Do immigrants “steal” jobs in South Africa? What the data tell us

On 1 May 2008, riots broke out against foreign nationals in Alexandra township, a densely populated settlement some two miles east of Sandton, sub-Saharan Africa’s financial centre. The iconic image of Mozambican national Ernesto Nhamuave being burnt alive shocked South African audiences and quickly spread across the world. Within two weeks, “xenophobic violence” had spread across South Africa. Those targeted were largely foreign African and South Asian nationals. After two weeks, more than 60 people had been killed and more than 100,000 displaced. The “charges” against immigrants are typical of anti-immigrant sentiment globally: a 2010 survey by the Southern African Migration Programme found that 60% of South Africans believe immigrants “take jobs”, whilst 55% believe that they worsen crime.

But in fact we know very little about the actual labour market effects of immigration to South Africa.

The best source of useful demographic information about immigrants in South Africa is the census, the most recent of which was in the 2011 census. There are, of course, problems with using this data. Demographic trends may well have changed in the six years since 2011. There is also a risk that immigrants may dodge census surveyors if they are scared of being victimised by the authorities, so that the data may underestimate immigration to South Africa. But, taking the data at face value, we can glean useful demographic features of immigrants compared to South African born workers.

The table below presents useful demographic, educational and labour market statistics for the locally-born population and immigrants. There were approximately two million immigrants recorded in the 2011 census, compared to 49.6 million South Africans. The vast majority of immigrants are black African (71.7%), with 4.2% of Indian or Asian ethnicity, reflecting the sizable Pakistani and Bangladeshi population in the country. Often small entrepreneurs (e.g. spaza shop owners), these groups have been consistently targeted in xenophobic attacks.

Zimbabwe is by far the largest immigrant country of origin, reflecting the influx of Zimbabweans during and after the country’s economic crisis in the early 2000s.

The large populations of people from Lesotho, Mozambique and Malawi reflect South Africa’s history as a destination for migrants to the gold and platinum mines.Immigrants are significantly more likely to hold a diploma or university degree.

A total of 1.5% of the immigrant population holds some form of post-school qualification, compared with only 0.3% of the South African population.

Finally, it is worth emphasising that more than 550,000 new immigrants moved into South African municipalities between 2001 and 2011. That is a sizeable influx of people for a country of South Africa’s size.

Table: Characteristics of locals and immigrants 

	Locals	All Immigrants	Recent Immigrants
Demographic Characteristics
Total population (in millions)	49.60	2.16	0.56
% of total SA population	95.81	4.19	1.08
% black	79.53	71.7	57.17
% white	8.51	16.88	34.69
% coloured	9.27	0.85	1.17
% Indian/Asian	2.41	4.20	3.42
% living in 4 largest metros	27.21	48.05	43.23
Education
average years of schooling	9.27	9.21	8.52
% with post-school qualification	0.31	1.54	1.62
Employment
employment rate (expanded)	58.78	77.17	77.73
% manager (CAT1)	4.80	7.47	8.88
% professional (CAT2)	4.31	5.64	6.24
% technician (CAT3)	5.88	5.56	6.40
% clerk (CAT4)	7.32	7.70	8.64
% sales and services (CAT5)	9.56	13.00	11.93
% skilled agriculture (CAT6)	0.55	0.68	.74
% craft and related trades (CAT7)	7.06	11.95	11.90
% plant/machine operator (CAT8)	3.98	4.91	5.58
% elementary (CAT9)	15.46	20.53	18.30
Immigrant Countries of Origin (%)
Zimbabwe	-	31.00	14.74
Mozambique	-	18.08	25.16
Unspecified	-	11.00	14.40
Lesotho	-	7.84	9.25
Malawi	-	4.02	2.12
Other	-	28.06	34.33

Difficulties measuring the true effects of immigration

Generally, economists use a technique called regression analysis to address questions like these. Regression analysis is a statistical technique that measures, crudely, how well one variable explains another. The intuitive way to estimate the labour market effects of immigration would be to divide South Africa into geographical areas (say, municipalities), and compare how different levels in the immigrant population correlate with employment rates. Unfortunately, it is extremely difficult to do this credibly, because immigrants may move to particular areas because of other economic forces that also affect labour market outcomes.

This is a common problem in economic analysis, known as endogeneity. The intuition behind this is simple: suppose Johannesburg experiences positive economic growth in a number of industries before the 2011 census. This causes employment to rise, but also attracts immigrants to settle in Johannesburg. Thus, a naïve correlation between the high proportion of immigrants in 2011 and employment might falsely suggest that immigration improves employment, when in fact the employment rates of both immigrants and locals were pushed up by other economic forces.

To get around this, the first step is to control for as many observable things as possible that might otherwise influence the job prospects of locals. Fortunately, the census reports a rich set of variables that can be included in a model, such as race, age, years of schooling, whether a person lives in a rural or urban area. Including these in a model means the risk of endogeneity is lower. However, there is still a risk that things you can’t observe affect the migration decisions of immigrants and the labour market prospects of locals. Getting around this problem is tricky.

My models follow an approach pioneered by leading international migration economist David Card. I break up the whole South African working population into municipalities where they live and job categories, based on those reported in Table 1. I include all the controls above and also use what is called a fixed effect to pick up unobservable features that might influence labour market outcomes. This means that I include a dummy variable for each municipality and job category, that picks up anything affecting wages or employment not captured in the controls. The full technical paper is available on request.

Results from the models

Results from my models suggest that immigration either has a very small positive effect on overall employment in South Africa, or none at all, depending on the version of the model used (I run several versions of the same model for robustness). The models predicting a positive employment effect suggest that it is extremely small in magnitude. A 1% increase in the relative foreign population is predicted to increase employment of South Africans by between 0.003% and 0.01%. This result, while unsatisfying to an economist, is relevant politically, in that foreigners do not harm the average employment chances of locals – at least according to the census data.

However, the results become significantly more interesting when we break the census sample into job categories. My models predict that immigrants have an unambiguously positive employment effect on locals in elementary categories – these include factory workers, cleaners and manual labourers. A 1% increase in the relative immigrant population is predicted to increase employment of South Africans by 0.04% in this category.

This effect is difficult to explain but may indicate the foreigners are creating menial jobs where they settle.

This supports the claim that immigrants to South Africa are generally highly skilled, in that they appear to be competing more with workers in skilled job categories.

What to make of these results

Based on the evidence we have available, we can conclude with reasonable confidence that immigrants do not take jobs from South Africans overall – in fact, a best-case scenario suggests that they are creating a small number of jobs where they settle. However, when we hone in on particular job categories, we do see a small, negative employment effect for workers in better-skilled job categories. If we believe that immigrants are systematically under-represented in the census data, this may reflect a larger negative effect in reality.In any case, the predicted effects on both wages and employment, taken at face value, are extremely small. Further research into the skill profile of immigrants is needed to determine whether immigration is a potential solution to South Africa’s chronic shortage of skilled labour, as Ellis and Segatti (URL) suggest. It would also be interesting to explore how employers tap into immigrant networks within the country to source labour, as well and how immigrants in South Africa network with family and associates elsewhere in the world.

---

Published originally on GroundUp.

---

© 2017 GroundUp. This article is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.