Affective neighbourly relations between migrant and local residents in Shanghai

ABSTRACT In urban China there is growing scholarly interest in neighbourhood social interaction, but most studies focus on overt neighbouring activities whilst less is known about the affective dimension of neighbourhood relations, such as mutual trust and care. By surveying 1,420 residents from Shanghai, this study examines the affective relationship between rural migrants and local urban neighbours and explores whether the frequency of neighbouring and contextual characteristics may affect this outcome. Our results show that residents who interact more with out-group neighbours also tend to describe their relationship with them as more caring and amicable. Furthermore, residents in working class neighbourhoods tend to rely on intergroup neighbouring as means of facilitating mutual trust. In contrast, residents of neighbourhoods with commodified housing stock already possess a strong affective relationship with out-group neighbours because of a shared identity as middle-class homeowners and, therefore, do not rely on neighbourly interactions as a facilitator of neighbourly trust.


Introduction
Millions of rural migrants are now earning their livelihood in Chinese cities but still struggle to integrate into urban society, and are often referred to as "economic sojourners" or the "floating" population (Fan, 2002;Solinger, 1999;Wu, 2012;Yue, Li, Feldman, & Du, 2010). As part of the effort to socially integrate rural migrants, the relationship between locals and migrants has moved more into scholarly focus, but most findings so far suggest that these relationships are truncated and transient (Chen et al., 2011;Liu, Li, & Breitung, 2012;Wang, Zhang, & Wu, 2015, 2016Zhang, Li, Fang, & Xiong, 2009). The Chinese government has also started to pay more attention to the importance of neighbourhood social relations in assisting migrants to become socialized to their new urban contexts in the host society (Wang, Shen, & Liu, 2008;Wu, 2012). The neighbourhood plays a crucial role in this matter as the government attempts to engineer social life at the local level through the policy of "community construction" (Friedmann, 2007;Shieh & Friedmann, 2008). Scholars have, therefore, called for a better understanding of social relations at the neighbourhood level and gradually more studies on neighbouring and place attachment in urban China are emerging (Forrest & Yip, 2007;Hazelzet & Wissink, 2012;Wissink, Hazelet, & Breitung, 2013;Wu, 2012;Wu & He, 2005;Zhu, Breitung, & Li, 2012). Existing findings indicate that the importance and frequency of neighbouring activities are generally declining whilst the neighbourhood attachment of residents also varies significantly across different localities (Breitung, 2012;Forrest & Yip, 2007;Wu, 2012;Yip, 2012;Zhu et al., 2012). There are also a growing number of studies examining the relationship between migrants and locals, indicating that migrants with more social connections outside of their migrant peer-group, or more "out-group" social ties, have better chances of successfully integrating into the host society (Yue, Li, Jin, & Feldman, 2013), and have better housing opportunities (Liu, Wang, & Tao, 2013).
The contribution of this paper is twofold. Firstly, this study sheds light on the importance of neighbourly relations in assisting rural migrants to socially integrate into China's urban society. At present, little is known as to whether and how the neighbourhood may be related to the success or failure of social integration of migrants in urban China. Existing neighbourhood studies have only explored neighbourly relations in general Yip, 2012), failing to take the opportunity to directly address the affective relationship between migrant and local urban residents. On the other hand, social integration studies on China so far have rarely considered the importance of neighbourly relations as a means of socially integrating rural migrants (Liu et al., 2013;Yue et al., 2013). However, experience from multi-ethnic contexts shows that neighbourly interactions and feelings of mutual trust can help to remove social barriers between majority and minority groups and foster social cohesion (Alesina & La Ferrara, 2002;Henning & Lieberg, 1996;Putnam, 2001), thus making this understudied intersection of neighbourhood affect, immigration experience and social integration ripe for investigation in China.
Our second contribution is the exploration of the relationship between neighbourly interaction (i.e. activities such helping and greeting) and the affective dimension of neighbourly relations more generally. In contrast to neighbourly interactions, characterized by overt forms of relations including mutual support or greeting, affective neighbourly relations refer to the level of trust or feelings of mutual regard between residents (Mann, 1954). Whilst scholars have examined the neighbouring activities of residents in urban China (Forrest & Yip, 2007;Hazelzet & Wissink, 2012;Wang et al., 2016;Wu, 2012) and general neighbourly trust under the large umbrella term of "sense of community" (Breitung, 2012;Yip, 2012), little is known about whether neighbourly interactions and affective neighbourly relations are associated, or the theory underpinning this relationship. The position explored in this paper is the contact hypothesis, which asserts that social interaction in an equal and cooperative environment facilitates tolerance and positive intergroup relations (Allport, 1954;Hewstone & Brown, 1986;Pettigrew, 1998).
