The hegemonic neoclassical theory of wages involves attributing an individual receiving relatively high wages to his or her "human capital" (education, training, inherent IQ, etc.) Wages may also be high due to an individual's job having tremendous amounts of responsibility. Some workers -- e.g., star athletes -- gain scarcity rents because of the uniqueness and nonreproducibility of their talents. In addition, a wage may be high because his or her job is unsafe (coal miners) or low because of non-pecuniary benefits (like tenure for professors), following Smith's theory of compensating wage differentials. But the main emphasis is on individual characteristics, with the neoclassical vision centering on two allegedly exogenous variables, individual tastes and technology. Market imperfections -- like labor unions, in this vision -- make the analysis more complicated, but don't change much. (See almost every labor econ. textbook.)
The alternatives to this theory do not deny the role of individuals, but put a much larger emphasis on the demand side of the market, the managerial organization of corporations, and the role of unemployment. Your wage doesn't simply depend on your individual skills, etc., but also what industry or company you work for. Secretaries in high-wage industries get paid more than those in low-wage industries even though the skills involved in their jobs are very comparable. Some authors point to segmented labor markets, or what Cairnes called "non-competing groups," to explain wage differentials. Others point to the roles of racial and gender discrimination.
I could go further in the description of the alternatives, but won't do so. The point of "unobserved skill" for the hardcore neoclassicals is (1) to ignore the alternative theory; and (2) when doing econometrics, call the unexplained residual "unobserved skill." But since they ignore the alternative theory, they are biasing their empirical work in favor of their own theory, "proving" what they wanted to prove in the first place, i.e., that it's all a matter of individual characteristics.
BTW, Rakesh, if you don't know much about econometrics, how can you praise Galbraith's CREATED UNEQUAL? The book seems very empiricist, using more statistical techniques than theory. Or maybe I misremember who it was who praised that book on lbo-talk.
Jim Devine jdevine at popmail.lmu.edu & http://clawww.lmu.edu/Departments/ECON/jdevine.html