I'm sorry but the following is mathematically provable: If (a) you have a perfect screen and then (b) your screen for what stocks to buy excludes certain companies b/c they are 'evil' or meet some other criterion, then your maximum return can only decline. (Your return from the narrower universe of stocks may be the same as the return from the wider universe of stocks if the 'evil' stocks happen to fail your screen - but otherwise the maximum return from the narrower universe will be lower.)
That said, the only possible conclusion is that fund managers in the ESG field have better screens. (One might posit that the ESG field attracts smarter managers who build better screens, but, attractive as that idea is, I don't see any data supporting it.)
Comments
Pragmatically, there is almost surely no "screen" that is perfect, i.e. no set of filters that has no false positives or negatives. Rather, you are hypothesizing an oracle, an enumerator of stocks independent of criteria. Given an oracle, then applying any screen, ESG or other, would indeed produce only false negatives.
But if we're hypothesizing an oracle, or as you put it, a perfect screen, then why would it ever produce more than a single security? After all, by the same logic, even a second security would reduce the return.