Mapping the invisibles: Using non-conventional point-level data to analyse residential patterns of deprived people in a mid-sized city

Abstract

Chinese

In urban geography it is a common practice to refer to censuses and other official sources of data to analyse residential segregation. However, there are limitations to those data, especially where particular groups of people are concerned. Often, official sources of data do not allow micro-level analysis of the characteristics, needs and residential patterns of socially disadvantaged residents. At the same time, measures based on income thresholds fail to fully take into account the complexity of multidimensional deprivation. This work uses the unconventional information coming from a voluntary organisation to investigate and understand the residential patterns of disadvantaged residents living in a mid-sized Italian city. The factors associated with the relative presence of deprived residents in city neighbourhoods are tested with a GLM/Poisson regression model. The results are differentiated among the sub-groups: the disadvantaged people pertaining to the Asian community are more residentially clustered than others. Their distribution, unlike that of other ethnic groups, is not significantly related to the economic characteristics at area level but to the presence of other Asian residents in an area.

Keywords

urban deprivation residential segregation unconventional data charity GIS Poisson modelling

Introduction

Residential segregation – the spatial concentration of particular minority populations (Neal et al., 2013) – is widely found in urban contexts. The nature of the phenomenon and its implications have been extensively studied since the fundamental contributions of the so-called ‘Chicago school’ (Park et al., 1925; Shaw and McKay, 1942). Newer contributions aim to better understand the causes of segregation and its impact, and to improve the methods employed to quantify it (Piekut et al., 2019).

Nevertheless, it is difficult to understand the causes that drive individuals from specific groups to locate in certain areas of the city by referring solely to official sources of data. The measures used to quantify segregation (such as the index of dissimilarity, Duncan and Duncan, 1955, and its many improvements, see Harris, 2014; or Simpson, 2004, for critical reviews) are commonly built upon censuses and other official sources of areal data. But there are limitations to those data. First, many official sources of data do not allow micro-level or point-level analysis (Oliver and Wong, 2003; Rosenbaum et al., 2002; Sager, 2012). The use of administrative areas as units of analysis has well-known shortcomings (Crowder et al., 2012; Hwang, 2014; Lee and Campbell, 1997; Lee et al., 2008; Massey and Denton, 1993; Wong, 2004) and the measures obtained from them might conceal asymmetry in the spatial patterns of different groups, and different levels of separateness at different scales (Dean et al., 2018). Since segregation can happen at various levels and for different groups of people asymmetrically, patterns of residential segregation can be diverse among sub-groups (Piekut et al., 2019). The use of area deprivation measures to classify the socioeconomic position of residents is subject to the fact that the aggregated information relating to a group of individuals may not reflect the characteristics of all individuals in that group (the so-called ecological fallacy, Macintyre et al., 1993).¹

This is especially true where the most vulnerable groups of people are concerned, such as the people seeking assistance from charities. In fact, deprived people are often invisible in official statistics for at least three main reasons: (1) those who do not have a permanent residence might escape census records;² (2) poverty measures based on relative income thresholds do not adequately capture non-monetary forms of deprivation, which is itself a multidimensional and complex concept not restricted to a lack of basic necessities nor of material benefits (Alkire and Santos, 2014; Fremstad, 2010); (3) some people may not reveal their actual condition because of social stigma but they may rely on assistance services for their basic needs (Pratschke, 2007).

This is why, within the social sciences, there has been a growing interest in the use of non-official sources of data to broaden our understanding of urban processes – as happens for the extensive use of Volunteered Geographic Information (VGI, Goodchild, 2007). Sometimes, these data fall under the vague but fashionable rubric of ‘big data’ yet the issue is less about size and more about having ‘alternative’– but information-rich – data and methods to extract useful information from them, to test and to develop urban theory.

In light of the above, this study aims to show how an unconventional source of data can be used to analyse the residential patterns of a particular group of disadvantaged residents and to understand the reasons behind the observed distribution. The group considered is made up of a non-homogeneous population. It is a group of people who, seeking the help of assistance centres, show various forms of deprivation: poverty, homelessness, illness, scarcity of relations, addiction, legal problems and so forth. Among this complexity, economic poverty is just one piece of the puzzle and conventional measures based on income thresholds fail to fully take into account such a ‘hard-to-count’ group of people (Saunders and Adelman, 2005; Townsend, 1987). Although this group is mostly made up of immigrants, it should be noted that there is no direct causal link between immigration and deprivation. Similarly, problems of loneliness or addiction concern not only disadvantaged residents but also the most affluent population groups.

The specific research questions this article aims to answer are: (a) is there evidence of segregation among the socially disadvantaged residents? (b) Do their residential patterns differ from the ones of the general population or according to their ethnicity? (c) Which area-level characteristics (social, economic or ethnic) are significantly related to the proportion of deprived residents in a particular area? The third aim implies that after observing different degrees of concentration, an investigation is needed to understand the causes underlying those differences, in particular since an interconnection exists between ethnicity, deprivation and segregation (Clark, 1991; Friedman, 2011; Piekut et al., 2019; Sager, 2012).

The research questions above can be answered thanks to the particular data used: they come from a well-established charity (Caritas Italiana, see section ‘Data, study population and methods’) which routinely collects information on visits to its assistance centres. By linking ecological data with personal, individual-level data this study offers a unique opportunity to understand the nature of residential segregation for interesting – yet often neglected – vulnerable groups.

The case study of Ancona, a medium-sized city located in Central Italy, extends the literature on urban deprivation in Italy, which is mostly devoted to metropolitan contexts (Benassi, 2005). Moreover, Ancona, as a railway junction central to the country, with maritime connections with Eastern Europe, represents an interesting case of attraction for a transient population that moves along the North–South axis of Italy or comes from Eastern countries. The original source of point-level data also represents a novelty in Italy and it indicates the way for future research exploiting data from charities and other organisations that – by collecting large, detailed and geo-referenced information on deprived people – can complement official measures effectively.

