What could make sense here is to 'weight' these allocations differently based on how much Alpha/Beta they are. BRKB is mostly Beta with some Alpha involved (for the reasons mentioned above), so perhaps I need to weight it as 75% beta and 25% Alpha. I got 12% in there so by 'Dalio weighting', my true exposure is 9%. PSH is mostly Alpha with a small Beta component. Perhaps 75% Alpha 25% Beta, I got 15% so by Dalio weighting it it would come out as 3.75% as my true 'stock exposure' In any event, it doesnt look like I'm that overweight equities by weighting things like that. What I am overweight is a few managers, especially Ackman, so I need to avoid adding to that position and I also need to avoid adding stocks that are in his portfolio (like FNMA, FMCC, VRX) by my own discretion. If I do that I will start to correlate my own results (my own Alpha 'fund') with his results which wouldn't be good. I plan to exit that MDLZ, no news have surfaced and its an Ackman holding, so it would help to rebalance my Alpha exposure
This 'alpha fund' that I manage for myself, I will try to keep small. I think it makes sense to weight it similarly to the other Alpha exposures that I have. That way if I'm off my game I wont ruin the performance of the portfolio. Except, when I have a lot of conviction in something. That's when I can go "soros" and make a bigger bet. Usually, I think its safer to not get too crazy on your own trading, except here and there
So I just sold half my TLT but bought ZN futures (10y) in a greater duration adjusted proportion. The net effect is for me to increase my bond exposure. As if I sold 4% in TLT but bought 8%. I'm doing this because: -that TLT was leaving me with a negative cash balance at IB (and IB rates are higher than libor from futures) -price action seems good -To increase my stock market hedge. I'm also running some swing strategies that are net long small caps (they showed up as 'other' in my January 1 breakdown), so this adds more needs for me to not be so underweight bonds. I think I got a good enough level of bond exposure for now and I'm not looking to increase it further
Media analyzing risk parity allocation to commodities using 23 years of data of one country "Including passive ownership of commodities can be a significant long-run drag on returns, as a risk-parity index constructed by Salient Index Management shows. It has sharply underperformed both stocks and a traditional 60% stock 40% bond index fund offered by Vanguard since 1993." http://www.wsj.com/articles/trump-rally-darkens-skies-for-all-weather-portfolio-1483117823 journos are so bad
I run some numbers to try to figure some things out. I compared the Daily, weekly and monthtly volatility of SPY to TLT and IEF. Yahoo has data going back to 2002 for all of them. So, the more you lenghten the time horizon, the less volatile the stock market is compared to the bond market. The same thing happened by only looking at the Standard Deviation of Negative Returns only as it can be seen bellow Whats interesting is that the improvement stops with yearly data but only for the SD, SD of Negative Returns still improve I dont have yearly data for the 30y for a big enough sample but looking at IEF and the 10y bond data from my US database only I'm getting this SD Ratio of SPY/S&P500 to IEF/10y Bonds Daily/Weekly/Monthly/Yearly SD 2.79 2.69 2.18 2.25 (it gets worse for the last one, the yearly data) SD Negative Returns 3.47 3.36 3.01 2.53 (improvement continued with yearly data) The stock market is consistently volatile, it gets less volatile as you lengthen the time horizon but that improvement seem to be limited to less negative years. The market will still be volatile to the upside. Its just that the more time you give, the less of a chance of a negative outcome for stocks Of course, all of this is pretty obvious but it has important implications to the use of the Sharpe Ratio vs Sortino Ratio The Sortino Ratio looks at SD of Negative Returns only, it goes to what people care about more, losses. The Sharpe Ratio looks at Standard Deviation in general. As a result an investment strategy that relies on it will be underweight equites compared to one that doesn't. This explains why on some of my tests, using the Sortino, the computer recommended a fair amount of stocks compared to the Sharpe, when the Sharpe was used as metric, the computer was more risk averse The people at Bridgewater are very smart and they probably know all of this, so why are they relying on the Sharpe instead of Sortino? I think that has to do with the fact that they are a hedge fund/investment manager, this leads them to prefer the Sharpe instead of Sortino for a few reasons -Clients tend to care about month to month, week to week volatility. Relying on the Sharpe will create a portfolio with less short-term volatility because it will have less equities in them and decrease that type of short-term volatility that can lead to clients complaints One way to see this is to look at the daily SD of negative returns/SD. The 3.47 from above divided by 2.79, 1.24 is the result. For yearly data, 1.12 is the result. So as time passes, even though the SD remains high for stocks, more of that SD is coming from upside volatility, not downside volatility. Downside volatility is coming down and more positive years become likely -The Sharpe is more well-known by clients and people in finance and they can understand it better. Most people never even heard about the Sortino -Because the future is