Predicting xenophobic attitudes: Statistical path models of objective and subjective factors

In September 2019 a wave of xenophobic violence hit Gauteng, reportedly leaving more than ten people dead and over a thousand homeless.

Specific outbursts of xenophobic violence are difficult to anticipate, but they are fuelled by underlying xenophobic attitudes that are more enduring and widespread. This Provocation investigates possible predictors of these xenophobic attitudes, using statistical analysis of data from GCRO’s Quality of Life survey. If these predictors can be identified, they can be targeted by policy interventions and thereby the likelihood of violent outbreaks could be reduced.

Three previous analyses, published in the social-scientific literature, were bemused to find that respondents’ objective background conditions – such as unemployment, poverty, lesser education or residence in an informal settlement – appeared variously not to be correlated with xenophobic attitudes when examined in multivariable regressions. The reason, as explained in this Provocation, is that a deeper statistical analysis is required. Statistical path models reveal that objective factors are indeed at work, but take effect via mediating subjective factors such as depression and dissatisfaction.

The implication of the analysis in this Provocation may be construed as potentially optimistic, in two respects. On the one hand, it implies that what needs primarily to be addressed to mitigate xenophobic attitudes are the underlying circumstances of economic disadvantage (notably unemployment, hunger and housing). On the other hand, potentially more intractable identity issues such as race and political alienation – which feature large in, for example, the mobilising discourses of some political parties – would seem to be less salient than the objective factors as causes of xenophobic attitudes when results from a large population survey are adequately analysed.


The GCRO sends out regular news to update subscribers on our research and events.