Shop ‘til you drop (off your kids at school): accurate representations of trip chaining in shopping travel simulation models 

Abstract
As e-commerce becomes more prevalent, many researchers and practitioners are asking whether delivery services have greater environmental impacts than the personal vehicle trips to local stores they may replace. An important input to this question is how much shopping travel is happening now. In the past, researchers have primarily answered this question in one of two ways. Some researchers have assumed that each shopping trip represents a round trip from home to the store and back. Others have used travel survey data to estimate shopping travel by summing the vehicle miles for trips where the respondents reported the purpose of visiting their destination was to shop.

Both of these methods give inaccurate results, because shopping trips are often chained—either with other shopping trips, or with other activities. We estimate the vehicle miles traveled (VMT) generated by both methods, and compare them to VMT estimates based on the marginal distance people travel for shopping—i.e. only the distance that they would not have traveled if it were not for shopping. For a stop on the way home from work, for instance, this may be minimal.

We use the geocoded version of both the 2009 and the 2017 National Household Travel Survey California add-on sample for this analysis, and compute shopping travel using three methodologies, all calculated using the same shortest path algorithm. Our favored method finds that shopping generated an average of 4.0 (2009) and 3.6 (2017) marginal vehicle miles traveled per household per day. If we assumed that every shopping event generated a round trip from home to the store, each household would generate approximately double the more accurate estimate of marginal vehicle travel generated by shopping. If we instead summed all trips with a reported trip purpose of shopping, we would arrive at an estimate of average shopping travel that is 20-30% lower than the more accurate estimate.

This analysis shows that properly calculating the VMT generated by in-person shopping is critical to analyses of the transport and environmental impacts of online shopping. Studies that assume a single round trip from home for each shopping trip drastically overstate the impacts of in-store shopping, making online shopping appear relatively more advantageous than it actually is. Studies that simply sum the mileage generated by trips to shopping destinations understate the actual mileage generated by shopping, making online shopping appear relatively less advantageous.