Co-ethnic concentration and trust in Canada’s urban neighbourhoods

Abstract

Chinese

This study investigates the relationship between the density of people’s ethno-racial in-group in their neighbourhoods (co-ethnic concentration) and trust in their neighbours. Previous studies demonstrate that ethno-racial diversity decreases trust in others, however, these studies rely on overly broad definitions of diversity and of trust, and often do not disaggregate the effects for Whites and ethno-racial minorities. Hence, this study examines the relationship between co-ethnic concentration and trust, focusing on how this relationship may change depending upon one’s ethno-racial status. Putnam’s (2007) analysis leads to a paradox in the sense that, according to the same principle that predicts declining trust amongst Whites, increasing diversity should lead to greater levels of trust for ethno-racial minorities whose share of the population increases with diversification. The findings demonstrate that there is a positive relationship between co-ethnic concentration and trust in neighbours and that this relationship holds for Whites as well as ethno-racial minorities.

Keywords

co-ethnic concentration neighbourhood diversity social capital trust

Introduction

This study investigates the relationship between the density of one’s ethno-racial in-group in one’s neighbourhood (co-ethnic concentration) and trust in one’s neighbours. Our study represents an outgrowth of an expansive literature that addresses the impact of ethno-racial diversity on social capital – trust being one of the attitudinal components of social capital (Putnam, 2000, 2007). Central to the many studies that find diversity to have negative impacts on trust (Alesina and La Ferrara, 2002; Costa and Kahn, 2003; Leigh, 2006; Letki, 2008; Putnam, 2007; Stolle et al., 2008; Uslaner, 2011) is the inverse finding that homogeneity has a positive impact on trust. Our study draws on the 2006 Canadian Census (20% microdata file) and Cycles 22 and 23 of the Canadian General Social Survey (GSS), to test the effect of co-ethnic concentration (homogeneity) on trust. Of particular interest is the question of whether co-ethnic concentration has the same effect on dominant group (White) and ethno-racial minorities, or whether there is a differential effect on these groups.

This study’s use of the term ‘ethno-racial’ is a result of the Canadian conventional use of ‘visible minorities’ as a catchall category for non-White, non-Aboriginal persons. Members of visible minorities are defined by the Canadian Employment Equity Act as ‘persons, other than Aboriginal people, who are non-Caucasian in race or non-white in colour’ (Government of Canada, 2016). In government statistics, visible minorities self-identify as one of the following groups: Blacks, Chinese, Filipinos, Japanese, Koreans, Latin Americans, Pacific Islanders, South Asians, and West Asians/Arabs (Statistics Canada, 2016). The listed categories are not all of the same type (one is a skin colour-based racial category, some are ethnicities, while others refer to regions of descent) and thus ethno-racial is used as an umbrella term throughout the paper, in spite of the important differences between race/ethnicity/nation.

This study posits that the meaning of living in an ethno-racially homogenous neighbourhood depends on an interplay between the ethno-racial composition of the neighbourhood and an individual’s ethno-racial status. For example, a neighbourhood where Whites are demographically predominant could be a place of abundant social ties for the White residents, but a place of social isolation for others (Hero, 2007). Hence, this study examines the relationship between co-ethnic concentration and trust in neighbours as well as how this relationship may differ among ethno-racial groups. Co-ethnic concentration refers to the size of an individual’s in-group in the neighbourhood. As the density of co-ethnics increases, the ethno-racial homogeneity of the neighbourhood increases for in-group members, but decreases for out-group members. Putnam’s (2007) observation that ethno-racial diversity has a negative effect on trust implies that homogeneity has a positive effect.

Our study contributes to the literature on social diversity and trust in several ways. First, it specifically studies co-ethnic concentration as one aspect of diversity. Surprisingly, given the prevalence of studies on diversity and trust, co-ethnic concentration remains largely unstudied. To avoid the ‘cacophony of empirical findings’ (van de Meer and Tolsma, 2014: 460) that result from the simultaneous measurement of several competing aspects of diversity, we focus on co-ethnic concentration as one central feature of diversity in Canada. Similarly, rather than examining generalised trust we focus on trust in neighbours. Diversity is often found to have different effects on different aspects of trust. This study avoids overgeneralisation by focusing on trust in neighbours. Given the study’s focus on co-ethnic concentration at the neighbourhood level, measures of trust should be related to the neighbourhood context. Third, the study provides a new national case study (Canada) that fits well with a model of diversification that entails increasing co-ethnic concentration. Finally, unlike the majority of studies on diversity and trust, we disaggregate ethno-racial minorities and Whites in order to test whether there is a differential effect on these groups. These four features make our study not only methodologically unique, but also provide us with the ability to gain insight into a complex social and political debate with relevance to any country facing demographic diversification similar to Canada.

Co-ethnic concentration and trust

For the purposes of our study we draw heavily on literature that emerged as a response to Putnam’s (2000, 2007) conceptualisation of trust as one component of the attitudinal aspect of social capital (the other main aspect of social capital being the structural social networks of individuals). Several scholars have persuasively argued that social capital needs to be measured along many other dimensions besides trust (Bécares, et al., 2011; Hooghe, 2007). Although we agree with this assessment, we wish to avoid the mixed results that often occur from building complex composite measures of social capital. Instead, our study makes no pretences about determining the impact of diversity on social capital in general, and focuses on a clear, contextualised, and carefully qualified analysis of the impact of co-ethnic concentration on trust in neighbours.

One of the major weaknesses of the existing literature is that it often fails to sufficiently specify for whom diversity has a negative impact on trust (for everyone?; for Whites?; for ethno-racial minorities?). Numerous prior studies demonstrate that trust is lower in ethno-racially diverse than in homogenous environments (Alesina and La Ferrara, 2002; Costa and Kahn, 2003; Leigh, 2006; Putnam, 2007). This negative effect of diversity includes a lower level of generalised trust, inter-racial trust, intra-racial trust, and trust in neighbours. These studies also show that the negative effect of diversity is independent of local-level socioeconomic disadvantage and exists across cultural environments. Most of these studies hypothesise mechanisms whereby diversity leads to social withdrawal, which in turn makes trusting others more difficult (Williamson, 2015). However, unlike diversity in the abstract, co-ethnic concentration means that there are more people ‘like you’ and thus should logically lead to higher levels of trust.

