Segregation and the effects of adolescent residential context on poverty risks and early income career: A study of the Swedish 1980 cohort

Abstract

Chinese

Will the consequences of residential segregation, that is, spatial concentration of marginalised populations on the one hand, and spatial concentration of affluent populations on the other hand, generate a situation where individual life trajectories are influenced by where individuals grow up? Our aim is to analyse how poverty risks and early income career at adult age are influenced by different neighbourhood contexts in early youth. We use Swedish longitudinal register data, and follow individuals born in 1980 until 2012. Residential context is measured in 1995 at age 15 by expanding a buffer around the residential locations of each individual and, by computing statistical aggregates of different socio-demographic variables for that population. The results show that poverty risks increase for individuals growing up in areas characterised by high numbers of social allowance recipients living nearby, whereas elite geographical context is favourable for both women’s and men’s future income.

Keywords

Income neighbourhood context neighbourhood effects poverty

Introduction

Today, measures of spatial variation need not be restricted to using aggregates for fixed geographical subdivision. Instead, it is possible to compute multi-scalar measures of geographical context to measure context using both multiple socio-demographic indicators and multiple geographical scales. These advances have been important both for the analysis of segregation patterns (Clark et al., 2015; Reardon et al., 2009) and for the development of composite measures of geographical context that can be used to assess neighbourhood effects (Andersson and Malmberg, 2015; Chaix et al., 2005; Holloway et al., 2012; Östh et al., 2014). In this paper, the questions about neighbourhood characteristics and scale will be addressed using an explicitly multi-scalar approach. Our aim is to analyse how poverty risks and early income career at adult age are influenced by different neighbourhood contexts in early youth.

Earlier research into neighbourhood effects on predominantly education and class, has indicated the need for an approach including some kind of indicator for elite context (Barthon and Monfroy, 2010). The importance of interaction with elite groups for the reproduction of advantaged social positions has been repeatedly stressed by scholars such as Bourdieu (e.g. Andersson and Malmberg, 2015; Bourdieu, 1984). His analysis, as discussed by Toft (2015), clearly suggests that residing in affluent neighbourhoods can help individuals to accumulate different forms of social capital that promote an elite career. Here the social mechanisms in play (Ainsworth, 2002; Friedrichs, 2015) are suggested to be tied to social capital, spatial capital and social support.

Another type of context often researched in neighbourhood effects literature is the foreign-born context. The roots of the theory that neighbourhoods dominated by foreign-born residents are not beneficial for individuals’ future work and for earning a reasonable income, can be found in American research about African Americans (Wilson, 1987). However, the American literature in this field of research is not directly comparable with the European situation, see for instance Wacquant (2008). In the Swedish research tradition, foreign-born neighbourhoods are often considered equivalent to suburbs, located in rather isolated positions away from the metropolitan centres, and built in the period 1964–1975.

Also debated in neighbourhood effect research are the influences on young people’s life trajectories of living in areas with high unemployment and poverty. This research has a long tradition in which high unemployment environments are considered negative for the opportunities of youth (Ludwig, 1999; Wilson, 1987). The reason why certain areas or rather larger urban districts are unsupportive of finding work is also referred to as the mismatch between workers and employers. The influence of poverty neighbourhoods on adolescents is probably the main theme researched in the neighbourhood effects literature, and it is probably unnecessary to say such influence is considered negative for children’s and adolescents’ future life course trajectories (Brännström, 2004; Brooks-Gunn and Duncan, 1997; Friedrichs, 1998; Galster et al., 2007; Holloway and Mulherin, 2004; Jencks and Mayer, 1990; McCulloch, 2001; Massey et al., 1994; South and Crowder, 1999).

Also, based on earlier research we can make assumptions about how housing can affect adolescents to take certain life trajectories. Single-family housing forming large areas with a middle-class population in the Swedish case will be a particular concern in our analysis. Earlier research has pointed to the will of policy makers to expect responsible, caring homeowners to stabilise neighbourhoods because of the investments they make in their housing (Doling and Ronald, 2010). In addition, competition in housing markets will result in only the middle-class and richer segments of the population being able to afford single-family housing and being able to enter the home-owning sector with a mortgage. Usually, neighbourhood effects from single-family housing areas have a positive influence on adolescents’ future achievements (Andersson and Malmberg, 2015) but the opposite is also possible (van Ham and Manley, 2010).

In order to construct measures of residential context we have selected indicator variables that can capture the different types of neighbourhood milieus that have been discussed in the neighbourhood effect literature. For elite environments, the presence of top income earners and individuals with tertiary education has been identified as a relevant indicator. The migrant status of neighbours is a natural measure of foreign-born context. As indicators of poverty context, this paper will employ indicators of employment and receipt of social allowance in combination with single motherhood. Finally, single-family housing will be used as an indicator of middle-class neighbourhoods.

To assess the influence of these indicator variables we will employ a novel approach to context measurement that is based on multi-scalar individualised neighbourhoods. This approach acknowledges the fact that individuals are not only influenced by the social context of their immediate surroundings; they also interact with, and can be influenced by, community-based institutions that serve smaller or larger geographical areas. Moreover, local employment opportunities can influence the expectations that individuals have with respect to different career choices. We would argue that a failure to acknowledge the multi-scalar nature of geographical context could result in a mis-measurement of neighbourhood influence on individual-level outcomes.

Taking advantage of the availability of individual-level data with geo-coordinates we construct aggregates for individualised neighbourhoods with different population counts. This approach to contextual measurement has previously been used by Bolster et al. (2007); Chaix et al. (2005); Macallister et al. (2001) and similarly in Andersson and Malmberg (2015). In our case we used the Equipop software developed by John Östh (Östh et al., 2014) to obtain aggregate information for neighbourhoods that vary in scale; their size being measured by the population count, k. In order to handle the complexity that results from combining multiple scales with multiple indicators, we will take advantage of the possibilities offered by a factor analysis. By applying factor analysis to the multi-scalar measures of neighbourhood composition, it becomes possible to identify the different types of neighbourhood context discussed above.

