Surveys

If the sample is large enough and drawn from a representative group of the population, inferences applicable to the entire population can be made. There are a number of sophisticated sampling techniques which can be used to capture rare populations, including probability and non-probability methods. In broad terms, probability samples mean that each sampling unit has an equal chance of being selected. One example of probability sampling is dual-sampling, which means locating sampling units with a higher proportion of persons from the target population, i.e., migrants, either using pre-existing data from a census or via sampling areas to discover them, and then oversampling in these units. This is particularly relevant for surveying migrants who tend to concentrate in specific geographic areas. Non-probability sampling methods include adaptative techniques such as snowballing, or asking respondents if they know any other people similar to themselves who might be interviewed. Non-probability sampling is useful for obtaining information on less prevalent or hidden migrant groups such as victims of trafficking or smuggling or irregular migrants but are limited in the generalizability of results due to the sampling bias (IOM, 2008).

Passenger surveys are face-to-face interviews with a sample of persons crossing an international border on a given day or during a given period. The sheer volume of movement that takes place between borders means that only a small portion of passengers sampled will be migrants as opposed to visitors or travellers. These data collection methods are usually more successful for island countries without land borders (IOM, 2008). 

The key strength of this source is the wealth of information collected relative to other data sources, which allows for in-depth analysis of the likely drivers and effects of migration (Global Migration Group, 2017). Surveys are an important tool for collecting migration data in countries where technical capacities for exploiting data from administrative sources are limited (IOM, 2008). They are usually designed and administered by NSOs which means the data comply with international statistical standards, are representative of the general population and can be analysed and disseminated in line with pre-existing conventions and protocols.

The low percentage of international migrants present in most countries means the sample size needs to be sufficiently large to identify enough migrants to allow for meaningful analyses and comparisons with non-migrants (Global Migration Group, 2017). Even with large enough sample sizes, the coverage of hard-to-reach migrant groups and gender-specific inequalities remain a concern. Depending on the country, these surveys might also be infrequent and/or costly (IOM, 2020a). Given that many surveys are either one-offs or planned on an ad-hoc basis, the data generated may not be comparable across surveys. Sample surveys are effective at measuring the characteristics and impact of migration, but less so at measuring the size of migration stocks and especially flows. A very large sample size is needed to measure specific country-to-country flows, and even then, depending on how weights are applied, invalid results are common (IOM, 2008). Household surveys also suffer from data quality concerns related to social desirability, respondent recall, respondent burden and question sensitivity, for example in relation to questions about earnings or sexual and reproductive health.  Regarding question sensitivity, it is important to note that respondents may be put at risk by simply responding to questions or by being seen with a survey enumerator (IOM 2020b). Emigration is particularly difficult to measure, since responses for household members living abroad are dependent on proxy respondents, as well as the possibility that entire households have migrated abroad.