Abascal and Baldassarri (2015) provide an extensive critique of the overly broad and uncritical use of diversity measures in the existing literature. They argue that studies using heterogeneity indexes as a measure of diversity ‘discount the analytic distinction between in-group and out-group contact, which is so critical to social psychological theories of intergroup relations’ (p. 751). Similarly, Fieldhouse and Cutts (2010) criticise the existing literature for studying the general population rather than disaggregating the population into ethno-racial groups that may be differently impacted by diversification. Our study aligns with the above scholars who call for a more nuanced disaggregation of concepts such as diversity.

Besides asking what it is about diversity that leads to mistrust, it is also important to consider what it is about homogeneous environments that fosters trust. The preference for trusting people who are similar or familiar suggests that there is less perceived risk in trusting people in environments with a high density of co-ethnics. Co-ethnics concentrate in neighbourhoods for several reasons, notably including reasons such as relating to mutual needs, interests, and values. For some, especially recent immigrants and ethno-racial minorities, this can represent an economic strategy since co-ethnic networks are a source of social capital, providing information, employment opportunities, and sources of instrumental support (Bauder and Sharpe, 2002; Chiswick and Miller, 2005). Co-ethnic concentration can also reflect preferences for living among co-ethnics or in a familiar cultural and linguistic setting. In Canada, high co-ethnic concentration often represents an ethnic group’s capacity to mobilise collective resources and establish ethnic institutions (Hou, 2006). These ethnic community neighbourhoods represent a common social space for interaction among people with similar values, needs, and cultural backgrounds.

These ethnic communities can help to build trust in two ways. First, when the community emerges as a response to systemic racial discrimination or economic exclusion, a sense of bounded solidarity (Portes and Sensenbrenner, 1993) can help create shared norms and a moral community. Uniform social norms create predictability and make trusting easier (Gundelach and Freitag, 2014). Second, these communities have a material mechanism for enforcing predictable behaviour that may improve trustworthiness of members. Portes (1998) observes that obligations are enforceable within co-ethnic networks ‘not through recourse to law or violence but through the power of the community’ (p. 9). Given the economic aspect of ethnic enclaves (Bauder and Sharpe, 2002; Chiswick and Miller, 2005), there can be real material consequences for a failure to fulfil an obligation. The network’s ability to monitor behaviour and disseminate information about shirkers motivates group members to act in expected ways, and thus lowers the potential cost of trusting others.

There are, of course, limitations and problems associated with the above ethnic community mechanisms for explaining trust in neighbours. This mechanism assumes actually existing ethnic or racially based groups in neighbourhoods. As Brubaker (2004) points out, actually existing groups and sociological categories do not always map neatly on to one another. To assume that self-identification in a census category leads directly to ethnic community participation and identification with a moral community, which in turn leads to greater trust, is a form of ethnic and racial essentialism that should be avoided. At the same time, the dominance in the literature of in-group homogeneity as an explanation for levels of trust makes this explanation an important hypothesis to explore. As Brubaker (2004) explains, ‘groupness’ may happen, but it also may not. In this study we do not treat ethnic and racial groups as bounded and essentialised, but rather examine the effect of own ethno-racial concentration on an individual’s trust of neighbours when individual characteristics such as socioeconomic, linguistic, and demographic factors are controlled for.

Despite the vast literature on diversity and trust, relatively little is known about the relationship between co-ethnic concentration and trust. Co-ethnic concentration of ethno-racial minorities is a specific form of diversification; in White settler colonies such as Canada, as diversity increases so does the co-ethnic concentration of minorities such as Chinese-, South Asian-, and Filipino-Canadians. Out of the 90 studies on diversity and social capital, van der Meer and Tolsma (2014) found only two studies that disaggregated majority/minority groups to analyse the differential effect of diversity on intragroup cohesion. Bécares et al. (2011) argue that this lack of study is the result of most studies’ focus on Whites as the baseline for analysis. Given the demographic trends in Canada, it is becoming increasingly dubious to frame the empirical narrative about trust around Whites as the standard. When they shifted the focus away from the majority toward racial minority groups, Bécares et al. (2011) found that own-group ethnic concentration increased trust and respect for ethnic difference. In a comparative study of the UK and the USA, the authors found that co-ethnic concentration for minorities ‘blunted’ the negative effect of diversity on trust in the USA and actually improved trust for minorities in the UK (Fieldhouse and Cutts, 2010). On the other hand, Siordia and Saenz (2013) found no relationship between co-ethnic concentration and positive neighbourhood perception.

Hypotheses

From the above discussion of the possible mechanisms that links co-ethnic concentration with trust, we can draw three hypotheses. First, if Putnam (2007) is correct that ethno-racial heterogeneity negatively affects all ethno-racial groups equally, then we should expect that co-ethnic concentration will have an equally positive effect on trust for Whites and ethno-racial minorities alike. Thus:

H1: Engagement hypothesis. Co-ethnic concentration will positively affect level of trust for Whites and racial minorities equally.

If, however, we focus on the specifics of immigrant ethnic enclaves – and the demographic trends in Canada suggest that many ethno-racial minorities will be recent immigrants – it may be the case that co-ethnic concentration matters more for ethno-racial minorities than for Whites. If ethno-racial enclaves are a defence against racial and economic exclusion (see Bauder and Sharpe, 2002; Chiswick and Miller, 2005; Portes and Sensenbrenner, 1993), then we might expect that co-ethnic concentration is more consequential for these groups. Thus:

H2: Minority bounded solidarity hypothesis. Co-ethnic concentration will positively affect trust for racial minorities more than Whites.

