Talk:Modern portfolio theory

graphics

I added a picture. I have the xfig source, so it is possible to modify it if this is too clunky. CSTAR 04:15, 1 Jul 2004 (UTC)

Oops I noticed in the text that you're plotting risk on the Y-axis, whereas I plotted it on the X-axis. Also I'm callin the measn \mu (which is common ampng probabilists) Well I can fix this, if you think it's worth it. CSTAR 05:07, 1 Jul 2004 (UTC)

I think it would tie in closer with the article if the axes were labelled as per the text -- ie. return and risk -- as well as per the notation as above.

I'll change the graphic, although it seems that yuu changed the text to have risk on X axis.

Is the assumption that between alternatives of equal risk choose one with largest expected return not necessary? (I haven't thought about the minimal number of assumptions so maybe this follows from risk aversion) CSTAR 12:42, 2 Jul 2004 (UTC)

Thanks for the graphic. The arguement is (ultimately) built around standard deviation as opposed to variance, could the graph reflect that?

I think the assumptions are equivalent and are used interchangeably....

Ooops you're right the CML is linear in the space of standard deviation, risk pairs. Oh well, I'll fix that.CSTAR 15:27, 2 Jul 2004 (UTC)

I adjusted the graphic; I also take note of the fact that in the article risk is measured by standard deviation, which is necessary in order for the capital market line to be a straight line! CSTAR 16:45, 2 Jul 2004 (UTC)

Shorting

Maybe you should something about shorting? Or at least point to some place in Wikipedia where this is mentioned. CSTAR 18:28, 2 Jul 2004 (UTC)

Great graphic. Will add discussion on shorting

MPT has nothing to do with shorting. It has everything to do with the balance of risk and return of assets classes and their relationship over time. Shorting stocks does not fall into that framework.

Black-Scholes

I corrected two incorrect references to the Black-Scholes model. That model does not imply or assume that stock prices follow a martingale or that the best forecast of tomorrows prices is the price today. —Preceding unsigned comment added by MathHisSci (talk • contribs) 22:01, 29 October 2009 (UTC)

=

It is my recollection that returns could be achieved above the Efficient Frontier, but constrained by the Capital Market Line, by leveraging, i. e. shorting the risk free asset. Also, couldn't being short an asset be seen as a synthetic asset that has an expected return with a (presumably high!) standard deviation. Since "shorts" would be expected to have low covariance with "long" assets "shorts" might be great assets to create efficient portfolios.X17bc8 (talk) 17:49, 19 May 2009 (UTC)

Risk

In order for the capital market line to be straight, risk should be standard deviation not (variance). I think. CSTAR 22:47, 14 Jul 2004 (UTC)

That is true - I see it has been changed. Zain Ebrahim (talk) 22:21, 17 May 2008 (UTC)

distribution about the mean

I see no discussion of what happens when there is an asymmetric distribution about the mean, which is always the case in financial models where the asset value is bounded on the low side by ZERO.

• Exactly. This all seems to reply on the Gauss-Markov assumption (returns are normally distributed => wake up, they aren't, just use HRH on Bloomberg for a few indices or stocks). As I understand it, the CAPM can outside of the Gauss Markov assumptions (significantly beyond the scope of the definition given on this page). I do not know of any 'Omega metric' pages on wikipedia (which the derivation comes off), and feel I am probably not the best person to write one, though I would be happy to have a stab and link it in somewhere?

• I shall write something on Omega metrics, and stick a reference in the page (Omega being a generalised treatment for asymmetric returns, though is it non-parametric - CAPM can still be derived from it).

Financial models don't a normal distribution of prices. They assume a log-normal distribution of prices, aka a normal distriubtion of returns. --206.248.177.127 (talk) 19:35, 2 January 2011 (UTC)

Expected beta

Note that the theory uses an historical parameter, volatility, as a proxy for risk while return is an expectation on the future. Is this an accurate description of the theory or of practice? I would think that the theory says that expected return and expected volatility should be used. I don't the the theory prescribes using historical beta. In practice people use historical return and historical beta, then adjust for their expectations. Since it is very hard to develop numerical estimates of beta, it is usually expected return and historical beta.

