A report of the BBC News Africa website describes the horrific burning to death of several suspected witches in rural western Kenya.
The prevalence of the belief in witchcraft in Sub-Saharan Africa hasn't received much attention in the development econ literature (it, of course is covered by other disciplines). Mark Koyama at the Oxonomics blog has a great post on the history of witch killings, as well as very convincing argument about the reasons behind the violence we see in modern SSA:
My view is that 'witches' comprised individuals from two groups:Mark's second point is related to the work of Edward Miguel on poverty and witch killings (PDF version of his 2004 paper here) which gave compelling evidence from rural Tanzania of the danger of being old and female during the lean season: retrospective data from villages revealed that witch killings were more likely to occur when food ability was low (in the lean season and during times of crop failure) and that the victims were overwhelmingly old women.
- Those individuals who refused to cooperate or failed to contribute their share in some communal risk-sharing activity.
- Those individuals who because they were old, infirm, widowed or without kin, were dependent on the support of others.
Miguel's reasoning is convincing: those with the least amount of social power and income-earning capacity are more likely to be violently ejected from the household. Yet I'm more concerned with Mark's first idea: that witch killings (or just accusations of witch craft) are a response to the reneging on some social contract, often a risk-sharing or income-pooling agreement.
When I lived in Malawi, the papers were full of witchcraft stories and accusations. The conventional wisdom that those that were accused were often those that were seen as the successful, the tallest of the poppys, that refuse to share their success with others. The standard risk-sharing network models predict that those that are extremely successful are more likely to break away from their social contracts. These models often posit that risk-sharing networks punish those that go back on their obligations, and it is reasonable that an accusation of witchcraft would be an effective punishment.
How could we test this? We'd need data on witchcraft activity and perhaps on rainfall data (risk-sharing networks are more likely to fall apart when there is a community level shock, usually through rainfall). The latter is easy to measure, but not the former.
At this point I'll delegate to a more nuanced discussion of witchcraft.
Other links: