designing a smoother ride

8/3/2019 Designing a Smoother Ride

1/36

Designing a Smoother RideBalancing Risk and Return Using Dynamic Asset Allocation

n Identifying changes in the risk/

reward trade-off as they occur

n Judging when adjustments to

asset allocation are warranted

n Smoothing portfolio volatility and

limiting the severity of losses

January 2010


2/36


3/36

Table of Contents

1Key Research Conclusions

2Introduction

Why Do Long-Term Investors Care About Short-Term Risks?

7How Much Risk Do You Really Have?

Measuring Volatility and Diversification Potential

14How Much Return Is Enough?

Measuring the Opportunity

21Achieving More Consistent Outcomes

The Portfolio Impact of Dynamic Asset Allocation

25Notes

26Glossary


4/36


5/36

Dynamic Asset Allocation 1

Key Research Conclusions

A well-designed long-term asset allocation is crucial to the

success of any investment program. But even a thoroughly

diversified portfolio is vulnerable to large losses, particularly

when a financial-market shock occurs. We have developed

dynamic tools that can be used to adjust an asset-allocation

strategy systematically as market conditions change.

Our dynamic asset-allocation research seeks to measure

short-term risks and returns more accurately in order to rein in

volatility and cut down on extreme outcomes, without giving up

return potential. We believe such an approach can deliver a

more consistent investment experience, regardless of the

capital-markets environment. Some of the key conclusions from

our research are:

n By focusing on controlling risk, and being skeptical about

making changes to portfolio weights based solely on expected

returns, dynamic asset allocation can smooth out volatility and

mitigate extreme outcomes, without sacrificing performance

in the long run.

nMarket risks can be more reliably forecast than returns, largely

because volatility trends tend to persist for extended periods

across all major asset classes.

nContrary to popular belief, periods of high volatility are often

not followed by large gains. Its crucial for investors to

measure how well they are being compensated for accepting

more risk.

n Return forecasting can help to indicate when markets are

most vulnerable and when they are likely to be most reward-

ing. Factors such as valuations, levels of corporate profitability,

the level and direction of interest rates, and credit spreads can

provide early warning signals.


6/36

2 AllianceBernstein.com

IntroductionWhy Do Long-Term Investors Care About Short-Term Risks?

The past 10 years have offered a stark reminder of just how

volatile the capital markets can be. Over this period, global

equities have twice suffered peak-to-trough falls of more than

45%, followed by sharp recoveries.1 Global investment-grade

corporate bonds underperformed government bonds by almost

17% in 2008, only to beat them by 15% in the first nine

months of 2009.2 Central banks have adjusted monetary policy

rates dramatically on several occasions, and commodity price

movements have been unprecedented, with oil fluctuating

between US$20 and US$150 a barrel in the past decade.

This type of volatility can be extremely unsettling to investors

and may even cause lasting damage to the growth of their

portfolios. The traditional way of mitigating these types of

violent capital-markets swings is portfolio diversification, in the

form of a well-balanced long-term asset allocation. Spreading

assets across a wide array of weakly correlated investments can

reduce the short-term volatility of returns without giving up

much performance in the long run. Many investors have

adopted such an approach, adding a wide array of asset types

and strategies to their mixnot only global stocks and fixed

income but also real estate, commodities and alternative

investments such as hedge funds.

A well-diversified long-term asset-allocation strategy is one of

the most important decisions an investor is ever likely to make.

But even a thoroughly diversified long-term strategy is vulnera-

ble to unusually large losses. During extreme and unexpected

financial-market shocks (sometimes referred to as tail events),

equity volatility soars and correlations between assets can

increase rapidly, making diversification less effective just when

investors need it most. The top chart in Display 1 shows how,

over the past decade, a portfolio invested in a balanced mix of

60% equities and 40% fixed income would have suffered large

fluctuations while generating hardly any real growth.

And diversifying by adding other asset classes would not have

made much difference to the outcome. Few asset classes

provided a safe haven during the technology, media and

telecommunications (TMT) collapse, and only government

bonds offered any protection during the credit crisis (Display 1,

bottom).

A Balanced Allocation Can Behave in Different Ways

The discomfort that investors suffer during market downturns

illustrates a broader problem: the tendency for the risk profile of

any fixed asset mix to stray materially from investors expecta-

tions. Over the past four decades, the 60/40 portfolioan asset

allocation designed to suit an investor with a moderate

tolerance for riskhas at times displayed the volatility of an

all-bond portfolio and at other times been as volatile as an

all-equity portfolio (Display 2).

Any major shift in volatility alters the range of returns that an

investor is likely to experience. For example, a portfolio with an

expected return of 7% and an expected volatility of 9% (which

is how the 60/40 mix behaves over the long run) should

generate returns somewhere between a gain of 25% and a loss

of 11% in a given year, only exceeding an 11% loss about once

in 40 years.

If volatility shot up to 15%, this would substantially increase

both the upside and downside potential of the portfolio, to a

gain of 37% on one hand and a loss of 23% on the other. 3 Its

unlikely that an investor with a moderate profile would be

comfortable with that degree of uncertainty.


