Indian immigration to Italy: concentration, internal mobility and economic crisis

The spatial concentration of immigrants has become a concern for local administrations in most European countries, as it is perceived as an obstacle to their better integration in host societies. Indian immigration to Italy began in the 1960s, but large-scale immigration of unskilled labour from Punjab and Haryana (NorthWest Indian states) began during the 1990s. This influx was formed by young men who entered Italy in search of economic opportunities and then concentrated spatially in the economically active regions. In this paper, using the municipal registers of inhabitants (Anagrafe), the residence permits records and the 2011 census data, firstly, I made a spatial analysis (LISA) of the settlement pattern of Indian immigrants in Italy; second, I explored the internal mobility pattern (through a gravity model) of Indian immigrants compared to other immigrant groups during 2005–2015; and finally, I studied the impact of the recent economic crisis on Indian immigration. ARTICLE HISTORY Received 2 August 2018 Accepted 14 February 2019


Introduction
At the dawn of the twenty-first century, Italy, which was traditionally perceived as an emigrant country, emerged as a major recipient of immigrants from 195 countries around the world, including India. According to the National Statistics Institutes (ISTAT) of Italy, the size of the immigrant population (on the basis of citizenship) in Italy increased from 1.12 million in the year 2000 to 5.03 (3.9 did not belong to the EU) in 2016. The immigrant population represents 8.3% of the total population of Italy. The largest immigrant communities in Italy are Romanians (22.9% of total immigrants), Albanians (9.3%), Moroccans (8.7%), Chinese (5.3%), Ukrainians (4.7%), the Philippines (3.2%) and Indians (2.9%). This large influx of immigrants in a very short time has brought a remarkable change in the composition of the host society. It has also posed challenges for the local administration, which was not fully prepared for the better management of immigrants (of different cultures and socioeconomic profiles) in the labour market, housing and the socio-cultural sphere (Einaudi 2007).
The Indian immigrants, which is the seventh largest group in Italy, represent only 2.99% of the total immigrant population. In 2016, there were 150,436 legal Indian immigrants living in Italy, of which 75% were Sikhs from Punjab and Haryana, 20% were Malayali, mainly Christian Catholics from the state of Kerala, and the remaining were Hindus from different parts of India (Lum 2012). After the United Kingdom, Italy has the largest number of registered Indian immigrants in Europe. Despite the rapidly growing number, this group received less attention in the current literature and academic research. The existing studies focus mainly on the description of this immigration process, the Sikh religion and the struggle for identity (Lum 2012;Gandolfi 2007;Denti, Ferrari, and Perocco 2005;Bertolani and Singh 2012), Punjabi culture, gender and family relations (Bonafanti 2015), and the economic contribution of the Indians to the Italian cheese industry (Sahai and Lum 2013;Bertolani 2005) and hospital services (Gallo 2005(Gallo , 2008. But even so, information on the demographic structure, spatial distribution and internal mobility of the Indians in Italy is scarce. In this paper, I will fill this gap in the current academic literature. Immigrants, mainly from developing countries like India, follow their social and kinship networks to settle in new destinations (Munshi 2014). It leads to their concentration in some parts of large cities, which are often characterised by low incomes and poor living conditions (Garha, López-Sala, and Domingo 2016). In previous studies, the minority concentration and residential segregation of the immigrant group have been perceived as an obstacle in their better integration in the host society (Wilson 1987;Peach 1997;Andersson 2007;Schönwälder 2007;Deborah 1998). According to the theories of socialisation, the concentration of immigrants made them less socialised in their new surroundings and due to the lack of good role models, they can isolate themselves from the host society (Musterd and Deurloo 2002). In addition, it prevents them from expanding their network of contacts by limiting them to their own group, which in turn contributes to their social isolation (Waldinger 1997;Jargowsky and Yang 2006). Several other authors also underlined the effect of the concentration of immigrants on their income, employment and socioeconomic status (Galster, Metzger, and Waite 1999;Clark and Drinkwater 2002). Moreover, in countries like Italy the recent economic crisis has an enormous impact on 'the dimensions and the characteristics of migratory flows, the occupational situation of foreign workers … , the processes of integration (Bonifazi and Marini 2011, 1). The precariousness caused by the economic crisis contributes to the concentration of immigrants in the deprived areas of big cities, which raises the concern of local governments regarding the formation of ghettos. As a result, the concentration of immigrants becomes a concern for local government in most European countries and it has become essential to study the settlement pattern of immigrants and their internal mobility to avoid the formation of immigrant neighbourhoods, especially when they are characterised with poor living conditions.
In this paper, my main hypotheses are: first, the settlement pattern of Indian immigrants in Italy guided by the availability of job opportunities in different parts of Italy and the presence of their social or kinship network, which provides assistance during the first phase of immigration and settlement; secondly, the internal migration of Indian immigrants during the last decade (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) worked as a double centre system, where the provinces of Brescia and Rome continued to be the centre of attraction and dispersion of Indian immigrants; and finally, the recent economic crisis has changed the size of influx and causes of Indian immigration to Italy, and the direction of their internal mobility. Therefore, the main objective of this paper is to explore the settlement pattern, level of concentration, internal mobility and the socio-demographic profile of Indian immigrants in Italy.
This paper is structured as follows: the second section presents the data sources and the methodology. The third section describes the main characteristics of the recent flow and population of Indian immigrants in Italy. The fourth section presents the spatial analysis of their settlement pattern in Italy. The fifth section compares the trends of internal migration of Indian immigrants with the entire immigrant population at the regional and provincial levels. The sixth section explores the impact of the recent economic crisis on Indian immigrants in Italy. And finally, the seventh section presents some key findings of this research.

