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
Thanks a lot for all your efforts. Could you please make a video on how to construct a portfolio of different asset classes, monitor its performance and weights over time, and how to rebalance it?
Thank you for your incredible work, truly captivating. Often times Risk Parity Funds use leverage to increase expected returns as risk remains somewhat the same. Would you be willing to demonstrate how funds manage leverage, such as the factors to increase or decrease borrowed capital? I'm assuming volatility and interest rates are generally what determine this but would enjoy hearing your explanation. Thanks again for your diligent work.
Hi Mitch, and thanks so much for your feedback and an excellent question! It is true that a risk parity portfolio can theoretically advise for a short position in a particular asset class but this happens much less frequently than with an optimised mean-variance portfolio for example. It is also quite common for funds to pretend they are generating alpha or enhance their information ratio by taking on leverage (a very witty term in the industry for such fund managers is "closeted indexers"), I do cover this issue when discussing appraisal ratio versus information ratio, check out this video if you are interested: th-cam.com/video/_CxGArdo5Ig/w-d-xo.html Hope it helps!
great video ! thanks for making this! best video on risk parity that I've found so far! Interesting to learn that this approach relays on the covariance matrix. The obvious shortfall is that in stress times/crisis most assets become almost perfectly correlated how does this approach handle those times? and estimating futures values of the cov matrix that must be challenging and maybe very sensitive to forecast errors? risk parity for the retail investor can be challenging to implement as well since it requires leveraging the bonds...maybe these topics could be tackled in a subsequent tutorial. The markowitz mean variance model gets criticized that looks good on paper but since returns and cov matrix are difficult to forecast and sensitive to estimation errors, in practical terms is produces non sensical results, I wonder how much better is risk parity? there's no need to forecast the returns so that aliviates the problem but the cov matrix estimation is still tricky I would assume
Hi Ricardo, and glad you liked the video! And a very relevant criticism of the concept! Yes, the framework relies on the covariance matrix being relatively stable through time which need not be the case especially during market turbulence. You can robustify the estimation against that in practice using conditional correlation at the distribution tail for market returns by implementing something I have showed in my hedges and safe havens video: th-cam.com/video/4TB2nD5-cDo/w-d-xo.html. Check it out if you are interested and hope it helps!
Hi! Is there a video you would suggest when wanting to understand which formulas to use? By 4:43 I'm a little overwhelmed but of course it seems super powerful!
Hi Dan, and thanks for the question! I am going in a bit more detail into how one might use matrix algebra functions in Excel to calculate portfolio volatility in my VaR video, check it out if you are interested: th-cam.com/video/SFESr6tal2o/w-d-xo.html. Hope it helps!
Hi Savva, here's a quick question. How can we know this solution is unique? I mean, could there be another combination of weights that results in the same RC for every asset? Thank you.
Hi Rafael, and thanks for the excellent question! This paper (www.optimization-online.org/DB_FILE/2013/10/4089.pdf) proves that there exists at most one solution per each orthant of the weighting space. In particular, there is at most one risk parity asset combination if no short-selling is allowed. More generally, there exists at most one risk parity portfolio if you impose sign constraints on each of the weights (i.e., you know which assets you want to long and which to short).
great video! However, when i use solver to do this, one of my assets goes to 0% allocation and the deviations from parity only gets to around 15%. I don't understand why the solver attempts push one of the assets away from parity?
i have 14 funds in the analysis. one is contributing negative covariance risk (similar to treasuries in your example), solver tries to move tis fund to a zero allocation, and all the others well beyond their appropriate risk weight. I've tried setting upper bounds for the other funds and tried to force a minimum on fund with negative covariance contribution. Solver cant make it work. Iam going to use the iterative method, but does anyone know how to avoid these solver errors?
Hi Sava, What are your recommendations for prices of assets opened on weekend and holidays (like Bitcoin) with assets opened during the regular 252 trading days ? This situation creates different volatility if one is to ''eliminate'' weekend data for Bitcoin or ''extrapolate'' weekend data for stocks. Thank you,
Hi, and thanks for the question! It is a very common issue indeed, and the most natural way of dealing with that would be to map your price data to the trading days of your benchmark - for example, if you are a US investor, you would have price data for each trading day of S&P 500 and "eliminate" weekends and holidays for the Bitcoin price. Hope this helps!
