He wrote a very interesting book, which Harvard published in 1984. It contains an extensive theoretical investigation of an open model with at least three possible closures. Here is a comment on some of that theory:
"There are, in short, two marginal-productivity theories. There is first of all a general theory that assumes nothing more controversial than continuous substitution, constant returns to scale, and competitive profit maximization. This theory, far from being unique to one school of thought, is - or at least ought to be - the common property of all three approaches ['neo-Keynesian', 'neo-Marxian', and 'neoclassical']. The second marginal-productivity theory adds the assumption that sooner or later, in the long run if not in the short, the wage rate clears the labor market and ensures full employment. This version of marginal productivity theory, the special theory, is manifestly the property of the neoclassical model. Neo-Marxians and neo-Keynesians wouldn't touch it with a ten-foot pole." -- Stephen A. Marglin (1984). Growth, Distribution, and Prices, Harvard University Press: 223.Marglin's "Neo-Keynesian" model contains the theory of distribution developed by Richard Kahn, Nicholas Kaldor, Luigi Pasinetti, and Joan Robinson in the late 1950s and early 1960s. Marglin's neoclassical model has intertemporal utility-maximization in an overlapping generations setting, but doesn't imply Marshall's principle of substitution or that prices are scarcity indices. That is, while logically consistent, it lacks the features that make the model intuitive and easy to apply.
Marglin looks at both long-run and short-run version of the models. Here he offers a judgement on the neoclassical model:
"Observe that even in neoclassical theory full employment alone is not enough to transform marginal-productivity relationships into a long-run theory of distribution. In long-run neoclassical theory, the capital:labor ratio is endogenously determined, so that the wage rate cannot be determined solely by marginal productivity of labor at full employment - not even in Chicago. Instead, distribution must reflect household preferences with respect to present and future consumption.
Thus, it is fair to conclude that there are two marginal-productivity theories. The first is a relatively innocuous, general theory that involves nothing more controversial than competitive profit maximization - and provides correspondingly little contribution to the theory of growth and distribution under capitalism. The second is more powerful, and very special, providing by itself a theory of distribution, for the short run at least, whose 'only' defects are (1) that it assumes full employment and (2) that it begs the question of accumulation. The wonder is that it is precisely this theory that so many students come away with from their study of economics. Only slightly more wondrous is that by and large they believe it!" - Stephen Marglin, ibid: 330-331
Marglin reports on some econometric evidence. Despite the differences in mechanisms driving value and distribution in the models and the difference in policy implications, it is sometimes difficult, Marglin notes, to draw different implications on the direction of effects in certain empirical applications, such as a shock to oil prices. He does find some contrasting implications when it comes to pensions. And he finds the neoclassical model does the worst, although he thinks his empirical findings are not decisive.
A couple of years ago, some Harvard students wanted to be seriously taught introductory economics, instead of the ideological propaganda that Marty Feldstein and now Greg Mankiw apparently teach. And they wanted to be able to take this course for credit. They asked Marglin to teach this revised Ec 10 course. I'm not sure how much Marglin planned to go into alternatives, or whether he would mention his empirical results. I do know that Harvard decided (again) no way would they permit future leaders of the United States to get an education about the sources of income and property.
Update: Originally posted 5 September 2006. The update is to highlight Ekaterina's comment.