Much development economics research is driven by econometrics—the statistical analysis of data sets to identify causal relationships. The debate on poverty’s relationship to economic growth has been led by this methodology, as has the aid effectiveness debate (with the protagonists engaged in close combat over such arcane, but vital, issues as ‘outliers’ in the data).
Now don’t get us wrong, we’ve been known to practice this art ourselves, but sometimes it reminds us of Shakespeare’s witches in Macbeth, gathered around their pot at midnight, happily tossing in more frogs and toads (otherwise known as ‘data mining’).
So innocent policymakers beware, don’t believe everything you read. That’s the message from a policy brief by Francisco Rodriguez from the International Poverty Centre (one of our favourite sites for poverty research). Correlation is not causation, sample sizes are often small, and statistical traps abound. You have been warned. Must get back to that pot to give it another stir.