During Apartheid, black South Africans were not allowed to permanently relocate to urban areas, but instead were forced to retain ties to rural ‘homelands’. These ties led to the development of oscillating or circular trends of rural-urban migration. Following the breakdown of mobility restrictions, homeland displacement, and other discriminatory laws that governed South Africa prior to 1994, these oscillating trends of rural-urban migration were expected to dissipate. However, following democracy, these patterns of circular migration have persisted, alongside trends of increased urbanisation and feminisation of the labour force. [1]

Rural-urban migration is often associated with improved access to healthcare, living standards, education levels, and favourable employment opportunities. However, the migration process may also present new physical health risks. These risks may be introduced through the failure to attain either urban employment or adequate housing.[2]

Methodology

The following blog reflects on the impact of rural-urban migration on the physical health of the sampled[3] South African migrant group. A Difference-in-Differences (DiD) approach was employed, in tandem with Propensity Score Matching (PSM).

The DiD model allows for the creation of a control and treatment group. The treatment group in this instance refers to individuals who have engaged in rural-urban migration. On the other hand, the control group refers to people who have remained in a rural area consistently over the timeframe of the study. An assumption of the DiD model is that the health outcome of the two groups trend in a parallel fashion prior to the intervention or migration efforts (see Figure 1a). The impact of the intervention or Treatment Effect (TE) is the variable of interest in this form of analysis. The TE is estimated by measuring the difference between realised health outcomes and projected health outcomes in the periods following the intervention.

The DiD model may be combined with PSM. PSM is a technique that allows matching between treatment and control individuals based on observed characteristics. In this way, one is better able to compare the health outcomes of like individuals following migration. Put differently, PSM ensures that the analysis is done for individuals who have similar demographic and socio-economic characteristics apart from their rural-urban migration status. This ensures that the analysis compares apples to apples. Significant overlap in the observed characteristics of the control and treatment groups is a key assumption of the PSM technique (see Figure 1b).[4]

Figure 1: DiD and PSM Assumptions

    (a)    (b)

Results

The analysis shows that reported physical health outcomes are adversely affected by rural-urban migration efforts, given the significant and negative TE or Difference-in-Differences variable of interest.[5] This result is consistent with an inverse relationship between migration and health found in various literature.[6]

Table 1: Subjective Physical Health

Significance: ***p<001, **p<0,05, *p<0,1

Findings

Based on this and further descriptive analysis, the following preliminary findings can be drawn:

• Rapid urbanisation could give rise to poor living conditions for the migrating cohort. A descriptive analysis of this sample shows that roughly 10 percent of rural-urban migrants go from living in a formal housing to experiencing some form of informal displacement following migration. Moreover, the importance of formal housing is confirmed by the significant and positive association between this variable and reported physical health.
• Migration is associated with an increased appetite for reckless behaviour (such as cigarette and alcohol consumption).[7] However, a descriptive analysis of this sample shows that demand for these goods decreases in the wave following rural-urban migration.
• The adoption of a Western Diet is often associated with rural-urban migration and this can have an adverse impact on physical health outcomes. There is evidence that suggests a “Western Diet” are likely to adopted by migrants[8]. This is usually linked to an increase in an individual’s real income and improved labour market prospects experienced by the migrating group in the period following their migration efforts. As income increases, so too does one’s ability to afford western diets, which includes more fat, sugar, and meat and less fruits and vegetables.

These results suggest that rural-urban migration in South Africa has had an adverse impact on reported health outcomes, in this sample. Such evidence should be used to better inform urban planning policies, and the provision of health care services in particular. Further research is required to assess whether these findings can be generalised. It is therefore alarming that migration studies remain somewhat neglected and that the impact of migration is largely under-investigated in the South African literature.

[1] (Wilson, 2001; Smith, 2003; Posel, 2004; Gong, et al., 2012; Mulcahy & Kollamparambil, 2016)

[2] (Smith, 2003; Gong, et al., 2012)

[3] Unweighted NIDS data sample

[4] (Angrist & Pischke, 2008)

[5] Behavioural, economic, demographic, and other factors are included in the model for the purpose of refining the accuracy of the estimated Difference in Differences variable (reduce the error variance) and for the further identification of the Treatment Effect. These are time-varying factors that affect the outcome but are not related to the intervention. Omission of which could potentially violate the parallel trend assumption. These are not to be interpreted directly. (Stack Exchange, 2015)

[6] (Lu Y. , 2010; Lu & Qin, 2014; Mulcahy & Kollamparambil, 2016)

[7] (Kristiansen, Mygrind, & Krasnik, 2007)

[8] (Salant & Lauderdale, 2003)