The Great Debate!

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 10:00 am

Herbert,
Perhaps we can agree on the following

If a person believes that the excess return will drop to zero, and they are a risk neutral individual (in other words they are indifferent to a 7% guaranteed return and a 50% chance of a loss of 3% and a 50% chance of a 17% gain) than they might not want to own CCF

But that also depends the assumption that the negative correlation to other portfolio assets will disappear as will the high volatility.

Now if you believe the excess return disappears and the negative correlation to other portfolio assets disappears then there is no reason to consider owning them. That is clear because the insurance nature of the asset goes away.

That to me are the conditions for not considering adding CCF. Or if you think the risks of those things occuring are high. And also you ignore the possibility for the other side to occur, the excess return rises and/or the negative correlation and high volatlity increase.

Personally I don't think these are LIKELY. Possible, yes. But so are the more favorable outcomes. For example, from 1970-2007 the GSCI actually outperformed the S&P 500 by 12% to 11.1%.So not only did you get the benefits of the negative correlation and high volatlity but you even had higher returns BEFORE that benefit. Could that happen again, sure. None of us has clear crystal balls, and in very dangerous world we live in, adding some portfolio insurance seems logical to me

Finally, you might consider the following, it seems prudent to me for everyone to own at least a small amount of CCF simply as hedge against their own personal expenditure risks. You can have for example low inflation (as we have had) but oil and perhaps food rising a lot. I am sure that there are lots of consumers that wished they had some hedge against commodity price increases.

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Mon Feb 04, 2008 10:06 am

Also, Rick both acknowledges the diversification (saying he thinks it is 2%) and denies it when he says the real return to CCF is expected to be zero. As they say "enquiring minds want to know which it is." Either it is 2% in his view (in which case he should retract his other statements and apologize for the error) or it is zero (and retract his 2% estimate and explain why that is the case).
Nice try at twisting things around. If there is a diversification benefit in the future, that benefit cannot be known today. So, why are you 'counting on it'?

I honestly cannot believe you are recommending that people give up the good and reliable return of stocks because one research team (Erb and Harvey) created a commodities trading strategy that showed a 4% real return in back-testing. They said that return was the "most reliable" part of the return from the trading strategy. But, what does "most reliable" mean? It does NOT mean "reliable." It does not mean a "sure thing". But you are taking that single statement claiming a MPT benefit is a guarantee. That is a dangerous presumption.

Your entire argument is based on the hope of earning an excess returns from MPT, not from investing in commodities in isolation. You need to see the forest through the trees. To RELY on MPT to make up for the lower returns from commodities is not a good idea. You can say it is until you are blue in the face, but IMO, it is foolish to give up the reliable real returns of stocks to hope that an MPT benefit by adding CCFs will make up the opportunity cost.

Why do Fama/French say no to CCFs? Why is DFA against adding CCFs as you recommend? Why have you refused to talk about the other side of the academic story?

Rick Ferri
Last edited by Rick Ferri on Mon Feb 04, 2008 10:20 am, edited 1 time in total.

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 10:18 am

Rick
First you continue to spout ridiculous comments. Like your latest accusation. You asked me to explain about FF and their views and I gave a very detailed explanation. On other hand you refuse to answer the many questions I have put forward

Second, I twisted nothing. You have made two contradictory statements. First you state that there is no expected diversification return then you clearly said it was 2% (if my memory is correct on that figure). So I ask again,which is it? Now we both know we cannot know exactly what it is going to be any more than we know what the ERP is going to be, but we make assumptions based on historical evidence and current valuations.

Third, there is no guarantee about the future of anything, including the returns to stocks which you seem to be treating as certain as well. We all have to make assumptions about expected returns. We have about 50 years of data on CCF excess returns and you have not presented a single argument in favor of why one should believe that all of a sudden the excess return will go from 3-5% to zero

Finally you comments about MPT amaze me, I won't even comment on them because IMO they are so absurd that they don't warrant comment. I have explained why they are wrong already so no need to comment further.

I should have stayed with my decision not to continue this debate because of both your refusals to address the issues and your absurd accusations and other comments. But you asked me about FF and I decided to respond, which btw I had already done long ago. I will now cease responding though.

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Mon Feb 04, 2008 10:28 am

Larry Said:

[Rick] You asked me to explain about FF and their views and I gave a very detailed explanation. On other hand you refuse to answer the many questions I have put forward
Larry, I went back and searched all your posts for a 'very detailed explanation' of why Fama/French and DFA are against your idea of adding CCFs to portfolio and why those very intelligent people are wrong. I could not find anything referencing this information. Can you re-post them please.

Rick Ferri

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Mon Feb 04, 2008 10:34 am

Finally you comments about MPT amaze me, I won't even comment on them because IMO they are so absurd that they don't warrant comment. I have explained why they are wrong already so no need to comment further.
Sorry, I do not understand that answer.

MPT is a benefit when it occurs, and IMO, to say investors should rely in it for return is the 'absurd' position.

Let's end this pre-debate and agree to disagree.

Rick Ferri

wearethefall
Posts: 248
Joined: Sun Jul 29, 2007 6:01 am

Post by wearethefall » Mon Feb 04, 2008 11:20 am

The new edition of Roger Gibson's Asset Allocation has a good section on commodities. He comes down firmly in favour of them, and his arguments convinced me.

HerbertSitz
Posts: 684
Joined: Fri Feb 01, 2008 8:52 pm

Post by HerbertSitz » Mon Feb 04, 2008 12:00 pm

larryswedroe wrote:Herbert,
I beg to differ on all counts--except I agree that E&H are smart people.
Interesting that you call them smart but then disagree with them regarding crucial points, like what level of expected excess return is required from CCFs to make them desirable additions to the portfolio.
larryswedroe wrote:That IMO is wrong perception of rebalancing.
Not sure you're in a majority there with your opinion. For example, Bill Bernstein says, "If the markets were truly efficient, then you shouldn't be able to make any money rebalancing. After all, rebalancing is a bet that some assets (the lowest performing ones) will have higher returns than others (the highest performing ones)." Four Pillars at 290.

I understand that part of the reason behind rebalancing is to maintain your AA. So what? Part of the reason is also to capture the diversification return. So with stock and bond assets there are (currently) two reasons for rebalancing. Except, contrary to what you say, most people think the extra return from rebalancing is anomalous, so that it would disappear in a truly efficient market, leaving maintenance of AA as the only benefit from rebalancing.

