2016年-世界发展银行全球_Land_Measurement_Bias_and_Its_Empirical_Implications___Evidence_from_a_Validation_Exercise_31页_1009kb

Land Measurement Bias and Its Empirical Implications: Summary

Core Content

This paper investigates the differences in land size measurements across three common methods: farmer self-reported (SR), Global Positioning System (GPS), and compass and rope (CR). It also explores the implications of these measurement errors on key agricultural relationships, including the inverse farm size-productivity relationship (IR) and input demand functions. The study utilizes data from the second wave of the Nigeria General Household Survey Panel (GHS-Panel) and a supplementary validation survey that includes CR measurements.

Main Findings

• Measurement Accuracy:

  • GPS measurements are generally more accurate than self-reported farmer estimates.
  • The CR method is considered the gold standard, but it is time-consuming and costly.
  • GPS measurements are less affected by human error and more efficient, though they can still have calibration errors.

• Bias in Self-Reported Estimates:

  • Farmers tend to overestimate the size of small plots and underestimate the size of large plots.
  • The over-reporting bias for the smallest tercile of plots is 83%, while the under-reporting bias for the largest tercile is over 20%.
  • The bias in SR estimates is nonlinear and affects the accuracy of econometric relationships.

• GPS Measurement Error:

  • GPS error is relatively consistent across plot sizes, ranging from -0.2% for large plots to -2.8% for medium plots and -2.0% for small plots.
  • GPS measurements are not significantly different from CR measurements, except for small plots where the error is more pronounced.

• Impact on Empirical Relationships:

  • SR estimates significantly distort the inverse land size-productivity relationship, leading to an overestimation of the negative correlation between land size and productivity.
  • SR estimates also underestimate the effect of land size on input utilization, including fertilizer, pesticide/herbicide, and hired labor.

Key Information

• Data Sources:

  • The GHS-Panel data includes 5,000 households, with 5,125 plots.
  • A validation survey was conducted on a subsample of these plots, measuring using SR, GPS, and CR methods.

• Methodology:

  • The study uses descriptive statistics and econometric analysis to compare the three measurement methods.
  • It also employs adjusted Wald tests to assess the statistical significance of differences in land size estimates.

• Sample Design:

  • The validation sample was stratified by plot size to ensure a representative sample across different sizes.
  • A total of 518 plots were measured, with 211 households participating.
  • Sample weights were calculated to ensure the validation sample was representative of the GHS-Panel population.

• Geographic Scope:

  • The validation survey was conducted in four states: Oyo and Osun (South West) and Benue and Kogi (North Central).
  • These states were selected for their geographical dispersion and safety.

• Implications:

  • The use of SR estimates in econometric models can lead to biased results and underestimation of input demand.
  • GPS measurements, while more accurate than SR, may still have non-trivial errors, especially on small plots.
  • The CR method, though the gold standard, is not always feasible due to time and cost constraints.

Structure of the Paper

1. Introduction: Highlights the importance of accurate land measurement in development economics and the potential for measurement error to affect empirical results.
2. Land Measurement: Background and Literature: Reviews existing literature on land measurement methods and their accuracy.
3. Data Description: Explains the data sources and sample design used in the study.
4. Descriptive Differences between Measurement Methods: Compares the three methods using descriptive statistics and visualizations.
5. Determinants of Differences between Methods: Examines the factors that influence measurement differences, such as plot size, rounding of estimates, and household characteristics.
6. Conclusion: Summarizes the findings and discusses the implications for policy and future research.

Policy and Econometric Implications

• The inverse land size-productivity relationship is significantly affected by measurement error, especially when using SR estimates.
• Input demand functions are also distorted by SR measurement error, leading to underestimation of the effect of land on input utilization.
• The paper emphasizes the need for accurate land measurement in policy analysis and econometric modeling, particularly in the context of agricultural productivity and input use.

Conclusion

The study concludes that while GPS measurements are more reliable than self-reported ones, they are not without error. The CR method remains the most accurate, but its implementation is resource-intensive. The paper underscores the importance of using accurate land measurement methods in empirical research and policy analysis, especially in developing countries where land size data is crucial for understanding agricultural productivity and input demand.