unknown, its possible that in the future things will be more volatile. An Extremistan type world requires a more conservative allocation. Since equities tend to be the most sensitive asset to this sort of issue, they would prefer a metric that punishes it (the Sharpe) -Other reasons I havent thought about it I'm not sure these reasons are good enough for me to switch to the Sharpe, especially given that I have no clients. I think it makes more sense and as a result, if one has the stomach to ignore day to day, month to month noise, using the Sortino and the computer recommeded portfolios is the way to go. That Extremeistan adjustment can be made by decreasing the equity allocation a little bit, but still, relying on the Sharpe leads to some big differences compared to the Sortino (35% vs 50% in stocks, plus the Sharpe wants you to load up in bonds like there is no tomorrow) I plan to do more thinking/research on this topic
There is one issue with my data here which is that the data is not standarized, I'm comparing SPY to IEF and the S&P500 to the 10y constant maturity bond return. IEF is a 7 to 10y bond ETF (8.5y Weighted Avg Maturity right now)The 10y constant maturity bond return is likely to have more duration and therefore, more volatility. I'm not statistics expert but I think that still proves my point. SD of Negative returns ratio went from 3.01 on monthly data (IEF) to 2.53 on yearly data (10y bond) and SD from 2.18 (IEF) to 2.25 (10y bond). The ratio of the ratio (3.01/2.18, 2.53/2.25) went from 1.38 to 1.12 So the negative returns from stocks went down quite a bit with yearly data, even when you consider that 10y bonds are more volatile than IEF If this looks all complicated its because it is but its also pretty obvious, bonds dont need that much time to improve their negative returns, because they dont have it very often. Stocks do, so they need more time, if you give them time, the ratio of their negative returns to negative returns in bonds, will improve
One of the issues with using the yearly Sortino ratio is that it doesn't capure the times you were up nicely (say 8%) and then had losses all the way down to 1%, for instance. its a positive year but there was a lot of downside vol (and even if it were to go negative to -1%, it wouldn't raise the Sortino much, because the path to the -1% wont be part of the data). This wont show up on the yearly Sortino (it doesn't show up on the Sharpe either though) which raises the possibility that the Sortino calculated off monthly returns is probably the right measure. It strikes a nice balance between long-term and short-term and captures those nasty drawdowns that yearly data dont So this is a next part for my project, to ask the computer to give me the perfect portfolio but judging by monthly data using the Sortino. I dont have monthly data on the S&P500, bonds, t-bills and gold going back to the 20's. If anyone know where to get it I would appreciate any help
But one thing is for sure, given the issues I raised about the yearly Sortino (Extremistan and not accounting for big gains that dissapear), the 50% historical allocation to stocks that the computer recommends is almost surely too high (if you are going for a "tangency" portfolio, that is). But on the other hand, the 35% from the Sharpe could also be too high for the same reasons. The difference is that, in my opinion, the Sortino makes sense while the Sharpe, doesn't make as much sense since big gains are counted as bad (as they increase the standard deviation) So I think its safe to say that for a Tangency portfolio (best return/vol portfolio), 50% is probably the limit that one can go for
Brazil is a good example of this Extremistan issue. The stock market volatility (SD of annual real returns) after hyperinflation was controlled is 43%, before that (during the hyperinflations) it was 91%. And that 91% lasted for long periods of time (with some years being more and some less). Having too much in equities for long periods can be pretty rough in a investor stomach. If one just takes the weighted average of those two numbers, that can lead to trouble as you might face a long strech (say 30 years) that the stock volatility is quite high. Could this also happen in bonds? I guess so, but it SHOULD (according to finance theory) happen more often in stocks, because it is an riskier asset. In theory, one would expect the volatility of volatility to be higher in high return assets and lower in lower return assets So because of that, one would have to bring down their allocation to stocks, to avoid getting in a situation where there is too much in equities for long streches of time (maybe forever, due the future looking nothing like the past) So, more and more, it looks like that 30-40% in stocks allocation that Dalio and risk parity funds use is the right one, and not the 50% recommended by the computer
All of this stuff is quite complex and mathematical, I'm working on all of this in order to create a paper talking about risk parity/balanced portfolios in Brazil. The technical paper wont be hard because you can just put technical terms, number and jargoons and it is all fine. The problem is that in addition to the technical paper (in english), I want to do an long article/mini book for the average person (in portuguese), how I'm going to explain all of this to them? I have no clue, I guess I will have to say 'trust me, I run the numbers' because I have no idea how to explain all of this in a way that mom and pop can understand it