Health Reports
Ethnocultural and socioeconomic disparities in exposure to residential greenness within urban Canada

by Lauren Pinault, Tanya Christidis, Toyib Olaniyan and Dan L. Crouse

DOI: https://www.doi.org/10.25318/82-003-x202100500001-eng

Living in a home surrounded by trees, gardens and natural vegetation (i.e., greenspace, or greenness) may confer numerous health benefits. Notably, studies in Canada, the United States and Europe have found inverse associations between residential greenness and all-cause or non-accidental,^{Note 1Note 2Note 3Note 4Note 5} respiratory^{Note 2Note 3} and cardiovascular mortality,^{Note 4} after adjusting for socioeconomic and demographic confounders. Associations with mortality are stronger among certain groups: in Canada, among women and people with higher income and education,^{Note 4} and, in England, among the lowest income group.^{Note 1} In addition to mortality outcomes, a Dutch study found that better self-assessed general health was associated with exposure to greenness within both 1 km and 3 km of a person’s home,^{Note 6} and a Toronto study found better perceived health was associated with higher street tree density.^{Note 7}

Proximity and access to residential greenness may act upon health outcomes in several ways. For example, greener environments may directly promote improvements in psychological health by reducing psychological stress,^{Note 8} severity of distress and poor self-perceived mental health.^{Note 9} In addition, greenness may provide an aesthetic experience, and it may also provide an opportunity to receive ample daylight by encouraging outdoor activity, thereby nurturing circadian rhythms.^{Note 8} Greener environments may also act on health indirectly, by promoting outdoor physical activity through physical spaces for recreation and community activities, thereby improving cardiovascular health and reducing obesity.^{Note 10Note 11Note 12Note 13} However, this benefit may depend on the type of greenspace available, as one study in Montréal and the city of Québec noted that cardiovascular benefits from greenspace were observed only if sports and recreational facilities were also present.^{Note 14}

Despite the benefits of living in a greener environment, these environments are not evenly dispersed across the landscape. Several studies have highlighted inequalities in the physical area, vegetation density (e.g., local Normalized Difference Vegetation Index [NDVI] value), quality and characteristics (e.g., manicured garden vs. open field), and accessibility of residential greenspace. One study of the three largest cities in Canada found an association between family income and vegetation cover.^{Note 15} In England, greater greenness was observed in the least economically deprived regions,^{Note 1} and another study noted that usable greenspace was least accessible for those of Indian ancestry.^{Note 16} In urban areas of the United States, greenness values were associated with higher concentrations of White residents,^{Note 17} and lower greenness values were associated with higher concentrations of low-income, Black and Hispanic populations.^{Note 18Note 19}

Environmental justice is a broad field of study that focuses on describing exposures that disparately affect specific socioeconomic and ethnocultural groups.^{Note 20Note 21} Inequalities in access to greenness have not been well documented in Canada. One study in Montréal found results similar to those of international studies, reporting that city blocks with higher proportions of low-income residents and, to a lesser extent, populations designated as visible minorities had lower greenness values and access.^{Note 22} An earlier study that focused on respondents to the 2001 Census of Canada^{Note 4} reported that more affluent and more highly educated adults living in the 30 largest census metropolitan areas of the country had greater exposure to residential greenness than those who were less affluent and less well educated. This study builds on that work by using data from the more recent 2016 Census, including respondents of all ages, and by considering differences in exposure according not only to age, education and income, but also to immigrant status, time since immigration, self-reported ethnicity and neighbourhood deprivation indexes.

Data and methods

Study cohort

Exposure to greenness by demographic and socioeconomic characteristics was explored using the 2016 Census of Canada long-form questionnaire. The long-form questionnaire was distributed to about one in four households in private dwellings, excluding collective dwellings.^{Note 23} It had a response rate of 96.7%.^{Note 23} Census weights were used to derive estimates that were reweighted to more approximately represent the full population distribution of Canada. Individual and household variables extracted for the study included respondent age, Indigenous identity, self-identified population group (White, South Asian, Chinese, Black, Filipino, Latin American, Arab, Southeast Asian, West Asian, Korean, Japanese, or other or multiple identities), immigrant status and year of immigration to Canada, educational attainment (for those aged 25 or older), labour force status (for those aged 25 or older­), household income, housing tenure (owner vs. tenant), and low-income status. Regarding self-identified population groups, the authors acknowledge the limitations of these disaggregated categories and are relying on available census data. The census defines visible minority based on the Employment Equity Act, as "persons, other than Aboriginal peoples, who are non-Caucasian in race or non-white in colour". The low-income status threshold is defined as half the median of the adjusted after-tax income, multiplied by the square root of household size.^{Note 24}

