Politics and Governance (ISSN: 2183–2463) 2021, Volume 9, Issue 4, Pages 133–145 https://doi.org/10.17645/pag.v9i4.4436 Article The Role of Emerging Predictive IT Tools in Effective Migration Governance Cristina Blasi Casagran 1, *,† , Colleen Boland 1,† , Elena Sánchez‐Montijano 1,2,† and Eva Vilà Sánchez 1,† 1 Department of Public Law and Legal History Studies, Autonomous University of Barcelona, Spain; E‐Mails: cristina.blasi@uab.es (C.B.C.), colleen.boland@uab.cat (C.B.), eva.vilas@e‐campus.uab.cat (E.V.S.) 2 Department of International Studies, Center for Research and Teaching in Economics (CIDE), Mexico; E‐Mail: elena.sanchez@cide.edu (E.S.‐M.) * Corresponding author † These authors contributed equally to this work Submitted: 29 April 2021 | Accepted: 25 July 2021 | Published: 28 October 2021 Abstract Predicting mass migration is one of the main challenges for policymakers and NGOs working with migrants worldwide. Recently there has been a considerable increase in the use of computational techniques to predict migration flows, and advances have allowed for application of improved algorithms in the field. However, given the rapid pace of technological development facilitating these new predictive tools and methods for migration, it is important to address the extent to which such instruments and techniques engage with and impact migration governance. This study provides an in‐depth examination of selected existing predictive tools in the migration field and their impact on the governance of migratory flows. It focuses on a comparative qualitative examination of these tools’ scope, as well as how these characteristics link to their respective underlying migration theory, research question, or objective. It overviews how several organisations have developed tools to predict short‐ or longer‐term migration patterns, or to assess and estimate migration uncertain‐ ties. At the same time, it demonstrates how and why these instruments continue to face limitations that in turn affect migration management, especially as it relates to increasing EU institutional and stakeholder efforts to forecast or predict mixed migration. The main predictive migration tools in use today cover different scopes and uses, and as such are equally valid in shaping the requirements for a future, fully comprehensive predictive migration tool. This article provides clarity on the requirements and features for such a tool and draws conclusions as to the risks and opportunities any such tool could present for the future of EU migration governance. Keywords European Union; forecasting; migration governance; predictive tools Issue This article is part of the issue “Migration and Refugee Flows: New Insights” edited by Inmaculada Martínez‐Zarzoso (University of Göttingen, Germany / Jaume I University, Spain). © 2021 by the authors; licensee Cogitatio (Lisbon, Portugal). This article is licensed under a Creative Commons Attribu‐ tion 4.0 International License (CC BY). 1. Introduction and Theoretical Framework demand for strategic and effective global migration gov‐ ernance (Robinson, 2018; Triandafyllidou, 2020). Migration is unpredictable. Since 2015, with the advent Indeed, international migration flows have become of the so‐called “migration crisis,” different institutions more diversified and globalised, and are notably marked within the European context have allocated extensive by increasingly restrictive admission policies, relative resources and funding in seeking migration forecast‐ change in labour market dynamics, issues of legal irreg‐ ing or predictive tools (European Commission, 2021). ularity and human trafficking, and new transnational The main objective is to respond to the continual networks and spaces (Arango, 2018). Both institutions Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 133 and academia have identified different drivers that EU‐commissioned assessment concluded that a forecast‐ could impact these migration flows (Castles et al., 2014; ing and early warning tool based on artificial intelli‐ Massey, 1999). Whereas some contend that factors such gence (AI) technology evaluating the intensity and direc‐ as conflict, the economy, and climate are the main rea‐ tion of these mixed migratory flows is feasible (European sons behind a decision to emigrate, others hold that sud‐ Commission, 2021). As the scope of this article por‐ den disruptive elements such as political events or social tends to considering mixed migration flows in light of unrest in a territory could lead to a decision to leave a addressing migration to the EU, it does not extend to country (Carling & Collins, 2018; Raleigh, 2011). As inter‐ all predictive analytics or tools for humanitarian action, national migration flows are complex and uncertain, the although there are extensive, global efforts underway in need for equally complex tools that help understand and this regard (Hernandez & Roberts, 2020). manage migration becomes imperative. Within this context, this study analyses how existing Meanwhile, incorporating algorithmic governance supranational predictive IT tools address these issues to has become common practice internationally, and today achieve effective migration governance in the EU. This new technologies are increasingly embedded in every‐ two‐fold inquiry first asks: What are the main predictive day decision‐making. The potential of computational migration tools and what is their scope? Here we focus power can aid in addressing various gaps in understand‐ on identifying the variables and data sources used to cre‐ ing migration flows, and thus benefit policy. For exam‐ ate the tool models, as well as their underlying objec‐ ple, new technologies enable, inter alia, conducting sub‐ tives and rationale, exploring how this relates to the gov‐ national forecasts within international migration, con‐ ernance of migratory flows to date in their respective sidering dimensions like labour force status and ethnic target countries. Additionally, this study further exam‐ groups, examining micro and macro factors involved in ines how existing prediction or forecasting tools might international migration (by harnessing the potential of affect migration governance within the EU. Hence, the micro models alongside macrosimulations), and simulat‐ second research question is: To what extent can these ing sets of scenarios to understand migration drivers for existing tools facilitate effective migration governance? policy (Wilson & Rees, 2005, pp. 340–341). It should be In essence, this study seeks to provide a comprehensive noted that from a temporal point of view, forecasting is understanding of relevant past and current IT prediction a process of predicting or estimating future events based or forecasting tools for migration, as well as to determine on past and present data, whereas predictions consist of the extent to which such tools can offer improved migra‐ estimating the outcomes for unseen data (i.e., new or tion governance and policy solutions in the EU. test data); for the purposes of this article, the two terms This study includes three main parts. Firstly, it (forecasting and prediction) are used interchangeably. describes the qualitative methodology used for this Most importantly, however, the way international investigation. Secondly, it conducts a comparative exami‐ migration governance is understood and practiced can nation of three IT prediction tools or projects in the field be mutually configured and shaped by technologies for of migration. Lastly, it suggests what a valid prediction migration forecasting. In this sense, international migra‐ tool for migration should include in terms of variables tion governance is a contested field with competing inter‐ of analysis, user interface, data accuracy, modelling, and ests and stakeholders, and predictive tools exercise the mechanisms for