Can you provide an uncontroversial theoretical reason why returns from rebalancing have persisted in the stock and bond markets? Can you give me a theoretical reason why that benefit would not diminish if more and more participants employed more and more aggressive strategies to capture the rebalancing return?

My main concern with your overall attitude toward CCF funds another that Rick has expressed. Analyses of real world data for CCF funds aren't here yet. Why not advise people to to wait until the real world data is analyzed? What I hear you saying is "Well, the historical data is favorable. There's not enough real world data in to make a judgment on that but I don't see any good reasons for it being different in the future so go for it!" Why not advise people to wait a few more years to see more real world data? Why is it not good to be cautious? Why not live with status quo until we have more definitive proof?
larryswedroe wrote:2) As I have stated an asset class with negative correlation can have return below riskless rate and still add value. The answer depends on how much below the riskless rate and how negative the correlation and how high the volatility and how much you add. As I have said (unfortunately I did not keep the example) I have seen cases with actually negative returns to an asset and its inclusion improved the returns of the portfolio--very small amount.
Right, by now I understand your opinion here. And I agree that it's possible that depending on the degree of negative correlation and degree of volatility it could be beneficial overall to add a lower-than-risk-free-rate-perofrming asset class to the portfolio. I'll just point out again that Erb and Harvey seem to think the degree of non-correlation and volatility of CCFs is insufficient to bring an overall benefit to the portfolio when they have zero expected excess return. They've apparently analyzed the historical data pretty well and advise not holding CCFs as permanent portfolio asset if they're projected to have only a riskless rate of return.

I guess we're just rehashing our differences over and over, so the time comes for me to stop.

I do come back to one of Rick's concerns, though. Even if I believed everything you say, it seems like the responsible thing is to admit that unreservedly advocating CCFs is somewhat reckless. The proviso should be added on, "Of course, my conclusions are drawn from historical data and while I theoretically see no reason for the data to change in the future, the management in CCF funds is adding variables to the mix that didn't exist in the period the historical data comes from. So it might well be wise to wait at least a few more years to see what the real world data for CCF funds looks like."

User avatar
craigr
Posts: 2696
Joined: Tue Mar 13, 2007 6:54 pm

Post by craigr » Mon Feb 04, 2008 1:18 pm

...
Last edited by craigr on Wed Feb 06, 2008 3:05 am, edited 4 times in total.

HerbertSitz
Posts: 684
Joined: Fri Feb 01, 2008 8:52 pm

Post by HerbertSitz » Mon Feb 04, 2008 1:25 pm

larryswedroe wrote:Herbert,
Perhaps we can agree on the following

If a person believes that the excess return will drop to zero, and they are a risk neutral individual (in other words they are indifferent to a 7% guaranteed return and a 50% chance of a loss of 3% and a 50% chance of a 17% gain) than they might not want to own CCF
Larry --

Is this "insurance" point anything more than saying you have to be aware of your position on the x axis of the return/risk plane, as well as the y? Isn't everybody working with that understanding?

I can only assume that Erb and Harvey concluded that the portfolio without CCF funds stands at more efficient point (or curve) in Return/risk plane when CCF funds have zero return. So an indvidual would in that case actually prefer the portfolio without CCF regardless of their level of risk aversion. That is, E&H believe there's a point on frontier curve of portfolio w/o CCFs that has equivalent return but lower s/d than w/CCFs? Thus, they believe you can get the same insurance cheaper w/o using zero return CCFs. Maybe there's some other explanation for why they assign zero allocation to zero return ccf's, but if you assume they're ignoring risk then you must think they're stupid.

Do you have web link to the G&R paper.

One last question: I'm not sure why, but Rick raised the issue of salesmanship. In the interest of full disclosure, do you or your firm receive revenue of any kind tied to CCF funds, whether in form of commissions, management fees, or any other way? [Edit: I just looked at the other thread you started on this issue and from that I assume Rick's comment may have been directed to Ibbotson, and to the high loads and fees of CCF funds generally. Seems like legitimate concern to raise.]
Last edited by HerbertSitz on Tue Feb 05, 2008 2:28 pm, edited 7 times in total.

User avatar
oneleaf
Posts: 2369
Joined: Mon Feb 19, 2007 5:48 pm

Post by oneleaf » Mon Feb 04, 2008 1:41 pm

HerbertSitz wrote: Have people actually read the Erb and Harvey paper? It seems to me Rick's position is closer to Erb and Harvey than Larry's. Throughout the paper Erb and Harvey raise cautions about extrapolating historical returns into the future, the same caution I hear Rick giving. Their final conclusion is that CCF's should _not_ play a prominent strategic role as a a permanent portfolio asset, but that they may have a tactical role as an asset to move in and out of depending on the situation. Here's the final paragraph of the paper:

"Overall, it is our sense that the strategic [emphasis added] case for long only commodity futures has been
oversold. Our results suggest that the traditional rationales for including these assets in a
diversified portfolio are suspect. Focusing on the role that commodity futures play in providing
commodity price insurance, our research suggests that commodity futures should be reserved as a
tactical source of alpha in the investment manager’s arsenal."
at p. 36 of their published paper (sorry I still can't post links)
From the "unabridged version" published Jan 12, 2006, I can't find that paragraph anywhere. It sounds like, from reading the unabridged version of The Tactical and Strategic Value of Commodity Futures, available on the SSRN website, they make no such conclusion.

I know you can't post links, but can you give some indication of where I can find the version you are quoting from?

Thanks.

User avatar
RobertH
Posts: 219
Joined: Thu Mar 01, 2007 11:35 pm
Location: Bay Area

Post by RobertH » Mon Feb 04, 2008 2:06 pm

The SD of individual commodity futures contracts is 30% and the average correlation among those contracts is 0.10. Plugging these numbers into Table 8 of Erb & Harvey's paper assuming 30 commodities shows a diversification return of about 4% (3.92% is the number in the table). If this is a trading strategy, it is simple, formulaic and based on hard statistics. That works for me.

Will arbitrageurs enter the commodities futures markets, driving correlations toward 1.0 and standard deviation toward zero, driving down the diversification return? Maybe, but until that happens sign me up for the arbitrageur camp.

So we have an investment that over the long haul is a little bit like TIPS: inflation protection (at least as measured by the increase in commodities spot prices), plus a real return equal to the diversification return. If there is a net return for providing insurance after backwardation/contango, so much the better.

Will the diversification return exceed the real return currently offered by TIPS? I think it's a good bet, but opinions differ. (And fwiw Ibbotson's MVO analysis suggested taking half of the commodities position out of TIPS for a 55/45 investor.)