7/36


Display 2

The Volatility of a Balanced Account Has Fluctuated Widely

Portfolio Volatility: 60% Global Stocks / 40% Global Government Bonds**

Percent

0

5

10

15

20

70 73 76 79 82 85 88 91 94 97 00 03 06 09

Long-Term Average

Equity-Like Volatility

Bond-Like Volatility

Through September 30, 2009Past performance is not indicative of future results.*Refers to 60% in the MSCI All Country World Index and 40% in the Barclays Global Aggregate Index, rebalanced monthly. Growth of US$1 is calculated on an inflation-adjusted basis. Returns to high-yield and investment-grade credit refer to excess returns over comparable-dated government bonds. See notes on page 25 for asset class definitions.**Throughout this paper, unless otherwise noted, global bonds refer to government bonds and global stocks refer to developed-country equities, with returns hedged into US dollars.Source: Barclays Capital, Bloomberg, FTSE NAREIT, Global Financial Data, MJK Associates, MSCI, Thomson Reuters and AllianceBernstein

Display 1

A Balanced Allocation Has Had a Bumpy Ride over the Past Decade

Growth of US$1 of a Global 60% Stock / 40% Bond Asset Allocation*

TMT Collapse: Cumulative Returns Mar 00Sep 02

GlobalEquities

Emerging-Market

Equities

REITs High-Yield

Credit

Investment-Grade

Credit

Govt.Bonds

CommodityFutures

ForeignCurrency

(47)%

12%

(27)%

(3)%

22%

2%

(7)%

(44)%

Credit Crisis: Cumulative Returns Oct 07Feb 09

US$1.01

US

Dollars

(54)%(67)%

(35)%

(17)%

9%

(19)%(11)%

(61)%

GlobalEquities

Emerging-Market

Equities

REITs High-Yield

Credit

Investment-Grade

Credit

Govt.Bonds

CommodityFutures

ForeignCurrency

0.50

0.75

1.00

1.25

00 01 02 03 04 05 06 07 08 09

(29)%

(36)%


8/36


It seems counterintuitive that a balanced portfolio should

behave in such different ways. The fact is that even though a

60/40 mix is balanced in terms of asset allocation, it is concen-

trated in terms of risk. One asset classequitiesdrives the

lions share of portfolio volatility. Since stocks are three to four

times more volatile than bonds, they generate an average of

about 90% of the performance variability of the typical 60/40

portfolio. So, when equity-market volatility ebbs and flows, it

tends to take the whole portfolio along with it, (See Equities

Drive Portfolio Volatility, page 5).

Mitigating the threat of a disconnect between investors

expectations and actual portfolio outcomes is most important

during periods of high or rising volatility, which usually coincide

with bear markets. At these times, investors are likely to be

feeling severe pressure in other areas that affect their invest-

ment plans. For example:

n Individuals are more likely to lose their jobs and income;

n Foundations and endowments are likely to face declines in

charitable contributions;

n Pension funds are likely to see their funding capacity decline

as public plans are hit by falling tax revenues and private plans

face underfunding due to declining corporate profits; and

n Assets become illiquid and access to credit dries up.

In short, equity-market misery is often compounded by other

factors that can make a bear market even more painful. These

realities argue for a more flexible approach to asset alloca-

tionone that can enhance a long-term strategy by providing a

smoother pattern of returns.

Dynamic Asset Allocation: Responding to PrevailingMarket ConditionsThe goal of our research was to find a systematic and durable

way to monitor changes in the market environment in order to

find a better balance between changes in market risk structure

and changes in return potential (Display 3).

This is not a new idea: as long as capital markets have existed,

investors have been seeking systems for buying low and selling

high. Unfortunately, the results of such systems have been

inconsistent at best. In our view, this is because most strategies

focus almost exclusively on returns, even though it is extremely

challenging to predict short-term turns in the markets with a

high degree of accuracy.

In the course of our research, we started to question whether

the focus on predicting market returns was too one-sided. After

all, risks can change significantly as well.

We found that risk could be forecast with considerably moreconfidence, and that improvements in forecasting could have a

significant impact on the efficacy of a dynamic strategy. This is

where our approach really diverges from traditional tactical asset

allocationit seeks to improve the risk/reward trade-off

primarily by mitigating risk, rather than by reaching for higher

returns.

Our tools measure the expected risks of a portfolio (by estimat-

ing asset volatilities and correlations) and the expected returns

available so that, when the risk environment changes, we can

determine whether investors are being paid enough to maintain

or increase their exposure.

Display 3

Weighing Risk and Return

Fundamental Oversight

Market Risk

VolatilitiesCorrelations

SentimentValuations

Interest RatesCredit Spreads

Return Potential


9/36


Equities Drive the Volatility of a Balanced Account

Contribution to Portfolio Volatility60% Equities / 40% Government Bonds

Percent

(5)

0

5

10

15

20

69 74 79 84 89 94 99 04 09

Equities Bonds

0.7%

8.3%

Long-Term Average

Equities

(92% of risk)

Bonds

(8% of risk)

9.0%

Through September 30, 2009Source: Barclays Capital, MSCI and AllianceBernstein

Understanding where a portfolios risk is coming from is

crucial in order for investors to manage portfolio fluctuations

effectively. Equities are three to four times more volatile than

government bonds, which means that they tend to play a

bigger part in overall portfolio volatility than their nominal

value suggests.