Data sources
To measure the size and scope of the international and internal migration of Indian immigrants in Italy, I used the data from the Residence Variation Statistics (Iscrizioni e cancellazioni all'anagrafe per trasferimento di residenza in Italian, RVS) from 2003 to 2015. RVS captures the influx of new immigrants (on the basis of citizenship) and registers all internal movements of immigrants and the native population. It also collects information on the main socio-demographic characteristics of immigrants, such as citizenship, place and date of birth, gender, marital status and level of education. It is the most relevant data source to study the immigrant population, but it has also some limitations. For example, it only captures the regular immigrant population, since most of the Indian immigrants in Italy entered irregularly or become irregular after reaching Italy (Garha and Paparusso 2018), a significant part of them always remained uncounted; secondly, the data is only available up to the provincial level, therefore, the information on intra-provincial migration (at the municipal level) is not available.
To study the stock and settlement pattern of Indian immigrants, I used the data from the Municipal Register of inhabitants (Anagrafe) and the Residence Permit register (Permesso di Soggiorno). Both sources are administrative registers that collect information of all legal residents of the municipality, including immigrants. At the individual level, municipal register collects information about the age, sex, place of birth, nationality, municipality of residence and time of registration of all inhabitants. For the study of the immigrant population, its main limitations are: the data is only collected on the basis of current nationality. Therefore, it becomes very difficult to separate the Indian immigrants with Italian nationality from the native Italian population. Secondly, it does not gather information about irregular immigrants that leads to the underestimation of the Indian immigrant population. And finally, the immigrant population is very mobile and changes their residence frequently for jobs and other purposes, and often does not register at all places of residence. Therefore, it creates a big problem for the administration to keep an accurate record of the resident immigrant population in each municipality. Despite all shortcomings, it is the most reliable data source over the stock of immigrant population in Italy. The Residence Permit register, which is maintained by the police authorities, provides information about the legal status of the immigrant population (longterm or short-term residents) and the reasons of immigration. The main limitations of this database are: it always shows the high number of immigrants compared to the municipal register because many immigrants after receiving the residence permit move to other countries to work and stay (Garha and Paparusso 2018); secondly, immigrants can lose their residence permit, if they lose their job in Italy; and finally, a small number of immigrants with a residence permit issued never come to Italy for work. Despite all these shortcomings, it is the only register that provides us the causes of immigration and an estimation about the Indians with Italian permits living outside Italy.
To study other socio-demographic characteristics, I used the 2011 census of population and housing data. In Italy, the census is held every 10 years, so the last census was held in 2011. The census data includes the people of all nationalities, with a legal residence permit and habitual residence in the country. The data collected contains information on age, sex, household structure, level of education, economic activity, place of residence, country of birth, and nationality.

