Crop yield is one of the most commonly used partial factor productivity measures. It is used to estimate the ratio of quantity of crop output, generally measured in kilograms or tons, to a sole input, land area. Ongoing EPAR research explores the policy implications of measuring yield by area planted versus area harvested. In this brief, we consider implications for crop yield estimates of other decisions in how to construct yield measures from household survey microdata. Using data from three waves of the Tanzania National Panel Survey (TNPS) and two waves of the Ethiopia Socioeconomic Survey (ESS), both part of the World Bank’s Living Standards Measurement Study-Integrated Surveys on Agriculture (LSMS-ISA), we calculate separate crop yield estimates across survey waves following different decisions on disaggregating yield by gender(s) of the plot decision-maker(s) and for pure-stand and mixed stand (intercropped) plots, on including crop production from multiple growing seasons, and on how to treat outlier observations.
Our technical brief helps to illustrate some of the potential effects of decisions of how to construct and report yield estimates, focusing on maize yield estimates. We observe large differences in maize yields between pure stand and mixed stand plots, though the differences vary between Tanzania and Ethiopia and depending on how outliers are treated. We observe strong patterns in maize yields by gender of the plot decision-maker in Tanzania, but the patterns differ in Ethiopia again depending on how outliers are treated. Conclusions around maize yield do not seem to vary across sub-populations of plots when short rainy season production is added to long rainy season production in Tanzania, though these yield estimates are larger. These findings demonstrate that interpretations of crop yield estimates should pay particular attention to the decisions made in arriving at those estimates and the directions of bias potentially introduced by those decisions.
The code used to calculate the various crop yield measures we examine and to generate the summary statistics in our technical brief is available in a GitHub repository. The code can generate estimates for a variety of crops beyond maize.
Additional code for generating agricultural development indicators from LSMS-ISA household survey microdata is available through our Agricultural Development Indicator Curation project.