Research Paper: The impact of COVID-19 on trips to urban amenities: Examining travel behavior changes in Somerville, MA

Research
,
1

Niki Kazahaya (SafeGraph) : :tada: Big shoutout to some our Community members on the recent publication - just saw the news! Congratulations , @Anne_Hudson_MIT , @Dylan_Halpern , Rounaq Basu, @Kloe_MIT , and Jorrit de Jong!

Be sure to check out their paper, The impact of COVID-19 on trips to urban amenities: Examining travel behavior changes in Somerville, MA, here: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0252794

This topic was automatically generated from Slack. You can find the original thread here.

For Community members that want to ask questions to the authors, feel free to respond to this thread!

2

Thanks for sharing this paper! I think this is a really interesting application of the SafeGraph data to examine stated vs. revealed consumer preference. I really enjoyed reading it and have a few questions if you don’t mind.

  1. You limit your analysis to clustered establishments due to limitations in mobile GPS accuracy. Is there reason to believe that establishments within these clusters are systematically different from non-clustered establishments? Some businesses, such as car dealerships, are necessarily more dispersed than others (i.e., coffee shops).
  2. Similarly, I notice that the number of observations in tables 2 & 3 varies considerably by month. My interpretation of this is that some trip data between CBGs and clusters was not available for that month, potentially due to few trips. This seems somewhat concerning as those excluded establishment clusters would not be included in your primary analysis. Have you tried limiting the analysis to CBGs/clusters that are observed in every time period? Are the results robust to doing this?
  3. It was interesting to read about the construction of a synthetic median cluster for each CBG. I understand how this addresses the reference category issue, but wonder if you considered a multinomial logit model?
  4. I’m not sure that land values are a valid proxy for the expensiveness of the cluster. Some large-volume, low-cost retailers might be located on valuable land but still offer a cheaper product than competitors. This seems like it may be an important factor as individuals behavior could have shifted as a result of the pandemic-induced recession (i.e., choosing to shop at a national supermarket vs. a local market). Have you used other proxies for expensiveness?
    Again, thank you for sharing this excellent work. This paper and the methodology it uses have many opportunities for further research into consumer behavior during the pandemic. It will be interesting to see how widespread vaccination (and thus an increase sense of personal safety) may have shaped consumer preference following March 2021.