explaining and communicating predic‐ potential to introduce or reinforce unequal power rela‐ tions, in order to translate outputs into effective gover‐ tions. In utilizing these tools, those states with more tech‐ nance policies. nological capabilities can further solidify their position in setting the international migration agenda (Beduschi, 2. Cases, Methodology, Data, and Categories 2020). Such tools can also be appropriated to further of Analysis securitize or bolster non‐entrée policies and human surveillance, at the expense of those rights protected In order to answer the two main research questions, by international human rights frameworks (Broeders & we reviewed an extensive number of models, projects, Dijstelbloem, 2016). This is important to bear in mind and tools with the potential to predict migration flows in seeking a truly effective global migration governance arriving to the EU. The IT predictive tools were selected oriented towards adequately benefiting all stakeholders, from a 10‐year time range from 2010–2020. Originally, especially migrants themselves. in the analysis, we identified 18 tools and projects incor‐ In the case of the EU, efforts are underway to achieve porating AI, which were relevant for EU migration gov‐ early warning of migratory movements, to forecast them ernance purposes. However, data were incomplete for or to predict mixed migration flows in particular (Sohst many of these, as some were no longer in use, further et al., 2020). Mixed migration flows are characterized information was not accessible, others were not predic‐ by their irregular nature (often in breach of regulatory tive tools as such, or those responsible for tool operation norms of countries of origin, transit, and destination), were unavailable for interview. As a result, we ultimately and consist of distinct types of migratory populations, selected three tools and projects that could be useful for including groups of refugees, asylum‐seekers, forced forecasting or predicting EU mixed migration flows as a and voluntary migrants, and others. Most recently, an sample for this comparative examination: Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 134 (1) the Jetson tool, funded and operated by UN High dictive tool or project was effective in predicting migra‐ Commissioner for Refugees; tion for effective migration governance. Furthermore, (2) the Early Warning and Preparedness System tool they informed the selection of the three tools for this arti‐ (hereafter EPS‐Forecasting), funded and operated cle’s in‐depth analysis. by the European Asylum Support Office; The other type of interview conducted included a (3) Foresight, currently funded and operated by the group interview with 13 representative European NGOs Danish Refugee Council (DRC). It was initially specialising in migration and based in Bulgaria, Denmark, funded by the Danish Ministry of Foreign Affairs, Greece, Italy, and Spain, on 20 January 2021. The NGOs with the model and user interface developed in col‐ were selected according to their expertise on the ground. laboration with IBM. The main objectives of this interview were twofold: firstly, to understand to what extent this kind of predic‐ It should be noted that the Internal Displacement Event tive tool can be useful for them as end‐users; secondly, Tagging and Clustering Tool (IDETECT), funded and oper‐ to learn what they would like and expect from this type ated by the Internal Displacement Monitoring Centre of tool. (IDMC), is also examined in‐depth in Section 4.1 below, These interviews enabled us to validate and com‐ although this tool is not considered predictive as such. plete the information examined via the document analy‐ Moreover, in listing the final selection of tools above, it sis technique (Corbetta, 2003), lasting between 40 min‐ is equally important to signal that the Jetson project does utes and two hours. The interlocutors were all informed not describe itself as a “tool” as such, but rather a proof about the purposes of our research and were given the of concept, as over the course of a year it conducted opportunity to review this work. The general sample has scoping, research, and piloting, but never produced a been sufficiently representative to enable saturation and user‐facing, interactive tool. At the same time, as it uses triangulation of the information obtained. machine learning for predictive purposes, and pertains We have predefined various categories of analysis to to all of the categories of analysis, it is used here as one improve the consistency of the research. As no previous of the selected “tools,” with these qualifications. literature has identified these, we have extracted differ‐ To answer the research questions, a qualitative ana‐ ent categories from the data collected via an inductive lysis has been carried out. The analysis used two types process. All the data collected via the document analy‐ of research techniques: document analysis and in‐depth sis as well as the interviews were systematically entered interviews. The document analysis provides access to into an Excel spreadsheet and organised by the different empirically objective, common, and reliable evidence categories according to the study’s two key objectives, as (Franzosi, 1998, p. 547). We focused on analysing all the Table 1 illustrates. publicly accessible information published about these tools, including user‐manuals, websites, reports, and aca‐ 3. Comparative Analysis of the Existing Prediction Tools demic publications, totalling approximately 50 sources. Regarding the interview analysis, two kinds of This section focuses on the study’s first research ques‐ in‐depth interviews have been conducted. We would tion as to the main predictive migration tools and their first highlight those carried out from February to March corresponding scopes. As explained in the methodology, 2021, where we interviewed five experts on forecasting we analyse: (1) the Jetson tool; (2) the EPS‐Forecasting or AI tools. The interviewees were selected according tool; and (3) the Foresight tool. To carry out this compar‐ to their expertise, using the snowball sampling method, ative analysis, we identified four categories of analysis in and included: the three developers of the forecasting Figures 1, 2, 3, and 4: target migration flow, timeframe, tools reviewed in this article, an IDMC representative variables of analysis, and data sources. regarding the AI tool IDETECT, and the founding devel‐ Much of recent literature acknowledges that there oper of the Global Database of Events, Language, and is no “one model as the ‘best model’ for all situations” Tone (GDELT) project (which monitors the world’s broad‐ (Bijak et al., 2019, p. 12; Sohst & Tjaden, 2020). In fact, cast, print, and web news in over 100 languages). These using partial or different models with the same data interviews first explored the scope of the different tools, would produce different forecasts. Providing a compre‐ and secondly inquired as to what extent the existing pre‐ hensive forecasting tool entails assessing the type of Table 1. Categories of analyses. Comparative descriptive analysis (Section 3) Component of a valid predictive tool (Section 4) Target flow Variables of analysis Timescale User‐friendly interface Variables of analysis Accuracy of migration data Data sources Best suited model(s) Source: Own elaboration. Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 135 migration flow analysing the corresponding hard data while a refugee or an asylum‐seeker is a legal term sig‐ available, and cross referencing with varied contextual nifying those individuals that lack protection from their data (qualitative and quantitative). The modelling should own country and are entitled to international rights or then be tailored to each of these different flows and protections. As Figure 1 demonstrates, while they are accompanying data to improve accuracy and minimise distinct, sometimes these terms and concepts may over‐ potential bias. lap or are combined in the migration flow being targeted As illustrated in Figure 1, none of these tools fore‐ for prediction. cast regular migration, presumably because this infor‐ Modelling corresponded to the type of migra‐ mation can be more easily obtained, as it can take tion flow under examination. The Jetson project used months to years in applying for admission to another machine learning gravity modelling and multivariate country. By contrast, irregular and forced migration can time series analysis (a total of 11 modelling tech‐ either occur quickly via a sudden event, which could niques), the Foresight tool uses machine learning and be detected and signalled with an early warning system Bayesian network models, and EPS‐Forecasting uses in place, or over longer periods of time with varying adaptive machine learning algorithms, further compared uncertain events that affect migration flows. The Jetson in Section 4.4. project formerly predicted forced, internally displaced The types of movement to be predicted or forecasted people as pertained to the case study of Somalia. relate to how each of the tools have different predic‐ Meanwhile, the Foresight tool focuses its forecasts on tion timescales, portrayed here in Figure 2. The Jetson only forcibly displaced asylum seekers and refugees from tool worked with both monthly predictions and addi‐ a given country. Finally, the EPS‐Forecasting tool seeks tional short time frames, including three‐month predic‐ to anticipate flows of refugees and asylum‐seekers arriv‐ tions, as they sought to test the assumption that sud‐ ing to the EU, which can be unpredictable if migrants den conflict events or external factors like drought and arrive via irregular routes. It should be noted that forced floods would cause population movement towards areas migration is referring to migration that is not voluntary, of humanitarian assistance. The EPS‐Forecasting tool TARGET GROUPS Asylum Seekers and Refugees Forcibly Internally Displaced Displaced Jetson EPS–Forecas ng Foresight Figure 1. Target migration flow by predictive tool. Source: Adapted from Carammia et al. (2020); R. Jimenez, Jetson project, interview, March 11, 2021; A. Kjærum, DRC interview, February 18, 2021; and C. Melachrinos, interview, April 23, 2021. FORECASTING TIME FRAME 1 Month 12 Months Jetson EPS–Forecas ng Foresight 1 Week 3 Months EPS-Forecas ng Jetson Figure 2. Timescale of the predictions according to each tool. Source: Adapted from Carammia et al. (2020); R. Jimenez, Jetson project, interview, March 11, 2021; A. Kjærum, DRC interview, February 18, 2021; and C. Melachrinos, interview, April 23, 2021. Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 136 provides weekly predictions, as well as offers short‐term It is of note that both the EPS‐Forecasting and the predictions of up to one month that can be expanded FORESIGHT tools take into account the area and/or via a user‐selected variable. Finally, the Foresight tool ini‐ country of origin’s economy or governance to refine tially predicted mixed migration flows one to three years their predictions. In addition, FORESIGHT uses data on in advance but was then redesigned (among other rea‐ geological, hydrological, and meteorological events to sons) for better accuracy to focus on forced displacement inform predictions. via one to three‐year predictions. Figure 4 shows that several data sources are used As displayed in Figure 3, these tools include differ‐ for the predictive tools’ models. Here it is notewor‐ ent variables of analysis; this relates to how selecting thy that Internet usage can assume an active role tool variables depends on the underlying migration the‐ in irregular migrants’ choices and routes (Lif, 2016). ory or assumption, and the identified migration flows or The EPS‐Forecasting tool incorporates real‐time online tool objective, further detailed in Section 4. The most data sources in its predictions. However, none of the common variable, used by all three tools, is information tools incorporate data from social media platforms like on violence or conflict in a particular territory, due to Twitter, Facebook, or Instagram; this is perhaps due to a consensus that this represents one of the most influ‐ an ongoing debate as to the ethical concerns behind use ential factors in migration flows. In addition, both the of migrants’ personal data. JETSON and EPS‐Forecasting tools use information on Figure 4 also demonstrates how all of the predic‐ political events to feed their predictions. In particular, tive tools rely mainly on open‐access data. When it was the EPS‐Forecasting tool identifies political events and in operation, the Jetson project used data from the social unrest, which could be caused by riots or rebellion Armed Conflict Location & Event Data Project (ACLED), within a specific country or territory. the Food Security and Nutrition Analysis Unit, the Food VARIABLES Polical Events Violent Conflict Economy Jetson Governance EPS–Forecas ng Foresight Market Prices Natural Disasters Social Unrest Figure 3. Variables of analysis included in each predictive tool. Source: Adapted from Carammia et al. (2020); R. Jimenez, Jetson project, interview, March 11, 2021; A. Kjærum, DRC interview, February 18, 2021; and C. Melachrinos, interview, April 23, 2021. DATA SOURCES Jetson EPS–Forecasng Foresight ED U O IM x a LT s Pr a C T m ict nk d A te at NA FA E en -D Ba ra fl L AL W e on SD GD og on AC m FS EM Tr SW ld Fr EP gle or ta al o s Go Da p p U Figure 4. Data sources for the predictive tools. Source: Adapted from Carammia et al. (2020); R. Jimenez, JETSON project, interview, March 11, 2021; A. Kjærum, DRC interview, February 18, 2021; and C. Melachrinos, interview, April 23, 2021. Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 137 and Agriculture Organization of the UN, and the Somalia UN, 2021), policy changes (European Country of Origin Water and Land Information Management Project. Information Network, 2020), or other specific events In addition to making use of GDELT, Google Trends and (GDELT, 2021). As illustrated in Figure 4, two of the three Frontex data, the EPS‐Forecasting tool focuses on predict‐ tools studied here use the public dataset ACLED, which ing asylum applications using their own weekly EPS‐Data offers the “dates, actors, locations, fatalities, and types acquired from exchange with EU+ countries (EU mem‐ of all reported political violence and protest events” in ber states and associated countries), the latter of which real‐time throughout the world (ACLED, 2021). provide data on 19 standardised, disaggregated indica‐ Additionally, a valid predictive tool working with tors, or the EPS‐Data. A subset of those data is then used quantitative data could benefit from a monitoring team for the EPS‐Forecasting tool (Albertinelli et al., 2020). that can simultaneously examine or provide qualitative The Foresight tool utilizes 18 different sources of data, data to validate or correct the tool’s quantitative results. with main sources including EM‐DAT, World Bank, the An example includes the work of the IDMC, which oper‐ Uppsala Conflict Data Programme, and ACLED. ates an AI tool and a real‐time monitoring platform as part of its work in investigating displaced individuals 4. Towards a Valid Predictive Tool for Migration worldwide. IDMC enlists a team of about eight to 10 Governance in the EU experts to monitor the raw quantitative data, ensuring it corresponds to real‐time displacements. Following this brief comparison of three main predic‐ Ideally, when assessing successful integration strate‐ tive migration tools, this study fleshes them out further gies, predictions of migratory flows should not only and assesses the most