By Justin Schon, Postdoctoral Associate, University of Florida

In March 2018, the Expert Group on Refugee and Internally Displaced Persons (IDP) Statistics (ERGIS) released detailed reports on the status of refugee and IDP statistics and challenges in compiling these statistics. The reports made many valuable recommendations for how to increase the quality and quantity of migration data, but several recent developments highlight the need to also be more transparent about the types of uncertainty that exist in our measurements.

Uganda announced in October that a recent census had revealed that it currently hosts 1.1 million refugees, not 1.4 million as had previously been believed. IOM data on displacement from Mosul in Iraq during the 2016-2017 military offensive to retake the city from ISIS forces show a sudden jump in the estimate of IDPs due to a counting adjustment. Fabrice Balanche notes that UNOCHA decreased its estimate of Syrian IDPs from 7.5 million to 6.5 million during the fall of 2015, simply due to blatant overestimates that it knew were being provided.

Uncertain estimates even exist in refugee camps, where there are large numbers of humanitarian personnel. Officials in Jordan’s Zaatari refugee camp have significantly revised its estimated population multiple times after new counts. For example, the REACH initiative conducted a camp census from December 30, 2014 through January 18, 2015, and counted 7 954 fewer people in the camp than during the June 2014 count. On July 10, 2018, UNHCR deactivated nearly 11 000 camp registrations due either because they were absent from the camp, they were bailed out, they had registered elsewhere in an urban location, or they had returned to their country of origin.

These events suggest that a significant portion of changes in migration statistics may be due to changes in how the numbers are compiled or calculated, as opposed to an actual change in events on the ground. Users of migration data who are not aware of these issues are thereby prone to draw inaccurate inferences about migration dynamics, which may also lead to inappropriate policy responses. Uncertainty is impossible to eliminate, but it is possible to ensure that analysts are aware of the types of uncertainty to expect.

Different kinds of uncertainty exist in refugee and IDP statistics derived from surveys, censuses, administrative registers or UNHCR registration lists. Unfortunately, it is often difficult for data users to determine which source provided which data. To illustrate the value of being transparent about the data source, consider estimates of Syrian refugees in Jordan. UNHCR has consistently reported an estimate of Syrian refugees in Jordan that is about half the number claimed by the Jordanian government. This discrepancy exists because UNHCR reports the number of registered refugees, whereas the Jordanian government is estimating the total of registered and unregistered refugees. Considering the reality of many political concerns and vulnerability challenges that deter some people from registering with UNHCR, registered and unregistered refugees are potentially two very different groups of people. If we do not transparently include this information in our datasets, we risk missing a wealth of valuable information.

Then, we know that counting is done differently in stable, secure contexts than it is during emergencies. In an emergency, humanitarian personnel are often scrambling to obtain resources. This easily leads to very rough estimates. Over time, greater stability provides opportunities for rigorous surveys or a population census. The Zaatari camp provides an example where more rigorous counts led to a decline in the estimate of the camp’s population. That’s not an issue of negligence. It’s just a reality of humanitarian operations.

Different kinds of uncertainty can also arise from whom is doing the counting, and where that counting occurs. The Population Movement Tracking initiative in Somalia encountered this challenge as they relied on counts done in IDP settlements. As the ERGIS Technical Report on Statistics of Internally Displaced Persons states:

Prior to 2015, virtually all statistics on IDPs in the urban area of Mogadishu were based on data collected by humanitarian actors. This meant that data collected on IDPs were exclusively focused on populations residing in so-called IDP settlements, where IDPs were known to concentrate. As a result, all persons residing in these areas were automatically included into IDP statistics. In 2015-16, however, a profiling exercise was undertaken in collaboration with local authorities, national authorities, and humanitarian actors, which altered this practice. In the new profiling approach, IDPs were singled out from other population groups living in settlement areas through analyses of migration histories. The profiling revealed that approximately 85% of these populations were actual IDPs, while the rest represented members of the local urban population, Somali economic migrants, returned refugees, and refugees and migrants from other countries. From a resource perspective, this profiling approach helped to obtain more accurate information on the settlement areas where the most vulnerable and in-need-of-assistance populations were expected to live, regardless of their displacement history. The profiling did not, however, aim to produce comprehensive figures on IDPs in the whole city: though it was known that a large amount of the population in the city had been displaced at some point in the past. Rather than aiming to cover the total of this population, the process concentrated on obtaining more targeted information about the settlement areas where the most vulnerable and in-need-of-assistance populations were expected to live.

In Syria, IDP counts have been collected from pro-government and anti-government sources based on who controls the given piece of territory. Pro-government and anti-government sources do not appear to have the same biases in their reporting, a dynamic that is also analysed in new research on dam resettlement. Lebanon stopped registering Syrian refugees in May 2015, so it is much more difficult to obtain statistics after that point in time.

From the migrant’s perspective, the reasons are plenty to not disclose that they are displaced. Jordan’s example of registered and unregistered Syrian refugees highlights this dynamic for refugees. Migrants arriving in the European Union have also attempted to avoid registering in points of arrival when they want to move to countries like Germany or Sweden (due to the Dublin Regulation). In El Salvador, many IDPs do not report their displacement because they fear repercussions from the gangs that originally motivated their displacement.

All migration data is imperfect. It is unreasonable to expect otherwise. Yet, as we work to maximise the quality and quantity of migration data, we will benefit from also making information about our uncertainty in existing data as easy to access as possible. While some organisations may fear a backlash for openly admitting uncertainty, they should instead embrace the fact that transparency will increase confidence in their work. It will also allow more effective responses to their data.