Finally, it is possible that the opposite of H2 is the case. Hero (2007) argues that the social networks that should lead to greater trust are not neutral and universalisable, but are rather the product of White privilege and racial exclusion. If this is the case, then Whites have a lot more to lose compared with ethno-racial minorities when their co-ethnic concentration decreases. It is debatable whether Hero’s study on ‘Blacks’ in the USA is generalisable to other minorities in Canada, but his study does put forward the important idea that co-ethnic concentration may be more important for Whites because of their long history of numerical and socio-political dominance in a country like Canada. Perhaps ethno-racial groups need to reach a numerical majority or a political dominance in order to reap the trust enhancing benefits of co-ethnic concentration. There is also the possibility that perception, rather than a network mechanism (Gundelach and Freitag, 2014), might explain a relatively greater importance of co-ethnic concentration for Whites. Since English and French speaking Whites have historically dominated the cultural landscape in Canadian cities, any decrease in co-ethnic concentration may be perceived as a symbolic loss of universal norms regulating behavior, whereas minorities, having never experienced cultural dominance in Canada, would not attach the same importance to changes in own group demographics. Thus:

H3: Declining White dominance hypothesis. Co-ethnic concentration will positively affect trust for Whites more than racial minorities.

Data and methods

Data sources

The data for the empirical analysis come from three sources: the 2006 Canadian Census (20% microdata file) and Cycles 22 and 23 of the General Social Survey (GSS). The census data were employed to construct the selected neighbourhood-level variables. This study defines neighbourhoods as census tracts, which is consistent with research conventions (see Alba et al., 2000; Hou, 2006). Although census tracts are imperfect representations of actual ‘lived’ or subjective neighbourhoods, these are accurate representations of the demographic and socioeconomic characteristics of residential environments. Prior research has consistently shown that contextual variables measured at the census tract level have a strong association with individual health and socioeconomic outcomes in Canada (e.g. Fong and Hou, 2015; Hou and Myles, 2005). In Canada, census tracts consist of approximately 4000 persons on average, and using the 20% microdata file reduces this number to about 800 respondents per neighbourhood. Using common geocodes, the census-derived variables were merged with the individual-level data from the GSS-22 and GSS-23 to generate a multilevel data set. These geographic identifiers allowed us to determine the residential neighbourhoods of the respondents in the GSS-22 and GSS-23.

The GSS-22 and GSS-23 were pooled and provide our individual-level data. The GSS-22 was conducted in 2008 and the GSS-23 in 2009 by Statistics Canada. The GSS-22 and GSS-23 are nationally representative surveys of the Canadian population aged 15 and older, excluding full-time residents of institutions and residents of the Yukon, Northwest Territories, and Nunavut. See Allan and Marchand (2010) and Burns and Williams (2011) for further details about survey and sample design and data collection procedures. The sample size of the GSS-22 is 20,401 with a response rate of 57.3%. The sample size of the GSS-23 is 19,422 with a response rate of 61.6%. We pooled these surveys for a total of 20,905 White respondents and 3141 racial minorities, after excluding cases with missing data. Except for income, there are few cases with missing data. To retain the cases where income is missing, we included a dummy variable for those who did not report income in all regression models.

The respondents in our target population reside in over 4000 urban neighbourhoods nested in Census Metropolitan Areas (CMA) and large census agglomerations (CA) in the ten provinces. A CMA has a total population of at least 100,000 persons and an urban core of at least 50,000 persons (Statistics Canada, 2013). A large CA has a core population of between 50,000 and 99,999 persons. About 95% of racial minorities in Canada live in these urban areas; their proportion ranges from a low of under 1% in some CMAs in Québec to a high of 37% in the Toronto and Vancouver metropolitan areas (Wu et al., 2011). At the CMA-level, there is also an uneven distribution of racial minorities across neighbourhoods.

Dependent variable

Our dependent variable of interest is a person’s trust in their neighbours. The respondents were asked: ‘Using a scale of 1 to 5 where 1 means “Cannot be trusted at all” and 5 means “Can be trusted a lot” how much do you trust … people in your neighborhood’. Although our dependent variable is an ordinal variable, we treat it as a continuous variable in the regression analysis (see more discussion in the methods section). Table 1 presents the coding and descriptive statistics of all of the individual- and neighbourhood-level variables in the analysis. It shows that the mean level of trust on the 5-point scale is 3.61 for Whites and 3.24 for racial minorities.

Table 1.

Coding and descriptive statistics of variables used in the analysis.

	Whites		Racial minorities
	M or %	S.D	M or %	S.D
Response variable
Trust neighbours (1 = cannot be trusted at all, …, 5 = can be trusted a lot)	3.61	0.87	3.24	1.10
Ethnic grouping
South Asian	–	–	23.2%	–
Black	–	–	16.0%	–
Filipino	–	–	9.5%	–
Latin American	–	–	7.2%	–
Southeast Asian	–	–	6.1%	–
Arab/West Asian	–	–	10.6%	–
Korean	–	–	1.5%	–
Japanese	–	–	1.4%	–
Chinese	–	–	24.6%	–
Individual-level variables
Age (range: 15–98)	45.21	15.07	37.84	15.49
Sex (Female = 1)	50.8%		50.2%
Marital status
Widowed	5.0%	–	1.7%	–
Separated or divorced	6.9%	–	4.9%	–
Never married	25.6%	–	36.3%	–
Married or cohabiting	62.5%	–	57.1%	–
Education
Some post-secondary	42.4%	–	35.3%	–
High school	13.4%	–	9.0%	–
Less than high school	16.0%	–	13.3%	–
Bachelor’s degree or higher	28.2%	–	42.3%	–
Family income		–		–
Lowest income (<Can$20,000)	4.4%	–	6.8%	–
Low middle income (Can$20,000–49,999)	10.8%	–	12.3%	–
Middle income (Can$50,000–79,999)	13.7%	–	15.5%	–
Upper middle income (Can$80,000–99,999)	23.6%	–	20.0%	–
Income not reported	18.3%	–	25.3%	–
Highest income (>=Can$100,000)	29.3%	–	20.1%	–
Immigrant status (immigrant = 1)	13.7%	–	78.8%	–
Home language (1 = not English or French)	3.4%	–	42.3%	–
Years lived in the current neighbourhood
Less than 1 year	6.9%	–	10.0%	–
1–3 years	12.1%	–	20.8%	–
3–5 years	10.4%	–	15.9%	–
Over 5 years	70.6%	–	53.3%	–
Homeownership (1 = yes)	79.0%	–	73.4%	–
Location
Toronto	18.0%	–	43.1%	–
Montreal	16.9%	–	11.5%	–
Vancouver	8.2%	–	18.1%	–
Other large CMAs with population over 500k	24.7%	–	18.1%	–
Small CMAs	25.5%	–	8.0%	–
Large census agglomerations	6.7%	–	1.2%	–
Neighbourhood-level variables
% of own ethnic group	0.82	0.15	0.14	0.17
Income inequality	190.49	97.54	197.58	121.48
Low-income rate	0.14	0.08	0.19	0.11
% with university degrees	0.21	0.10	0.24	0.11
% of non-movers	0.58	0.11	0.53	0.15
% of seniors	0.13	0.05	0.11	0.06
Logged population density	7.18	1.30	7.91	1.06
N	20,905		3141

Sources: the 2006 Census 20% sample microdata file; the 2008 and 2009 General Social Survey.