relevant criteria in determining include parameters related solely to countries of origin, what a valid predictive tool for effective EU migration but also variables pertaining to the country of destina‐ governance should look like. As such, this section par‐ tion, such as macroeconomic indicators (unemployment, tially responds to the second research question: To what job vacancies) or migration and integration policies extent can those existing tools facilitate effective migra‐ (migration caps, visa regulations, etc.). This considera‐ tion governance? In particular, the following aspects are tion is adequately incorporated into the EPS‐Forecasting analysed: necessary variables for the models, interface model, and Foresight tool developers also tested this of the tool, data and prediction accuracy, and adequately origin–destination relationship when initially looking at suited models. mixed migration in its preliminary modelling. In using such variables, both macro and micro synergies must be 4.1. Necessary Variables to Be Included in the Tool considered to sufficiently address populations with mul‐ tiple characteristics. What follows is further exploration of how and why Big data has demonstrated advantages in that it could to select variables and establish parameters for these allow for identifying a particular phenomenon affecting tools or their models, providing context for why the migration flows, can permit measurement of variables tools analysed above did so. First of all, a valid pre‐ at a regional or even local level, and provide input in dictive tool should include variables of analysis related updating a corresponding algorithm (open‐access data to the demography of the selected countries of ori‐ sources usually provide information only at the national gin. For instance, variations in population size within level). For instance, if a conflict occurs in a specific region, countries of origin, transit, and destination can point this could be accurately measured by identifying users’ to the existence of migration flows in those territo‐ reactions in that particular location via social media. ries, and even provide data regarding migration routes. In recent years, several universities and research centres There are many public sources of information that pro‐ have been working with big data in forecasting displace‐ vide quantitative data related to migration, including, ment globally, examining social media for sentiment ana‐ among others, population statistics (e.g., City Population, lysis, and in evaluating economic and social variables 2021; UN, 2021), number of displaced individuals (e.g., (European Commission, 2017; Singh et al., 2019). In this International Organization for Migration, 2020; OECD, sense, an ideal model could include behavioural and sen‐ 2020), number of asylum applications (e.g., Eurostat, timent analysis collected online. In utilizing data on sen‐ 2021; UN High Commissioner for Refugees, 2020), and timent, for example, a tool could identify posts, “likes” or main migratory routes (e.g., Frontex, 2020). Other rele‐ interactions shared by migrants on their social networks, vant data sources could provide information on the given which could then be properly anonymised (according to theoretical drivers of migration, including armed conflict the terms of the EU General Data Protection Regulation, and violence (ACLED, 2021), climate disasters (EM‐DAT, as further described in Section 4.2) and codified as 2021; European Centre for Medium‐Range Weather positive or negative reactions towards potential desti‐ Forecasts, 2021), low levels of development (The World nation countries, routes, and relevant migration topics. Bank, 2021), food insecurity (Integrated Food Security Although several models that incorporate data on indi‐ Phase Classification, 2020), irregular governance (Rulers, vidual preferences and opinions have been proposed in Elections, and Irregular Governance, 2021), unafford‐ recent years (Alam et al., 2020; Papakyriakopoulos et able food (Food and Agriculture Organization of the al., 2018, p. 9), there is no evidence that this type of Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 138 data has been used to feed an actual predictive migra‐ Lastly, potential end‐users find it important that such tion mechanism to date. EPS‐Forecasting uses Google a tool could provide predictions automatically, without Trends in generating their predictions, and a valid pre‐ having to rely on inputting their own data, which is often dictive tool could similarly incorporate insight from pre‐ insufficient or even non‐existent. In this regard, the tool vious studies exploring how variables from “Google could have a mechanism in place via machine‐learning queries” can predict migration flows (Böhme et al., 2020; techniques, or a human monitoring unit responsible for Lif, 2016). updating information. Such a mechanism would guaran‐ Finally, a valid predictive tool could also benefit tee, for example, that one year after the tool is opera‐ from GDELT data, which provides real‐time, open source, tional, predictions and functionalities of the tool remain spatio‐temporal data sets on crises, drawn from the accurate, considering all recent phenomena impacting world’s news media. The project codifies news and infor‐ migration flows. mation about events via the (political science discipline) To ensure the effectiveness of the tool, it is essential Cameo system and provides a range of users with the var‐ to engage with experts or conduct pilot tests among end‐ ious methods with which to analyse this georeferenced, users in real‐world environments, as part of the valida‐ globally contextualised data. As mentioned above, the tion process. Two of the three tools examined and still in EPS‐Forecasting tool already uses GDELT data, selecting operation receive formal validation from in‐house versus a set of 240 events potentially driving migration and external users. displacement, divided into five macro‐categories (politi‐ cal events, social unrest, conflicts, economic events, and 4.3. Accuracy of Migration Data and Predictions governance‐related events). While there are many sources of information, at the 4.2. Tool Interface and End‐User Involvement same time the inherent uncertainty and complexity of the migration phenomenon, and its study, present sev‐ A valid predictive migration tool must be designed so that eral issues with data. The first problem encountered its selected end‐users can easily operate it. In this regard, includes access: Given the nature of migration, as well as we arrive at three main conclusions based on our analy‐ the multiple actors involved, there may simply be a lack sis of the interviews conducted with 13 NGOs based in of data, it could be incomplete, or it could be impossible Bulgaria, Denmark, Greece, Italy, and Spain. to obtain (Felkai Janssen, 2020; Kjærum, 2020). For exam‐ Firstly, end‐users should be able to select and dese‐ ple, as some of the events being anticipated are sudden lect the main demographic criteria. In this sense, at the or unpredictable, a frequent possibility with mixed migra‐ very least, an origin–destination–age–gender migration tion flows, the short life of the process makes data col‐ matrix should be made available. In addition, potential lection itself difficult. Another example includes how, ini‐ end‐users in our study find it relevant to have criteria tially, Foresight sought to use household level informa‐ such as nationality, language spoken, ethnic group, and tion, but ended up using national level data, as obtaining skills of migrants. Although it would be ideal to include the former dataset was not feasible. all these criteria, concerns could arise in terms of data Moreover, raw data are not always accurate. protection and ethical requirements. Particularly, arti‐ Although accuracy of migration data