Yes, CCFs have high volatility, but adding CCFs in sensible quantities lowers the overall volatility of the portfolio. Larry posted a thread yesterday giving an example where this is even true if they displace 1-year treasuries.

It's possible that an upswell in demand from investors for diversifying assets over the last few years lowered the risk premia for CCFs, possibly even driving certain commodities into contango, but if so commodity futures were not IMO the only asset class affected.

I respect the argument that investors who have a hard time maintaining a value tilt would also have a hard time maintaining a commodities position, but for informed investors who understand what they are buying and why, it seems to me that a commodities position can make sense.

Robert

HerbertSitz
Posts: 684
Joined: Fri Feb 01, 2008 8:52 pm

Post by HerbertSitz » Mon Feb 04, 2008 2:23 pm

oneleaf wrote:I know you can't post links, but can you give some indication of where I can find the version you are quoting from?
Sorry, they have a couple different versions posted. One was published, one they are apparently revising as work in progress. Can get them by going to Campbell Harvey's website at duke dot edu slash tilde charvey slash and then go to his page with "Downloadable research papers"

User avatar
oneleaf
Posts: 2369
Joined: Mon Feb 19, 2007 5:48 pm

Post by oneleaf » Mon Feb 04, 2008 2:33 pm

HerbertSitz wrote:
oneleaf wrote:I know you can't post links, but can you give some indication of where I can find the version you are quoting from?
Sorry, they have a couple different versions posted. One was published, one they are apparently revising as work in progress. Can get them by going to Campbell Harvey's website at duke dot edu slash tilde charvey slash and then go to his page with "Downloadable research papers"
thanks i'll take a look at it.

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 3:28 pm

Rick
My apologies
I wrote the response and was sure I posted it but for some reason it does not appear
So here it is. Again.

First, I have the highest regard for Ken French and certainly would prefer to be on the same side of a debate. In this case I am not because I am convinced by the data and unconvinced by his arguments, which I will go into. Having said that it also shows how absurd the comment was by Rick (and condescending) that I treat all academics as gospel.

Now, French made the presentation and I found problems with it.
a) He used monthly data to state that CCF were to volatile to act as inflation hedge. Now never have I seen monthly data used, typically annual data. Why he chose to use monthly data I don't know. But I don't care about monthly data. The correlation of CCF to inflation is not only positive but it increases in correlation the longer the maturity. And this also misses the issue that with negatively correlating assets high volatility is a good thing (if you use a small amount). I agree though that because of the volatility CCF is not a good monthly hedge of inflation.
BTW-the data also shows that CCF correlates not only with inflation but with the change in rate of inflation

b) French looked at the equity like return from the G&R paper and said that could not be expected--lucky outcome. Reason commodities have no expected real return. That however does not include the diversification return that explains the "excess" return found in the G&R paper. To be fair the Erb and Harvey paper was not out yet.

c)French stated the data used in the G&R paper (he did not btw that they are top academics he respects) was too short a period, you needed 50 years of data to be "sure"--that I found interesting since the G&R data was pretty close to 50 years and certainly far longer than the data set for the famous three factor model paper which was about 30 years.

d)French did not consider the portfolio benefits, the negative correlation with other assets-and that is just on the stock side. He did not consider the hedge they bring to nominal return bonds either. he just addressed the issues of the return and the volatility.(if my memory serves correctly)

so the bottom line is I respectively disagree with French's conclusion. Even though we completely agree that one should not expect backwardation (nor contango) to be the natural order of things.

As I said because of your comments I will not debate this-I will add only that you make a similar mistake on the equity issue here as you do on junk bonds when you stated that the worst case for junk is that they will have the same returns as investment grade. IMO there is hard to find a more incorrect statement than that. That outlaws risk. And if you truly believed it then it would not make sense to own anything but junk bonds--as the worst case is they have investment grade returns and best case is they do far better. So why add anything else? The same issue applies in similar manner here.

Sure if I knew equity returns were going to be a consistent 6% a year then there is no need for CCF and if I knew what inflation would be there would be no need for TIPS. Now, sorry to say this but the academic literature suggests TIPS should be owned even if the expected return is significantly less than that of nominal return bonds--because of the insurance nature they play --hedging unexpected inflation. The same is true of CCF-even though they might have lower expected returns than stocks doesn't mean you should not own them as hedge against the risks that the ERP will be less or dramatically less. And the same is true for their value hedging the bond side (assuming it is nominal bonds you own). Since CCF hedges unexpected inflation then you can take more duration risk than if you don't own them. That certainly should not be ignored as the incremental returns from the bonds adds to portfolio returns. It is the insurance/hedge nature of commodities that is the reason to own them. They not only have negative correlation to US stocks, but int'l stocks and nominal bonds. And the correlations tend to turn negative when needed most. That is why they make such excellent insurance. Now if you even pay say 1% more for a fund, and it makes up even 5% of a portfolio that raises your total costs of the portfolio by 5bp. The benefit of the insurance hedge is far more than that.


I hope everyone finds the above helpful

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 3:43 pm

herbert
As everyone else knows I receive no incentives to sell any investment products. I have no incentive to advise people to do anything that is not in their best interest
Also I invest the same way I advise, and as I have stated I own the PIMCO fund and have since 03.

As to the Erb and Harvey comments--
An asset can have less than riskless return and still be efficient addition to portfolio. The answer is it depends on
a)how much below
b) the correlation (the lower the better, or the more negative the better)
c) the volatility (if negative the higher the better)
it is not simple black and white answer

The existence of the insurance industry demonstrates that assets with negative correlation have utility even if they have less than riskless return. We buy insurance (hedge) against our labor capital with life insurance and disability insurance because their returns have perfect negative correlation with our labor capital. We buy longevity insurance in the form of immediate pay annuities as its return is perfectly negatively correlated with life span (both products transfer expected profits to the insurance companies). Similar concepts apply to health insurance, travel insurance, etc.

