Research Paper: Dine in or Take out? Trends on Restaurant Service Demand amid the COVID-19 Pandemic

Hi everybody,

I am happy to share a recently finalized research that applies SafeGraph foot traffic data from the Washington metropolitan area to understand the evolving trends of restaurant service demand through the COVID-19 pandemic. The preprint of our paper can be found at Dine in or Take out? Trends on Restaurant Service Demand amid the COVID-19 Pandemic by Linxuan Shi, Zhengtian Xu :: SSRN.

The outbreak of the COVID-19 pandemic has caused unprecedented damage to restaurant businesses, especially for indoor dining services, due to the widespread fear of coronavirus exposure. In contrast, the online food ordering and delivery services, led by DoorDash, Grubhub, and Uber Eats, filled in the vacancy and achieved explosive growth. This research sheds light on both the magnitudinal and structural changes in restaurant service demand.

We first analyzed the aggregate foot traffic volumes to reveal the disruptions to restaurant services across the different stages of the pandemic. Then, a probabilistic learning model was proposed to decompose the aggregate foot traffic by service modes into those for dine-in and takeout, respectively. The transitions in demand structures were identified for restaurants of various service types, price levels, and locations. (Note: Takeout traffic in this paper considers those visits relevant to customers who order online or offline and then either choose to take out themselves or have someone else pick up meals on their behalf.)


  1. In general, our results evidence that the overall restaurant demand still drifted around half of the pre-pandemic level in the Washington metropolitan area by June 2021, far from a complete recovery, one year after the mandatory lockdown was ended. Meanwhile, given their comparative advantages in takeout channels, limited-service and budget restaurants were hit less severely than full-service counterparts. For the urban-rural division, restaurants in exurban/rural areas top the race in recovery, followed by those in suburban and urban areas.
  2. Before the pandemic, the percentage of takeout traffic remained fairly stable around 50% to 60% of the total traffic, as opposed to the growing shares for food delivery services. The takeout portion has converged to a slightly higher level at 65% during the pandemic. With the retreat of epidemic severity, the trend suggests that customers’ preference regarding dine-in and takeout was not fundamentally overturned.
  3. The separate trends by restaurant types show dichotomous paths of recovery in demand structure. After the termination of mandatory lockdown, the percentage of takeout visits at full-service restaurants gradually trended downward toward the pre-pandemic level. In contrast, limited-service restaurants and drinking bars observe a mild upward trend in the share of takeout services, possibly due to the popularization of online food ordering and delivery.
  4. The comparative changes of the two service modes (dine-in versus takeout) exhibit conspicuous heterogeneity among different regions along the path of recovery. In the exurban areas, the demand for dine-in services hesitated around half of the pre-pandemic level, whereas the takeout traffic almost returned fully back to normal. But in the urban areas, there lacks a clear sign of either mode advancing faster in the recovery.

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

Congratulations @Zhengtian_Xu_George_Washington_University and @Linxuan_Shi! This is fantastic - so excited to see this out! :tada:

And great timing as some Community members were recently asking about any research using SafeGraph to look at visits to restaurants.

Going to resurrect this previous conversation from general-discussion a few weeks back:

Hey @Alexander_Audet_American_Enterprise_Institute - have you seen this research by @Zhengtian_Xu_George_Washington_University and @Linxuan_Shi? I remember you were interested in other publications using SafeGraph to look at foot traffic to restaurants. Be sure to look at this paper! You might get Linxuan and Zhengtian up-to-speed on what your work is currently looking.

Hey @Shooshan_Danagoulian_Wayne_State_University - not sure if your graduate student is on our Slack, but flagging this for your team. I remember you’re also looking at foot traffic to restaurants! Have you checked this publication out?

Hey @Irwin_Mier_San_Diego_State_University! Also going to loop you into this conversation since you were asking about something similar previously! Do you have any questions for Zhengtian and Linxuan?

@Zhengtian_Xu_George_Washington_University and @Linxuan_Shi - you touched on it slightly towards the end of the publication. However, in an ideal world, have you thought about an arrangement where all parties, including OFD company, restaurant owners, etc, all win? Curious to hear your thoughts now that you have a better understanding of the foot traffic and revenue that these different restaurants generate.

Thanks, @Niki_Kaz ! We have not seen this, and definitely will take a look.

@Niki_Kaz This is a really great question. The arrangement of a win-win solution will largely depend on the profit margin of restaurants joining the third-party delivery platform, which will further associate with how the demand transforms. I believe most of the existing empirical efforts cautioned that OFD services do not bring in many new sales, but significantly substitute the existing sales that would occur offline, especially in urban areas. So simply bringing the offline menu to the virtual market seems to be unsustainable (financially) for restaurants, maybe they should restructure their products to somehow differentiate between online and offline channels. But this aspect perhaps overlooked the fierce competition among restaurants. If we consider the possible migration of customers between different restaurant establishments, restaurants will forcibly enrich the online menu to match those in offline channels to not fall behind. We haven’t looked deeply into the potential policies, but it is for sure a challenging one considering the slim-profit margin restaurants currently hold.

Currently, it is restaurants that absorb most of the commissions to OFD platforms (but we read that some platforms have already started charging customers commissions or platform fees). In the future development, not only should OFD platforms lower commission rates, but OFD customers may need to “absorb” a portion of the commissions as well, i.e., there will commonly be price differentiations of the same product between online and offline channels. This is an interesting research topic to look into.