Over the course of this research, from January 1st until May 3rd, 1090 migrants died in the Mediterranean according to the Missing Migrants Project, or 1092 according to the UNHCR (IOM 2017c, UNCHR 2017). The databases that go back into 1990s reveal that migration to and across the European continent is in fact not contemporary. However, today's numbers of migrants' fatalities are registered as record-breaking. Accordingly, the attention on migrant mortality has gained momentum, especially since 2016, which has been marked as the deadliest year in the Mediterranean. A plethora of actors record migrant fatalities (academic, NGO, non-state, state and international organizations), working with diverse definitions, in a variety of regions, using different methodologies and pursuing various goals with their count. For this paper, six primary counts (BorderDeaths, United against Racism, the Migrants' Files, Watch the Med, Missing Migrants Project and UNHCR) were analysed in depth, representing evidence-based and incident-based databases, as well as mixed methodologies.
With the aim to shed light on the production of the different databases, this paper analyzes the work of the six projects in terms of definitions, goals, motivations and representations and discusses the links between the databases. To this end, first traditional methodologies for counting are outlined and how these are implemented in the specific context of migration. Six aggregators are then presented by describing the organization, the goals of the count, the geographical area, who was counted, who counted and the methodology before comparing the different definitions of the counts. For each aggregator that published the data, the raw data is then analysed. This includes the IOM, United, Migrants' Files and Borderdeaths. Each section on these four contains visualizations and explanations of the death count and how it varies over time, the location data, main causes of death and a description of the sources. Since the data is so radically different, many means of visualization had to be used: unclustered data is reflected in word clouds, refined numerical data is displayed alluvial diagrams or tree charts. These methods have been chosen to extract the most meaningful trends from the raw data. All word clouds are drawn with R and the default English stopword list based on the refined data. Anything but letters have been stripped off and the text corpus had to be stemmed where indicated. All diagrams are created with Rawgraphs.io based on the refined data. If value-judgments had to be made while refining it is indicated in the text. Ultimately, definitions are examined to understand the commonalities and differences in the building blocks of counting. Then, incident-based approaches are compared with the evidence-based datasets, as well as the connotative incident-based recording with alternative connotative incident based recording. Finally, there is an analysis on the differences in numbers produced by state and non-state actors.
From the fields of conflict and crime, incident based recording is designed to capture distinct incidents with a clear date and location. In theory it is a means by which information can be derived from direct witnesses who have been in close proximity to the incident. It is useful as the information gathered from the reports have the potential to be documented and verified. In terms of implementation today, more often than not the incidents recorded are gathered from media outlets, NGO reports, hospital records, morgue files, official figures, etc before being collated into a database. The total in the count based on incident-based reporting is not an estimate but rather an attempt to reach a real number for each given instance by collecting as much testimony on the incident for the actuality of the data. This may later reveal trends. Evidence-based reporting on the other hand does not use published media as its source but instead relies on official sources and tangible objects to verify its numbers. This method of counting requires physical artefacts which represent whatever it is that is being counted. Other counts employ retrospective surveys with survivors after incidents have occurred in a methodology known as survey-based estimation. Survey estimations aim to understand perception of numbers based on those who were involved or affected in the incidents by distributing questionnaires to a mass and aggregating data from their answers.
The five aggregators that have chosen to make their raw data public cover different time periods from 1990 onward. Here is their time span, the minimum count of incidents and deaths and a relative expression of deaths per day. The count is minimum because missing migrants are excluded.
|First dated Record||Last dated Record||Time Span||Total Incidents||Total Deaths||Deaths per Day|
|Border Deaths||01/01/1991||16/07/2014||23 years, 6.5 months||3188||3188||0.37|
|Migrants' Files||01/01/2000||20/06/2016||15 years, 5 months||3193||4216||0.75|
|United Against Racism||01/01/1993||23/04/2015||22 years, 4 months||2985||22092||2.71|
|Missing Migrants||06/01/2014||26/03/2017||3 years, 2.7 months||567||4524||4.02|
Although the IOM claims to produce a minimum estimate of deaths in the Mediterranean (see Missing Migrants Project description), their toll is much higher than that of the other three aggregators. An average of 6 migrants per day lost their lives accordingly. Yet, it should also be taken into account that the IOM counted during the "deadliest" year in the Mediterranean (2016), during which the other counts had already been ended. From this data, there also appears to be a small-time window where all five aggregators overlap. From January 7th 2014 until June 15th all had their eyes on the Mediterranean. This coincidence shall be later used for an in-depth comparison of these counts.
Borderdeaths is an incidents-based dataset. Each row corresponds to a dead body and an autopsy. Here are the chief causes of death:
The Deaths at the Borders Database (BorderDeaths) is an evidence-based database covering migrant deaths in the region of the EU external maritime borders. It accounts for migrant mortality in the Mediterranean from 1990 to 2013, starting when border deaths at the southern EU border emerged. The database collected evidence from death management systems of Spain, Gibraltar, Italy, Greece and Malta. As the first database that used official documentation, BorderDeaths brought together individualized information about each person who died attempting to reach the Southern shores of the EU without authorization and whose body has been found in or brought to one of the Southern EU member states. The database recorded 3188 people found dead in the accounted time period. BorderDeaths has been a five-year research project of the Vrije Universiteit Amsterdam, funded by NWO, the Netherlands Organisation for Scientific Research (BorderDeaths 2017a; Last, Spijkerboer and Ulusoy 2016: 5f).
BorderDeaths aims to serve as a complementary resource to existing incidence based databases that are based on media reports or interviews. It was less interested in finding the exact number of border deaths, but rather collected the data to share it enabling other people to work with it. Therefore, the information has been collected in a way that made it accessible, while at the same time containing as much raw data as possible. Although not a direct aim of the project, BorderDeaths contributed to give a more complete and more nuanced account of the situation at the Southern EU borders (Interview Tamara Last).
