South Asian Diaspora in Italy: settlement patterns and locational factors

The spread of international migration is one of the most salient social and political issues in a growing number of OECD regions facing self-reinforcing demographic shrinkage and population ageing (Hugo and Morén-Alegret, 2008). Policymakers and academics debate the opportunities and challenges posed by this unprecedented mass migration, particularly about the ability to replace the working-age population, whose loss threatens the viability of health, welfare and production systems (Strozza et al., 2016; Livi Bacci, 2018).

Some scholars argue that international migration enhances economic growth, productivity (Michelangeli et al., 2019), innovation (Poot, 2008) and entrepreneurship (Livi Bacci, 2018); it fills in for labour shortages in agriculture, construction, manufacturing and tourism sectors and provides care/domestic services (Mingione, 2009). Migrants also reduce territorial imbalances and keep basic services such as schools, hospitals and childcare facilities afloat in sparsely populated areas (Hugo and Morén-Alegret, 2008).

Another stream of thought is more sceptical about the long-term replacement effect of immigration. In this perspective, foreigners represent only a temporary and unstable presence (Bayona-i-Carrasco and Gil-Alonso, 2012) that increases the existing territorial imbalances because of their propensity to locate in urban settings or dynamic labour markets (Martín Martín and Jiménez Aguilera, 2017). Moreover, immigrants fuel social resentment among the poorer segment of the native population, which perceives them as competitors both in the labour market (Kim, 2020) and in access to public assistance (Jakubiak, 2019). It also perpetuates obsolete industries, based on low-skilled, low-salaries jobs, delaying their restructuring toward more efficient models (Garzón, 2015). Finally, foreigners belonging to culturally distant communities are a source of concern about social and cultural cohesion (Simpson et al., 2008), especially if concentrated in few settlements (Brezzi et al., 2010).

Such debate cannot ignore the increasing complexity of contemporary migration as compared to earlier waves (McAreavey and Argent, 2018). Some scholars emphasize the superdiversity (Vertovec, 2007) of current migratory flows, which are fragmented in a wide range of countries of origin and have spread out to new destinations, as Southern Europe (King, 2002), small towns (Garzón, 2015) and rural areas (Hugo and Morén-Alegret, 2008). The re-routing of migratory flows even beyond traditional urban gateways poses unprecedented challenges to local communities (McAreavey and Argent, 2018). Further complexity stems from recent shocks such as the 2008 economic crisis, which impacted foreigners particularly hard compared to natives (Colombo and Dalla-Zuanna, 2019), therefore changing the location patterns of some migrant communities (Kahanec and Guzi, 2017).

The spatial distribution of foreigners has therefore become a relevant research issue, which encompasses new destinations and migrant origins, and strongly influences their integration process in the host country (Lichter et al., 2020). Migrants’ location is determined by the interplay between local characteristics (Maza et al., 2013), and network effects exerted by the already-settled immigrants (Brezzi et al., 2010; Beine et al., 2011).

In this perspective, South Asian People (SAP) in Italy represent an intriguing case study, as they emphasize the benefits and drawbacks of immigration. They are generally considered hardworking, with strong entrepreneurial attitudes but scarcely integrated with the natives because of their concentrated settlement (Garha et al., 2016), the greater cultural barriers and the political mistrust after the events of 9/11 and the refugee crises (Hangartner et al., 2019).

On the other hand, Italy represents a meaningful field of study due to the large territorial imbalances, the polycentric settlements and the high aging population (Livi Bacci, 2018) which strongly pose the issue of replacement migration. It is the second destination in Europe after the UK for all the major SAP national groups, which are Indians, Sri Lankans, Bangladeshis and Pakistanis. Their size has strongly increased since 2000 and has persisted even after the 2008 crisis which asymmetrically affected some migrant groups (Venturini and Villosio, 2018). Besides, the lack of cultural, colonial, religious, linguistic ties or geographical proximity between Italy and South Asia allow to better conceptualise the role of local pull factors and ethnic networks in the location choices of migrants.