Methodology
The methods of the study had two parts. In the first part, a descriptive analysis of Indian immigration to Italy was done to describe the main characteristics of recent flow and stock of the Indian immigrant population to Italy. In the second part, I did spatial analysis with 'Global and Local tools of Spatial Autocorrelation'. Data on the immigrant population are available up to the municipal level, so in the present work, I consider municipality as the basic unit of spatial analysis. In 2016, Indian immigrants were settled in 4068 Italian municipalities. To examine the local autocorrelation, I used the index proposed by Luc Anselin (1995), i.e. LISA (local indicator of spatial association), which can be seen as the local equivalent of Moran's I. The sum of all local indices is proportional to the (global) value of Moran's statistic. The local value of a LISA is computed as: For each location, LISA values allow for the computation of its similarity with its neighbours and also to test its significance. As an operational definition, LISA is any statistics that satisfies two requirements: first, the LISA for each observation gives an indication of the extent of significant spatial clustering of similar values around that observation; second, the sum of LISAs for all observations is proportional to a global indicator of spatial association. Five scenarios may emerge: first, locations with high values with similar neighbours (high-high), also known as 'hot spots'; second, locations with low values with similar neighbours (low-low), also known as 'cold spots'; third, locations with high values with low-value neighbours (high-low), potential 'spatial outliers'; fourth, locations with low values with high-value neighbours (low-high), potential 'spatial outliers'; and fifth, locations with no significant local autocorrelation. These specific configurations can be first identified from a scatterplot showing observed values against the averaged value of their neighbours. This so-called Moran scatterplot is a useful exploratory tool. Once a significance level is set, values can also be plotted on a map to display the specific locations of hot spots and potential outliers.
To analyse the internal mobility, I used a 'gravity model'. In the previous studies, it is widely applied in the empirical analyses of the flow of goods and services, particularly within the field of international trade (Fotheringham and O'Kelly 1989). It also provides satisfying results in studies on internal and international migration flows to describe and predict the degree of interaction between geographical areas Zagaglia 2008, 2013;Casacchia et al. 2010b;Kim and Cohen 2010). The model considers the migratory flows, as directly proportional to the product of the masses (measured in terms of origin and destination resident population) and inversely proportional to the distance (or to a function of the distance) between the place of origin and the place of destination. The classic formulation of the gravity model is the following: where i is the area of origin and j is the area of destination, F ij is the migratory flow between i and j, P i and P j are the respective population masses, and d ij is a measure of the distance between i and j. We have computed the distance between the provincial geographical barycentre, adopting the triangular and spherical definitions of distance. The gravity model with multiplicative errors is then specified as follows: Considering the natural logarithm of both sides of the equation, and considering lnα equal to β 0 , the linear model is It may be estimated by a linear regression using the ordinary least squares method (OLS). Furthermore, the model has been modified considering in a simultaneous model by using dummy variables gender (men/women) and period of stay (before crisis or after crisis). With this modification, I obtained a simultaneous and comparable estimation of the effects of gender and economic crisis on the internal mobility of Indian immigrants.

Indian immigration to Italy: an overview
The pioneer Indians entered Italy as a British imperial army during the Second World War (Bedi 2011), but they did not settle permanently in Italy. In the 1960s, a regular flow of businessmen (related to the expansion of the commercial links of the Italian automotive industries in India), students of theology (in churches and catholic convents), priests, nurses and domestic workers entered Italy and began to settle here permanently (Sahai and Lum 2013). In the coming decades, due to their high demand, the influx of Malayali Catholic nurses (residents of the state of Kerala, India) increased significantly and they settled in Rome (Gallo 2005). The massive immigration of unskilled labour from the states of Punjab and Haryana began during the 1990s (Garha and Domingo 2017). Most of them were low-skilled irregular immigrants, who after the imposition of strict borders controls in the United Kingdom and other countries in northern Europe, decided to settle in southern Europe (Saha 2009). Due to the possibilities of regularisation, easy access to the labour market for undocumented immigrants in low-skilled occupations, fewer border controls, and the absence of deportation schemes, most of them attracted to Italy, Spain, Greece and Portugal (Garha and Domingo 2019). According to Sahai and Lum (2013), India immigration to Italy can be divided into three phases, i.e. leather and textile worker's phase , asylum seekers (1985)(1986)(1987)(1988)(1989)(1990)) and agricultural workers (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010). In all these phases, we can add a recent phase of family reunification (2010-present).