BorderDeaths collected data in the Southern EU member states, where irregular migrants have arrived or where bodies might be brought to. This precisely includes Spain (along the southern coast from Portugal to Valencia, at the coasts of the Baleares and the Canary Islands and in the Ceuta and Melilla), Gibraltar, Italy (Sicily coast and inland, southern coasts of Calabria and Sardinia, southern and eastern coast of Puglia, the coast of Foggia, the international port of Ancona, the major international port of Naples), Greece (the Cyclades, and eastern coasts of Evoia, the Dodecanese, Crete, the North Aegean islands, and the Greek side of the land borders with Albania and Turkey (Evros)) and Malta (BorderDeaths 2017b).
The project had been constructed with a European, or rather EU, bias, that has not been enlarged to other Mediterranean countries for mainly two reasons. First, BorderDeaths applied for research funding for a project with this scope and could not include other regions for financial reasons. Second, it started the research shortly after the Arab Spring and did not include non-EU countries for safety reasons. Following the methodology, the data collection has to be systemic and due to the political circumstances, large areas, for example in Libya, would have needed to be cut out. Yet, if done again, Turkey, Morocco, Mauritania and Senegal could be included (Interview Tamara Last).
The data base accounts for border deaths between 1990 and 2013 without the objective by the researchers to continue until the present date. Nevertheless, it would be easy to continue collecting the data. The death registries are all digitalized and centralized today. Since the method is already established, it would only be necessary to write a search program that searches for the criteria applied in the research to collect the data (Interview Tamara Last). Every person falling into the definition of border death was counted. Different degrees of certainty of a person being a border death were established. For this project, border refers to the physical borders of the EU, which included the high seas between southern EU states and North and West Africa (Last et al 2017: 710). It includes “[p]eople who have died attempting to migrate irregularly to Europe by crossing the southern external borders of the EU without authorization, whose bodies were found on or brought to the territories of Spain, Gibraltar, Italy, Malta or Greece” (BorderDeaths 2017b). This definition excludes people who died on the European continent, for example in detention, due to the limitations of the data collection method.
Border deaths were classed into three different categories: confirmed, likely and possible border death, depending on different levels of certainty that the person was a border death (BorderDeaths 2017b). The level of certainty was also established according to the amount of information available – the more information existed on a person, the more definite one could say, if he or she was a border death. The final count and the visualization include all the categories without distinctions.
Generally, a state does not establish whether a person is a border death, except for rare cases where it is marked in the death certificate. Therefore, criteria have to be set up to decide when a person counts as a border death (Interview Tamara Last). Counting border deaths based on death certificates has the advantages that they are easy to access, reliable in terms of avoiding double-counts and provide a useful summary of the identity and the death of the deceased (BorderDeaths 2017b).
The data BorderDeaths uses was collected in the death management systems of the geographical regions defined above. Since BorderDeaths was an academic project, the deaths were counted by twelve researchers and Tamara Last from 563 civil registries and additional offices in the period of April 2014 until February 2015 (BorderDeaths 2017b). A common methodology to search registries and extract data of border deaths was followed to collect data in a uniform way across the countries. The common methodology consisted of: a set of tools for collecting and recording data, a working definition of border death, a step-by-step manual on collecting the data in the archives and using the instruments (BorderDeaths 2017b, Last et al 2017: 699). The step-by-step manual clarified the steps when searching the books of the death registry, clarifying that researchers had to check every death certificate of the given time period. It also provided guidance on how to identify a border death, what details to look out for that hint at a border death. Identifying a border death was a process of exclusion. Persons with European nationalities, or with the place of birth or last residence in Europe were directly excluded. Death certificates of persons to whom these criteria did not apply were looked at more thoroughly (especially place of death, age and description of the deceased) and when nothing in the death certificate excluded the person from being a possible border death, it was recorded (Interview Tamara Last). Due to the logistics of data collections and some problems in the field, slight deviations from the common methodology were necessary in two cases.
After the data collection, all collected cases were reviewed and the degree of certainty of a case being a border death was established (BorderDeaths 2017b). The database contains of 43 categories. In many cases, not all of the categories could be filled in with the information of the death certificates because the data was not available or did not exist (Interview Tamara Last).
Naturally, the evidence-based dataset has the fewest total deaths recorded. A trend of increasing deaths with a controlled time frame can be observed: since 2010 the evidence-based count is moving closer to one dead migrant per day.
|Interval||Total Death Count||Deaths per Day|
|01.01.1990 - 31.12.1999||649||0.18|
|01.01.2000 - 31.12.2009||1708||0.47|
|01.01.2010 - 16.06.2014||831||0.65|
As every row in the dataset corresponds to one medical examination, the causes of death are accordingly accurate and elaborate. "Other" are therefore all detailed descriptions that could hardly be summarized.
Borderdeaths provides us different location descriptors. Where a migrant died may be different from where the case was registered and where she was ultimately buried. Let us look at the link between these three categories. In other words, this alluvial chart answer the question: If a migrant dies in place x is she also registered there and buried there?
Most deaths are counted in the three countries Italy, Greece and Spain with similar proportions. A small number of deaths has also been recorded in Malta and Gibraltar. BorderDeaths is the only dataset with comprehensive location descriptors, from country of death, village of death, place of filing the case and place of burial. These different locations allow us to compare flows between them. In other words, if a migrant dies in Sicily, will the death also be recorded in Sicily and the body buried there as well? Here we display the twelve most common places of death (village/town), where these cases have been filed and where the bodies have been buried. The burial site has been generalized by country with the help of OpenRefine's cluster function.
The data collection in Alexandroupoli (GR), Lampedusa (I) and Agrigento (I) seems problematic. Most cases that occurred in the two towns lack the location of registration. Ceuta (ES), Tarifa (ES), Lesvos (GR) and Brindisi (I) fare better: most cases have been filed in the same place and deaths occurring in Ceuta and Valetta get largely buried there as well.
The Border Death dataset is unique by including cases that have been verified by coroners. The sourcing can therefore be assessed through the "Death Certificate" column. Combined with the cause of death and the identification status, it allows us to illustrate the link: Are there causes of death where the person is rarely identified and if a person is not identified, can a death certificate still be given?
