We estimate the number of migrants and refugees that died while trying to enter the European Union, during a period of 25 years. Only a subset of attempts with at least one casualty are reported by at least one media source. In order to obtain the estimate, we propose a regression-extrapolation approach, for joint estimation of population size (here, the number of deadly individual or group attempts) and the sum of an accompanying trait (here, the number of deaths) over the population. The trait is measured only for a biased sample of individuals, that are repeatedly observed. Closed-form expressions are derived for the estimator and its standard error. Our findings are that about 40,000 have died from January 1993 to March 2019, during about 5500 attempts to enter the European Union. The number of deaths has been steadily increasing over time, and so has the number of deaths per attempt. About 20% of attempts with at least one casualty have not been recorded by any media source, and slightly less than 10% of deaths have thus been overlooked by media.
Introduction.Migration regulation is a controversial and politically sensitive theme. Border control is central in the political debate (Buonfino (2004), Marino (2016), Celata and Coletti ( 2016)), which is also strongly shaped by media (e.g., Herbers (2016), Vieira ( 2016)). According to the 1951 Refugee Convention, every human has the right to look for a shelter safe from war, ungrounded persecutions, life threats for their beliefs, political views, or for love. In this work, we focus on the specific phenomenon of refugees and migrants trying to enter "Fortress Europe" (Marino and Dawes (2016), Junemann, Scherer and Fromm (2017)). Restrictive policies of European states make it difficult for many refugees and migrants to enter Europe legally and safely. This makes movement of refugees also a public health issue (Smith and Daynes ( 2016)). Data from UNITED for Intercultural Action (see Section 2) collect information about attempts to enter the European Union (EU) that lead to at least one casualty, the number of casualties of each event, and the media sources that reported the event.In our application, unlike similar problems, our aim is not only to estimate a population size (the number events in which there was at least one death in trying to reach the EU), but also to estimate the total of an accompanying covariate: the number of deaths. Here, we define total as the sum of the covariate over observed and unobserved elements of the population. To the best of our knowledge, capture-recapture methods have never been used for this scope, and qualitative and quantitative measurements taken during sightings have only been used to improve population size estimates (e.g., by conditioning on covariates).Population size estimation methods are on the other hand not new to be used for investigating social phenomena. A recent review can be found in Silverman (2020). Examples of population size estimation methods for social phenomena include human rights violations (Mitchell ...