has improved over cle 5(1)(c) of the EU General Data Protection Regulation time (Wilson & Rees, 2005, p. 339), some of this avail‐ establishes that the processing of personal data must able information can often be inconsistent or incomplete, be “adequate, relevant and limited to what is neces‐ especially due to existing bias in the way data is collected sary in relation to the purposes for which they are pro‐ (for example, there is currently a lack of gender disaggre‐ cessed” (European Union Regulation of 27 April 2016, gated information in migration research). Assuming that 2016, p. 35). In essence, migrant’s personal data should data sources are sufficiently accurate, the next challenge be processed at a minimum, only as needed for the pur‐ lies in achieving accurate predictions. Many scholars pose. In particular, as migrants meet the definition of have sought to minimise uncertainty through their mod‐ “vulnerable and minority groups,” any tool processing els and studies (Bijak et al., 2019; Wilson & Rees, 2005). their personal information should ensure that access to However, there is still a large margin of error and various this information is limited to specific selected end‐users, biases in migration predictions, due to events that are and that this information access’ purposes comply with very difficult to predict (Bijak et al., 2019). An example the EU Charter of Fundamental Rights. includes the economic cycle of a country. Although many Secondly, according to the surveyed potential end‐ studies have aimed to identify vulnerable economies fac‐ users, the interface of the tool could include interactive ing risk in recent years (Brei et al., 2020; Stamer, 2019), user features like the potential to compare previous peri‐ in general, economists are not always capable of accu‐ ods, or scenario‐building capacities. For instance, users rately predicting country recessions (Thomson‐DeVeaux, find it particularly attractive if a tool can allow them to 2020). Consequently, emigration driven by an economic break down migration to the EU by routes, and to iden‐ downturn is even harder to forecast. tify and visualise population groups according to spe‐ In the last 10 years, many methods have studied how cific attributes. to explicitly identify errors (Wilson & Rees, 2005, p. 340) Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 139 and improve predictions. One strategy for reducing high debates, and offering long‐term perspectives. Both at levels of error includes assessing data sources according the EU and international level, scenario‐building consti‐ to the concept of “true flow.” True flow is understood as tutes a form of anticipating and preparing for future the number of migrants a flow would amount to accord‐ migration patterns, or understanding alternative future ing to a given definition of migrant, if immigration could flows (Kjærum, 2020). This involves using models in an be monitored perfectly. Another strategy could focus on explanatory capacity, with experts predicting how differ‐ assessing the margin of success by pulling everything ent scenarios affect the composition of migration flows backwards (e.g., before 2015 in the case of migration in a systematic process. However, disagreement among predictions), and looking at the ratio of error to the suc‐ such experts and the difficulty of providing actionable cess of “past” forecasts. Through this method, the model conclusions complicates this method. A strategy to mit‐ could include benchmarking against the actual change in igate lack of consensus and best incorporate expert opin‐ migration flows from one year to the next, or benchmark‐ ion includes the Delphi survey, which surveys experts in ing against the previous prediction technique in place. multiple rounds to reduce individual biases and promote consensus (Sohst & Tjaden, 2020). 4.4. Best Suited Model(s) 5. From Predictions to Policy Finally, it is important to recognise how lack of or errors in data, as well as differences in conceptions and the‐ In addition to the considerations outlined in Section 4, ories of migration, relates to how a tool is designed or the feasibility and effectiveness of predictive tools and models should be chosen. The fact that a tool relies on projects for migration governance require extensive only one forecasting model could also lead to uncertainty assessment of who manages them and how. These new (Bijak, 2016; Disney et al., 2015), as any single model will technologies present a set of tools to understand and have only a limited number of variables. Even if a par‐ anticipate migration, as the information they provide can ticular model works well for a certain period, one sole be analysed to inform long‐term, good governance efforts. event might change everything, and from that point on However, there are still several potential challenges and the predictive tool might have to consider a different gaps in first providing these tools for policymaking, and degree of uncertainty. Advocates for the advantages of then in translating predictions into policy decisions. Bayesian models, a model used in Foresight, argue that For one, economic and political considerations they theoretically can provide for the three elements of remain inextricably linked. Policy makers may view eco‐ uncertainty inherently characteristic to migration predic‐ nomic costs as a barrier to even utilizing such tools tion: the uncertainty of future events, of migration data, in the first place. This could be addressed by incorpo‐ and of different forecasting models producing different rating open‐source data and engaging in resource and results. Moreover, the Bayesian approach can allow for knowledge sharing, as the Jetson project did by using incorporating historical trends, expert judgements, and open‐source code and volunteer developers (R. Jimenez, various model specifications (Bijak & Wiśniowski, 2010). Jetson project, interview, March 11, 2021). Furthermore, Meanwhile, a gravity model like that used in the sufficient political will could assist in obtaining funds. Jetson experiments employs demographic variables, Still, the earlier described competing interests of migra‐ using population size to map people over spatial dis‐ tion governance do present continued concerns with tance; this type of approach may offer insight into future political risk, and multi‐stakeholder collaboration is a flows’ structures, rather than magnitude (Bijak et al., complicated negotiation. As a result, these tools may 2019). In fact, it has been increasingly used in forecast‐ remain for internal use by the developer and funder to ing numbers of people who may want to move along mitigate political sensitivities. with their potential destination, as has been conducted For example, while the EPS‐Forecasting algorithm is in a World Bank report exploring potential future climate both replicable and transparent, it remains internal in migration (Rigaud et al., 2018). Finally, the literature compliance with the dissemination policy of some of has identified how agent‐based modelling approaches its data, which is provided in a two‐way exchange with in combination with machine learning are promising governments under the condition it is not used publicly to provide predictions based on indicators or decisions (Carammia et al., 2020). However, in addition to being to migrate at the individual or household level (corre‐ used by its initial, internal stakeholder, the Foresight sponding to micro‐level theories of migration; Searle & tool is also used by external stakeholders who have pro‐ van Vuuren, 2021). These agent‐based models are often vided positive feedback and is made available to other used to explain migration, rather than predict (Klabunde stakeholders or governments (A. Kjærum, DRC interview, & Willekens, 2016). February 18, 2021). As a way to mitigate the uncertainty of some of In all cases, translating the available and significant the variables, model‐based estimations could be com‐ wealth of analysis, information, and technological poten‐ plemented by expert judgment. Experts are especially tial to actual decisions for migration governance requires necessary for assessing past forecast errors, identify‐ continuous efforts towards enhanced communication, ing random variations in the models, enriching policy participation, and transparency, as Figure 5 illustrates. Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 140 FEEDBACK Quesons & Input Form Policymakers & Media TOOL Back to Tool N PREDICTIONS and Analysts ATIO IC UN PARTICIPATION M Data PROACTIVE M Producon CO POLICY SHARING DECISIONS TR Research & AN SP Parcipatory AR E NC Events Y ANALYSIS w/ Policymakers Tool Team Makes Media Sense of Data Figure 5. From tool predictions to policy decisions. Source: Own elaboration. To utilise the explanatory and empirical value of migra‐ demonstrate the impact that this technology and data tion predictions for governance, the nature of translating offer, and it would act as a check on decision‐makers: research into policy necessitates both human analysis in It would either monitor that they do not assert their the form of a research team, as well as a policy unit or agenda or self‐interest in the direction of one tool or team corresponding to a given tool, in order to ensure monitor that they do not compare and select different the tool is useful for governance. In effect, while the tool tools to simply match tool results to their desired policies. analysts and even invited outside experts make sense of Moreover, recent literature suggests that elites across tool results, the engagement unit could serve as the liai‐ countries—including those drawing from the business, son with policymakers. The latter team can ensure infor‐ media, and civil society in addition to those from govern‐ mation is provided in a palatable, non‐technical format, ment and research—are moderately open to and confi‐ as overly scientific explanations can ultimately not prove dent in select global institutions (Scholte et al., 2021); useful (Albertinelli et al., 2020). Such an engagement or increased transparency and engagement with such elites policy team would also maintain an established link with in this prediction to policy dialogue could in turn further policymakers and stakeholders, so that the tool’s analysis legitimise global migration governance. unit could communicate results in a two‐way dialogue. Finally, the entire process of creating policies for For example, the European Asylum Support Office has populations would be incomplete without visibility into maintained such an ongoing, two‐way conversation and the lived experiences of migrants themselves, either engagement with EU+ countries that seek early warning obtained by a tool’s analysis unit in tool design or inter‐ and forecasting of asylum seekers, and provides them pretation of results, or by the policy team in engag‐ with up to date, limited release, comprehensive reports ing with decision‐makers. The Jetson project, for exam‐ regarding the state of early warning and forecasting to ple, incorporated this at the analysis stage, conducting date or the impacts of recent trends like the Covid‐19 qualitative interviews among the displaced individuals pandemic; they are working towards a joint predictive being monitored, to validate variables utilized in running exercise among such stakeholders in 2021 (European their simulations. Asylum Support Office, 2017; C. Melachrinos, interview, Rather than reactive, short‐term decision‐making, April 23, 2021). long‐term policy provision that incorporates interrelated Evidently, the dialogue between scientists and poli‐ policy areas beyond those strictly related to migra‐ cymakers is not simply a matter of submitting palatable tion and integration measures are key (Szczepanikova desk research resulting from the tool, and the proposed & Van Criekinge, 2018). For example, rather than solely engagement team’s role would mean organizing partici‐ anticipating and preparing for migration in the short‐ patory, collective, ideally public, events like workshops, term, translating migration predictions could mean focus groups, and webinars, where policymakers could reforming or bolstering public health or welfare policy present questions and input. This transparency would for inclusive growth, as was indicated in this study’s Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 141 interviews with relevant NGOs; in short, it is key to tions, properly communicating the uncertainty of migra‐ account for all stakeholders in human mobility. Given tion forecasting. that migration governance should resemble a horizontal Most importantly, the way international migration process, the tool and its proposed units should engage all governance is understood and practiced can be mutu‐ relevant multilevel governmental and non‐governmental ally configured and shaped by technologies for migra‐ actors, including international organisations, EU institu‐ tion forecasting. The literature to date has pointed to the tions, states, regional or local governments, civil society risk of new technologies and IT predictive tools intensify‐ representatives, academic experts, and even the private ing global or regional asymmetries, and curtailing human sector. Finally, such a long‐term view could mean also rights, which is at odds with effective migration gover‐ engaging origin and transit countries, enacting a specific nance (Beduschi, 2020; Broeders & Dijstelbloem, 2016; policy or developing capacity‐building projects, rather Crépeau & Atak, 2016). than focusing exclusively on host countries. At the same time, this analysis of current predictive tools and related projects has demonstrated that if devel‐ 6. Conclusion oped and operated in a transparent and accountable process, such tools can alternatively be leveraged as an This study explores existing developments in predictive equalizer in the field of migration management. Firstly, tools for migration. After reviewing the literature and knowledge and resource sharing via open source and available documents, following up with tool developers access can reduce costs or barriers that otherwise exac‐ and interviewing potential end‐users, three predictive erbate power differentials in global migration manage‐ migration tools and projects were studied and compared, ment. Secondly, these tools boast the capacity to incor‐ namely, the Jetson tool, the EPS‐Forecasting tool, and porate a broad and diverse range of actors—individual the Foresight tool. The analysis demonstrates the chal‐ states, intergovernmental organisations, civil society, lenges in providing for effective interaction and feedback interdisciplinary researchers, and technical experts, as among tool developers and end‐users, and how each of well as migrants themselves—that can offer the most these tools has a different scope, data sources, models, comprehensive data and insight for effective governance. and validation mechanisms, according to their goals. In conclusion, this study indicates that there are cur‐ The discussion fleshes out the difficulties to date rently several initiatives underway in pursuit of an IT pre‐ encountered by those managing predictive tools in offer‐ dictive tool that could assist in migration management. ing predictions that could serve the totality of rele‐ The EU and international organisations are working with vant stakeholders in their intent to develop strategic more data and insight than ever before understanding migration governance (Robinson, 2018; Triandafyllidou, migration flows for governance purposes. Nonetheless, 2020). In illustrating this, the study emphasises four main to render them both useful and relevant, predictive tools aspects