Refugee dynamics — what does the data say?

Click here if you prefer to read this post on Wordpress.

Let me present you the visual exploratorium of refugee flows over the world using dynamic chord diagrams! Perhaps one of the hottest topics today is that of refugees and immigrants. There are conflicting views about stately attitudes towards unwanted immigration and the debates end nasty. Spanning from Europe through the Middle East, Central Africa and all the way to Eastern Asia, refugees are topic of constant uproar — and unavoidably a handy political (populist) tool. But what is the reality behind the movement of the world’s refugees? Without taking any political stance, let us see, what does the data say about refugee dynamics!

The visual exploratory of refugee flows over the world using dynamic chord diagrams

The visual exploratorium of refugee flows over the world using dynamic chord diagrams

While it is known that natural disasters and wars (especially civil wars — read my previous post about insurgent dynamics) are the main causes of pushing somebody into jumping the border, these people-flows can also act as a catalyst for igniting other conflicts in the receiving regions. The linkages between the Second Congo War and the Rwandan and Burundian civil wars are prime examples for cross-border conflict overspill. Therefore, most countries are reluctant to receive refugees in large numbers, nonetheless in culturally integrated regions, the process is unavoidable. However, when talking about the receiving countries, you might think that the wealthy, developed nations take in the most refugees. You couldn’t be more wrong… It is the developing nations, oftentimes those with the smallest GDPs in the world taking in the largest number of people. In fact, in 2013, there were no developed countries among the top 25 refugee receiving countries. However, based on recent cryses in the Andaman sea, and repeatedly over the Mediterranean, Western refugee intake is expected to soar… a bit. But we will investigate how really off the scale the mainstream media is, when talking about refugee and illegal immigrant invasion in Europe for example.

The United Nations High Commissioner for Refugees (UNHCR) has a well established reporting mechanism for tracking refugees and asylum seekers from and in countries across the world (except for Palestinian refugees, the largest refugee body in the world and, kept under account by a separate agency, United Nations Relief and Works Agency for Palestine Refugees in the Near East (UNRWA) . Using this data, many data visualization interfaces have been written over time in order to present in an aesthetic and easily understandable way to the general public. I would like to mention The Refugee Project, in particular, applaudable for their world map-style visualization of all refugee flows since 1975 (the entire database span), connecting it to major world events, as well as the Migration Policy Institute for including UNRWA numbers for Palestinian refugees, otherwise not present in the main UNHCR database, and finally UNHCR’s own dataviz app.

Inspired by the above, I have decided to build my own refugee flow diagrams. Since UNRWA does not have a publicly accessible database, I have estimated the number of Palestinian refugees using the World Bank totals for Jordan, Syria, Lebanon and Palestine, discounted by the UNHCR numbers. While this methodology neglects Palestinian refugees living in countries other than the above 4, it counts dual citizens (especially Jordanians). I have decided on using a chord diagram layout, useful for visualizing bi-directional flows. Inspired by the visual storytelling with a chord diagram of visual cinnamon and’s dynamic chords, I have created the visual exploratorium of refugee flows over the world using dynamic chord diagrams (full screen — interactive, static):

The visual exploratorium of refugee flows over the world using dynamic chord diagrams — tutorial screen (full screen — interactive, static)

The visual exploratorium of refugee flows over the world using dynamic chord diagrams — tutorial screen (full screen — interactive, static)

In the beginning, there is a short tutorial that explains how to read the chord diagrams — visual cinnamon style — and the main app features two chord diagrams, side-by-side. One of them is a target-based, while the other one is a source-based accounting. This means they display the receiving (initially on the left) or the sending (right) countries. You can filter out some countries from the diagram by clicking on their label. And you can also filter out all countries with flows smaller than a threshold (set to 100 000 people by default). All data is dynamic — spanning over the time period 1975–2013! Drag the time slider, scroll with your mouse or use the arrow keys to change the year.

Chord diagram representation of refugee flows over the world in 2013 (full screen — interactive, static)

Chord diagram representation of refugee flows over the world in 2013 (full screen — interactive, static)

At a glance, in 2013, the largest refugee source countries are Palestine, Afghanistan and Syria by far, followed by Somalia, Sudan, Viet Nam, DRC, Iraq, China, and other Central African nations. The first 3 countries give more than 2/3rds of the global refugee population. On the receiving side, Pakistan and Iran take the Afghans, the largest flow in the world, while the Syrians are split almost equally between Jordan, Lebanon and Turkey, while the Palestinian refugees reside in Palestine, Jordan, Syria and Lebanon. The Somalis flee to Kenya and Ethiopia, while the Sudanese to Chad and South Sudan.