Surprisingly, over a third of all drownings are identified. More counter-intuitively, the majority of people who die of dehydration and suffocation are not identified. Moreover, in the majority of unidentified cases death certificates are still issued.
Founded in 1993, UNITED for Intercultural Action has the most extensive incident-based recording of deaths of "refugees, asylum seekers and undocumented migrants who have died through their attempts to enter Europe". As a European Network against nationalism, racism, fascism; aligning themselves in support of migrants, refugees and minorities, UNITED's goal regarding the undertaking of the project of counting the dead is to "monitor the human impact of the policies building so called Fortress Europe" (UNITED 2017). Geographically speaking United’s count includes deaths that occurred on European soil, but also of anyone who died on the journey to Europe (not just the physical border region). The description of UNITED's count as the most comprehensive is reflected in the definition of who is counted: any refugee, asylum seeker and undocumented migrant who fits into at least one of the following categories makes the count
The collective failure of an organisation to provide an appropriate and professional service to people because of their colour, culture, or ethnic origin. It can be seen or detected in process, attitudes and behaviour which amount in discrimination thought unwitting prejudice, ignorance, thoughtlessness and racist stereotyping which disadvantage minority ethnic people.
As of June 2015, the count for their 22 year-long project concludes that 22,092 people have died during migration. Following the criteria listed above, their database is then compiled by way of information gathering from other organisations including, government reports and testimonies. UNITED does not collect the data on the ground, but relies on other counters for its numbers. They also conduct "constant media monitoring" (UNITED 2017). Since there is no methodology available, United's methodology based on the dataset was analysed.
For its long history of collecting data, United's numbers allow for trends to be drawn from the early 1990's. Three intervals can be identified:
|Interval||Total Death Count||Deaths per Incident||Deaths per Day|
|01.01.1993 - 31.12.1999||2003||3.36||0.78|
|01.01.2000 - 31.12.2009||12349||6.53||3.38|
|01.01.2010 - 23.04.2015||7646||15.42||3.89|
According to the counts of United, significantly more people died within the same time frame after 2000 compared to other counts, as each incident brought double the number of deceased. After 2010, the amount of deceased with time controlled stayed roughly the same, whereas the deadliness of each incident has doubled once more. United posed a significant problem when determining causes of death and location: both descriptions are wrapped into an incoherent "Cause of Death" column. A first, unstructured look at a word cloud generated from this will help.
Words such as "capsized", "trying", "stowaway", "river", "deportation", etc. are describing circumstances around the death rather than actual causes - counter the column title. Those more technical causes are less frequently mentioned: "Hypothermia", "starvation" and "suffocation" are such examples.
Circumstances were therefore grouped together, which might often hint at a cause of death, together with some chief causes of death. With OpenRefine, 10 chief categories of death causes could be identified from the description column. Whenever the description column hinted at specific circumstances, these were taken as a category. For example, the presence of the word "himself" leads in all cases to a description of a suicide - and therefore the column cause of death is filled with "suicide". "Stowaway" is by no means a cause of death but was often the only information available in the description column. Broad categories like "firerelated" are of course technically no death causes, but represent the most meaningful categorization that can be extracted from this messy column. The table summarizes the ten main causes of death/circumstances of death. Similarly, the location column has also been filtered for country names as place markers. This yields the this table where detailed location descriptions (such as Lesvos, Lampedusa, Granada) have been ignored.
In the source column the United data fares much better: only six incidents have no source attributed at all. Nonetheless, the sourcing is only noted with abbreviations, without URLs, or article titles and publication dates. A word cloud will yield a pattern.
Three source categories stand out here: for once, AFP (Agence France Presse), DPA (Deutsche Presse Agentur), the German left-wing newspaper taz (die tageszeitung) and BBC are wire services or established media organizations. A second category of sources is that of more field-based activist groups. Notable among them are Mugak, a Basque activist group from San Sebastian helping migrants, Picum an activist network for undocumented migrants and APDHA, the association for human rights of Andalusia. The last and perhaps lion's share is attributed to less known watchdogs and think tanks specialized in migration. The MNS (Migration News Sheet), ZAG - an anti-racist magazine from Berlin, statewatch, the UK-based EU critical news-aggregator, or the IRR (Institute of Race Relations) are such notable examples. The amount and impact of these activist groups on the ground and the news sites dedicated to migration reveals: a main cause of anti-racist activism in the 1990s has revolved around undocumented migrants and border deaths. Compared with the struggle against racism in other parts of the world, this was not obvious. Moreover, the frequent occurrence of sources from Spain, Germany and the UK may explain why these three countries appear as the main theaters of migration-related deaths in the table above.
Perhaps the most outstanding feature of this data set is the incoherence. The date column required cleaning (turning "dates" such as "in 2009" into an actual date). The "name" column only contains names in around 860 cases. Where the deceased has not been identified this column attempts to give information about the demographics. An estimation of age, sex and origin are given here in an unstructured manner - making it nearly impossible to process this information.
Between 2013 and 2016, the Migrants' Files was a conglomerate of journalists, statisticians and software developers who aimed to aggregate deaths of migrants into a concrete number. The data collection however, has a wider timeline by counting the deaths of migrants which occurred between 2000 and 2015. The project fell under the sixth and seventh research framework of the European Union's funded body of research programs (Kayser-Bril 2017). The EU research program's primary objective at the time was to produce in-house content which could lead to political mechanisms to aid in discontinuing migration to Europe. The Migrants' Files was funded through the European Council's own initiative to better understand the total cost of maintaining "Fortress Europe" across all member states and how to reduce the excess expenditures (The Migrants' Files 2016). The primary costs that the council sought to lower included deportation, detention, inter-state coordination, security and military costs on top of the cost of processing the dead themselves (recovering sunken ships, death certification, burial) (Kayser-Bril 2017) (The Migrants' Files 2013). Analysis of various incidents were supported by other EU organs including the European Space Agency (The Migrants' Files 2016). The deduction from the Migrants' Files understood that the number of deaths directly correlated to the rate of migration. Meaning that the project concluded that less deaths entailed a decrease in overall migration to the Union. The deaths database was then correlated with a "money trails" count to understand the trends of migration rate vs. impact on the border regime's budget.