Geographic location of residence was estimated using the six-character postal code reported on the census questionnaire. The Postal Code Conversion File Plus (PCCF+) version 7C was used to determine an approximate latitude and longitude, as well as census geography characteristics. The PCCF+ is most accurate for geocoding within urban centres, with representative points assigned a median of 100 m to 200 m from the actual home, usually along the block face of the same street.^{Note 25}

In addition to individual and household characteristics, neighbourhood deprivation indexes were assigned by attaching the Canadian Index of Multiple Deprivation (CIMD) based on the 2016 Census to the individuals in the dataset. The CIMD considers four dimensions of deprivation: residential instability, economic dependency, ethnocultural composition and situational vulnerability. It is derived at the dissemination area (DA) level, which is a small, relatively stable area of census geography composed of about 400 to 700 people.^{Note 26}

All of urban Canada was included in this study. To focus on the urban population of Canada, people who did not reside in census metropolitan areas (CMAs) or census agglomerations (CAs) were excluded from analyses. CMAs are defined as having a total population of at least 100,000 people, of whom at least half must live in the core, while CAs must have a core population of at least 10,000 people.^{Note 27} Urban form was further considered to distinguish those who may reside within the rural portions of CMAs or CAs.^{Note 28} Those living in a rural or exurban urban census tract (i.e., fewer than 150 people per km²) were also excluded from analyses. Approximately 3.0 million respondents were excluded because they lived outside a CMA or CA or in an area with low population density. This left a total study population of 5,306,800 respondents, representing 21,925,200 people after applying census weights (counts are rounded for confidentiality).

Exposure assessment

Greenness estimates were derived from the remotely sensed NDVI, which is an index of vegetation ranging from -1 (water) to 0 (bare soil) to 1 (thick vegetation). NDVI values between 0.01 and 1.00 were used to estimate the residential mean NDVI within a 500 m buffer around a representative point of the respondent’s postal code. Other buffer sizes were considered in preliminary analyses, with similar results (results not shown). Annual maximum greenness values were taken from images at 30 m resolution from the United States Geological Survey’s Landsat satellites during the growing season for vegetation (May 1 to August 31).^{Note 4Note 29} Data were derived from the most recent year available (2012 or an older year if 2012 was missing). Of the original 8.65 million respondents to the census long-form questionnaire, 253,700 were excluded because they were not successfully linked to greenness estimates. Further context for the concept of greenness is provided in Figure 1, which describes the four quartiles of greenness estimates within urban Canada and provides examples.

Statistical analyses

Descriptive statistics were used to determine differences in average exposure to greenness by selected individual and neighbourhood characteristics, using census weighting. To compare differences in greenness values among groups, Student’s t-tests were used for significance testing, and Cohen’s d effect size was used to interpret the magnitude of the difference between groups, since it provides a measure independent of sample size.^{Note 30} In addition to the full population across urban Canada, the same analyses were conducted within the three largest cities in Canada: Toronto, Montréal and Vancouver. Effect size estimates can be interpreted as small (d = 0.2), medium (d = 0.5) or large (d = 0.8). Analyses were conducted in SAS 9.4 software.

Results

Descriptive statistics of the study population are provided in Table 1. Approximately 31% of the 5.3 million respondents were immigrants, 34% self-identified as Indigenous or as belonging to a group designated as a visible minority, 31% rented their homes, and 15% were identified as being of low-income status. Across urban Canada, the mean residential greenness value within 500 m of the home was 0.44 (standard deviation [SD] = 0.18). Descriptively, greater greenness was observed for children younger than 5 or older adults, White populations, non-immigrants, homeowners, higher income quintiles, and people not of low-income status. At the neighbourhood level, residential greenness was incrementally greater among DAs with increasingly lower levels of residential instability, lower ethnocultural composition and lower situational vulnerability.

When the analysis focused on the three largest cities in Canada, similar trends were observed between individual-level covariates and exposure to greenness (Table 2). However, the magnitude of inequalities was greater in Montréal than in Toronto or Vancouver, with residents reporting White or Indigenous ancestry having mean greenness values 0.05 higher than residents reporting Arab or Latin American ancestry (p < 0.001; d = 0.25), and non-immigrants having mean values 0.04 higher than immigrants (p < 0.001; d = 0.23). Notably, homeowners in Montréal reported mean values 0.08 higher than tenants (p < 0.001; d = 0.44), and, similarly, the mean NDVI of the highest income quintile was 0.08 higher than that of the lowest (p < 0.001; d = 0.41).

