, can also inhibit it. For rural development efforts to succeed, they need to take these dynamics into account, crafting policies that recognize both the power and the pitfalls of social networks.
Luisa Cefalà is a job market candidate at the University of California, Berkeley. You can learn more about their work here. This paper is coauthored with Franck Irakoze (University of Burundi), Pedro Naso (SLU), and Nicholas Swanson (Stanford University).
Join the Conversation