The Schengen Zone, plus the United Kingdom was considered to be "Europe." This was included in the methodology as per the EC's stipulations on the funding (Kayser-Bril 2017). Therefore, only migrants who died in a Schengen state or in attempt to reach a Schengen state on route were counted. The boundaries for "on route" are also explicitly defined and are all in reference to a direct entrance to a Schengen member state. Those include,
The Migrants' Files had a clear framework of which migrant deaths would be counted prior to the start of the project. A migrant was considered to be anyone who did not possess EU citizenship, who was attempting to reach Europe- they had to have been found on one of the routes listed above- or a person attempting to stay in Europe- the person either died attempting to declare asylum in Europe or recently was deported died as a result of the deportation process. The other set of migrants who were counted were those with the intent of going to Europe. This was defined as a person who died just outside of the legal Schengen border as an attempt to cross. Direct deaths from state officials (shot) or indirect deaths (exhaustion) were counted.
The total death count was generated through an incident-based causality methodology by collecting data from a variety of sources and cross referencing them. The main sources for the Migrants' Files data were from Puls, The Puls Project is hosted by the University of Helsinki, Finland. It is a collaboration based consortium which conducts web-scale surveillance of News Media in the arenas of epidemic surveillance, security-related and business news. Puls does not have an active death count of migrants, however it does have a subheading under "security news" deemed "the migrant crisis" which is a composite of media reports on migration worldwide. United for Intercultural Action, Fortress Europe, IOM and UNHCR (Kayser-Bril 2017) (The Migrants' Files 2013). A consistent methodology with OSINT was employed to collect data. OSINT collects data from publically available sources including media publications, government reports and grey literature. OpenRefine, an opensourced analysis algorithm was then applied for data-cleaning and fact checking. Duplicates were manually removed by those working on the project. Migrants' Files also conducted sporadic interviews across the Schengen Zone in formal camps and urban spaces with migrants. These testimonies were included in the dataset. Early in the process (2013-2014), students from the University of Bologna were doing the fact verification and manual duplication cross-checking, supervised by Dr. Carlo Gubitosa (Kayser-Bril 2017) (The Migrants' Files 2013) (The Migrants' Files 2016). Yet, the majority of the counters of the dead were officials from the media, various member-state governments and International NGOs, such as the Red Cross. Those aggregating, fact checking and developing the data included EU journalists and statisticians. Most of whom had not counted the dead prior to the commencement of the project (Kayser-Bril 2017).
The Migrants' files' dataset contains two separate counts: dead and missing numbers have been recorded separately. The difference is striking
With the column death only taken as base then the picture looks different. Especially the low number of deaths before 2010 may indicate a problem of verification. Migrants’ Files has started actively aggregating data in 2013 - all incidents before had been taken from other aggregators. This reflects in the numbers above and even more so in the scatter distribution of time versus death count where clusters of high density are highlighted in blue.
|Interval||Dead and Missing||Deaths per Incident||Deaths per Day|
|01.01.2000 - 31.12.2009||17949||7.70||4.92|
|01.01.2010 - 20.06.2016||16910||19.70||8.48|
|Interval||Dead||Deaths per Incident||Deaths per Day|
|01.01.2000 - 31.12.2009||108||0.05||0.03|
|01.01.2010 - 20.06.2016||4106||4.79||2.05|
Migrants' Files had been collecting large numbers of incidents since 2000 - as the darker blue indicates. Only since the project took off in 2013, however, incidents with high casualty number have been recorded. In the causes of death, the category "authorities-related death" seems to echo the United dataset - where Migrants' Files' journalists have taken the lion share of the data from.
These incidents occurred largely in and around the Mediterranean and come with an accurate name of a village/town, the Frontex migration route and, most importantly, a GPS location. Based on this, the following map could be plotted with Khartis.
The map indicates, that on continental Europe most recorded incidents involve dead migrants. Migrants go missing chiefly in the Mediterranean. Sources have been given with names of outlets and organisation and URLs in a second column, where half of the incidents have URLs, some even with links to the Internet Archive. However, producing a word cloud of all source names produces an unsettling result.
The pro-bono activist oriented project Watch the Med (WTM) originally stemmed from the Boats4People campaign (2012) in the central Mediterranean. The on-going project was founded out of outrage following the famous 2011 Lampusa boat sinking which resulted in at least fifty deaths and was a quintessential example of the European Border Regime's massive human rights violation campaign for those stranded at sea (Stierl 2017). The project was founded to support migrants attempting to illegally cross into the EU while also defending their inalienable rights by pressuring authorities to respect their obligations under the International Law of the Sea. WTM itself has two core features in relation to water migration to the European continent: the online mapping tool of deaths and violations of migrants' rights and the Alarm Phone in which migrants directly call in for support during situations of distress. Online mapping involves reconstructing past events which involved migrants' rights violations and at times, deaths. The Alarm Phone acts to inform migrants of their rights and security and connect them with EU coast guards during times of deep distress in order to save lives (Stierl 2017) (Watch The Med 2012). Also, WTM supports and regularly participates in ongoing campaigns by relatives of the dead and missing at sea and in legal cases involving the violation of rights of migrants. A death count is a by-product of WTM's violations records rather than an intentional development of statistics to be released to the public (Stierl 2017).
