Understanding Ethnic Enclaves

According to Burgess’ concentric zone model (illustrated), immigrants typically move directly into the zone in transition before gaining the economic and social capital to move towards the outer edges, forming ethnic enclaves.

While there has been significant Chinese immigration to Canada since the 1850s (with the construction of the Canadian Pacific Railway), the scale and other characteristics have changed over time. Chinese immigration grew exponentially in the 1960s and 1970s with the loosening of immigration laws. Beginning in the 1980s, more and more Sino-Vietnamese immigrants settled in Toronto’s Chinatown, with this diversification reflected through the 1990s and early 2000s, with more Vietnamese owned businesses and restaurants appearing towards the outskirts of Chinatown.

Greektown was established as a major place of settlement for Greek immigrants starting in the 1960s. While by the 1990s much of the Greek settlement had shifted towards the suburbs, many Greek businesses remained. This incongruity is present not only in Greektown, but in some of Toronto’s other ethnic enclaves as well, attributing much of the remaining culturally-specific neighborhood character to an attempt to market the space by Business Improvement Areas (BIAs).

In working with these two neighborhoods, we initially used Toronto’s default neighborhood shapefiles, which were found on the City of Toronto’s Open Data portal. However, The resulting datasets were too general, exhibiting limited variability throughout our selected dataset years. 

We find that the City of Toronto's definition of 'Kensington-Chinatown' neighborhood lacked sufficient consideration of their important retail corridors.

Finding Toronto’s default neighbourhood borders inadequate for our project, we chose to create our own study areas with reference to Toronto’s dissemination areas for the years of 2001, 2006, and 2016, which match our chosen census years (to be elaborated upon later).

In creating our own study area, we looked to focus in and around the neighbourhood’s main intersection at Spadina Avenue and Dundas Street West, while also including nearby residential areas off of the intersection in order to extract needed census data.

As seen in figure, we were able to successfully eliminate most of Kensington Market, which based on site visits, is an area with an identity that differs greatly from the Chinese-influenced Chinatown. In order to capture commercial spots, we aimed to have our borders cover both sides of Dundas and Spadina.

Our second study area is The Danforth, which Toronto divides in half between two neighborhoods, as seen in figure. Although the “Danforth Village Toronto” captures stores on the northern side of Danforth Avenue, the southern side is left unaccounted for.

Furthermore, neither neighbourhood captures the length of the Greektown retail strip, which stretches roughly from Broadview Avenue in the west through to the suburb of Scarborough in the east. 

The DMTI restaurant data were manually coded for the following criteria:

• Culturally-specific to the neighborhood of interest (Y/N)
• Chain (Y/N)
• Cultural Cuisine (Country)

Sources used for manual coding include:

• online resources such as Yelp, Zomato, Google Maps, and Yellow Pages,
• personal knowledge,
• and etymology in restaurant names.

Census data were gathered using the Computing in the Humanities and Social Sciences (CHASS) portal.

Census data were acquired for the years of:

• 2001,
• 2006,
• and 2016.

Census variables acquired includes:

• Median Household Income,
• Average Monthly Shelter Costs for Rented Dwellings,
• Chinese and Greek Ethnic Origin (20% data for 2001 and 2006, 25% for 2016),
• and Language(s) at Home (20% for 2001 and 2006, 25% for 2016).

The hot spot analysis tool in ArcGIS online uses the Getis-Ord Gi* statistic to determine statistically significant hot and cold spots.

We performed cluster analysis on the restaurant data for each year (2002, 2006, and 2018) in both of our neighbourhoods to determine where the culturally specific restaurants were concentrated within the study areas. 

Cluster analysis allows us to gain a clearer picture of the spatial relationships between restaurants and the dissemination areas they reside in within each neighborhood.

ANOVA is a statistical method used to test differences between two or more group means with respect to a single variable.

We performed ANOVA on both the response variables- the percentage of Chinese restaurants in Chinatown and the percentage of Greek restaurants in Danforth, and the independent variables- language spoken at home, average rent, and median income, for both neighbourhoods. 

ANOVA allows us to answer the first research question: how have demographics and restaurant composition changed in Chinatown and Danforth over time?

The output of our hot spot analysis using Gi* reveals differing results for Chinatown throughout the three study years.

In 2002 there was significant clustering of Chinese restaurants (C.I. = 99%) on the stretch of Dundas between Spadina Ave and Beverly St, and a cluster of non-Chinese restaurants (C.I. = 95%) in the northeast portion of our study area (see figure 5.2)

In 2006, the hot spot analysis of the restaurant data found no significant clustering in the neighborhood

In 2018 there was some clustering of Chinese restaurants along the same stretch of Dundas between Spadina Ave and Beverly St as in 2002 (C.I. = 90%) 

In 2002 there was some clustering of Greek restaurants right at the intersection of Danforth Ave and Pape Ave (C.I. = 95%)

In 2006 we saw slightly more intense clustering concentrated slightly further west along Danforth 

Similarly in 2018, Greek restaurants were heavily clustered along Danforth from Hampton Ave. to Pape Ave 

How have the proportion of ethnic restaurants changed in these neighborhoods over time?

Since both p-values are smaller than the significance level 0.05, the percentages of Chinese and Greek restaurants in Chinatown and Danforth, respectively, do not fluctuate on statistically significant levels, over census years. 

How have demographics changed in these neighborhoods over time?

For language spoken at home, since both p-values are greater than the significance value 0.05, we fail to reject the null hypothesis. The proportion of the population speaking Chinese and Greek at home in Chinatown and Danforth, respectively, do not fluctuate over census years on statistically significant levels.

For average rent and median household income, since all p-values are smaller than 0.05, we reject the null hypothesis. Average rent and median income in both Chinatown and Danforth increase significantly over at least two of three census years. 

We fit two identical regression models for Chinatown and Danforth, using Chinatown and Danforth datasets, respectively. The generalized linear mixed regression models are illustrated on the left side of the screen.

Response Variable: Percentage of Chinese/Greek restaurants

Random Effect: Census Year

Predictors: Median Household Income; Average Monthly Rent; Language(s) at Home

The regression models answer our second research question:

How has the number of culturally-specific restaurants changed with census variables?

The results for the Chinatown model show that average rent and language at home are statistically significant in predicting the percentage of Chinese restaurants in Chinatown.

• Given all other variables remain constant, for every 1% increase in rent, we anticipate a 0.63% increase in percentage of Chinese restaurants;
• For every 1% increase in the proportion of population speaking Chinese at home, we anticipate a 1.03% increase in the percentage of Chinese restaurants in the same dissemination area.
• Median household income does not have a significant effect on Chinese restaurants.
• Lastly, census year explains close to 0% of the variation we see in the model.