The display below shows how much of a 60/40 portfoliosperformance variability has been driven by equities and how

much has been driven by bonds over the past 40 years. Over

the period that we studied, equities contributed an average

of 92% of the overall risk8.3% out of the 9.0% total

and bond volatility accounted for the remaining 8%.

Contributions to risk can fluctuate depending on each asset

classs volatility at the time, and the extent to which the

performance patterns of the asset classes are correlated. For

example, in the 1970s, bonds contributed more to portfolio

volatility than usual because concerns about inflation

undermined the fixed payments of bonds and made their

prices very volatile. And, because stocks were also adversely

affected by inflation fears, the correlation between the two

asset classes rose above its normal level. In that inflationary

context, bonds not only became riskier; they also provided

less portfolio diversification.

During the 2000s, the opposite scenario has unfolded.Government bond volatility has been extremely low, and

stocks and bonds have displayed strong negative correla-

tions as concerns about the sustainability of real economic

growth and the risk of deflation have loomed larger than

worries about inflation. As a result of this negative correla-

tion, in recent years bonds have acted as a powerful

diversifier against equity risk. But in cases of extreme equity

volatility, such as the escalation of the credit crisis in late

2008, a portfolio would have needed a much larger

weighting in bonds and much less in equities in order to rein

in overall portfolio volatility. n

Equities Drive Portfolio Volatility


10/36


When applied in a systematic way over time, we believe that

dynamic asset allocation will produce measurable benefits,

namely:

n Less portfolio volatility;

n Fewer extreme negative outcomes, reducing the probability of

large losses; and

n Comparable long-term returns.

A reduction in tail events both mitigates outsize losses and

reduces outsize gains (Display 4). This tends to result in outper-

formance in bear markets and underperformance in recoveries.

In the sections that follow, we discuss the building blocks of our

approach: how we analyze market risks, how we assess

potential returns and our mechanism for integrating these

forecasts into asset-allocation recommendations.

Display 4

Dynamic Allocation Seeks to Improve Distribution of Returns

Dynamic Asset Allocation

Conventional Asset Mix

Returns

FewerLargeLosses

FewerLargeGains

Less Volatility

Lower volatility

Fewer tail events

Comparable returns

n Even a well-diversified long-term asset allocation can suffer high volatility and heavy losses.

n In the short term, shifting market risks can cause portfolio outcomes to disconnect alarmingly from what long-term averages

might suggest.

n Our dynamic asset-allocation approach responds to short-term market changes, with the goal of providing a smoother

pattern of returns.

Chapter Highlights


11/36


How Much Risk Do You Really Have?Measuring Volatility and Diversification Potential

The goal of our dynamic risk tools is to identify shifts in

volatilities and correlations in their early stages so that we can

adjust portfolios in time to mitigate damage or respond to

changing diversification opportunities.

For example, as the technology, media and telecommunications

(TMT) bubble was collapsing in April 2000, simulation results

suggest that our forecasts would have pointed to high equity

volatility, while our correlation model would have indicated a

good opportunity to diversify risk, with both real estate

investment trusts (REITs) and bonds showing below- average

correlations with equities (Display 5). All else being equal, this

would have called for less exposure to equities and increased

exposure to bonds and REITs.

In September 2008, as the credit crisis escalated, our volatility

forecasts likely would have signaled high risk in the equity and

real estate markets. The forecast correlation between REITs and

equities, at +0.76, would have signaled a below-average

diversification opportunity from real estate.

By contrast, the simulation results suggest that our model

would have highlighted the exceptional diversification benefits

available from government bonds at the timeas shown by the

strong negative correlation of (0.34) between bonds and

equities. All else being equal, this would have called for lower

equity and real estate exposure and higher exposure to bonds.

Volatility Is Easier to Forecast than ReturnsGiven that volatility fluctuates so much, how can we forecast it

with any confidence?

Surprisingly, our research shows that volatility can be predicted

with reasonable accuracy. We found that over the past 40 years,

the short-term volatility factors in our model could have

explained 30%50% of the variability in global equity, bond,

currency, commodity futures and real estate volatility. This is a

very good fit: in the world of return forecasting, a model that

explains more than 10% of future returns is considered quite

powerful.

The main reason why we believe that volatility can be more

accurately forecast than returns is that recent levels of volatility

tend to persist for extended periods before slowly trending back

toward their long-term averages.

The finding that volatility trends persist makes intuitive sense.

For example, if investors expect an economic shock to cause a

Display 5

Risk Forecasting Provides Early Warning Signals

TMT CollapseApr 2000

+0.55

Correlation withEquities

+0.76+0.37Global REITs

+0.15 (0.34)+0.12Global Bonds

Volatility

14% 26%19%Global Equities15% 28%15%Global REITs

5% 5%4%Global Bonds

Long-Term Avg.Credit CrisisSep 2008

12Month Forecast

Through September 30, 2009. Source: Barclays Capital, FTSE NAREIT, GlobalFinancial Data, MSCI and AllianceBernstein


12/36


recession, they are likely to have doubts about corporate cash

flows and the health of companies balance sheets. Asset types

that are most affected by the turmoil, such as equities, corpo-

rate bonds and industrial commodities, are likely to become

more volatile as a result. Monetary, fiscal or regulatory authori-

ties may step in with measures to address the fallout, but it is

often not immediately clear to the market how well these

measures will work. Volatility is unlikely to ease back toward its

long-term average until investors gain confidence about the

economic outlook and its impact on the value of their invest-

mentsand this process takes time.