The online mapping platform provides an understanding of the maritime environment in which migrants cross. The scope of the map is directly dependent upon wherever boats contact WTM through the Alarm Phone. The geographical span insofar has split the Mediterranean into three sections for which operators on the Alarm Phone answer to:
According to WTM's work, a migrant is any person crossing international borders illegally. The definition is not contentious because the project strictly records violations and deaths of those who died at sea or those who die during a rescue or just after through a direct death in relation to a violation by the authorities (Stierl 2017). For example, if a boat of migrants calls the Alarm Phone, WTM informs the Italian Coast Guard. During the mission to make contact with the migrant vessel in distress, the Italian Coast Guard chooses to push the boat out of Italian waters rather than conduct a rescue, it is considered to be illegal abandonment in international waters. If that boat sinks and 200 people die, or if the boat eventually makes it to the shore with no deaths, the event is recorded on the map as a human rights violation. If migrants are rescued and shot by the authorities during the rescue, it is also included as a rapport or on the map (Stierl 2017). The sources of the majority of the violations stem from anonymous Alarm Phone recordings and interviews conducted directly by WTM. There are however other sources for reconstructing a past event including direct contact with state authorities, alternative NGOs and activist groups conducting similar work (Stierl 2017) (Watch The Med 2012).
The incident-based methodology of WTM is exclusive particularly because the organization approaches the recreation of events through counter-mapping (Stierl 2017). This technique is not aiming for an exact (or even general) number of the total migrants who have died at sea. WTM acts in the opposite of other counts by tracking the actions of state actors rather than routes of migrants in order to pinpoint an accurately coordinated location of where the law of the sea was broken and to recreate the event in which human rights violations occurred. More precise procedures include that of tracking SAT phone signals moving irregularly across the Mediterranean (compared to the typical routes of the coast guard) to track the movement of a migrant vessel in comparison to a state one (Stierl 2017). Qualitative interviews in a postevent scenario are generally quite challenging to conduct because migrants will dispose of their SAT phones upon rescue out of fear of being identified as a smuggler (Stierl 2017). They also continue to move away from the landing destination quite quickly, making any post-Mediterranean contact rare. WTM works directly with coast guards to understand incidents further, however there is generally limited communication when it comes to human rights violations inflicted by said authorities themselves (Stierl 2017).
The Missing Migrants Project is a joint initiative of the IOM's Global Migration Data Analysis Centre and Media and Communications Division. The research began following behind the October 2013 shipwrecks near Lampedusa in which 368 migrants lost their lives. The project started its work in 2014 and counts migrant deaths on a current day basis (IOM 2017a). Apart from the database and the website, the Missing Migrants Project also publishes infographics, policy briefs and an annual report on migrant deaths, holds seminars and builds networks for better data collection and to advocate for better management of the deceased, improved identification and family support (GMDAC 2017).
The project pursues various goals with their count. One aim is to acknowledge the fact of the deaths, since a lack of data on migrant mortality contributes to a lack of public awareness, a lack of public concern to the safety of the migrants and a process of dehumanization of migrants. With the availability of data, it is easier to assist migrants and prevent their loss of life. Collecting data on migrant mortality can help to understand the causes of death and their possible link to migration control policies. Furthermore, a death count is a first step in improving to identify those who die. Identify the deceased acknowledges and respects their death while providing closure for the family who might otherwise continue live in uncertainty (IOM 2017b).
The Missing Migrants Project accounts for migrant fatalities worldwide, dividing the count into twelve regions (IOM 2017c). One part of the project focuses entirely on the Mediterranean, splitting up the region into the Western, Central and Eastern Mediterranean, accounting not only for the deaths, but also for the arrivals in Europe (IOM 2017d). The IOM defines a migrant as "any person who is moving or has moved across an international border or within a State away from his/her habitual place of residence, regardless of (1) the person's legal status; (2) whether the movement is voluntary or involuntary; (3) what the causes for the movement are; or (4) what the length of the stay is." (IOM 2017b).
A person accounted for in the database is a migrant who died or went missing at external state borders or during the migration process towards an international destination. Persons dying related to migration, but without crossing external borders, are not counted. This excludes for example people who died in detention centers, during deportation or after being expulsed, as well as migrants who died in their "new home", i.e. in refugee camps or refugee shelters. This definition has been chosen because deaths at physical borders of persons moving can be more clearly defined. They inform also on the danger of migration routes. The Missing Migrants Project relies on a variety of sources, depending on the region and the availability of data. This is also the case for the Mediterranean region. Data is obtained from national authorities (i.e. Coast Guards and Medical Examiners) who share their information with IOM field missions. Other data is obtained by interviews with survivors of shipwrecks in the landing points in Italy and Greece. Furthermore, media reports are used. The data is coordinated between IOM and UNHCR to ensure consistency (IOM 2017b).
The Missing Migrants Project aggregates data from different sources to account for migrant mobility on the globe and in the Mediterranean. Since sources vary across the different regions, this section focuses on the methodology applied for the Mediterranean region. People working on the ground, like forensic specialists, humanitarian workers or coroners, send their data to the project who makes it publicly available in a somewhat sensitized form (Interview Julia Black). To get access to media reports, the project uses google alerts and a social media crawler for twitter and facebook.
The received data is entered into a database consisting of 23 columns, accounting for details of the incident, such as location, number of people, age, origin, what happened, cause of death, survivors or the GPS location. Columns have been added as the project further developed. The data on the incidents is challenging to collect, due to the irregular nature of journeys where deaths are likely to occur. Even with an incident based database, the data on the gender or the origin of the deceased therefore remains very limited and all data rather remains an estimate than a complete figure. The threshold of verification for an incident depends on the source. If the source is IOM staff, the UNHCR or other organizations in the form of an official report then the incident is considered as verified (Interview Julia Black, IOM 2017b).
As the IOM dataset includes incidents from all over the world, a few assumptions had to be made before processing: The column region of incident has been filtered so that only Europe, Mediterranean, Middle East and North Africa are left. The category Middle East only contains two incidents that are definitely within our scope as the column "migrant's route" indicates that migrants were on their way to Europe. In the place-category "Europe" most incidents revolve around Calais and the Balkan Route. The short time span allows to draw only limited trends. Yet, as the table indicates, the trend is towards less deadly incidents and less deaths per day.
|Interval||Total Dead Count||Deaths per Incident||Deaths per Day|
|19/01/2014 - 31/12/2015||2967||9.45||4.18|
|01/01/2016 - 26/03/2017||1758||5.60||3.86|
With openrefine the 75 initial causes of death have been generalized to 47. Out of these, the following top 10 causes of death explain 93% of all incidents. The migration route has been linked to the cause of death, as per the dead count per category.