Nevertheless, SAP in Italy are quite neglected in economic literature. Analyses generally focus on sociological features as religion, identity, gender and family relationships. Little is said about the spatial distribution, and above all, comparative works between the national groups are missing.

The paper has three main purposes. First, we compared the spatial distribution of the four major SAP communities in Italian municipalities between 2004 and 2014. Then, we investigated the key drivers of settlement choices of SAP in Italy, distinguishing between a set of pull factors and network chains. Finally, we checked if the settlement pattern has changed over time, before and after the economic crisis.

Our paper seems to bring some novelties to the literature on migrant location. Firstly, we provided new insights on the relevant but under-researched issue of the SAP diaspora in Southern Europe. Secondly, we took a holistic approach to address the spatial, ethnic and temporal complexity of contemporary migratory settlements. We discussed Indian, Pakistani, Bangladeshi and Sri Lanka location models separately, identifying the key factors behind their settlement patterns. Thirdly, we expanded our analysis at a fine spatial scale, for all the Italian municipalities, to grasp the physical, functional and cultural interdependencies between rural and urban areas. Fourthly, we checked if and how the SAP spatial distribution changed over the 2008 crisis. Finally, the SAP location patterns are compared through a set of different indexes. We also incorporated a rich set of explanatory variables, from various disciplinary fields, and addressed the issue of spatial dependence, often overlooked in migration studies (Maza et al., 2013).

The rest of the paper is structured as follows: Section 2 introduces the main characteristics and the role of SAP diaspora in Italy; Section 3 starts the empirical analysis using indices of SAP location and segregation in Italy; Section 4 explores the factors driving the location of SAP in Italy through econometric models; Section 5 comments empirical findings and concludes.

The acronym SAP commonly refers to people originating from the Indian subcontinent. Most studies focus on the largest national groups which are Indians, Pakistanis, Bangladeshis and Sri Lankans. Similarities stem from their belonging to the British Empire until 1947, when Pakistan became independent, while Bangladesh emerged as a sovereign state in 1971. Besides their historical roots, SAP share the affiliation with transnational networks, which is enforced through remittances, social norms and cultural practices. The presence of a collective identity and ongoing ties with the country of origin is a distinctive feature of a diaspora (Constant and Zimmermann, 2016; Epstein and Heizler, 2016).

However, the term “South Asians” also promotes a misleading sense of homogeneity among communities sharing some social structures but with substantially different settlements and economic profiles (Peach, 2006).

The increasing polarization between social and economic integration (Peach, 2006) makes SAP a paradigmatic example of the challenges and opportunities of immigration. They combine a high employment rate, due to their hard-working and entrepreneurial attitude, with a huge risk of segregation in low-skilled occupational niches and in poor, unsafe and unqualified jobs (Fellini, 2018). SAP low integration in the host society is attributed to the more distinct and visible ethnicities, a closed and traditional social structure, settlement concentration and higher residential segregation compared to other foreigners (Garha et al., 2016).

Up till the 1970s, SAP were mainly headed to the United Kingdom, then the tightening of immigration policies reduced the entry opportunities and legal residence (Peach, 2006). As a result, growing flows of undocumented SAP have shifted across Europe. They have turned to Italy since the 1990s (Abenante, 2017) and have had a dramatic increase in the following decades ^[1]. According to ISTAT (Italian National Statistics Institute), the number of SAP in Italy has multiplied by almost seven times in the last 21 years, passing from 86 thousand residents in the 2002 to 595 thousand in the 2023. This marked growth involves all four major citizenships (Figure 1). SAP increased especially between 2002 and 2009, due to both new arrivals from abroad and to the regularisation of undocumented migrants who were already present in Italy. This steady growth slowed down but did not stop in the post-crisis period, unlike the total number of foreigners (Table 1). This is noteworthy evidence, especially if we consider the increase in Italian citizenship acquisitions.

Italy became a popular destination for SAP because it provided a “back door entrance” in Europe due to its flexible migration policy, lax entry controls and the periodic regularisation drives (Colombo and Dalla-Zuanna, 2019).

