Over the last year and a half, EPAR has constructed over 150 indicators related to agricultural development, and shared our decisions on data cleaning and indicator construction. We do this work and share our outputs both to facilitate analyses of these rich datasets and to make estimates of relevant indicators available to a broader audience of potential users. There are numerous decision points in the process of constructing indicators that affect the final estimates. These choices and the resulting estimates can inform policy and decision making both in-country and among international development stakeholders. This post is part of a blog series examining the potential consequences of different measurement choices.
Calculating crop yields using area harvested vs. area planted
Increasing agricultural productivity is an important agricultural development goal. Crop yield is a commonly used indicator of land productivity. Crop yield for any particular crop can be calculated as a measure of crop production weight (in kg) per area of land harvested or area of land planted (in hectares). Our question is, does the choice of area harvested or planted in the denominator of yield have important implications for interpreting patterns in land productivity?
Empirical evidence
Household-level survey data from the World Bank’s Living Standards Measurement Study - Integrated Surveys on Agriculture (LSMS-ISA) provide information on both area planted and area harvested, making it possible to compare yield estimates based on the choice of denominator. Figure 1 compares estimates from the LSMS-ISA Tanzania National Panel Survey household survey estimates of maize yield using both area planted and area harvested to FAOSTAT estimates of maize yield in Tanzania. FAOSTAT defines yield as “the harvested production per ha for the area under cultivation” and area under cultivation as “the area that corresponds to the total sown area, but after the harvest it excludes ruined areas (e.g. due to natural disasters).” Thus, FAOSTAT yield estimates are calculated using production quantity over area harvested. The differences in estimated LSMS-ISA yield values using area planted versus area harvested are statistically significant.
Figure 1. Comparison of Tanzania maize yield estimates (kg/ha) by area measure and data source
Note: “arhv” denotes estimates of yield by area harvested, and “arpl” denotes yield by area planted. TNPS estimates are means for rural households only. In all estimates, area planted and area harvested on a given plot are constrained to not exceed the plot size, as measured by GPS when available. FAOSTAT yield is the reported harvested production for the total crop area under cultivation. Source EPAR estimates (TNPS); FAOSTAT (Metadata http://www.fao.org/faostat/en/#data/QC, 3_5_2019)
Where does this difference come from?
The difference between area planted and area harvested occurs when farmers experience a reduction in crop area between planting and harvesting, also known as “pre-harvest losses”. These losses may occur due to unforeseen shocks such as floods, drought, theft, and fire. In addition to pre-harvest losses, other reasons such as labor constraints or changes in crop prices may result in farmers choosing to harvest less than the full cultivated area.
What are the implications?
Some pre-harvest losses can be mitigated by farmer management practices such as using pesticides and herbicides or using improved variety seeds that can be more resistant to pests or drought. Land management choices and constraints to adopting practices that could mitigate negative effects of unforeseen shocks may systematically occur in certain subgroups. For example, using data from Wave 4 of the Tanzania LSMS-ISA, we find that agricultural households with per capita consumption under $1.90 (2011 PPP) are statistically significantly less likely to use improved seed. If certain sub-populations systematically experience more pre-harvest crop loss, using yield measures based on area harvested rather than area planted may disproportionately overestimate productivity across distinct subpopulations. In the Tanzania National Panel Survey data, plots planted with more than one crop were significantly less likely to significantly less likely to experience area loss: approximately two thirds of mixed crop plots do not experience any area loss, whereas two-thirds of monocropped plots do. This suggests that there is the potential for the choice of yield metric to mischaracterize the relative productivity of farm management strategies and under-value contributions of plot agrobiodiversity to productivity from reducing area losses.
By Isabella Sun
Summarizing research by C. Leigh Anderson, Travis Reynolds, & Pierre Biscaye