As the alluvial chart illustrates, two thirds of the deceased in the central Mediterranean die of drowning or asphyxiation. The Eastern Mediterranean route appears to be better monitored, as most deaths are attributed to a specific cause and less die of an "unknown" reason. Naturally, the route from Calais to the UK claims most of its deaths in train and vehicle accidents.
These sources account for 60% of all incidents, the remainder are other sources that would be too many to enumerate. This graph indicates, most of these incidents are sourced by IOM-in-house reports and the UNHCR (in blue). Press reports (in yellow) take another share, whereas few reports come from the NGOs that had been so important for the sourcing of United (in red).
The United Nations High Commissioner for Refugees (UNHCR) officially began counting in November 2013. The numbers that are published daily on the online data portal, "Mediterranean Situation" are however interestingly enough, the most cited and relied upon for accuracy. The only count that may compete in attention is IOM, but the two work together regularly to compare and contrast counts. The UNHCR plays a monitoring role in all of the states being affected by migration, including in Turkey, the Balkans, North Africa and across the European Union (Mijovic 2017) (Toskovic 2017) (Veide 2017). The primary goal of the agency is to support state apparatuses in handling the migrant situation through direct assistance (Mijovic 2017). The UNHCR itself however does not make direct decisions in procedures when it comes to migrants- this is strictly state based (Toskovic 2017) (UNHCR 2013). The Mediterranean Situation aims to aggregate all of the efforts across the various states to better understand the crisis at hand. Arrivals are the UNHCR's primary concern, as it will directly impact their camp assistance efforts. Recordings also span from tracking financial response plans to recording the nationalities of migrants and refugees and those deceased (UNHCR 2016).
Although the UN organ operates across all states involved in the migration crisis, the count is in accordance to three main routes in the Mediterranean Sea which is described in accordance to the arrival state:
The UNHCR only records water-based direct and indirect causes of death (UNHCR 2017). This includes, drowning, exhaustion, hypothermia, starvation and/or dehydration. Deaths as a result of state interference are not recorded on the Mediterranean Situation (UNHCR 2017). Land based deaths do not count, even if they are connected to the sea voyage (Toskovic 2017). This count is fascinating because it conducts separate counts of deaths for "refugees" and "migrants," as well as a total count for the two combined. The UNHCR is explicit about these definitions:
Refugee: persons fleeing armed conflict or persecution and unable to return to home state/state of residence. Nationalities considered as refugees in the count include The Syrian Arab Republic, The Republic of Iraq, The Islamic Republic of Afghanistan and State of Eritrea. (UNHCR 2013) (UNHCR 2017)
Migrant: persons who choose to cross international borders illegally in order to improve their lives by finding work, education, family reunion, etc; they are not fleeing conflict or persecution and have the ability to return with protection from their home government. The top migrant nationalities recorded in the count include The Federal Republic of Nigeria, The Republic of Guinea and "Others." (UNHCR 2013) (UNHCR 2017).
The UNHCR is not open to the public regarding the details of its methodology behind the aggregation of the Mediterranean Situation death count. This is primarily because the sources of the data are connected to states and therefore, politics. There are a few details behind the project which are interesting to understanding where the numbers come from and how they are interpreted by the wider public. All figures are considered to be estimates and are compiled from surveys conducted in UNHCR refugee camps which are commissioned by UNHCR staff or government officials, Coast Guard statistics, Navy Vessels, Ministries of Interior and national police authorities. (UNHCR 2016) The agency itself does not conduct any count itself when corpses are discovered on the premises it operates on (Veide 2017). These situations are transferred directly to the appropriate state authorities who then record the death in their own manner (Mijovic 2017) (Veide 2017). The UNHCR is explicit in disclaiming reliability when it comes to the safety and well-being of migrants and refugees in this scenario of mass movement (UNHCR 2016). The UNHCR also cites "New Media" and "Civil Society" as other important sources of information, however these categories remain undefined. Sources for nationality and demographic data is simply stated as being "based on latest available information that may be partial and may change" (UNHCR 2013) (Toskovic 2017).
The starting place for each of the aggregators is undoubtedly the most precarious point of all: who is to be counted? Although there are overlaps in definitions, each aggregator is counting this general set of human deaths on very different terms. For the evidence-based count BorderDeaths from the University of Amsterdam a death is taken into the database of "people who died while attempting to reach southern EU countries from the Balkans, the Middle East, and North & West Africa, and whose bodies were found in or brought to Europe." Migrants' Files tries to count "the number of migrants who die while seeking refuge in Europe," with a geographical point of death changing the status of "intent to stay vs. seeking refuge." The activist network United against Racism "lists those refugees, asylum seekers and undocumented migrants who have died through their attempts to enter Europe, and as a result of European immigration policies including deportation procedures, detention conditions and the inadequacies of asylum application processes" and is thereby perhaps the most extensive definition. The IOM counts the deaths of "any person who is moving or has moved across an international border or within a State away from his/her habitual place of residence, regardless of (1) the person's legal status; (2) whether the movement is voluntary or involuntary; (3) what the causes for the movement are; or (4) what the length of the stay is." The UNHCR separates "refugees" from "migrants" with strict definitions, but determining such political labels after death is done through nationality indicators.
From these definitions, certain expectations emerge: BorderDeaths' count appears closely linked to the body that has to be on European territory - constraining the number of counted cases in this evidence-based dataset severely. The Migrants' Files link their count to the condition of "seeking refuge" which appears limited at first glance, but must not be confused with the severely restrictive legal term "asylum seeker." The Migrants' Files' open notion along the lines of "seeking shelter" might yield a large count. United similarly claims to include all forms of migrants and places special emphasis on authorities, whereas the IOM simply assumes movement across borders to be a determining factor. These latter definitions are more extensive and reflect the overall sum of dead migrants.