Residential segregation and the geography of deprivation

Residential segregation refers generally to the spatial separation of social groups within a specified geographic area, such as a municipality, neighbourhood or census tract. It is ‘the degree to which two or more groups live separately from one another, in different parts of the urban environment’ (Massey and Denton, 1993: 63). Groups can be residentially segregated on the basis of any characteristic (such as religion, ethnicity, socioeconomic status, etc.) and at any level of geography (such as residential blocks, provinces or regions). Since most children attend the school nearest to their home, residential segregation is often reproduced in school segregation, which in turn can lead to labour market segregation (Harris, 2014) and further forms of physical or socioeconomic separation (e.g. rental constraints, Turner et al., 2002; crime, Ellen, 2000).

Residential segregation of ethnic groups is dependent on people sorting into neighbourhoods based on their individual characteristics, the affordability of housing and preferences towards certain types of housing or neighbours (Johnston et al., 2002; O’Sullivan and Wong, 2007; Reibel and Regelson, 2011; Wong, 2003). Sometimes residential sorting arises as an unintended consequence of individual decisions, even when people prefer a degree of integration (Dean et al., 2018). The spatial concentration of low-income groups in neighbourhoods with affordable housing, for example, arises because of their limited financial resources, which restricts options in the housing market (Iceland et al., 2005), as well as a consequence of real estate agents and government officials closing down some markets and making them inaccessible to those groups. It has also been observed that vulnerable residents are prone to be ‘displaced’ from their neighbourhoods by gentrification, in-migration and redevelopment (Desmond and Gershenson, 2017). The quantification of displacement, however, has been shown to be problematic because of the lack of spatially and temporally disaggregated data (Easton et al., 2020). Physical barriers that reduce interaction and mobility among a city’s sub-areas – such as railway tracks, highways, natural obstacles – can also sometimes provide unintended starting conditions for segregation (Hipp et al., 2014).

At other times, segregation is somewhat promoted by dominant social groups, following the desire to avoid ‘certain neighbourhoods’ in the light of (still) common racial stereotypes (Borjas, 1997; Charles, 2000, 2003; Farley and Frey, 1994; Krysan, 2002). When assessed in terms of racial segregation, the issue has been inevitably subject to ethical criticisms (Simpson, 2004). Interestingly though, some minority groups proclaim to prefer to live in a segregated neighbourhood because they feel more comfortable among members of their own racial and ethnic group, not conforming to the theories that predict a progressive integration of communities (e.g. Park et al., 1925). At the same time, it is not unusual that wealthy populations concentrate within segregated and gated communities as well (Clark, 1991; Clark and Fossett, 2008; Mulder, 2007; Thernstrom and Thernstrom, 1997).

The clustering of people with similar characteristics has both positive and negative dimensions. Owing to physical proximity, people living in the same local community are assumed more likely to interact and cooperate with each other and form various groups of interest (Piekut et al., 2019). Within urban neighbourhoods, social networks and imitation among peers tend to be stronger (cf. peer group effect, Ellen and Turner, 1997; and neighbourhood effect, Sampson et al., 1997). That, in turn, was shown to be an important resource for inclusion and for the creation of social capital (Daconto, 2014). Sometimes, tight communities also have enhanced self-regulation and resilience capacities (Bauder, 2002).

On the downside, spatial unevenness reproduces and can aggravate social inequalities, being detrimental to the life chances of both adults and children (Galster and Sharkey, 2017). The fact of living in a segregated neighbourhood may have an impact, first, by exposing adolescents – more sensitive to external factors – to dysfunctional patterns and lifestyles; second, by reducing economic and employment opportunities, even out of the area of origin because of the negative reputation of disadvantaged neighbourhoods; and third, by undermining quality of life because of the scarcity of services (such as transport, Hine and Grieco, 2003; see also Brannstrom, 2004; Leventhal and Brooks-Gunn, 2000; Urban et al., 2009). Residential sorting might transform itself into ‘social exclusion’ when individuals do not take an active part in the community or cannot have access to good occupational positions (Harris, 2014; Massey and Fischer, 2000; Wong, 2003). Sometimes, the interaction between individuals reinforces negative outcomes (e.g. higher crime rates, Sampson et al., 1997) where poverty is highly concentrated (Atkinson and Kintrea, 2001; Galster, 2003; Jencks and Mayer, 1990).

Today, in spite of mixed evidence, urban housing in central neighbourhoods of many cities has become generally less affordable and accessible in North America and Europe, as a result of processes of urban transformation (Monkkonen et al., 2018; Tammaru et al., 2016). So, the share of low-income households has gradually diminished, leading to either the decentralisation of deprived people towards peripheral suburbs, or the presence of pockets of extremely deprived residents in inner-city neighbourhoods characterised by older social housing (see Piekut et al., 2019, and references therein).

There is, then, an important link between forms of deprivation and spatial segregation (Wacquant, 2004; Wilson, 1987). Multi-dimensional deprivation provides a wider view than income-based poverty measures. It consists of various forms of inadequacy of individuals with respect to the social system in which they live, which leads them to a state of suffering or, in general, to the absence of wellbeing (Alkire and Santos, 2014; Atkinson and Kintrea, 2001). Multi-dimensional deprivation estimates may include people who are income poor and would be considered to be in poverty by traditional uni-dimensional income measures but they also include people who may not be income poor but face hardships or deprivations in other areas of their lives. Multidimensional deprivation indices have been proposed by national governments and international institutions, encompassing dimensions such as health, education and living standards (housing, electricity, etc.).³ An extensive literature has investigated the individual and psychological characteristics that underlie multidimensional deprivation. Multidimensional deprivation can be the result of a reduced ‘endowment of capacities’ (Sen, 1997) or health-related problems (Allgood and Warren, 2003)⁴ but it is also related to other, less predictable, factors, such as the subjective perception of one’s own condition (Stranges, 2007) or institutional setting (Kazepov, 2010).