Consequently, this paper sets out to investigate the current affective relationship between rural migrant and native-born urban neighbours and analyses whether more neighbourly interactions, such as mutual support can improve its outcome. Furthermore, we aim to explore how contextual factors may be related to intergroup neighbourly trust and care. Our study attempts to answer questions including: How many migrant and native-born residents would consider their affective relationship with "out-group" neighbours as positive? How much of this affective relationship is related to how often neighbours interact with each other? What neighbourhood factors are associated with the affective relationship between migrant and local neighbours? To answer these questions, we analyse a 1,420 questionnaire dataset collected from both migrant and local residents across a variety of Shanghai's neighbourhoods.
The paper is structured as follows: The next section reviews existing theories on the affective dimension of neighbourhood relations and its underlying dynamics, as well as gives an overview of neighbourhood studies in urban China. The following section discusses our methods of data collection and analysis, followed by the presentation of our analysis. Finally, the last section offers a discussion on the implications of our study for the integration of rural migrants in China.

Affective dimensions of neighbourhood social relations
Many studies contend that neighbourhood social relations play an important role in fostering the cohesion between different social groups (Forrest & Kearns, 2001;Kearns & Parkinson, 2001;Putnam, 2001Putnam, , 2007. According to Mann (1954) neighbourly relations can be categorized as either manifest or latent forms of "neighbouring." Manifest neighbouring refers to visible forms of social interactions, such as visiting each other, whereas latent neighbouring is characterized by "favourable attitudes to neighbours which result in positive action when a need arises, especially in times of crisis or emergency" (Mann, 1954, p. 164). Some studies also consider neighbourly relations as an important part of the wider concept of sense of community (Nasar & Julian, 1995;Talen, 1999;Unger & Wandersman, 1985) and neighbourhood cohesion (Buckner, 1988). A positive affective relationship between neighbours can break down prejudices and assists in forming harmonious communities and is thus of great importance to migrants and marginalized social groups (Henning & Lieberg, 1996;Nannestad, Lind Haase Svendsen, & Tinggaard Svendsen, 2008;Putnam, 2007). In line with these studies, our paper focuses on two dimensions of neighbourhood social relations, namely neighbourly interactions (or neighbouring), which includes activities, such as mutual support or visiting each other. In addition, we examine indicators, such as mutual trust or reciprocal care, which reflect the intangible and affective relationship between neighbours. Henceforth, we will use the terms "affective relationship" and "affective neighbourly relations" interchangeably to refer to the levels of trust and care between migrant and local residents.
With respect to the determinants of the affective relationship between out-group neighbours, most studies state that ethnic diversity and the level of poverty within the area negatively affect intergroup neighbourly relations (Laurence, 2011;Li, Pickles, & Savage, 2005;Stolle, Soroka, & Johnston, 2008). Whilst there is no unanimous support, many studies operate in accordance with the conflict theory that living in ethnically more diverse areas increases social distrust amongst individuals (Gundelach & Freitag, 2014;Putnam, 2007;Stolle et al., 2008). Similarly, neighbourhood deprivation leads to social isolation and poorer relationships between residents (Laurence, 2011;Letki, 2008). For both strands of research, contention over limited resources, which is more likely to occur in poor neighbourhoods, is the largest cause of conflict and alienation between the majority and minority groups (Laurence, 2011;Letki, 2008;Putnam, 2007).
Conversely, a growing body of literature contends that more frequent neighbourly interactions between recent immigrants and local residents can help foster inter-ethnic social trust (Gundelach & Freitag, 2014;Laurence, 2011;Stolle et al., 2008). Stolle et al. (2008, p. 61) speculate that it is diversity without contact that is most detrimental to intergroup relations whilst more frequent interaction helps mediate this effect. Much of the explanations for this outcome are based on the fundamentals of contact theory, which asserts that pleasant and cooperative interactions between different social groups can assist in reducing social tensions and create a stronger sense of shared social identity (Allport, 1954;Hewstone & Brown, 1986;Pettigrew, 1998).

Migrant-local social relations and neighbourhood interaction in urban China
Scholarly interest in migrant-local relations has grown considerably in urban China (Roberts, 2002;Solinger, 1999Solinger, , 2006Wang et al., 2015Wang et al., , 2016Zhang et al., 2009). The general consensus is that rural migrants in particular are facing discrimination and hostility from local residents due to stigmas of crime and unemployment (Solinger, 1999). More recent studies suggest that those who successfully acquire social ties with native urban citizens are often rewarded with better chances of integration (Yue et al., 2013) and also have higher likelihood of living in a more affluent neighbourhood and be homeowners (Liu et al., 2013). According to research, usually those who have lived longer in the city and have received better education also tend to possess more local social contacts and better integration chances (Li & Wu, 2013;Liu et al., 2013;Yue et al., 2013).