Also, our selection of outcome variables is based on the idea of different links between neighbourhood milieus and specific adult outcomes. Thus, we will use both poverty risks and indicators of early career success. The latter will be measured based on incidence of high income. The former is measured by receipt of social allowance and persistent low income.

We hypothesise that at age 15 living in elite areas will positively affect the income career as well as protect individuals from poverty in early adult life. A previous study has shown positive effects of elite areas on attainment of tertiary education (Andersson and Malmberg, 2015) as have the results of earlier international research inspired by Bourdieu (1984), among others.

If growing up in neighbourhoods with a high proportion of foreign-born residents exposes individuals to a higher proportion of newly arrived people and less helpful social networks, when it comes to finding well-paid work one would expect this factor to be associated with poorer outcomes. Because we will construct a factor that is independent of contexts with high education and income, or high unemployment, any such effect should be independent of the socio-economic characteristics of the foreign-born (Andersson et al., 2007)

Concentrations of social allowance in small-scale neighbourhoods should also be associated with poorer outcomes concerning both income and social allowance in early adulthood (Mood, 2010). Neighbourhoods are thus characterised by high proportions of non-employed and single mothers, which are traditionally associated with poorer outcomes in later life (Wilson, 1987).

Neighbourhoods of 25,000 individuals or fewer, characterised by a high proportion of residents living in single-family housing, are hypothesised to be associated with higher educational attainment (Andersson and Malmberg, 2015). Based on the same arguments as above, we might expect growing up in such neighbourhoods to be associated with positive effects on income, and a reduced risk of receiving social allowances.

In order to measure how different neighbourhood types affect early career success and poverty risk in adulthood we will use a cohort design based on longitudinal data that measures neighbourhood exposure in adolescence (around age 15) and outcomes at the end of young adulthood (around age 30). If, instead, we had used cross-sectional data it is likely that correlations between individual income and neighbourhood characteristics would have reflected the effects of income on where individuals live, not the other way around. Also, our design focuses on neighbourhood effect mechanisms that operate during adolescence, which is considered an important formative period in an individual’s life (Ainsworth, 2002).

However, the number of studies is relatively limited, especially studies using a long follow-up period (see e.g. Brännström, 2004; Oreopoulos, 2003). Moreover, most longitudinal studies focus on poverty outcomes and only a few consider individual income career. Lastly, some earlier studies have focused on particular metropolitan areas, which limits the variation in segregation for which contextual effects are measured (Andersson et al., 2007; Brännström, 2004). In comparison, studies that consider neighbourhood effects on educational achievement are more numerous, possibly because such studies require a shorter follow-up period (Andersson and Malmberg, 2015; Andersson and Subramanian, 2006; Galster et al., 2007; Sykes, 2011).

Studies using a longitudinal design to analyse neighbourhood effects on adult income from residential context in adolescence and childhood include Datcher (1982), Corcoran et al. (1992), Galster et al. (2007), Vartanian and Buck (2005) and Islam (2013). The first two did not find significant neighbourhood effects on adult income. Galster et al. (2007) found significant effects that were not robust when an instrument variable approach was used, whereas Islam (2013) found significant negative contextual effects on adult income from the neighbourhood-level poverty rates. All these studies used data from longitudinal studies (Andersson and Malmberg, 2015) with relatively small sample sizes. In the Swedish context, Brännström’s (2004) study is a prominent example.

Our study differs from the above studies in that we use a complete national sample based on register data and we employ a complete longitudinal design that relates adult outcome to neighbourhood conditions during adolescence. In addition, it can be argued that our contextual measure is an improvement on earlier approaches (see Andersson and Malmberg, 2015). We believe that this combination provides an opportunity to obtain more precise estimates of neighbourhood effects on adult income and poverty risks than earlier studies.

Empirical design, data and methods

The outcome variables in this study are indicators of poverty, low income, and high income at age 28–32 for a cohort born in 1980, and we will analyse the effect of neighbourhood exposure at the age of 15. Moreover, we will use contextual data from 1995, based on individualised neighbourhoods ranging in scale from populations including the 12 nearest neighbours to areas including the 25,600 nearest neighbours for Sweden. However, in the contextual effect estimations we will not use statistical aggregates obtained for individualised neighbourhoods directly. Instead, we will use factor scores resulting from a factor analysis carried out on the individualised neighbourhood-based data. We used ten factors, which we named according to their characteristics, as presented below. Since the outcome variables are categorical, a logit model will be used in order to assess the impact of individual-level and contextual-level variables.

Individual-level, cohort and household data

The dependent variable social allowance is a dummy variable with a value of one if the individual had social allowance income in at least one year between 2008 and 2012 (see Table 1). Four different dummy variables based on the maximal disposable income for the individual in the period 2008 to 2012 were used to measure income career for the study population: Below 25th percentile, Below 50th percentile, Above 75th percentile and Above 90th percentile. Percentiles were computed separately for each year and based on the study population. The coding based on the maximum percentile values implies that it is sufficient to have one year of income above the 75th and 90th percentiles for the dummy variables Above 75th and Above 90th to be coded as one. For the Below 25th and Below 50th dummies to be coded as one, on the other hand, it is necessary to have a disposable income below the 25th and 50th percentile for every year between 2008 and 2012. A motive for constructing the outcome variables as categorical is that low income and high income outcomes can be seen as a result of socialisation processes that direct the life course of individuals towards specific paths.

Table 1.

