Some argue that bad outcomes are to be explained by reference to signalling, rather than human incompetence. On this view, people under-perform not because they couldn't do better, but because they weren't rewarded for good performance by their peers.
But as it stands, this explanation is incomplete. It raises the further question: why are the peers impressed with poor performance, and why don't they reward good performance?
The main answer is that they are incompetent. So the signalling explanation of bad outcomes, far from ruling out human incompetence, rather relies on it.
I think much/most of the time, the explanation is that 'performance' is hard to estimate directly (or at least hard to estimate directly at the right time), regardless of the competence of the judges.
The lack of access to counterfactuals in most scenarios is likely a large barrier a lot of the time: i.e. we often don't know how things would be going if someone else were to have done something differently. If someone successfully signals competence, but things seem to be going badly, we might reasonably conclude that things couldn't have gone better.
Also while I quite like the 'quality of the audience' explanation for under-performance, the fact that the audiences have signalling games of their own to engage in seems like it may be a major problem. I worry that often elites might know better (than to accept a given idea or policy, let's say), but accept it anyway due to signalling considerations. At least one example of this would be business leaders knowing that they should do x, but not being able to x without bringing in outside consultants to say they should do x.
In such cases, it might be more important to change the incentives for elites more than to raise their competence. For example: when they vote (or deliberate), make their votes blind; give them more opportunities to pass the buck of decision-making to other independent/technocratic bodies or committees and build a norm encouraging them to do this. (Of course, in public policy such moves may be seen as undemocratic and anti-transparent and so unpopular). It's interesting to note that whether you want more or less transparency will depend quite heavily on your views on this competence/signalling explanation and who you think the audience would be in cases of transparency (if you think it would be a less competent audience, presumably you want less transparency).
Note that this is incompetence of the peers or third parties judging the action, not incompetence of the person undertaking the action. And yes, limited observability and other factors may affect third parties' judgements. But still, truly competent third parties would incentivise good performance. So human incompetence is still a key cause of bad outcomes, even if the signalling explanation is true.