So I would respectfully disagree with the conclusion by Erb and Harvey--unless you added other caveats they did not. But the important thing as we discussed is there is no logical reason I can think of that one should expect the excess return to be 0. Again they found it to be between 3-5% depending on what index one used. And as I said it seems to me if anything the odds favor a higher number. But I don't count on it. BTW- I also believe that the odds of the excess return being above zero are far greater than the odds of the ERP being above 6%. But just my opinion. The key is not whether they will be but, there is virtual certainty we will go through periods where the ERP will be lot lower, it will be negative , and it might even be well below that over our lifetime, and it might even be negative (see Japan since 1990). That IMO is the mistake, or one of them, that Rick makes, treating the likely as if it was certain. If it was certain, the 6% ERP I would not recommend CCF, there would be no need. On other hand if it was certain it would not be 6%--it would be 0, as there would be no risk (:-))

Also note that in withdrawal phase the volatility of a portfolio becomes extremely important--and getting rid of the fat left tail, or reducing it is important. Failing to do so if the equities Rick expects to return 6% don't ( or do but not in right order) then you can easily have a portfolio with lower return but lower SD provide a portfolio that will not outlive you while the higher returning one won't. Great example of this is the person who retires in 1973. That is one real example of Rick's statement about not being able to spend SD is based on false concept.

We have run MC simulations using conservative estimates of CCF returns (like around 5%, which is well under their return even from 91, let alone from 1973) and find that CCF addition positively impacts the odds of success of a portfolio.

I hope that the above is helpful

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 3:49 pm

Rick
knew I was not crazy--just wasted time duplicating post
Here is my response to your question-your challenge

http://www.diehards.org/forum/viewtopic.php?t=12370

HerbertSitz
Posts: 684
Joined: Fri Feb 01, 2008 8:52 pm

Post by HerbertSitz » Mon Feb 04, 2008 4:32 pm

larryswedroe wrote:herbert
As everyone else knows I receive no incentives to sell any investment products. I have no incentive to advise people to do anything that is not in their best interest
Also I invest the same way I advise, and as I have stated I own the PIMCO fund and have since 03.
Larry -- Thanks for clarifying. Obviously not everyone knows, so I'm glad you stated it again.

I just looked at the other thread you started on this issue and from that I assume Rick's comment regarding salesmanship may have been directed at Ibbotson, or just generally to salesmanship in financial industry that arises around funds that have the high loads CCF funds generally do.

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Mon Feb 04, 2008 4:45 pm

Larry,

I did not check the simultaneous discussion going on about CCFs. Thanks for answering that post also. I am not going to argue the points that French made, but I will reitirate that the use of CCFs in a portfolio is not universally accepted by all leading academics.

The only other issue I would address is this one:
negatively correlating assets and high volatility is a good thing (if you use a small amount).
Two items. 1) using a small amount of CCFs is only useful if there is a return benefit, and that is questionable, i.e. Ken French. 2) Commodities are not always negatively correlated with stocks. The correct statement is that they have negative correlation at times. That is clear when viewing a rolling 36 month correlation chart. Rolling correlation looks at correlations over independent periods rather than taking one static view based on one specific period.

Rick Ferri

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 7:15 pm

Rick since you said it nicely I will respond-
First, I never said that all academics agree, in fact the issue of French's view has been discussed before. However, there are many that do agree.

Second, reasonable people can disagree, my problem with French is we disagreed over interpretation of the data and what it means. No problem there. I stated his view and why I disagreed. In your case I have had different issue. I believe your views are based on the wrong set of ideas to make decisions on. I stated why and I am certain French would agree with me that the reasons you give are wrong (the idea to consider asset in isolation as one example).

BTW-on that subject, if you don't like the idea of taking it from high expected returning stocks take the allocation away from low expected retirurning bonds and you still get the same result, more efficient portfolios. And there is no doubt that CCF and bonds are negatively correlated.

Third, of course I know that correlations are not static. we have discussed this before. I have very clearly even discussed how the correlation with equities and CCF goes up occassionally (even stating the type of period when it occurs). However, the right way to look at this issue is to look at the long term average correlation and also when the tendency of the correlations is to rise and fall. Now with bonds the correlation is great, so they act as great hedge of nominal bond risks. You have completely ignored this in the discussion --again because you make mistake of thinking of things in isolation. But they also act as very good hedge of equities risk ON AVERAGE--when stocks down they average up 23% and up 75% of time--that is excellent record. Bonds down they average up 30% and up 100% of time. thus as I have stated you can take more duration risk with your nominal bonds if you also own some CCF as hedge. (French also ignored this issue totally--he never looked at it, at least to my knowledge)

What is also important is to look at whether something makes sense--do the correlations move logically. And in this case they clearly do because CCF is positively correlated to inflation and stocks and bonds are negatively correlated. The issue for stocks and CCF is some event risks correlated negatively (say wars or oil blockade) and some positively (9/11).

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Mon Feb 04, 2008 8:50 pm

BTW-on that subject, if you don't like the idea of taking it from high expected returning stocks take the allocation away from low expected returning bonds and you still get the same result, more efficient portfolios. And there is no doubt that CCF and bonds are negatively correlated.
The correlation between bonds and commodities is not negative. It also varies over time. See the top figure on page 9 of Asset Class Correlations. It is the rolling 36 month correlation between the CRB and stocks, and the CRB and bonds. There was positive correlation in the late 1990s as bonds did poorly and commodities did poorly.

Nonetheless, I agree that if a person were to invest 10% in CCFs, both bond and stock allocations should be reduced. For example, in a 50% stock and 50% bonds portfolio, 5% would come from both stocks and bonds so that the allocation is 45/45/10. That was a 'debate' item I believe.



Rick Ferri

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Mon Feb 04, 2008 9:21 pm

Rick
Correlations refers to the long term average of the correlation, over monthly or quarterly or annual terms. We all know that the drift--but the important points are just two

A) what is the long term average
b) when do they tend to rise and when do they tend to fall.

If they tend to rise during periods of crisis when both would produce below average returns that is bad, if they tend to fall during crisis so that when one produces above average returns while the other is producing below average returns that is good.

There is no asset classes with constant correlations thus there is risk of drift and we all know that. So we build portfolios based on the long term averages and when the correlations tend to be high and low--because that impacts the dispersion of returns.

In the case of CCF they are significantly negative over the long term (and if you are an investor than you should only be concerned about the long term and the other issue I mentioned) and also they tend to be negative when needed most.

User avatar
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

CCFs

Post by Robert T » Mon Feb 04, 2008 11:04 pm

.
Additional thoughts FWIW.

Some of my reasons for excluding collateralized commodities futures (CCFs) [as listed in one of my earlier posts – with some small additions].
  • (i) Limited space in tax-advantaged accounts. Most of the returns of CCFs are from collateral returns and internal and external rebalancing – leading to low tax efficiency. IMO this is large enough to erode any portfolio mean-variance benefits.