The S&P 500, the equity index for which we have the longest

series of historical data, shows how sticky volatility has been

since 1929. We ranked past levels of annualized stock-market

volatility4 from highest to lowest and then calculated the future

levels of volatility over the subsequent month, quarter and year

(Display 6). Following the 20% of periods when volatility was at

its highestabout 35% on averageit was still well above its

long-run average a year later. Likewise, in the quintile of periods

when volatility was at its lowest, volatility was slow to climb

back toward its long-term average.

We found that the same pattern occurred, with remarkable

consistency, across a range of global asset classes and countries.

In other words, recent history is typically a good indicator of the

level of volatility we are likely to experience in the months

ahead. We studied numerous methods for incorporating recent

market data in developing our shorter-term risk forecasts. Forexample, there are short-term volatility forecasts that are traded

in the markets, such as the Chicago Board Options Exchange

Volatility Index (VIX). This represents expectations of US equity

market volatility over the next 30 days, implied by the pricing of

options on the S&P 500 futures contract. However, market-

based forecasts are not available for most asset classes, do not

have long histories and can become distorted during crises just

when they are most needed. We believe that a model based on

Display 6

Volatility Is Sticky Across Asset Classes

Annualized Volatility: Top and Bottom Quintiles

S&P 500

19292009Global Stocks

19702009

Months Forward

0

10

20

30

40

0 3 6 9 120

10

20

30

0 3 6 9 12

Currencies

19742009Commodities

19702009Global Fixed Income

19702009

0

2

4

6

8

10

0 3 6 9 125

7

9

11

13

15

0 3 6 9 120

5

10

15

20

25

0 3 6 9 12

Percent

Past (Realized) Volatility Forward Volatility Average

Through June 30, 2009Each observation is ranked by past volatility at the end of the month and sorted by quintile (in-sample) except currencies, which are sorted by tercile. Past volatility is anexponentially weighted average using daily data with a three-week half-life (5% decay per day). All metrics are annualized.Source: Barclays Capital, Bloomberg, FTSE NAREIT, Global F inancial Data, MJK Associates, MSCI, S&P, Thomson Reuters and AllianceBernstein


13/36


daily data and driven by recent volatilities is the most robust and

flexible. It can be applied over a wide range of asset classes and

countries and can be used to forecast volatilities over different

time horizons.

Its worth noting that the aim of our volatility forecasting is not

to predict shocks before they happena difficult task at the

best of timesbut to give us fair warning when the market risk

structure starts to change.

One analogy is hurricanes: Even the best meteorologists havetrouble predicting the exact number of hurricanes that will

happen in a given year, but once a storm begins, it is possible to

measure how it is building and changing. It is rare that a

hurricane escalates from a category one to a category five

overnight, so there is often time to take cover.

Simulation results for the period from 1970 to 2009 suggested

that our volatility forecasts would have been quite sensitive to

changing risk environments, displaying the ability to capture

extremes. For example, our equity volatility forecasts ranged

from 9.8% to 33.7% and our bond forecasts ranged from

2.2% to 12.1% (Display 7, left).

Most importantly, we found that our volatility forecasts were

quite accurate. The bar charts to the right in Display 7 isolate

the 20% of cases when volatility was forecast to be highest and

the 20% of cases when volatility was forecast to be lowest,

comparing our 12-month forecasts with the actual levels of

volatility that occurred. There was a good fit between theforecast and realized levels. Simulations showed that, in the

lowest-volatility periods, realized global equity volatility aver-

aged about 12%, compared with an average forecast of 11%

by our model. In the highest-volatility periods, realized volatility

was about 18%, compared with forecast volatility of 20%.

For more on our volatility forecasting techniques, see Stormy

Weather: How Our Volatility Model Works, page 10.

Display 7

Our Forecasts Are Built to Capture Changing Risks

Global Volatility: One-Year Forecasts19702009

70 73 76 79 82 85 88 91 94 97 00 03 06 09

0

10

20

30

40 Global Stocks

Global Bonds

Lowest-VolatilityQuintile

18%20%

12%11%

Highest-VolatilityQuintile

6%6%

3%3%

Lowest-VolatilityQuintile

Highest-VolatilityQuintile

RealizedForecast

RealizedForecast

Nifty50 1982

Recession

BlackMonday

Savings& LoanCrisis

CreditCrisis

TMTBubbleAsian

Crisis

Global

Stocks

Global

Bonds

Percent

As of September 30, 2009Source: Barclays Capital, Global Financial Data, MSCI and AllianceBernstein


14/36


We use an adaptive risk modeling framework to forecast

volatility and the correlations between asset classes. First we

estimate the volatilities and correlations of the major market

risk factors, such as global equity price movements and

interest rates. We can then estimate any assets sensitivity to

those risk factors, and any residualthe amount of volatility

and correlation that is unexplained by those market risk

factors.