Descriptive statistics, outcome variables (based on 2008–2012 data) and individual background variables (1995 data).

At least one year from 2008 to 2012 with:	Men	Men	Women	Women	Total	Over all
At least one year from 2008 to 2012 with:	N	Proportion	N	Proportion	Total	Proportion
At least one year with social allowance	49,123	0.08	46,520	0.07	95,643	0.07
Maximum disposable income below 25th percentile (all years)	49,123	0.04	46,520	0.09	95,643	0.07
Maximum disposable below 50th percentile (all years)	49,123	0.18	46,520	0.36	95,643	0.27
Maximum disposable income above 75th percentile	49,123	0.57	46,520	0.36	95,643	0.47
Maximum disposable income above 90th percentile	49,123	0.31	46,520	0.16	95,643	0.24
Parent social allowance					95,643	0.11
Earned income percentile					95,643	60.84
Disposable income percentile					95,643	50.39
Parent university education					95,643	0.38
Parent not employed					95,643	0.24
Single mother household					95,643	0.19
Single father household					95,643	0.04
Swedish-born parent					95,643	0.88

As can be seen in Table 1, the proportions of men and women having a maximum value above the 75th or 90th percentiles are strikingly different; therefore, we use separate models for men and women. This is also true when it comes to low (below 25th and 50th percentiles) incomes for men and women.

To account for individual-level influence on poverty and income career our statistical model includes eight indicators of parental background: parent(s) with social allowance, family type (single mother households and single father household), parent(s) with university education, not employed parent(s), parents’ earned income in percentiles, parents’ disposable income in percentiles, and Swedish-born parents (one or two). ‘Not employed’ is not the equivalent of being registered as unemployed at the national public agency for employment but instead means not having paid work in November. That is, a person can be on parental leave, retired, travelling, a homemaker, etc., as well as unemployed. The same variable is used below for the contextual measure. Descriptive statistics for these variables are given in Table 1. The data source is MONA (Microdata-ON-Line-Access) provided by Statistics Sweden.

Our final sample consists of 95,643 individuals. This can be compared with the 102,601 individuals born in 1980 that lived in Sweden in 1995. The main reason that the final sample is smaller is that 6046 of these individuals have missing information on the outcome variables. An additional 480 individuals have been excluded because we have not been able to match them with contextual data in 1995. Finally, 314 individuals have been excluded because they have missing data on parental background.

Contextual measurement

With respect to the use of individual background variables, the empirical design of this study is conventional. This is not the case, though, with our approach to context measurement. Here, instead, our study introduces two important approaches: first, and most importantly, we introduce contextual measures that are based on individually defined and scalable neighbourhoods. Second, we introduce a factor analysis-based representation of the spatial variation in a socio-demographic context as a means to manage the wealth of information resulting from scalability and as a means to obtain orthogonal, clearly separated factors for our analysis.

Individualised and scalable neighbourhoods

We measure neighbourhood population compositions using individual centred neighbourhoods based on population size. Thus, we have used register data containing information about individual residential location to compute contextual variables based on the population composition among an individual’s nearest 12, 25, 50, 100, 200, 400, 800, 1600, 3200, 6400, 12,800 and 25,600 neighbours for 1995 (for the population older than 25 years and the total population, depending on variables), see Table 2.

Table 2.

Variables used with Equipop to measure residential context in 1995.

Variable	Description	Population	Number of neighbours (k)
Education	1 = university/college, 0 = not university/college	>25 years	12, 25, 50, 100, 200, 400, 800, 1600, 3200, 6400, 12,800, 25,600
Social allowance	1 = social allowance, 0 = not social allowance	All	(same as above)
Family type	1 = single mother, 0 = not single mother	>25 years	(same as above)
Disposable income	Disposable income, average percentile	>25 years	(same as above)
Foreign-born	1 = foreign-born, 0 = born in Sweden	All	(same as above)
Non-employed	1 = non-employed, 0 = employed	>25 years	(same as above)
Housing (1996)	1 = Single-family housing, 0 = other types of housing	All

Traditional measures of segregation such as the isolation index are strongly dependent on the size of the statistical units for which the segregation index has been computed (Östh et al., 2014). In the Equipop software, the individualised neighbourhoods are obtained by expanding a circular buffer around each residential location until the population encircled by the buffer corresponds to the population threshold chosen. When this threshold is reached, the program computes aggregate statistics on a selected socio-economic variable for the encircled population. For example, we used the proportion of single mothers among the 100 closest neighbours. For the contextual measurements, we have used data from the PLACE¹ database of Uppsala University. This body of data contains register-based, individual-level data for the population in Sweden from 1990 to 2010 with geocodes of the residential location by 100-m squares. From these data, seven different socio-demographic indicators have been extracted and used as input for Equipop; see Table 2.

Factor analysis, contextual variation

With seven different characteristics and 12 different levels of neighbourhood scale we obtain a total of 84 contextual indicators. These indicators should be seen as characteristics of a detailed location rather than as characteristics of a larger geographical subdivision.

We subjected the 84 contextual indicators to a principal components factor analysis, and varimax rotation. The best fitting model, as evidenced by proportional declines in eigenvalues and factor loadings above 0.40 was a seven-factor solution, capturing 67% of the original covariation. The seven factors demonstrate that both scale (number of neighbours) and substantive characteristics combine to describe the dimensions of an individual’s neighbourhood. The factors are orthogonal, thus clearly non-correlated for our further analysis. Moreover, the factors have been normalised to have zero means and a standard deviation of one.

Based on the factor loadings, we can characterise the seven dimensions (see Figure 1) as follows:

- Elite (factor 1): High proportions of people with tertiary education and disposable income above the 90th percentile. Factor loadings range from 0.61 to 0.93 for education and 0.44 to 0.63 for income across neighbourhood sizes, see Figure 1.