Geographical clustering of deprived people is commonly associated with economic, ethnic or physical segregation. A commonly used approach to identify individuals who live in socioeconomically deprived geographical areas is to use area-based measures, which incorporate multiple aspects of deprivation (Baker et al., 2013). These measures classify small areas using aggregated data about the characteristics of residents but they have various shortcomings (Noble et al., 2006). Moreover, technical difficulties pertaining to the endogeneity of choice variables, spurious correlations and unobservable factors weaken the validity of statistical models (Jencks and Mayer, 1990; Rivkin, 2001).

Recently, the increasing availability of geo-referenced data and computational power have facilitated the development of Geographic Information System (GIS) applications in the social sphere, and many maps of deprivation have been proposed at national or local level (Norman, 2010; Steele et al., 2017). Educational level (Minot and Baulch, 2004), health quality (Ravallion and Datt, 2002), wage differentials and household composition (Madden, 2003) explained the spatial clustering of deprivation at various scales. Kearns et al. (2000) used non-census-based indicators at small-area level; at point-level, Iwata and Karato (2007) represented the GIS network of homeless people living in streets and parks of the city of Osaka, Japan. In Italy, the spatial representations of deprivation are scarce – practically absent for mid- or small-sized cities – as a result of the limited availability of localised data. The spatial segregation of immigrants has been analysed with spatial autocorrelation techniques using municipal administrative data (Borruso and Murgante, 2012). Pratschke (2007) constructed a social deprivation index at the micro-area level using census data, showing differences in spatial patterns of deprivation in Milan (isolated spots), Rome (peripheral areas) and Naples (spread throughout the city). The analysis of urban deprivation was deepened by means of specific surveys in Milan, integrating data from official sources (Benassi, 2005; Braga and Corno, 2009; Zajczyk, 2003).

Data, study population and methods

The data used in this study came from the archive of a local Caritas listening centre (‘Centro d’Ascolto’– CdA). Caritas Italiana is an organisation of the CEI (Italian Bishops’ Conference, the permanent union of Catholic bishops in Italy) for the promotion of charity. Founded in 1971, it is the biggest charity in Italy, with the aim to educate people, families and communities towards the Christian sense of solidarity.⁵ It publishes an annual report on poverty and migration alongside other studies (e.g., see Caritas-Migrantes, 2015, or Caritas Italiana, 2019). The Caritas organisation is articulated in over 200 diocesan centres. Local listening centres and Observatories on Poverty and Resources routinely collect data on people who seek assistance. The reference assistance centre is located in Ancona, which is the chief town of Marche, a region located on Italy’s Adriatic coast. Ancona has about 100,000 inhabitants, an area of 125 km² and an average population density of 811 inhabitants per km². Its boundaries delimited the study area.

The sample obtained was made up of people who asked for assistance regarding multiple forms of deprivation: shelter, food and basic needs, income aids or supporting services such as schooling, job seeking assistance, help with the immigration process, etc. During interviews with the listening centres, Caritas volunteers recorded information on demographics, family conditions, needs and problems, and the main reasons for which help was requested. Individual records were anonymised before statistical analysis; residential addresses were geocoded and then analysed in aggregate form. More than 2000 people (2470) visited the CdA of Ancona, during 2014. However, it was possible to geocode only a small sub-sample (about 600 users) after having excluded residents outside the municipality (including transient population) and missing/inconsistent data.⁶

The people who sought help from the charity were mostly non-Italian citizens (72.7%) (Table 1). Those non-Europeans were Africans (28% from Northern Africa and 23% from the rest of Africa), Asians (14%) and Latin-Americans (8%). The Italians were older (with a mean age of 50.7), while the Africans were the youngest (39.3). The proportion of females was exceptionally high for some Eastern European or African countries (Ukraine 94%, Nigeria 80%, Ghana 78%) and low for some Asian countries, such as Bangladesh, 27%.⁷

Table 1.

Number and share of deprived people by nationality, gender and mean age.

Country of origin	Number	%	Female	Female %	Mean age
Italy	155	27.3	70	45.2	51.0
Morocco	56	9.9	34	60.7	41.8
Tunisia	47	8.3	17	36.2	43.0
Romania	44	7.8	27	61.4	38.6
Bangladesh	29	5.1	8	27.6	38.1
Albania	25	4.4	18	72.0	38.9
Ghana	23	4.1	18	78.3	36.8
Nigeria	20	3.5	16	80.0	36.8
Peru	18	3.2	13	72.2	46.3
Ukraine – other 40+ states	17	3.0	16	94.1	48.8
Total (non-ITA)	415	72.7	225	55.8	40.7
Total	570	100	298	53.0	43.6

Source: Author’s calculation from Caritas data.

The data showed that there was a multidimensional mix of needs and problems that included poverty and unemployment, housing, family, health conditions and addiction, and legal problems. The Africans and the Asians – more than other ethnic groups – showed problems in the immigration process (information on visas, rights of stay, general orientation, etc). The Asians also demanded educational services, mostly targeted at learning the Italian language. The above needs were related from a statistical point of view. For example, poverty, addiction and legal problems were correlated with one another (see correlation Table A1 in the Appendix). Those were also correlated with family problems, which, on the contrary, were inversely correlated with educational problems. So, in terms of statistical associations, problems could be grouped as, first, economic problems (poverty and unemployment); second, immigration process and education; third, family problems, legal problems and addiction; fourth, housing.

