Here is a passage from a new paper from Lant Pritchett and Eric Werker:
One of the puzzles to be reconciled is that although “institutions” are associated with long run growth rates, their predictive power for short-to medium-run (5 to 10 years) growth, or for growth accelerations is very weak. (Khan, 2007) makes the distinction between “market supporting governance” versus “growth promoting governance.” He points out that, while “governance” measures are correlated with levels of GDP per capita, there is little or no predictive power of the current level of institutions for future growth rates.We illustrate this by comparing the level of income and a measure of “governance” versus the growth rate and that same measure of governance and the growth rate and the change in the measure of governance. Whereas the first is strong the second is quite weak and the third near zero.…Essentially all rich countries have reasonably high “quality of government” and all countries with high “quality of government” are rich.…But among countries with the same governance there are massive differences in growth (e.g. China versus Cote d’Ivoire) and countries with rapid growth (above .6) the ‘quality of government’ ranking ranges from .15 (Indonesia) to .9 (Singapore).
…Figure 7c shows there is no link at all between the improvement in ‘quality of government’ and economic growth 1984 to 2004. A country like Uganda has massive improvement but exactly average growth, China has massive growth and no improvement at all, Malaysia saw QOG worsen but growth well above average, etc.
The relevant figures start at around p.30. Was it Jeff Sachs who put it this way?: Go back to 1960 and try to measure the quality of institutions any way you wish, knowing of course in advance which countries end up doing well. Can you find any measure at all which predicts subsequent growth? This is a tough problem for we economists.
There is another sense in which institutions have to be the causal factor, in which case we are very far from having a sense of how to measure good institutions.