THE PROGRAM'S AIMS AND DEVELOPMENT UNDER THE PROJECT PERIOD
This research program has had two overall objectives. On the one hand, it has sought to develop the theoretical and methodological foundations of analytical sociology. On the other hand, it has sought to improve our knowledge of segregation processes in contemporary Sweden. The research carried out within the program reflects this dual objective. A number of publications from the program concern issues on the borderline between sociological theory and the philosophy of the social sciences. Other publications concern the causes and consequences of segregated residential areas, workplaces, and schools. The purely theoretical and the empirical parts of the research program are closely interlinked, however, in the sense that the empirical analyses have to a considerable extent sought to concretely illustrate how the general theoretical principles can be translated into practice. This applies in particular to the connection between micro and macro; how macro conditions such as segregation patterns can be empirically related to the actions at the micro level that jointly generated them. We have addressed this issue through extensive empirical analyses of Swedish registry data that have sought to identify the factors that affect individuals' actions, such as their tendency to choose certain schools or residential areas rather than others. We then have combined this with large-scale counterfactual simulation models developed on the basis of the empirical results.

The research group has had its emphasis in sociology but has included researchers from other disciplines such as philosophy, mathematics, statistics, and business economics throughout the program period. This interdisciplinary composition has been an important factor behind the success of the research program.

SOME WORDS ABOUT THE IMPLEMENTATION
The program was originally housed at the Institute for Futures Studies in Stockholm, but when the opportunity was offered to build up a permanent research institute around the program's core concerns, the program was moved to the Institute of Analytical Sociology at Linköping University (Campus Norrköping). Although the research environment was excellent at the Institute for Futures Studies, the move to Norrköping was a great advantage for the program. The reason for this is partly that it is always advantageous to carry out major research programs in university environments, but in our case the added value was primarily due to the possibility of founding a permanent institute devoted to the type of research that was developed within the program. This means that RJ's investment and the research carried out within the framework of the program came to have far greater and longer-term effects than would otherwise have been the case.

THE MAIN RESULTS OF THE PROGRAM AND CONTRIBUTION TO THE RESEARCH FRONT
In an extensive research program like this, it is difficult to highlight certain specific contributions that are clearly more important than the others. The selection is therefore necessarily very subjective, but with this reservation in mind, the following publications can be singled-out:

1. Meta-theoretical contributions: Ylikoski (2013) clarifies the distinction between causal and constitutive explanations and shows how important this distinction is when trying to explain how micro processes generate different types of macro outcome. Another important contribution is Hedström & Ylikoski (2014) that discusses meta-theoretical differences between analytical sociology and rational-choice theory. Keuschnigg, Lovsjö & Hedström (2018) further discuss in detail how analytical sociology relates to "computational social science" (see also Boelaert & Ollion 2018).

2. Methodological contributions: Spaiser et al. (2016) shows how machine-learning algorithms can be used to better understand social processes; in their case, how such algorithms allow the data to speak more freely in order to empirically establish how schools' ethnic compositions affect the outflow of Swedish-born children from the schools. An unpublished essay by Arvidsson et al. (2019) furthermore shows that a combination of statistical analyses and large-scale counterfactual simulation analyses constitutes a very powerful approach for better understanding how the macro states to be explained were brought about.

3. Contribution to research on housing segregation: Müller, Grund & Koskinen (2018) use advanced network methods to better understand the flows between residential areas in Stockholm. They find significant support for the "ethnic avoidance" hypothesis but limited support for the "ethnic flight" hypothesis. Another important contribution to research on housing segregation is a not yet published work by Jarvis, Mare and Nordvik, where they analyze the mutual interdependencies that exist between the housing market and the marriage market, and how these affect and reproduce ethnic residential segregation from one generation to another.

4. Contribution to the research on school segregation: Brandén, Birkelund & Szulkin (2018) analyses the extent to which pupils’ school results are affected by the student groups' ethnic composition. They find that pupils in schools with many children with immigrant backgrounds perform on average lower, but that this depends to only a limited extent on the composition of the student population as such. Another important contribution to the research on school segregation is a still unpublished work by Eduardo Tapia where he uses large-scale counterfactual simulation analyses to better understand how the segregation of the school system changes over time. The above mentioned article by Spaiser et al. (2016) also makes important contributions in this area by empirically estimating so-called tipping points that are of considerable importance for understanding why the segregation of the school system develops as it does.

NEW RESEARCH QUESTIONS GENERATED BY THE PROGRAM
In a large and extensive research program such as this, lots of new research issues are always generated because the studies being conducted not only answers already known issues but also lead to new issues and approaches. It does not appear apt to try to catalogue all such new questions here. Instead we focus on new types of research initiatives that the program has made us see the value of. Perhaps the most important of these concerns contemporary developments in data science and their social science significance.