At present, Italy hosts the largest numbers of Indians (Garha, 2020), Pakistanis (Abenante, 2017), Bangladeshis (Morad and Sacchetto, 2020) and Sri Lankans (Benassi et al., 2023) in continental Europe. These main SAP national groups in Italy share similarities in the size of the population (Figure 1), the arrival time, the fast increase of the flows, the young age range and a male-oriented gender ratio ^[2], except for Sri Lankans, who are more gender-balanced (Benassi et al., 2023). SAP in Italy also share a poor level of education (Abenante, 2017), large remittances to the homeland, confirming their strong transnational ties (Morad and Sacchetto, 2020), the settlement concentration, the segregation from natives (Benassi et al., 2023) and the fragmented integration in the host society, limited to the economic sphere (Garha, 2020).

However, both the settlement models and employment structure considerably differ between the groups examined. Pakistanis are predominantly employed in industry and commerce, while Bangladeshis work in the hospitality and commercial sectors (Morad and Sacchetto, 2020). Both communities have high propensity to self-employment whereas they are not involved in elderly care and domestic services, which especially represents the main occupation for Sri Lankans (Benassi et al., 2023). Indians work in industry, in the services and in agriculture (Lum, 2012).

Despite its relevance, the SAP settlement in Italy has been poorly studied so far, and no one deals with this issue by means of a comparative perspective between the different countries of origin, unlike the UK (Peach, 2006) or Spain (Garha et al., 2016). Current studies focus on a single national group, as Indians (Garha, 2020), Sri Lankans (Benassi et al., 2023), Bangladeshis (Morad and Sacchetto, 2020) and Pakistanis (Abenante, 2017), or some selected Asian communities (Conti et al., 2023).

In addition, foreign settlement is mainly examined at national or regional levels, whereas the municipal scale has so far been explored through individual case studies, which provide detailed but fragmentary insights on few contexts, generally urban areas (Strozza et al., 2016; Benassi et al., 2023). Systematic and detailed comparative analyses of settlement patterns of national groups over time are still remarkably absent in literature.

In order to implement the empirical analysis, we collected data for the 8.091 Italian municipalities at 2011, which is our reference time. The municipal scale was preferred for data availability and because it is the smallest political unit with administrative power. Since samples change over time, due to municipal splitting or amalgamation, the population data were reclassified at the municipal boundaries relating to the year 2011. Variables on the geographical characteristics of the municipalities and their belonging to economic districts were also derived from the municipality classification of 2011 ^[3].

The analysis concerns two specific years: 2004 to 2014. This period is over the 2008 crisis and encompasses the rise, maturation and relocation of SAP communities in Italy (Figure 1). The years 2004 and 2014 are chosen based on data availability (and proximity to census surveys of 2001 and 2011, which provide exhaustive socioeconomic information at a small scale ^[4]. At the time of writing, detailed data for the 2021 census are not yet available.

The first part of the analysis concerns the computation and reading of spatial distribution indices. The Location Quotient helps to analyse the concentration of a specific ethnic group in each municipality (Brown and Chung, 2006):

where $xi$ represents the number of individuals belonging to an ethnic group x in the municipality i, while $popi$ is the population of municipality i, X is the number of individuals belonging to the ethnic group at national level and POP is the national population. The higher the index value, the higher the concentration of the community under analysis. However, this index suffers from the fact that some municipalities attract more migrants than others in general.

Hence, we also compute a standard Balassa Index:

where $immi$ is the number of immigrants in municipality i and IMM is the number of immigrants at national level. Unlike the previous one, this index assesses the “specialization” of municipalities in “attracting” the ethnic groups of interest.

The maps in Figures 2–7 in Appendix show these indices for migrants in general, for all SAP, and finally for each SAP national group. In the six figures, the colour gradations for the values of each of the three measures (Percentage on population, Location Quotient, Balassa Index) are kept constant to facilitate the reading of the data.

In summary, the maps show that the SAP are a small part of the foreigners present in Italy; however, their share has increased over time. We also note that individual SAP groups are highly localized in certain municipalities (much more than other migrants in general). However, different national groups follow different settlement patterns, reason why analysing SAP as if they were a unicum would be misleading.

