Applied Economics - Exploring the complementarity between traditional econometric methods and machine learning - an application to adoption and disadoption of conservation practices

Aug 2, 2026

Upload by Riley

This study examines how traditional econometric methods and machine learning can complement each other when analyzing agricultural adoption and disadoption of cover crops, an important conservation practice. Using a unique longitudinal panel survey of Iowa farmers, the analysis contrasts logistic regression with a Random Forest model to explain both initial uptake and withdrawal. Results balance interpretability and prediction: logistic regression supports economic-theory explanations while Random Forest improves predictive accuracy and uncovers nonlinear relationships. SHAP further pinpoints key disadoption drivers, informing policy design for more sustained conservation adoption.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Conservation practice adoption versus continued use

Limits of traditional econometric methods

Role of machine learning and complementarity

APPLIED ECONOMICS [https://doi.org/10.1080/00036846.2025.2462792](https://doi.org/10.1080/00036846.2025.2462792) Exploring the complementarity between traditional econometric methods and machine learning – an application to adoption and disadoption of conservation practices Zhushan Dua, Hongli Fengb and J. Arbucklec a Department of Economics, Iowa State University, Ames, Iowa, USA; bDepartment of Economics and Center for Agricultural and Rural Development, Iowa State University, Ames, Iowa, USA; cDepartment of Sociology and Criminal Justice, Iowa State University, Ames, Iowa, USA ABSTRACT This study explores the potential complementarity between traditional econometric methods and machine learning in analysing the adoption and disadoption of a key conservation practice in agriculture, cover crops. While the adoption of conservation practices has been widely examined, the literature on their disadoption is limited. Using a unique longitudinal panel survey of Iowa farmers, we compare logistic regression models with a Random Forest algorithm to examine factors driving conservation adoption and disadoption. Our findings show that while traditional logistic regression models offer interpretability grounded in economic theory, Random Forest provides superior predictive power and reveals complex, non-linear relationships among key factors such as past adoption behaviour, cost-share participation, and farmer perceptions. SHAP (SHapley Additive exPlanations) analysis identifies adoption scale, past adoption behaviour, and environmental factors as primary drivers of disadoption. Farmers with larger previous cover crop acreage and consistent adoption history are significantly less likely to disadopt, while the longterm impact of cost-share programmes on continued use appears limited. By combining machine learning’s predictive power with the interpretability of traditional econometrics, the study provides a deeper understanding of the drivers behind conservation decisions, which are crucial for informing policy design that promotes more sustainable adoption of conservation practices. KEYWORDS Adoption; conservation practices; disadoption; logistic regression; machine learning JEL CLASSIFICATION Q18; Q20; Q28 I. Introduction Conservation practice adoption in agriculture is a widely discussed topic, with most literature focusing on initial adoption rather than continued use (Claassen et al. 2019) . However, it is crucial to recognize that adoption decisions are not always permanent, and users may not always maintain the technology or practice due to various factors, such as liquidity and labour constraints, or unsatisfactory yield response (Razafimahatratra et al. 2021) . Understanding the circumstances related to discontinuing or alternating adoption behaviours could help with policy-making to promote more widespread long-term adoption, particularly for practices like cover crops that provide both private benefits to farmers and public benefits (Mitchell et al. 2017; Wulanningtyas et al. 2021) . Despite the importance of post-initial adoption behaviour, research related to conservation practice disadoption remains limited. For the few studies that examine the adoption and disadoption of conservation practices, the primary data analysis methods are traditional econometric methods, such as logistic regression and bivariate probit models (Dunn et al. 2016; Neill and Lee 2001). However, these methods often struggle to fully capture nonlinear relationships and the complexity of factors influencing farmers’ decisions due to their reliance on strict assumptions. Machine learning (ML) models have become extensively used in many areas, from physics to genetics to social sciences. These models identify patterns and relationships through iterative processes and adjusting internal parameters to minimize prediction error, making them well-suited for capturing nonlinear relationships and modelling interactions that traditio

FAQs

What conservation practice and farmer data does the study focus on?

The study focuses on cover crops and uses a unique longitudinal panel survey of Iowa farmers to analyze both adoption and disadoption behavior over time.

How do logistic regression and Random Forest differ in this analysis?

Logistic regression provides interpretability grounded in economic theory, while Random Forest delivers superior predictive power and captures complex, non-linear relationships among decision factors.

What does SHAP contribute to understanding disadoption drivers?

SHAP identifies the main drivers of disadoption, highlighting factors such as adoption scale, past adoption behavior, and environmental factors.

Aug 2, 2026

Upload by Riley