- Foreign-born (factor 2): High proportions of foreign-born for all neighbourhood sizes (loadings range from 0.45 to 0.93); high proportions of unmarried mothers and social assistance recipients load marginally (0.4 to 0.5) for the larger neighbourhood sizes, see Figure 1.

- Small-scale non-employment (factor 3): High proportions of non-employed, low proportions for incomes above the 90th percentile for smaller-size neighbourhoods.

- Large-scale non-employment (factor 4): High proportions of non-employed, low proportions for incomes above the 90th percentile for larger neighbourhoods.

- Social allowance nearby (factor 5): High proportions of individuals receiving social assistance and high proportions of unmarried mothers in smaller neighbourhoods.

- Single-family housing (factor 6): High proportions of persons living in single-family houses, across all neighbourhood sizes (loadings range from 0.41 to 0.79). This factor is not only an urban–rural dimension; when mapped, even small towns include enough renters and apartment-dwellers to show lower scores for larger sizes of closest neighbours (k > 6400).

- Medium-scale non-employment (factor 7). High proportions of non-employed persons in neighbourhoods of the medium range (400–1600), see Figure 1. (If this factor is excluded, factors 3 and 4 do not as clearly differentiate the small- and large-scale neighbourhoods in terms of non-employment and income.)

Figure 1.

Factor loadings at different scales. loadings on the y-axis and k number of closest neighbours on the x-axis. (To reduce clutter, these graphs only show factors that for at least one k-level have a loading higher than 0.2 or lower than −0.2.).

If these factors are compared with the different neighbourhood types discussed in the introduction it can be seen that there is a clear correspondence. Factor 1 Elite can clearly be seen as capturing environments that can promote long education and advantageous careers. Factor 2 Foreign-born will pick up environments with many migrants. Factor 3, factor 4, and factor 5 can be seen as related to poverty contexts and hence can be expected to increase the risk of poverty outcomes. Factor 5 Single-family housing can be interpreted as a measure of middle-class environments, and thus can be expected to promote longer education and increase the probability of high earnings. An advantage of a multi-scalar approach demonstrated by these results is that multi-scalar measures of geographical context allow a separation of different types of neighbourhood context that often cannot be reached if a single scale level is used (see Andersson and Malmberg, 2015). This separation of different dimensions in a neighbourhood context using multi-scalar measures is made possible by variation in the correlations between indicator variables aggregated at different scale levels. If a single scale level is used, a similar separation is hindered by high levels of multi-collinearity.

Results

Family background variables

For the family background variables, the effects on risk of having social allowance, low disposable income and on having a high disposable income between age 28 and age 32 are very much as expected (see Table 3) and correspond well with the results in earlier studies referred to above both concerning social allowance and income variables (Brooks-Gunn and Duncan, 1997; Mood, 2010).

Individuals that, at age 15, lived in households that received a social allowance have a strongly increased risk of receiving a social allowance themselves at least once between age 28 and 32. For men the odds ratio is 1.980 and for women 1.736 compared with individuals not having a background in such households (model 1 in Table 3). Also the risk of having a disposable income below the 25th and 50th percentiles is higher, models 2 and 3, Table 3. The individuals from households with a social allowance also have a lower chance of receiving a high disposable income in the same age span, irrespective of whether high is defined as above the 75th or 90th percentiles.

The same pattern is visible for individuals that at age 15 lived in single mother and single father households or in households where at least one parent was not employed. This increases this risk of receiving a social allowance between age 28 and 32 and decreases the chances of receiving a high disposable income (all estimates not significant) in the same age span, Table 3.

Having at least one parent with tertiary education has the opposite effect. This radically reduces the risk of receiving a social allowance at age 28–32 and having a low disposable income. It also has a strong positive effect on the chances of receiving a high disposable income. The estimates in Table 3, moreover, indicate that the effects of parents’ education are stronger for women than for men, and a statistical test (available from authors upon request) confirms that this is the case.

The effect of household disposable income and earned income at age 15 is similar in direction to that of having at least one parent with tertiary education. It reduces the risk of receiving a social allowance and having a low income at age 28–32, and has a positive effect on the chances of receiving a high disposable income. Note that the effect stated in Table 3 shows the result of an increase in disposable and earned incomes from the 25th to the 75th percentiles.

Having Swedish-born parents is not a significant background variable for social allowance. However, Swedish-born parent(s) are significant for reducing the risk of having a low disposable income in adulthood.

Table 3.

Parameter estimates, exp(β), for five models for poverty and income variables, 1995.

