Brian Romanchuk has a post on portfolio allocation, including its representation in SFC models. One area he discusses compares different ways of weighting portfolios.
In Monetary Economics, Godley & Lavoie use portfolio allocation functions that assign fixed weights to different asset classes (subject to changes in relative expected returns). In the context of international allocations this means, for example, that investors in Country A allocate x% of investment to domestic assets and (100 - x)% to Country B assets.
If we ask ourselves what kind of value x might be, it should seem that it depends to some extent on the relative size of the economies of Country A and Country B. If the economy of Country B is only 10% the size of that of Country A, it is unlikely that Country A investors are going to be investing 50% of their assets there, or even 25%.
We should expect investors to show some home preference, but beyond that we should expect the amount that investors wish to invest in a particular country to be influenced by the relative size of its economy.
How do we compare the size of different economies? We have to translate GDP's into a common currency using suitable exchange rates. Therefore, if we wanted to have portfolio allocation functions that reflected the different sizes of economies, we would need to include exchange rates in these functions. This is not the route that Godley & Lavoie take.
This may seem like a relatively unimportant point but, as I have stressed before, portfolio preference actually has a big impact. How international weightings are determined can make a big difference to the response to changes in trade propensities, including whether that response is primarily an exchange rate response or a trade balance response.
Unfortunately, it's very difficult to identify portfolio allocation behaviour from the data, so it's hard to form a view on this. My guess is that weightings are based on relative sizes of economies, but only in the longer run. That is, weightings do not adjust to short-term movements in the exchange rate, but over time they come back into line. This would seem to fit with a rule-of thumb approach to dealing with Brian's observation that it is best to adjust weightings for trend movements, but not for fluctuations.