DNA fingerprinting is increasingly used to measure adoption of improved crop varieties in low-income countries, often revealing high rates of misidentification attributed to farmer error. We show that this interpretation overlooks important sources of mismatch between survey- and fingerprinting-based identification. Using a field experiment in Costa Rica, we find that selecting respondents with higher varietal knowledge improves identification by 19 percent. This effect reflects variation in farmer knowledge and is not solely driven by intra-household respondent selection. We find that peer learning and on-farm variety testing play a role in explaining higher levels of knowledge. Further, we show that inherent uncertainty in genomic methods accounts for a substantial share of the observed mismatch, which could be mistakenly attributed to farmer misidentification. Farmer and DNA fingerprinting based mismatch rates increase as precision of the DNA reference library variety categorization declines. Notably, roughly one-third of observed matches occur in cases where genomic methods identify varieties with high confidence. We explain the implications of relying on DNA fingerprinting adoption information that omits consideration of breeding- and social science-specific methodological choices. We conclude that, in the current quest for methodological rigor which is driving the surge in fingerprinting adoption studies, considerations of farmers’ knowledge and fingerprinting uncertainty should be prioritized.
- Replicate figures and descriptive, which contains the survey data and the code to replicate Fig 2 and Fig A2 of the manuscript
- Replicate Table 1-2-3, which contains the fingerprinting data and the code to replicate Table 1-2-3 in the manuscript
- Estimation proximity matrix, which contains data and code to replicate the spatial proximity matrix used as input in Table 3
Please contact Martina Occelli [mo386(at)cornell.edu] and Luis Sanchez [ls2252(at)cornell.edu]