With regard to neighbouring in Chinese cities, studies suggest that neighbourhoodlevel social interaction has declined compared to the pre-reform era (Hazelzet & Wissink, 2012;Whyte & Parish, 1984). The transition from a planned to market-led economy has fundamentally changed how Chinese urban citizens interact with each other (Hazelzet & Wissink, 2012;Wu & Logan, 2015). The burgeoning middle class, who are the primary residents of commodified housing stock, rely instead on social ties outside of the neighbourhood and are less involved with neighbours (Li et al., 2012, p. 249). In comparison, older neighbourhood types, such as courtyard houses, work-units and relocation settlements, which are mostly occupied by lower income and working class residents, still retain a fairly high share of neighbourhood-based social activities (Forrest & Yip, 2007). In relation to neighbourly interactions between migrants and locals, migrants are more eager to interact with their local neighbours largely due to greater needs for local support (Wu, 2012). In comparison, local residents in lowincome areas have difficulties adapting to the different life-styles of their migrant neighbours (Wu, 2012).
Research so far has paid considerable attention to the frequency of local interactions ranging from visiting each other, mutual support and neighbourly communications (Forrest & Yip, 2007;Wang et al., 2016;Wu & Logan, 2015). However, less focus has been placed on the affective side of neighbourhood relations. Existing evidence indicates that the relationship between migrant and local neighbours suffers from discrimination, prejudice and hostility, and both groups of neighbourhood residents are more reliant on social networks with fellow in-group members (Chen et al., 2011;Roberts, 2002;Solinger, 1999;Whyte, 2010). Consequently, our study first hypothesizes that the affective relationship between migrant and local neighbours is truncated and more strained as compared to the relationship between fellow in-group members (H1). However, the literature further indicates that frequent neighbourly interactions may contribute to a better relationship (Gundelach & Freitag, 2014;Stolle et al., 2008). Thus, our second hypothesis assumes that neighbourly interaction has a positive effect on the affective relationship between migrant and local neighbours (H2). Moreover, we hypothesize that in comparison to native Shanghai residents, migrant residents are more likely to have a positive relationship with their local Shanghai neighbours (H3). This hypothesis is based on the existing literature that migrants in urban China are more willing to establish social ties with local residents since migrants have a stronger need to bridge social support networks in order to overcome their limitations in the city (Liu et al., 2013;Yue et al., 2013). Finally, we hypothesize that the underlying dynamics of affective neighbourly relations is significantly different between commodified-housing neighbourhoods and older and less affluent neighbourhoods such as traditional courtyards, workunit housing and relocation housing settlements (H4). This hypothesis is based on the existing knowledge that in comparison to older neighbourhoods (such as courtyards and work-units), residents in privately developed commodified neighbourhoods are less likely to interact with their neighbourhoods but are still very emotionally attached to the neighbourhood and consider their neighbours as equals (Forrest & Yip, 2007;Pow, 2007).

Methods and data
The data used for the analysis of this study comes from a survey conducted in August 2013, in Shanghai. There are several reasons for selecting Shanghai as a case study. Firstly, Shanghai is amongst the cities most challenged by the large influx of migrants, as 40% of the total population of 23 million residents are non-locals (National Bureau of Statistics [NBS], 2010). Shanghai's migrant population is also very diverse, possessing varying degrees of education, income and skills level and coming from rural and increasingly also urban areas (National Bureau of Statistics [NBS], 2010). Furthermore, in contrast to Guangzhou and Shenzhen, where migrants primarily congregate in urban villages, migrants in Shanghai live in a variety of neighbourhoods, ranging from older and dilapidated traditional courtyards and work-units to gated communities developed through the private market (Migrant Population Commission, 2012). This diversity of residential neighbourhoods can be beneficial to the exploration of different contextual effects and varying practices of neighbouring activities.
