Research projects are the heart of the research incubator. During the summer schools, small teams work through the entire research process on a topic of their choice. These projects are usually leaded by two experienced researchers who supervise about three junior scholars on the topic.

The following projects will be offered at the Summer School on Migration 2019: 

  1. Social integration and the Urban Form: Measuring immigrants integration through urban mobility patterns (Urban social integration)
  2. Behavioural Consequences of Group Threat – Ethnic Derogation on YouTube in the Aftermath of Threatening Events (Ethnic derogation)
  3. Analyzing the Social Integration of refugees through Text Analysis of Social media Data (Integration and social media)
  4. A test of the Schelling model of spatial segregation (Testing the Schelling model)
  5. Digital and computational approaches to study migration (Digital migration reserach)
  6. Global Migration Modelling with MIDAS (Global MIDAS)

Social integration and the Urban Form: Measuring immigrants integration through urban mobility patterns

Project leader: Marta C. González, Luis E. Olmos

Short title: Urban social integration

Research question:

  • How do mobility patterns, based on activity-space, travel times, and number of trips vary for lower-wage workers compared to trips originating from higher-income areas?
  • How those segregation-related mobility patterns are affected by the different demographic distributions in different cities?
  • How social cohesion influence political outcomes and crime statistics?

Methods: Having deployed this data-driven modeling framework, we propose a hands-on workshop to explore the daily mixing effects on political support and crime statistics. The argument is as follows. Across countries, greater amounts of ethnic division strikingly reduce the levels of social spending (welfare state) as a share of GDP. At the individual level, people who live among poor people of the same race are more likely to support welfare, but those who live among poor people of a different race are less likely to support welfare. As the welfare implementation depends on the ideologies of the political parties (i.e. left-wing supports it and right-wing not), analyzing the relationship between racial exposure and the voting preferences at the tract level could give us some insights of the well-being of immigrant integration. Segregation-related mobility patterns will add a new dimension in the evaluation of immigrant integration.

A data-driven methodology to study these processes is crucial for designing policies that ease the social integration of immigrant communities, as well as prompting the need for improvement of social support systems. Human migration creates the political challenge of designing integration policies to allow newcomers to settle in unfamiliar environments, as well as to promote their socioeconomic development and well-being. As this process can be affected by social polarization and spatial segregation, we want to enrich the classical notions of racial/economic segregation through the lenses of time and human mobility.

Data: Mobility models in three cities based on mobile phone records, census data, crime records, political outcomes.

Description: Driven by globalization, ethnic diversity is increasing in major cities around the world. This process creates the political challenge of designing policies that ease the social integration of immigrant communities, as well as promoting the need for improvement of social support systems, and so reduce the chances of ethnic and political frictions. The natural question is how to evaluate the immigrant integration, we argue that integration can be reflected by segregation-related mobility patterns. In this summer school, we have two goals. First, we will present a data-driven methodology that extends segregation metrics to incorporate daily mobility and also quantify the effects of the urban form. We will analyze the dynamics of mixing and also discuss how it can be disentangle from residential segregation, policentricity or monocentricity of cities. This methodology also allows us to quantify the racial exposure of residents at a neighborhood level (census tract), the substrate of the second goal. We propose to explore how this daily exposure predicts political support and crime outcomes. Our hypothesis is that supporting welfare state, and thus voting left-wing political parties, could be a good proxy of social cohesion. We will use data from Los Angeles, Boston and Bogota. Less is clear about crime statistics with respect to segregation.

Aditional material:  Project proposal

Behavioural Consequences of Group Threat – Ethnic Derogation on YouTube in the Aftermath of Threatening Events

Project leader: Christoph Spörlein, Stephan Dochow

Short title: Ethnic Derogation

Research question: To what extent is the spread of ethnic insults in comment sections of YouTube videos driven by social contagion and social selection effects and how do these patterns change during periods of heightened intergroup conflict?

Methods: Regression analysis

Data: Comment data from YouToube

Description: This projects aims to extent our prior work on the diffusion of ethnic insults in the comments of YouTube videos of German political talk shows. Using ideas from the social contagion and social norms literature, we contested that ethnic insults propagate faster through online comment networks in periods of heightened intergroup conflict (i.e., in the aftermath of sexual assaults and terrorist attacks). Results suggest that ethnic insults appear socially contagious but are likely due to processes of social selection.