In order to corroborate these first impressions, we then calculate other indices. The Dissimilarity Index is a measure of the evenness with which two groups are distributed across some geographic areas. A group is evenly distributed when each spatial unit has the same percentage of group members as the total population (Brown and Chung, 2006).

where $xi$ and $yi$ represent the number of individuals in each municipality i belonging to ethnic groups x and y, while X and Y represent their totals in Italy.

The Segregation Index is a particular specification of the Dissimilarity index, but with reference to the whole population, thus it measures how unevenly the population of an ethnic group is distributed in relation to the total population:

The results shown in Tables 2 and 3 suggest that SAP follow very different settlement patterns compared to both Italians and the whole migrant population. Moreover, each SAP community is concentrated in different cities. Hence, SAP are quite segregated. This dissimilarity persists over time, highlighting a substantial inertia in the settlement patterns (Conti et al., 2023). The location has not changed despite the 2008 crisis; this is a quite different result compared to other proveniences and migrants in general (Kahanec and Guzi, 2017).

The second target of the paper is to analyse the factors driving the settlement of SAP. The evidence provided in the previous section suggests that treating SAP as a unicum would be misleading, hence the analysis will be carried out separately for the four main communities.

Based on the reference literature, we use two approaches: the Settlement approach allows to explore the role of pull factors whereas the Inflows approach aims to investigate the role of the network chain. The analyses will be carried out for two years (2004 and 2014) to check whether the location factors have changed over time.

The Settlement approach follows Maza et al. (2013) ^[5], in which the percentage of residents of an ethnic group in the population depends on a set of local characteristics:

where:

1. Y_i is the share of the ethnic group under analysis over the total population in municipality i.

2. $μi,γi$ are the regional and province fixed effects of the municipality i, respectively.

3. θ is the vector of the C regressors, which are municipality features. Some regressors are time invariant (class of municipality, regional capital, provincial capital, coastal, mountain, district specialisation), others vary over time. Regarding the latter, the regressors used for the analysis in which the dependent variable is relative to 2004 were those from 2001 and for the analysis in which the dependent variable is relative to 2014 those from 2011. As disclosed, this is due to data availability.

4. Finally, $ρ$ is the spatial correlation of the dependent variable, and $λ$ the spatial correlation of the errors. By estimating these parameters, we further check the robustness of the results with respect to potential spatial autocorrelation biases. In particular, the models without $ρ$ e $λ$ are normal regressions; models with $ρ≠0$ and $λ=0$ are SAR (Spatial Autoregressive model); models with $ρ≠0$ and $λ≠0$ are SAC (Spatial Autoregressive Combined model); $wi,j$ is the “weight” between municipality i and j, where the weight is the inverse of the distance between the two municipalities (Darmofal, 2015)

The set of plausible explanatory variables (⁠$Xic$⁠) can be divided into several blocks: demographic variables, which include population density (Garzón, 2015), elderly dependency ratio (Colombo and Dalla-Zuanna, 2019), masculinity ratio and average household size; economic variables, which consist of the activity rate, the unemployment rate (Maza et al., 2013), families in economic distress, the economic district specialization (Martín Martín and Jiménez Aguilera, 2017), the employment share in agriculture, industry and services (Mingione, 2009), and the percentage of employment in different skill levels; among the social variables we consider education (Beine et al., 2011) and housing characteristics (Strozza et al., 2016); while the locational factors include province or region capital (Garzón, 2015), coastal or mountain municipalities (Viñuela et al., 2019), centrality or peripherality of the municipality (Royuela, 2015) and public transport usage.

To assess the role of network chains (i.e. the tendency of migrants to choose municipalities where individuals of the same group are already present), we will also estimate models based on an Inflows approach. For this specification, we follow Brezzi et al. (2010), in which the variation of the presence of an ethnic group is linked to a series of contextual characteristics and, where possible, we consider the variation over time for the regressors. “More robust inference of the effect of location characteristics on inflows could be obtained by studying how changes in levels of inflows are affected by changes in regional characteristics” (Brezzi et al., 2010, p. 10). We use the changes of the independent variables, apart from time invariant characters.

