Post-modern portfolio theory explained: Sortino ratio and volatility skewness (Excel)
HTML-код
- Опубликовано: 20 ноя 2021
- Post-modern portfolio theory (PMPT) is a refinement of the conventional modern portfolio theory approaches with an emphasis on downside risk, return targeting, and asymmetries of the return distributions. Today we are covering the most commonly used concepts from PMPT, such as downside beta, Sortino ratio, and volatility skewness, and applying them to portfolio evaluation and optimisation, while comparing the results to those of the conventional MPT.
Don't forget to subscribe to NEDL and give this video a thumbs up for more videos in Investment!
Please consider supporting NEDL on Patreon: / nedleducation
You can find the spreadsheets for this video and some additional materials here: drive.google.com/drive/folders/1sP40IW0p0w5IETCgo464uhDFfdyR6rh7
Please consider supporting NEDL on Patreon: www.patreon.com/NEDLeducation
Fantastic video. No nonsense, no gimmicks, just pure, crisp logic. Thanks a lot.
Hi, and many thanks for such kind words! Stay tuned for more videos on portfolio management!
Excellent clarity on the topic and very well explained without wasting a word!! Thanks a lot. You are making it very easy to grasp the concepts and the calculations for everyone!!
Thank you so much! Best practical explanation of PMPT by a margin
Could you please make a video on dynamic mean-variance asset allocation considering transaction costs?
Hi, and glad you liked the video! Thanks for the suggestion, will be making a video on portfolio optmisation with transaction costs in the nearest future!
@@NEDLeducation Thanks! Your videos are really helping me with my thesis! 😁
This is great. And feels a lot more applicable to real life than the markovitz portfolio.
I'd love to see an implementation of it in python, and also a video on ways of estimating future returns to use with this optimization method.
Hi Jorge, and happy you enjoyed the video! I might do a series of videos on portfolio management applications in Python at some point in the future! As for the expected return estimations, I do provide examples of CAPM and downside-CAPM in the video, but if you wish to implement a more active approach (buy undervalued stocks, for example), then the expected return can be derived from your target prices or fair valuations. This is really the essence of your stockpicking strategy, and what portfolio management models such as this one do is optimise allocation when your investable universe is a given :)
Excellent video. A paucity of talk, a plethora of explanation & visuals.
Merry Christmas Sava! Thank you so much for your amazing content and for keeping so high your channel's quality!!
Hi Peter, and Merry Christmas! Many thanks for your kind words, glad you liked the videos. :)
Great video; we use a version of Post Modern Portfolio Theory using skewed returns and Copulas. Attilio Meucci has some excellent papers on it as well.
Excellent explanation on the topic!
Thank you very much. Excellent explanations. God bless you!
This was so so helpful. Thank you :)
Amazing video, thank you. The only thing is the daily rebalancing you are indirectly making with the portafolio returns.
Hi Franco, and thanks for the comment! Here, we are not necessarily simulating a portfolio investment for this period but optimising allocations based on historical data sampling. However, it is prudent to distinguish between fixed and drifting weights as you correctly say, I have got a video on this here: ruclips.net/video/fGov9fvug8o/видео.html
@@NEDLeducation Understood. Thanks for the clarification and for all the videos.
thanks and congratulations for the video. In the minute 7:16 where yoy explain how you construct the formula for the downside beta, the first part is: "take into account only the security values when the market is down. but in the second part you mention: here we apply the same... but the negative part is missing. Am I understanding something different?
Hi Alexander, and thanks for the great feedback! The second IF is actually unnecessary given how the SLOPE function works - it omits any pairs where one of the observations is blank. So it does work correctly but the second IF can be omitted entirely.
Dear Sava, thanks for video! Which methodology did you use to calculate the downside beta?
Hi, and thank you again for the great content! Wanted to ask why I get different values for Downside Beta in Excel when hitting Ctrl+Shift+Enter as opposed to just hitting Enter? Thank you.