Individual-level variables

The analysis controls for a selection of individual-level variables that have a well-known influence on interpersonal trust (see Hou and Wu, 2009; Putnam, 2007). The individual-level controls include several demographic variables: ethnic/racial grouping, age, sex, marital status, and immigrant status. For ethno-racial grouping, we consider ten groups: Whites and nine groupings for racial minorities. It is unfortunate that the GSS-22 did not collect information on ethnic groupings for Whites. Table 1 shows that Chinese, South Asians and Blacks are the three largest racial minorities, representing 25%, 23%, and 16% of the minority population, respectively. Age is measured in years, with a range of 15 to 98 years and a mean of 45 years for White respondents and 38 years for racial minorities. Sex is measured using a dummy variable (1 = female); about 50% of all respondents are female. Marital status is measured using a four-level categorical variable: widowed, separated or divorced, never married, and married or cohabiting (reference group). The majority of Whites (63%) and racial minorities (57%) are married or cohabiting and the next highest response is never married. A small proportion is separated/divorced or widowed. Immigrant status (foreign-born) is measured using a dummy variable (1 = immigrant). About 14% of Whites and 79% of racial minorities are immigrants.

The analysis also controls for several indicators of socioeconomic status: education, household income, and homeownership. Education is measured using a four-level categorical variable: less than high school, high school diploma, some post-secondary, and a Bachelor’s degree or higher (reference group). In general, racial minorities have higher educational attainment than Whites. About 42% of racial minorities have a Bachelor’s degree or higher in comparison with 28% of Whites. At the lower end, 13% of racial minorities and 16% of Whites have less than high school. About 42% of Whites and 35% of racial minorities have some post-secondary education. Household income is measured with a six-level categorical variable, ranging from lowest income (less than Can$20,000) to highest income (greater than Can$100,000). As noted, this variable includes a category for ‘income not reported’ because of the high number of missing data for it. Over 18% of Whites and 25% of racial minorities failed to report their household income. Homeownership is measured with a dummy variable (1 = yes); 79% of Whites and 73% of racial minorities are homeowners.

In addition, the regression models control for several other individual-level variables. Home language or whether the respondent’s primary language used in the home was either English or French (1 = no), which are Canada’s official languages. Over 42% of racial minorities report using a home language other than English or French compared with less than 4% of Whites. The study also controls for time lived in the neighbourhood: less than 1 year, 1–3 years, 3–5 years, and 5 years and longer. Over 70% of Whites and 53% of racial minorities have lived in the same neighbourhood for 5 years or longer. A larger proportion of racial minorities have also lived in the same place for less than 3 years. Finally, the study controls for regional location. This variable includes Toronto, Montréal, Vancouver, other large metropolitan areas, small metropolitan areas, and large census agglomerations. Over 90% of racial minorities live in Toronto, Montreal, Vancouver, or other large metropolitan areas. This compares with 68% of Whites.

Neighbourhood-level variables

The regression models include several measures of the demographic and socioeconomic characteristics of neighbourhoods. Our primary interest is in the ethno-racial composition of neighbourhoods. It is measured as the level of co-ethnic concentration in the neighbourhood. This represents the percentage of the neighbourhood that consists of a respondent’s ethno-racial in-group. This measure considers ten major ethno-racial groups: Whites and a nine-category disaggregation of the ethno-racial minority population. The ethno-racial minority groups include: Chinese, South Asians, Blacks, Filipinos, Latin Americans, Southeast Asians, Arab/West Asians, Koreans, and Japanese. Table 2 indicates the sample size for each ethno-racial group. This table also provides the average co-ethnic compositions of neighbourhoods. On average, Whites reside in neighbourhoods that are 83% co-ethnic. Chinese and South Asians, Canada’s two largest ethno-racial minority groups, also reside in neighbourhoods with large concentrations of co-ethnics. Chinese Canadians reside in neighborhoods that are 24.5% co-ethnic and South Asians in neighbourhoods that are 20.1% co-ethnic. All other groups live in neighbourhoods that have co-ethnic concentrations less than 10% of the neighbourhood population.

Table 2.

Ethno-racial groups in Canada.

	Sample size	Mean % own group in the community	Mean level of trust
Chinese	745	24.5	3.4
South Asian	661	20.1	3.3
Black	610	8.7	2.9
Filipino	298	7.3	3.2
Latin American	236	2.7	3.1
Southeast Asian	172	2.2	3.4
Arab/West Asian	319	4.2	3.1
Korean	43	2.7	3.4
Japanese	57	0.7	3.8
White	20,905	83.0	3.6
Total	24,046

Source: the 2008 and 2009 General Social Survey.

The analysis controls for several indicators of neighbourhood socioeconomic context, given that prior studies observe that the concentrated disadvantage could confound the relationship between diversity and trust (Letki, 2008; Sampson and Graif, 2009). Socioeconomic context is measured using several variables: neighbourhood income inequality (using coefficient of variation), the low-income rate of the neighbourhood (using Statistics Canada low-income cut-offs), and percentage of university graduates in the neighbourhood. Other ecological-level variables include: the percent of non-movers (residential stability), i.e. the percent of people who have lived in the neighbourhood for 5 years or longer; the percent of seniors (people aged 65 and older) in the neighbourhood; and logged population density.