that should be taken into consideration to create, require continual monitoring and re‐assessment, con‐ or transform a predictive tool into, a valid predictive tool sistent and expanded multi‐stakeholder collaboration, for effective migration governance. These main elements and further efforts toward sufficient communication and provided here include guidance on variables to incorpo‐ translation of outputs, in order to better assist concrete rate into the models, ways to involve end‐users in the pro‐ policy decision‐making and outcomes. cess, adequate levels of accuracy, and tailoring modelling to the prediction or governance objective. Furthermore, Acknowledgments mechanisms to convert predictions into policy decisions were ultimately emphasized. The authors are grateful to Alexander Kjaerum, Kalev Again, there are unresolved challenges related to pre‐ Leetaru, Constantinos Melachrinos, Rebeca Moreno dictive migration tools at all stages of the development Jimenez, Teddy Wilkin, and Sylvain Ponserre, for their key process: While some existing tools have managed to val‐ inputs as part of this study, and to Daniel Morente and idate their predictions and achieve an acceptable level Christopher Neeson for their relevant feedback on ear‐ of accuracy, others still struggle to obtain accurate pre‐ lier drafts of this article. This study is part of a project dictions even over a few weeks’ time. Even if this first that has received funding from the EU’s Horizon 2020 issue is overcome, the later stage of converting such pre‐ research and innovation programme under grant agree‐ dictions into decisions for governance remains an ongo‐ ment No. 882986. ing process for all pre‐existing tools. The ultimate goal of predicting migration flows for governance should be Conflict of Interests to enable policymakers and appropriate stakeholders to make prudent and robust decisions, by illustrating a clear The authors declare no conflict of interests. causal relationship between migrant arrivals and neces‐ sary policies for managing future migration. Section 5 References explores this crucial step of engaging with policymak‐ ers, noting how it is essential to offer clear explanations Alam, M., Gesese, G. A., Rezaie, Z., & Sack, H. (2020). of comprehensive policy solutions based on the predic‐ MigrAnalytics: Entity‐based analytics of migration Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 142 tweets. In K. Taylor, R. Gonçalves, F. Lecue, & J. ments in the modern world. Ethnic and Racial Yan (Eds.), Proceedings of the ISWC 2020 posters Studies, 38(13), 2355. https://doi.org/10.1080/ and demos track, co‐located with 19th International 01419870.2015.1050048 Semantic Web Conference (pp. 74–78). ISWC. City Population. (2021). Population statistics for coun‐ Albertinelli, A., Alexandrova, P., Melachrinos, C., & tries, administrative divisions, cities, urban areas Wilkin, T. (2020). Forecasting asylum‐related migra‐ and agglomerations—interactive maps and charts. tion to the European Union, and bridging the gap https://www.citypopulation.de between evidence and policy. Migration Policy Prac‐ Corbetta, P. (2003). Social research theory methods and tice, 10(4), 35–41. techniques. SAGE. Arango, J. (2018). Explaining migration: A critical view. Crépeau, F., & Atak, I. (2016). Global migration gover‐ International Social Science Journal, 68, 105–117. nance: Avoiding commitments on human rights, yet https://doi.org/10.1111/issj.12183 tracing a course for cooperation. Netherlands Quar‐ Armed Conflict Location & Event Data Project. (2021). terly of Human Rights, 34(2), 113–146. https://doi. Disaggregated data collection, analysis & crisis map‐ org/10.1177/016934411603400203 ping platform. https://acleddata.com/#/dashboard Disney, G., Wiśniowski, A., Forster, J. J., Smith, P. W. F., & Beduschi, A. (2020). International migration manage‐ Bijak, J. (2015). Evaluation of existing migration fore‐ ment in the age of artificial intelligence. Migration casting methods and models: Report for the Migra‐ Studies. Advance online publication. https://doi.org/ tion Advisory Committee. ESRC Centre for Population 10.1093/migration/mnaa003 Change. Bijak, J. (Ed.). (2016). Migration forecasting: Beyond the EM‐DAT. (2021). The international disaster database. limits of uncertainty. Global Migration Data Analy‐ https://www.emdat.be sis Centre. https://gmdac.iom.int/sites/gmdac/files/ European Asylum Support Office. (2017). Quantitative gmdac_data_briefing_series_issue_6.pdf assessment of asylum‐related migration: A survey of Bijak, J., Disney, G., Findlay, A. M., Forster, J. J., Smith, methodology. https://easo.europa.eu/sites/default/ P. W. F., & Wiśniowski, A. (2019). Assessing time files/publications/Quantitative_assessment_of_ series models for forecasting international migra‐ asylum_related_migration_1.pdf tion: Lessons from the United Kingdom. Journal of European Centre for Medium‐Range Weather Forecasts. Forecasting, 38, 470–487. https://doi.org/10.1002/ (2021). Advancing global NWP through international for.2576 collaboration. https://www.ecmwf.int Bijak, J., & Wiśniowski, A. (2010). Bayesian forecasting of European Commission. (2017). Big data and alternative immigration to selected European countries by using data sources on migration: From case studies to pol‐ expert knowledge. Journal of the Royal Statistical icy support. European Comission Knowledge Centre Society, 173(4), 775–796. https://doi.org/10.1111/ on Migration and Demography. https://knowledge4 j.1467‐985X.2009.00635.x policy.ec.europa.eu/migration‐demography/big‐ Böhme, M. H., Gröger, A., & Stöhr, T. (2020). Searching data‐alternative‐data‐sources‐migration_en for a better life: Predicting international migration European Commission. (2021). Feasibility study on a with online search keywords. Journal of Development forecasting and early warning tool for migration Economics, 142, Article 102347. https://doi.org/ based on artificial intelligence technology. Publi‐ 10.1016/j.jdeveco.2019.04.002 cations Office of the European Union. https://op. Brei, M., Borio, C., & Gambacorta, L. (2020). Bank inter‐ europa.eu/en/publication‐detail/‐/publication/ mediation activity in a low‐interest‐rate environment. 5afa29f0‐700a‐11eb‐9ac9‐01aa75ed71a1 Economic Notes, 49, Article e12164. https://doi.org/ European Country of Origin Information Network. 10.1111/ecno.12164 (2020). About ecoi.net. https://www.ecoi.net/en/ Broeders, D., & Dijstelbloem, H. (2016). The datafication about/about‐ecoi.net of mobility and migration management: The mediat‐ European Union Regulation (EU) No 2016/679 of the ing state and its consequences. In I. Van der Ploeg European Parliament and of the Council of 27 April & J. Pridmore (Eds.), Digitizing identities: Doing iden‐ 2016 on the protection of natural persons with tity in a networked world (pp. 242–260). Routledge. regard to the processing of personal data and on https://doi.org/10.4324/9781315756400‐13 the free movement of such data, and repealing Carammia, M., Iacus, S., & Wilkins, T. (2020). Forecast‐ Directive 95/46/EC (General Data Protection Regula‐ ing asylum applications in the European Union with tion). (2016). Official Journal of the European Union, machine learning and data at scale. ArXiv. L 119. Carling, J., & Collins, F. (2018). Aspiration, desire and Eurostat. (2021). Asylum and first‐time asylum applicants drivers of migration. Journal of Ethnic and Migration by citizenship, age and sex—monthly data (rounded) Studies, 44(6), 909–926. https://doi.org/10.1080/ [Data set]. https://appsso.eurostat.ec.europa.eu/ 1369183X.2017.1384134 nui/show.do?dataset=migr_asyappctzm&lang=en Castles, S., de Haas, H., & Miller, M. J. (2014). The Felkai Janssen, Z. (2020). Migration monitoring and age of migration: International population move‐ preparedness at the European Union level: The need Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 143 for multi‐stakeholder cooperation. Migration Policy McCusker, B., Heuser, S., & Midgley, A. (2018). Practice, 10(4), 45–47. https://publications.iom.int/ Groundswell: Preparing for internal climate migra‐ es/books/migration‐policy‐practice‐vol‐x‐number‐ tion. World Bank. https://openknowledge. 