The type of issues that have been most central to this research program is about how individuals in interaction with each other generate different intended and unintended collective outcomes. Many of the central issues in sociological theory are also about such issues, but the data and methods that are traditionally used in the social sciences are not suitable for addressing such issues. Traditional qualitative approaches are well suited for analyzing interactions in small groups but not for analyzing interactions between hundreds or thousands of individuals. This is a significant limitation if, as in this program, one intends to explain large-scale social phenomena such as segregation. However, traditional quantitative analyses of survey data are also not suitable for this purpose. Such analyses are excellent for describing distributions of and relationships between variables even in large populations, but the random selection that is at the heart of the survey approach means that one cannot capture the social interactions and the social networks that the individuals are embedded in.

From a social science methods perspective, developments concerning the analysis of "big data" is of particular interest, especially when the data refers to large groups of interacting individuals. The development in this area goes extremely fast and the combination of powerful computers and highly efficient computer algorithms makes it possible to now address issues that would have been impossible to address empirically just 5-10 years ago. The development of different machine learning methods also opens up the possibility of a type of "quantitative ethnography" where data is allowed to speak more freely to the researchers without first having to be filtered through existing and not always fully reliable theoretical lenses. Within this program, we used such inductive methods to try to determine the ethnic tipping points of schools. The experience from this was very promising, and we have recently submitted an application to the Swedish Research Council (VR) to further explore this.

A related development that we have increasingly realized the importance of during the course of the program concerns the emergence of effective and useful methods for analyzing large amounts of text. This development is of great importance for both qualitatively and quantitatively oriented traditions, but may perhaps be considered particularly important for qualitatively oriented research because it enables researchers to analyze non-numeric text data with a precision that previously only was possible for quantitative numerical data. In addition, some of these methods offer a methodological rigor that many qualitative traditions lack. These traditions are often rich in theories and generate many interesting ideas and concepts, but they often leave much to be desired when it comes to reliability and reproducibility. The new types of methods that are now being developed are considerably more transparent than traditional qualitative approaches, which means that the results become less dependent on who performs the analyses and the research results thus becomes more reproducible. This development, combined with the ongoing digitalisation of our cultural heritage and the increased access to digitalized data from the web, is extremely interesting from a social science as well as a humanistic perspective. The programme's focus on ethnically-related segregation led to an interest in how these types of methods and analyses can be used to better understand the societal discourse on immigration and the emergence of anti-immigrant sentiments. The Swedish Research Council recently granted support for such research at the Institute for Analytical Sociology.

THE PROGRAM'S INTERNATIONAL DIMENSIONS, EXAMPLE CONTACTS, MATERIALS AND SO ON
The research group has been internationally composed as well as focused. Researchers from countries such as Germany, the United States, the Netherlands, Finland, Norway and Peru have been included in the research group and the research results have been published in international scientific journals. The Swedish researchers who have been part of the program have also been internationally oriented and have had significant international networks. The internationalization was also strengthened by the international workshops and the regular seminar series that were part the program. All this meant that the internationalization aspects never had to be problematized because the research group's international composition naturally solved all such problems.

As far as the research material is concerned, however, we have almost exclusively used Swedish data. The reason for this is not that the type of processes that we studied in any way would be unique to Sweden. The reason is instead that Swedish registry data is unique internationally. If, as has been the case with this research program, one is interested in issues relating to the connection between micro processes and large-scale macro outcome, better data simply is not available anywhere else.

HOW THE RESEARCHERS HAVE DISTRIBUTED THE RESULTS TO OTHER RESEARCHERS AND GROUPS OUTSIDE THE SCIENCE SOCIETY, AS WELL AS AND WHAT THE COOPERATION HAS BEEN TAKEN
The researchers have disseminated their results in a variety of ways. First and foremost, it has been done through a large number of publications, some of which have appeared in journals with very high impact scores. The researchers have also presented their research at a large number of international conferences, and have also presented their research at research seminars in Sweden as well as abroad.

In terms of collaboration and dissemination of research results to groups outside the scientific community, different types of channels have been used. New publications and other notable events have been regularly published on Facebook and Twitter and often also on our website. Parts of our research has also been recognized by traditional media. This applies in particular to a study concerning the Sweden Democrats (a large anti-immigrant party), which was presented at DN Debatt (an op-ed page of the leading Swedish daily paper Dagens Nyheter) and led to considerable mass media attention. Our segregation research also has received attention from local and national mass media, radio as well as TV and newspapers.

LIST OF PUBLICATIONS
The publications that are in print or have been accepted for publication are found in a separate publication list which contains over 75 publications. In addition, a considerable number of papers currently are under review at different journals. Open access has been ensured through a combination of measures. For some publications we have chosen to pay for OA and for others we have ensured OA through parallel publishing of preprints on-line. For the latter, we have to a great extent relied on the DiVA portal of Linköping University.