Great piece.
Two minor comments:
1. You argue against "market fetishism" yet use some attributes of that fetishism, namely incomprehensible jargon. What does "prominence of M&A in the stock landscape" mean? How about "The University of California system, whose endowment is run on indexing principles by a small staff of civil servants, was long the stock and lost a bundle. Harvard, whose endowment is run by a large staff of professionals paid on a Wall Street scale, made a bundle through shorting?"
As C Wright Mills aptly observed, a successful demystification involves a "translation" of that jargon into plain English - which he quite effectively did with the notoriously obtuse writings of Talcott Parsons. One would thus expect the same from a piece written for general audience.
2. You write: "UBS recently issued a research note that gathered statistics from seven rich countries on the size (relative to GDP) and investment allocations of pension funds. When combined with World Bank data on savings and investment rates some interesting results emerge. A ranking of the seven countries (Australia, Japan, Netherlands, Sweden, Switzerland, the UK, and the U.S.) by pension fund size shows an unimpressive correlation of .35 with national savings rates, and just .03 with gross fixed capital formation. When ranked by the stockholdings of pension funds, the signs change: -.89 for savings and -.46 for capital formation: that is, the bigger the wad of pension money committed to the stock market, the lower the level of national savings and investment. The sample size is small, for sure, but it does make the economic case for prefunded pensions a little harder to prove."
Technically speaking, this is not a sample but a "population," unless we take a difficult to maintain position that the seven richest countries "represent" a larger population of countries. Therefore, the correlations you mention are the actual correlations as found in the population, not encumbered by a sampling error.
To make a long story short - sampling error reflects the possibility that the observed in the sample relationship deviates from that of the population due to atypical composition of the sample. Statistical significance in that case would reflect the probability that the said correlation is different from zero due to sampling error (i.e. atypical composition of the sample vis a vis the population) - and is a function of two factors: the magnitude of the correlation itself and the sample size. The smaller the sample size, the greater the value of correlation is needed for the result to be "statistically significant" which translated into plain English means "the probability that the correlation of the size we observed in the sample is due to atypical composition of the sample is very low, i.e. less than 5%" (i.e. if we drew 100 different samples of the same size from that population, we would get the correlation of that magnitude only in 5 such samples).
However, when we are dealing with a population rather than a sample - the concept of sampling error is meaningless - it simply does not apply. Consequently, the observed correlation, no matter how small, is the actual correlation not an estimate encumbered by a sampling error. This means that the caveat "the sample is small" is unnecessary and dilutes the strength of your argument: the figures you cite positively show a negative correlation between stockholding of pension funds and savings/capital formation in the set of countries under the scrutiny.
Wojtek