Randomized trials and corruption

There is a debate in the analysis of poverty between those in favour of small incremental reforms based on randomized trials and those in favour of big non-marginal reforms that change the structure of society. I believe that the latter is the spirit of the critique of Martin Ravaillon to the work of Duflo and Banerjee. Of course, small reforms are welcome, but there is always a question about their external validity and there are doubts that they can lend support to any particular theory that may help change things at a larger scale. Perhaps something similar can be said about the reform of politics and the reduction of corruption. Actually the same Banerjee and other authors have some articles that I am reviewing where they introduce similar randomized trials, not only to reduce corruption but also to improve politics by reducing patronage and ethnic-based appeals. Again, all these small reforms would be welcome, but given that corruption reforms depend on the solution to a collective action problem (people only tend to act cleanly if others do so), it seems difficult that one can fix the problem without sweeping reform and a societal ethical commitment to fight it. Many well intentioned reform proposals on corruption based on incentive theory have failed. There are positive experiences with large scale corruption reform, such as those in many US states in the XXth century or those in Sweeden in the XIXth century. These reforms were not brief nor simple. In the case of the US, it is likely that corruption did not disappear, but it mutated into something different: from machine politics to the revolving doors, summarizing in the smallest possible number of words something that took decades and different mechanisms in different places. I am not sure that corruption and patronage can be fixed by appointing an expert committee that concludes that the main problem is in political parties, when political parties are needed to fix de problem. Corruption is a political and economic problem that affects the structure of society, and that is linked to inequality and to excessive political power of big economic interests. Like with poverty, it is hard to imagine that corruption is going to be fixed one experiment at a time.