I appreciate authors’ efforts to improve the paper, especially the analysis of the supplier selection.
The core of the paper is simple. From a system’s perspective, the shared structure enjoys two advantages: 1) the cost reduction because of production consolidation, instead of ; 2) a monopolist supplier instead of two competing supplier, which further improves system efficiency (the monopolist’s centralized decision dominates the decentralized ones by two independent suppliers; he further differentiates qualities to soften the competition). Therefore, the shared structure (suppliers + retailers) is more efficient and profitable than the independent structure. In particular, we have property (P): the quality and profit gaps between two retailers are larger in the shared structure than in the independent structure.
The main flaw of the paper is in its attempt to use the property (P) as the main rationale to explain the retailers’ channel structure choice. But the channel structure is far more complicated than the simple property (P) can explain. The reason is that (P) is derived under the exogenous structure. but endogenous structure choice requires more than the quality profit gap comparison: instead of playing a given game, the structure choice is a question of how to design the game (of two choices).
The property (P) proves to be insufficient for explaining the structure choice. Indeed, the authors have to resort to ad hoc elements to patch up the story: last round is searching cost, and this round it shifts to reputation parameter . I am fine with such ad hoc approach, if it can deliver a compelling argument. But this is not the case. In both attempts, the ad hoc elements come only as a means to rescue the main logic (P).
Here lies my problem: if the basic model cannot answer the main question the authors ask, and has to rely on the ad hoc fore later on, then what is the point to have that model in the first place? If the ad hoc element is indeed the driving force, then it should be build into the main model at the outset.
Also, after two rounds revisions, the analysis is still sloppy, containing technical errors (see the detailed comments below). This casts doubts and further compromises the validity of the conclusions.
The bottom line is, I struggle to find a conclusion that is sufficiently solid and compelling for publication.