This paper (for INFO-616 Programming Interactive Visualizations) analyzes the visualizations presented in the Washington Post article The most common restaurant cuisine in every state, and a chain-restaurant mystery by Andrew Van Dam. The article utilizes data from a project analyzing the geographic frequency of chain restaurants in the US by Clio Andris and Xiaofan Liang of Georgia Tech’s Friendly Cities Lab. In their project, Andris and Liang attempt to understand what factors determine what they term chaininess (a high density of chain restaurants) in certain geographic locations, which includes political party affiliation, car-dependency and walkability.
The Washington Post article uses the same chain restaurant dataset used by Andris and Liang, which was sourced from a marketing data firm called Leads Deposit, to create its own visualizations with the data. The WP article also uses Census data to add additional context on the population of workers who commute by car.
There are 7 visualizations in this article: three bar charts (Fig. 1, 2, 3), three maps (Fig. 4, 5, 6) and one scatterplot (Fig. 7). All but one (Fig. 4) are interactive.
Bar Charts
The bar charts in this article visualize the following: states with the most chain restaurants by percentage of all restaurants (Fig. 1), metro areas with the most chain restaurants by percentage of all restaurants (Fig. 2), and percentage of car commuters per state (Fig. 3).
The same design is used for each bar chart: vertical bars sorted in descending order from largest percentage to smallest, with the states and/or metro areas on the y-axis and omitting the x-axis in favor of numerical labels on the bars themselves. This design choice is effective because it enables the user to read the bar charts vertically, from geographic location to percentage, without having to absorb the moderate amount of information while limiting the number of directions the eyes need to move around.
The color choices of blue (Fig. 1 and 3) and red (Fig. 2) are safe for most users with colorblindness (Fossheim, 2020). However, because the article does discuss how the Andris/Lang chaininess project discovered a correlation between areas with a high density of chain restaurants and areas of a high percentage of Trump voters in the 2020 election, I did have the initial interpretation that the red/blue was encoding for Republican/Democrats. A closer look allowed me to determine that the color choices were unrelated to political affiliation. Additionally, the use of gray in these bar charts is successful in moving the values not being emphasized to the background. This allows the user to focus on what the design is intending to highlight.
The bar charts use three interactive elements, primarily a text box to allow the user search for a location of their choice. Second, there is a sort tool that allows the user to quickly see the highest and lowest values. This sort element does not technically sort from descending to ascending, as you may see in a spreadsheet, but from first page to last. The other element is an arrow button tool to scroll through the bar chart similar to a slideshow, with the values in descending order. This layout breaks down the visualization into smaller, digestible chunks.
Maps
The maps in the Washington Post article visualize the following: metro areas with proportion of restaurants in that are chains (Fig. 4), the most common restaurant cuisine by state (Fig. 5) and the percentage of population per county that commutes by car.
The first map (Fig. 4) appears to be a quite complex dot density map, but is static for reasons that are unclear. This map uses the same red and blue colors as the bar charts, but also uses size and color saturation to encode for size of metro area (dot size) and percentage of chains (color/saturation). I will add that it is my interpretation that the dot size represents the size of the metro area, but that aspect of the design is not described outside of a statement in the caption that narrow strokes were added to make smaller metro areas more visible. This map could be improved using more contextual information that could be included in interactive elements, such as tooltips that appear when hovering over a dot.
The second map (Fig. 5) of the most common restaurant cuisine in each state is interactive and somewhat interesting to see (surprisingly, pizza is not the most popular food in New York). When the user hovers over a state, a tooltip appears with the state name and the most common cuisine. As a design choice, this is successful in limiting the amount of visual clutter that could result in labeling each state. The use of color to encode each type of cuisine is less successful in that it seems like a random color scale that does not show any meaningful representation of cuisine by color (i.e. dark gray for Italian, purple for pizza). The use of green, yellow and red may also make this more difficult to read for people with colorblindness.
The third map (Fig. 6) combines the successful elements of the first two maps discussed above. It employs a clear use of color, aligned with blue and red in the previous visualizations, and a use of contextual tooltips that display county name, population and percentage of car commuters when the user hovers over a county. The map also uses clear stroke lines around each county and state. Using a thicker stroke weight to outline the states and a lighter stroke weight outline counties creates a clear visual hierarchy. Even though there are a lot of data being displayed on this map, it manages not to be visually overwhelming thanks to the interactive elements and use of color and stroke weight.
Scatterplot
The scatterplot in this article (Fig. 7) visualizes the percentage of chain restaurants (x-axis) using the USDA’s rural-urban continuum (y-axis) to represent the type of county population areas from rural to suburb to city.
Color is used to encode which county voted for Trump (red), Biden (blue), and overall (gray). While the concept is interesting, this visualization is a bit more difficult to read than the rest. For example, it is not clear what ‘overall’ represents. There is an interactive element that allows the user to highlight the data points related to Trump, Biden or overall. When selecting one dot in that category, the others fade to the background and the data points are labeled by percentage. Visualizing this data may have been more successful using a map.
Rather than having a strong continuous narrative, the nature of this Washington Post article by Van Dam is exploratory and the visualizations support that. The conclusion, which brings in US Census data on commuters, does work to support the findings of Andris and Liang chaininess project, which is that there are a higher percentage of chain restaurants in areas that are more reliant on cars. Van Dam’s exploration here is enhanced by bringing in the Census data to supplement the findings of the chaininess study
As previously mentioned, it would be interesting to see the static dot density map (Fig. 4) as an interactive map so the user can see the contextual information. Andris and Liang do have an interactive map of each chain restaurant as part of their project. One could see how Van Dam may have intended to use the static map as an extension of Andris/Liang’s map to include population density. It would be interesting to see how the integration of both maps would work.
___________________________________________________________________
Fossheim, S. (2020, May 20). An introduction to accessible data visualizations with d3.js by Sarah L. Fossheim. An introduction to accessible data visualizations with D3.js by. Retrieved October 18, 2022, from https://fossheim.io/writing/posts/accessible-dataviz-d3-intro/
Liang, X., & Andris, C. (2021). Measuring McCities: Landscapes of chain and independent restaurants in the United States. Environment and Planning B: Urban Analytics and City Science, 49(2), 585-602. https://doi.org/10.1177/23998083211014896
Van Dam, A. (2022, October 6). Analysis | The most common restaurant cuisine in every state, and a chain-restaurant mystery. The Washington Post. Retrieved October 18, 2022, from https://www.washingtonpost.com/business/2022/09/29/chain-restaurant-capitals/