In this case, the model can be written as follows:

where:

1. $ΔSi$ is the inflows of the ethnic group under analysis in the municipality i, between 2004 and 2014;

2. $ϕiand ηi$ are the regional and province fixed effects of the municipality i, respectively;

3. $ω$ is the vector of the F regressors (municipality features) that do not vary over time. Among these regressors, we also consider, in some model specifications, the share of Indians, Pakistanis, Bangladeshis, Sri Lankans and other migrants among the population of municipality i at the beginning of the period (i.e. in 2004) because, following Brezzi et al. (2010), these initial quotas are the variables that capture the chain effect;

4. $φ$ is the vector of the C regressors (municipal features) varying over time (for the same reasons highlighted above, in this case, the variation is between 2001 and 2011). Obviously, caution is needed in interpreting the relationship between a change that occurred between 2001 and 2011 (varying regressors) and one that occurred between 2004 and 2014 (dependent variable). Nevertheless, what we are interested in analysing here is the “chain” effect while the other regressors should be considered more as “controls”. In fact, we will not comment on their results.

5. Finally, $ρ$ is the spatial correlation of the dependent variable, and $λ$ the spatial correlation of the errors, as in the Settlement approach.

More precisely, the dependent variable (⁠$ΔSi$⁠) is defined by the inflows of migrants of a given ethnic group between 2004 and 2014 in the municipality i, relative to the total population of the municipality i in 2004, for example:

It is useful to specify that the second model (Inflow approach) is not a “in first differences” version of the first model (Settlement approach). These are two distinct approaches, both found in literature: the first focuses on a level-based analysis and the second centres on an analysis of recent flows. From an interpretative standpoint, the first approach (Settlement) better captures long-term dynamics, while the second one (Inflows) focuses on short-term aspects in the time frame in which the variation in the population under analysis is studied.

From a technical point of view, our approach only allows us to analyse possible correlations whereas, in theory, it does not allow us to say anything about the presence of causality and, possibly, its direction. The limited municipal data would not allow the implementation of more advanced methodologies such as dynamic panel models or instrumental variables (in our specification we already use almost all the variables available at the municipal level). However, our approach is consistent with what is usually done in the literature concerning the issue of endogeneity and causality (Brezzi et al., 2010; Melguizo and Royuela, 2020). In the presence of a correlation, it is much more logical to think that the characteristics of municipalities attract certain migrants rather than to assume that the presence of those migrants may influence the overall social and economic characteristics of a municipality, especially with regards to the quite small SAP communities. The Inflow approach still includes some aspects of dynamics as it considers the variation of the dependent variable, the variation of some regressors and the lagged value of other regressors.

Moreover, with the Settlement approach many regressors are time-invariant and exogenous by definition (province capital, region capital, coastal and mountain character, degree of centre-periphery of the municipality, district economic specialization) and the others are lagged (2001 for 2004 analysis and 2011 for the 2014 analysis) due owing to the availability of data.

Tables A1 and A2 in Appendix A show summary statistics for the 8,091 municipalities included in the sample. From these tables, it is also possible to know the regressors included in the analysis in detail (⁠$Xic$ for the Settlement approach, $Xic$ and $Xif$ for the Inflow approach).

Detailed findings for the four communities are shown in Appendix B (Tables B1-B4). Table 4 provides a summary of the results, to facilitate their uniform reading and interpretation.

The analysis following the Settlement approach provides coherent results with those emerging from Section 4. Each SAP group shows different settlement factors. The only common feature is that SAP generally avoid mountain areas.

For Bangladeshis and Sri Lankans, what emerges is an urban model (they show a positive and significant correlation with capitals and population density), but with relevant differences. Sri Lankans locate in post-industrial cities, characterized by higher activity rates, public transport usage, less buildings in poor condition, more trade services, belonging to paper and printing districts. Such cities (i.e. Napoli, Milano, Verona, Firenze, Messina, Napoli and small municipalities in the surroundings) demand both medium-high skilled and low-level professions, and deputies to manual jobs. Bangladeshis prefer coastal, crowded, old industrial cities, with low average house areas, specialized in both traditional (leather and footwear) and chemical industries and with a higher share of employment in professions requiring a low level of competence. Among them, we note Bologna, Venezia and Jesolo, Ancona, Porto Recanati and Palermo.