Flow and stock of Indian immigrants in Italy
According to ISTAT, the annual flow of Indian immigrants to Italy doubled in the last 14 years, from 5345 immigrants in 2003 to 11,762 in 2016 (Figure 1). This flow has witnessed a significant increase in the period 2003-2005, which was partly due to direct immigration from India, and partly due to the process of regularisation of irregular immigrants, who were already living irregularly in Italy. In 2011, the flow reached its peak with 15,067 immigrants. After 2011, the flowbegan to decrease and now it has been stabilised around 11,000 immigrants per year. The most important characteristics of this flow were its young age structure and male predominance. As in the North Indian society, males are still considered as bread winners, therefore, the initial flow was mainly composed of young males of working age groups. The number of females started to increase when these males started to reunite their families in Italy.
This regular influx of Indian immigrants has contributed to the formation of the seventh largest immigrant community in Italy. According to ISTAT, in 2003 the number of registered Indian immigrants in Italy was 35,518. It increased rapidly to 121,036 individuals in 2011. After 2011, under the impact of the economic crisis and the shortage of job opportunities, the size of the stock declined to 118,409 individuals in 2012. After this small decrease, it began to grow rapidly in the ensuing years and  Figure 1). Like the flow, the stock of Indian immigrants was also characterised by the predominance of young men. In 2011, the proportion of women in the total population was 40.5%. According to the age structure, two-third of the population was in the working age groups (16-49 years) and the proportion of youth (0-15 years) and the elderly (over 50 years) was of 30% and 3.9%, respectively ( Figure 2).

Socio-demographic profile of Indian immigrants
The Indians in Italy form a very diverse immigrant community in terms of origins, religions, causes of emigration, education, occupation, legal and civil status. The majority of the Indian immigrants in Italy had their origin in three states of India, namely, Kerala in the south, Punjab and Haryana in the north. Most emigrants from Kerala belong to the Catholic religion and migrated through the church to serve in the Christian institutes (Kodoth and Jacob 2013), while most of the emigrants from Punjab and Haryana belong to the Sikh religion and their main objective of immigration was economic (Garha and Paparusso 2018). According to the 2011 Census, 20% of the Indian immigrants in Italy were illiterate, and almost 40% had an education below the secondary level. The proportion of illiterates and university graduates among women was greater than men. 62% of the total Indian immigrant population (over 15 years old) was economically active, in which 90% had permanent or temporary employment and 10% were unemployed and looking for work. In the total active population, 28.8% were employed in agricultural, forestry and fishing activities, 37.8% in industrial jobs and 33.4% in services. The share of women in the total active population was only 14.3% and they were mainly dedicated to the service sector. In 2011, half of the Indian immigrants in Italy were married. The proportion of women in the never married category was 37.6% and that of divorced, separated and widows was less than 3%. The number of mixed couples (Indian men or women married to Italian groups or other immigrant groups) was less than 1% of the total population.
In Italy, the registration of immigrants in municipal register depends on their legal status, therefore, we do not have information about irregular immigrants of Indian origin. According to the Residence Permits data, on 1 January 2016, there were 169,394 residence permits granted to Indian immigrants, in which 45% were short-term and 55% were long-term permits. For the Indian immigrants, the process of naturalisation is very difficult. It requires a proof of uninterrupted regular legal residence for 10 consecutive years in Italy (only 18,420 Indian immigrants were naturalised in the last 4 years).