	1. Social allowance		2. Disposable income below		3. Disposable income below		4. Disposable income above		5. Disposable income above
	Men	Women	25th percentile		50th percentile		75th percentile		90th percentile
			Men	Women	Men	Women	Men	Women	Men	Women
Constant	0.077***	0.075***	0.086***	0.200***	0.334***	0.952	0.804***	0.290***	0.243***	0.087***
Parent social allowance	1.980***	1.736***	1.082	1.245***	1.229***	1.263***	0.809***	0.783***	0.781***	0.813***
Earned income, 25th–75th pct	0.834***	0.709***	0.725***	0.825***	0.772***	0.857***	1.268***	1.199***	1.190***	1.206***
Disposable income, 25th–75th pct	0.881**	0.903*	0.844**	0.743***	0.868***	0.819***	1.194***	1.285***	1.300***	1.402***
Parent university education	0.721***	0.657***	0.905	0.748***	0.902***	0.708***	1.145***	1.453***	1.307***	1.534***
Parent not employed	1.488***	1.456***	1.200**	1.194***	1.149***	1.130***	0.891***	0.929*	0.938*	0.950
Single mother household	1.674***	1.573***	1.160*	0.974	1.119**	1.041	0.923**	0.918**	0.922*	0.956
Single father household	1.834***	1.847***	1.103	1.077	1.179**	1.260***	0.805***	0.854**	0.733***	0.781**
Swedish-born parent	0.960	1.065	0.749***	0.852**	0.913*	0.945	1.060	0.971	0.993	0.860**
Elite (factor 1)	0.974	0.945**	0.981	0.880***	0.971**	0.856***	1.074***	1.196***	1.159***	1.258***
Foreign-born (factor 2)	1.004	1.012	0.994	0.971**	1.001	0.964***	1.023***	1.064***	1.055***	1.084***
Small-scale non-employment (factor 3)	0.995	1.031	1.036	1.073***	1.043***	1.041***	0.947***	0.945***	0.939***	0.934***
Large-scale non-employment (factor 4)	1.022	1.015	1.095***	1.064***	1.059***	1.065***	0.927***	0.923***	0.911***	0.893***
Social allowance nearby (factor 5)	1.095***	1.075***	1.035*	1.041***	1.036***	1.043***	0.954***	0.936***	0.939***	0.923***
Single-family housing (factor 6)	0.949***	0.918***	0.970	0.970*	0.960***	0.988	1.046***	1.011	1.049***	1.023*
Medium-scale non-employment (factor 7)	1.049**	0.980	0.968	1.011	0.992	1.002	1.012	1.021*	1.027**	1.054***
Chi-square family background variables	2031.395	1784.109	503.568	1055.893	1198.210	1986.799	1873.499	2405.275	2073.827	1936.414
Chi-square contextual-level variables	127.558	120.408	42.491	124.864	109.468	397.774	271.308	635.215	516.521	687.592

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

Contextual-level factors

The purpose of this study is to investigate how poverty risks and early income career at adult age are influenced by the different neighbourhood contexts in early youth. The effects of the contextual factors that are reported in Table 3 are therefore of special interest in relation to this aim. Given that the values for the contextual factors have been normalised, the parameter estimates presented in Table 3 show the effect of an increase in the factor score by one standard deviation. Our first observation here is based on the chi-square values for the family background variables and the contextual-level variables. The chi-square values are based on the differences in log likelihood that result from introducing the two sets of variables in addition to an intercept-only model. Thus, the values give an indication of how important each set of variables is for determining the outcome variables. Judging from the chi-square values it is clear that the family background variables in combination are more important than the contextual-level variables, most clearly so for the risk of receiving social allowance between 28 and 32 years of age and for the risk of having a low income in the same age span. The chi-square values, however, also indicate that the contextual-level variables are far from unimportant. This is a first indication that the approach to identifying contextual-level effects that we have taken in this paper is fruitful.

A second observation is that the contextual-level effects, judging from the chi-square values, appear to be stronger for more advantageous outcomes, disposable income above the 75th percentile and disposable income above the 90th percentile, than for poorer outcomes, social allowance and disposable income below the 25th percentile, Table 3. This is also reflected in the significance levels of the estimated effects for contextual factors. For the 90th percentile outcome, all the estimated effects are significant, whereas for social allowance and disposable income below the 25th percentile only about half of the estimates are significant. Looking at the sign of the point estimates, and ignoring significance levels, it is also clear that, in general, contextual factors that have a positive effect on advantageous outcomes have a negative effect on disadvantageous outcomes, see Table 3.

Contextual-level effects on disposable income

The estimates for contextual-level effects on both the high disposable incomes and low disposable incomes presented in Table 3 are to a large extent in line with what one might expect from theoretical discussions about how neighbourhood context during adolescence can influence life course trajectories.

Factor 1 Elite

Growing up in an elite neighbourhood (that is, having high values on factor 1) does, in general, have a positive effect on an individual’s chances of receiving a high income. The effect is stronger for women than for men. Our evaluation of the point estimates for women (in the range 1.20 to 1.26) is that these represent substantial effects for women that grow up in areas with high scores on the elite factor. From earlier studies it is known that elite contexts play a role in educational attainment. The stronger effect for women could result from education being more important for the income career for women than for men. For example, according to the estimates in model 5 of Table 3, women growing up in an area with an elite factor score of three would have a 100% increase in their odds of receiving a disposable income above the 90th percentile relative to the baseline value of zero. This is stronger than the effect of having a parent with tertiary education (53% increase). Given the level of socio-economic segregation in Sweden, there are few individuals that grow up in such high elite areas and, therefore, contextual-level effects on most individuals will be weaker. However, this does not change the fact that the estimated statistical effect gives support to the idea that neighbourhood context can strongly influence individual life course trajectories.

For women, coming from areas with a strong elite (factor 1) presence, on the other hand, gives some protection against the risk of low income. This can be seen as a mirror image of the positive effect of elite neighbourhoods on the chances of getting a high income, but at the same time it is interesting to note the gender difference in the contextual effect. One interpretation of this finding is that an elite environment can strengthen the career orientation of young women and lessen their orientation towards a career that involves part-time work and, therefore, the risk of low income. A support for this interpretation is that having a parent with tertiary education also has a stronger protective effect against low income for women than for men.

Factor 4 Large-scale non-employment

This is the contextual-level factor that has the second largest effect on high income. The parameter estimates are in the range 0.89 to 0.93 (models 4 and 5, Table 3), indicating a decrease of around 10% in the odds of receiving a high income when the score for this factor increases by one unit, see Table 3. The same caveat as above concerning the number of individuals affected by effects of this size also applies here; that is, rather few individuals live in extreme areas.

