Asset Allocation Didn't Fail: The Allocators Did

There has been a lot said and written lately about the failure of asset allocation, supposedly proven by the large losses experienced in 2008. We don’t believe asset allocation failed. We believe asset allocation practitioners failed in their execution.

Too much slicing and dicing had crept into the working definitions of asset classes — pushing granularity beyond reasonable limits. Categories were being confused with classes. Just because two categories sound different, doesn’t mean they are different in the asset class sense.

Assets are primarily aggregated into classes, and classes distinguished one from the other, based on the correlation of their returns. Diversification, while important to minimize issue selection risk, is not allocation. Allocation is putting assets into asset category groupings that do not respond to economic and other conditions in the same way, to the same extent, at the same time as each other — that mean’s low correlation. Merely different category descriptions does not make for different asset classes.

Putting some money into US stocks and other money into EAFE stocks, for example, is a form of diversification, but not effective asset allocation. The correlation of EAFE to the S&P 500 over 1, 3 and 5 years has been 0.923, 0.901 and 0.902.

If you accept correlation as a key to asset class differentiation, then US stocks and EAFE stocks have not been separate asset classes for some time long before 2008, and it would be inappropriate to blame the asset allocation concept for use of those two categories as separate classes when they have not been for some time.

They may be separate classes some time in the future, but they are not now.

The correlation between emerging market stocks and US or EAFE stocks is lower but still above 0.800 — not particularly attractive for allocation purposes.

Similarly, if you put money into the short-term, intermediate-term and long-term subcategories of the US aggregate bond index, the duration difference has not been effective for allocation purposes, with correlations in the 0.850 to 0.950 range.

We believe there was adequate information prior to 2008 to make appropriate risk controlling asset allocation decisions. Blaming bad portfolio returns on the failure of a basic investment concept sounds like scapegoating to us.

The chart shows the 3-year correlation data that was available, for example, for iShares ETFs with at least 3 years of history as of year-end 2007 (before 2008).

click image to enlarge

However, those expressing concerns about possible flaws in asset allocation as a concept are correct in stating that there was surprising convergence of equity correlations during 2008.

On the other hand, you can’t toss out a half century proven concept just because a couple of quarters took everybody by surprise. Regardless of some convergence, there was not universal convergence, and a well distributed allocation did not experience crippling portfolio damage.

This chart from the iShares site shows the rolling correlations for several ETFs. It demonstrates that while equity categories tended to converge, fixed income remained below 0.500 correlation and produced positive returns that would have moderated losses for portfolios with significant fixed income allocations.

click image to enlarge

We know some will say who wants bonds — bonds have low returns. Well, yes, but they also have low volatility, a higher percentage of positive years, and lower maximum draw-downs.

The issue in life and investing, is that you can have anything you want — you just can’t have everything you want. If you want maximum returns from equities with massive equity allocations, then some very painful years will come around multiple times in an investing career. If you can’t take the rough water, bonds are the calm water in which to paddle. If you are somewhere in between, you need significant allocations to each of equities and bonds. And, we would add, that cash is not trash — from time-to-time, as the situation deteriorates, take a break, stand aside, watch the action, but don’t play — wait for positive price action to be restored before taking on risk.

So what might have worked? We think the answer lies in a less granular approach — staying with big categories.

Consider this approach to asset allocation. You begin the process with your base currency (USD for US investors, Euro for German investors, Yen for Japanese investors, etc). Cash in the form of your base currency is one asset class. Other than spending or donating your money, there are only a few generic things you can do with it. You can lend it. You can use it to be an owner, or you can exchange it for some other currency which you hold in reserve instead of your base currency.

Key representatives of US “lending” are bonds (VBMFX proxy used here). Key representatives of “owning” are stocks (VFINX proxy used here). A dominant representative of cash is a money market fund (VMMXX proxy used here). One general representation of US investors holding a currency other than their base currency, would be the inverse of a Dollar index (UDN proxy used here).