In order to link individual-level data with the contextual features of the areas in which the subjects live, the data obtained from the charity were integrated by linking them with census tract data (the smallest areal unit for which official data were released). Two common factors were extracted from census data by means of a factor analysis (see Greene, 2008). The extracted factors could be tagged as a social deprivation factor (share of old people in the tract, divorced or widowed, living alone, out of the labour force, low level of education) and an economic deprivation factor (people unemployed, with a large family, living on rent, in buildings in bad condition) (see Tables A2 and A3 in the Appendix for factor analysis results).

A range of quantitative methods were adopted to analyse the residential patterns of the various ethnic groups and to understand how they were related to the characteristics of the neighbourhoods. From a descriptive point of view, the locations of the deprived residents were mapped using distributional mean centres and Standard Deviation ellipses to obtain measures of ‘compactness’ of their distribution (see O’Sullivan and Unwin, 2003). The visual analysis was then confirmed by comparing the concentration profiles (share of population for each census tracts decile) among the various ethnic groups and general population (see Johnston et al., 2002).

Finally, a Generalised Linear Model (GLM) regression with Poisson family and logarithmic link⁸ was computed to estimate the expected count of deprived people in each census tract. To compare the regression coefficients in terms of percentage-scale impacts, the Incidence Rate Ratio (IRR)⁹ was also reported. In conclusion, common post-estimation and goodness-of-fit tests were conducted (for a fuller description of the methodologies and further insights, see Greene, 2008).

Locational patterns

The majority of the people who sought assistance from the charity were located around the centre of the city (the elbow-shaped area to the north), with two agglomerations in hamlets to the west (Figure 1).

The number of points in the outer areas was very low. With respect to the total population, deprived residents were more compactly agglomerated in the central neighbourhoods (Figure 2). Their presence is particularly high there because of lower property and rental prices, and proximity to the infrastructural hubs such as the railway station and the bus terminal. With respect to the general population, the relatively unpopulated area surrounding the harbour (north of the map) contains a comparatively high number of deprived residents (see Figure A.1, available online). Overall, out of the total 722 census tracts of Ancona, 288 of them (33.2%) contained at least one of the deprived residents recorded by Caritas.

Other densely populated neighbourhoods, such as the north-eastern area along the coast or the residential south, had a lower density of deprived people.

With respect to ethnic and non-Italian citizen groups, the spatial distribution of Africans was slightly more clustered than that of Europeans (including Italians) or the overall sample, as it emerged from the size of the distributional ellipses of Figure 3. Asians were compactly clustered near the train station (their distributional ellipse was the smallest: 4.5 km long × 2 km wide). The South Americans were dispersed along the coastline.

The visual information was confirmed by the analysis of concentration profiles (share of population living in each census tract’s decile) for the various ethnic groups, in comparison with the total population (Table 2). Again, since half of the census tracts (5th decile) contained 100% of the Asian population (while, for example, it is necessary to reach the 8th decile to find 100% of the Europeans), the Asian ethnic group was the most concentrated. The results were proportionally similar for the deprived population: 52.9% of the deprived Asian population were in the first decile of the tracts, while approximately 30% of the deprived Africans, Europeans or Latin Americans were found here (Table 2).

Table 2.

Concentration profiles of ethnic groups (general population and deprived subjects).

Rank of census(deciles, %)	African		Asian		European		American
	Population	Deprived	Population	Deprived	Population	Deprived	Population	Deprived
10	53.8	30.1	62.6	52.9	29.1	27.4	45.9	28.1
20	80.5	57.0	83.1	66.7	50.4	43.2	70.3	43.8
30	94.9	74.2	93.9	76.5	67.8	60.0	85.7	50.0
40	100.0	81.2	99.1	84.3	79.7	75.8	95.6	65.6
50	100.0	84.4	100.0	86.3	88.8	84.2	100.0	68.8
60	100.0	86.6	100.0	86.3	95.7	88.4	100.0	68.8
70	100.0	89.8	100.0	88.2	99.4	92.6	100.0	78.1
80	100.0	94.1	100.0	92.2	100.0	96.8	100.0	78.1
90	100.0	96.8	100.0	96.1	100.0	97.9	100.0	93.8
100	100.0	100.0	100.0	100.0	100.0	100.0	100.0	100.0

Modelling residential segregation

The previous section showed that the spatial distribution of deprived residents was more clustered than that of the general population. Moreover, the residential patterns differed according to ethnic background. What could explain this? Were there any features at area level that might be associated with a higher presence of deprived residents? First, it was assessed whether the deprived people resided in the most socially and economically disadvantaged areas of town (buildings in bad condition, high-unemployment areas and so forth; compare the section ‘Data, study population and methods’ and Tables A2 and A3 in the Appendix for a full list of variables included in the deprivation factors). In addition to the characteristics at area level, it was also evaluated whether the presence of deprived people in a neighbourhood was affected by the relative shares of the various ethnic populations living in the same area.

To test the hypothesis above, four paired multivariate regression models were built, using the count of the number of users of the charity’s services in each census tract and for each ethnic group as the recurrent dependent variable. Independent variables indicating the relative share of people from the various ethnic groups were also included in the model. Where positive and significant coefficients are found, a relationship exists among the outcome variables (e.g. the count of those of deprived people of non-Italian African descent) and the relative share of people of the same nationality in the tract. A demographic control (relative share of females) was also included, together with the extracted deprivation factors (see section ‘Data, study population and methods’). According to the model, a unit increase in the coefficient of the economic deprivation factor corresponds to an increase of 0.15 in the expected log-count of deprived African and 0.23 of deprived Europeans, and so forth (Table 3).