Increases across deciles of median household income were associated with marginal but nearly consistent increases in exposure to residential greenness across all age categories examined (Figure 2). However, greenness was lower for those aged 18 to 39 across all income deciles, except the highest. Outside the highest and lowest deciles, seniors had the highest residential greenness compared with those in other age groups.

A positive relationship was observed between median household income and residential greenness in all ethnocultural groups examined (Figure 3). However, across all income deciles, White populations had markedly higher greenness than any other population group. Indigenous populations also had greater greenness than every other group, except among those in the lowest household income decile. People of Filipino ancestry had lower greenness than any other population group, across all income deciles. The lowest exposure to greenness was among the lowest-income Filipino group (mean = 0.40; SD = 0.16), and the highest exposure to greenness was among the highest-income White group (mean = 0.48; SD = 0.18; difference: p < 0.001; d = 0.46).

Figure 4 examines population groups by immigrant status and period of immigration to Canada. For most population groups who were also immigrants (except those of Chinese ancestry), residential greenness increased with longer residence in Canada, sometimes reaching and surpassing the estimated exposure of Canadian-born populations after a few decades. For Chinese immigrants, by contrast, greenness was greatest for those who had recently immigrated to Canada and was lowest among those who had been in Canada longer than 50 years. Non-immigrant White populations had higher greenness values (mean = 0.46; SD = 0.18) than any immigrant group, including White immigrants. The lowest greenness values were observed among people of Filipino ancestry who had immigrated to Canada in the past 10 years (mean = 0.40; SD = 0.14; difference between highest and lowest groups: p < 0.001; d = 0.35).

The association between homeownership, immigrant status and greenness was examined across household income deciles in Figure 5. Non-immigrant homeowners had greater greenness than any other group, at all income levels. Among homeowners who were not recent immigrants, greenness rose with increasing income. Tenants at all income levels had lower greenness, especially among recent immigrants. Among tenants and recent immigrant homeowners, there was no clear association between income and greenness.

Discussion

Residential greenness was distributed unequally among the 5.3 million urban residents of Canada included in this study, with inequalities observed across dimensions of income, immigrant status, ethnocultural identity, housing tenure and neighbourhood socioeconomic characteristics. In general, lower residential greenness was observed for people with lower income, immigrants (especially recent immigrants), young adults, populations designated as visible minorities (especially those of Filipino ancestry) and tenants. These trends were observed in the three largest cities in Canada to a greater or lesser degree. Particular strengths of this study were the large population sample (i.e., the weighted census) and the use of a wider variety of variables that cross social, ethnocultural and economic dimensions, as well as neighbourhood-level covariates.

Particularly high or low greenness values were observed for some combinations of these population characteristics. For example, some of the lowest greenness was observed for people of Filipino ancestry who were either recent immigrants to Canada (in the country for less than 10 years) or in the lowest income decile. Low-income young adults (aged 18 to 39) and tenants who were recent immigrants also had notably low greenness around their place of residence. Conversely, greater greenness was observed among non-immigrant homeowners, White non-immigrants and White populations with higher income. In some of these comparisons, Cohen’s d (which measures the effect size for the difference between the mean NDVI) was estimated between 0.40 and 0.48, indicating a nearly “medium” effect size. The finding that immigrants had lower greenness levels was consistent with those of a study in purban Montréal, which found a negative association between immigrants and street trees,^{Note 31} and a study across the United States, where immigrant status was negatively correlated with NDVI greenness.^{Note 32}

Generally, these findings were consistent with those of previous studies in Canada and elsewhere, which observed greater greenness among more socially and economically advantaged populations. Most city-scale studies, particularly in the United States, have noted that greenspace, parks and tree cover are more likely to be concentrated near populations with higher income and less likely to be located in regions with lower-income populations, tenants and specific groups designated as visible minorities.^{Note 8Note 33Note 34Note 35} Similarly, a European review on promoting environmental justice through urban greenspace access indicated that walkable and accessible green infrastructure was frequently clustered in wealthy neighbourhoods.^{Note 10}

The observed inequalities in greenness values are attributable to many historical, sociocultural, economic and demographic factors. While an examination of the underlying causes of these inequalities was beyond the scope of the current study, the role of housing availability, historical settlement patterns and housing policies, as well as the distribution and maintenance of rental housing stock, merit further study, particularly in Canada, where such studies are lacking. A study in Fredericton, Halifax and Winnipeg found that different populations, namely older people and women, rate the importance of urban forests more highly, and this may influence their housing choices.^{Note 36}