Our aim is to extend this research in a number of ways: first, we propose to explore the generalizability of these findings by following a cross-national comparative design and analysing YouTube comments from political talk shows of other European countries which experienced terrorist attacks or sexual assaults by minority members in recent years (i.e., Belgium, France, Great Britain or Sweden).
Second, we intent to explore aspects of political mobilization as drivers of social selection by using national elections to study changes in the content of social media comments. We plan to integrate a party political perspective by investigating the link between ethnic insults and the reference to parties and their political programs in the lead-up to national elections. During the relevant time frame (2015-2018), several important elections took place (i.e., Brexit or the German election in 2017).

And third, we intent to abandon the assumption that YouTube comments are closed systems and explore the contribution of cross-platform sharing of YouTube videos to the volume of comments as well as their content. Our guiding hypothesis would state that the text accompanying the video as part of a tweet will contribute the social selection and thereby to the prevalence of ethnic insults.

Since this project aims to analyse text data from non-German languages, non-German native-speakers are strongly encouraged to apply. The project will mainly work with textual data and use various regression-based methods to answer its research questions.

Aditional material:  Project proposalPaper Spörlein & Schlüter

Analyzing the Social Integration of Refugees through Text Analysis of Online Social Media

Project leader: Claire Kelling, Mohsen Mosleh

Short title: Integration and social media data

Research question: What are the main barriers that refugees face in regards to integration into host communities? How do these barriers differ between sanctuary cities and cities that have introduced bans (temporary or not) against refugees?

Methods: Natural Language Processing and Text Analysis, Short Text Topic Models, (Structural) Topic Modeling and Sentiment Analysis

Data: The primary data of this study includes users’ comments on news channels on the topic of refugees posted on Facebook. As the secondary data source, we investigate the generalizability of our results on data collected from other online social media such as Twitter.

Description: In order for policymakers to assess the social integration of refugees into their communities and to decide where to intervene, they must understand the reactions of the current residents. Social media provides a way to gauge the public reaction to several kinds of events, whether it be international, national, or local. Public reaction to refugees and asylum seekers is a subject that has been of recent interest to political scientists and those who study international relations. In this project, we study the reaction of current residents as presented by their comments on articles from news channels about refugees, through Facebook and Twitter data. We will analyze the sentiment of the comments and fit topic models to find the most prevalent topics with regard to refugees. We also investigate the emotion (e.g., fear, anger, sadness) of the comments and the interactions with the use of topics. We compare the results across cities that instituted a ban with those that were declared as a sanctuary. Using the results of these analyses, we hope to examine the integration of refugees into host communities. For example, we hope to understand if members of a community are fearful or welcoming, for example, of refugees.

A test of the Schelling model of spatial segregation

Project leader: Carolina Zuccotti, Jan Lorenz

Short title: Testing the Schelling model

Research question: How can the Schelling model help us understand the evolution of spatial segregation in different contexts?

Methods: The research project is divided into two parts: Data analysis and agent-based modeling/simulation. The first part consists of estimating tipping points for different cities/contexts. Following Card, Mas et al. (2008), tipping points can be estimated regression discontinuity techniques. The second part consists of using the information on tipping points for the development of simulations that can inform us about the evolution of spatial segregation in those cities. By focusing of comparisons, and using the theories available to us, we will discuss the different elements that may play a role in the evolution of spatial segregation in different contexts.

In terms of the cases to be studied, an option would be, for example, do a comparative study of cities in different countries; or cities that are known for being more tolerant than others (e.g. Card, Mas et al. (2008) show for the US that tipping points are higher in more tolerant cities). For this we would need detailed neighbourhood Census data for different years. Or we could also go for a more in depth analysis, perhaps focusing on one country only, and work with Census microdata or panel data at the individual level (e.g. the Sample of Anonymized Records or the UKHLS in the UK) in combination with aggregated Census data. This would allow using additional information in the creation of tipping points such as the role of migrant background, individuals’ socioeconomic characteristics or even the neighbourhood of origin, which would then lead to developing more informed simulations.

Data: The analysis requires that we have detailed information on neighbourhoods’ characteristics for different years. This is obtained, in general, though aggregated Census data. If focusing on one country, we could also work with more detailed information, and combine individual-level data from Census microdata or panel data (e.g. the Sample of Anonymized Records or the UKHLS in the UK) with aggregated Census data.

