Volatility Balancing

[Sam]

Your strategy of volatility weighing is interesting

Have a read through my Balanced Diversified Volatility web pages at http://www.jfholdings.pwp.blueyonder.co.uk/ and it might help throw some light on the matter (or just confuse you more

If you understand the concepts behind the BDV then you'll get a better feel for how a Japan like exception can occur when using equal weightings.

I was browsing this forum to learn more about PP and found Clive's discussion of Volatility Balancing particularly interesting. (It was in the thread about PP in Japan.)

The normal weightings in the PP is 25% each for Stocks, Long Term Bonds, Gold and Cash.

However, based on each category's historical volatility, Clive came up with different set of weights that seem to reduce the overall portfolio's volatility.

Using Simba's backtesting spreadsheet, I balanced the weights based on volatility in the following securities and came up with:

17.97% Small Cap Value
11.64% Emerging Markets
32.79% Long Term Goverment Bonds
25% 2 Year ST Treasury (Used instead of Cash)
12.60% Gold

1972 - 2009 rebalanced yearly
11.18% CAGR
0.76 Sharpe
2.94 Sortino
2.19% Down SD
6.20% Up SD


Compare this to:

---

PP
25% TSM
25% LTGB
25% Money Market
25% Gold

9.10% CAGR
0.46 Sharpe
1.69 Sortino
2.59% Down SD
6.96% Up SD

---

Modified PP
25% TSM
25% LTGB
25% 2 Year ST Treasury (instead of Cash)
25% Gold

9.53% CAGR
0.50 Sharpe
2.01 Sortino
2.39% Down SD
7.08% Up SD

---

Paul Boyer PP (from MadMoneyMachine.com)
12.5% SCV
12.5% EM
25% LTGB
25% 2 Year ST Treasury (instead of Cash)
25% Gold

11.26% CAGR
0.68 Sharpe
2.72 Sortino
2.42% Down SD
6.79% Up SD

---

The most surprising result is how smooth the graph is from 1972 to 2009 by only adjusting the weights based on relative volatility. I guess this makes since they have low correlation.

Anyone else considering volatility balancing for PP?

[craigr]

Keep in mind that the larger allocation to international stocks introduces a level of currency risk to the portfolio. If we get a period like the 1970s inflation this could be good as the dollar fell sharply. But if we get a period with a rising dollar this could be bad for the strategy. Also be aware that the optimum asset allocation based on a backtest is going to change each year as the markets move and is highly time dependent. What is the optimum allocation today won't be the same one five years from now.

Lastly, be sure you understand the tracking error that could come with a stock allocation that varies from the Total Stock Market fund. There could in fact be some very good years but there are also some very bad years as well. In 2008 the EM index fell -52% compared to the TSM index of -37%. In 2000 the EM index fell -27% vs. TSM -10.5%. In 1982 TSM was up +20.5% with EM showing a loss of -1.51%. Etc. Be sure to consider how you may emotionally handle such a wild ride if the markets are going crazy.

With the above said, if you want to mix up the stock portion in the portfolio I would advise only using index funds to do it.

[Sam]

Good points about the extra risk from the emerging markets component.

I'll see if I can create a Volatility Balanced portfolio with the two basic versions of PP later tonight after work.

PP - TSM, LTGB, Money Market (Cash), Gold
PP Modified - TSM, LTGB, STGB (Cash), Gold

[Sam]

BTW, I was looking into using the ETF's VBR for Small Cap Value and VWO for Emerging Markets.

Is there an index equivalent for these or is this basically what you meant when you said to consider using index funds when mixing up the stock component of the portfolio?

[Sam]

Ok so I did the same thing using the standard PP assets.

---

PP - Volatility Balanced

23.14% TSM
37.47% LTGB
25% Money Mkt
14.41% Gold

8.8% CAGR
0.498 Sharpe
2.048 Sortino
1.95% Down SD
5.78% Up SD

---

Modified PP - Volatility Balanced

23.14% TSM
37.47% LTGB
25% 2 Year ST Treasury
14.41% Gold

9.22% CAGR
0.536 Sharpe
2.340 Sortino
1.89% Down SD
6.15% Up SD

---

Basically the Volatility Balanced portfolios have about 0.3% CAGR less than the original portfolios.

The Down SD is lower and the Sortino Ratio is higher which means that the yearly returns of the portfolio has less downside risk.

From these results, I think it's worth thinking about volatility balancing when putting together a permanent portfolio.

[rickb]

As Craig mentions above,

Also be aware that the optimum asset allocation based on a backtest is going to change each year as the markets move and is highly time dependent. What is the optimum allocation today won't be the same one five years from now.

Optimizing your current portfolio based on even a long backtest is like driving your car by looking in your rearview mirror.  It works great if you're driving on the same roads you drove on before, but one thing that's nearly guaranteed is the future is not going to be the same as the past.  Relative to this optimization in particular, weighting long term bonds at 37% at the current moment is saying you think the next 30 years will be more or less like the last 30 years where long term bond rates dropped from about 12% to about 4%, see http://fixedincome.fidelity.com/fi/FIHi ... ield.  It seems to me the chances of long term bond rates dropping from 4% to -8% over the next 30 years are less than or equal to 0.