    (ii) Unexpected inflation protection can be provided by other instruments: Inflation protected securities provide, by definition and construct, protection against unexpected inflation – so CCFs do not provide unique protection against unexpected inflation.

    (iii) There are other (although less precise) instruments to hedge event risk: The ability of CCFs to hedge (non financial) event risk is IMO one of the strongest arguments for their inclusion in a portfolio. However, adding emerging market equities (which includes resource rich countries) can provide some protection to the longer-term impacts of particular events (e.g. wars in oil rich states) although with some lag (one year correlations of EM stocks with CCFs is about 0.11, five year correlations increase to about 0.5). US treasuries can hedge financial event risk (e.g. 1998 financial crisis in Asia).

    (iv) Lower expected future return. As per the Bernstein’s article, the demand for commodities futures as insurance against unexpected inflation has increased relative to demand for commodities futures as insurance against unexpected deflation. This has possibly reduced the historical backwardation and hence expected return. As a result the previous back-tested results may not be indicative of the future.

    (v) Lower expected rebalancing return. As trade barriers decline, and trade increases, with globalization, it could be argued that the volatility of commodity price is expected to decline (holding everything else constant). As a result the rebalancing return may be lower with a more integrated global economy, than with the less integrated economies of the past (which shows up in the historical backtest). As above the results suggest that the previous back-test results may not be indicative of the future.
What do the simulated data say when returns and SD are modified downwards? [from table below] The baseline impact of CCFs in the historical backtest when the allocation was taken from stocks was reduced volatility of portfolio returns, but with no return enhancement (using my portfolio benchmark as the baseline with data from 1970-2006). Adding 5% GSCI (taken from stocks) did not change the portfolio annualized returns but lowered their volatility. All modifications to the historical CCFs series (lowering returns, lowering volatility, and raising correlations) reduced portfolio returns relative to the stock:bond baseline – although still retained a higher Sharpe ratio (I did not add TIPS or EM stocks to the stock:bond portfolio to try to partially mimic the inflation protection and event risk hedge attributes of CCFs). Interesting results nevertheless. Adding CCFs seems to be more a volatility reducer, leading to a higher Sharpe ratio, but likely at the expense of returns (at least in the context of my value and small cap portfolio benchmark), and this result seems fairly consistent across the range of different assumptions used. The impact of the CCFs allocation when taken from bonds was both reduced volatility of portfolio returns, and enhanced returns, although the portfolio efficiency effect seems to be lower than when the allocation was taken from stocks (although the impact was still positive in all cases except when the CCFs return is assume to be zero).

I agree that the opinions on the value added of CCFs in the finance profession today seems to vary widely, more so than almost any other ‘asset class’.

Code: Select all

1970-2006
                                          Annualized   Standard
                                            Return     Deviation  Sharpe

75:25 stock:bond portfolio (no CCFs)         14.14       13.44     0.663

5% CCFs (GSCI) taken from stocks (baseline)  14.14       12.23     0.716 
….with the following assumptions:
  CCFs with 0% annualized return             13.50       12.23     0.665
  CCFs with 20% lower volatility             14.05       12.29     0.707
  CCFs with 0 correlation with stocks:bonds  14.04       12.62     0.690   
	
5% CCFs (GSCI) taken from bonds (baseline)   14.47       13.07     0.703 
….with the following assumptions:
  CCFs with 0% annualized return             13.83       13.06     0.655
  CCFs with 20% lower volatility             14.38       13.13     0.694
  CCFs with 0 correlation with stocks:bonds  14.36       13.45     0.679   
Definitions:

- CCF with 0% annualized return: Simulated series with 0% annualized return but with the same volatility and correlation with stocks and bonds as the GSCI.

- CCF with 20% lower volatility: Simulated series with 19% SD (rather than 24% for the GSCI), but with the same annualized return, and correlation with stocks and bonds as the GSCI.

- CCF with 0 correlation with stocks:bonds: Simulated series with 0 correlation coefficient (rather than the negative correlations of the GSCI), but with the same annualized return and volatility as the GSCI.

Robert
.

User avatar
Random Musings
Posts: 5339
Joined: Thu Feb 22, 2007 4:24 pm
Location: Pennsylvania

Post by Random Musings » Mon Feb 04, 2008 11:19 pm

Robert,

Very interesting and like the table summary. Big picture- over that time frame the changes (returns, volatility and Sharpe) are a little different - but who knows what will happen over the next XX years? Bond yield curve different than 1970, relative valuations probably different.

My take - 3-5% in CCF's probably won't hurt, but would the average investor even notice? After the current commodity run, I'd at least wait for weakness if I was starting CCF's. And I'm still on the fence.

RM

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Mon Feb 04, 2008 11:23 pm

Rick
Correlations refers to the long term average of the correlation, over monthly or quarterly or annual terms. We all know that the drift--but the important points are just two:

A) what is the long term average
b) when do they tend to rise and when do they tend to fall.
I completely disagree with "A". Correlations have tendencies, but there is DEFINITELY not a consistent long-term average. In other words, no long-term average would pass a T-test (a statistical test of the accuracy of the believed average).

As to "B", that is more accurate although not entirely sound. To say this with any certainly, you must find independent periods when like circumstances occurred, and then have enough evidence to assume it will happen that way again. But that is often difficult because things change and relationships change.

For example, there is supposed to be a tendency for oil to rise as stocks fall. However, over the last few years the price of oil and the price of stocks have both risen. Why? Because oil is not as important to GDP as it once was. It takes half the amount of energy to earn $1 dollar of GDP today than it did 30 years ago.

This divergence continued into 2008. So far stocks fell and commodities fell, including oil. That should not have happened according to your model. What you are missing is the tendency for 'tendencies' to change.

Rick Ferri
Last edited by Rick Ferri on Mon Feb 04, 2008 11:54 pm, edited 2 times in total.

mikenz
Posts: 780
Joined: Sat Mar 10, 2007 8:33 pm

Post by mikenz » Mon Feb 04, 2008 11:29 pm

Again, backtesting basically equates to 100% rebalancing every year. So to achieve the extra return you'd have to IMO backtest rebalancing strategies, as a 25/5 scheme, or calendar-based rebalancing might give different results. Obviously CCF will do nothing for your portfolio if you merely buy and hold, so rebalancing is key, and attracts costs.

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Tue Feb 05, 2008 8:28 am

Rick's statements on correlation are to me amazing

Correlations are exactly the way I define them

His statements are the exact equivalent of saying the ERP is not 8% because it drifts. The ERP is the long term average. And correlations are the long term average.