Factor 3 Small-scale non-employment and factor 5 Social allowance nearby

These contextual factors have similar effects on the odds of receiving a high income. The effects are in the range 4.6% to 7.7% lower odds for receiving a high income (see models 4 and 5, Table 3) but both have appreciably smaller overall effects than factors 1 and factor 4. The gap in factor score values between the 10th and 90th percentiles is about three for both factors, corresponding to a difference in effect that is of similar size to the combined effect of growing up in a household with a social allowance and an unemployed parent. However, for factor 5 there are a few locations that have very high scores, above 5 (97.5th percentile) (compare Mood, 2010). It could be of interest to investigate whether the effect of such high levels of exposure is linear, but this falls outside the scope of this paper.

With respect to the risk of experiencing low income in young adulthood, poor employment prospects in the area where individuals grew up stand out as important. The Large-scale non-employment factor (4) is especially important for increasing the risk of low income, but small-scale non-employment (3) and factor 5 Social allowance nearby are also associated with risk increases.

Factor 6 Single-family dwelling

For men, this factor has significant positive effects on the odds of receiving a high income, stronger for men than for women. The effects are relatively small (2–5% range, models 4 and 5, Table 3). Also, living in single-family housing dominated areas is somewhat protective against low incomes in adult age, especially for men’s disposable income below the 50th percentile, model 2 in Table 3.

Factor 2 Foreign-born

This factor has significant positive effects on the odds of receiving a high income in adult age for the 1980 cohort. The result is contrary to expectations in that growing up in a neighbourhood dominated by migrants could have adverse effects on the prospects for a good income career. One reason could be that foreign-born composition as such has little effect when socio-economic composition is controlled for. Another possible reason is that our variable (foreign-born) is too broad to account for effects related to, for example, discrimination. However, the result is in line with the result in Andersson et al. (2007: 656) who did not find foreign-born neighbourhood mix to be important to income; rather, the socio-economic mix of neighbourhoods was important.

For women, factor 2 Foreign-born has a protective effect on the risk of receiving a low income.

Factor 7 Medium-scale non-employment

This factor has no consistent significant effect on the odds of receiving a high or low income in young adulthood.

Contextual-level effects on social allowance

The estimated contextual effects on poor outcomes are, as pointed out above, in general inverse to the effects of advantageous outcomes, and less strong. For receipt of social allowance two contextual factors stand out as especially important: Social allowance nearby (factor 5) and Single-family housing (factor 6). The former increases the risk for social allowance recipiency and the latter decreases the risk. This is in line with the theoretical expectations and with earlier findings in the literature (Mood, 2010). Growing up in areas with a concentration of poor households can affect expectations and opportunities negatively and, thus, increase the risk of poor adult outcomes. Here it can be noted that neighbourhoods characterised by Social allowance nearby (factor 5) tend to increase the risk of having a low income, and this also supports the idea that a background in areas with a concentration of poor households can have negative effects on economic prospects. Conversely, the estimates for factor 6 Single-family housing can be seen as suggesting that a background in neighbourhoods dominated by owner-occupied single-family housing (factor 6) can foster attitudes and expectations that make individuals more capable of avoiding and addressing economic difficulties.

Contextual effects and the level of segregation

The extent to which adult outcomes vary between different neighbourhoods will be determined by the level of segregation, as shown by the estimates of contextual factors in Table 3. High levels of segregation imply high variation in factor scores (Manley et al., 2015) and, thus, large differences in expected outcome across neighbourhoods. Examples of variation in scores between neighbourhoods for factor 1 Elite, and factor 4 Large-scale non-employment scores in different commuting regions in Sweden are given in Figure 2.

Figure 2.

Percentiles of factor 1 Elite and factor 4 Large-scale non-employment in the labour market regions of Munkfors, Trollhättan, Jönköping and Malmö.

Thus, the Munkfors labour market region (first row, Figure 2) is an example of a region with low segregation and thus low variation in factor scores. The upper left graph shows the low variation in the Elite factor, and there are almost no positive scores for the Elite factor in Munkfors. There is also low variation for the Large-scale non-employment factor but the scores are positive, indicating fairly high non-employment in the region (upper right graph).

The next example is Trollhättan, a region with moderate levels of segregation. Compared with Munkfors, Trollhättan has a wider variation of values for both factors (see graphs in the second row of Figure 2). This higher variation implies that at the extreme ends of the distribution there will be some contextual effects on the income careers for young adults, more so than in Munkfors. In Munkfors the effects of location on income career will be negligible.

Even wider variation in factor scores and, thus greater differences between neighbourhoods, is found in the Jönköping labour market region. The presence of a university college, and the larger city size are possible explanations for the positive incline of the Jönköping curves (row three, Figure 2). The increase in the factor 1 Elite, from about the 90th percentile is evidence of neighbourhoods with high elite representation, which in turn lead to larger effects on income careers.

The last example is the labour market region of Malmö (bottom row, Figure 2) a metropolitan area in the Swedish context (about 1 million inhabitants). In this region we find high levels of variation in factor scores across neighbourhoods. The Malmö labour market region includes both Lund, a university town with a long academic tradition, and areas with manufacturing industries employing comparably unskilled labour. Large variation in factor scores across neighbourhoods implies that location will strongly affect the income careers of individuals who grow up in the Malmö region.

In conclusion, this means that apart from family background effects on adolescents’ future poverty and income careers, the segregation in regions and cities helps to form the life course trajectories of Swedish youths, and that these differences are greater in some regions than others, as shown by the region examples above.

Concluding discussion

In this paper we have applied a new approach to contextual measurement in order to assess the effect of different neighbourhood characteristics on early adult income career and poverty outcomes. In our theoretical framework we hypothesised that it would be possible to identify different dimensions in neighbourhood context that could influence such outcomes. The empirical results that we have presented above largely support this idea.

