I am implementing the Black Litterman model for a few assets, in particular I am using five ETF:

  • EFA (EAFE stock index: developed markets outside US and Canada)
  • EEM (stocks from Emerging Markets)
  • GLD (Gold)
  • TLT (Long Term (20+ years) US Treasury Bonds)
  • IYR (US Real Estate Investments Trusts)

My questions are the following: how can I compute the market capitalisation weights for these assets? Is it sufficient to divide the capitalisation of each ETF by a market index capitalisation? Should I normalize the weight in order that they will sum up to one?

Thank you for your answers.


1 Answer 1


Do not use the capitalizations of the ETFs, they hold a small and variable proportion of the securities outstanding.

Instead base your weights on the papers by Swinkels et al. which have tried to estimate the actual amounts invested by all investors (not just ETFs!) in various assets. Their latest paper is Historical Returns of the Market Portfolio, 2019 SSRN 297809

Here is a picture of the weights they use

enter image description here

Commodities 1%, Equities 45%,Real Estate 6%, Govt Bonds 29%, non Govt Bonds 19%.

You will have to make some adjustments, for ex. I see you do not include non-Govt Bonds. "Non government bonds" generally consist of Corporate Bonds, Mortgage Securities and some supra-national or sub-national govt organizations. (I suggest you use that 19% share for an additional allocation to Govt Bonds or split it into a mix of the other assets (mostly govt bonds but also a little bit of equities and perhaps real estate), so that the total weights still add to 100). For commodities you can substitute Gold.

  • $\begingroup$ Thank you very much. Since I am using a dataset that ranges from 2004 up to now, is it a big approximation to use these fixed weights for all the considered time? Anyway, you've been very exhaustive. $\endgroup$
    – Matteo
    Commented Jun 6, 2020 at 6:18
  • $\begingroup$ Swinkels in some other papers also has the history of these proportions, if you want historical data instead of current. $\endgroup$
    – nbbo2
    Commented Jul 11, 2020 at 19:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.