If not disagreements in the next few days, I'll make the change. Thanks. Chris vLS 20:44, 19 Nov 2004 (UTC)

OK for the change --Pgreenfinch 22:30, 19 Nov 2004 (UTC)

Ideally, one should use some kind of time series analysis to estimate these parameters. If one makes an assumption about the nature of the underlying time series process, then there is some rational justification for making estimates of the various parameters characterizing the process. I don't have a clue in fact how the nuts-and-bolts financial analysts actually compute these numbers and whether they would take exception to the characterization given in the article, but I think the person that put that material in (fintor) is knowledgeable about practices in this area. CSTAR 22:40, 19 Nov 2004 (UTC)

Oh, I'd absolutely agree about it as a statement of practice. I thought that it currently reads like a criticism of the abstract theory itself, claiming the CAPM nonsensically requires you to mix past and future. In practice, there are tons of historical calculations of beta, but only a few proprietary models for predicting beta (see Barra, who sells both [1]]).(Indeed, all theories tell you to focus on good estimates of the future for all modeled parameters, but in practice, detailed analysis of past perfomance gets a lot of weight.) Chris vLS 08:17, 20 Nov 2004 (UTC)

I totally withdraw my comment. The text is correct. Nearly all forms I can find of the formula use E() to denote what is expected, and do not show E(beta). Hence, the description is correct. Sorry for the trouble. Chris vLS 19:02, 30 Nov 2004 (UTC)

Risk free asset

If you're going to change notation for rf, change it everywhere. CSTAR 18:40, 30 Nov 2004 (UTC)

Distribution about the mean and other objections

Right now, there are some comments that come close to attacking the theory in the same paragraphs that explain it. While that's fine for the sophisitcate, for the novice I think it's confusing -- kind of like saying "Note that general relatively is not accounted for in F=ma" in the middle of the description of Newton's force formula. I think it would be better to create a new section, called "Shortcomings and challenges" or something like that and discuss where modern portfolio theory has gone since the CAPM.

Thoughts? -- Chris vLS 21:10, 7 Dec 2004 (UTC)

---

Do you mean something like the following assertion?

(Here again, the theory accepts in its assumptions that a parameter based on past data can be combined with a future expectation.)

I don't know who put this in, but in the model as formulated this assertion is meaningless and should be removed. The MPT model described in this article is a probabilistic model, but the random variables have no time dependency whatsoever, and although such models are used in finance, this isn't one of them.CSTAR 21:25, 7 Dec 2004 (UTC)

Yes and I agree. Here are the statements I see that could be moved:

"The investor is indifferent to other characteristics of the distribution of returns, such as its skew."

"Note that the theory uses an historical parameter, volatility, as a proxy for risk while return is an expectation on the future."

"(Here again, the theory accepts in its assumptions that a parameter based on past data can be combined with a future expectation.)"

--Chris vLS 21:17, 10 Dec 2004 (UTC)

---

"The investor is indifferent to other characteristics of the distribution of returns, such as its skew."

This is a meaningful assertion about this model, so I don't see any reason to change it.

On the other hand, the article should say something intelligible about parameter estimation somewhere. I suppose this was the intent of whoever adding the assertion of the parameter is based on past performance. It is a truism that historical data is the only data available to estimate parameters of time series. But this requires a time series model.

I would wait to whoever wrote that part of the article to respond before deleting anything. CSTAR 22:41, 10 Dec 2004 (UTC)

---

Oh, I agree with the statement about skew. I am not looking to delete, just clarify. All I'm saying is that we should consider moving and expanding the discussion of shortcomings into its own section, as is done in the article on the CAPM. --Chris vLS 06:09, 12 Dec 2004 (UTC)

Agreed: we need a seperate "shortcomings" section. I also think that the article could use an explicit "assumptions" section - I would think that the two should to some extent dovetail... Fintor | talk |January 9 18:49 UTC

---

For those talking about distributions about the mean, please note that the CAPM works _whatever_ the distribution of returns it is derived from is - you could even have a classic lottery distribution of returns and it'll work, but you cannot use the variance-covariance method to work it out (so while it can be derived from raw data, standard deviation cannot be used as a probability).

