Do immigrants have better labour market outcomes than South Africans?

ABSTRACT

We use data from the ten percent sample of the 2011 Census to explore labour market outcomes among immigrants and locals in South Africa. We show that naturalised immigrants and foreigners engage more successfully with the labour market than locals on average. The ability to access social networks improves labour market access for immigrants, but sequentially controlling for observable characteristics, including networks and location, decreases immigrants’ participation and employment advantage over locals. The conditional immigrant earnings gap is negative, but because immigrants typically work in low quality jobs, their relative earnings disadvantage is entirely explained by differences in workers’ occupations and industries. Our attempts to control for the possible endogeneity of immigrant status suggest that the direction of selection bias may be different for naturalised and foreign immigrants in South Africa, reinforcing the importance of distinguishing between different immigrant groups in research of this nature.

KEYWORDS:

• Immigration
• labour market participation
• employment
• earnings
• South Africa

JEL CLASSIFICATION:

• J21
• J61

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 All respondents are asked “What is the income category that best describes the gross monthly or annual income of (name) before deductions and including all sources of income?” Twelve possible response categories are provided, ranging from no income to R204 801 or more per month.

2 The main exception is the sparsely populated Mier local municipality in the far Northern Cape, which borders on Namibia and has a higher proportion of naturalised immigrants than foreigners.

3 In 2001 prices.

4 We control for gender throughout our analysis but do not present our findings separately by gender. The pattern of the results across labour market outcomes and model specifications is identical for both men and women, although the magnitudes of the coefficients differ somewhat by gender. In general, social networks play a slightly larger role for women than for men.

5 We thank an anonymous reviewer for the suggestion to refine the network quantity measure for locals using information on province of birth. We use data on approximately 233 municipalities and 176 reported birth countries.

6 Following Bertrand et al (2000), we do not calculate $\overline{EMPL}$ at the municipal level. Doing so may introduce omitted variable bias if the individual has unobserved characteristics in common with others from the same birth region living in the same municipality.

7 On average, among naturalised immigrants, 1.7 percent of other working age residents in their municipality were born in their same birth country, while for foreigners this figure is 2.7 percent. An immigrant’s network size is significantly negatively correlated with the length of time since they immigrated to SA, suggesting that new immigrants in particular locate themselves in close proximity to others from their birth country.

8 Note that this variable takes on a value of zero for locals.

9 Table A1 in the Appendix includes the results for the full set of control variables for the models in Table 4. The full results for other models may be requested from the authors.

10 Specifically, we instrument for the two categories of immigrant status, social network access for the two immigrant groups, and number of years in South Africa and its square.

11 We extend our thanks to an anonymous reviewer for the suggestion to instrument for immigrant status using macroeconomic indicators.

12 For each of the GDP per capita and POLITY2 measures we use an average calculated over the 3 years prior to immigration for immigrants. For locals, we use GDP per capita and the POLITY2 index in the year the individual turned 16 years old. Our distance measure is coded as zero for locals.

13 Probit estimates for participation and employment for the migrant sub-sample similarly show smaller penalties to immigrant status and smaller network effects, compared to the full sample. These results may be requested from the authors.

14 The Hansen-J test results suggest that the validity of the instruments is similarly doubtful whether we use macro-level instruments only, individual-level instruments only, or instruments at both the macroeconomic and individual levels. However, the results of the IV estimation are more stable when we use the combined set of instruments, and this is therefore our preferred set of estimates.