The household survey was conducted at a citywide scale by a team of professionally trained surveyors, and the team leader was a former survey officer of the Shanghai Statistical Bureau (SSB)'s urban livelihood survey team. Similarly, all surveyors were former staff members of the SSB and have experience in conducting surveys in the neighbourhood they were assigned to. The survey was carried out following the principle of random sampling, and we specifically used a two-stage sampling strategy in order to enhance the representativeness of the data and also reflect the spatial characteristics of different localities, such as an area's poverty rate and the presence of migrant residents. The data was sampled at residential committee (juweihui) level, which is the next level down from the sub-district level (jiedao). Building blocks and streets naturally delineate the boundaries of residential committees and can be useful in identifying different types of localities. For each chosen area, 40 copies of questionnaires were distributed. Although the population of each juweihui is fairly similar in general, there are still outlier cases; thus, in order to avoid any biases in our analysis that may occur due to varying population sizes of residential committees, our final data sample includes weighting variables that account for the total population in each respective neighbourhood. With regard to the sampling strategy, at the first stage we randomly selected 35 neighbourhoods based on a set of criteria including the location of the neighbourhood (inner city, middle ring and outer ring areas of Shanghai), the GDP per resident, population density and the share of native hukou population at the subdistrict jiedao level. At the second stage, households in each selected juweihui were randomly picked for the survey questionnaire starting from a random street number and chosen at a fixed interval. The purpose of this strategy was to approximate the distribution of the sample within the neighbourhood as closely as possible to the actual characteristics of the locality's population. There are two reasons for choosing an address-based approach rather than using a selection based on the official registration list. Firstly, any temporary and non-local residents are not available in the official registration list and secondly an address-based approach in practice performs better and enables more randomness. Regarding the interview procedure, the survey required the head of household to be interviewed. The survey was aided by members of the residential committee who introduced surveyors to interview households and thus ensured a very high success rate (95%). In total, 1,420 valid samples were produced. From the survey samples, 1,046 are native residents holding an urban hukou, 128 local rural hukou holders 2 and finally 244 migrant residents (17.18%) amongst which 86 were urban migrants and 158 were rural migrants. The reason for the low migrant ratio is due to their irregular and long working hours, which resulted in many migrant households being unavailable for interview during the day or even weekends. To rectify this shortcoming we interviewed another 100 migrant respondents after the initial survey in order to ensure that there is no systemic lack of any migrant group. Consequently, we are confident that the lower number of migrant residents will not affect our analysis. The comparison of our survey data and official statistics also reveals a considerably good degree of similarity (see Table 1). In order to control for the unaccounted clustering at a higher level, this paper employs a mixed effects linear regression, also known as multilevel modelling, utilizing the Stata 13 program. A multilevel model significantly reduces correlation errors and biased estimates of parameters caused by the grouping of variables at higher levels, which a conventional linear regression model is unable to take into account (Raudenbush & Bryk, 2002). Furthermore, by allowing random effects to vary it is possible to explore the contribution of both individual and neighbourhood indicators separately (Raudenbush & Bryk, 2002). Given the substantial advantages of mixed effect models in estimating neighbourhood variations, previous studies researching contextual determinants have also adopted the multilevel approach (Gundelach & Freitag, 2014;Laurence, 2011). Using a mixed effects model is, therefore, reasonable for the purpose of this study, which is to control for neighbourhood variations including housing type and migrant concentration, and to assess the effect of neighbouring activities.

Measuring the affective relationship between neighbours and intergroup neighbouring activities
The dependent variable of this study is the affective relationship between residents. We follow the approach by Buckner (1988) and Mann (1954) but make some amendments to the index of affective neighbourly relations in order to render it more comprehensible and relevant for the Chinese context. The index consists of four subcategories measuring the levels of mutual amity, care, trust and familiarity. Each sub-question is measured on a scale of 0-5 whereby 1 is highly disagree, 5 is highly agree; and 0 means not applicable. We asked migrant residents to describe their relationship with nativeborn neighbours whilst local residents were asked about their affective relationship with migrant neighbours (see Appendix for the specific questions asked). Similarly, the independent variable "neighbouring activities between migrant and local residents" was also created using an index of three sub-questions which included the frequency of visiting each other's home, helping and receiving support from neighbours (we did not specify what kind of help although helping neighbours take care of children or pick up children from school were mentioned as examples) and finally exchanging greetings. The index for neighbouring activities can be seen in Table 2. Again, migrant residents were asked about their neighbouring activities with locals whilst native residents had to answer about their frequency of interacting with their migrant neighbours. With respect as to how we conceptualized the category of "out-group" and formulated the question, we used the expression "locals" (bendiren) for native-born Shanghai residents and "non-local" (waidiren) for migrants since these terms are widely used for describing natives of and migrants in Shanghai by the general population.

