What can localized unemployment rates within these cities tell us about the spatial distributions of Starbucks and McDonalds?
To help illustrate what the results discussed below are about, shown above are the Unemployment Rate profiles of Toronto and Vancouver organized on choropleth maps with the same ratio scale. Just by looking at the maps you can observe Vancouver’s spatially uniform distribution in its unemployment rate, while Toronto has a slightly more spatially heterogeneous distribution.
In Vancouver, the unemployment rate does not appear to have a strong correlation with McDonald’s or Starbucks locations . Each business has a slightly negative correlation coefficient, meaning when the unemployment rate is higher, then there should be less McDonalds and Starbucks in that Census Tract. More specifically, the spatial density of McDonalds and Starbucks locations should be less. However the p-values for both businesses were roughly 0.3 indicating the correlation coefficients are likely insignificant and the presence of an area with a higher rate of unemployment is not a good method of indicating a higher or lower amount of Starbucks and McDonalds . Similarly, Toronto has two slightly negative correlation coefficients, but instead displays two statistically significant p-values in regards to the unemployment rates. The McDonald’s p-value is 0.00003 and Starbucks’ p-value is 0.00013. Interestingly, the McDonald’s correlation coefficient is a larger negative value than Starbucks’, meaning when an area in Toronto experiences less unemployment, they may have a higher density of Mcdonalds locations in that given area.
To sum it up, if you were to pass by a few McDonald’s or upwards of four or five Starbucks in Toronto in a given area, that indicates you are likely in an, albeit specific, part of the city that experiences proportionally less unemployment. This inherently makes sense as the points representing businesses have a second order effect on the unemployment rate. In contrast with Vancouver, the unemployment rate of a given area is not useful to indicate whether or not that area has more or less Starbucks or Mcdonald’s locations.
What can localized median household income within these cities tell us about the spatial distributions of Starbucks and McDonalds?
Shown above are choropleth maps of household median income, again shown with same ratio scale between cities. Vancouver and Toronto are both generally wealthy cities, but each city still exhibits spatial income inequality as seen above.
Beginning with Toronto, you can see three general clusters of census tracts with very high median household incomes. Notably very few McDonalds are located by these areas, whereas more areas of densely located Starbucks are. This shows in the regression analysis as median household income has a slightly negative correlation coefficient with the spatial density of McDonald’s locations and is statistically significant with a p-value of 0.00016. Thus, in Toronto, when the median household income is higher in and around a census tract, then there is likely a lower amount of McDondald’s in and around that census tract, and furthermore, if there is a McDonald’s then it is likely to be proportionally further from other McDonald’s locaitons. On the flip side, the regression shows Starbucks in Toronto experience a slight positive correlation coefficient with median household income. However, the p-value is significantly higher than the p-value for McDonalds at 0.15, and thus median household income in a given area is not a good indicator of the density of Starbucks in a given area.
On the west coast, Vancouver’s median household income indicates a slight negative correlation coefficient with both McDonald’s and Starbucks. However the p-value for the Starbucks’ correlation coefficient is 0.59. This means that median household income is a horrid indicator of the spatial density of Starbucks in Vancouver. McDonald’s p-value with regards to its median household income correlation coefficient is about 0.06, which is better, yet still not statistically significant. Thus, median household income in Vancouver is a much worse indicator of the spatial densities of McDonalds and Starbucks than in Toronto.
What can localized population densities within these cities tell us about the spatial distributions of Starbucks and McDonalds?
Again, shown above is a choropleth map using the same ratio scale of population, except for the ratio of the most dense region because Toronto has an outlier census tract with a very high population density. Judging from these maps there apperas to be a strong positive correlation as both population density, McDonalds, and Starbucks are all highly concentrated in the downtown regions of the cities.
Unsurprisingly, the regression for both cities and businesses exhibit positive correlation values, all with p-values of 0, implying the strongest possible statistical significance in the coefficients measured. These coefficients notably differ between each business to a similar degree within each city. McDonald’s has a smaller positive correlation value, indicating that as population per square kilometer rises, the presence and density of McDonalds locations rise only slightly. Whereas Starbucks’ correlation coefficient is roughly fivefold that of McDonalds in both cities, which indicates more significant increases in their spatial density with increases in population per square kilometer in an urban area. Thus, we have our smoking gun as both the spatial distributions of McDonalds and Starbucks reflect, to their own extents, the population per square kilometer in these Canadian cities.