When looking at the proportion of refugees reported to the total national population, we find that 1 in 23 people in Jordan is a refugee (highest in the world even without Palestinians — that would make it less than 1 in 2!), while Lebanon, Syria and Central African nations come close to the 30–50s. This ratio is 1/1206 in the US, 1/417 in the UK, 1/933 in Italy, 1/292 in France and 1/140 in Germany — one of he highest in Europe by far — and 1 in 56 in Montenegro. In the Asian developed nations, there are practically no significant refugee populations, with Japan at 1/50 000 and Korea at 1/100 000!

We can see that most of the chords look like wedges, which means that most of the refugee flows are unidirectional! That is, people are only escaping from or purposefully going to a certain country. It is almost exclusively when two neighbors are both in conflict when the flow turn bidirectional such as Iraq-Syria in the early 2010s or Azerbajian and Armenia or Rwanda, Burundi and the DRC in the beginning of the 1990s. When browsed in sync with the The Refugee Project and Insurgent Dynamics websites, the exploratorium reveals some interesting patterns. Let us scroll back to the beginning of our time interval, 1975, and look at the largest (scale of hundreds of thousands) refugee flows.

  • In the mid 70s, the bulk of the global refugees comes from Palestine, but there is also a lack of data and a lot of countries are lumped together under the Other category. Ivory Coast, the US and the DRC are the largest, Non-Middle-Eastern asylums, mostly sprouting from the civil wars in Ivory Coast’s neighbor Guinea and DRC’s neighbor Angola.
  • In 1977, Somalia invades Ogaden, resulting in vast numbers of refugees fleeing Ethiopia for Somalia and Sudan for years to come, eventually totaling around 3 million.
  • 1979 marks the start of the 10-year Soviet invasion of Afghanistan, resulting in the largest ever refugee flow in history to neighboring Iran and Pakistan, totaling more than 4 and half million people. Since the Afghan state did not really manage to return to a stable government ever since, a lost of these people are still displaced.
  • In the early 80s, from the Western nations the US and Australia host hundreds of thousands of refugees, with the UK and France each also hosting 100 000+. By the end of the decade Germany becomes the major Western host (500 000+).
  • The conflict in Ogaden continues well into the 80s, and combined with a terrible famine, the number of Ethiopian refugees in Somalia and Sudan hits 1.7 million.
Refugee flows in 1980 (full screen — interactive, static)

Refugee flows in 1980 (full screen — interactive, static)

  • Towards the end of the 80s, the Mozambique civil war takes a turn for the worst and 1.4 million people flee into Malawi and Zimbabwe.
  • The first Gulf War of 1991 forces 1.2 million Iraqis to Iran. The situation in Afghanistan does not improve, there are still 4.5 Afghan refugees in Iran and Pakistan.
  • After being a receiver country for a decade, 700 thousand people flee Somalia as the civil war breaks out. The worsened war situation in Liberia, the start of the civil war in Sierra Leone, as well as the power struggle in Eritrea displaces 500 thousand people from each of the countries. The receivers are Guinea, Sudan and Ethopia.
  • As the Soviet Union collapses, hundreds of thousands of people cross-flow between Azerbaijan and Armenia as the countries enter into conflict.
  • There are hundreds of thousands of Vietnamese refugees in China, ever since the aftermath of the Vietnamese war in the late 70s.
  • In 1994, Palestinian refugees in total pass 3 million, becoming the largest refugee group. At this point, there are still 2.7 million Afghans displaced.
  • More than 2 million desperate Rwandans flee in every neighboring country (DRC, Burundi, Tanzania, Uganda) to escape the genocide. A few years later, the overspill starts the lengthy Burundian civil war. 500 thousand flee to the only stable country of Central Africa of the time, Tanzania.
  • 800 thousand Bosnians seek refuge in Croatia and Germany. At this point, the US is still a large target country, with more than 400 000 refugees. Germany, however, has 1.3 million. 200 thousand Russians seeks refuge in America after the fall of the Iron Curtain. Vietnamese are 200 thousand.
  • Liberians are still by far the largest Western African refugee group with 200–200 thousand people in neighboring Ivory Coast and Guinea.
Refugee flows in 1994 (full screen — interactive, static)

Refugee flows in 1994 (full screen — interactive, static)