When modeling each component, we consider a combina-

tion of its recent realized volatility and correlations, as well

as its long-term averages, so that we can best capture the

changing nature of the risks as capital-markets conditions

evolve.

Short-Term Risk: This measure helps us gauge very recent

changes in market sentiment so that our forecasts can pick

up sudden shifts in market risk. Volatility observations in the

past three weeks count for half the weight in our short-term

measure.

Medium-Term Risk: Our medium-term factor has a slightly

longer look back period, with the goal of understanding

whether we are operating in a generally high- or low-risk

environment. This measure is important since the short-term

risk factor can at times become quite volatile, swinging

above and below the long-term average.

Long-Term Risk: This factor in our model is based on our

analysis of very long-term capital-markets return data from

each asset class. The long-term measure captures the

tendency of all asset classes to revert toward their long-term

averages over time.

To illustrate how these different measures work in concert to

form our dynamic risk forecasts, we sampled four time

periods. The display on the right breaks out the readings for

each of the three factorsshort-, medium- and long-term

riskand shows our resulting one-year global equity-market

volatility forecasts.

The Credit Crisis, 20072009The credit crisis, which started in mid-2007, gathered momen-

tum in February and March 2008, around the time of the

collapse of the investment bank Bear Stearns. In historical

simulations for this period, our short-term volatility readings

rose above 20%.

But soon after the Bear Stearns failure, short-term equity-market volatility began to fade, falling below its average.

Nevertheless, the initial pickup in volatility sent our medium-

term measure upward, highlighting the fact that we were in an

environment of high investor anxiety.

As a result, our volatility forecast remained at or above the

long-term average throughout the credit crisis. Once volatility

began to escalate again in September with the collapse of

Lehman Brothers, our short-term volatility measure quickly

registered the spike in market anxiety and caused our forecast

to shoot up simultaneously.

As the markets began to recover in 2009, volatility edged back

down toward its long-term norm. But by the end of September

2009, our tools were still counseling caution as volatility

remained above normal.

The TMT Bubble, 19982002The technology, media and telecommunications (TMT) bubble

peaked in March 2000. Its subsequent collapse was punctuated

by a series of shocks to the financial markets, including the

September 11, 2001, terrorist attacks, the bankruptcy of Enron

in December 2001, the collapse of WorldCom in July 2002 and

the demise of Arthur Andersen, Enrons audit firm, in August of

that year.

In simulations during this period, the short-term volatility

measure in our model fluctuated significantly, spiking and

subsiding several times. In concert, the factors in our model

worked well, highlighting the elevated risks and the need for

caution.

Stormy Weather: How Our Volatility Model Works


15/36


The 1990s Bull MarketThe mid-1990suntil the collapse of the Thai baht marked the

start of the Asian crisis in July 1997was a period of below-av-

erage global equity volatility. Historical simulations showed that

this lull would have been reflected in the short- and medium-

term volatility factors in our model. But our long-term metric

would have indicated that volatility was likely to move back up

somewhat toward its average. The resulting forecast would

have allowed for increased risk-taking given the low-volatility

environment, but would have sounded a note of caution giventhat depressed volatility was unlikely to persist indefinitely.

The 1987 Stock-Market CrashBlack Monday, when global stock markets plunged on

October 19, 1987, was not successfully signaled by our model

in the historical simulation, although our risk forecasts adapted

reasonably well to the sharp rally that followed. The crash

was preceded by a period of low volatility. The simulation

showed that our risk forecasts for the period would have

been moderate, reflecting reasonably low short- and

medium-term volatility, qualified by the assumption that

below-average volatility was likely to correct upward over

time.

The suddenness of the stock-market crash meant that our

risk forecasts would not have given early warning of thespike. However, we found that our return forecasts were

well below normal, reflecting increasingly expensive

valuations and rising interest rates, resulting in an under-

weight in equities. This illustrates the importance of

incorporating additional tools besides volatility forecasting in

measuring the risk/return trade-off. n

Volatility Forecasting: Global Equity Case Studies

Credit Crisis TMT Bubble

Mid-1990s Bull Market and Asian Crisis 1987 Market Crash

Short-Term VolatilityMedium-Term VolatilityLong-Term Volatility AllianceBernstein One-Year Forecast Volatility

Percent

Percent

Percent

Percent

0

20

40

60

Jan 08 May 08 Sep 08 Jan 09 May 09 Sep 09

0

10

20

30

40

Jan 99 Jan 00 Jan 01 Jan 02 Jan 03 Jan 04

0

20

40

60

Jan 87 May 87 Sep 87 Jan 88 May 88 Sep 88 Jan 89

0

6

12

18

24

Jan 95 Jul 95 Jan 96 Jul 96 Jan 97 Jul 97 Jan 98

Through September 30, 2009At times, our one-year forecast can be greater than its three components. This is mainly due to an adjustment for autocorrelation in daily returns.Source: Global Financial Data, MSCI and AllianceBernstein


16/36


Measuring the Diversification OpportunityVolatility is not the only aspect of market risk. Correlations can

also shift dramatically over time, either reducing or increasing

the diversification opportunities available across a portfolio.

For example, the correlations between equities and other asset

classes have historically moved in a surprisingly large range, as

illustrated by the three-year rolling correlations between equities

and various other assets since the 1970s (Display 8).

