Tom Moultrie ANALYSIS: What do we really know about international migration to & from SA?

Demography is regarded as the most quantitatively strong discipline in the social sciences but there are limits to what can be known and measured, especially about migration in South Africa. Demography professor Tom Moultrie explains.

The analysis of migration is the “Cinderella” of demographic measurement. Unlike data on births and deaths, data on migration are notoriously hard to collect.

In South Africa’s censuses and national surveys, two forms of questions are asked to identify migrants. The first type asks about the place of birth – the province if born in the country, or the country of birth if born outside South Africa.

This provides a measure of lifetime migration but provides no information on the timing of the migration, or on movements that have occurred between birth and one’s current place of residence.

The second type asks whether the respondent was living in the same area at some point in the past (usually one, five or ten years before the present), or – alternatively – the date and origin of their last move.

This question provides a more current picture of population movement but suffers from similar imprecision relating to multiple movements in a short-ish period of time.

Producing measures of migration from such limited data is no easy task, and these measures are routinely compromised by structural and individual-level biases and distortions.

Immigrants may not honestly declare their country of birth (perhaps for fear of xenophobic identification, or because they are undocumented immigrants). Emigrants, by definition, are not captured in a census, because they have left the country.

At best, demographers’ ability to quantify migration with any accuracy from such partial and incomplete data is extremely limited. Because the estimation of recent migration is less straightforward than that of lifetime migration, much popular analysis of migration focuses on measures of lifetime migration.

Different sources of data show different numbers

Two internal data sources provide the bedrock for estimating international migration to and from South Africa: the decennial census – the most recent conducted in 2011 – which seeks to enumerate the entire population of the country, and which asked questions on both countries of birth and citizenship.

There is also a more ambitious exercise in the form of the 2016 Community Survey, which asked the same questions as the census as well as attempting – for the first time – to quantify emigration from South Africa by asking about members of households who had left the country.

Data from the 2011 census suggested approximately 2.2 million foreign-born people in South Africa. The recently-released 2016 Community Survey put that number at around 1.6 million.

Agencies outside of South Africa have also weighed in with widely disparate estimates. The United Nations Population Division estimates there were 3.14 million foreign-born people in South Africa in mid-2015, with 840,000 South Africans living outside the country.

The organisation’s refugee agency, the UNHCR, incorrectly suggests that there are more than a million unresolved asylum claims in South Africa. These estimates are also publicised by another UN agency, the International Organisation for Migration.

In addition to the conceptual difficulties mentioned before, two additional factors contribute to the uncertainty surrounding the number of international migrants. First, Statistics South Africa’s own data is contradictory and second, the methods used to estimate numbers of migrants by various UN agencies are not without their own problems.

Aggregate census picture sensible, constituent parts make little sense

The 2011 census covered some 15 million households. Because of the large size of the data captured, and the need for to ensure that individuals cannot be identified from the census data, census bureaus do not typically offer a complete set of individual-level data to researchers.

Instead, in order to check the integrity of all the census data, an official (representative and randomised) 10% sample is drawn from the census data, further anonymised and released.  

The official tally reported in the 2011 census stated that there were 2.19 million foreign-born people in South Africa. The 10% sample of the same census agreed with this, at 2.18 million. However, when this data was observed at a regional level a number of discrepancies emerged that can not be explained.

The number reported in the full sample as born in the UK and Europe is less than half of that indicated by the 10% sample. The number reported as coming from SADC countries is 17% lower. And the number reported as “not specified” (but not born in South Africa) is nearly three times greater than in the 10% sample!

Evidently, the 10% sample has been subjected to extensive cleaning and editing, in a way that was not done with the original data released to the public.

What this means is that we do not have a strong basis on which to rely on data from the 2011 census. Patterns and trends of migration that appear to make sense on a national level fall apart when evaluated at a provincial level, or by population group. The aggregate picture is sensible, but the constituent parts make little sense.

Also, the census 2011 numbers do not square with either those prepared by the UN Population Division nor with the data from the 2016 Community Survey.

The UN estimates are derived from their own mathematical model, although country-level data are incorporated into their projections. Importantly, because the UN population division methodology is done at a global level, some indication of where South African emigrants are now resident can be divined from the country of birth data recorded by receiving countries.

Striking differences are evident in the numbers of people born in Europe and North America thought to be resident in South Africa.