The lack of correlation with regional capital or population density highlights that Indians settle outside of cities, often in rural areas characterized by higher average households and house size, higher masculinity ratio, lack of public transport and lower activity rate. In 2004, Indians chose to settle in industrial districts or municipalities with employment in agriculture while in 2014 the latter variable became not significant. Indians are the only SAP located in areas specialized in metallurgy, home goods or food industries such as the Parmesan cheese production in Viadana and Suzzara or Latina’s greenhouses in Anzio and Sabaudia.

Pakistanis prefer traditional industrial districts, located in medium towns, not rural (positive relation with population density) or urban settings (negative relation with masculinity ratio). The labour market is fully pivoted on manufacture, especially in the traditional or mechanical sectors in Brescia, Gallarate, Prato, Carpi, Novi di Modena and Corridonia. They locate in industrial districts with cheap housing, established ethnic networks, job opportunities in sectors with low linguistic, educational or financial barriers to entry.

Finally, the parameters related to the spatial correlation of the dependent variable and/or errors are always insignificant. This result aligns with the tendency of individual SAP communities to be concentrated in a few specific municipalities, which are not necessarily in close geographical proximity. Hence the dependent variable (presence of the ethnic group under analysis in neighbouring municipalities) and the error term (which generally captures what is not explained by the model’s regressors) exhibit no spatial correlation. This interpretation is also suggested by the indices and maps already presented, which reveal both a considerable dispersion of the four communities across the country and a concentration in specific municipalities.

The non-significance of the spatial correlation parameters results in highly similar values of the regressor parameters in models that include them or not. The standard errors, on the other hand, are slightly different in some cases. This may be due to the fact that some of the regressors still have a “spatial” dimension, i.e. they are linked with the geographical location of the municipalities. The spatial correlation parameters could therefore capture shared characteristics among some regressors, thereby influencing their standard errors and significance. This reading seems to be confirmed by the fact that in our results often (though not always) the regressors that see the greatest increase in standard errors are those relating to the peripheral or central condition of the municipality, whereas this is more rarely the case, for example, for regressors relating to sectoral employment characteristics.

Tables C1–C4 in Appendix C show detailed results for the Inflow approach Analysis. Table 5 presents only the results for the “chain” (correlation between inflows of a certain ethnic group and the presence of the various ethnic groups at the beginning of the period) because our model has some shortcomings due to the data availability: while the variation of the dependent variable is relative to the period 2004–2014, the variation of the dependent variables “time variant” can only be relative to the period 2001–2011, raising problems of interpretation. While it makes sense in the Settlement approach to consider regressor values relative to 2001 (2011) for a dependent variable of 2004 (2014), it makes less sense to regress a 2004–2014 variation on the 2001–2011 variations. We think, however, that these time variant variables can still be good “controls”. As already mentioned in the presentation of the two models, the Settlement approach is more effective in capturing long-term dynamics (which are the primary focus of our analysis). In the Inflow approach, however, the interpretation of the parameters is more complex as the variable “share of migrants already present” tends to encapsulated many of the effects linked to the characteristics of the municipality that attract the ethnic group under study. For all these reasons (and in the interest of article length) we limit our commentary to the regressors related to the “chain” (although others do not appear to contradict the results of the settlement approach either).

The Inflow approach shows the role of chain networks for all the four SAP groups. Indians and Sri Lankans have the tendency to locate close other migrants already present, thus reducing segregation. Pakistanis and Bangladeshis show a specific settlement model and the segregation from main migratory flows. We therefore confirm that the existing diaspora strongly attracts SAP, by providing them with basic information, employment, financial, linguistic and psychological assistance (Ukrayinchuk and Jayet, 2011; Beine, 2016).

The differences between the models with and without the spatial correlation parameters reflect the same observations discussed in relation to the Settlement approach.