  • By 2000, Palestinian and Afghan refugees are both approximately 4 million. There are about the same number of Palestinians living in Palestine and Jordan (1.5 million) and Syria and Lebanon (400 thousand). Iran and Pakistan are the largest refugee hosting nations, mostly Afghans and Iraqis.
  • As the Sudanese civil war rumbles on, 400 thousand flee to all neighboring countries (Ethiopia, Uganda, DRC, Kenya). Angola is still not stable and 400 thousand Angolans live in exile.
  • In 2006–2007, violence in Iraq and the repercussions of the US invasion in 2003 force 700 thousand people to flee to Syria and 500 thousand to Jordan. At this point, Jordan is hosting 2.4 million refugees, the largest number of any country by far. By 2007, the number of Iraqi refugees in Syria rises to 1.5 million. This makes Syria temporarily the second largest target country after Jordan, overtaking Palestine, Iran, Pakistan and Lebanon.
Refugee flows in 2007 (full screen — interactive, static)

Refugee flows in 2007 (full screen — interactive, static)

  • The sustained conflict in Afghanistan keeps the number of Afghan refugees high in Iran and Pakistan and the renewed conflict in Somalia displaces 500 thousand to Kenya, Yemen and Ethiopia. By 2010, Afghan refugees again reaches 3 million, while the number of Palestinians surpasses 5 million. In 2011 and 2011 the situation in Somalia worsens a lot, with refugees totaling 1 million now.
  • In 2013, as a result of the civil war, Syrian refugees total 2.3 million: 850 thousand in Lebanon and 600–600 in Jordan and Turkey, while around 200 thousand in Iraq and Egypt each.
  • The conflict in Central Africa is ongoing with the DRC, the Central African Republic and more recently Burundi all experiencing a power struggle. Their refugee populations number hundreds of thousands.
  • In 2013 Germany had 326 thousand refugees, France 42 thousand, the UK 34 thousand and Italy a mere 11 thousand.

One can see most of the refugees are flowing between 3 hot spots: Levant, Afghanistan and Central Africa. As conflicts are spilling over (let us disregard geopolitcal interest and/or tactical foreign intervention for now), people are desperate to move to the nearest country. In most of the cases that country is also fragile, since it has also recently been or currently is in a state of crisis. This significantly raises the possibility resurgence and creates a conflict deadlock. The number people involved in the refugee flows within these regions is at least one order of magnitude larger than people involved in conflicts and disasters in other developing parts of the world and they are roughly two orders of magnitude higher than refugee flows to developed nations.

While the per capita GDP of the developing Middle Eastern nations is roughly 5–10 times lower than that of Western Europe, their refugee populations, as well as population shares are incomparably (50–100) higher — not even talking about the Central African states with average per capita GDPs up to 100 times lower than in Western Europe, but almost comparable refugee flows to the Middle East. One might say that the cultural differences within these regions are smaller than with Europe, but this still does not justify the European nations’ recent attitude towards rejecting refugee flows, insignificant in size compared to global flows, in great deal resulting from European assistance in creating crisis situations anyway… Going back to the visual exploratorium, for the year 2013, when filtering out flows smaller than 100 000 people, no European nation shows up on the diagrams!

Refugee flows in 2013 (full screen — interactive, static)

Refugee flows in 2013 (full screen — interactive, static)

In terms of technical construction, the main page index.html hosts only one iframe. Then, the source attribute is dynamically changed, based on which page are we displaying:

  • demo.html, a page hosting the tutorial
  • parent.html, a page hosting the chord diagrams

The parent.html is also only a container for for two other iframes. The one initially on the left, #iframe1 sorts and displays the data based on target countries. This data is loaded with d3.js from the data.csv, and displayed using’s dynamic chords and updated over time and filtered with angular.js. The frame on the right, #iframe2 is a replica of the first one, the only difference being that this one loads the data2.csv, a file identical to data.csv, but with the source and target attributes swapped. The time slider is hosted in parent.html and it is built using dragdealer. Whenever a skidding or scrolling event is fired, the new year is passed along to the update functions of both iframes and the chord diagrams change smoothly.

This post describes a web-app that I have created that visualizes refugee flows over the world using dynamic chord diagrams and the following visual data analysis. The open data was taken from UNHCR and the World Bank and processed them into chord diagram input CSV types with this IPython notebook, using pandas. Visualizations were entirely made with d3.js and they all extensively use svg format. The webb app tutorial was inspired by visual storytelling with a chord diagram of visual cinnamon and the main layout by’s dynamic chords with angular.js. The main outcome is the visual exploratorium of refugee flows over the world using dynamic chord diagrams (full screen — interactive, static). If you liked this post or have any questions or thoughts, Like, Share, Comment, and Subscribe! If you think my work is cool and you would like to support it, please consider a small Donation.