Those very basic categories exhibited meaningfully different correlation of return over the past year.

click image to enlarge

We don’t have correlation data for a basket of currencies against the USD for longer periods, but by definition a basket of foreign currencies will have not have a high correlation with your base currency. That is because currencies are denominated in terms of each other. Thy cannot all go down, and they cannot all go up. Either nothing happens, or some must rise and some must fall. It’s axiomatic.

The next table shows that money market cash, total US taxable bonds and the S&P 500 exhibited meaningfully different correlations over 1 year, 3 years, 5 years and 10 years.

That suggests to us that it was not the asset allocation concept that failed in 2008, but rather ineffective or inappropriate allocation decisions that failed, or too short-term thinking in deciding allocation doesn’t work.

Instead of correlations tables, let’s look at the charts.

These charts plot the percentage change from the beginning of the chart period to enable a clearer comparison between securities. The currency basket versus the USD (UDN) is plotted in red. Aggregate US taxable bonds (VBMFX) is plotted in blue. The S&P 500 (VFINX) is plotted in black. StockCharts.com does not provide plots for money markets, so we represent cash with a horizontal green line (an approximation of reality).

The actual returns for the money fund (VMXX) through June 2009 was 1.48% for 1 year; 3.65% for 3 years; and 3.39% for 5 years.

Clearly, a simple asset allocation based on those four classes, would not have failed in 2008 or YTD in 2009.

click images to enlarge

Calendar 2008

YTD 2009

The Dow Jones US asset allocation index uses US Dollar cash, US bonds and US stocks to create five different allocations from conservative to aggressive portfolios. The charts below present the results of those five allocation mixes for calendar 2008 and YTD 2009 (most aggressive labeled $P100US and the most conservative labeled $P20US — see the DowJones site for details of the mix).

These also show that asset allocation worked in 2008, and it is working now (”working” does not mean no losses). Contrary to some articles we’ve read, there was not a simultaneous collapse of all asset classes (yes, for a few days all Hell broke loose in October 2008, but things quickly sorted themselves out for major asset categories). We don’t include gimmicky structured products in our consideration. We didn’t buy them, don’t buy them and won’t buy them.

click images to enlarge

Calendar 2008

YTD 2009

The key,we believe, is keeping it simple, sticking to clearly different classes based on correlation, and not placing too much in one class or too little on the other. Too little in one class has too little effect. Too much in one class has the potential for ruin — that is what happened to many investors in 2008. Someone with 80+% of their money in stocks, even if divided between US and various foreign stocks, or US stocks divided by market-cap, style or sector, generally got hammered. If they didn’t realize their concentration in highly correlated assets, their steep losses probably came as a rude surprise.

That fact that many investors confused category diversification with asset allocation and did not know they were not practicing asset allocation, and instead were placing over-sized class bets, has resulted in a flurry of claims that asset allocation doesn’t work anymore. Really, it works fine if you play by the rules.

• seek low correlation
• don’t make huge bets
• don’t make tiny bets
• rebalance
• monitor changing correlation and weights accordingly.

At its core asset allocation is about risk control — assembling a spectrum of assets with an expected long-term (not 12-month) portfolio mean return and volatility. Overall, you need to know the mean return, volatility and correlation of each asset (along with individual securities diversification) to have a reasonable shot at controlling risk.

Optionally, some might want to add a bit of strategic timing by using catastrophic stop loss orders to get out of a collapsing class before the worst sets in (that’s what we did in July 2008), and then holding the cash generated until the class stops it precipitous decline and resumes an upward course (that’s what we are doing in stages now).

For those US investors who don’t want to include currency in their process and who prefer to stay close to home, here is series of plots for US cash, US bonds and US stocks for 10 years monthly, 3 years weekly, and daily for calendar 2008 and for YTD 2009.

click images to enlarge

10 Years Monthly

3 Years Weekly

2008 Daily

2009 YTD Daily

Asset allocation is not dead. It just needs to be practiced more carefully and rationally, and given time to work.