Description: While the literature considering the consequences of ethnic concentration and that evaluating the size and scale of concentration/segregation is vast, there is a more limited understanding of the drivers and reasons behind the persistence of ethnic minority concentration and of its association over time with spatially concentrated deprivation. This is especially the case in the European context. Explanations for this phenomenon are varied, but the literature has pointed to two in particular: discrimination in the housing market (also defined as place stratification processes: see Logan and Alba 1993) and individuals’ preferences to live close to members of the same group. In 1971, a paper published by Schelling (1971) made, in fact, an important advancement in acknowledging that individuals’ preferences for in-group members are an important factor in explaining why the spatial segregation of different groups (Black Africans and Whites, in his study) persists over time. Using agent-based modelling, the author argues that mixed neighbourhoods are unstable in nature, and that a tendency towards segregation is always present. Within this model, the concept of “tipping” is fundamental. Tipping refers to the maximum level of concentration of members of another group that an individual can tolerate in the neighbourhood, and which, if surpassed, will theoretically push the individual to leave such neighbourhood. In its original version, the model suggests that segregation will persist – and will become inevitable – also in the presence of individuals with relatively “open-minded” views about sharing their neighbourhood with members of other groups (i.e. individuals with high tipping levels). Although in practise there is no such thing as an “absolutely segregated society” – nor mixed neighbourhoods are necessarily unstable (Card, Mas et al. 2008, Ong 2017) – this theoretical model is useful for explaining why segregation levels persist over time. While discrimination in the housing market is for sure a valid explanation behind the persistence of spatial inequalities, the model demonstrates that preferences play a major role. Even with no housing discrimination, the spatial segregation of groups will remain. The general objective of the research project, to be developed during the summer school, is to identify tipping points for different cities (to be chosen) and use these to simulate trends in spatial segregation (like in Clark and Fossett 2008). Specifically, the idea is to develop simulations that are specific to each city and which can provide us with a comparative perspective on how segregation might evolve in different contexts.

Aditional material: Project proposal

Digital and computational approaches to study migration

Project leader: Ingmar Weber, Emilio Zagheni, Kiran Garimella, André Grow

Short title: Digital migration research

Research question:

  • How are internal and international migration related to educational attainment?
  • Are there processes of self-selection of migrants to sub-national regions of the destinations country?
  • Can we use online data to now-cast or even predict flows of refugees during disasters?
  • Can we map the intra-Europe migration of skills? – Does the assimilation of a given migrant group vary across different countries/regions of a country?
  • Are patterns in Google searches related to migration movements?
  • How does public opinion react to immigration? …

Methods: Data analytics (e.g., regresson models, text analysis, etc.), agent-based computational modelling


  • Information from advertising platforms (e.g., Facebook and LinkedIn)
  • Geo-located Twitter data
  • Census data
  • Survey data

Description: The unexpected surge in refugee flows into Turkey and Europe since 2015 has highlighted the need for timely data and new analytical tools to better understand migration dynamics and the underlying decision processes. Recent advances in computational social science (CSS) have shown that digital trace data, such as tweets, email logs, call records, and advertising data can be used to estimate stocks of migrants and their assimilation. Such data can also be incorporated in increasingly detailed models of migration flows that make it possible to better understand the dynamics and complexity of migration decisions.

In this project, we offer participants of the BIGSSS Summer Schools in Computational Social Science the opportunity to build on our experience in the use of non-traditional data for migration monitoring and the use of new analytical approaches to study migration to answer their own migration-related research questions. We are flexible as to the specific topics that participants want to explore and will develop the details of the research question and design together with them.

Depending on the interest of the participants they could either collect their own data or use existing data collections of Facebook and LinkedIn advertising data. These data sets include information on how many Facebook/LinkedIn users match certain targeting criteria including things such countries previously lived in (for Facebook) or countries where a university degree was obtained (for LinkedIn). Alternatively, we could aid participants with formulating their own questions and collecting the necessary data via the Facebook and LinkedIn advertising platforms or via Google Trends/AdWords. Similarly, participants could collect a subset of tweets from Twitter (e.g., based on a certain hashtag) and conduct a content analysis of the resulting text corpus, or use one of the data sets that we have already produced. Finally, based on existing traditional data sources (e.g., census data), participants could adopt existing models of migration and study aspects of the migration decisions, such as how migration is related to educational attainment.