Changing correlations can leave investors exposed to more riskthan they realize if volatilities and correlations rise at the same

time. For example, a common rationale for holding foreign

currencies and commodities is their near-zero long-term average

correlations with global equities (0.07 and 0.04, respectively).

Adding them to a portfolio is typically not perceived as adding

much in overall risk. But if correlations riseas might happen if

fears of a global recession caused investors to flee all economi-

cally sensitive assetsthen benefits from diversification might

disappear, causing portfolio risk to increase.

On the other hand, there are times when correlations fall or

even turn negative, making the diversification opportunity

better than average.

By far the most important correlation for investors is the

relationship between global equities and interest ratesthe two

primary sources of risk in most portfolios. In the long run these

assets are very weakly correlated, but their relationship has at

times been both strongly positive and strongly negative.

For example, during the 1970s, when inflation was thedominant concern, the correlation between stocks and bonds

was well above its long-term average. This is typical of periods

of high inflation, when the fixed payments of bonds become

less attractive and the future cash flows from stocks less

valuable, simultaneously putting pressure on both asset classes.

When we back-tested our model by simulating how it would

behave in this environment, we found that during the 1970s it

would have indicated rising correlations and a diminished

diversification benefit from holding bonds (Display 9).

Display 8

Diversification Benefits Have Fluctuated Widely

Rolling Three-Year Correlations with US Equities

(1.0)

(0.5)

0.0

0.5

1.0

High

Low

Average

0.60

0.65 0.62

0.17

0.48

0.04 0.07

Non-USEquities

Emerging-MarketEquities

REITs USTreasuryBonds

Investment-GradeCredit

Commo-dity

Futures

ForeignCurrency

As of September 30, 2009. Periods under observation are: non-US equities, USTreasury bonds, commodity futures and foreign currencysince 1972; REITssince1976; emerging-market stockssince 1990; and investment-grade creditsince1991. See notes on page 25 for asset class definitions.Source: Barclays Capital, Bloomberg, FTSE NAREIT, Global Financial Data,MJK Associates, MSCI, Thomson Reuters and AllianceBernstein

Display 9

Capturing Changing Diversification Opportunities

Forecast Correlations: Global Bonds and Global Equities

69 73 76 79 83 86 89 93 96 99 03 06 09

(0.6)

(0.3)

0.0

0.3

0.6 Inflation Concerns

Economic Growth Concerns

Through September 30, 2009Source: Barclays Capital, Global Financial Data, MSCI and AllianceBernstein


17/36


In a bear market, when concerns about economic growth and

deflation are running high, equities and bonds tend to move in

opposite directions, with investors stampeding out of equities

into the safety of the fixed payments of government bonds. For

example, during the Russian crisis and the collapse of Long Term

Capital Management (LTCM) in 1998, the TMT crash of 2000

2002 and the credit crisis of 20072009, back-testing showed

that our model would have picked up signals of increasingly

negative correlations between stocks and government bonds,

indicating that bonds were offering a better hedge than usual.

At the same time, rising positive correlations between stocks

and other risky assets, such as REITs, would have signaled that

economically sensitive assets offered fewer diversification

benefits.5

Chapter Highlights

n Volatility fluctuates, but its trends tend to persist, allowing us to predict future volatility with a reasonable level of confi-

dence. In most situations, this helps us to identify and react to changing volatilities in time to adjust portfolio risk exposure.

n Different asset classes offer different degrees of diversification over time. By identifying when those changes are most likely

to happen, we can increase or decrease diversifiers to smooth out fluctuations in the portfolio.


18/36


How Much Return Is Enough?Measuring the Opportunity

While it is important to have effective measures of risk, in order

to make informed asset-allocation decisions, it is also crucial to

have a well-researched perspective on future returns. Even

when volatility is low, return forecasts can provide early warning

of building market pressures. And even when volatility is high,

return forecasts may be signaling improving conditions. The

question is, how much return is enough?

To accept more risk, an investor must have the potential to earn

more in return. Interestingly, the additional compensation

required is not linear. This is because investors are loss

aversethe pain they feel from incurring a large loss will

typically outweigh the pleasure they get from generating a

comparable portfolio gain (Display 10, left). So expected returns

need to increase at a faster rate than expected risks for investors

to be content to maintain their long-term asset-allocation

strategy. The chart to the right in Display 10 illustrates the

return requirement for an investor with 60% of his or her

wealth in an investment that has a long-run expected volatility

of 10% and an expected return over cash of 3% (a Sharpe

ratioin other words, extra return per unit of riskof 0.3).

If an investor is determined to keep a 60% weighting in the

asset, what risk/return trade-off is required if expected volatility

rises from 10% to 20%? With twice the volatility, the potential

for large losses is high, so the investor would need to generate

Display 10

Investors Are Averse to Large Losses...

Loss Gain

Pleasure

Small Pleasure

Big Pain

Pain

...so They Should Ask More in Return for Taking Large Risks

Risk/Return Trade-Off Required to Maintain Long-Term Allocation

0

3

6

9

12

15

5 10 15 20 25

Expected Volatility (Percent)

Expected

Excess

Re

turn

(Percent)

2 Risk

4 Return

Source: Amos Tversky and Daniel Kahneman, Advances in Prospect Theory: Cumulative Representations of Uncertainty, Journal of Risk and Uncertainty (1992)


19/36


four times the amount of returna 12% excess return

requirement, or a Sharpe ratio of 0.6.6

If the investor does not believe that such high returns are likely,

he or she should reduce risk by cutting the allocation to

equities, or look for more attractive investment alternatives

elsewhere.