Related ETF securities: SPY, IVV, IWV, VTI, EFA, VEA, EEM, VWO, UDN

Disclosure: from time-to-time, we own some of the securities named in this article.

Comments

[Lightway:]

Great foundational article that clears up alot of misconceptions and the confusion between asset allocation and diversification.

Alot of the traditional advise on diversification would have simply had investors allocate between equities: large cap, mid cap, small cap, and foreign. Bonds would have been treated as a nice to have and cash not even mentioned.

Thanks for the back to basics explanation with the supporting data. I think this is another article everyone should bookmark, regardless of their trading style.

Jul 14 10:53 AM

[Analyste de...:]

Asset allocation per se doesn't dictate "fully-invested, all the time." That's an erroneous presumption.

But yes there is a story here: traditional asset allocation (the industry groupthink, as practiced) certainly failed investors. Just look at the consistency of returns in the Lifecycle products: the average long term lifecycle fund (75/25) lost -47% between 10/9/07-3/9/09. Likewise, VERY FEW diversified mutual fund managers escaped > -40% declines, so yes the failure 'asset allocation as practiced' was systemic and across the board.

I don't know why advisors & investors began to assume the black-box algorithms would work forever - especially considering how poorly they'd done in the 2000-02 Bear. Did we already forget that, too? It's as though collective amnesia has set in... and now we're playing a Blame Game. Mutual fund investors nearing retirement were left holding the bag ("Buy & Hold... Invest for the long Term!"), so the John Bogles are now blaming .... ETFs! Hilarious, pay no attention to the Man Behind the Curtain...

If you want to find the REAL CULPRIT, look instead at the Monte Carlo simulators that predicted commodities would do poorly vs. stocks while presented the rosiest of scenarios for equities, etc. These software products were (mostly) garbage in/garbage out; since almost everyone was using them, everyone got burned. Shouldn't bunky software foibles - duping the entire industry & its clients - be what's discredited and abandoned, now?

Good old fashioned common-sense would have protected more investors : there are times to strategically allocate more to cash & fixed income, and/or hedge with shorts. Those periods are called Bear Markets!

Will we ever learn & remember?

Jul 14 01:01 PM

[Ernie Gulden:]

Your work is excellent, Richard.

Looking at your 10-year chart of VFINX, cash, and VBMFX shows the latter gaining 74%. BigCharts plot of VBMFX shows it going from $9.80 to $10.25. Even if dividends are included in the return I don't see 74%. Do I have a bad chart?

Jul 14 02:05 PM

[Grecian Gringo:]

great article.

Jul 14 04:00 PM

[Richard Shaw:]

Ernie:

Great point. I do believe that StockCharts plots at least some mutual fund data differently than they plot individual stocks or indexes. This is apparently one of the instances.

Your chart is OK by me, but it is different than the StockCharts chart. Somehow, although they will not say so, StockCharts seems to be including some aspect of total return that causes that difference.

The conclusions about cash vs stocks vs bonds, however, would not be different with other chart services in this instance, although the pictures would be less severe.

I have been in contact with StockCharts.com about the differences between their charts and those from BigCharts, Yahoo Finance, and MetaStock.

Their answers have been opaque to me. I cannot explain why or how they are different. I have suggested to them that if everybody else agrees but them, they must be wrong.

Their response is not helpful. The only part I can understand clearly is their assertion that for technical analysis their way is best.

Fortunately, the discrepancies are are not universal. Mutual funds tend to be more of a problem.

If I could find a service with the same charting features as StockCharts.com (MetaStock has more) with the image sizing and printing features of StockCharts (MetaStock is not good for that), I would not use StockCharts. But for blogging purposes StockCharts is most often the best choice, but not in this instance.

Thank you for your comment. Don't worry about your BigCharts chart. I'll try not to use mutual funds in my future examples, unless I use BigCharts, MetaStock or some service other than StockCharts.

Jul 14 04:24 PM

[User 229008:]

As a Fin101 student, your work is most appreciated.

Jul 14 07:00 PM