Statistical methods

Our empirical analysis uses ordinary least squares (OLS) regressions. Although the OLS model assumes a continuous response variable and our dependent variable is ordinal, we chose the OLS model for two reasons. First, the measure is on a five-point scale and the underlying concept (trust) is continuous. Prior research demonstrates that OLS estimates are generally robust when the ordinal (dependent) variable has at least five categories and the underlying concept is continuous and unidimensional (e.g. Bollen and Barb, 1981; Carifio and Perla, 2008). Second, we experimented modelling it as an ordinal variable using ordered logistic regression models. The results are very similar to those reported using OLS models, indicating that it is not unreasonable to treat the dependent variable as a continuous variable (see Appendix A1). In addition, we also examined several other OLS model assumptions, particularly multicollinearity, a ubiquitous phenomenon among contextual variables. We computed the zero-order correlations of all neighbourhood-level variables (see Appendix A2) and estimated variance inflation factors for all independent variables (most are less than three and none is over ten). We found no evidence of violations of the key OLS assumptions in our regression models.

Besides selection of statistical models, two methodological issues require brief discussion, however. First, there is the issue of clustering. In multilevel models, cluster effects (i.e. correlated errors within neighbourhoods and unequal variances across neighbourhoods) could bias our regression estimates. To address this issue we computed robust standard errors for the regression models, which account for the clustering effects that can arise when working with multilevel data (Steenbergen and Jones, 2002). This approach is akin to a fixed-intercept model with level-1 covariates in hierarchical linear models (Raudenbush et al., 2000). In essence, this approach first estimates the mean outcome for each cluster (neighbourhood), adjusted for differences in individual-level characteristics across neighbourhoods, and then regresses the mean outcome onto neighbourhoods-level predictors. Multilevel models can also be used to conduct cross-level analyses, examining, for instance, whether the effect of gender varies across neighbourhoods. But these questions are beyond the scope of this study.

Second, there is the issue of self-selection (endogeneity) in neighbourhood choice. In certain cases, a person’s neighbourhood choice could reflect their attitudes toward others. For example, a prejudiced person can self-select themselves out of neighbourhoods with large numbers of racial minorities or out-group members. To address this issue, we experimented using an instrumental variable (IV) approach to test this proposition. Following Dustmann and Preston (2001), we used co-ethnic concentration at the municipality level (with an average population of 110,000) as the instrument for the same variable at the neighbourhood level. The theoretical underpinning of this procedure is the assumption that most residential moves take place between neighbourhoods and that mobility of individuals across neighbourhoods within a municipality does not alter the overall diversity of the municipality. The results from the IV models are reported in Appendix A3. In brief, the results from the Davidson-MacKinnon test confirm that the regressor in question (co-ethnic concentration) is not endogenous, suggesting that the IV approach is unnecessary for our purposes.

Results

Although our interest is on individual-level outcomes, there could be group-level differences in trust in neighbours. Table 2 presents the means levels of trust in neighbours for each ethno-racial group considered in the analysis. For the most part, most ethno-racial groups report a fair level of trust in their neighbours. The mean level of trust among ethno-racial minorities ranges from a low of 2.9 for Blacks to a high of 3.8 for Japanese. The mean level of trust in neighbours for Whites is 3.6. The mean level of trust in neighbours for all ethno-racial minority groups (except Japanese) is somewhat lower than the mean level of trust among Whites. These group-level differences in trust could reflect either the ecological conditions where an individual lives or dispositional differences in the general propensity to trust others. On the one hand, since most ethno-racial minorities live in White-dominated neighbourhoods, the aggregate level of trust for these groups could be lower than for Whites because living in a White-dominated neighbourhood associates with weak ties to neighbours for most ethno-racial minorities. On the other hand, it is also possible that some ethno-racial groups are less trusting of others, regardless of neighbourhood contexts.

Table 3 presents the OLS regressions of trust in neighbours among Whites and ethno-racial minorities. As noted earlier, the study conducts separate empirical analysis for Whites and ethno-racial minorities, examining whether the effect of co-ethnic concentration is conditional on ethno-racial status. For Whites, the bivariate model demonstrates that there is a positive relationship between co-ethnic concentration and an individual’s trust in neighbours. With each unit increase in the percentage of co-ethnics in the neighbourhood, the level of an individual’s trust in neighbours increases. Model 2 adds individual- and neighbourhood-level covariates. When these covariates are controlled for, the magnitude of the correlation between co-ethnic concentration and trust in neighbours declines, but the level of significance remains unchanged. The findings from this model also demonstrate that the correlation between neighbourhood co-ethnic concentration and trust is independent of factors such as an individual’s demographic attributes (e.g. gender, age, and marital status) and socioeconomic status as well as ecological factors, such as the low-income rate of the neighbourhood, neighbourhood stability, and population density.

Table 3.

Ordinary least squares regression of trust in neighbours (with robust SEs): Canada, 2006–2009

	Whites		Racial minorities
	Model 1	Model 2	Model 1	Model 2
% of own ethno-racial group	0.607***	0.324***	0.196	0.287†
Individual-level variables
Ethnic grouping
South Asian		–		−0.021
Black		–		−0.263**
Filipino		–		−0.068
Latin American		–		−0.057
Southeast Asian		–		0.155
Arab/West Asian		–		−0.148
Korean		–		0.084
Japanese		–		0.307*
Chinese (reference)
Age		0.013***		0.010***
Sex (1 = female)		0.052**		−0.097*
Marital status
Widowed		0.096*		−0.191
Separated or divorced		−0.128***		−0.159
Never married		−0.047		−0.154*
Married or cohabiting (reference)
Education
Some post-secondary		−0.082***		−0.152**
High school		−0.088**		−0.291**
Less than high school		−0.084**		−0.036
Bachelor’s degree or higher (reference)
Family income
Lowest income (<Can$20,000)		−0.198***		−0.080
Low middle income (Can$20,000–49,999)		−0.132***		−0.230*
Middle income (Can$50,000–79,999)		−0.124***		−0.222**
Upper middle income (Can$80,000–99,999)		−0.078**		−0.061
Income not reported		−0.106***		−0.127
Highest income (>=$100,000) (reference)
Immigrant status (1 = immigrant)		−0.020		0.050
Home language (1 = not English or French)		0.011		−0.077
Years lived in the current neighbourhood
Less than 1 year		−0.172***		−0.153†
1–3 years		−0.102***		0.017
3–5 years		−0.109***		0.082
Over 5 years (reference category)
Homeownership (1 = yes)		0.306***		0.135*
Location
Montreal		−0.067*		−0.007
Vancouver		0.139***		0.003
Other large CMAs with population over 500k		0.000		−0.018
Small CMAs		0.067*		0.156†
Large census agglomerations		0.019		0.432*
Toronto (reference)
Neighbourhood-level variables
Low-income rate		−0.738***		−0.594*
% with university degrees		0.675***		0.556*
% of non-movers		0.268**		0.046
% of seniors		0.079		0.499
Logged population density		−0.033***		−0.029
Intercept	3.110***	2.678***	3.216***	3.215***
R square	0.012	0.140	0.001	0.111