There are positive contextual effects on early adulthood high income from adolescent residential areas with elite characteristics, and there are negative effects on high future incomes from poverty and non-employment contexts. Also, single-family housing contexts promote high incomes and protect from having social allowance in early adulthood. An exception to the expected results, but not unknown from earlier research, is the positive effects on high incomes of growing up in contexts characterised by foreign-born residents. This last result, as well as the other results, is found when at the same time controlling for individual/parental socio-demographic background.

In the literature, effects of neighbourhood context on individual-level income have been assessed as either weak or close to non-existent (Bolster et al., 2007; Van Ham et al., 2012). Our results agree with earlier studies that for most individuals, variation in neighbourhood context has little impact on their income career. However, if effect sizes are computed not for individuals that grow up in average neighbourhoods but for individuals from the most advantaged or most disadvantaged neighbourhoods, the contextual-level impact turns out to be comparable with the effects of important elements of family background such as parental income, parental education or social allowance recipiency. These relatively strong effects of residential context are found for the probability of having a high income around age 30 but not to the same extent for the risk of receiving social allowance and having low income. From this we draw four tentative conclusions.

First, the Swedish welfare state used to be effective in protecting individuals from poverty, as measured by social allowances. When the contextual effects on poverty are compared with those found in an earlier longitudinal study by Brännström (2005) on individuals born in the 1950s, where no effects on either poverty or social allowance were found, we do find stronger effects. One reason could be that the mid-1990s was a period of economic crisis and welfare state retrenchment. When poverty was measured by low incomes we found large-scale non-employment areas increased the risk of the 1980 cohort having below the 25th and 50th percentiles of disposable income in adult age.

Second, segregation levels have increased with time, especially in the metropolitan areas, partly because of a greater diversity in the total Swedish population resulting from a larger inflow of international migrants. Therefore, the 1980 cohort that we studied had been exposed to higher levels of segregation compared with the cohort studied by Brännström (2005).

Third, and in accordance with Chetty et al. (2014), neighbourhood effects in combination with high levels of segregation pose a risk of impeding social mobility. From the results of strong effects (negative and positive) on income careers in areas deviating from factor means in labour market regions we can conclude that there is a risk that intergenerational income mobility is also hindered in Sweden.

Moreover, that the estimated contextual effects are important implies that parents who make residential choices on the basis of ideas about how different neighbourhoods might influence their children’s future are not necessarily mistaken. In this study we found that the factor of single-family housing did have a protective effect against social allowance for adolescents growing up in such areas. To reside in single-family housing areas is thus to be socialised into adopting behaviours and developing capacities that promote finding a job and earning an income to support oneself. Of course, it could be argued that such parental choice, if linked to unobserved characteristics of the families that influence children’s life courses, can imply that estimated contextual effects are not truly causal.

Fourth, in earlier studies of neighbourhood effects in Sweden (Andersson et al., 2007; Brännström, 2005) there are suggested policy implications to consider. When no neighbourhood effects are found, strong arguments for mixing policies are lacking, and general welfare policies can be seen as more effective than area-based policies. Our results, on the other hand, point in the opposite direction. Both area-based policies and mixing can be seen as important for hindering the effects on adolescents of growing up in disadvantaged areas. Area-based programmes are important for creating equal opportunities, and mixing policies are important for preventing strict sorting and strong residential segregation.

Moreover, we conclude that the empirical design of this paper, which combines a longitudinal design, the use of a large, register-based sample, and measures of context based on individualised, scalable neighbourhoods with a combination of indicators, has resulted in more precise estimates than earlier studies. We believe that our improved contextual measure is one reason why we find comparatively strong neighbourhood effects. Another possible explanation for why we find neighbourhood effects is that the use of a national sample that includes individuals from regions with low levels of segregation may have increased the variation in residential context among individuals with similar parental background. Moreover, the longitudinal design is important since children and adolescents are considered to be extra susceptible to the social and physical surroundings. The result of this study is an image of two strands. Individuals coming from affluent and less affluent residential areas in Sweden are already under strong influence of differences in family background. But because of contextual influences between living in the least advantaged areas and the most advantaged areas, there is a risk that high levels of segregation will induce individuals with different backgrounds to drift even further apart.

Footnotes

Funding

This research was supported by the Stockholm University Linnaeus Center on Social Policy and Family Dynamics in Europe, SPaDE, grant 349-2007-8701 of the Swedish Research Council and Riksbankens Jubileumsfond (DNR M12-0301:1). The Equipop software used for computing contextual variables based on individualised neighbourhoods was developed by John Östh, Uppsala University.

Notes

References

Ainsworth

(2002) Why does it take a village? The mediation of neighborhood effects on educational achievement. Social Forces 81(1): 117–152.

Andersson

Subramanian

(2006) Explorations of neighborhood and educational outcomes for young Swedes. Urban Studies 43(11): 2013–2025.

Andersson

Malmberg

(2015) Contextual effects on educational attainment in individualised, scalable neighbourhoods: Differences across gender and social class. Urban Studies 52(12): 2117–2133.

Andersson

Musterd

Galster

et al . (2007) What mix matters? Exploring the relationships between individuals’ incomes and different measures of their neighbourhood context. Housing Studies 22(5): 637–660.

Barthon

Monfroy

(2010) Sociospatial schooling practices: A spatial capital approach. Educational Research and Evaluation 16(2): 177–196.

Bolster

Burgess

Johnston

et al . (2007) Neighbourhoods, households and income dynamics: A semi-parametric investigation of neighbourhood effects. Journal of Economic Geography 7(1): 1–38.

Bourdieu

(1984) Distinction: A Social Critique of the Judgement of Taste. Trans. Nice

. Cambridge, MA: Harvard University Press.

Brännström

(2004) Poor places, poor prospects? Counterfactual models of neighbourhood effects on social exclusion in Stockholm, Sweden. Urban Studies 41(13): 2515–2537.