4‐september‐december‐2020 worldbank.org/handle/10986/29461 Food and Agriculture Organization of the UN. (2021). Robinson, C. (2018). Making migration knowable and Domestic price warnings. http://www.fao.org/giews/ governable: Benchmarking practices as technolo‐ food‐prices/en gies of global migration governance. International Franzosi, R. (1998). Narrative analysis‐or why (and how) Political Sociology, 12(4), 418–437. https://doi.org/ sociologists should be interested in narrative. Annual 10.1093/ips/oly020 Review of Sociology, 24, 517–554. https://doi.org/ Rulers, Elections, and Irregular Governance. (2021). 10.1146/annurev.soc.24.1.517 Rulers, elections, and irregular governance dataset Frontex. (2020). Migratory map. https://frontex.europa. [Data set]. https://oefdatascience.github.io/REIGN. eu/along‐eu‐borders/migratory‐map github.io/menu/reign_current.html Global Database of Events, Language, and Tone. (2021). Scholte, J. A., Verhaegen, S., & Tallberg, J. (2021). Elite The GDELT story. https://www.gdeltproject.org/ attitudes and the future of global governance. Inter‐ about.html national Affairs, 97(3), 861–886. Hernandez, K., & Roberts, T. (2020). Predictive analytics Searle, C., & van Vuuren, J. H. (2021). Modelling forced in humanitarian action: A preliminary mapping and migration: A framework for conflict‐induced forced analysis (K4D Emerging Issues Report 33). Institute migration modelling according to an agent‐based of Development Studies. approach. Computers, Environment and Urban Sys‐ Integrated Food Security Phase Classification. (2020). tems, 85, Article 101568. https://doi.org/10.1016/ The IPC population tracking tool. http://www.ipcinfo. j.compenvurbsys.2020.101568 org/ipc‐country‐analysis/population‐tracking‐ Singh, L., Wahedi, L., Wang, Y., Kirov, C., Wei, Y., Mar‐ tool/en tin, S., Donato, K., Liu, Y., & Kawintiranon, K. (2019). International Organization for Migration. (2020). Dis‐ Blending noisy social media signals with traditional placement tracking matrix. https://displacement. movement variables to predict forced migration. In iom.int A. Teredesai & V. Kumar (Eds.), KDD ’19: Proceedings Kjærum, A. (2020). Foresight: Using machine learning of the 25th ACM SIGKDD international conference on to forecast and understand forced displacement. knowledge discovery & data mining (pp. 1975–1983). Migration Policy Practice, 10(4), 26–30. https:// Association for Computing Machinery. publications.iom.int/es/books/migration‐policy‐ Sohst, R., & Tjaden, J. (2020). Forecasting migration: A practice‐vol‐x‐number‐4‐september‐december‐2020 policy guide to common approaches and models. Klabunde, A., & Willekens, F. (2016). Decision‐making in Migration Policy Practice, 10(4), 8–13. https:// agent‐based models of migration: State of the art publications.iom.int/es/books/migration‐policy‐ and challenges. European Journal of Population, 32, practice‐vol‐x‐number‐4‐september‐december‐2020 73–97. Sohst, R., Tjaden, J., de Valk, H., & Melde, S. (2020). The Lif, O. (2016). Migration in the internet age. The 21st future of migration to Europe: A systematic review of century: When refugees went online (Lund Stu‐ the literature on migration scenarios and forecasts. dent Papers). Lund University. https://lup.lub.lu.se/ International Organization for Migration. student‐papers/search/publication/8873467 Stamer, M. (2019). Country Risk—Which countries Massey, D. (1999). International migration at the dawn are the most vulnerable to the global slowdown? of the twenty‐first century: The role of the state. Euler Hermes. https://www.eulerhermes.com/en_ Population and Development Review, 25(2), 303–322. global/news‐insights/economic‐insights/Country‐ https://doi.org/10.1111/j.1728‐4457.1999.00303.x Risk‐Which‐countries‐are‐the‐most‐vulnerable‐to‐ OECD. (2020). International migration database. https:// a‐global‐slowdown.html stats.oecd.org/Index.aspx?DataSetCode=MIG Szczepanikova, A., & Van Criekinge, T. (2018). The future Papakyriakopoulos, O., Hegelich, S., Shahrezaye, M., & of migration in the European Union: Future scenarios Medina Serrano, J. C. (2018). Social media and micro‐ and tools to stimulate forward‐looking discussions. targeting: Political data processing and the conse‐ Publications Office of the European Union. https:// quences for Germany. Big Data & Society, 5(2). doi.org/10.2760/274425 https://doi.org/10.1177/2053951718811844 The World Bank. (2021). World development indica‐ Raleigh, C. (2011). The search for safety: The effects of tors. https://databank.worldbank.org/source/world‐ conflict, poverty and ecological influences on migra‐ development‐indicators tion in the developing world. Global Environmen‐ Thomson‐DeVeaux, A. (2020). Even without a pandemic, tal Change, 21(Suppl. 1), S82–S93. https://doi.org/ it’s hard to forecast a recession. FiveThirtyEight. 10.1016/j.gloenvcha.2011.08.008 https://fivethirtyeight.com/features/even‐without‐ Rigaud, K., de Sherbinin, A., Jones, B., Bergmann, a‐pandemic‐its‐hard‐to‐forecast‐a‐recession J., Clement, V., Ober, K., Schewe, J., Adamo, S., Triandafyllidou, A. (2020). Decentering the study of Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 144 migration governance: A radical view. Geopoli‐ UN. (2021). United Nations population division. https:// tics. Advance online publication. https://doi.org/ www.un.org/development/desa/pd 10.1080/14650045.2020.1839052 Wilson, T., & Rees, P. (2005). Recent developments in UN High Commissioner for Refugees. (2020). Refugee population projection methodology: A review. Pop‐ data finder. https://www.unhcr.org/refugee‐ ulation, Space and Place, 11, 337–360. https://doi. statistics/download/?url=pwN2 org/10.1002/psp.389 About the Authors Cristina Blasi Casagran is assistant professor in EU law at the Autonomous University of Barcelona, and she holds a PhD in Law at the European University Institute (Florence, 2015), in which she specialised on EU privacy and data protection law. She is currently coordinating the H2020 project IT Tools and Methods to Manage Migration Flows (ITFLOWS) and the Erasmus+ Jean Monnet Module EUCONAS. Colleen Boland is a postdoctoral researcher at the Autonomous University of Barcelona, where she works on the EU Horizon2020 funded ITFLOWS project investigating the management of migra‐ tion flows to the EU and co‐coordinates the project’s Gender Committee. She holds a PhD in Sociology and Anthropology from the Complutense University of Madrid. She is currently a 2021–2022 re:constitution fellow, a programme funded by the Forum Transregionale Studien and Democracy Reporting International. Elena Sánchez‐Montijano is research professor at the Center for Research and Teaching in Economics (CIDE, Mexico) and a researcher at the Autonomous University of Barcelona, where she collaborates on the EU Horizon2020 funded ITFLOWS project. She holds a PhD in Political and Social Science from the Pompeu Fabra University (Barcelona). Currently, she is co‐coordinating the project Towards an Integration Policy for Migrants in Mexico, funded by the Open Society Foundation and is a partner coordinator at the Atlantic Network 2.0 project, funded by the European Jean Monnet programme. Eva Vilà Sánchez is a graduate student at the Autonomous University of Barcelona. She is currently conducting the MA on European Integration and has completed an internship at the EU Horizon2020 funded ITFLOWS project. She holds an undergraduate degree in European Studies and French from Royal Holloway, University of London. Politics and Governance, 2021, Volume 9, Issue 4, Pages 133–145 145