Potato Potahto in the FAO-GAEZ Productivity Measures? Nonclassical Measurement Error with Multiple Proxies

Abstract: The FAO-GAEZ productivity data are widely used in Economics. However, the empirical literature rarely discusses measurement error. We use two proxies to derive novel bounds around the effect of agricultural productivity in a setting with nonclassical measurement error. These bounds rely on assumptions that are weaker than the ones imposed in empirical studies and exhaust the information contained in the first two moments of the data. We reevaluate three influential studies, documenting that measurement error matters and that the impact of agricultural productivity may be smaller than previously reported. Our methodology has broad applications in empirical research involving mismeasured variables.

• The paper identifies and addresses the issue of nonclassical measurement error in FAO-GAEZ agricultural productivity data using a novel econometric approach with multiple proxies.
• It proposes a partial identification method that relaxes standard assumptions about measurement error, utilizing different versions of FAO-GAEZ data to derive bounds for true effects.
• Applying this method suggests that prior influential studies may have overstated the economic impacts of agricultural productivity, emphasizing the need to account for measurement error in empirical economics.

Analysis of Nonclassical Measurement Error in FAO-GAEZ Data and Its Impact on Economic Outcomes

The paper "Potato Potahto in the FAO-GAEZ Productivity Measures? Nonclassical Measurement Error with Multiple Proxies" by Rafael Araujo and Vitor Possebom addresses an important empirical challenge in economic research: the presence of nonclassical measurement error in agricultural productivity data, specifically FAO-GAEZ data. The researchers propose a novel econometric approach to partially identify the effect of agricultural productivity, accommodating measurement errors by leveraging two versions of FAO-GAEZ data as proxies. This paper reevaluates existing studies and suggests that previous estimates might have overstated the economic impacts of agricultural productivity due to unaddressed measurement errors.

Key Contributions

1. Identification of Measurement Error: The paper begins by acknowledging a rarely addressed issue in empirical economics—the nonclassical measurement error in the FAO-GAEZ crop productivity measures. It documents discrepancies between different versions of the data, which suggest the presence of errors due to differences in model parameters and data sources.
2. Methodological Innovation: To handle the nonclassical measurement error, the authors propose a partial identification method. Unlike traditional approaches that either ignore measurement error or assume it is classical and uncorrelated with the true variables, this method relaxes these assumptions. It utilizes two proxy variables derived from different versions of the FAO-GAEZ data, allowing for potentially correlated measurement errors between them.
3. Reevaluation of Influential Studies: The paper applies this methodology to reevaluate previous influential studies by Nunn and Qian (2011), Bustos et al. (2016), and Acharya et al. (2016). The findings suggest a reduced impact of agricultural productivity on economic outcomes than reported by these studies when accounting for measurement errors.
4. Implications and Applications: The implications of this research extend to various fields within economics, such as development, economic history, and political economy, where agricultural productivity data play a crucial role. Moreover, this approach can be applied to other empirical contexts with mismeasured variables, thus having broader relevance beyond the current application.

Technical Insights

• Bounding Approach: The authors derive bounds for the true effect of productivity on economic outcomes, which utilize all the information in the first two moments of the observed data distribution. This methodological approach is rigorous and provides a credible alternative when full identification is not possible due to measurement errors.
• Practical Testing Framework: The paper offers a testing mechanism for their assumptions about measurement error. By comparing the similarity of estimated bounds with standard OLS estimates, researchers can assess whether measurement error plays a significant role in their empirical context.

Conclusion

This research highlights the critical role of accounting for measurement error in economic data analysis. By developing a robust econometric framework and testing it against empirical data, Araujo and Possebom demonstrate that previously estimated effects of agricultural productivity might have been overstated. The results emphasize the importance of considering measurement error in empirical economics and suggest that applying this novel methodology could lead to more accurate and reliable estimates. Researchers are encouraged to adopt this approach in other areas where measurement error is suspected, thereby improving the robustness of empirical findings across disciplines.