SAP in Italy are more concentrated in some specific locations fairly distant form the natives (Garha, 2020). This settlement model reminds of the decentralised clustered distribution of the Chinese communities (Brown and Chung, 2006), which results from a mix of contextual pull factors and ethnic networks with a strong local character (Ukrayinchuk and Jayet, 2011). Among the former, the local economic specialization is a main pull factor for SAP (Benassi et al., 2023), who often filled the dramatic shortage of manual jobs nowadays unattractive for natives (Colombo and Dalla-Zuanna, 2019). SAP’s concentrated settlement is also a result of their wide cultural, geographical, religious and lifestyle distance from Italy.

However, the national communities exhibit striking differences in the location patterns, which are determined by different pull factors. Pakistanis move towards industrial districts (especially in some sectors such as leather and textiles), Indians also spread in rural areas, Bangladeshis settle in several “Banglatowns” usually close to major North Eastern industrial centres (Morad and Sacchetto, 2020) and Sri Lankans locate in the cities, even in the Southern regions.

In short, we found two settlement patterns: an urban model, split in a post-industrial (Sri Lankans) or old industrial (Bangladeshis) cities; and an industrial districts model, in medium towns (Pakistanis) or rural areas (Indians). Urban gateways have a powerful attraction over the newcomers, especially those involved in domestic services or commercial activities, because of more opportunities arising in the cities (Royuela, 2015). However, a noteworthy feature of the Italian case is the strong SAP presence in industrial districts, where they provide a cheap and efficient workforce or create ethnic firms ancillary to the local companies. Rather than outsourcing to low labour-cost countries, they play a key role in the onsite outsourcing strategy within the industrial districts, keeping their competitiveness on the global markets (Mingione, 2009).

We found evidence of the overall persistence of drivers over time, which generates stable location patterns (Ukrayinchuk and Jayet, 2011). The economic crisis seems not to have affected the geographical distribution of SAP national groups (Garha, 2020). The stability of spatial distribution is related to the rising role of chain networks, feed from family reunions and the persistent specialization in some occupational niches. Despite the concerns about possible competition between immigrant and local workers (Mingione, 2009), the labour market remained highly segmented even in the post-crisis period (Venturini and Villosio, 2018). Many foreign workers reacted to the higher unemployment risk by moving into different jobs and places (Kahanec and Guzi, 2017). On the contrary, the SAP were less penalized during the Great Recession: the specialization in some low-paid, labour-intensive jobs complementary to the natives (dairy activities for Indians, textile and leather for Pakistanis) intensified settlement concentration but also social segregation. High-skilled natives have a favourable attitude toward low-skilled SAP, but just in the economic sphere (Kim, 2020). In Italy, the crisis has thus reinforced the “low unemployment risk-no access to skilled jobs” pattern and settlement segregation (Fellini, 2018). Our findings are consistent with some recent evidence, according to which the Asian immigrants have shown falling unemployment and low inter-sectoral mobility during the Great Recession (Kahanec and Guzi, 2017).

The substantial inertia in the settlement pattern of SAP communities is also due to the high natural growth of young and prolific communities (Simpson et al., 2008) as well as the new arrivals, driven by the ethnic network and family reunion (Garha, 2020), less sensitive to the economic climate (Ukrayinchuk and Jayet, 2011). The presence of an existing diaspora plays a major role in shaping the SAP employment profile, spatial distribution and everyday life, where many key choices are a family, rather than an individual affair. The chain network favoured the inflows of unskilled immigrants, to whom it provided easier access to the job market (Beine, 2016). However, a thick ethnic network only helps to avoid but not to bear the linking costs such as language acquisition and knowledge of local customs and culture. So, the employment opportunities are achieved at the risk of being confined to occupational niches or weak self-employment (Venturini and Villosio, 2018) blocking the individual upward mobility, raising deep-rooted stereotypes (Indian-cattle, domestic Sri Lankans, Bengali-trade), promoting social segregation and, in the worst cases, forms of exploitation (Biggeri et al., 2022) and gang mastering.