Spatial distribution of Indian immigrants in Italy
According to ISTAT, in 2016 most of the Indians were settled in the north-western part of Italy, where notably the Lombardy region was home to a third of the total number of Indian immigrants registered in Italy. This concentration correlated positively with the availability of low-skilled jobs in this region. The second main concentration was in the north-eastern part, where the Emilia-Romagna and Veneto regions were their major settlements. In the central part, the Lazio region, which includes the capital city of Rome, has a significant number of the Indian immigrants. Due to their poor economic condition and high unemployment, very few Indians have settled in the southern part of Italy. In the south, Campania and Calabria regions were home to a small number of Indian immigrants. Some of them also settled in the island provinces of Sicily and Sardinia (Table 1).
If we compare the territorial distribution of Indian immigrants with that of the native population and the total immigrants at the provincial level, in Latina the proportion of Indian immigrants was 20.7% of the total immigrant population and 1.74% of its total resident population. The provinces of Mantua and Cremona, with a very prosperous dairy and cheese industry, also had a significant proportion of Indians in their total immigrant population, i.e. 17.9% and 16.8%, respectively. Brescia, Bergamo, Vicenza and Parma regions also have a significant proportion of Indian immigrants in their total immigrant population (Table 1), the concentration of small industries in these provinces is the main reason for the high proportion of Indian immigrants in their total population. At the municipal level, the municipality of Rome was at the top with 9669 registered Indian immigrants, followed by the municipalities of Brescia (2197), Terracina (1922), Sabaudia (1835) and Aprilia (1211) (Figure 3). Except Brescia (in the Lombardy region), all other municipalities in the first five were part of the Lazio region. It shows the high concentration of Indian immigrants in a few municipalities in the centre of Italy. Recently, with a greater demand for manual labour in agriculture, the province of Latina emerged as a major destination for Indian immigrants.

Spatial analysis with LISA
To measure the concentration of municipalities with a high population of Indian immigrants, I applied a widely used measure of spatial autocorrelation, namely the Moran Index. At the national level, in 2016, the value of the Moran index was 0.127, which shows a positive spatial autocorrelation between municipalities with a high or low number of Indian immigrants. Therefore, we can expect groups of municipalities with high (hot spots) and low (cold spots) numbers of Indian immigrants in different parts of Italy. In the LISA maps, I have identified two large clusters of municipalities (hot spots) with a high concentration of Indian immigrants in Italy (Figure 4). The first cluster, which was in northern Italy, I called 'Po Valley cluster' (because it was located on the plains of the Po River). It was composed of the provinces of Brescia, Bergamo, Cremona and Mantua in the Lombardy region, and the provinces of Reggio nell'Emilia and Parma in the Emilia Romagna region. The second cluster, 'Rome-Latina cluster', includes the province of Rome and Latinain the Lazio region of central Italy.
In 2016, the Po valley cluster was the largest and home to 35% of the total Indian immigrants in Italy. The province of Brescia, which was made up of 206 municipalities, had the largest number of Indian residents. The value of the Moran index for the province of Brescia was 0.23, indicating a positive spatial autocorrelation among municipalities with a high number of Indian immigrants. Most of the Indians settled in the southern part of Brescia, where the municipalities of Vobarno, Leno, Ghedi, Montichiari and Palazzolo sull'Oglio form a mini cluster. The number of Indian immigrants settled in the north of Brescia was very small. After Brescia, Bergamo was the second province with a high concentration of Indians. It had the Moran index of 0.496, which shows a highly positive spatial autocorrelation. In Bergamo, almost all the Indians were concentrated in the south-eastern municipalities of Castelli Calepio, Telgate, Romano Di Lombardia, Trescore Balneario and Villongo. In the Po valley cluster, the province of Mantua and Cremona had the largest proportion of Indians in its total immigrant population, but they are dispersed in many municipalities. Therefore, the values of the Moran Index for Mantua and Cremona were very low, i.e. 0.22 and 0.01, respectively. In Mantua, most of the Indians settled in the south of the Municipality of Suzzara, Viadana and Luzzara. In Cremona, the Indian population was equally distributed in municipalities not very close to each other, so there was no clustering ( Figure 5).
The Roma-Latina cluster had 18.8% of total Indian immigrant population in Italy. In 2016, the province of Rome had 121 municipalities, but most of the Indian immigrants were concentrated in the capital region. The municipality of Rome had the highest number of Indian immigrants (70%) in the province of Rome. The Moran's scatter plot shows a very weak negative spatial autocorrelation of −0.03. The negative spatial autocorrelation for the geographic distribution of any variable across a map shows that high values of that variable tend to be geographic neighbours of low values. In the province of Rome, the LISA cluster map shows that the municipality of Rome was surrounded by many municipalities with low values, which makes a low-high outlier around the municipal limits of Rome. The province of Latina had the highest degree of clustering with a  Moran Index value of 0.48. Here, the municipalities of Sabaudia, Terracina, Fondi and Aprila form a big hot spot in the southwest of the province, which can be seen in the LISA cluster map (Figure 6).

