Guest post by Elias Steinhilper and Rob Gruijters. Elias is a lecturer in political science at the University of Freiburg and PhD candidate in Political Science and Sociology at Scuola Normale Superiore in Florence. While drafting this blog post, Elias was a visiting scholar at the Institute for Societal Issues at UC Berkeley. Rob is a postdoctoral researcher at the Department of Sociology, University of Oxford.

The term ‘refugee crisis’ is almost exclusively used to refer to the number of arrivals and EU’s (in)ability to accommodate them, rather than the risks and suffering faced by people fleeing war and deprivation. According to the dominant portrayal, the ‘refugee crisis’ has started around spring 2015, reached its climax in summer and autumn 2015 and has been gradually fading out since the closure of the Balkan route and the implementation of the Turkey-EU deal in March 2016. The dominant crisis narrative has obscured the very real humanitarian crisis, which continues unabated at Europe's Mediterranean Sea borders. The annual number of border deaths has increased continuously over the past three years, reaching more than 5,000 in 2016 alone, a number never recorded before. Yet, the highly polarised and politicised debate on migration and migrant deaths is often devoid of a deeper understanding of migrant mortality. Both policymakers and public opinion react to incidents, and empirical data is often absent or presented in a selective and misleading way. In this post, we present analyses from a number of new and recent data sources on migrant arrivals and migrant deaths, covering the period between 2010 and 2016. We seek to address three questions:

  • How does the number of deaths vary between routes and over time?
  • How does the risk involved in crossing vary between routes and over time?
  • Do Search and Rescue (SAR) operations encourage more and riskier crossings (‘pull factor hypothesis’)?

The findings presented here are derived from a larger study, which is available upon request. We also provide links to the relevant data and analytical templates for researchers who wish to replicate our results or conduct further analyses.

Question 1: How does the number of deaths vary between routes and over time?

In 2016, more than 5,000 migrants died during their attempt to reach Europe by sea. This was the highest number ever recorded, bringing the total number of casualties between 2000 and 2016 to 31,799. Figure 1 shows that the vast majority of these deaths took place on the Central route, e.g. the routes to Italy and Malta.

Figure 1: Migrant fatalities in the Mediterranean region, by year and route, 2000–2016

Source: The Migrant Files (2000–2013) and the Missing Migrant Project (2014–2016).

The above figure also shows that the number of fatalities varied greatly between routes and over time. For example, while 2010 was a year with comparatively low mortality, 2011 saw a massive increase, particularly on the Central Mediterranean route. In the last three years (2014–2016) the number of recorded deaths remained consistently very high.

Question 2: How does the risk involved in crossing vary between routes and over time?

Part of the variation in death tolls, both over time and between routes, is due to fluctuations in the number of migrants. The total number of deaths in a given region or period therefore does not provide much information about the actual risk involved in crossing. By combining data on the number of fatalities and the number of arrivals, we can calculate relative risks or migrant mortality rate (MMR): the number of reported dead and missing persons per 1,000 attempted crossing (also see Carling 2007). The findings, by route, are plotted in Figure 2.

Figure 2: Migrant mortality rates, by year and route

Source: The Migrant Files (2010–2013), the Missing Migrant Project (2014–2016) and Frontex.

Figure 2 shows that mortality rates differ greatly between routes. The Eastern route has consistently been the least dangerous, and the Central route the most so. In 2015, the risk of dying on the central route was over 19 times higher than on the Eastern route (15.4 vs. 0.83 out of 1,000 crossings). We find no clear time trend in overall mortality rates. Instead, it appears that crossing has always involved a high and persistent risk, particularly on the Central and Western routes. It is particularly worrying that the mortality rate on the Central route appears to have increased substantially in 2016, which saw a relatively high number of arrivals. During this year, 26 out of 1,000 migrants (or one in 39) did not survive the attempt to reach Italy or Malta. The overall mortality rate was lowest in 2015, mostly because the high number of arrivals during that year was concentrated in the comparatively safer route from Turkey to Greece (also see this report by the IOM).

Question 3: Do Search and Rescue (SAR) operations encourage more and riskier crossings?

One of the most controversial arguments in the migration debate relates to the role of Search and Rescue (SAR) activities as a potential 'pull factor' for irregular migration. Several leading politicians, including senior EU leaders, have claimed that the presence of rescue boats close to the Libyan coast serves as a ‘magnet’ or even a ‘bridge to Europe’, tempting smugglers to send larger numbers of migrants in increasingly unseaworthy vessels, thereby increasing the overall death toll.  Most recently this view was expressed in a (secret) Frontex communication obtained by the Financial Times, which criticised NGO rescue operations. Others have strongly criticised this claim, arguing that it is the insufficiency of SAR capacity that causes the large number of deaths, and that migration is driven by factors unrelated to SAR.