The aggregators considered count deaths in and around Europe. "Europe" however, is a contentious point as each aggregator geographically defines "Europe" differently according to capacity, political interests and goals in counting. Unlike a conflict which is contained to a set of borders and is therefore mostly traceable, the migrant crisis surrounding the European Continent is viewed through many different lenses- this impacts the final numbers being produced, making data comparison, confrontational. The area to be counted is therefore solely dependent on the aggregator. Most databases require a migrant's connection to the European continent however for United, deaths considered are much more global including those that have occurred in Africa or the Middle East after departing Europe forcibly. This differs greatly from the Migrants' Files which is narrowed to only count the Schengen Zone and deaths on the borders or BorderDeaths which only counts deaths occurring on the EU's Southern Mediterranean States.
Definition is further complicated through the aggregator's choice to count deaths by context, in this case context being land, water-based deaths or both. Water only counts are difficult to gage for accuracy because two thirds of recovered corpses are unidentifiable even to basic markers, such as gender or skin color. Aggregators have also stated that their counts represent the most "minimal" estimation of total deaths at sea because there is a staggering number of migrants that are never recovered after vessels sink. These ghost deaths make the aim for a "real" number quite idealist. This is compared to land-based deaths which are still difficult to track. Both question aggregators' definition of "verification" and how much information and sources are considered to be enough.The differentiating numbers between each count are largely dependent on the starting point of definitions. Final numbers are further impacted by general methodological choices by aggregators, on top of their connotations.
Whether a database is based on an incident or an evidence-based count, relates to another discussion around counting - counting things compared to people (Martin and Lynch 2009). As Martin and Lynch describe, counting is not a simple form of enumeration, but rather contains of two determining steps. It is necessary to decide what to count and what counts as something, therefore determining membership (Martin and Lynch 2009: 246). For evidence it is deaths related to migration for the "what" and membership relates to the who is counted as a death related to migration. For incident-based reporting, the what becomes more objectoriented as events that included migrant deaths are what is actually being counted over individual human deaths. The events that determine membership are what become differentiated through definitional choice of the incident aggregator prior to the start of the count. For example, counting events with deaths in the Schengen Zone vs not counting events with migrant deaths in the Sahara. Counting therefore puts things and people into categories (counting as) and precedes the actual process of enumeration.
This categorization becomes visible when looking at the databases and the way they are set up. Counting as is hereby determined by the definition of a migrant death, the area that is being taken into account and the methodology used. As discussed before, BorderDeaths and UNHCR only count water-based deaths, IOM, due to its global focus, includes any person dying when crossing external borders, while UNITED as well as the Migrants' Files account more generally for deaths related to migration, also including persons deceased in detention centers, refugee camps or victims of homicide as well as suicides. In terms of geographical area, the areas included and therefore also the counts vary for almost every database, depending on the definition of Europe or the Mediterranean for this purpose.
Ultimately, the methodology has a large impact on the count, relating to whether a database is incident or evidence-based. BorderDeaths, the only evidence-based database who used death certificates to count border deaths, had a significantly lower count than the other databases. When asking Tamara Last about this, she emphasized:
"Some activists got very angry with us. We were accused of working for the state interest in bringing the numbers down. But that's not the point. I mean, you can see it in a different way that our database revealed a lot more problems than the news, a more complete picture, although of course it's also not a complete picture. But the bigger numbers don't tell you as much. […] Actually, I would say my response to when we got that kind of negative press for our database, was: Stop only looking at the numbers. We're dealing with people and I think it doesn't really matter if the numbers are less, we learn something new and that's the point of collecting data." (Interview Tamara Last).
As Tamara Last underlines, their primary interest in making an evidence-based database was not to have an exact number of border deaths, but rather to create nuance by finding out more about who the people are that drown in the Mediterranean, where they come from, how old they are, etc.
In contrast, incident based databases and also databases applying mixed methodologies aim at having a correct count, therefore leaving the individual and the individual's story more out of sight. For example, the Missing Migrants Project counts incidents, but with the objective to get close to the real dead count. Whereas some incident-based databases (especially UNHCR) do not underline the weaknesses of their data collection methods, BorderDeaths gives a precise account of what they included or left out (for example regarding the geographical area). This approach clarifies what the data can show, instead of trying to account for a total number. Yet, this also relates to BorderDeaths' goals and their academic nature.
As incidence-based databases also rely on media reports, there is a risk of overcounting, although this is improbable and counts are more likely to be underestimates. For example, an incident might be reported with 20 persons missing, and a couple of weeks later two bodies might be washed up at the coast. They could be two out of the 20 missing, but one cannot be sure about this. Should these bodies therefore be added to the count or counted as two out of the 20 missing? These double counts are incredibly unlikely to happen with evidence-based databases, since only one death certificate is attributed to one person.
Although there is a high chance for evidence-based databases to be below the real number of migrant fatalities, due to bodies that have never been found, their databases provide much more detailed information about the deceased and enable families to get certainty about members that went missing. Additionally, evidence-based databases can also be related to incidents: BorderDeath grouped persons into incidents, if there was information on shipwrecks or if several bodies washed up together (Interview Tamara Last).
As the Missing Migrants Project underlines, they encourage identification of deceased migrants as a moral imperative, which is what Borderdeath does with its methodology. Consequently, the databases can actually be used in a complementary manner.
|Region||Town/Village||GPS||Frontext Route||Reported||Found||Died||No. Missing||No. Dead||No. Survivors||Country of Origin||Age||Sex||Cause of Death||Intention of going to Europe||Reliability/ Verification||Source||URL||Reviewed||Object ID|
A scale of verification is a decision that each aggregator decided upon prior to counting. How much information was required for each row in the datum to be made and which sources did it consider to be verified? This question relates to who is actually doing the counting of the deaths and for what purpose. These aggregators presented are not physically on the ground counting corpses that they encounter. Therefore, the data must start from someone on the ground and then be passed through at least one other set of hands prior to the aggregator including it in their individual database. Moreover, who is doing the actual aggregating? The UNHCR utilizes state numbers and compresses them into their own statistics. It regularly does so with other context situations- however the organization has a breaking point wherein it will stop counting if numbers become "unreliable" by the state, such as with Syria. How the UNHCR measures this level of verification is curious- especially in comparison to BorderDeaths who requires actual evidence for an individual to be included in their own count. This difference relates to the goals of state vs non-state counters because their approach to the matter and then, final number produced is for different purposes and shows in the datum.
