I recently got an interview question for a junior analyst role asking if risk could be infinite in CAPM, and I wasn't sure how to answer it. I don't see how an asset could be infinitely more volatile than the market. I understand a theoretical 0-beta asset, but not infinity. So how high can beta really be? In what case would it be infinite (if any)?

I apologize if this question is very elementary and I'm missing a key point here.

  • $\begingroup$ Theoretically, with an outrageous amount of leverage and without the usual negative impact. (Taking this approach to the financial product world, there are plenty of highly leveraged ETFs with betas of 5.) Obviously you'd need to suspend some sensibility, like the imposition of borrowing costs. $\endgroup$ – Peacock Sep 15 '17 at 7:20
  • 1
    $\begingroup$ It is a strange question. Hard to understand the point behind it IMO. $\endgroup$ – noob2 Sep 15 '17 at 19:51

Infinity is rather non-sensical. A better question perhaps is whether you can put some theoretical bounds on an asset's market beta.

An asset's volatility bounds its market beta

Let $R_i$ be the return of security $i$ and $R_m$ be the return of the market. Market beta would be given by:

$$ \beta_i = \frac{\operatorname{Cov}(R_i, R_m)}{\operatorname{Var}(R_m)}$$ Let $\rho \in [-1, 1]$ be the correlation coefficient, $\sigma_i$ the standard deviation of return $i$, and $\sigma_m$ the standard deviation of the market return. Since $\operatorname{Cov}(R_i, R_m) = \rho_{im} \sigma_i \sigma_m$ we can rewrite the above expression as:

$$ \beta_i = \rho_{im} \frac{\sigma_i}{\sigma_m} $$

$\rho_{im} \in [-1, 1]$, hence if we know $\sigma_i$ and $\sigma_m$, we can put an upper bound on the market beta: $\beta_i \in [-\frac{\sigma_i}{\sigma_m}, \frac{\sigma_i}{\sigma_m}]$. To have a high market beta, you need high volatility. This is perhaps rather obvious.

Another constraint: value weighted average beta must be 1

Let $w_i$ be security $i$'s share of the market portfolio. The market portfolio return is then:

$$R_m = \sum_i w_i R_i $$

Take covariance of both sides and divide by variance of the market:

$$ \frac{\operatorname{Cov}(R_m, R_m)}{\operatorname{Var}(R_m)} = \sum_i w_i \frac{\operatorname{Cov}(R_i, R_m)}{\operatorname{Var}(R_m)}$$

Observe that the first side is 1 and the second side are market betas:

$$ 1 = \sum_i w_i \beta_i$$

The value weight mean market beta of all securities in the market portfolio must be 1! Speaking loosely, the larger $i$'s weight is in the market portfolio, the more market beta is pulled toward 1.

On the other hand, there's no theoretical reason the security must have positive net supply (eg. derivatives don't have positive net supply).

Last comment

Whenever people talk about market betas, there's some tendency to say "CAPM." Resist the urge. You can estimate market betas and run the following regression whether the CAPM is true or not.

$$R_t - R^f_t = \alpha + \beta (R^m_t - R^f_t) + \epsilon_t$$

The CAPM is an economic theory that implies that $\alpha_i$ in the above regression is zero. The CAPM, while simple and beautiful, is an empirical failure. It doesn't work. That said, you can still estimate market betas and use them in sensible ways. Just don't use them as a sufficient statistic to forecast returns.

  • 1
    $\begingroup$ As hinted, you can indeed reach any real number with an asset in zero supply. Assume that you have a derivative that pays you $10\times R_m$, then $\beta=10$. If you have a derivative paying $100\times R_m$, then $\beta=100$. Of course, you can keep going forever. $\endgroup$ – fni Sep 15 '17 at 17:53
  • $\begingroup$ @fnic Yes! Thanks for making that point clear and explicit. $\endgroup$ – Matthew Gunn Oct 2 '17 at 14:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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