Factors shaping the spatial distribution of Indians in Italy
The settlement pattern of the Indian immigrants in Italy was guided by the concentration of employment opportunities and their social networks. At the regional level, the concentration of industries in northern Italy was a major attraction for Indian immigrants. The Parmesan cheese industry in the Po Valley region now depends heavily on Sikh workers involved in all stages of cheese production (Sahai and Lum 2013). Similarly, the leather industry in Vicenza has also attracted a large number of Indian workers. While in the Centre (the capital region of Rome) and in the Islands regions, where the tourism industry is the main source of employment, half of the Indian immigrants were employed in the hospitality and service sector. Especially, in the Island region, three quarters of the total population were employed in different services. In addition to tourism, in southern Italy agriculture has also attracted many low-skilled and irregular Indian immigrants.
Here the provinces of Caserta, Salerno and Reggio de Calabria had the highest concentration of Indians (Figure 7).
The second most important factor that affected the settlement process of Indian immigrants was social or kinship networks. This is one of the main reasons behind the rapid growth of the Po Valley and Roma-Latina clusters, where the total number of Indian residents increased from 18,462 and 4371 in 2004 to 52,156 and 23,705 in 2016, respectively. If we compare the growth of both clusters and the level of concentration of the Indian community, we can conclude that the Po valley cluster was the largest, but the Roma-Latina cluster was growing faster. Second, the decline in the Moran Index value from 0.14 in 2004 to 0.12 in 2016 shows a slight decrease in the concentration of Indian immigrants at the national level.

Internal mobility of Indian immigrants in Italy
The spatial distribution of any foreign population depends on its direct flow and internal mobility in a host country. Recent studies have shown that foreigners have higher internal mobility than the native population (Casacchia et al. 2010a;Bonifazi 2013). It is partly because immigrants have little or no affiliation to a particular place or city in the host country. In addition, international immigrants have already experienced a migration and, therefore, it is easier for them to migrate again (  with the increase in the number of families, the total population increased, but internal mobility decreased. In 2012 alone, due to the adverse effects of the economic crisis, the rate of internal migration increased, as many people lost their jobs and began to move in search of job opportunities in other areas (Figure 8). At the regional level, throughout the study period, the north-western region of Italy had the highest number of registered movements (44%) of Indian immigrants, followed by the north-eastern region (30%). It is directly related to the high concentration of the Indian population and better employment opportunities in these regions compared to the rest of Italy. Most immigrants from the northwest region migrate to the northeast region, it is understandable since people do not like to migrate to faraway places. A significant number of immigrants also move to the northwest region from central and southern Italy. If we compare with the total immigrant population of Italy, Indian immigrants were relatively more mobile in the northern regions and less in the Island region (Figure 9).
At the provincial level, Indian immigrants have also observed the traditional pattern of native internal migration from southern to northern Italy. During the last decade, all provinces in the south and centre showed negative net internal migration, while in the north, except for Brescia and Verona, all other provinces showed a positive net internal migration. The availability of unskilled jobs in industries was the main factor of attraction for immigrants, on the contrary, high unemployment in the south was the main reason behind negative net migration. The calculation of net internal migration at the provincial level during 2005-2015 shows that the north-western province of Mantua had the highest net positive migration and the southern province of Reggio de Calabria had the highest net negative migration ( Figure 10). The province of Mantua has received most of the immigrants from its nearest neighbour i.e. Brescia, which had functioned as a gateway and a place of dispersion for new Indian immigrants. On the contrary, the province of Reggio de Calabria had sent the largest number of Indians to Mantua, Brescia and Reggio nell'Emilia.