Notes: *** significant at p < 0.001; **p < 0.01; *p < 0.05; †p < 0.10.

Sources: the 2006 Census 20% sample microdata file; the 2008 and 2009 General Social Survey.

The covariates do, however, have some notable correlations with trust in neighbours among Whites. Trust in neighbours increases with age and tends to be higher among females than males. Lower individual socioeconomic status correlates with lower levels of trust. Respondents who have lived in the same neighbourhood for over five years trust their neighbours more than respondents who have lived in their neighbourhood for shorter periods of time. This suggests that time spent in the neighbourhood is an important dimension of forging ties with neighbours. At the ecological-level, trust in neighbours is lower in neighbourhoods with a higher low-income rate and in neighbourhoods with higher in- and out-migration. Lower levels of education and higher population densities within neighbourhoods also associate with lower levels of trust. These findings suggest that ecological conditions such as local-area poverty, neighbourhood instability, and over-crowding are barriers to the development of trust between neighbours.

Table 3 also presents these OLS models for ethno-racial minorities. Model 1 demonstrates that the correlation between co-ethnic concentration and trust in neighbours is non-significant for ethno-racial minorities. A marginally significant relationship emerges when the covariates are added (p = 0.067). This demonstrates that the non-significant relationship observed in Model 1 is somewhat misleading. Similar to Whites, trust in neighbours among racial minorities increases as the concentration of co-ethnics increases, after controlling for the selected covariates, although the level of significance is lower for racial minority groups than for Whites. Indeed, the difference in the regression estimates on co-ethnic concentration between the two models is nonsignificant (p > 0.05), suggesting that the positive effect of co-ethnic concentration is similar between Whites and racial minorities, which is consistent with H1.

Despite the intuitive appeal of the minority bounded solidarity hypothesis, which suggests co-ethnic concentration will be more significant for ethno-racial minorities (H2), we do not find convincing evidence for this hypothesis in our findings. Without controlling for individual-level variables and ecological variables, the relationship between co-ethnic concentration and trust is non-significant for ethno-racial minorities, whereas it is significant for Whites. When the control variables are added to the model, the relationship, although marginally significant, is clearly not stronger for ethno-racial minorities than for Whites (see Model 2, Table 3).

Finally, the results presented in Table 3 do not fully contradict the declining White dominance hypothesis (H3). One must be careful in interpreting the results since the relationship between co-ethnic concentration and trust is on the cusp of being statistically significant (p = 0.067) for ethno-racial minorities. It appears that while there is some evidence for H1, there is also the potential to interpret the results as supporting H3. The relationship between co-ethnic concentration and trust is clear cut and significant for Whites, but is harder to interpret for ethno-racial minorities. The stronger significance of the relationship for Whites can be seen as tentative, but weak, evidence that co-ethnic concentration matters more for White members of the population.

Turning to the control variables, there are ethnic variations in trust. The level of trust is higher among Japanese Canadians and lower among Black Canadians. There is little variation among the other ethnic groups. Like Whites, trust increases with age, is lower among women than men, and is lower among the never married than the married/cohabiting. Trust is also comparatively low for racial minorities in the low-middle and middle-income categories, and non-homeowners. Trust is also lower for racial minorities living in the three largest Canadian cities (Toronto, Montreal, and Vancouver) or other large CMAs than those who reside in other urban localities. Thus, it is possible that the proportion of respondents in these categories is responsible for suppressing the positive relationship between co-ethnic concentration and trust. It is also possible that living in neighbourhoods with lower proportions of residents with university degrees and high low-income rates suppresses this relationship. None of the other selected neighbourhood-level covariates has a significant correlation with trust for ethno-racial minorities.

Conclusion

Because of the importance of the ethno-racial demographic shift in Canada, our study examined the potential differential impact that co-ethnic concentration might have on ethno-racial minorities and Whites. Our study provides support for the hypothesis that co-ethnic concentration has a similarly beneficial impact on trust in neighbours for Whites and ethno-racial minorities. While the literature on diversity and trust suggests some negative impact on overall levels of trust in neighbours, our findings reinforce the importance specifying what type of diversity one is referring to (co-ethnic concentration), and for whom this has a negative impact (Whites). The bad news story of declining levels of trust as a result of increased diversity is really only a story of declining trust for Whites as a result of declining co-ethnic concentration. Given the fact that a handful of immigrant source countries are consistently in the top sending countries (Statistics Canada, 2013), Canada’s growing diversity will likely coincide with increased co-ethnic concentration of ethno-racial minorities leading to increased levels of trust amongst these populations. As major urban centres become less White-dominated, overall levels of trust should stabilise. When diversity is conceptualised through the lens of co-ethnic concentration, even the most robust aspect of Putnam’s declining trust theory is refuted.

Co-ethnic concentration appears to matter for levels of trust in neighbours. Because of this, we conclude that previous, often highly influential studies, only tell us part of the story when they find that diversity is associated with a decline in trust. Diversity cannot be measured as a single variable as it has different forms, such as co-ethnic concentration. For example, Putnam’s influential research (2007) does not decompose diversity into co-ethnic concentration. Alesina and La Ferrara (2002), like Costa and Kahn (2003), do not examine the effect of co-ethnic concentration or the potential differential impact on various ethno-racial groups. It may be the case that ethno-racial heterogeneity (as a different demographic aspect of diversity from co-ethnic concentration) does, in fact, lead to a decline in trust in neighbours, but these influential studies do not adequately decompose the concept of diversity to tease out this relationship. To be fair to Putnam, his study (2007) does call for more nuanced decomposition of diversity in future work. Our study is one step in this direction. Future work could examine co-ethnic concentration in tandem with ethno-racial heterogeneity indexes to further our knowledge of how these two aspects of diversity interact with one another.

