The Reverse Side of Online Shopping: Examining Sociodemographic and Built-Environment Determinants of Delivery Returns

Term Start:

June 1, 2024

Term End:

May 31, 2025

Budget:

$150,000

Keywords:

E-Commerce Growth, Online Purchases, Vehicle Miles Traveled

Thrust Area(s):

Data Modeling and Analytic Tools, Equity and Understanding User Needs

University Lead:

The University of Texas at Austin

Researcher(s):

Chandra Bhat

E-commerce growth has transformed retail, offering unparalleled convenience but causing a surge in product delivery returns. Industry reports show 30% of online purchases are returned, compared to 9% for brick-and-mortar stores, resulting in an $817 billion financial burden in 2022, with online retail accounting for a quarter. The impacts extend beyond finances, straining logistics and transportation networks, increasing vehicle miles traveled, emissions, and urban truck traffic. Retailers may need to expand urban logistics infrastructure, but strategic planning is required to mitigate pressures on congested areas. Despite these implications, delivery returns remain under-researched, with studies focusing on product details or retailer policies rather than consumer-level factors driving return decisions, particularly in the U.S. market. This study aims to bridge this gap by examining how sociodemographic factors (e.g., age, gender, income) and built-environment characteristics (e.g., urbanity, access to return options) influence the frequency and channel choice for returning online purchases. We analyze responses from the NHTS 2022 survey regarding the frequency of returns through four channels: home pickup, post office/UPS/FedEx, Amazon drop-off center, and direct store returns.


The approach contributes to existing knowledge in four ways. First, we use a joint multivariate ordinal-response modeling approach, which is suited for this analysis as the data indicates that reported frequencies of returns per 30 days are generally less than 4 times for each channel. Second, through joint modeling, we can control for unobserved factors that lead to associations among the counts of the four return channels. For example, if an individual values convenience, they may be predisposed to selecting home pickup while also prioritizing returns at local mail carriers but be less likely to make dedicated trips to Amazon drop-off centers. Ignoring this correlation could lead to inaccurate estimates and inappropriate policy implications. Third, in addition to individual and household characteristics, we include zip code-level built-environment variables (including the density of residential location, land-use mix, number of mail carriers, and number of Amazon drop-off centers). Such built-environment factors reflect accessibility conditions that are important determinants of return choices. It is important to note that while factors such as return policies and shipping costs are critical in return decisions, the study design precludes their inclusion because we are analyzing aggregate counts of returns over a 30-day period rather than examining each return occasion individually. While we acknowledge the significant impact these factors have on return behavior, the methodology does not allow for their incorporation. Future research could potentially address this limitation. Fourth, we estimate the magnitude effects of variables that quantify how much increasing or decreasing a variable would actually affect the delivery return frequency by channel. This analysis allows policymakers and industry stakeholders to quantify the potential impact of interventions or changes targeting specific variables

Project scope

Scroll to Top