And what people should care about is the long term average (the longer the data the more confidence we have that it is meaningful) and also when the risks tend to show up. In the case of equities we should care how the ERP correlates with our earned income and the other portfolio assets. With correlations we should care when they TEND to go up and down and how that relates to the other portfolio assets.

As to his point about point B. the way I said it is exactly correct again. It is the TENDENCY which is what I said. I did not say it always happened. In fact I have explicitly stated examples when it did not. I have even said historically it is about 75% of the time favorable for stocks (not 100%) --so how can I be wrong as he said--but also that it has been 100% for bonds (a point he continues to ignore)

I have also said that given that CCF and stocks can be highly correlated when there are events that lead to deflationary type recessions that is why one should consider adding some duration to the fixed income side of the balance sheet--that hedges that risk and allows one to pick up additional yield (and the risk of that is hedged by the addition of CCF). Another example of why it is so important to think of assets in terms of how their addition impacts the risk of the portfolio.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Tue Feb 05, 2008 10:26 am

Larry and Rick,

To make a long story short, much depends on sample sizes and on what test value one is using for a t-test, so I think you're both half right on correlations. :D

All best,
Pete

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Tue Feb 05, 2008 11:30 am

peter

The correlations are negative --now whether they are statisically significant or not is another issue. But with correlations if you have 38 years of data you have far more than needed for statistical significance. What Rick is getting wrong is the difference between negative correlation and perfectly negative correlation (which I did not state, in fact I pointed out that it was not the case). The correlations are for stocks at about -0.3.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Tue Feb 05, 2008 11:46 am

larryswedroe wrote:peter

The correlations are negative --now whether they are statisically significant or not is another issue. But with correlations if you have 38 years of data you have far more than needed for statistical significance. What Rick is getting wrong is the difference between negative correlation and perfectly negative correlation (which I did not state, in fact I pointed out that it was not the case). The correlations are for stocks at about -0.3.
Hi Larry,

I think it would also depend whether you had good metrics over that 30 year period, but I agree that given a coefficient of -0.3 you could probably get a result that's statistically different from a (default) test value of 0. What I thought Rick was saying, however, is that if you did a separate T-test with -0.3 as the test value you might get a very wide confidence interval (and, as I think you both agree, you might get either better or worse results for commodities over a particular period).

All best,
Pete

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Tue Feb 05, 2008 11:46 am

Rick's statements on correlation are to me amazing
Correlations are exactly the way I define them
:lol: Sorry, I couldn't contain myself.

"Correlations are exactly the way I define them." THAT is amazing.

.

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Tue Feb 05, 2008 11:55 am

Question: What is the correlation between stocks and bonds?
Answer: It depends on the decade. Some decades it is positive and some decades it is negative.

Question: What is the correlation between REITS and stocks?
Answer: It depends on the decade. Some decades it is positive and some decades it is negative.

Question: What is the correlation between CCFs, stocks and bonds?
Answer: It depends on the decade. Some decades it is positive and some decades it is negative.

Question: What absolute benefit should be expected from shifting correlations that have no statistical significance?
Answer: NONE (but it would be nice if it happened).

MPT is a benefit that will likely occur to some degree over the long-term, but no absolute benefit should be relied upon.

Rick Ferri

learning
Posts: 125
Joined: Fri Dec 07, 2007 4:02 pm

Post by learning » Tue Feb 05, 2008 2:20 pm

From what i can make out from this "pre debate", one answer as to whether to include commodities in your portfolio depends on what kind of investor you are, and what kind of investment time frame you have.

Mr. Ferri seems to be suggesting that there exist already the classes of investments -- stock and bonds -- that most individuals will need to meet their investment goals. And that those asset classes have proven themselves over the long term. So why, he seems to be suggesting, stretch for something that has yet to show it can take the knocks and tumbles of real world investing over the long-term and deliver reliable results?

As someone who is investing for retirement, I don't doubt that I will be able to reach my goals with a well constructed portfolio of stocks and bonds. Adding commodities may theoretically improve the efficiency of my portfolio -- or maybe not -- but it certainly is not needed to achieve a real world investment goal.

Perhaps if I were an institution with a time frame of infinity, I might look more closely at commodities and at maximizing the theoretical efficiency of my portfolio.

Or maybe I have that wrong?

Thanks for an incredibly lively and fascinating debate.

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Tue Feb 05, 2008 2:51 pm

Pete
Yes, since the correlations are not perfectly negative that is correct. For different periods you will get different results.
That is true of all assets, there is no asset class with perfectly negative correlation.
What investors should care about is, as I stated, what the long term average is PLUS when the correlations tend to rise and when they tend to fall.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Tue Feb 05, 2008 3:10 pm

larryswedroe wrote:Pete
Yes, since the correlations are not perfectly negative that is correct. For different periods you will get different results.
That is true of all assets, there is no asset class with perfectly negative correlation.
What investors should care about is, as I stated, what the long term average is PLUS when the correlations tend to rise and when they tend to fall.
hi larry,

just to clarify, the size of the confidence interval isn't strictly a matter of the negative correlation being -0.3 rather than -1, as sample size also comes into it, nor is it the case that you couldn't get a very narrow confidence interval given, e.g., 100 -0.3 observations and a single 0.1 outlier. as you say, though, given a bull market for equities/bonds one actually /wants/ the correlation coefficient for that period to turn positive, so one would probably want to break out the figures for bull and bear periods.

all best,
pete

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Tue Feb 05, 2008 4:49 pm

Peter
I agree, but as I noted the sample size for CCF is quite large now, almost 50 years of data and all kinds of markets

Again the key issues are not just the correlation figure itself--which is important, but also WHEN they tend to rise or fall.

As I noted in every single down year for bonds CCF was up-averaging 30%. And of course some years both stocks and bonds are down, making this a very valuable asset for risk averse investors.
Also in the down years for stocks CCF was up 75% of the time, not bad. With average return of 23%.

Now what is also key is that there were no years with all three down. Showing how well CCF works in conjunction with other portfolio assets. And as I noted, importantly, the addition of CCF allows one to take more duration risk (there are no periods when both are negative, and logically so) and that enhances portfolio return--a point completely ignored.