Table 3.

GLM/Poisson regression model. Dependent variable: Count of deprived people in census tracts for each ethnic group.

	African (1)			Asian (2)			European (3)			American (4)
	Coeff.	IRR (std)	S.E.	Coeff.	IRR (std)	S.E.	Coeff.	IRR (std)	S.E.	Coeff.	IRR (std)	S.E.
Deprivation factors:
f1_Social	0.23*	1.26	0.09	0.41**	1.51	0.15	0.29*	1.33	0.12	0.29	1.34	0.21
f2_Economic	0.15*	1.17	0.07	0.19	1.21	0.13	0.23**	1.25	0.08	0.19	1.21	0.15
Peer group (share of people in census tract):
African (%)	3.14**	1.13	1.06	0.16	1.01	3.06	1.49	1.06	1.74	−2.02	0.92	4.71
Asian (%)	3.44***	1.26	0.70	5.26***	1.43	1.08	2.45*	1.18	1.02	3.92**	1.31	1.52
European (%)	1.91**	1.16	0.59	2.26*	1.20	1.14	2.11*	1.18	0.82	1.53	1.13	1.63
American (%)	2.32	1.09	1.58	4.21	1.17	2.46	2.99	1.12	1.86	4.48	1.18	2.88
Female (%)	−0.57	0.57	0.78	0.27	1.31	1.60	0.27	1.30	1.19	−0.96	0.38	1.79
_constant	−1.54***	0.29	0.40	−3.50***	0.04	0.86	−2.61***	0.09	0.63	−3.06**	0.06	0.94

Notes: N = 722 census tracts, *p < 0.05, **p < 0.01, ***p < 0.001. Incidence Rate Ratio (IRR) refers to standardised coefficients.

Results for Model 4 (American) are significant at the 5% level only.

The social deprivation factor was significantly related with the expected number of the charity’s clients. In terms of IRRs, a unit increase of the social deprivation factor in a tract corresponded to a 26% increase in the expected count of deprived Africans, 33% of deprived Europeans and 51% of deprived Asians. The results for the Latin Americans were not statistically significant because of the low number of observations.

The economic deprivation factor was significant for deprived people of African and European origin (Table 3, columns 1 and 4) but it was not significant for deprived Asian residents (column 2). Instead, the relative share of the Asian population in the tract was highly significant: the expected number of deprived residents with Asian origin was 43% higher in a census tract in which the share of the Asian general population increased by 1%.

So, while the number of disadvantaged Africans and Europeans was influenced by the generic presence of other foreigners (not only by people of the same ethnic group), the presence of deprived Asians was influenced only by the share of other Asians in the tract. Moreover, there was no statistical association between the number of deprived Asians and the economic features of the neighbourhood.

Discussion and conclusions

Spatially disaggregated data are increasingly needed to study urban deprivation at small-area level. Several methodological issues suggest that micro-level phenomena might be concealed by analysing larger units of analysis, because of the nature of aggregation (Sager, 2012). Official sources often fail to meet the information needs of researchers, social operators and social policy makers, especially where disadvantaged subjects are concerned and in non-metropolitan contexts, where research applications are lacking. In Italy, the attention of researchers has, in fact, focused on metropolitan areas. However, the local context remains crucial to have both a comprehensive view of the phenomenon and to design policies, since social services are organised and implemented at the local institutional level.

Consequently, there is potential to make use of other sources of ‘non-official’ data. In this sense, voluntary-based data are a precious resource: charities which are spread throughout territories are able to intercept a large number of people in need. Such data do not meet the scientific standards of unbiased sampling but they do have the potential to offer fresh insights on otherwise ‘invisible’ populations. The tension between statistical orthodoxies and gaining knowledge from unconventional data is an increasing issue in the social sciences where ‘big data’, administrative data and other forms of harnessed or volunteered information challenge the boundaries of historical practices.

The self-selected and non-randomised nature of the sample and the low number of geo-referenced data represent limitations to this study, undermining causal inference. It is also possible that the information was ‘filtered’ by the operators’ view during the recording phase, and the detection of needs was dependent on the range of services offered, excluding the bearers of unexpressed needs from the sample. On the upside, Caritas volunteers are particularly open to the needs of clients and, thanks to the frequency of dialogue and interviews, they can build up relations and trust. So, more than municipal centres or other state offices, they are able to assess relevant information on needs and problems of the disadvantaged population they encounter.¹⁰ Hopefully, the quality of data is also likely to improve in the near future when all the Caritas centres will be provided with smart cards to automatically record user visits, thus eliminating errors in manual data entry.

The results of this case study first confirm the multidimensional character of deprivation, showing diversity in the needs and socioeconomic conditions of the disadvantaged people intercepted.¹¹ From a spatial perspective, deprived people reside in specific neighbourhoods, mostly located in the old central area of the city rather than in the suburbs (cf. section ‘Residential segregation and the geography of deprivation’, and Piekut et al., 2019). The ‘deprivation spots’ highlighted in this study are similar to those found in Milan, in spite of the different context (see Benassi, 2005; Pratschke, 2007). The compact residential clustering of the Asian community, already found elsewhere (Johnston et al., 2002), appears to hold in the case of the disadvantaged people analysed here as well. In this respect, the results of this study seem not to support the classic theories that predict the residential assimilation of longer established communities (e.g. Park et al., 1925) but to follow the idea that people in need search for affordable housing, which is often clustered in specific low-income, already highly populated neighbourhoods (Wessel and Nordvik, 2019).¹²

One assumed reason behind the observed patterns of segregation is economic (see Sager, 2012): the census tracts in which the sample of deprived residents live have in fact higher values in several deprivation-related indicators (i.e. areas with lower property and rental prices and closer to infrastructural hubs; cf. Hipp et al., 2014). However, for some of the ethnic groups analysed, such as non-Italian Asians, the economic conditions of the neighbourhood are not significant. According to the regression model, the proportion of deprived Asians is significantly related to the presence of other Asian residents in an area. Asians who live in more densely populated clusters appear to have a preference for living close to other community members. A sort of peer group effect might exist because of language, word-of-mouth, lifestyle habits, transfer of information and so forth. Furthermore, it is interesting to notice that deprived people of Asian origin asked the charity for educational services (language schools, etc.) more than other ethnicities, and language and education were shown to be factors related to segregation (Bayer et al., 2004). However, a comprehensive statistical investigation into the above reasons is not allowed by the available data.