[Sam]

The only thing I used to come up with these weighting are the relative volatility of each asset. I did not optimize the weightings based on historical returns.

I agree that the probability of rates going from 4% to -8% is nearly impossible, but these weightings balances out the ups and downs of the individual components to each other. Wasn't that that point of the choosing assets with low correlations in the first place?

[Sam]

In Simba's spreadsheet, on the "Data_72_09" tab, there are the Standard Deviations of each asset listed on row 49.

I was actually calculating this myself using the excel stdev function on the data in rows 66 to 103 but just now realized that the returns being used in the portfolio calculations were in rows 8-45. I realized this when I was comparing your Std Dev numbers with mine. Anyways, the numbers were close enough.

I used the method you suggested in the other thread.

Std Dev numbers:
21.12 SCV
32.58 EM
11.59 LTGB
30.09 GOLD

After setting EM to 1, the ratios are:
1.54 SCV
1.00 EM
2.81 LTGB
1.08 GOLD

Adding it up I get 6.44.

Then I convert to portfolio weights with the assumption that cash (or 2 Year ST Treasury in my case) is 25%

1.54/6.44 * 75% = 17.978% for SCV
1.00/6.44 * 75% = 11.655% for EM
2.81/6.44 * 75% = 32.750% for LTGB
1.08/6.44 * 75% = 12.617% for GOLD
and 25% for 2 Year ST Treasury

11.19% CAGR
0.761 Sharpe
2.938 Sortino
2.19% Down SD
6.21% Up SD

---

Just out of curiosity, I calculated the weights for a portfolio with 0% in cash.
23.970% SCV
15.539% EM
43.667% LTGB
16.823% GOLD

12.35% CAGR
0.729 Sharpe
2.545 Sortino
3.05% Down SD
7.64% Up SD

It has higher returns, but also higher risks compared to the portfolio with 25% 2 Year ST Treasury.

[Sam]

Optimizing your current portfolio based on even a long backtest is like driving your car by looking in your rearview mirror.  It works great if you're driving on the same roads you drove on before, but one thing that's nearly guaranteed is the future is not going to be the same as the past.

Volatility balancing is more of an art than a science.  We can however estimate likely general volatilities as the PP does in estimating likely general correlations.

Gold for instance historically has sometimes been highly correlated with stocks and at other times highly inversely correlated with stocks.

Volatilities are somewhat similar in that ST's are likely to be less volatile than LT's which in turn are less volatile than stocks.  Gold is a form of Forex play and equally might maintain historic levels of volatilities in the forward time direction.

Whilst we can't predict the exact figures, generally we can make assumptions as to likely possible volatility comparisons.

Volatility balancing is just a method to have the assets weighted to levels where the capital losses in one are more equally balanced with capital gains in another, which helps smooth out the overall ride.  The historic evidence suggests that such a smoother ride whilst depleting rewards, doesn't reduce rewards that much overall.

If stocks are perhaps 3 times as volatile as LT's and there's a degree of inverse correlation between stocks and LT's, then having stocks up perhaps $3000 and LT's down -$1000 might subsequently encounter a reverse case of stocks down $3000 and LT's up $1000 in the conventional PP case.  In contrast the volatility balanced approach might have stocks up $1000 and LT's down $1000 at one time and then stocks down $1000, LT's up $1000 over another time.  That smoother ride (lower combined volatility) adds to compound benefits in a manner similar to how a 10% average yearly gain with a 20% standard deviation produces less annualised reward than a 10% average gain with a 10% standard deviation.

If you have individual assets that all achieve the same longer term reward then rebalancing using constant value (weighting) type methods (adding to those that are lagging, reducing from those that are leading) will generally provide a higher reward than not rebalancing.  Whilst the equal weighted PP might add more in rebalance benefits through having more disparate volatilities, that gap is somewhat closed down by a higher (relative) annualised gain instilled by the lower overall volatility approaches case uplifting compound average gains.

The other benefit of the balanced volatility case is that you're much less likely to overpay for the set at any one time.  In the case of Japan for example there was a period in time when the PP had performed relatively well, but had extended into over-bought (relatively expensive) territory which was followed by relatively poor performance subsequently.  The balanced volatility approach in contrast was much less over-priced in comparison.  A significant factor in overall long term rewards is the price paid for the investment in the first place.

Thanks Clive! That's a great explanation of why I was interested in volatility balancing the portfolio. Using assets with low correlation, I wanted to "normalize" the size of the returns so that they balance each other out, which I hoped would smooth out the returns of the total portfolio. It seems to be working by looking at the results in Simba's spreadsheet.

I was looking at the Decision Moose webpage you mentioned today. It looks like the perfect thing to use in a variable portfolio. I wonder how far into the future the DM signal will be available though.