In the long run, SAP’s location patterns could become a source of vulnerability. Concentrated settlements are less resilient to health and economic crises (Gaia and Baboukardos, 2023), feed the frustration of people who are relegated to stigmatizing and downgraded jobs (Della Puppa and King, 2019), hamper the acquisition of language and educational skills (Conti et al., 2023) and promote a precarious and fragmented integration (Fellini, 2018), restricted to the economic status. A large ethnic concentration may imply a separation, or even marginalization, from the host society (Epstein and Heizler, 2016), segregation in the deprived neighbourhoods (Ukrayinchuk and Jayet, 2011) and dependency on the local welfare system (Pellizzari, 2013).

The paper identified settlement patterns and its determinants of the SAP in Italy. We found a substantial persistence of location models over time, mainly due to the ethnic network or job specialization. This could imply a stabilization process that could provide long-term benefits to the host municipalities but also reinforce the existing spatial imbalances and slow down social integration. The persistence of SAP’s location patterns could even introduce some rigidity in the Italian labour market, forced to increasingly rely on foreign workers due to its aging population (Maza et al., 2013). In addition, SAP will have to face the next evolution of industrial districts toward a sustainable and inclusive development; such process will imply job upgrading and the need for more integration into the host community (Biggeri et al., 2022). The current severe work conditions and job exploitation experienced by SAP will not be accepted by the second generation of migrants. This could result in mounting frustration and lack of interest in permanent stay leading to joining the larger diaspora in English-speaking countries (Della Puppa and King, 2019).

Location concentration also fosters a fragmented social integration, where easy access to the job market is associated with more mistrust upon the issues of social and housing insertion. The poor integration harshly affects SAP women, who are doubly marginalised in both the domestic and social spheres, as evidenced by very low female employment rates, especially for Pakistanis and Bangladeshis (Abenante, 2017; Morad and Sacchetto, 2020).

We stressed how SAP have different settlement patterns and drivers, so they cannot be treated as a unicum. Our empirical findings thus help to inform effective and inclusive policies addressed to the multiple challenges of a superdiverse migration which calls for “place-based” responses, tailored to the specific needs of individual communities. For instance, we raised some potential concerns about segregated ethnic minorities, which are more vulnerable to economic crises. This is the case of Pakistanis or Bangladeshis, which are more segregated in industrial districts and thus they need appropriate policies aimed to bring out them from a social and occupational niche.

Our research is a preliminary step that calls for further extensions. Dataset could be updated at 2021 and include more variables; it could be even useful to separate first arrivals and internal displacement of SAP. These kind of mobilities have different settlements and drivers. What still needs to be understood is whether the increase in the foreign population mainly depends on the natural balance between already established people or on the newcomers from abroad (Simpson et al., 2008). The former hypothesis signals an ongoing stabilization process while the latter shows intense transnational connections.

Other limitations lie in the grounded methodological nationalism, which makes us to refer to national citizenship as a single homogeneous group. The national SAP communities consist of different migratory waves, highly heterogeneous in ethnic origins, skills and backgrounds. Settlement patterns can be conceptualised according to religion (Muñoz, 2010), caste affiliation or along ethnic lines such as the Indians, which divide among Punjabis, working in agriculture, and the Malayalees, more employed in domestic services (Lum, 2012).

In summary, SAP in Italy provide peculiar, significant opportunities and challenges. They are a young, stable presence, economically inserted in labour-intensive sectors but still largely culturally and socially alien to the host society. However, public opinion and politics are not yet able to understand the extent of the phenomenon, often abandoned to captious or stereotyped readings.

A qualitative leap is needed to overcome the social fragmentation that feeds economic vulnerability and pushes towards marginalization. SAP can play a key role not only in the preservation of the current industrial districts but also in their evolution towards a sustainable and inclusive model. Recognizing the role and the potentialities of immigration in Italy also means the introduction of active migratory policies, aimed to achieve a full integration. SAP’s location cannot only rely on the ethnic network, which often provides rigid and inefficient support, or on local job markets, which frequently result in exploitation of individual resources. Both the strategies are ineffective and unequal because they increase the vulnerability of the system, feed frustration and resentment which leads to re-emigration or protests, social separation and sometimes marginality. SAP in Italy need a full and not fragmented integration, through policies aimed to value the individual, as well as the worker.

The supplementary material (Appendices) for this article can be found online.