Does ethics pay?
One way to operationalize the question: Analyze all of the 144 possible matrices representing ordinal utility in one-shot 2 x 2 games.
If the analysis in the attached powerpoint is correct, a "social" player who plays the highest joint value outcome for the two players narrowly outperforms a "selfish" or "non-social" player who plays a dominant strategy where applicable and mixed Nash otherwise. The edge of the social player in S-S interactions over the non-social player in N-N interactions narrowly overcomes the edge of the non-social player in S-N interactions.
A more fine-grained breakdown: The significant edge of the N player in the 15 PD matrices is narrowly overcome by the considerable edge of the S player in the 9 Stag Hunt and 8 Battle of the Sexes matrices, along with a tiny edge for the S player in the 17 Chicken matrices. There are plenty of nuances--two of them are that the edge of the S player is dependent on the S player and not the N player being able to employ Schelling focal points in 4 Battle of the Sexes matrices, and that the edge of the S player is substantial in the 7 symmetrical PD, Chicken, Stag Hunt, and Battle of the Sexes matrices.