Understanding the geographies of the deprived residents is vital for planning social services at the urban level and useful for policy in broader terms. In order to complement official information, charities, intercultural centres and other organisations could produce geo-referenced data routinely, possibly respecting adequate standards of quality. For example, since Caritas data included information on people’s needs and problems, they uniquely allow for problem co-location analyses, such as exploring geo-patterns of alcohol or drug abuse, disability, etc. (see Macdonald et al., 2018). Yet, even in its own right, understanding the causes of the agglomeration of the various ethnic groups is particularly useful for researchers, with point-level mappings particularly effective as a tool for representing deprived people’s needs and putting the ‘invisibles’ on the map.

Footnotes

Appendix

Table A4.

Descriptive statistics of variables used in the GLM regression model.

	Mean	Std. Dev.	Min	Max
Share of foreigners on total population, by origin (ISTAT census):
- Africans	0.015	0.039	0	0.56
- Asians	0.029	0.068	0	0.55
- Europeans	0.048	0.071	0	0.82
- American	0.017	0.038	0	0.50
Share of females on total population	0.501	0.137	0.20	0.83
Count of Caritas users in each census tract, by origin:
- Africans	0.258	0.670	0	6
- Asians	0.071	0.340	0	5
- Europeans	0.132	0.486	0	5
- Europeans (including Italians)	0.332	0.840	0	6
- Americans	0.044	0.237	0	3

Acknowledgements

The author wishes to thank Prof. Richard Harris for his contribution to an earlier draft of this article, and for his support and motivation during a visit to the School of Geographical Science, Bristol, UK. The author is also grateful to Andrea Tondi, Simone Breccia and all the volunteers at the Caritas centre of Ancona for the availability of data and for the important work that they do.

The edits of Prof. Phil Hubbard have been fundamental to realising the final version of the paper.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Francesco Balducci

Notes

References

Alkire

Santos

(2014) Measuring acute poverty in the developing world: Robustness and scope of the multidimensional poverty index. World Development 59: 251–274.

Allgood

Warren

(2003) The duration of homelessness: Evidence from a national survey. Journal of Housing Economics 12(4): 273–290.

Atkinson

Kintrea

(2001) Disentangling area effects: Evidence from deprived and non-deprived neighbourhoods. Urban Studies 38(12): 2277–2298.

Baker

Mitchell

Pell

(2013) Cross-sectional study of ethnic differences in the utility of area deprivation measures to target socioeconomically deprived individuals. Social Science & Medicine 85: 27–31.

Bauder

(2002) Neighbourhood effects and cultural exclusion. Urban Studies 39(1): 85–93.

Bayer

McMillan

Rueben

(2004) What drives racial segregation? New evidence using census microdata. Journal of Urban Economics 56: 514–535.

Benassi

(2005) La Povertà Come Condizione E Come Percezione. Una Survey A Milano. Milan: Franco Angeli.

Borjas

(1997) To ghetto or not to ghetto: Ethnicity and residential segregation. Journal of Urban Economics 44(2): 228–253.

Borruso

Murgante

(2012) Analyzing migration phenomena with spatial autocorrelation techniques. In: Murgante

, et al. (eds) ICCSA 2012, Part II, LNCS 7334. Berlin and Heidelberg: Springer, pp. 670–685.

10.

Braga

Corno

(2009) Being an homeless: Evidence from Italy. Working Paper No. 2009-17, Università Degli Studi Di Milano, Dipartimento Di Scienze Economiche Aziendali E Statistiche.

11.

Brannstrom

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

12.

Caritas Italiana (2019) Flash Report Su Povertà E Esclusione Sociale. Available at: http://s2ew.caritasitaliana.it/materiali/studi_ricerche/flash_report_2019/Flash_Report_Caritas_Italiana_2019.pdf (accessed 17 April 2020).

13.

Caritas-Migrantes (2009–2015) RapportoImmigrazione. Latest report (in italian) available at: https://www.migrantes.it/xxviii-rapporto-immigrazione-caritas-migrantes/

14.

Charles

(2000) Neighborhood racial-composition preferences: Evidence from a multiethnic metropolis. Social Problems 47: 379–407.

15.

Charles

(2003) The dynamics of racial residential segregation. Annual Review of Sociology 29: 167–207.

16.

Clark

WAV

(1991) Residential preferences and neighborhood racial segregation: A test of the Schelling model. Demography 28: 1–19.

17.

Clark

WAV

Fossett

(2008) Understanding the social context of the Schelling segregation model. Proceedings of the National Academy of Sciences (PNAS) 105(11): 4109–4114.

18.

Crowder

Pais

South

(2012) Neighborhood diversity, metropolitan constraints, and household migration. American Sociological Review 77(3): 325–353.

19.

Cutter

Boruff

Shirley

(2003) Social vulnerability to environmental hazards. Social Science Quarterly 84: 242–261.

20.

