Appeals to authority are a substitute for argument, reasoning and evidence. One such appeal, frequently offered by social (especially economic) modellers, is to Occam's Razor - the aphorism to the effect that a simpler model is a better model. There is, of course,a constraint on simplicity. Either, as Einstein is said to have had it, a theory should be as simple as possible but not simpler or, as Occam (William of Ockham, the 14th century scholastic who does not actually appear to have written the phrase) is claimed to have written, entities should not be multiplied beyond necessity, there is a notion of a limit on how simple a theory should be.
There is no doubt that, in the physical sciences, the principle of Occam's razor has been enormously productive and was a key aesthetic consideration in Einstein's development of relativity theory, for example. Matters could hardly be more different in the social sciences.
There is here an important issue of epistemology. The objective of physical scientists to predict physical outcomes of specified events has been shown to be appropriate because it has been broadly successful. The failure of the social scientists systematically to predict social outcomes of specified events has never led to a consideration of why that has been the case but rather to the elaboration of failed theory - frequently in violation of the principle of Occam's razor. (Edmonds, 2007) In general, social scientists seek to exploit existing theory rather than explore for a better theory and, indeed, a better epistemology.
Now a tendency towards exploitation of what is known rather than an exploration of what is not already well known is a common feature of intellectual advance and social development. It certainly characterises the history of technological (Rosenberg, 1976) and organisational (Chandler, 1977) change and business strategy (Penrose, 1959). However, as exploitation becomes less beneficial and increasingly expensive, progress takes the form of exploration - trying new ways to achieve goals or perhaps new ways of using prevailing knowledge, experience and physical or technological resources. It would seem hard to justify the proposition that, as a result of continued exploitation, social theory has been getting either simpler or better at informing social policy or predicting the course or outcomes of social processes.
So far, I have taken the meaning of the words ``simple'' and ``simpler'' in their vernacular senses of being uncomplicated and less complicated, respectively. Many social modellers have what amounts to another meaning of simplicity as a justification for relying on assumptions that distort the reality of the social phenomena they seek to capture with their models. A search of almost any issue of any leading economics journal will turn up at least one assumption made ``for the sake of simplicity''. Typically, this means that a straightforward formalisation of observed social relationships would make the application of some economic theory or conventional modelling technique either infeasible or intractable. Assumptions of linearity and normality are common ``simplifications''. However, other examples are not hard to find.
To distort the description of a problem or assume it has unobservable properties when those assumptions do not lead to empirically validated predictions or validated specifications of social processes should reasonably render models or theories based on such assumptions inapplicable to policy analysis or social explanation. Such ``simplifying'' assumptions are such patently bad science that this form of simplicity and simplification will not be considered further here.