Again it is not just the correlations which we know drift but it is understanding how they are LIKELY to drift, favorably or unfavorably--IMO the clear answer is that at least historically the drift is favorable, and IMO quite logically so do to nature of correlations to inflation and to event risks.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Tue Feb 05, 2008 4:56 pm

hi larry,

all of the above sounds fine (though of course debatable given the metrics) -- the below quote is certainly an attention-grabber and it'd be interesting to hear rick's thoughts:

Also in the down years for stocks CCF was up 75% of the time, not bad. With average return of 23%.

all best,
pete

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Tue Feb 05, 2008 5:09 pm

Peter
That was not a quote but a fact
You can find the tables on pages 274-5 of The Only Guide to a Winning Investment Strategy You'll Ever Need

We all know that correlations drift. But that doesnt mean they should be ignored--they are extremely important to understanding how to best construct a portfolio. But as I said, the long term correlations are not sufficient, you also need to analyze the issue of when they tend to drift and in which direction. This is the same issue we have discussed re junk bonds--which fail the same test because their correlations tend to go in wrong direction at the important moment--their worst returns tend to show up when the best returns are needed most.

BTW- would you agree that 456 data points are sufficient to have statistical significance for correlations? I would assume so. Well you have that in monthly correlations between CCF and both stocks and bonds. You also have it for the 132 data points for quarterly correlations which are even more negative than the monthlies. And that should be more than sufficient I believe. And of course you have it for the 38 years as well.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Tue Feb 05, 2008 5:18 pm

larryswedroe wrote:Peter
That was not a quote but a fact
You can find the tables on pages 274-5 of The Only Guide to a Winning Investment Strategy You'll Ever Need

We all know that correlations drift. But that doesnt mean they should be ignored--they are extremely important to understanding how to best construct a portfolio. But as I said, the long term correlations are not sufficient, you also need to analyze the issue of when they tend to drift and in which direction. This is the same issue we have discussed re junk bonds--which fail the same test because their correlations tend to go in wrong direction at the important moment--their worst returns tend to show up when the best returns are needed most.

BTW- would you agree that 456 data points are sufficient to have statistical significance for correlations? I would assume so. Well you have that in monthly correlations between CCF and both stocks and bonds. You also have it for the 132 data points for quarterly correlations which are even more negative than the monthlies. And that should be more than sufficient I believe. And of course you have it for the 38 years as well.
Hi Larry,

Just clarifying that I was "quoting" that sentence from your post (having been too lazy to delete all around it). . . I got a pm earlier today asking me about whether a particular number of observations is sufficient for statistical significance and I'm afraid I sent a rather rambly reply, the
bottom line is that the stronger the magnitude of the relationship (i.e., the absolute value of the correlation coefficient) the fewer observations you need. For an r of +/- 1 you might only need 10, for an r of +/- -0.1 you might need well over 1000, but it's not linear and so yes, I think 456 observations would be sufficient given an r of +/- 0.3 (i.e., given a two-tailed test at the 95% confidence interval).

All best,
Pete

User avatar
Cb
Posts: 274
Joined: Wed Feb 21, 2007 12:39 am

Post by Cb » Tue Feb 05, 2008 5:33 pm

larryswedroe wrote:...would you agree that 456 data points are sufficient to have statistical significance for correlations? I would assume so. Well you have that in monthly correlations between CCF and both stocks and bonds. You also have it for the 132 data points for quarterly correlations which are even more negative than the monthlies. And that should be more than sufficient I believe. And of course you have it for the 38 years as well.
Larry, what commodity dataset are you referring to here, and how similar is it to PCRIX since it's inception?

Cb

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Post by Rodc » Tue Feb 05, 2008 5:36 pm

456 data points are sufficient to have statistical significance for correlations?
Independent data points or not?

If you had 456 data points corresponding to daily stock market prices all in about 1998-1999, or you had 456 data points corresponding to daily stock market prices all in 2001-2002 you would get rather different and equally bad notions on how the stock market behaves.

The more important fact is the 38 years and if one believes that captures at least a few cycles or whatever it is that drives CCFs.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Tue Feb 05, 2008 5:38 pm

BTW- would you agree that 456 [CCF] data points are sufficient to have statistical significance for correlations?
Larry,

You throw out the acronym CCF as if all funds follow the same strategy, and that the index created on that strategy has actual data going back 50 years. That is absolutely false information. These data points were from hypothetical back-tested trading strategies, not investible portfolios.

You would have a cow if an active stock manager proposed doing the same thing using a hypothetical stock trading strategy. How can you insist that it is different with back-tested commodity trading strategies?

Rick Ferri

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Tue Feb 05, 2008 5:48 pm

Rodc wrote:
456 data points are sufficient to have statistical significance for correlations?
Independent data points or not?

If you had 456 data points corresponding to daily stock market prices all in about 1998-1999, or you had 456 data points corresponding to daily stock market prices all in 2001-2002 you would get rather different and equally bad notions on how the stock market behaves.

The more important fact is the 38 years and if one believes that captures at least a few cycles or whatever it is that drives CCFs.
Hi Rod,

I thought someone might mention that! As the pm'er could attest i was careful to say independent observations in my rambly e-mail, so I agree. Now, I think with monthly return data on an asset we're kind of positing a priori is volatile it's sort of ok to set non-independence aside, but, i /strongly/ agree that the real question is whether or not those 456 observations represent periods bogleheads actually care about. Better to have data that you're 80% confident isn't due to random chance over relevant periods than data that you're 99.9999% confident in over meaningless ones!

all best,
pete

rwwoods
Posts: 1312
Joined: Thu Jun 07, 2007 7:54 pm
Location: The Villages, Florida

Post by rwwoods » Tue Feb 05, 2008 6:15 pm

You are entitled to your opinion. But you are not entitled to your own facts.
- Daniel Patrick Moynihan
How many sets of facts do we have in this debate? Until only one set is mutually agreed upon, there is no way a debate of reasonable merit can proceed.

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Tue Feb 05, 2008 8:47 pm

AS I have pointed out many times one of the benefits of passive strategies is that you can back test them. You cannot do that with active managers.

As I have also pointed out many times Rick uses this argument when it suits him and forgets it when it does not.

Also while it might be a matter of semantics I don't believe there are any trading strategies at all going on with the CCFs we are talking about. They are based totally on passive indices that simply rebalance or reconstitute--that is not IMO a trading strategy. But as I have said it is irrelevant what you call it.

HerbertSitz
Posts: 684
Joined: Fri Feb 01, 2008 8:52 pm

Post by HerbertSitz » Wed Feb 06, 2008 9:29 pm

larryswedroe wrote:Also while it might be a matter of semantics I don't believe there are any trading strategies at all going on with the CCFs we are talking about.
Larry -- Any reaction to this argument?:

Rebalancing has two benefits: (1) maintenance of desired asset allocation and (2) capturing a "bonus" return by exploiting "sporadic buy-low/sell-high opportunities". (quote from Daryanani article below).

