2017年-世界发展银行全球_Could_the_Debate_Be_Over____Errors_in_Farmer-Reported_Production_and_Their_Implications_for_the_Inverse_Scale-Productivity_Relationship_in_Uganda_76页_1mb

Summary of "Could the Debate Be Over?" by Sydney Gourlay, Talip Kilic, and David Lobell

Core Content

This study investigates the inverse scale-productivity (IR) relationship in Ugandan maize farming, focusing on whether this relationship is a statistical artifact caused by errors in farmer-reported production data. The research is based on a two-round household panel survey conducted in Eastern Uganda, comparing self-reported production with more objective measures such as crop cutting and remote sensing.

The IR relationship suggests that as plot size increases, productivity decreases. However, the authors argue that this relationship may not be a true economic phenomenon but rather an artifact of data measurement. They find that when using more accurate, objective methods for measuring yield (e.g., sub-plot crop cutting, full-plot crop cutting, and remote sensing), the relationship shifts from diminishing returns to constant returns to scale.

Main Findings

• Farmer-reported production consistently supports the IR relationship.
• Objective yield measures (based on crop cutting and remote sensing) suggest constant returns to scale, both at the mean and across the distribution of yield data.
• The core finding is driven by persistent overestimation of farmer-reported maize production and yield, particularly in smaller plots.
• The IR appears to be a statistical artifact, not a true economic relationship, when production data is based on self-reports.

Key Implications

• The IR may be influenced by measurement errors in farmer-reported data, including:
  • Recall bias due to limited memory.
  • Rounding of values, which is more problematic for smaller plots.
  • Social desirability bias, where farmers may overstate their production.
• The use of non-standard units and incomplete information on crop condition and state can further distort production estimates.
• Conversion factors for production to kg-equivalent are often unavailable or outdated, making it difficult to standardize measurements.

Methodology

The study uses panel data from the Methodological Experiment on Measuring Maize Productivity (MAPS), which includes:

• Self-reported production data.
• Crop cutting data (sub-plot and full-plot).
• Remote sensing data for plot area and yield.

The data was collected by the Uganda Bureau of Statistics, with support from the World Bank LSMS. The survey involved three visits per household: post-planting, crop cutting, and post-harvest. These visits collected detailed information on plot size, cultivation patterns, inputs, and management practices.

Data Collection

• Sampling Design:
  • 75 enumeration areas (EAs) were selected, divided into three strata.
  • In each EA, households were categorized based on whether they cultivated pure stand or intercropped maize plots.
  • A total of 900 households were surveyed in Round I, with 440 households having data from both rounds in Round II.
• Fieldwork:
  • Enumerators used GPS devices to measure plot areas and record boundaries.
  • Crop cutting was conducted using a random subplot placement method, with strict supervision to prevent interference from farmers.
  • Post-harvest data was collected to obtain farmer-reported production and inputs.

Key Measurement Domains and Methods

1. Plot Area Measurement:

   • GPS-based measurement was used to determine plot areas accurately.
   • Plot boundaries were identified with the plot manager, and areas were recorded in square meters.

2. Maize Production Measurement:

   • Crop cutting is considered the gold standard for yield measurement.
   • Subplots were laid out in a random and controlled manner, and harvests were weighed and barcoded.
   • Moisture content was standardized to 12% for all samples.

3. Soil Fertility Measurement:

   • Soil samples were collected during the post-planting phase.
   • These samples were used for objective soil fertility testing.

Conclusion

The study concludes that the inverse scale-productivity relationship observed in many previous studies may be due to measurement errors in farmer-reported data. When using objective measures, the relationship disappears, suggesting that the IR is not a fundamental economic relationship but rather a statistical artifact. This finding has important implications for policy formulation, as it challenges the basis for land redistribution and agricultural development strategies that rely on the IR. The authors also emphasize the need for high-quality, objective data in agricultural research and policy analysis to ensure accurate assessments of productivity and land use.

Contributions

• This paper is among the first to demonstrate that the IR could be a statistical artifact.
• It highlights the systematic biases in farmer-reported data and the importance of objective measurement methods.
• The results are robust to the inclusion of objective soil fertility, genetic heterogeneity, and edge effects in the analysis.

Keywords

• Maize
• Yield Measurement
• Plot Area Measurement
• Inverse Scale-Productivity Relationship
• Crop Cutting
• Remote Sensing
• Household Surveys
• Uganda
• Sub-Saharan Africa