Neighbourhood control level variables
This study controls for three contextual variables at the juweihui level (obtained from the respective juweihui) which is the de facto local government and is also responsible for collecting accurate statistics at the neighbourhood level. We firstly control for the percentage of migrant residents in the neighbourhood and secondly in order to measure neighbourhood poverty, we adopt Wu, He, and Webster (2010) method and utilize the number of recipients of the Minimum Living Standard Support (MLSS) within the neighbourhood. Finally, we also included the prevalent housing type of the neighbourhoods in order to control for varying levels of neighbouring practices that are due to elements of the built environment. In general, it is possible to categorize neighbourhood-housing types into traditional courtyard housing, relocation settlements, workunit housing, urban villages and neighbourhoods with privately developed housing meant for sale on the market. Considering that migrants do not receive MLSS, urban villages 1 (chengzhongcun) were also included to account for migrant poverty to a certain extent (Wu et al., 2010, p. 140).

Individual-level control variables
Individual-level variables include education level, income, age, tenure, length of residency and hukou status. There are four categories of hukou status: native-born urban (local non-agricultural), native-born rural (local agricultural), rural migrant (non-local agricultural) and urban migrant (non-local non-agricultural). The reason to include four hukou categories is to account for the heterogeneity of the migrant population, as urban migrants from other cities may be very different from rural migrants in terms of education level and employment opportunities. For the same reason, we also added two interaction terms between hukou status, education and income of survey respondents, since the underlying dynamics for having local social interaction may be different for native and migrant residents with differing levels of income and education.

Initial data findings
Comparing in-group and out-group affective relations amongst residents Tables 3 and 4 show the current level of in-group and out-group affective relationship of migrant and local residents. We included results regarding the feelings migrant and local-resident neighbours hold towards their own social group in order to put our findings into perspective.  Table 3 reveals that both native and migrant residents have a mixed to positive affective relationship with their fellow in-group neighbours, as less than 10% of residents report that they distrust or do not care for their in-group neighbours. Instead, more than half of the respondents describe their relationship with their ingroup neighbours as amicable, and almost half of residents find that they are familiar with other fellow in-group neighbours. In contrast to in-group affective relations however, residents in Shanghai feel more distanced towards their out-group neighbours (see Table 4). Over 20% of residents state that there is no mutual care and trust between themselves and out-group neighbours and that they are unfamiliar with their out-group neighbours. Another finding is that a large share of residents have chosen a neutral value to describe their current relationship with both in-group and out-group neighbours. Over half of respondents reported mixed feelings towards out-group members, whilst around 30-45% expressed similar sentiments towards their in-group neighbours. This outcome may reflect the growing degree of indifference and apathetic feelings amongst neighbours and the decline of neighbourhood level social relations as noted by earlier studies (Forrest & Yip, 2007).

Affective relationship between neighbours by hukou status and across different neighbourhood types
Figures 1 and 2 present the current affective relationships of residents with in-group and out-group neighbours by hukou status, respectively. Figure 1 shows that more than half of native urban Shanghai residents reported a positive to fairly positive relationship with fellow local Shanghai neighbours and are thus the highest scoring hukou group. In comparison, Figure 2 reveals that less than a third of native Shanghai residents, both urban and rural, feel trustful and familiar towards their migrant neighbours. Moreover, the share of local residents who describe their relationship towards migrant neighbours as alienated and distrustful is two to three times higher than the share who report negative relations with fellow Shanghai neighbours. In contrast, urban and rural migrant residents feel relatively similar towards both out-group and in-group neighbours in terms mutual trust and care. These outcomes support our first hypothesis and also indicate that the differentiation between the "us" and "them" is especially accentuated amongst Shanghai residents. This is no surprise as negative stereotypes and the institutional divide of the hukou system are reinforcing the sense of superiority of local residents over migrants. Figure 3 shows how the affective relationship between migrant and local neighbours varies across different types of neighbourhoods. The affective relationship between migrant and local residents is by far the most positive in privately developed, commodified housing neighbourhoods, as more than a third of respondents describe their intergroup neighbourly relationship as positive to fairly positive. Residents living in work-unit neighbourhoods and urban villages also report that their relationship with out-group neighbours is fairly trustful and amicable, closely followed by relocation settlements. In contrast, traditional courtyard homes have the lowest share of residents who would describe their affective relationship towards out-group neighbours as familiar or trusting.  This initial result indicates firstly that the affective relationship between migrant and local neighbours varies considerably across different types of neighbourhoods whereby residents living in affluent neighbourhoods tend to feel the most trustful and caring towards their out-group neighbours. Secondly, in the case of Shanghai stronger relations between existing local residents may be a reason why traditional courtyard neighbourhoods are not responsive to intergroup neighbourly relations. Since local residents in traditional courtyards neighbourhoods tend to have lived in the area for a long time, they may already be part of an existing community with fellow Shanghai residents. Consequently, local residents may be less inclined to establish any deeper relations with migrant neighbours.