The first benefit (maintaining AA) is the only one that is integral to MPT. The second benefit is in fact a matter of historical chance, exists only insofar as markets are not perfectly efficient, and can be captured only by adopting appropriate trading strategies.

The benefit of stable AA is easily gained by simply rebalancing at some definite interval, say quarterly, semi-annually, or annually.

Nevertheless, there has been considerable research on what strategy to adopt for rebalancing. These investigations of what is the best strategy for rebalancing all revolve around maximizing the bonus return from the buy-low/sell-high opportunities. Thus, rebalancing policy is primarily driven not by MPT consideration of maintaining AA, but rather by trading strategies designed to maximize gains by exploiting the market inefficiency that yields the bonus return.

Here are a few papers that discuss advantages and disadvantages of different rebalancing strategies, all of which focus primarily on maximizing the bonus return:

Optimal Rebalancing Frequency for Bond/Stock Portfolios
by David M. Smith, Ph.D., CFA, and William H. Desormeau, Jr., CFP
(can be found at fpanet dot org)

Opportunistic Rebalancing: A New Paradigm for Wealth Managers
by Gobind Daryanani CFP®, Ph.D.
(at fpanet dot org)

THE REBALANCING BONUS: Theory and Practice
William J. Bernstein
(at efficientfrontier dot com)

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Thu Feb 07, 2008 9:11 am

Herbert
Yes I don't agree with the statements at all, except for the first part.

Second, IMO rebalancing should NOT be a time dependent issue. It should be a % issue that should be checked regularly. We in fact have system that allows us to check it daily. Why would it be time dependent if the purpose is to maintain portfolio allocations. Swensen for example rebalances daily.

Third, as I stated before the majority of time when you are rebalancing you are REDUCING the expected return of the portfolio. This will be true any time you sell a high expected returning asset class to buy a lower expected returning asset class. Say stocks and bonds. After period of stock outperformance and you sell stocks to rebalance you are lowering the expected return of the portfolio. Same true if selling value to buy growth or small to buy large, or EM to buy developed country.

The only time you would get some expected benefit in terms of higher expected returns is if you are rebalancing between asset classes with the same expected return. Then when you sell the outperformer (whose expected return relative to the other is now lower because of the outperformance) you are raising the expected return.

The diversification return from rebalancing is not higher expected returns from a portfolio but the impact of the rebalancing-you get higher portfolio return than weighted average of the individual components.

User avatar
Topic Author
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Post by Rick Ferri » Thu Feb 07, 2008 9:44 am

Larry wrote:
IMO rebalancing should NOT be a time dependent issue. It should be a % issue that should be checked regularly. We in fact have system that allows us to check it daily. Why would it be time dependent if the purpose is to maintain portfolio allocations. Swensen for example rebalances daily.
I agree that % work better than rebalancing on the calendar. That is the way our firm does it also. The issue with individual investor's doing % is the introduction of behavioral risk. Portfolio management companies are contractually obligated to rebalance because that is what we are hired to do. Individuals have no obligation to themselves. If an individual is constantly looking at the allocations, and watching their account value drop on a daily basis, there is a large risk that they will not rebalance when it is time to, or, sell the asset that is falling rather than buy more. That is why I advocate that individuals rebalancing annually rather to hawk thier portfolio every day or more.
I stated before the majority of time when you are rebalancing you are REDUCING the expected return of the portfolio. This will be true any time you sell a high expected returning asset class to buy a lower expected returning asset class.
This is true also. Although MPT shows that rebalancing could produce a return benefit over the very long-term, rebalancing at its heart is a risk control mechanism, not a return enhancement mechanism.
The only time you would get some expected benefit in terms of higher expected returns is if you are rebalancing between asset classes with the same expected return. Then when you sell the outperformer (whose expected return relative to the other is now lower because of the outperformance) you are raising the expected return.
I will only add that since commodities have no real return and stocks have a real expected return of about 6%, then any long-term return benefit from including commodities is expected to be negligible. Adding commodities to a portfolio lowers risk, but also lowers expected portfolio return. Caveat: the trading strategies in CCFs may produce Alpha, but the high cost of those products will likely chew up whatever Alpha the trading straetgy generated.

Thank you for making my case for not including commodities, Larry.

Rick Ferri
Last edited by Rick Ferri on Thu Feb 07, 2008 9:49 am, edited 2 times in total.

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Thu Feb 07, 2008 9:46 am

Herbert
As demonstration of the point
1990-99
S&P 60% and 5 year T bonds 40%.
First case rebalance annually--ending portfolio growth 268.
Second case, never rebalance--ending portfolio growth 289.

Clearly rebalancing reduced returns, not added to them.

However the return of the portfolio was higher than the weighted average of the two returns
The portfolio returned 13.93 vs weighted average of 13.78--the difference is the diversification return.

User avatar
ken250
Posts: 1894
Joined: Mon Feb 26, 2007 8:50 pm
Location: US-101

Post by ken250 » Thu Feb 07, 2008 11:46 am

Sorry to get academic...

Doesn't the decision to add or not add commodities come down to REQUIRED return?

To conclude commodities are not beneficial based on return alone dismisses the benefits of the low correlation. Doesn't CAPM permit us to determine the required return?

(The required return should be low because beta will likely be negative or low.)

larryswedroe
Posts: 15727
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Post by larryswedroe » Thu Feb 07, 2008 12:38 pm

ken
I showed period selectively chosen to show how viewing CCF the way Rick does leads to incorrect answers. The period was 91-07.
Now Rick himself forecasts ERP of about 5%. During this period the GSCI underperformed the S&P by about what Rick forecasts the ERP to be--

Thus this might be considered a good example of what would happen if the ERP was 5% and CCF provided no return above the risk free rate (the diversification benefit which Rick himself has estimated at 2% would basically be 0, not the 3-5% found by E&H).

during this very period CCF added value --a 5% allocation would have left one with the same return but significantly lower SD. And that is in very bad period for CCF, sort of reverse data mining.

Now go back further and you find CCF adding to returns and lower SD a lot.

BTW-Rick has made the incorrect statement that you cannot eat SD, only returns. Well that is wrong because in withdrawal phase SD matters a lot--it lowers odds of success of portfolios not running out of money. You can easily have portfolios with lower returns but also lower SDs surviving while the higher returning portfolio fails.

Post Reply