Perhaps the derivation section (as it stands now, discussing parametric specification of CAPM) can be moved to a subsection, or 'appendix' under the discussion of what CAPM actually means, I am happy to add a section on how to derive CAPM for _any_ distribution of returns (whatever the skew, kurotsis, or any other partial moments of conditions of a distribution), its really pretty simple stuff if you know how: uses 'Omega Metrics', mentioned above.

To repeat, the result of the SML/CAPM do not depend features of the underlying distribution if the SML/CAPM have been derived correctly (unbiased) w.r.t. that distribution.

zhte415 | talk |January 16 01:37 GMT

Practical guide

This article was a nice introduction for me, but for those who plan on using this theory in structuring a portfolio, it's not very practical. I was reading the same page on investopedia (http://www.investopedia.com/terms/m/modernportfoliotheory.asp) and they said there were four distinct steps:

1. Security Valuation
2. Asset Allocation
3. Portfolio Optimization
4. Performance Measurement

Is there any chance of getting a section describing in practical terms how this theory is used? --Jens Schriver 19:10, 11 January 2006 (UTC)

Merge with Portfolio Theory, more APT needed here

The Article Portfolio theory is a stub on the same topic. I suggest that it should be merged into this article, and will do so fairly quickly if there are no objections.

Also I think that an article named modern portfolio theory needs to have something more than just a reference to arbitrage pricing theory. Smallbones 09:49, 27 March 2006 (UTC)

I don't see why the word "modern" is needed at all. There's nothing more modern about portfolio theory than other theories in finance and economics, or in science generally, and the term isn't used by specialists in the field. All theories change over time, but that's to be expected. So, I think the correct course is to merge this into Portfolio theoryJQ 02:36, 28 May 2006 (UTC)

Errors in the article ?

Please check. In Diversification Section: "An investor can reduce portfolio risk simply by holding instruments which are not perfectly correlated." Shouldn't it be "negatively correlated" Also, "From the formulae above: if any two assets in the portfolio have a correlation of less than 1..." shouldn't it be "a correlation of less than 0"?

Agreed. Perfectly correlated assets implies that if one goes down, so does the next in the series. Reduce risk by holding varied, negatively correlated assets. 66.9.159.226 22:30, 25 January 2007 (UTC)Steve

For a two asset portfolio, the statement in the article is correct generally, if one allows selling short (e.g. allowing the weights w for portfolio composition to be negative).

Moreover, if κ is the correlation between asset 1 and 2 and

$\sigma_1 > \kappa \sigma_2$

$\sigma_2 > \kappa \sigma_1$

then the minimum variance portfolio is achieved by positive weights for both assets; (i.e., no shorting).

--CSTAR 02:53, 26 January 2007 (UTC)

Disagre!!!

"From the formulae above: if any two assets in the portfolio have a correlation of less than 1..." shouldn't it be "a correlation of less than 0"? NO!!

The article was right, now it's wrong. "if all assets of a portfolio have a correlation of 0, the portfolio variance and hence volatility will be the weighted average of the individual instruments' volatilities..." Volatility will be the weighted average of volatilities if correlation is 1 and not 0.

(σp = ωa.σa + ωb.σb ↔ σp^2 = ωa^2.σa^2 + ωb^2.σb^2 + 2.ωa.ωb.σa.σb ↔ ρab=1).Davivalle 21:34, 2 September 2007 (UTC)

Please correct me if I am wrong!

