If the native or raw ability to acquire wealth is distributed randomly, we ought to see a bell curve of abilities, whether of individuals or groups. However, in a dynamic system, where the ability to acquire wealth is positively enhanced by wealth already acquired, it is a truism that them that has, gets, so that we can expect any ability to acquire wealth to be quickly amplified by its own effects unless the effect of existing wealth on wealth acquisition is vanishingly small, which does not seem to be the case. Again, this effect would apply to groups (families, communities, classes) as well as individuals and would grow exponentially over time.
If all this is so, then in any example of a system of this type, we should soon see a very small number of individuals or groups having almost all the wealth, while the remainder were entirely or almost entirely deprived of it. This is what Marx seems to forsee in the _Manifesto_. But polemic to the contrary notwithstanding, this is not what we observe; if it were so, most people would now be starving to death, because food is a form of wealth and the ability to acquire it is another. Why?
I have been trying to think of reasons why we do not observe the radical concentration of wealth I provisionally hypothesize above.
1. The distribution of wealth-acquiring ability, or the aquisition-enhancement quality of existing wealth, may be vanishingly small, contrary to appearances.
2. It is possible that wealth is not a substance, that is, a well-defined material entity. Instead, it might be a mentation, a social construction or relationship, a complex in which material entities play only a part, a fiction, or something else -- something immaterial.
3. The influence of random events may disturb the system so much that the expected behaviors do not occur (or cannot be separate from noise).
4. Economic processes may be frequently, even customarily, overruled by other social processes, like politics.
But some of these cast a certain amount of doubt on the extent to which a physics-like approach is appropriate to the question. Perhaps one needs to define the domain over which the approach is valid.
Well, I said it was a dumb question.