Useful Links

Sam · 08-09-2007, 07:01 PM

I played around with mixing and found that the combo 61% US Small Value and 39% LT Corporate Bond gives the highest SWR. 100% success at 4.21%, for a 30 years plan.

Have you found a better combination in FIRECalc?

CCdaCE · 08-09-2007, 08:42 PM

That's what worked in the past, Wellington's proved that.

Now what's gonna work in the future?

-CC

FIRE'd@51 · 08-09-2007, 08:43 PM

Very interesting.

I assume you only tried combinations of two asset classes, right?

If my math is correct, that would be 28 combinations of asset classes. Then you would have to test each of the 28 for the optimal mix to find the overall best. Is this what you did?

FIRE'd@51 · 08-09-2007, 08:48 PM

Quote:

Originally Posted by CCdaCE

That's what worked in the past, Wellington's proved that.

Wellington uses Large Value.

CCdaCE · 08-09-2007, 08:58 PM

Quote:

Originally Posted by FIRE'd@51

Wellington uses Large Value.

Doh! Sorry, I misread US Value.

-CC

haha · 08-09-2007, 09:54 PM

[quote=Sam;545256]I played around with mixing and found that the combo 61% US Small Value and 39% LT Corporate Bond gives the highest SWR. 100% success at 4.21%, for a 30 years plan.

No, but I plan extensive data mining operations just in case something turns up.

Ha

Sam · 08-09-2007, 11:14 PM

Quote:

Originally Posted by FIRE'd@51

Very interesting.

I assume you only tried combinations of two asset classes, right?

If my math is correct, that would be 28 combinations of asset classes. Then you would have to test each of the 28 for the optimal mix to find the overall best. Is this what you did?

No, I have not tried all possible combos. That would take years, unless I rewrite the program. Hey, may be we can submit a suggestion to Dory?

cute fuzzy bunny · 08-10-2007, 12:05 AM

All you need is the highest returning stock component (which is small cap value), and the highest returning bond component, which is LT corporate bond.

Then all you have to do is find the optimum balance.

Then all you have to ask is if you can manage the gut wrenching volatility, and if those asset classes will perform going forward like they have previously.

Sam · 08-10-2007, 01:47 AM

Quote:

Originally Posted by cute fuzzy bunny

All you need is the highest returning stock component (which is small cap value), and the highest returning bond component, which is LT corporate bond.

Then all you have to do is find the optimum balance.

If the above is true, then I have already found the optimal combination.

Quote:

Originally Posted by cute fuzzy bunny

Then all you have to ask is if you can manage the gut wrenching volatility, and if those asset classes will perform going forward like they have previously.

Hmm... Using the same logic, then what good is FIRECalc at all? Whether you choose this particular mix, or the default, or any other combination, the same question remains.

FIRE'd@51 · 08-10-2007, 06:01 AM

Quote:

Originally Posted by cute fuzzy bunny

All you need is the highest returning stock component (which is small cap value), and the highest returning bond component, which is LT corporate bond.

Then all you have to do is find the optimum balance.

This would only be true if the correlations (covariance) were the same.

CCdaCE · 08-10-2007, 07:44 AM

Quote:

Originally Posted by cute fuzzy bunny

All you need is the highest returning stock component (which is small cap value), and the highest returning bond component, which is LT corporate bond.

Then all you have to do is find the optimum balance.

Then all you have to ask is if you can manage the gut wrenching volatility, and if those asset classes will perform going forward like they have previously.

Are we talking about the same thing that Nassim Taleb has already caused an uproar with? The "smidgen of really risky stuff" and mostly bonds?

Everything is starting to become intermingled that it's all becoming so clear to me.

-CC

cute fuzzy bunny · 08-10-2007, 09:36 AM

Quote:

Originally Posted by FIRE'd@51

This would only be true if the correlations (covariance) were the same.

That would require having the magic tablet of correlations. I'm not seeing a lot of predictable migration over the last few decades. Seems to be a lot of unpredictability.

Sam - firecalc isnt really a tool for finding optimal future asset mixes. It simply cant know that. Its just telling you what WAS optimal in the past. And for those purposes, yes, you found the optimal combination.

What the tool has been and will be good for is taking your portfolio size, your spending demands, and your overall plan and telling you if it would have worked throughout history.

Persistence is a pretty good predictor. Not perfect, but pretty good.

Where theres some problem is in the volatility and risk assumed for return given. Long term bonds can eke out a little more return, but as Bernstein quite ably charted in The Four Pillars, you're taking on a shitload of risk and volatility for what is a somewhat nominal extra slice of return. Not worth it.

I think you COULD do fine with SCV as a primary asset class...but you might want to mix a little something else in there to substantially reduce risk without dampening returns too much.

FIRE'd@51 · 08-10-2007, 10:00 AM

Quote:

Originally Posted by cute fuzzy bunny

That would require having the magic tablet of correlations. I'm not seeing a lot of predictable migration over the last few decades. Seems to be a lot of unpredictability.

Those correlations, whatever they were, are implicitly contained in FireCalc. I was challenging your assertion that Sam could have, a-priori, picked the highest return stock component and the highest return fixed income component without doing the analysis, and know he had the optimal solution.

cute fuzzy bunny · 08-10-2007, 10:09 AM

Yet he did.

Small Cap Value is the highest returning stock component. Long term corporate is the highest returning bond component. Historically they havent had a lot of correlation and acted as decent diversifiers.

Anyone with a modicum of investing knowledge would be able to answer that question correctly, at least academically.

The problem being, he's chosen the historical optimal solution. Not necessarily the future one.

Perhaps I'm not understanding what it is that you're 'challenging' and why its important.

FIRE'd@51 · 08-10-2007, 10:52 AM

Quote:

Originally Posted by cute fuzzy bunny

Perhaps I'm not understanding what it is that you're 'challenging' and why its important.

Maybe I misunderstood his answer to the question I posed in post #3. I thought he had tried pair-wise combinations in his analysis to come up with the two asset classes US Small Value and LT Corporate, and had eliminated the other equity classes and fixed income classes through trial and error.

The point I was trying to make to you (and maybe it's stating the obvious) is you couldn't just pick the highest returning equity class (perhaps US Micro Cap?) and combine it with the highest returning fixed income class without trying them all in FireCalc first because the historical covariances were different. I thought he had run FireCalc to eliminate the other equity classes.

Perhaps I totally misunderstood his methodology. :confused:

Sam · 08-10-2007, 10:55 AM

Quote:

Originally Posted by cute fuzzy bunny

The problem being, he's chosen the historical optimal solution. Not necessarily the future one.

CFB, I think everyone understands that the past is the one and only thing FIRECalc (or any other calculator) depends on. So, again, whether you pick this particular combo, or the default combo, or any other combo, the same logic applies. Since the future is unknown to all, the past is all we have.

twaddle · 08-10-2007, 11:15 AM

Hey, Sam. I did a similar experiment a while back, but for 40 years.

I have no idea what the bunny is rambling about, but the solution is not obvious since volatility matters when you're withdrawing. Not only that, but you'll get different "optima" for different withdrawal time periods.

For example, over a 40 year withdrawal period, try this:

61% ScV + 39% 1-month treasuries gives a 100% SWR of 4.46%

60% microcap + 10% LcV + 30% 1-month treasuries gives a 100% SWR of 4.87%

Raddr did a study a while back that indicated that adding commodities into the mix could bring the SWR up to around 7%....