Daconto

(2014) Living in disadvantaged neighbourhoods: Only negative effects? The role of the spatial dimension in the French urban sensible areas. Sociologia Urbana E Rurale 103: 99–114.

21.

Dean

Dong

Piekut

, et al. (2018) Frontiers in residential segregation: Understanding neighbourhood boundaries and their impacts. Tijdschrift voor Economische en Sociale Geografie 110(3): 271–288.

22.

Desmond

Gershenson

(2017) Who gets evicted? Assessing individual, neighborhood, and network factors. Social Science Research 62: 362–377.

23.

Duncan

(1955) A methodological analysis of segregation indexes. American Sociological Review 20: 210–217.

24.

Easton

Lees

Hubbard

, et al. (2020) Measuring and mapping displacement: The problem of quantification in the battle against gentrification. Urban Studies 57(2): 286–306.

25.

Ellen

(2000) Sharing America’s Neighborhoods: The Prospects for Stable Racial Integration. Cambridge, MA: Harvard University Press.

26.

Ellen

Turner

(1997) Does neighborhood matter? Assessing recent evidence. Housing Policy Debate 8(4): 833–866.

27.

Farley

Frey

(1994) Changes in segregation of Whites from Blacks during the 1980s: Small steps toward a more integrated society. American Sociological Review 59(1): 23–45.

28.

Fremstad

(2010) A Modern Framework for Measuring Poverty and Basic Economic Security. Washington, DC: Center for Economic and Policy Research.

29.

Friedman

(2011) Bringing proximate neighbours into the study of US residential segregation. Urban Studies 48(4): 611–639.

30.

Galster

(2003) Investigating behavioural impacts of poor neighborhoods: Towards new data and analytic strategies. Housing Studies 18(6): 893–914.

31.

Galster

Sharkey

(2017) Spatial foundations of inequality: A conceptual model and empirical overview. Russell Sage Foundation Journal of the Social Sciences 3: 1–33.

32.

Goodchild

(2007) Citizens as sensors: The world of volunteered geography. Geojournal 69(4): 211–221.

33.

Greene

(2008) Econometric Analysis. 6th edition. Upper Saddle River, NJ: Prentice Hall.

34.

Harris

(2014) Measuring changing ethnic separations in England: A spatial discontinuity approach. Environment and Planning A 46: 2243–2261.

35.

Hine

Grieco

(2003) Scatters and clusters in time and space: Implications for delivering integrated and inclusive transport. Transport Policy 10(4): 299–306.

36.

Hipp

Corcoran

Wickes

, et al. (2014) Examining the social porosity of environmental features on neighborhood sociability and attachment. PLoS One 9: 1–13.

37.

Hwang

(2014) The social construction of a gentrifying neighborhood: Reifying and redefining identity and boundaries in inequality. Urban Affairs Review 52: 98–128.

38.

Iceland

Sharpe

Steinmetz

(2005) Class differences in African American residential patterns in U.S. metropolitan areas, 1990–2000. Social Science Research 34: 252–266.

39.

Istat (2014) La Ricerca Nazionale Sulla Condizione Delle Persone Senza Dimora In Italia, Collana: Letture statistiche – Metodi. Available at: https://www.istat.it/it/archivio/127256 (accessed 17 April 2020).

40.

ISTAT (2016) La Povertà In Italia. Statistiche report. Available at: https://www.istat.it/it/files//2017/07/Report_Povert%C3%A0_2016.pdf (accessed 17 April 2020).

41.

Iwata

Karato

(2007) Homeless networks and geographic concentration: Evidence from Osaka City. CIRJE Discussion Papers. Available at: http://www.cirje.e.u-tokyo.ac.jp/research/dp/2007/2007cf527.pdf (accessed 17 April 2020).

42.

Jencks

Mayer

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

McGeary

MFH

(eds) Inner City Poverty in the United States. Washington, DC: National Academy Press.

43.

Johnston

Forrest

Poulsen

(2002) Are there ethnic enclaves/ghettos in English cities? Urban Studies 39: 591–618.

44.

Kazepov

(2010) Rescaling Social Policies: Towards Multilevel Governance in Europe (Vol. 38). Farnham: Ashgate.

45.

Kearns

Gibb

Mackay

(2000) Area deprivation in Scotland: A new assessment. Urban Studies 37(9): 1535–1559.

46.

Krysan

(2002) Whites who say they’d flee: Who are they, and why would they leave? Demography 39(4): 675–696.

47.

Lee

Campbell

(1997) Common ground? Urban neighborhoods as survey respondents see them. Social Science Quarterly 78: 922–936.

48.

Lee

Firebaugh

Matthews

, et al. (2008) Beyond the census tract: Patterns and determinants of racial segregation at multiple geographic scales. American Sociological Review 73(5): 766–791.

49.

Leventhal

Brooks-Gunn

(2000) The neighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin 126(2): 309.

50.

Macdonald

Olsen

Shortt

, et al. (2018) Do ‘environmental bads’ such as alcohol, fast food, tobacco, and gambling outlets cluster and co-locate in more deprived areas in Glasgow City, Scotland? Health & Place 51: 224–231.

51.

Macintyre

Maciver

Sooman

(1993) Area, class and health: Should we be focusing on places or people? Journal of Social Policy 22(2): 213e234.

52.

Madden

(2003) The changing spatial concentration of income and poverty among suburbs of large US metropolitan areas. Urban Studies 40(3): 481–503.

53.

Massey

Denton

(1993) American Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.

54.

Massey

Fischer

(2000) How segregation concentrates poverty. Ethnic and Racial Studies 23(4): 670–691.

55.

Minot

Baulch

(2004) Poverty mapping with aggregate census data. Research Paper No. 2004/38. World Institute for Development and Economics Research (WIDER).

