COVID19 and restaurant’s closure in Pittsburgh

The coronavirus 2019 has disrupted people’s livelihoods and economic growth around the globe. Dinning industries, like many other industries, has received a drastic economics impacts due to coronavirus pandemic, Work-From-Home (WFH) policy, as well as the COVID-19 lockdowns. Normal activity such as eating out or social gathering was not possible for a significant period, resulting in loss of revenue for many restaurants and cafes.

Pittsburgh, like many other cities in the US, have seen restaurant closure due to this reason. Pamela’s Diner in Squirrel Hill, for instance, has been confirmed to be closed permanently on Dec 28, 2021 [1], following by Squirrel Hill Eat'n Park on Jan 4, 2022 [2]. Some places are still operating but with different method, like Roundabout Brewery in Lawrenceville that is not close but now offer only to-go[3].

This GIS analysis aims to provide more insight of how restaurants, bars, or cafes were affects differently based on their spatial location during the frame from 2019 to 2021. The primary target audience of this article is a group of business owners who currently running a restaurant in Pittsburgh, Pennsylvania.

The main datasets for this project are historical business data from Data Axle Reference Solutions (Formerly known as ReferenceUSA). This access was possible via Carnegie Mellon University Library portal. Due to the selected time scope, most of restaurants/café/bar have 3 consecutive rows in the dataset. If there is no record for a certain year, that row will not be found. Some places also change their business’s name result in having multiple rows in the same year. For a case of recently close places like Pamela’s Dinner mentioned above was still counted as active business in 2021 due to the very recent closure (relative to the time of this study being conducted)

The data was merged, cleaned and rearranged with Python3. For more information, please follow the coding on GitHub at

•  GIS-Restaurant_COVID_Walkable_PGH/RestaurantSeperator.pyat main · korawichkavee/GIS-Restaurant_COVID_Walkable_PGH (github.com) 
•  GIS-Restaurant_COVID_Walkable_PGH/PivotYear.ipynbat main · korawichkavee/GIS-Restaurant_COVID_Walkable_PGH (github.com) 

After preprocessing the data, the mapping and geocoding was done within ArcGIS Pro 2.8.

The first stage of this analysis is to explore active restaurants in Pittsburgh. The duration scope of this analysis focus on the year 2019, 2020, and 2021, therefore the following map are placed according to emphasis the changes throughout the timeline.

Based on the 3 maps above, Figure-1 was made to illustrate the top 10 neighborhoods and their changes before and after the pandemic. (Sorted by the 2019 statistics)

Another way to look the changes is to calculate the different. Figure-2 emphasize such different between 2020-2019, 2021-2020, and 2021-2019.

From the number of closures during the pandemic in 2020 & 2021, the next step is to normalize the statistics. Although the number of restaurants can already be meaningful to the topic, understand the statistic of closure relative to the remaining places as a percentage would provide density and level of competitive in each neighborhood. If a restaurant is in a remote area, it would be the only option for consumers in that area there wouldn’t much of a local consumer. In contrast, a high-density commercial zone would be a great catchment of customers, but it wouldn’t be the only restaurant there in the market. To understand how much competition effort (e.g. marketing campaign/uniqueness of food & drink) a business owner have put in the work for each neighborhood, Figure-3 shows the ratio of no. of close business out of restaurants counted in 2019, revealing that although CBD has the highest decreases as discussed earlier, that lost is actually 20% = 40/201.

Crawford-Roberts and South Oakland face a dramatic loss, 3 out of 4 and 2 out of 2 accordingly. The reason can be that these neighborhoods have low populations[5], [6] and geographically surround by other well-known neighborhoods.

Still, the CBD is dramatically impacted during the year of pandemic. This is sensible founds since there is a lower demand for making a trip into downtown (especially driving dur to lockdown and WFH). The pattern would result in people visiting their restaurants in a walkable distance from their household instead of performing eat-out activities in the core of the city. Besides walking, active transport such as biking can be another great way to draw more revenue to the restaurant[4].

In this analysis, historical business datasets are examined and geo-processed to find if there are neighborhoods that have been affected by COVID more critical than other neighborhoods in Pittsburgh. The following statements can be concluded.

• Originally, regardless of the pandemic, many restaurants are located in the CBD (Downtown) area, as well as in suburbs e.g., Strip District, North & Central Oakland, Shadyside and very few in other areas.
• Although there are some new businesses born in the 2020-21, the overall number of restaurants in Pittsburgh is now decreased.
• By the number, CBD is the most affected area. The main reason is likely to be the change in transportation pattern during the pandemic.
• By the percentage, Crawford-Roberts and South Oakland are the most affected area. The main reason is likely to be the lack of customers in the area and demands being pulled to other nearby area.
• In terms of sidewalk network near the closed business, any restaurants within these area must find a way to compete with others, specifically about 10 other places in a sq.mile of walking proximity.

A future work should be able access more in-depth data collection. A survey of the reason of closure and the exact date of closure would be more robust than this project. The collection of how the restaurants is serving food such as dining-in, take home, or delivery via food applications (Uber Eat, Door dash, Grub Hub, in-house delivery) is also important as it is related to the pandemic. A web scrapping from social food-critics platform like Google Map, TripAdvisor, Yelp, etc. can also filter out some places that may have other problems and not just COVID.

1. Cover image was taken by Korawich Kavee at  Kiin Lao & Thai Eatery  (Dec 2021) ( Also available on Google Map )
2. Data of Restaurants was downloaded from the U.S. Historical Business Database by Data axle reference solutions. ( Access via CMU Library license )
3. Shapefile of Neighborhood and other relates GIS feature classes of Pittsburgh are provided within the coursework and  ArcGIS Pro 2.8 tutorial’s material .