To add to this, without Ctrl+Shift+Enter, Beta=Downside Beta.
Hi Alex, this is probably due to an older version of Excel. In these, you have to enforce all array/matrix functions with Shift+Ctrl+Enter.
Best. Channel. Ever!!! One quick question on this video...I replicated this with monthly returns for 4 mutual funds, and the sum of my upside variance and downside variance does not equal the total variance. Is this possible, or have I made a mistake somewhere? Thanks!!!!
Hi there. First of all, excellent material, thanks for sharing! I do have a few questions. How does this sort of portfolio behave in general in out of sample tests? How does this compares to for example M. Lopez de Prado Hierarchical Risk Parity? How does it perform against Black-Litterman? Also, in a particular problem I have, from an investable universe, I have a machine learning approach that classifies stocks as high probability of over performing a given benchmark n days ahead, either going long (classified as 1, p > 0.5 ) or going short. The OOS performance of the classifier is acceptable to me, but portfolio construction using traditional methods is not effective. Any idea/suggestion for a methodology that could overcome the "disconnect" between the in-sample and out-of-sample performance of MPT for long and short portfolios of short to mid term duration? Thanks again and best regards.
Hi, and glad you are enjoying the channel! Thanks for the wonderful questions, it is indeed the case many MPT methods are performing very poorly out-of-sample, particularly if you focus on historical estimations of return and (to much lesser extent, though) risk. Using some expected return measure (e.g., CAPM, or D-CAPM as presented in this video) makes optimised portfolios much more robust in an instance. You can theoretically implement forward-looking measures of risk like option-implied volatilities, however this does not translate nicely into covariance matrices (some work has been done recently on option-implied correlations but it is, IMO at least, not enough to estimate robust forward-looking covariance matrices). There is a pretty good paper that compares performance of some of the most common allocation optimisations (papers.ssrn.com/sol3/papers.cfm?abstract_id=2529944) so you can get some inspiration from there as well. It is from 2014 though so the market environment could have changed somewhat since then.
@@NEDLeducation Thanks!
Great video!
Just a question, the second "IF" in "Downside beta" formula has no "
Seems like it's just a mistype and there is should be, of course, "
Hi Ale and Kirill, and thanks for the comments! Well spotted, there indeed is a typo in the formula, have updated the spreadsheet on the Google Drive. Coincidentally, it did not affect the downside beta calculation results due to the way SLOPE function works in Excel (it works correctly as long as there is an IF condition for at least one array of the two).
Hi, In minute 9:07 when you are simulating the portfolio daily return and you lock the weights you are assuming a daily rebalance, right? Wouldn't be better/more realistic to simulate a portfolio with an specific rebalance date?
Also in minute 15:34 I think that this formula is Roy's safety first ratio. The sortino ratio is (port- risk free)/downside risk
Hi Luis, and thanks for the question! Generally, when optimising a portfolio, we effectively perform a historical simulation of risk, so we are not simulating a ten-year investment (where the daily rebalance assumption would be questionable), we are using ten years worth of data to simulate a number of alternate scenarios (where the fixed weights is not a daily rebalance but rather the starting weights). I discuss the application of fixed versus drifting weights here: ruclips.net/video/fGov9fvug8o/видео.html
Як же ти харош
Hi Savva, I was trying to compute PMPT but I got stuck with this problem : the historical return of market is negative , so I can't go on. Do you know how to solve this?
Hi Angelo, and thanks for the question! Ultimately you can go for a longer time period of use the expected return of the market (from Damodaran for example) rather than historical.
Can this be done with weekly/monthly data or must it be daily?
Sorry for my stupid question. Why was variance-covariance matrix not used when calculating the portfolio variance?
Hi Jia, and thanks for the question! As we are not using the matrix itself in the optimisation here, it was simpler to calculate portfolio variance directly. However, we could have still used the matrix and the result would be the same.