Learning and the Adoption of High Yielding Variety Seeds : Panel Data Versus Cross-Sectional Data

To date, due to the lack of panel data, most micro-level empirical studies of technology adoption have used cross-sectional data. These studies cannot examine the dynamic processes of adoption such as learning. This paper uses panel data to study the adoption of a new high-yielding variety seed. First, it establishes that learning is an important variable in the adoption process. Second, it establishes that cross-sectional estimates of a dynamic process are potentially biased. The extent of the bias is examined by comparing cross-sectional estimates with those form the panel. Third, it illystrates the econometric methods needed to estimate a dynamic model when controlling for unobserved household heterogeneity.