In an article in Prospect magazine William Easterly claims that "the most persistent myth in economic development" is that "autocratic governments create growth miracles".
This is far from clear to me as there are plenty of other candidates for most persistent myth about economic development, but what is clear is that Easterly's evidence does little to refute the argument that autocrats are good for growth. The regression work he describes seems to explore associations between changes in leaders and growth spells. He finds that changing leaders is barely associated with growth. Leaders don't matter, he suggests.
Now this result could hold because leaders don't try to change anything, or they try but fail, or the things they do change don't matter for growth - either because they are not interested in growth or because they back the wrong growth driver. The possibility of the last explanation is important because it suggests leaders could matter in the future even if we think they have not in the past - they just need to do different things.
So we should be much more interested in estimating the association between growth and the things leaders can control --either individually, or collectively over time--such as the extent of repression. Even better, we should estimate the associations between the dynamics of freedom and growth to test Easterly's very plausible assertion that it is a relaxing of repression -- whether within a leader spell or across spells -- that leads to an increase in growth rates.
When assembling evidence it is important to avoid replacing one myth (autocrats are good for growth) with another (leaders don't matter for growth).