Both academics and practitioners agree that the
mean-variance analysis is extremely sensitive to small
changes and errors in the assumptions. We therefore take
another approach to the asset allocation problem, in which
we estimate the weights of the asset classes in the market
portfolio. The composition of the observed market portfolio embodies the aggregate return, risk, and correlation
expectations of all market participants and is by definition
the optimal portfolio.
I think “by definition” is wrong here, but there are strong theoretical reasons to think that the global market portfolio is the optimal portfolio for all investors (investors with different risk tolerance should differ only in how much exposure or leverage they use). However that theory makes some unrealistic assumptions, and here’s a fuller picture, from Q&A: Seeking the Optimal Country Weighting Scheme :
Financial theory suggests that a global value-weight market portfolio is the logical default position for an equity investor seeking the optimal allocation scheme across countries.
What are the implications for this approach if we take structural factors into account that encourage a home bias? Australia, for example, offers tax incentives applicable only to local investors, so their citizens earn higher returns than foreign investors do holding the same stocks. Brazil accomplishes the same thing by imposing additional taxes on foreign investors.
[...]
One can argue that asset pricing theory always makes unrealistic assumptions, which are irrelevant if a model does a good job describing observed average returns. True, if one is only concerned with describing average returns. But this argument does not imply that a simplified model can be used as a prescription for optimal portfolios. Here one must face the implications of real world frictions in international investment.
One might argue that the effects of all frictions are captured by the aggregate portfolio of local and foreign assets held by the investors of a country. This is, in a limited sense, true, and it is reasonable to argue that this portfolio is a starting point for investment decisions (perhaps the best we can do). But there are caveats. Within a country, there are taxable and non-taxable investors, and if the data are available, it makes more sense to start with separate aggregate portfolios for the two groups, which also seems a reasonable starting point. There are real problems, however, because not all taxable investors face the same tax rates. Etc. Etc.
So it seems like the default/optimal portfolio should be the average portfolio of investors like me, and not the global market portfolio. If that information is not directly available, we can try to infer it from other data, e.g., if some investors unlike me are known to overweight some asset class (e.g., treasury bonds) then I should underweight that asset class.
References/tools for portfolio optimization
An Asset Allocation Primer: Connecting Markowitz, Kelly and Risk Parity
Portfolio Visualizer
Both of these references fall under the “mean-variance” paradigm, but according to Strategic Asset Allocation: Determining the Optimal Portfolio with Ten Asset Classes:
I think “by definition” is wrong here, but there are strong theoretical reasons to think that the global market portfolio is the optimal portfolio for all investors (investors with different risk tolerance should differ only in how much exposure or leverage they use). However that theory makes some unrealistic assumptions, and here’s a fuller picture, from Q&A: Seeking the Optimal Country Weighting Scheme :
So it seems like the default/optimal portfolio should be the average portfolio of investors like me, and not the global market portfolio. If that information is not directly available, we can try to infer it from other data, e.g., if some investors unlike me are known to overweight some asset class (e.g., treasury bonds) then I should underweight that asset class.