In our analysis we focus on the Central Mediterranean route, where most accidents and deaths take place. Over the past few years, we can identify three different phases in Search and Rescue activity in this area. Mare Nostrum (October 2013–October 2014) was a comprehensive SAR mission with an explicit humanitarian objective, managed by the Italian navy. During its operational period, the mission rescued around 130,000 people, often operating close to the Libyan coast (Italian Navy 2014). Following the criticism outlined above, Mare Nostrum was discontinued and succeeded by the new Frontex operation Triton. Triton was primarily a mission to secure the EU's external borders and not explicitly designed as an SAR mission (Andersson 2014). Moreover, it had a significantly smaller budget and rescue area compared to Mare Nostrum. Following two large-scale accidents in April 2015, Triton was substantially upscaled and expanded (Triton II). Around the same time, a number of NGOs started operating SAR missions in the area, so that as of June 2016 rescue capacity returned to about the same level it had been during Mare Nostrum.

In order to assess the ‘pull factor’ hypotheses, we compare the low-SAR period (November 2014–May 2015, just after the end of Italy's Mare Nostrum operation and before the upscaling of Triton) to the period before and after, during which time there was a comparatively high number of SAR operations. Because migration is seasonal (there are more arrivals in summer than in winter, other things being equal), we only look at arrivals in the equivalent months (November 2013–May 2014 and  November 2015–May 2016). If SAR operations do encourage more arrivals and increased risks (e.g. due to overcrowding or the use of lower quality boats), we would expect more arrivals and higher mortality risks in the high-SAR periods.

Figure 3: Total number of arrivals (left) and mortality rate (right) by Search and Rescue (SAR) period

Note: Arrivals (left) based on equivalent months only (MN: November 2013–May 2014, Triton I: November 2014–May 2015, Triton II: November 2015–May 2016). Mortality rates (right) based on the full periods (MN: October 2013–October 2014, Triton I: November 2014–May 2015, Triton II: June 2015–December 2016)

The findings (Fig 3) show that the number of arrivals in the low-SAR period was not higher than in the equivalent high-SAR periods, as predicted by the pull factor hypothesis. In fact, arrivals were highest in the low-SAR period. Moreover, we can observe that mortality rates were substantially higher in the low-SAR period (Triton I) than in the periods before and after.

The high mortality rate during Triton I is largely the result of two large accidents on 13th  and 18th  April 2015, with estimated casualties of 400 and 750 people respectively. However, it would not be appropriate to consider these accidents outliers that were unrelated to the (absence of) SAR capacity. The excellent ‘Death by Rescue’ investigative report by the University of London’s Forensic Oceanography department analysed the circumstances of both accidents, using multiple sources such as photos, interviews with shipwreck survivors, rescue vessel crews, statistical data, GIS locations and internal reports by national authorities. It concluded that the deaths could have been prevented, had a more intensive SAR mission been in place: ‘[the EU's] policy of retreat from state-led Search and Rescue (SAR) operations shifted the burden of extremely dangerous search and rescue operations onto large merchant ships, which are ill-fitted to conduct them. In this way, EU agencies and policy makers knowingly created the conditions that led to massive loss of life in the April shipwrecks’.

In combination, these results strongly suggest that SAR operations reduce mortality risks (or conversely, the absence of SAR operations leads to more deaths), and has little or no effect on the number of arrivals.


Our findings point out the continuing failure of European authorities to provide a comprehensive answer to the humanitarian crisis. More than 30,000 mostly young people have died in the past 16 years, and a resolution appears nowhere in sight. Torn between its legal and moral obligations towards refugees and its intent to maintain control over its external borders, the EUs border policy has been both inconsistent and ineffective. As a result, the humanitarian crisis in the Mediterranean is worsening and, given the current geo-political circumstances, unlikely to end anytime soon.

Any comments about this post? Get in touch with us! Send us an email, or post a comment here or on Facebook. You can also tweet us.


How to cite this blog post (Harvard style)

Steinhilper, E. and Gruijters, R. (2017) Border Deaths in the Mediterranean: What We Can Learn from the Latest Data. Available at: (Accessed [date]).