“…The first reason is the observation that sharing or compensation principalagent agreements in reality regularly take a linear form, e.g. see Chao (1983) on sharecropping; Lafontaine (1992) or Sen (1993) on franchising; Caves et al (1983) on licensing; Masten and Snyder (1993) on equipment leasing, among others. 9 The second reason concerns complexity-given that our paper is about modeling learning about the best compensation contract, the linearity restriction simplifies the principal's problem to learning about two numbers, s and f, as opposed to learning about a general non-linear function c(y), possibly without an analytic closed form, which could make the setting much harder to adapt in.…”