Determinants of inter-provincial migration of Indian immigrants
I have applied the 'gravity model' to analyse the internal migratory flows of Indian immigrants and other immigrants, using three explanatory variables: the size of the total resident population in the province of origin, the size of the total resident population in the province of destination and the distance between the two provinces. The model hypothesises that the masses, in this case two populations, affect the size of the migratory flows. It is assumed that the distance between provinces has a negative effect on the size of migratory flows. An intuitive conclusion is that the effect of distance should be less for foreigners, since they are less linked to the territory.
The gravity model applied to Indian and all immigrants' internal migratory flows shows a determination index (R-squared)of 0.496 and 0.441, respectively. The results of the models in both cases confirm the hypotheses on the role of the explanatory variables: the total population at origin and destination have a positive coefficient and therefore a direct positive effect on the number of migrants, while the negative coefficient of distance reveals the inverse relation between the number of migrants and the distance between provinces. It is interesting to note that for Indian immigrants the distance had less effect than the total immigrants, and the effect of the total origin population was also lesser than the total destination population. While comparing with 'all immigrants' the effect of all explanatory variables was weaker on the total internal migration during the study period ( Table 2).
The gravity model does not explain a significant part (50%) of the variability in the matrix of the migratory flows of both Indian immigrants and all other immigrants. This means that other variables play an important role. Then I decided to add two other dummy variables i.e. sex (male/female) and period of stay (before crisis, 2005-2009/after crisis, 2010-2015), to control the effect of 'gender' and 'time' on the internal migration. With the addition of these two variables, the determination index (Rsquared) improved to 0.502 and 0.504, respectively. By taking the reference category females, it shows that males were more mobile than females throughout the period of study. It was expected because the number of males is much higher than females, and they migrate more often for work reasons. The economic crisis had restricted the internal migration of the Indian immigrants; the internal migration was higher during the 'before crisis period' as compared to 'after crisis period' (Table 3).

Indian immigration during economic crisis
The recent economic crisis (2008) has negatively affected the entire population of Europe, but immigrants were the most affected by unemployment and cuts in social benefits in all countries (Castles and Miller 2010). In case of Indian immigrants, on the one hand, the total influx of immigrants declined, and on the other hand, immigrants who settled in Italy began to emigrate to other countries in search of work and a better life (Figure 11).

Effects on flow and stock
Indian immigrants in Italy who were employed in unskilled jobs suffered a major blow from the recent economic crisis. In the ensuing years, a significant number of Indians migrated to other countries and some of them also returned to India. As a result, for the first time in the history of Indian immigration to Italy, the total number of resident Indians decreased by 1% in 2012. The huge difference between the number of residence permit holders (169 thousand) and the actual residents (150 thousand) of Indian origin in 2016, also shows that a significant number of Indian immigrants (19 thousand) with Italian residence permits were not actually living in Italy. Not only the direction and size, but also the causes of immigration have also changed. According to the Residence Permit data, the main reason for issuing residence permits to Indian immigrants in 2007 was work (58%), followed by other reasons such as the family reunion (34%) and education (3%). In 2009, the proportion of workers in the total flow increased to 74%, but later, when Italy entered the period of economic crisis, the proportion of workers decreased to 23%, and immigration for family reasons increased to 61% of the total influx ( Figure 12). This dramatic change, on the one hand, showed the poor condition of the Italian economy to absorb more immigrant labour, and on the other hand, showed the success of family reunification among Indians in Italy.