Limits to what can be known & measured

Of greater concern, though, is that the number of immigrants reported in the 2016 Community Survey is half that estimated by the UN and about a quarter less than reported in the 2011 Census. Certainly, it is improbable that the number of immigrants to South Africa has fallen by this amount.

This points to a potentially severe problem with the 2016 Community Survey data (a point conceded by Statistics SA in their report on the Community Survey), although the underlying data have yet to be put into the public domain. External checking of this data, therefore, is not yet possible.

While the report on the Community Survey does not provide any hard numbers on emigration from South Africa, the manner in which the survey asked questions about migration was not consistent with the definition of lifetime emigrants used by the United Nations. Comparisons therefore are not recommended.

The very curious distribution of the timing of recent emigrations that Statistics SA’s data reveals, together with their own cautions about interpreting these data, would militate against using the Community Survey data at all to understand recent patterns of immigration to and emigration from South Africa.

Demography is regarded as the most quantitatively strong discipline in the social sciences, and demographers pride themselves on being able to draw sense from census and survey data, even from limited, defective and deficient data. But there are limits to what can be known and measured.

How many immigrants are in South Africa? We simply do not know, except to say that the likely answer lies between one and three million.

Tom Moultrie is professor of demography and director of the Centre for Actuarial Research (CARe) at the University of Cape Town.  

© Copyright Africa Check 2017. You may reproduce this piece or content from it for the purpose of reporting and/or discussing news and current events. This is subject to: Crediting Africa Check in the byline, keeping all hyperlinks to the sources used and adding this sentence at the end of your publication: “This report was written by Africa Check, a non-partisan fact-checking organisation. View the original piece on their website", with a link back to this page.

Comment on this report

Comments 1
  1. By Diego Iturralde

    As the Stats SA official responsible for demography, which includes migration I am only to happy to respond to this blog. Let me start by thanking Prof Moultrie for this enlightening piece which explains how migration is captured in Census so succinctly. There are however some comments that his piece warrant.

    The most important comment I wish to make is that the 10% sample did not under go any further editing after the sample was drawn. It was based on the final version of the full dataset. The reason why the full dataset and the sample may not match by country of birth is that the sampling domain used for weighting the persons data set was age group, sex, population group and municipality. With this data set covering so many themes it was not possible to make it representative across all domains and hence the above were considered to be the most important ones. A document with details around the method of sampling is available from Stats SA. When compared to the UN estimates of 2010 the Census matches up very well to it (2.0 mill vs 2.18 million).

    Regarding the 2016 CS, I share the concerns raised by the author around the declining number of foreign born persons and I can confirm that Stats SA has been addressing the issue with various migration experts around the country to ensure that we understand was drove the generation of such numbers. Once complete a report in this regard will be released. The CS data is incidentally in the public domain albeit under the 2011 geographic demarcation and it will soon be updated with the 2016 demarcation.

    With regards to the emigration module, this was done on recommendation of the UN Population division at a workshop held for all anglophone African countries in Addis Ababa. The intention was never to measure lifetime emigration but rather period emigration (10 years) for persons who were household members in this time but are no longer in SA. This includes SA born persons as well as others born outside SA who left SA during this period. Many lessons were learned about this module and improvements will be added to it for the next opportunity when we can ask it.

    Migration is, as can be seen, a complex thing to measure at the best of times, but with improved data collection strategies and better integration of administrative data, demographic surveillance site data and the opportunities that big data presents, we will only get to have a better understanding of this complex but very critical phenomenon.

    Diego Iturralde

    vote
    Reply Report comment

Leave a Reply

Your email address will not be published. Required fields are marked *

*

Africa Check encourages frank, open, inclusive discussion of the topics raised on the website. To ensure the discussion meets these aims we have established some simple House Rules for contributions. Any contributions that violate the rules may be removed by the moderator.

Contributions must:

  • Relate to the topic of the report or post
  • Be written mainly in English

Contributions may not:

  • Contain defamatory, obscene, abusive, threatening or harassing language or material;
  • Encourage or constitute conduct which is unlawful;
  • Contain material in respect of which another party holds the rights, where such rights have not be cleared by you;
  • Contain personal information about you or others that might put anyone at risk;
  • Contain unsuitable URLs;
  • Constitute junk mail or unauthorised advertising;
  • Be submitted repeatedly as comments on the same report or post;

By making any contribution you agree that, in addition to these House Rules, you shall be bound by Africa Check's Terms and Conditions of use which can be accessed on the website.

*