Results of the mixed effects linear regression
In order to assess the interwoven relationship between neighbourhood characteristics and the frequency of neighbouring activities, we entered the variables in a stepwise manner. Model 1 (see Table 5) includes all independent variables except for the intergroup neighbouring variable, which is added in model 2 (see Table 6) in order to see how it alters the effects of all other determinants.

Modelling the dynamics of affective neighbourly relations between migrant and local residents
We conducted a likelihood ratio test in order to verify whether using a mixed effects model is justified in our case. The multilevel model performs better than the ordinary least squares model as the result significantly rejects the null hypothesis (p < 0.001  further justification for the mixed effects model is that more than 30% of the dependent variable's variation can be explained through neighbourhood level determinants. Finally, we had to drop the interaction term between education and hukou status due to multicollinearity. Table 5 shows that out of the neighbourhood level determinants, neighbourhood type is the most significant factor. In comparison to urban villages, residents living in work-units, relocation housing and commodified housing are all more likely to trust and care for their out-group neighbours. The outcome may appear surprising at first given that urban village was ranked second place in terms of trust and care with outgroup residents (see Figure 3). However, considering that the mixed effect model has control for the likelihood of clustering and random variations, this result indicates that with the exception of a few cases, living in urban villages in fact results in a poorer affective relationship between migrant and local residents. This may be due to two reasons pertaining to the characteristics of urban villages. Firstly, the residential composition of urban villages, which mainly consists of rural migrants and local villagers, could be a reason. Previous research already showed that local villagers have a very paradoxical relationship with rural migrants, whom they consider as inferior but also as a necessary source of tenancy and rental income (see Chung, 2010 for a discussion on the Chinese context of urban villages as migrant enclaves). As a consequence, local villagers and migrants rarely interact and their relationship is superficial and distant. Secondly, compared to other neighbourhood types, urban villages usually have an extremely high share of rural migrants. In our sample the concentration varied between 60% and 76%. However, this is by far not the most extreme as the share of residents of some urban villages amount to 80% rural migrants (Chung, 2010). Although the variable migrant concentration suggests that areas with more migrant residents are 0.9 times more likely to have a positive affective relationship between out-group residents, such extreme concentrations may be an exception. The reason could simply be that the disproportionate ratio of local residents prevents any consistent intergroup relationship from emerging. For the remaining neighbourhood types, there are other reasons that lead to a stronger affective relationship between migrant and local neighbours. For example, stronger ties between fellow staff may be the reason in work-unit neighbourhoods. For a majority of residents living in work-unit neighbourhoods, their residences were allocated by their respective work-units. Consequently, the likelihood that work-unit residents are also fellow staff from the same department is also very high and may foster their neighbourly trust. On the other hand, residents in relocation settlements feel more trustful and familiar towards their out-group neighbours possibly due to their lower income level, which is often related to more frequent social interactions at the local level. With respect to individual-level factors, only hukou status is significantly related to the affective relationship between migrant and local neighbours. Compared to local urban residents, urban migrants and rural migrants are 2.7 times and 1.7 times, respectively, more likely to describe their relationship with native-born neighbours as caring and trustful. This signals that many native Shanghai residents still refuse to have any social interactions with rural migrant neighbours due stigmatisation. It is also important to mention that many factors that were significant predictors of general neighbouring activities, such as length of residency or age, do not affect the affective dimension of neighbourhood relations between migrant and local residents. The reason could be that in contrast to general neighbourly interactions, contextual factors play a more significant role for intergroup relations. For most residents, interacting or having a trustful relationship with out-group neighbours is not a necessity. In areas where the presence of out-group members is low for instance, native-born residents have the option to entirely rely on their in-group neighbourly ties. Thus it is possible for residents to remain involved in the neighbourhood whilst isolating themselves from their out-group neighbours. This is particularly true for local residents who already have an established social network, whereas migrant inhabitants still need to reach out to native-born neighbours in order to strengthen their local support ties. Consequently, it is understandable that the underlying dynamics of one's willingness to interact locally differ from one's inclination to feel familiar and trustful towards their out-group neighbours.

