Knowledge Hoarding and Diffusion of Profitable Technologies

Study Context

A large literature in development economics posits that “social learning” can achieve widespread diffusion of profitable technologies, especially when individuals share social ties, such as in rural villages. This is the idea underlying large-scale extension programs aimed at technology diffusion. However, when incumbents (those with early access to technology) derive rents from exclusive ownership of knowledge, they might have incentives to “hoard knowledge”, i.e. actively prevent others from learning. We study this question in the context of modern agricultural planting techniques in Burundi, which we have shown to be profitable in another CEGA-sponsored project. In this setting, skilled laborers derive rents from exclusive ownership of knowledge in the form of higher labor market earnings. Indeed, they report fearing their own displacement (losing relationships with current employers), and social sanctions from other skilled laborers if they trained unskilled laborers, aimed at least in part at preventing downward pressure on wages (as in Kaur et al. 2023).

Study Design

To test for the existence of knowledge-hoarding motives, the researcher team uses a Becker-Marschak-De Groot method to elicit skilled laborers’ willingness to accept to train an unskilled laborer. Crucially, the team varies whether the unskilled laborer directly competes with the skilled laborer for jobs, while holding other observable characteristics constant. The team validates this measure, showing that it correlates with fear over potential earning losses from knowledge diffusion, and that it predicts real-life behavior.

Next, the team asks whether these fears are warranted. They randomly train randomly selected unskilled laborers in the village, and look at the effects on the earnings of skilled laborers that are connected to them. To test for general equilibrium effects on wages, the team randomly assigns villages to a treatment that reduces incentives to “hoard knowledge”. This design also allows the team to quantify the distributional consequences of knowledge diffusion (who wins, who loses, and by how much).

Finally, the research team designs a series of experiments to understand the role of misinformation, biased beliefs, and fear of social sanctions in the persistence of this low knowledge-sharing equilibrium.

Results and Policy Lessons

Results forthcoming.