What Do the Roma in Italy and Commercial Sex Workers in Kenya Have in Common? Reversing Conditional Probabilities and its Consequences
By Fiorenzo Conte
The discussion about sensitive issues regarding marginalized groups is often misinformed and misled by a common and persistent cognitive bias known as Reversing Conditional Probabilities. Bill Easterly repeatedly insisted on the role that this cognitive bias plays in fuelling ethnic and religious stereotypes. Recently, I came across two new examples of this bias. In what follows, I illustrate these examples and shed light on how these cognitive mistakes feed into misguided public policies.
There is a heated debate in Italy about the chances to integrate the Roma population ( an ethnic group originally from the India sub-continent which started to arrive in Europe at the beginning of the 14th century) in Italian society. One of the challenges that stands out is the inability – and unwillingness – of the Italian government to offer stable and decent housing solutions that could replace the temporary camps where they are currently living. One of the main arguments used against the provision of housing for the Roma is that they are “nomads” and do not reside in the same place for a continuous period of time. The general population and also part of the political class in Italy believes that the chances that IF you are nomad THEN you are a Roma is high. Unfortunately, the relevant probabilities ( in terms of defining appropriate policies for the integration of Roma in Italian society) is the chance that IF you are a Roma THEN you are a nomad. And this second probability is far lower that the first one insofar as the majority of the Roma defines themselves as sedentary. In fact the majority of the Roma population are Italian citizens (60 %) and almost all of the Italian Roma are sedentary.
The second example is drawn from my experience working in Kenya. In designing a questionnaire to find out more about practices and behavior amongst commercial sex workers, some of my colleagues proposed to include amongst the background characteristics the category: tribe. The reason being that tribes play a role in the allocation of resources in Kenya so it could be that for example some tribes have been marginalized by the government, hence their livelihood strategies are limited, hence women are forced to enter prostitution. By including tribe in the survey you can get what is the chance that IF a woman is commercial sex worker THEN she belongs to the tribe X ( as opposed to tribe Y or Z). Unfortunately, this is not what you are looking for. If you want to understand whether belonging to a specific tribe increases the likelihood of practicing commercial sex the relevant probability is however the chance the IF the woman belongs to tribe X ( instead of Y or Z) THEN she is a prostitute. Reversing conditional probabilities would lead to draw wrong conclusions about the power of the tribe to predict the likelihood of becoming a commercial sex worker.
Both cognitive biases have a real impact in the real world insofar as they are used by governments to hedge difficult political decisions, as in the first case, or to design mistargeted interventions, as in the second case. To identify these mistakes is not only an intellectual exercise but also necessary to inform our practices.