Modelling the mediating effects of intergroup neighbouring activities on the affective relationship between migrant and local neighbours
In model 2, we added the intergroup neighbouring variable in order to investigate how the frequency of neighbourly interactions can affect the affective relationship between migrant and local neighbours (see Table 6). We conducted a Wald test to confirm whether adding intergroup neighbouring improves the model fit. The result shows that the variable adds one more degree of freedom and based on the significant p-value (p < 0.001) we can therefore reject the null hypothesis. With regards to the effect of intergroup neighbouring, if a resident were to increase his or her intergroup neighbourly activities by one unit, his or her affective relationship with out-group neighbours would improve by 0.6 units. Furthermore, as expected the significance of other variables also changed considerably after including intergroup neighbouring. Compared to model 1 where the concentration level of migrants was significant, in model 2, there is no longer a statistically significant relationship between an area's migrant concentration and the affective relationship between migrant and local neighbours. This result indicates that residents living in more diverse areas have a stronger affective relationship with out-group neighbours because there is a higher chance of encountering and interacting with each other. Moreover, the significance of the hukou variable also dropped considerably for both rural migrants (p < 0.05) and urban migrants (p < 0.01). This outcome implies that one key reason why migrant residents tend to have a better affective relationship with native-born neighbours is because they interact more with them, which in turn generates feelings of mutual trust and care. In addition, both migrant groups may regard their local neighbours as an important source of social support and therefore have a better neighbourly relationship with them. These results confirm our second hypothesis that more frequent interaction between migrant and native residents can lead to a more positive affective relationship between migrant and local neighbours. After controlling for neighbouring activities, the significance of neighbourhood types has also changed as relocation settlements are no longer significantly associated with better neighbourly sentiments and the significance of work-unit neighbourhoods has decreased considerably (p < 0.05). This outcome may imply that in addition to frequent neighbourly interactions, residents in work-unit neighbourhoods have closer relationships with out-group neighbours partly due to sharing the same work place or same occupation. Residents in relocation neighbourhoods on the other hand feel trustful towards each other because they tend to interact more frequently with each other. Consequently, model 2 also provides statistical evidence to verify our third hypothesis that living in older neighbourhoods and being a migrant is significantly associated with frequent neighbouring activities and thus results in a more trustful and caring relationship between migrant and local residents. Only the category of commodified housing neighbourhoods has retained its high significance level (p < 0.001), which we believe is related to the fact that residents of privately developed housing share a strong common identity as fellow residents of a gated community. Other studies on sense of community in urban China also found that residents living gated neighbourhoods generally have a strong sense of community due to a shared social identity (Breitung, 2012;Yip, 2012;Zhu et al., 2012). Being homeowners and perceiving other neighbours as fellow members of the same class may have strengthened residents' shared identity.
Finally in model 2, income has become a very significant determinant. After controlling for intergroup neighbouring the results show that the higher the income of households the more likely residents are to feel distanced from their out-group neighbours. The finding that higher income itself leads to social alienation and distrust between migrant and native-born neighbours is not surprising, as previous studies have come to similar conclusions . The largest reason could be that those with a higher socioeconomic status are less involved locally and thus also feel more alienated and distrustful towards their neighbours. This would also explain why the variable has gained significance after controlling for the frequency of intergroup neighbouring, implying that affluent households, which are more locally involved, are also exempt from this negative effect and have a stronger affective relationship with their out-group neighbours.

Conclusion
Much research focus has been dedicated to social interactions at the neighbourhood level, and the government has placed high hopes on the role of neighbourhoods in alleviating existing problems of integrating rural migrants Shieh & Friedmann, 2008;Wang et al., 2008). Most studies so far have concentrated on the level of neighbourly interactions (Forrest & Yip, 2007;Hazelzet & Wissink, 2012;Liu et al., 2012;Whyte & Parish, 1984;Wu & He, 2005) and general sense of community (Breitung, 2012;Yip, 2012), but there is less empirical research specifically investigating the affective dimension of neighbourly relations between migrant and local residents. This paper has sought to fill this gap and explored the affective relationship between migrant and native-born residents in Shanghai and how intergroup neighbouring activities and neighbourhood factors may affect this outcome. In general, our findings suggest that the affective relationship between migrant and local residents is relatively weak since only a third of respondents describe their relationship to out-group neighbours as trustful and caring. In contrast, the relationship is considerably stronger between in-group neighbours, whereby especially native urban Shanghai residents feel familiar and trusting towards their fellow local Shanghai neighbours. Comparatively, migrant residents remain relatively similar with regards to their affective relationship towards both local and non-local neighbours.
With respect to the underlying dynamics of the affective relationship between migrant and local residents, three key factors can be identified. Firstly, native residents are significantly less likely to have a trustful and caring relationship with out-group neighbours as compared to migrant residents. This result implies that interpersonal relations in urban China are affected by the stigmatisation of rural residents. Due to the common perceptions of low education and income and working in poorly paid jobs, many locals tend to avoid rural migrants. The government's hukou system exacerbates this problem firstly by providing an "official" label that underlines the difference between rural and urban residents. Furthermore, the welfare entitlements of the hukou status prevent migrants from attaining a better social status and escaping their stigmatisation.