Hi Tom, and glad you liked the video! Risk parity makes little sense when considering diversification across assets and much more sense when considering diversification across asset classes. The argument is that a portfolio exposed to equal risk shares from stocks, bonds, real estate, etc. has more attractive risk-return properties and captures risk premia associated with systematic risk on these different markets. The main assumption here of course is that these risks are leading to higher expected returns, which is not always true (think of most commodities like oil of coffee). Chaves et al. (2011) paper in Journal of Investing explain this argument to greater extent. Hope it helps!
Great Video. Thx. You describe asset weights at a specific single day. If you want to backtest this risk parity portfolio, what are you suggesting to do in excel in terms of calculations, creating many many covariance matrix, etc ? It could be interested to compare with equal-weight, 60/40, or any benchmark. Thx
Hi, and glad you liked the video! Yes, you could in theory apply this optimisation technique on a rolling window basis, calculating the risk parity weights on 5-year overlapping samples and checking for the stability of coefficients for example. This is way easier to implement in Python though. Might do something on portfolio management in Python at some point in the future.
Рад что видео понравилось! Бэктест такого портфеля на Питоне сделать довольно просто, рекомендую функции np.matmul для нахождения весов риска и scipy.optimize.minimize для нахождения паритетного портфеля эффективными численными методами оптимизации.
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
Thanks alot, Sava.
By far the best video on risk parity in internet ( and I have seen them all)
That is some God level excel skills
Thanks a lot for all your efforts. Could you please make a video on how to construct a portfolio of different asset classes, monitor its performance and weights over time, and how to rebalance it?
A Video on rebalancing would be great
Thanks for this video, it really helped me to find a breakthrough with practical implementation of risk parity portfolio building technique.
Sharp, concise, and precise
Why I can only minimize the deviation from parity from 85% to 24%, but can not make it zero? Or one of my asset weight must be negative %.
Great video, congratulations! You just won a well-deserved "like" and an enthusiastic new subscriber 🙂
thanks a lot for the clear explanation!
Absolutely wonderful!!!!
Thank you for the video, crystal clear!
Thank you for your incredible work, truly captivating. Often times Risk Parity Funds use leverage to increase expected returns as risk remains somewhat the same. Would you be willing to demonstrate how funds manage leverage, such as the factors to increase or decrease borrowed capital? I'm assuming volatility and interest rates are generally what determine this but would enjoy hearing your explanation. Thanks again for your diligent work.
Hi Mitch, and thanks so much for your feedback and an excellent question! It is true that a risk parity portfolio can theoretically advise for a short position in a particular asset class but this happens much less frequently than with an optimised mean-variance portfolio for example. It is also quite common for funds to pretend they are generating alpha or enhance their information ratio by taking on leverage (a very witty term in the industry for such fund managers is "closeted indexers"), I do cover this issue when discussing appraisal ratio versus information ratio, check out this video if you are interested: th-cam.com/video/_CxGArdo5Ig/w-d-xo.html Hope it helps!
@@NEDLeducation You've exceeded already high expectations, I thank you.
Thanks a lot Sava, this is helpful
Ur a wizard. Super helpful. Wish it was python tho
When calculting the Covariance Matrix, you didn't De-Mean the Returns - is that correct ?
great video ! thanks for making this! best video on risk parity that I've found so far! Interesting to learn that this approach relays on the covariance matrix. The obvious shortfall is that in stress times/crisis most assets become almost perfectly correlated how does this approach handle those times? and estimating futures values of the cov matrix that must be challenging and maybe very sensitive to forecast errors? risk parity for the retail investor can be challenging to implement as well since it requires leveraging the bonds...maybe these topics could be tackled in a subsequent tutorial. The markowitz mean variance model gets criticized that looks good on paper but since returns and cov matrix are difficult to forecast and sensitive to estimation errors, in practical terms is produces non sensical results, I wonder how much better is risk parity? there's no need to forecast the returns so that aliviates the problem but the cov matrix estimation is still tricky I would assume
Hi Ricardo, and glad you liked the video! And a very relevant criticism of the concept! Yes, the framework relies on the covariance matrix being relatively stable through time which need not be the case especially during market turbulence. You can robustify the estimation against that in practice using conditional correlation at the distribution tail for market returns by implementing something I have showed in my hedges and safe havens video: th-cam.com/video/4TB2nD5-cDo/w-d-xo.html. Check it out if you are interested and hope it helps!
really super. thank you so much
Great video! Now it could be the time for the Relaxed Risk Parity strategy, could you make a video about it? Thanks, you are the best!