The finding that people of Filipino ancestry had the least exposure to residential greenness may be explained, in part, by high concentrations of this population in large cities (namely Toronto, Vancouver, Montréal and Winnipeg) and more recent waves of immigration than for many other groups, beginning in the 1960s,^{Note 37} with larger waves in the 1990s and 2000s.^{Note 38} In Montréal and Winnipeg, Filipino immigrants have tended to cluster in specific neighbourhoods that are frequently characterized by low-rise apartments and less greenspace.^{Note 38} An analysis of settlement patterns in Toronto found that Filipino immigrants favoured affordable rental housing and mixed-used neighbourhoods, selecting areas with accessible public transit or workplace proximity,^{Note 38} all of which may correspond to lower residential greenness. However, when examined at the individual city level (Table 2), people of Filipino ancestry did not have markedly lower greenness than some of the other groups designated as visible minorities considered in this study.

Limitations

This study has several limitations, mainly associated with the use of the NDVI to approximate greenness and the cross-sectional nature of the study. This study did not assess the quality or accessibility of greenspace, only the average NDVI within a buffer around the home. No distinction was made between greenness resulting from publicly accessible parks and greenness from private, inaccessible spaces, such as golf courses or agricultural land. However, limiting the study area to suburban or core areas of the city likely removed most areas that would have included agricultural lands. Maintenance of greenspace is known to vary across income and demographic gradients in cities,^{Note 10} and examining the overall quantity of greenness may overestimate any real health benefits from greenspace. For example, in Baltimore and Los Angeles, overall park acreage was greater in predominantly Black or Latino neighbourhoods, but these parks had lower levels of maintenance and fewer sports facilities, and the acreage was overall less accessible to the population.^{Note 39Note 40} Similarly, a study in Montréal showed that overall availability of park spaces did not differ by neighbourhood socioeconomic status, but the parks in higher-income neighbourhoods had more recreational amenities and were better maintained.^{Note 41} Alternative greenness measures used in smaller-scale studies include tree density^{Note 7} and proportion of city blocks, streets or residential yards covered by trees and shrubs,^{Note 22Note 31} although these measures are not yet available nationally.

Greenness measures alone may not represent a benefit, as greenness may sometimes indicate areas of foreclosure or abandonment,^{Note 17} as well as reflect poor maintenance of nuisance trees, as was the case in Baltimore along property lines.^{Note 40} Other possible detrimental attributes of greater greenness may be higher pollen counts, possible pest habitats,^{Note 10} and real or perceived crime and insecurity in forested areas.^{Note 11} Future development of geographic datasets for these purposes, such as maps of accessible parks and recreational areas, as well as park quality indexes, may assist researchers in further exploring inequalities in access to these neighbourhood benefits.

Some spatial error is inherent in assigning geographic coordinates based on postal codes. Point estimates of residential location from the PCCF+ program are estimated as having a median spatial error of 160 m for ordinary urban households and 110 m for large urban apartment buildings,^{Note 25} which are the most typical postal code types within this study. However, error of this magnitude is unlikely to affect NDVI estimates, as these were calculated as a mean within a larger-sized buffer, representing values across the neighbourhood. Furthermore, only a single spatial scale (500 m) was included in this study, although it is possible that other elements of greenspace, such as park access or acreage of green areas, may be better captured at a different scale.^{Note 40} However, previous research on the health effects of greenness found similar effects using different-sized buffers (i.e., 250 m, 500 m and 1 km buffers).^{Note 14}

Changes in exposure to greenness over time may be an interesting avenue for future study, as historical greenness and census data could be used to provide a snapshot of how inequalities have evolved over time.^{Note 17} Studying changes over time in exposure to greenness could highlight another issue: the urban greenspace paradox.^{Note 42Note 43} This paradox is that the action of greening a neighbourhood may lead to unintended consequences, where property values may increase, leading to gentrification and driving out groups that were meant to benefit from the original greening efforts. As a result, efforts to improve greenness equity must be conducted in a manner that ensures that other neighbourhood characteristics are preserved.

Conclusion

In Canadian urban areas, residential greenness (the NDVI) was unequally distributed across socioeconomic and population groups. Greenness estimates were lower around the residences of people with lower income; recent immigrants; tenants; and most members of groups designated as visible minorities, particularly respondents identifying as Filipino. By contrast, residential greenness was greater for non-immigrant White populations and those with higher income. These findings may be of particular interest to urban planners or those who are involved in reforestation efforts within cities. Given the increasing body of literature relating residential greenness to health outcomes,^{Note 1Note 2Note 3Note 4Note 5Note 9Note 14} differential patterns of household greenness in Canada may represent an indirect mechanism by which health inequities are further perpetuated.