11-12: Gathering the data needed to assess large farm productivity
Root for the tubers: extended-harvest crop production and productivity measurement in surveys
1The World Bank, Italy; 2University of Malawi; 3Consultant
To document the relative accuracy of survey methods for cassava production measurement a field experiment was implemented in Malawi over a 12-month period, randomly assigning households to one of four approaches: daily diary-keeping, with semi-weekly supervision visits (D1); daily diary-keeping, with semi-weekly supervisory phone calls (D2); two 6-month recall interviews, with six months in between (R1); and a single 12-month recall interview (R2). The analysis reveals that compared to D1, the average household-level annual cassava production is 295 kilograms higher (and assumed to be closer to the truth) under D2. While the difference between R1- and D1-based estimates is statistically insignificant, R2 underestimates annual production, on average, by 221 and 516 kilograms, compared to D1 and D2, respectively. For improved microdata on root and tuber crop production, the findings support the use of (i) D2, if deployed in a broader mobile-phone based survey, or (ii) R1, as a second-best alternative.
How much can we trust farmer self-reported data on crop varieties? Experimental evidence using DNA fingerprinting of cassava varieties in Malawi
1The Living Standards Measurement Study, Development Data Group, the World Bank, Italy; 2CGIAR Independent Science and Partnership Council’s Standing Panel on Impact Assessment, Italy; 3University of Canberra, Australia; 4International Institute of Tropical Agriculture Malawi; 5Chitedze National Agricultural Research Institute, Malawi; 6CAVA2, Malawi
This paper empirically estimates the extent of measurement error associated with alternative approaches for collecting data on the adoption of improved varieties. A range of non-rival data collection protocols were implemented for the same sample of 1260 cassava producers in Malawi, and the accuracy and relative cost-effectiveness of these protocols are compared to a benchmark of DNA fingerprinting using the DArTSeq platform. The results show that only 35% of the farmers could correctly identify their varieties. Identification achieved through using a photo-based survey protocol of morphological attributes achieved correct identification in only 5% of cases. Farmer self-reported data overestimates adoption of improved varieties by a factor of 19. Based on these findings, we recommend that any empirical study in which crop varietal status is an important variable should make the marginal investment of approximately $25 per sample per household for including DNA fingerprinting.
Land measurement bias: comparisons from GPS, self-reports and satellite data
1Northwestern University, USA; 2Asian Development Bank, Philippines
Nonclassical measurement error from farmer self-reports for plot size has been well documented primarily in comparison to using Global Positioning System (GPS). Our study investigates the reliability of Google Earth (GE) for plot size measurement and its impact on the inverse land size–productivity (IR) relationship and input demand functions. Comparing across four Asian countries, we find significant differences between GPS and GE only in Vietnam, where plot sizes are small. The magnitude of farmers’ self-reporting bias relative to GPS measures is nonlinear and varies across countries, with the largest magnitude in Lao PDR relative to Vietnam. Except Vietnam, the IR relationship is upwardly biased for lower land area self-reported measures relative to GPS measures. In Vietnam, the intensive margin of organic fertilizer use is negatively biased by self-reported measurement error. As remote sensing data becomes publicly available, it may become a less expensive alternative to link to survey data.
Assessing the impact of systematic measurement error in farmer-reported crop production on the scale-productivity relationship: evidence from a survey experiment in Mali
The World Bank, Italy
We contribute to the renewed debate on the inverse scale-productivity relationship (IR) by using primary survey data from a representative cross-section of sorghum-producing households in Koulikoro, Mali, and show that the IR can be explained by (systematic) measurement error in farmer-reported crop production. The analysis compares plot-level sorghum yields based on (1) farmer-reporting, (2) crop cutting, and (3) high-resolution imagery-based remote sensing. We find that with respect to crop cutting, sorghum yields based on farmer-reported crop production are overestimated across the board, ranging from an average overestimation of 270 percent in the first quintile of plot areas to 71 percent in the fifth quintile. By switching from farmer-reported to objective yield measures in production function estimations, we show that IR exists only if yield measurement is based on farmer-reported sorghum production. Further, we profile the measurement error in farmer-reporting, and expand on the implications for policy and household/farm surveys.
From the ground up: integrating survey and geospatial data for improved soil fertility measurement at scale
World Bank, Italy
Key to agricultural production, and often omitted from data collection efforts due to cost and complication, is soil health and quality. This paper sets out to validate the use of Africa SoilGrids 250m geospatial soil data by (i) comparing plot-level soil sampling results for several properties with that extracted from the Africa SoilGrids 250m database, and (ii) analyzing the ability of Africa SoilGrids 250m data to predict plot-level soil properties and indices of soil quality when integrated with household survey data. This is possible using data collected through the Methodological Experiment on Measuring Maize Productivity, Soil Fertility and Variety conducted in Eastern Uganda. Preliminary results highlight statistically significant differences in many soil properties as measured at the plot-level and as extracted through Africa SoilGrids 250m. For key soil properties, such as organic carbon and cation exchange capacity, geospatial data paints a more optimistic picture of the state of soils.