1
$\begingroup$

Would it make sense to use a regression to calculate beta for returns on a foreign exchange currency (regressed on a market average of all currencies)?

Would the beta make sense? (why/why not)

Should I worry about CAPM assumptions?

Thanks for your assistance!

$\endgroup$
2
  • $\begingroup$ Make sure you express the N exchange rates in consistent terms, for example 'value in USD of one unit of foreign currency' and then yes you can do this kind of regression. Which will show that the USD does play a special role in the system. But CAPM is not really considered a satisfactory model of exchange rates, to be honest. In a better model there would be 2 factors: a USD vs everybody else factor and a 'high interest currency' vs 'low interest currency' factor. $\endgroup$
    – nbbo2
    Commented Oct 9, 2017 at 13:09
  • 3
    $\begingroup$ It seems to me that FX is not really an "investment" in the sense that the investor base is not being asked to collectively invest. Instead , risk is bilateral. For example , a US investor might consider being long Yen as a risk, whereas a Japsnese investor sees long USD as a risk. If these natural forces balance, then the expected return on FX should not be substantially different from the forward rates. Risk premiums should be small. $\endgroup$
    – dm63
    Commented Oct 12, 2017 at 10:16

2 Answers 2

2
$\begingroup$

I don't think it is making sense to choose the CAPM approach, since in the FX-Market there is no appreciation and the market average of the returns of every currency should be 0.

$\endgroup$
1
$\begingroup$

beta is a measurement of correlation and risk. Calculating beta with a regression can be a solution.

Beta = covariance / variance

Since Beta represents strength and level of correlation between two different symbols, it is correct to make a linear regression to calculate it.

CAPM is influenced by Beta, since the formula is:

CAPM = Rf + Beta * (Rm - Rf)

but I do not think you will affect negatively your CAPM if you compute Beta with a linear regression.

I do not see any problem related!

$\endgroup$

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.