Are investors justified in assuming that a rise in volatility will be

accompanied by an even greater rise in return? Market theory

suggests that when risk rises, assets should reprice immediatelyso expected returns adjust upward. But our research shows that

this is not something that can be taken for granted. Historically,

there is little evidence to support the idea that returns go up

following a phase of high volatility.

We looked at the major asset classes to see what returns were

generated in the 12 months after periods when volatility was at

its highest and its lowest (Display 11). For example, in the equity

market, we found that, on average, investors received less,

rather than more, compensation for exposure to a high-volatility

environment.

Historically, in the 20% of occasions when the S&P 500 Index

was at its most volatile, its average annualized volatility was

about 35%. In the 20% of occasions when the index was least

volatile, it averaged 8%. In the subsequent 12-month periods,

the average excess return over cash following low volatility was

7.6% while the return after high volatility was only 5.3%.

Investors received less compensation following periods of high

riskthe opposite of what theory would predict.

The same pattern emerged in most other asset classes. Govern-

ment bonds were the only asset class where low volatility wasfollowed by weaker returns and high volatility resulted in slightly

above-average returns. But even there, the increase in return

was far less than the increase in volatility. The same analysis over

three- and five-year periods showed no basis for believing that

investors should expect to get paid more for persevering

through periods of high volatility.7

In other words, investors would have done better to start with

the assumption that they should reduce exposure when volatility

is high and add exposure when volatility is low. Of course, its

not that simple. There are times when the odds point to higher

Display 11

High Volatility Is Not Necessarily Followed by High Returns

34.5

8.0

23.9

6.8

8.7

2.1

13.1

7.0

21.8

6.2

S&P 500

Global Equities

Global Government Bonds

Foreign Currency

Commodity Futures

5.3

7.6

4.2

6.5

2.4

1.0

1.5

2.0

3.74.6

Past Volatility (Percent) 12-Month Forward Excess Returns (Percent)

As of June 30, 2009. Past volatilities are sorted on a monthly basis into quintiles (terciles in the case of foreign currency); the bars refer to the highest and lowest quintiles (orterciles). Past volatility is an exponentially weighted average using daily data with a three-week half-life (5% decay per day). Excess returns refer to returns over cash.Periods are since: S&P 5001928; global equities1970; global government bonds1970; foreign currency1974; commodity futures1970. See notes on page 25 for assetclass definitions. Source: Barclays Capit al, Bloomberg, FTSE NAREIT, Global Financial Data, MJK Associates, MSCI, S&P, Thomson Reuters and AllianceBernstein


20/36


returns and investors are more likely to be rewarded for taking

risk. The aim of our dynamic asset-allocation tools is to identify

these occasions and determine how much of an adjustment to

the asset allocation is warranted.

In order to decide whether the return opportunity calls for a

change in portfolio exposure, other factors, such as valuation,

need to be taken into account. As in the previous example,

Display 12 shows excess returns to the S&P 500 Index, but

focuses just on the excess returns that were generated in the 12

months after periods when volatility was at its highest.

While the average excess return to equities following the

highest-volatility periods was about 5.3%, sorting these

observations by earnings to price (the markets earnings yield)

showed considerable variation. When valuations were least

expensive (earnings were high relative to the market price),

investors were well rewarded over the following year, earning

an average of 15%. But when volatility was high and valuations

were not already low, exposure to equities resulted in losses of

5.5% on average.

A Multilayered Approach to Return ForecastingValuation is just one of many factors that should be considered

when trying to predict returns for any given risk level. In

constructing our return forecasts, we believe it is important to

take a multifaceted view of the opportunity, taking into account

the markets current view of expected returns, historical

risk/return relationships, and an array of market and economic

indicators that help to explain the likely path of future returns.

We begin by assessing the markets view of the return opportu-

nity. This is a useful starting point because market pricingincorporates a large amount of information about investors

views on the attractiveness of various asset classes. When the

market value of an asset class moves higher (or lower), investors

are voting with their wallets, indicating that, all else being

equal, they expect future returns to be higher (or lower).

To arrive at an estimate of the markets expected return for each

asset class, we gather data on the market values of all publicly

traded assets (this gives us each assets weight in the market

portfolio), the volatility and correlations of the assets, and an

estimate of the average investors risk profile. Based on this

information, we estimate the returns that the market requires in

order to be comfortable holding its current portfolio.

Of course, the market is not always right about future returns. It

has a tendency to get ahead of itself at times, becoming overly

optimistic or pessimistic. We try to improve on the market view

by counterbalancing it with two additional components of our

expected return model: an estimate of the typical compensa-

tion for risk available for each asset class and a multifactor

model of returns.

Our estimate of the typical compensation for risk takes account

of long-term risk/return relationships and current volatilities and

correlations in order to provide an estimate of returns at each

point in time. The purpose of this estimate is to pull back our

forecasts toward long-term averages at times when the market

seems to be overly optimistic or pessimistic about an asset class.

