Essays on the Digital Transformation of Retail Grocery Industry

Abstract

This dissertation looks at the profound impact of digital transformation in the grocery retail industry. It specifically focuses on the competitive repercussions of Amazon's foray into the grocery market, the effects of its Fulfillment Centers (FCs) on local labor markets, and the integration of image and text information from digital sellers into demand estimation. Spanning three interconnected chapters, the research presents a comprehensive examination of the ongoing shifts revolutionizing the retail landscape. Chapter 1 probes the entry of Amazon Fresh and its ensuing effects on local grocery stores. An intricate analysis of quarterly grocery scanner data, combined with Amazon Fresh's entry information, reveals intriguing dynamics of retail competition. Interestingly, Amazon Fresh's debut prompts a significant negative volume response and a slight positive price response from incumbent grocery stores, specifically within the juice products category. The competition extends beyond mere volume and price adjustments, revealing that incumbent stores expand their product assortments and enhance service quality to differentiate from digital retailers. This nuanced understanding of the competitive entry effect sheds light on the strategic responses of traditional brick-and-mortar stores to digital competition in a rapidly evolving marketplace. Chapter 2 broadens the scope to the county level, examining the impacts of Amazon FCs on local labor markets. Employing robust econometric methods, this chapter unveils that while FCs stimulate higher wages, they also seemingly contribute to a reduction in the overall retail workforce. Additionally, the analysis notes a decrease in juice sales volume and inconsistent price changes across the county. These findings hint at a complex interplay between labor market changes, retail market adjustments, and the ongoing digital transformation. Chapter 3 offers a comprehensive exploration of the integration of deep learning-derived features into traditional demand estimation methodologies. A comparison of econometric models with machine learning models using juice product data illuminates the efficacy of different techniques in capturing diverse product characteristics. The study further explores advanced methods like double machine learning and convolutional neural networks to address challenges inherent in high-dimensional and sparse data. Notably, we successfully incorporate deep learning into traditional demand estimation, demonstrating a substantial improvement in model performance, particularly in neural networks. This work not only enhances the representation of product features but also bolsters the models' capacity to accurately capture underlying demand dynamics.