Machine Learning Applications for Agricultural Economics

Kennedy Rodgers Odongo

doi:10.7273/000005142

This dissertation utilizes machine learning to answer questions in agricultural economics in three related but independent essays. Machine learning and data science are increasingly being adopted in interdisciplinary work providing complimentary analytical methods and data tools for economics research. I use machine learning to investigate how COVID-19 and the resulting media coverage affected specialty crop markets dynamics and to develop insights how attributes of a new apple variety can be utilized in an advertising campaign to derive demand. The first paper of my dissertation investigates how COVID-19 and related social and traditional media coverage affected shipping point prices of specialty crops. I use Twitter data to estimate how the prevalence COVID-19 topics affect crop demand. The results show that crops that are usually consumed as food away from home (FAFH) were the most affected by COVID-19 relative to crops usually consumed as food at home (FAH). The impact of the pandemic was heterogenous across specialty crops with crops whose usage is concentrated in FAFH settings experiencing a decrease in demand compared to crops used mostly in FAH settings. The second compares the performance of two time series forecasting techniques in the context of event studies. The event in this paper is the economy-wide COVID-19 shutdown. The results show that the prices in strawberry and apples markets were higher during the pandemic than they should have been. In comparing the two forecasting methods, the neural network outperforms the ARIMA on error metrics such as the Mean Absolute Error. The third paper evaluates how attributes for newly developed WA 38 apple sold under the Cosmic Crisp brand can be used for to accelerate demand. I identify the market segments where marketing is effective and identify the attributes of the brand that most appeal to consumers. The results show that sentiment on Cosmic Crisp brand is positive with an overall compound score of 0.2263. The online conversation on the brand revolves around the history and novelty of the variety, farm tours to drive demand, the taste and appearance of the apple and, its affiliation to the university where it was developed.

Machine Learning Applications for Agricultural Economics

Files and links (1)

Abstract

Metrics

Details