56.

Monkkonen

Comandon

Escamilla

JAM

, et al. (2018) Urban sprawl and the growing geographic scale of segregation in Mexico, 1990–2010. Habitat International 73: 89–95.

57.

Mulder

(2007) The family context and residential choice: A challenge for new research. Population, Space and Place 13(4): 265–278.

58.

Murgante

Borruso

(2012) Analyzing migration phenomena with spatial autocorrelation techniques. In: Murgante

, et al. (eds) Computational Science and Its Applications – ICCSA 2012, Part II, LNCS 7334. Berlin and Heidelberg: Springer, pp. 670–685.

59.

Neal

Bennett

Cochrane

, et al. (2013) Living multiculture: Understanding the new spatial and social relations of ethnicity and multiculture in England. Environment and Planning C: Government and Policy 31: 308–323.

60.

Noble

Wright

Smith

, et al. (2006) Measuring multiple deprivation at the small-area level. Environment and Planning A: Economy and Space 38(1): 169–185.

61.

Norman

(2010) Applied spatial analysis & policy: Special issue: Poverty and deprivation mapping. Applied Spatial Analysis and Policy 3(2): 77–79.

62.

Oliver

Wong

(2003) Intergroup prejudice in multiethnic settings. American Journal of Political Science 47: 567–582.

63.

O’Sullivan

Unwin

(2003) Geographic Information Analysis. Hoboken, NJ: John Wiley and Sons.

64.

O’Sullivan

Wong

DWS

(2007) A surface-based approach to measuring spatial segregation. Geographical Analysis 39: 147–168.

65.

Park

Burgess

Mckenzie

(1925) The City. Chicago, IL: University of Chicago Press.

66.

Piekut

Pryce

Van Gent

(2019) Segregation in the twenty first century: Processes, complexities and future directions. Tijdschrift voor economische en sociale geografie 110(3): 225–234.

67.

Pratschke

(2007) L’articolazione Territoriale Dello Svantaggio Sociale In Italia: Una Nuova Misura Multidimensionale Per Piccole Aree Geografiche. In: Brandolini

Saraceno

(a cura di) Povertà E Benessere. Una Geografia Delle Disuguaglianze In Italia. Bologna: Il Mulin, pp. 123–146.

68.

Ravallion

Datt

(2002) Why has economic growth been more pro-poor in some states of India than others? Journal of Development Economics 68: 381–400.

69.

Reibel

Regelson

(2011) Neighborhood racial and ethnic change: The time dimension in segregation. Urban Geography 32: 360–382.

70.

Rivkin

(2001) Tiebout sorting, aggregation and the estimation of peer group effects. Economics of Education Review 20(3): 201–209.

71.

Rosenbaum

Reynolds

Deluca

(2002) How do places matter? The geography of opportunity, self-efficacy, and a look inside the black box of residential mobility. Housing Studies 17: 71–82.

72.

Sager

(2012) Residential segregation and socioeconomic neighbourhood sorting: Evidence at the micro-neighbourhood level for migrant groups in Germany. Urban Studies 49(12): 2617–2632.

73.

Sampson

Raudenbush

Earls

(1997) Neighborhoods and violent crime: A multilevel study of collective efficacy. Science 277(5328): 918–924.

74.

Saunders

Adelman

(2005) Income Poverty, Deprivation and Exclusion: A Comparative Study of Australia and Britain. Sydney: Social Policy Research Centre.

75.

Sen

(1997) Choice, Welfare and Measurement. Cambridge, MA: Harvard University Press.

76.

Shaw

McKay

(1942) Juvenile Delinquency and Urban Areas. Chicago, IL: University of Chicago Press.

77.

Simpson

(2004) Statistics of racial segregation: Measures, evidence and policy. Urban Studies 41(3): 661–681.

78.

Steele

Sundsøy

Pezzulo

, et al. (2017) Mapping poverty using mobile phone and satellite data. Journal of the Royal Society Interface 14(127): 20160690.

79.

Stranges

(2007) Il Divario Tra Percezione E Condizione Di Povertà In Italia. Rassegna Italiana Di Sociologia 48(2): 315–342.

80.

Tammaru

Marcińczak

van Ham

, et al. (eds) (2016) Socio-Economic Segregation in European Capital Cities: East Meets West. London: Routledge.

81.

Thernstrom

(1997) America in Black and White: One Nation, Indivisible. New York, NY: Simon and Schuster.

82.

Townsend

(1987) Deprivation. Journal of Social Policy 16(2): 125–146.

83.

Turner

Ross

Galster

, et al. (2002) Discrimination in Metropolitan Housing Markets: National Results from Phase I HDS 2000. Washington, DC: The Urban Institute.

84.

Urban

Lewin-Bizan

Lerner

(2009) The role of neighborhood ecological assets and activity involvement in youth developmental outcomes: Differential impacts of asset poor and asset rich neighborhoods. Journal of Applied Developmental Psychology 30(5): 601–614.

85.

Wacquant

(2004) What is a ghetto? Building a sociological concept. Revista De Sociologia E Política (23): 155–164.

86.

Wessel

Nordvik

(2019) Mixed neighbourhoods and native out-mobility in the Oslo region: The importance of parenthood. Urban Studies 56(5): 885–905.

87.

Wilson

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

88.

Wong

DWS

(2003) Implementing measures of spatial segregation in GIS. Computers, Environment and Urban Systems 27: 53–70.

89.

Wong

DWS

(2004) Comparing traditional and spatial segregation measures: A spatial scale perspective. Urban Geography 25(1): 66–82.

90.

Zajczyk

(2003) La Povertà A Milano. Distribuzione Territoriale Servizi Sociali e Problema Abitativo. Milan: Franco Angeli.