Brännström

(2005) Does neighbourhood origin matter? A longitudinal multilevel assessment of neighbourhood effects on income and receipt of social assistance in a Stockholm birth cohort. Housing, Theory and Society 22(4): 169–195.

10.

Brooks-Gunn

Duncan

(1997) The effects of poverty on children. The Future of Children 7(2): 55–71.

11.

Chaix

Merlo

Subramanian

et al . (2005) Comparison of a spatial perspective with the multilevel analytical approach in neighborhood studies: The case of mental and behavioral disorders due to psychoactive substance use in Malmö, Sweden, 2001. American Journal of Epidemiology 162(2): 171–182.

12.

Chetty

Hendren

Kline

et al . (2014) Where is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States. Cambridge, MA: National Bureau of Economic Research. Available at: http://www.nber.org/papers/w19843.

13.

Clark

Anderson

Östh

et al . (2015) A multiscalar analysis of neighborhood composition in Los Angeles, 2000–2010: A location-based approach to segregation and diversity. Annals of the Association of American Geographers 105(6): 1260–1284.

14.

Corcoran

Gordon

Laren

et al . (1992) The association between men’s economic status and their family and community origins. Journal of Human Resources 27(4): 575–601.

15.

Datcher

(1982) Effects of community and family background on achievement. The Review of Economics and Statistics 64(1): 32–41.

16.

Doling

Ronald

(2010) Property-based welfare and European homeowners: How would housing perform as a pension? Journal of Housing and the Built Environment 25(2): 227–241.

17.

Friedrichs

(1998) Do poor neighborhoods make their residents poorer? Context effects of poverty neigborhoods on residents. In: Andress

H-J

(ed.) Empirical Poverty Reseach in a Comperative Perspective. Aldershot, Brookfield USA, Singapore, Sydney: Ashgate, pp. 77–99.

18.

Friedrichs

(2015) Neighbourhood effects: lost in transition? Analyse & Kritik 38(1): 73–90.

19.

Galster

Marcotte

Mandell

et al . (2007) The influence of neighborhood poverty during childhood on fertility, education, and earnings outcomes. Housing Studies 22(5): 723–751.

20.

Holloway

Mulherin

(2004) The effect of adolescent neighborhood poverty on adult employment. Journal of Urban Affairs 26(4): 427–454.

21.

Holloway

Wright

Ellis

(2012) The racially fragmented city? Neighborhood racial segregation and diversity jointly considered. The Professional Geographer 64(1): 63–82.

22.

Islam

(2013) Childhood neighborhood conditions and the persistence of adult income. Regional Science and Urban Economics 43(4): 684–693.

23.

Jencks

Mayer

(1990) The social consequences of growing up in a poor neighborhood. In Lynn

McGeary

MFH

(eds) Inner-city Poverty in the United States. Washington DC: National Academy Press, pp. 111–186.

24.

Ludwig

(1999) Information and inner city educational attainment. Economics of Educational Review 18(1): 17–30.

25.

Macallister

Johnston

Pattie

et al . (2001) Class dealignment and the neighbourhood effect: Miller revisited. British Journal of Political Science 31(01): 41–59.

26.

McCulloch

(2001) Teenage childbearing in Great Britain and the spatial concentration of poverty households. Journal of Epidemiology and Community Health 55(1): 16–23.

27.

Manley

Johnston

Jones

et al . (2015) Macro-, meso- and microscale segregation: Modeling changing ethnic residential patterns in Auckland, New Zealand, 2001–2013. Annals of the Association of American Geographers 105(5): 951–967.

28.

Massey

Gross

Shibuya

(1994) Migration, segregation, and the geographic concentration of poverty. American Sociological Review 59(3): 425–445.

29.

Mood

(2010) Neighborhood social influence and welfare receipt in Sweden: A panel data analysis. Social Forces 88(3): 1331–1356.

30.

Oreopoulos

(2003) The long-run consequences of living in a poor neighborhood. Quarterly Journal of Economics 118(4): 1533–1575.

31.

Östh

Malmberg

Andersson

(2014) Analysing segregation with individualized neighbourhoods defined by population size. In: Lloyd

Shuttleworth

Wong

(eds) Social-Spatial Segregation: Concepts, Processes and Outcomes. Bristol: Policy Press, pp. 135–161.

32.

Reardon

Farrell

Matthews

et al . (2009) Race and space in the 1990s: Changes in the geographic scale of racial residential segregation, 1990–2000. Social Science Research 38(1): 55–70.

33.

South

Crowder

(1999) Neighborhood effects on family formation: Concentrated poverty and beyond. American Sociological Review 64(1): 113–132.

34.

Sykes

(2011) Spatial order and social position: Neighbourhoods, schools and educational inequality. PhD Thesis, University of Amsterdam. Available at: http://dare.uva.nl/record/365250.

35.

Toft

(2015) High-class living: Truly affluent neighbourhoods in Norway. Paper presented at the European Network for Housing Research Conference 2015, Lisbon 29 June–1 July.

36.

Van Ham

Manley

(2010) The effect of neighbourhood housing tenure mix on labour market outcomes: A longitudinal investigation of neighbourhood effects. Journal of Economic Geography 10(2) 257–282.

37.

Van Ham

Manley

Bailey

et al . (2012) Neighbourhood Effects Research: New Perspectives: Dordrecht, Netherlands: Springer.

38.

Vartanian

Buck

(2005) Childhood and adolescent neighborhood effects on adult income: Using siblings to examine differences in ordinary least squares and fixed-effect models. Social Service Review 79(1): 60–94.

39.

Wacquant

(2008) Urban Outcasts A Comparative Sociology of Advanced Marginality. Cambridge: Polity Press.

40.

Wilson

(1987) The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago, IL and London: The University of Chicago Press.