There are two points that alternate between non-state counters and state counters: if accuracy in counting matters and relationship to politicization. Three aggregators are state countersUNHCR, Migrants' Files and IOM- while the other three are non-state- WTM, United and BorderDeaths. The only outlier is BorderDeaths because it is academia oriented however in regards to accuracy and politicization, it aligns with WTM and United.
State counters aim for accuracy, for a "real" number of how many people have died during migration to Europe. Their discrepancies arise from definitions, as earlier discussed. This correlates to the goals behind their counts including the reduction of costs and quelling migration for the Border Regime. IOM tends to be more humanitarian oriented but is still a UN organ project. What is ironic is that all three of these counters aim to be non-political in the numbers they produce, meaning that the number itself is the core of the work rather than what comes after. Their datum isn't about the individual who died but rather the incidents in which people died and how this can be avoided in the future. Migrants' Files shifts slightly on this stance, as it is non-state counters working to count on behalf of the state. However, the project itself was employed by the EC to gather datum on costs and aid in making new policies to further reduce said costs. The core issue that is attempted to be addressed by these state counts is how to stop illegal migration.
The three non-state actors presented are more politicized in their counts. Accuracy for the "real" number is not the upmost concern. The actual concern is what is done with the numbers- policies towards migrants. WTM considers its death count to be a mere by-product of its recording of human rights violations precisely because it understands that 1) illegal migration will always occur, so long as socio-economic inequality and conflicts persist and 2) migrants are not to blame for their actions, but it is the Border Regime who commits violations of inalienable rights that is to blame. BorderDeaths attains accuracy in their evidence-based count, despite the low number because of the humanized aspect behind the count. The identity markers are extensive, similarly to United. The goal is to influence policy to radically change behaviour regarding migrants. The numbers produced are therefore a reflection of these groups' understanding that the real issue at hand is how and why people are dying.
An interesting comparison may be derived between the UNHCR and Watch the Med. The two over-lap generally, in geography- in the counting process- which includes water routes from Morocco to Spain, Tunisia/Libya to Italy/Malta and Turkey to Romania/Bulgaria/Greece. Both only count water-based deaths. The major differing factor is over who is being counted and why. The UNHCR differentiates between refugees and migrant according to nationality, generally. Whereas WTM does not consider the reason for crossing the Mediterranean to be of concern because the primary goal is to record human rights violations of any person in this particular sea. The two counts are deeply interrelated, as both are counting incidents- things, not individual people. They differ however in that the UNHCR presents incidents of water-based deaths in an aim for a "real" number while WTM counter-maps the claims from UNHCR sources (states) to record violations from government apparatuses during the incidents. The numbers vary greatly between thousands of deaths recorded by the UNHCR and only hundreds by WTM. Which number is more powerful is clearly dependent upon the reader and their interest in the crisis. The difference in these two exemplify that the number published by each aggregator is a reflection of that actors' perception of the migrant crisis and which consumers it is targeting.
This document presents the different databases that account for migrant mortality in relation to the European continent, analyzes and compares the counts of the dead.
BorderDeaths sheds light on the importance of what happens after a migrant dies by utilizing an evidence-based approach to conglomerate trends. Figure v indicated that many deaths are not registered, not to speak of being buried. The treatment of dead bodies is part of how migrants are being treated in Europe. As BorderDeaths suggests from an academic perspective, Europe is not faring too well in this respect.
The United dataset demonstrated the importance of activists and civil society in over two decades of counting the dead. Yet, with Migrants' Files foundation in 2013 their sourcing shifted drastically to established media organisations and INGOs. The IOM dataset seems to complete this trend where a dead migrant is only a dead migrant if governmental reports or established media organisation claim so. Counting the dead has therefore shifted further away from activists to more mainstream, well-established and well-funded organisations.
Migrants' Files highlights a problem with respect to retrospective counting: 108 dead migrants in ten years is indeed unsettling low. One may wonder if it is even possible to reconstruct previous counts from other sources or if "live" counting, i.e. counting while a controversy is under way, is imperative.
As it becomes clear from the different definitions and methodologies applied, no standard approach to collecting data in this particular context exists, nor is there a common set of definitions that is shared by the different organizations. Every choice that is made by an aggregator during the process is a direct reflection of the political perception of the situation; bias is entirely unavoidable.
Most of the databases examined are incident-based, except for BorderDeaths, the only evidence-based count. In contrast to incident-based counts in other situations such as conflicts or disasters, migrant related counts are all estimates. A real number is impossible to attain, due to the difficulties to account for migration related deaths such as the thousands of estimated ghost deaths. Therefore, the aggregators all acknowledge that their counts rather present low estimates over a reliable number of actual migrant mortality to Europe.
One question that arises, especially following the discussion on state and non-state actors, is whose responsibility it should be to count migrant fatalities and what the purpose of doing so is for policy and for protecting migrants themselves.
Further research could analyze how different media outlets, NGOs and aggregators frame the death causes of migrants and what impact this has on death counts. A wide gulf opens between United's rather graphic and detailed descriptions to BorderDeaths sober-medical categories. What impact does this have on public opinion and the trends in migration itself? The verbs and adjectives that are assigned to a migrant's death could tell a lot about the framing of these deaths in the European public sphere and, beyond.
This report has been produced between January and May 2017 for a course at Sciences Po Paris, led by Thomas Tari. The aim of the course was finding and mapping controversies surrounding the practice of enumerating the deceased and dying. Our project group decided to look into the different counts focussing on deaths in the Mediterranean Sea.