Our study lends support to previous studies that find that diversity does not have the same impact on all ethno-racial categories. Knowing that co-ethnic concentration positively affects trust in neighbours for all ethno-racial categories goes some way to help explain these findings. For example, Stolle et al. (2008) found that ethno-racial minorities’ level of trust was less affected by a context of diversity than Whites. Similarly, Fieldhouse and Cutts (2010) found that the negative relationship between diversity and neighbourhood norms (including trust) held for Whites but was not clear-cut for ethno-racial minorities. We can speculate that perhaps while one aspect of diversity (heterogeneity) is responsible for lower levels of trust, for minorities, the other aspect of diversity (co-ethnic concentration) blunts this effect. Our study provides the important insight that co-ethnic concentration matters for all groups, and thus provides an impetus for future research to examine the above speculation.

Our study contributes to a small, but growing, body of literature that investigates the relationship between co-ethnic concentration and trust. Our findings are consistent with Fieldhouse and Cutts (2010) who found that co-ethnic concentration has a positive impact on community norms (including trust) in both the USA and England. However, one of the very few studies that examines co-ethnic concentration and a specific measure of trust in neighbours (Bécares et al., 2011) found mixed results. For most ethno-racial categories in the UK, co-ethnic concentration increased trust, but for Black Caribbeans the relationship was reversed. Given the limited number of studies on this specific research question and level of statistical significance in the current study, it is important not to overstretch the findings. Nevertheless, there appears to be a trend in research that finds a positive relationship between co-ethnic concentration and trust.

Our findings provide an important empirical basis for the assessment of demographic trends in Canada and beyond. If immigration-led population growth continues to draw on a small number of source countries, then we can expect increasing co-ethnic concentration for these groups, and thus increasing levels of trust, which will partially offset overall declines in trust caused by diminishing White co-ethnic concentration. If, however, immigration streams are highly diversified we might expect increasing diversification without minority co-ethnic concentration to offset the overall decline in trust because of declining White dominance. This implication could have significant consequences for immigrant receiving countries, and suggests that immigration scholars and policy makers should pay particular attention to co-ethnic concentration as a central feature of diversification in the future. By highlighting the relevance of the current research to immigration and diversity policy, the authors are not advocating a policy position in one way or another on questions related to neighbourhood segregation or diverse immigration streams. We simply want to emphasise that diversity is a complex concept and that co-ethnic concentration does not warrant the current fearful and simplistic political discourse of ethno-racial minorities living ‘parallel lives’ (Pratsinakis et al., 2015) that inevitably lead to lower levels of trust. Whatever else co-ethnic concentration does for a society, it does not appear to erode trust in neighbours.

Footnotes

Appendix

Appendix A3.

Instrumental variable regression (with robust standard errors) of trust in neighbours: Canada, 2006–2009.

	Whites		Racial minorities
	b	Robust standard error	b	Robust standard error
% of own ethno-racial group	0.307***	0.076	−0.031	0.260
Individual-level variables
Age	0.013***	0.001	0.010***	0.002
Sex (Female = 1)	0.052**	0.018	−0.100*	0.045
Marital status
Widowed	0.096*	0.037	−0.192	0.161
Separated or divorced	−0.128***	0.030	−0.157	0.099
Never married	−0.048	0.026	−0.152*	0.064
Married or cohabiting^a
Education
Some post-secondary	−0.082***	0.020	−0.151**	0.053
High school	−0.088**	0.029	−0.288**	0.095
Less than high school	−0.085**	0.029	−0.035	0.082
Bachelor’s degree or higher^a
Family income
Lowest income (<Can$20,000)	−0.198***	0.047	−0.074	0.104
Low middle income (Can$20,000–49,999)	−0.132***	0.034	−0.227*	0.090
Middle income (Can$50,000–79,999)	−0.124***	0.028	−0.220**	0.078
Upper middle income (Can$80,000–99,999)	−0.079**	0.024	−0.063	0.072
Income not reported	−0.106***	0.030	−0.123	0.075
Highest income (>=Can$100,000)^a
Immigrant status (immigrant = 1)	−0.020	0.027	0.053	0.064
Home language (1 = not English or French)	0.011	0.056	−0.070	0.051
Years lived in the current neighbourhood
Less than 1 year	−0.172***	0.039	−0.161	0.084
1–3 years	−0.102***	0.028	0.015	0.060
3–5 years	−0.109***	0.029	0.080	0.069
Over 5 years^a
Homeownership (1 = yes)	0.305***	0.026	0.143*	0.061
Location
Montreal	−0.064*	0.032	−0.030	0.080
Vancouver	0.138***	0.040	0.013	0.068
Other large CMAs with population over 500k	0.002	0.028	−0.039	0.068
Small CMAs	0.070*	0.031	0.128	0.087
Large census agglomerations	0.021	0.043	0.400*	0.193
Toronto^a
Neighbourhood-level variables
Income inequality	0.000	0.000	0.000	0.000
Low-income rate	−0.749***	0.127	−0.488	0.284
% with university degrees	0.674***	0.084	0.529*	0.223
% of non-movers	0.270**	0.084	0.041	0.168
% of seniors	0.086	0.145	0.413	0.439
Logged population density	−0.033***	0.007	−0.028	0.025
Intercept	2.691***	0.111	3.274***	0.271
R square	0.140	0.101
Davidson-MacKinnon test	n.s.			n.s.

Notes :*** significant at p < 0.001; **p < 0.01; *p < 0.05.

Model 2 for visible minorities includes South Asians, Blacks, Filipinos, Latinos, Southeast Asians, West Asian/Arab, Korean, Japanese, with Chinese as the common reference.

Sources: the 2006 Census 20% sample microdata file; the 2008 and 2009 General Social Survey.