Effects on internal migration
The pattern of net internal migration at the provincial level before the crisis period was very similar to the internal migration pattern of other immigrants and the native Italian population, that is, from the economically poor southern provinces to the industrialised northern provinces. All the central and southern provinces with a significant number of Indian immigrants, that is, Rome, Latina, Caserta, Salerno, Bari and Reggio de Calabria had negative net migration before the crisis period. While, in the north, the province of Bergamo, Cremona, Brescia, Mantua, Parma, Reggio Emilia and Pordenone had positive net migration during the same period. In the Northeast region, only the province of Vicenza had a negative net migration before the crisis. But after the crisis period (2010)(2011)(2012)(2013)(2014)(2015), the trends were reversed and the province of Rome, which was dispersing immigrants to other provinces, emerged as the great recipient of immigrants and registered positive net migration. By contrast, the province of Brescia took the place of Rome and began to disperse people to other provinces, therefore, the net internal migration becomes negative. During this period, Indian immigrants entered many other northern provinces such as Florence, Torino, Cuneo, Milan, Genoa, Pavia, Lodi, Novara, Bolzano, Rovigo and Treviso ( Figure 13). Contrary to the Indians, the pattern of internal migration for total immigrants remained the same from south to north, with the exception of the province of Modena, which had positive net migration after the crisis period.

Conclusion
During the last two decades, with a steady influx of low-educated and unskilled immigrants from the north Indian states of Punjab and Haryana, the size of Indian immigrant population increased five-fold, but the socioeconomic profile was drastically lowered. Even the well-educated Indians were engaged in the manual jobs in agriculture or industry. Moreover, the employment rate for women was much lower than men and most of them were engaged in domestic services. The high concentration of Indian immigrants in the Lombardy and Lazio regions, which were the economic centres of Italy and first place of settlement for pioneer Indian immigrants, proves that the spatial distribution was mainly guided by the availability of manual jobs in small scale industry and agriculture, and the spread of their social networks. The province of Brescia in the Lombardy region has the largest number of Indian immigrants, but the province of Latina in the Lazio region had the highest proportion of Indians (21%) in its total immigrant population. The Moran global index shows a positive autocorrelation among the municipalities with a high concentration of Indians in 2016, but the value of the index has been decreasing, which shows the dispersion of the population at the municipal level. With the LISA maps, we have identified two clusters of municipalities with a high number of Indian immigrants, i.e. the Po Valley cluster and the Roma-Latina cluster. The Po valley cluster was the largest, but the Rome-Latina cluster was growing faster with the immigration of unskilled Indian workers to the agricultural regions of Latina. The gravity model applied to internal migration shows that the internal migration of the Indian immigrants in Italy has been declined during the last decade. The recent economic crisis had reduced the influx of Indians to Italy, and changed the size and direction of their internal migration. The main causes of immigration also changed from 'work' to 'family reunion'.
As the level of concentration of Indian immigrants is still high around the Po Valley and Latina clusters, and most of them are occupationally segregated in manual jobs, it can negatively affect their socioeconomic mobility (Musterd and Deurloo 2002) and integration into the host society (Schönwälder 2007) in the near future. As the LISA index has begun to show the dispersion of Indian immigrants to other municipalities, we can expect that it will increase their contact with the host society and reduce their occupational and residential segregation. Due to the lack of data sources, the study of spatial distribution is limited to the municipal level, which is not sufficient to measure the degree of residential segregation. Provided by the infra-municipal population data, we can go deeper into explaining the situation of residential segregation or the dispersion of Indian immigrants in different Italian cities. Similarly, the internal mobility data provided by ISTAT was also limited to the provincial level, which hides most of the movements among the municipalities of the same province, again provided details at the municipal level we can identify the municipalities that were working as a gateway to attract immigrants from India and disperse to other municipalities in Italy.

Disclosure statement
No potential conflict of interest was reported by the author.

Notes on contributor
Nachatter Singh Garha is a Post Doctorate researcher (POP) in the Centre for Demographic Studies (CED). He holds PhD in Demography from the CED and the Autonomous University of Barcelona (UAB), Master in Demography from the European Doctoral School of Demography (EDSD) and Master in Territorial and Population Studies from the CED and the UAB.