Hi! Is there a video you would suggest when wanting to understand which formulas to use? By 4:43 I'm a little overwhelmed but of course it seems super powerful!
Hi Dan, and thanks for the question! I am going in a bit more detail into how one might use matrix algebra functions in Excel to calculate portfolio volatility in my VaR video, check it out if you are interested: th-cam.com/video/SFESr6tal2o/w-d-xo.html. Hope it helps!
Thanks a lot!
Amazing work
Hi Savva, here's a quick question. How can we know this solution is unique? I mean, could there be another combination of weights that results in the same RC for every asset? Thank you.
Hi Rafael, and thanks for the excellent question! This paper (www.optimization-online.org/DB_FILE/2013/10/4089.pdf) proves that there exists at most one solution per each orthant of the weighting space. In particular, there is at most one risk parity asset combination if no short-selling is allowed. More generally, there exists at most one risk parity portfolio if you impose sign constraints on each of the weights (i.e., you know which assets you want to long and which to short).
Hi Sava,
Any chance that you could implement Hierarchical Risk Parity in excel? I know, stupid question 🤕
Quick question, why is covariance and the covariance matrix needed? Thank you so much!
Top G Sava.
great video! However, when i use solver to do this, one of my assets goes to 0% allocation and the deviations from parity only gets to around 15%. I don't understand why the solver attempts push one of the assets away from parity?
i have 14 funds in the analysis. one is contributing negative covariance risk (similar to treasuries in your example), solver tries to move tis fund to a zero allocation, and all the others well beyond their appropriate risk weight. I've tried setting upper bounds for the other funds and tried to force a minimum on fund with negative covariance contribution. Solver cant make it work. Iam going to use the iterative method, but does anyone know how to avoid these solver errors?
Hi Sava, What are your recommendations for prices of assets opened on weekend and holidays (like Bitcoin) with assets opened during the regular 252 trading days ? This situation creates different volatility if one is to ''eliminate'' weekend data for Bitcoin or ''extrapolate'' weekend data for stocks. Thank you,
Hi, and thanks for the question! It is a very common issue indeed, and the most natural way of dealing with that would be to map your price data to the trading days of your benchmark - for example, if you are a US investor, you would have price data for each trading day of S&P 500 and "eliminate" weekends and holidays for the Bitcoin price. Hope this helps!
Good video! What would be the reasons to aim for equal risk parity? It seems more of a byproduct of diversification rather than a goal
Hi Tom, and glad you liked the video! Risk parity makes little sense when considering diversification across assets and much more sense when considering diversification across asset classes. The argument is that a portfolio exposed to equal risk shares from stocks, bonds, real estate, etc. has more attractive risk-return properties and captures risk premia associated with systematic risk on these different markets. The main assumption here of course is that these risks are leading to higher expected returns, which is not always true (think of most commodities like oil of coffee). Chaves et al. (2011) paper in Journal of Investing explain this argument to greater extent. Hope it helps!
Is there a use of phyton?
Great Video. Thx. You describe asset weights at a specific single day. If you want to backtest this risk parity portfolio, what are you suggesting to do in excel in terms of calculations, creating many many covariance matrix, etc ? It could be interested to compare with equal-weight, 60/40, or any benchmark. Thx
Hi, and glad you liked the video! Yes, you could in theory apply this optimisation technique on a rolling window basis, calculating the risk parity weights on 5-year overlapping samples and checking for the stability of coefficients for example. This is way easier to implement in Python though. Might do something on portfolio management in Python at some point in the future.
@@NEDLeducation Much needed.
Супер! Было бы интересно на Python посмотреть, бэктест сделать.
Рад что видео понравилось! Бэктест такого портфеля на Питоне сделать довольно просто, рекомендую функции np.matmul для нахождения весов риска и scipy.optimize.minimize для нахождения паритетного портфеля эффективными численными методами оптимизации.
hah, lost from the get go