Depending on the particular topic choice, these projects include the possibility to contribute to ongoing collaborations with the International Organization for Migration (IOM) or the Office of the United Nations High Commissioner for Refugees (UNHCR).

Additional material: Project proposal

Global Migration Modelling with MIDAS

Project leader: Stefano Balbi, Helen Adams

Short title: Global MIDAS

Research question: What are the main root motivations of global migration?

Methods: Agent-based modelling and international migration data

DataThe UN Population Division’s estimates for Total Migrant Stock

Description: This projects aims at building a global model of migration based on the MIDAS (Migration, Intensification, and Diversification as Adaptive Strategies) framework (Bell, 2019). This framework draws on the ‘push-pull-mooring’ (PPM) theory of migration to integrate the influences of social networks, climatic shifts, and opportunities for livelihoods diversification on migration in a single framework. So far MIDAS has been applied to Bangladesh and to the US-Mexico Corridor, and employs a data structure that perfectly matches the UN Population Division’s estimates of international Total Migrant Stock, adaptable to a global model during the two weeks of the summer school. We think that such a global model could help us better investigate the components of utility that can adequately explain international migration flows. A global migration framework built on freely available data for pushes, pulls, and moorings (such as censuses and integrated household surveys) would be an invaluable tool in directing future research (and potentially policy focus) in migration, by highlighting i) the common bases across which migration in different regions and states can be compared, ii) the degree to which existing datasets help explain variation in global migration flows, and iii) the key missing pieces for targeted data collection.
MIDAS is a novel agent-based model of livelihoods and migration, which represents locations as nodes with place-specific opportunities for agents to derive utility—from regular sources of income, as well as from assets and local social and cultural amenities. Some opportunities are common to many places (e.g. teaching jobs, food marketing, etc.) while others are location-specific (e.g. access to mountains or lakes, etc.). Accessing opportunities may carry costs, which may give the agent access to the same opportunity in many places, or only in some (e.g. purchasing a home in a specific place). Agents occupy nodes, accessing a portfolio of different utility sources (described by the cost of access, time commitment, and per-time period utility), and are embedded in a social network. Agents share both income and information across their social network, with the likelihood and cost of sharing varying both with the strength of the link (i.e. closeness of the relationship) and the distance across which the relationship is enacted. At each time step, with agent-specific likelihoods, agents participate in social interactions across their networks, and (again, with agent-specific likelihoods) re-evaluate the appropriate portfolio of utility sources for them to pursue.

Aditional material: Project proposal

Migration Data Innovation for Evidence-Informed Policymaking

Project leader: Michele Vespe, Lorenzo Gabrielli

Short title: Migration data innovation

Research question:

  • How can non-traditional data fill migration data (stock, flows, skills etc.) gaps (timeliness, granularity), complement official statistics and provide anticipatory capacity?
  • What insights can be extracted from data on migrant communities in cities across Europe for integration policies?
  • What is the nexus between mobility and migration?

Methods: Data-driven modelling techniques

Data: Datasets that to be used for the module/project will be social media advertising platforms (migrant stocks estimates, skilled migration), micro-census maps of migrant communities (residential segregation and integration at local level), air traffic passenger derived data (mobility and migration nexus). One or more datasets will be prepared depending on the selected project (to be discussed), with the aim to complement official migration statistics using non-traditional data. Official statistics will also be used for training/validation, including from Eurostat, OECD, UNDESA and the World Bank.

Description: Adequate data is key for evidence-informed policymaking and for a better understanding of the rapidly changing scales and dynamics of migration. One of the Knowledge Centre on Migration and Demography (KCMD) lines of activity is relevant to improving access to migration data, promoting their use across multiple policy areas, as well as exploiting non-traditional data sources to address data gaps. The KCMD and IOM GMDAC launched the Big Data for Migration (BD4M) Alliance, aiming to advance discussions on how to harness the potential of big data sources for the analysis of migration. Concrete research lines using social media, micro census, mobile phones and air traffic data to improve migration data have recently started and this project will further investigate some of them in order to answer pressing policy question.
A first week shall be dedicated to an overview of the data landscape at European and global level and to framing and analysing the relevant migration research questions, which will be further explored during the second week by hands-on activities using data data-driven modelling techniques.

Aditional material: Project proposal