The interesting thing is that for perfect correlations $\rho=1$ the portfolio volatility (i.e. standard deviation) is the weighted sum of the assets' volatilities - as you have said Davidalle. However, for uncorrelated assets $\rho=0$ the portfolio variance is the weighted sum of the assets' variances. Tomeasy (talk) 00:47, 2 March 2008 (UTC)

in Mathematically section "the where w is a vector of portfolio weights. Each wi>=0 "

IMHO the condition wi>=0 is true with 2 assets but wrong in general. With more than 2 assets efficient portfolios can have wi<0. With more than 2 assets the condition wi>=0 is in fact an additional constraint that prohibites short selling and as any additional constraint will reduce the efficient frontier. check out Investments 7th edtion Bodie Kane Markus pg 226 —Preceding unsigned comment added by 137.111.130.201 (talk) 00:45, 30 July 2008 (UTC)

Criticism by Taleb

I was reading a recent FT article by Taleb that was pretty critical of some of this, and I was wondering why there isn't any criticism in the article. Taleb's book, "The Black Swan" is fairly popular, should we bring in some of his (and others'?) criticisms? Some articles like CAPM have "shortcomings" sections, but I'm not sure that is enough. What would be a good way to bring in some criticism? Smmurphy^(Talk) 19:29, 27 October 2007 (UTC)

Also Benoît Mandelbrot's The (Mis)Behavior of Markets —Preceding unsigned comment added by Fintor (talk • contribs) 08:03, 30 October 2007 (UTC)

I agree that a criticism section should be added--one of Taleb's criticisms is that modern portfolio theory is a fraud because it pretends that the data follows a Gaussian distribution when it doesn't, and thus events like the October 1987 crash occur far more frequently than the model predicts. The link to Mandelbrot's book is no good, but looking at the Amazon reviews of that book, it looks like an excellent source of material for a criticism section as well. Lippard (talk) 20:23, 22 May 2008 (UTC)

In a bloomberg interview on 7th Nov 2008, Taleb called Modern Portfolio Theory "hogwash" and said that academics like Scholes and Merton should be discredited. [2] —Preceding unsigned comment added by Vynbos (talk • contribs) 08:16, 13 November 2008 (UTC)

Efficient Frontier Concavity

The article claims that the graph of the efficient frontier is convex. This seems wrong. Convex functions are U-shaped. Convex function . Shouldn't it say "concave"? A Concave function is frown-shaped, upside-down U. I don't want to change it myself because maybe the wording comes from a reliable source. Eyal0 (talk) 06:22, 27 May 2009 (UTC)

It could also be "upper convex" or "convex upward". I think the use of the word "convex" in this article corresponds to ordinary English usage (and probably standard usage in the field) rather than mathematical jargon. --Doradus (talk) 17:00, 16 October 2009 (UTC)

Nice job

I just want to congratulate the authors of this article on doing a good job of explaining some very complex concepts in a concise manner. It may be confusing to the ordinary Encyclopedia reader, but then a topic this arcane can hardly avoid that. --Doradus (talk) 16:58, 16 October 2009 (UTC)

Criticism section

I've added a "criticism" section. Currently it has two parts, assumptions and shortcomings, both largely cribbed from the CAPM article (with which there is much overlap, an area ripe for cleanup.) To this I've added all of the problems with MPT I could think of or find here in the talk page.

Much more work on the criticism section is needed. Your improvements would be very welcome! --Jonathan Stray (talk) 08:01, 8 November 2009 (UTC)

Further, there is growing evidence that investors are not rational and markets are not efficient.[1][2] - I don't think this is true in general. As most people seem to know, it is actually very hard to make "above-average" money in financial markets, to add alpha or whatever you want to call it. If it was so inefficient as some people seem to espouse, it wouldn't be that hard. I don't understand why people bother having this discussion. uhh-hmm i will end there. --129.240.146.32 (talk) 16:54, 11 November 2010 (UTC)

"An empirical proof of this is the price-hike that stocks typically experience once they are included in major indices like the S&P 500" - I completely disagree here. Is there a proof this is due to MPT portfolios? As far as I know the main effects are a) Index funds & index arbitrageurs having to rebalance their holdings, b) speculation that a) will drive the price up. — Preceding unsigned comment added by LaurentP (talk • contribs) 16:55, 9 December 2010 (UTC)