Secondly, in cases where residents engage in frequent intergroup neighbouring, the affective relationship between them also tends to be more trustful and amicable. This outcome signals that the social distance between migrants and locals in urban China can be overcome by frequent interactions in a more intimate and consistent context, such as the neighbourhood. Our research also finds that certain neighbourhood characteristics foster intergroup neighbouring and therefore contribute to strengthening the affective relationship between migrant and local neighbours. Residents living in areas with a higher migrant presence tend to engage more in intergroup neighbourly interaction, which in turn leads to stronger mutual trust and care between residents. The exception to this trend is urban villages, where residents are the least likely to feel trustful and amicable towards their out-group neighbours. This may be due to the fact that in some migrant enclaves the share of migrant residents can reach up to 80% (Liao & Wong, 2015) and results in a lack of native-born residents to create any meaningful social connections. This disadvantage of urban villages is by no means limited to the Chinese case. Residents in migrant enclaves in other societies also suffer from a shortage of out-group social ties and run risk of being isolated from members of the host society (Light, Sabagh, Bozorgmehr, & Der-Martirosian, 1994;Logan, Zhang, & Alba, 2002). Nevertheless, urban villages provide other more immediate kinds of relief, especially to newly arrived rural migrants such as affordable housing and employment opportunities.
Finally, our third major finding implies that the social class of neighbourhoods significantly affects the affective neighbourly relationship between migrants and locals. Middle class residents living in commodified housing neighbourhoods seldom interact with their out-group neighbours but perceptions that other neighbours also belong to the same social class ensure that residents have a strong affective relationship with each other. Moreover, the advertised image of privately developed neighbourhoods being exclusive "civilised communities" (Pow, 2007) reinforces this shared sense of social class. This signals that the social identity of China's middle class is no longer shaped by institutional classification (i.e. hukou), which mattered more during the socialist era, but instead depends on homeownership in a private estate. The drawback of marketized neighbourhoods, however, is that tolerant attitudes are mainly reserved to fellow residents whilst non-residents, often depicted as poor rural migrants, are regarded as a threat (Pow, 2007). On the other hand, working class neighbourhoods, such as work-units or relocation housing estates not only tend to rely more in intergroup neighbouring activities as an important source of social support but also as a facilitator of neighbouring trust and amity.
This study has shed some light on the social relationship between migrant and local residents in urban China. Despite the general trend of declining neighbourly relations in Chinese cities (Forrest & Yip, 2007;Hazelzet & Wissink, 2012), rural migrants continue to rely on neighbouring activities as a crucial means to establish social ties with native-born residents. Interactions between migrants and locals at the neighbourhood level help to create an affective relationship based on mutual trust and care, which according to Mann (1954) can be of particular use during times of emergency and need. More importantly, intergroup neighbourly relations also help to reduce distrust and stigmatisation that are exacerbated by institutional discrimination and negative media coverage. Our study, therefore, contributes to the emerging research on the social integration of rural migrants (Liu et al., 2013;Wang et al., 2015;Yue et al., 2013) and identifies the importance of the neighbourhood in socially integrating migrants. With regard to the wider debate surrounding what constitutes positive intergroup contact, our research shows that neighbourly relations can be considered as a "pleasant" and "cooperative" form of social interaction advocated by the contact hypothesis (Hewstone & Brown, 1986;Pettigrew, 1998). The finding that the contact hypothesis is more applicable to the Chinese context also indicates that the social distance created by an institutional divide can be more easily overcome than compared to the social distance invoked by ethnic differences. Notes 1. Urban villages are formerly rural villages that have been encroached by urban developments and now primarily function as accommodation for rural migrants. The housing quality tends to be sub-standard as the rural landlords often built them to a very high density in order to maximize housing profit (for more information, see Chung, 2010;Wu, Fangzhu, & Webster, 2013). 2. In our case, native rural hukou holders refer to local villagers who hold a local agricultural hukou from Shanghai and live in the rural parts of the city. Compared to migrants, local villagers still hold a "local" Shanghai hukou and have access to welfare entitlements provided by the rural collective. Moreover, local villagers have received much more attention from the government. For instance, being collective landowners native rural residents receive various forms of compensation, similar to an urban hukou (which entitles the holder to a range of welfare resources, monetary compensation, alternative housing, etc.) from the government in the event that their land is used for urban development . For more information regarding local villagers see Wu et al. (2013) and Chung (2010).