UC Davis

The COVID-19 pandemic rapidly accelerated the adoption of online shopping in early 2020. This increase raises questions about whether there are changes in e-shopping behavior, the characteristics of online shoppers, and the frequency of online purchases by different types of delivery services during that period. The e-shopping patterns during the COVID-19 pandemic also bring into question whether everyone has the accessibility to engage in online shopping and use various delivery services. This study addresses the above questions through exploratory data analyses, the estimation of binomial logistic regression models, and market segmentation using the Latent Class Cluster Analysis (LCCA) on the Spring 2020 COVID-19 Mobility Study survey data. The results show that age, educational background, type of neighborhood where the respondents live, household income, and attitudes toward technology strongly influence e- shopping behavior. The LCCA results revealed three well-defined latent classes: 1) Occasional Shoppers, who shopped not very frequently, generally used both fast and standard delivery services more than the other delivery options and often live in suburban areas, 2) Non-Shoppers, who made very few or no online purchases at all, tend to belong to the older age group, have lower education and income level, and have negative attitudes toward technology, and 3) Super Shoppers, who made more frequent purchases (usually three or more purchases in a month with any type of delivery services) tend to be younger and wealthier, live in urban areas, and usually have positive attitudes toward the adoption of technology. The results indicate that e-shopping and some delivery services during the early phase of the COVID-19 pandemic might only be available and benefit some groups, as people who had lower digital literacy and did not live in urban areas had less access to those services.

Keywords: E-shopping, COVID-19, Shopping Behavior, Delivery Services, Exploratory Data Analysis, Logistic Regression, Latent Class Cluster Analysis