Funding

The authors gratefully acknowledge financial support from the Social Sciences and Humanities Research Council of Canada.

References

Abascal

Baldassarri

(2015) Love thy neighbor? Ethnoracial diversity and trust reexamined. American Journal of Sociology 121: 722–782.

Alba

Logan

Stults

(2000) The changing neighborhood contexts of the immigrant metropolis. Social Forces 79: 587–621.

Alesina

La Ferrara

(2002) Who trusts others? Journal of Public Economics 85: 207–234.

Allan

Marchand

(2010) General Social Survey, Cycle 22: Social Networks Public Use Microdata File Documentation and User’s Guide. Ottawa: Minister of Industry.

Bauder

Sharpe

(2002) Residential segregation of visible minorities in Canada’s gateway cities. The Canadian Geographer 46: 204–222.

Bécares

Stafford

Laurence

. (2011) Composition, concentration and deprivations: Exploring their association with social cohesion among different ethnic groups in the UK. Urban Studies 48: 2771–2787.

Bollen

Barb

(1981) Pearson’s R and coarsely categorized measures. American Sociological Review 46: 232–239.

Brubaker

(2004) Ethnicity Without Groups. Cambridge, MA: Harvard University Press.

Burns

Williams

(2011) General Social Survey, Cycle 23: Victimization Public Use Microdata File Documentation and User’s Guide. Ottawa: Minister of Industry.

10.

Carifio

Perla

(2008) Resolving the 50-year debate around using and misusing Likert scales. Medical Education 42: 1150–1152.

11.

Chiswick

Miller

(2005) Do enclaves matter in immigrant adjustment? City & Community 4: 5–35.

12.

Costa

Kahn

(2003) Civic engagement and community heterogeneity: An economist’s perspective. Perspectives on Politics 1: 103–111.

13.

Dustmann

Preston

(2001) Attitudes to ethnic minorities, ethnic context, and location decisions. The Economic Journal 111: 353–373.

14.

Fieldhouse

Cutts

(2010) Does diversity damage social capital? A comparative study of neighbourhood diversity and social capital in the US and Britain. Canadian Journal of Political Science 43: 289–318.

15.

Fong

Hou

(2015) Neighbours helping neighbours in multi-ethnic context. Population, Space and Place 23: 1–14.

16.

Government of Canada (2016). Employment Equity Act. Ottawa: Government of Canada. Available at: http://laws-lois.justice.gc.ca/PDF/E-5.401.pdf.

17.

Gundelach

Freitag

(2014) Neighbourhood diversity and social trust: An empirical analysis of interethnic contact and group-specific effects. Urban Studies 51: 1236–1256.

18.

Hero

(2007) Social Capital and Racial Diversity: Equality and Community in America. Cambridge: Cambridge University Press.

19.

Hooghe

(2007) Social capital and diversity generalized trust, social cohesion and regimes of diversity. Canadian Journal of Political Science 40: 709–732.

20.

Hou

(2006) Spatial assimilation of racial minorities in Canada’s immigrant gateway cities. Urban Studies 43: 1191–1213.

21.

Hou

Myles

(2005) Neighbourhood inequality, neighbourhood affluence and population health. Social Science & Medicine 60: 1557–1569.

22.

Hou

(2009) Racial diversity, minority concentration, and trust in Canadian urban neighborhoods. Social Science Research 38: 693–716.

23.

Leigh

(2006) Trust, inequality, and ethnic heterogeneity. Economic Record 82: 268–280.

24.

Letki

(2008) Does diversity erode social cohesion? Social capital and race in British neighborhoods. Political Studies 56: 99–126.

25.

Portes

(1998) Social capital: Its origins and applications in modern sociology. Annual Review of Sociology 24: 1–24.

26.

Portes

Sensenbrenner

(1993) Embeddedness and immigration: Notes on the social determinants of economic action. American Journal of Sociology 98: 1320–1350.

27.

Pratsinakis

Hatziprokopiou

Labrianidis

. (2015) Living together in multi-ethnic cities: People of migrant background, their interethnic friendships and the neighbourhood. Urban Studies 54: 102–118.

28.

Putnam

(2000) Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster.

29.

Putnam

(2007) E Pluribus Unum: Diversity and community in the twenty-first century: The 2006 Johan Skytte Prize Lecture. Scandinavian Political Studies 30: 137–174.

30.

Raudenbush

Bryk

Cheong

. (2000) HLM5 Hierarchical Linear and Nonlinear Modeling. Lincolnwood: Scientific Software International.

31.

Sampson

Graif

(2009) Neighborhood networks and processes of trust. In: Cook

Hardin

Levi

(eds) Whom Can We Trust? How Groups, Networks, and Institutions Make Trust Possible. New York: Russell Sage Foundation, pp. 182–216.

32.

Siordia

Saenz

(2013) On the relationship between neighborhood perception, length of residence and co-ethnic concentration. Applied Spatial Analysis 6: 267–284.

33.

Statistics Canada (2013) Immigration and Ethnocultural Diversity in Canada, NHS 2011. Ottawa: Statistics Canada.

34.

Statistics Canada (2016) Visible Minority of Person. Ottawa: Statistics Canada. Available at: http://www.statcan.gc.ca/eng/concepts/definitions/minority01.

35.

Steenbergen

Jones

(2002) Modeling multilevel data structure. American Journal of Political Science 46: 218–237.

36.

Stolle

Soroka

Johnston

(2008) When does diversity erode trust? Neighborhood diversity, interpersonal trust, and the mediating effect of social interactions. Political Studies 56: 57–75.

37.

Uslaner

(2011) Trust, diversity, and segregation in the United States, and the United Kingdom. Comparative Sociology 10: 221–247.

38.

van der Meer

Tolsma

(2014) Ethnic diversity and its effects on social cohesion. Annual Review of Sociology 40: 459–478.

39.

Williamson

(2015) Mechanisms of declining intra-ethnic trust in newly diverse immigrant destinations. Journal of Ethnic and Migration Studies 41: 1725–1745.

40.

Hou

Schimmele

(2011) Racial diversity and sense of belonging in urban neighborhoods. City and Community 10: 373–392.