So, when the market seems to be pricing in extremely high

expectations about the likely reward per unit of risk, we inject a

level of skepticism into our forecast. Likewise, when the market

seems overly conservative in its expectations, our forecast is

likely to be more optimistic.

Display 12

Risk-Taking Is Better Rewarded When Valuations Are Low

S&P 500 12-Month Forward Excess Returns

During the 20% of Most Volatile Periods19282009

Quintile of Earnings to Price

(5.5)%

5.3%

15.0%

Most Expensive

Average

Least Expensive

Through June 30, 2009Source: S&P and AllianceBernstein


21/36


While historical risk/reward relationships sharpen the focus of

our forecasts, they do not tell the whole story. We use a

multifactor return model to refine our expected return forecasts

further, capturing a range of important factors from across the

global capital markets. We incorporate information from the

equity markets on valuations, corporate profitability, earnings

quality and market momentum; we look at bond-market

information, such as interest rates and the shape of government

bond yield curves; and we include information from other

markets, such as credit spreads, real estate valuations and key

economic indicators.

For each asset class, we then seek to determine which factors

will be most important in explaining future returns and the

relative weight of each factor in our model. The more extreme

the value of a particular indicator becomes, the greater its

influence will be on our return expectations. This allows the

main controversies of the day to be expressed in our forecasts.

Display 13 shows the most decisive factors in our expected

return model for equities, bonds and REITs, at five-year intervals

since 1980. Weve found that most asset classes are sensitive to

factors from multiple markets. For example, when forecasting

equity and REIT returns, we not only incorporate equity and REIT

factors, but we also include bond-market variablessuch as

short-term interest rates and recent trends in 10-year govern-

ment bond yields. By the same token, our fixed-income

forecasts consider stock-market and economic indicators, which

provide additional perspectives on future economic growth and

inflation.

Some factors influence multiple asset classes, but in different

directions. For example, very low short-term interest rates tend

to signal the bottom of an economic cycle, which implies that

their next move is upward as accelerating growth and inflation

pressures trigger central bank interest-rate increases. This

scenario tends to be bad for government bonds but good for

stocks and REITs. The opposite applies when short-term interest

rates are unusually high.

Display 13

Key Drivers of Asset Return Forecasts Vary over Time

Largest Values in Our Expected Return Calculation

Yield Momentum REIT Valuation REIT Valuation Yield Momentum REIT Momentum REIT Momentum

REIT Valuation

REITs

Return on EquityEquity Valuation Equity Valuation Return on EquityYield MomentumReturn on Equity

REIT Valuation

Yield Momentum Equity Valuation Equity Valuation Equity Valuation

Equities

SlopeInflationEquity MomentumShort RatesShort RatesSlope

Equity MomentumSlopeShort Rates Slope Slope Short Rates

Yield Momentum REIT Momentum Real Short Rates Real Short Rates

Bonds

200520001995199019851980

Positive for Expected Returns Negative for Expected Returns

Yield MomentumYield Momentum

Credit Spread

Return on Equity

Credit Spread

Slope

Short Rates

REIT Valuation

2009

As of September 30, 2009. Factors are as follows: equity valuationprice/book, price/earnings and dividend yield; equity momentumequity total return over past year; returnon equityreported trailing earnings/book value; credit spreaddifference in yield between non-government bonds and government bonds of comparable maturity; slopedifferencebetween three-month and 10-year yields; short ratesthree-month yields; real short ratesthree-month yields adjusted for inflation; yield momentumrecent trend in 10-year

government bond yields; inflationconsumer price index; REIT valuationdividend yield; REIT momentumrecent trend in REIT prices relative to equities.Source: Barclays Capital, FTSE NAREIT, Global Financial Data, MSCI and AllianceBernstein


22/36


23/36


market that was speaking particularly loudly was credit. This

was not surprising given that credit had been an early victim of

the subprime mortgage crisis in 2007. Spreads on corporate

bonds had moved sharply wider as credit sold off, preceding the

most dramatic of the falls in equity valuations and interest rates,

and acting as a potential warning signal of a large disconnect in

the markets.

The fact that credit-market factors had become such a domi-

nant number in our results meant that our return forecasts were1.7% lower than they might otherwise have been. Accordingly,

driven more by our return expectations than by our volatility

forecasts, our dynamic asset-allocation tool set would have

recommended an underweight in equities in August 2008.

By August 2009, six months after the low point in the credit

crisis, our return forecast was reflecting a different picture.

Corporate bond spreads had tightened following a strong

credit-market rally, equity valuations had risen, central bank

policy rates remained extraordinarily low and long-dated

government bond yields were below average. Collectively,

these factors significantly increased our estimate of the

excess return opportunity in equities, to 7.6%.

Given the increase in expected returns, we would have beenadding to our equity exposure by this stage. But, with risk

levels still well above their historical norms, our tool set

would still have been calling for a modest underweight in

this case. n

Return Forecasts Seek to Identify Gathering Storms and Improving Conditions

Forecasts for Global Equity Excess Returns

2.1%

7.9%

3.2%

7.6%

5.2% 5.2% 5.2% 5.2%Long-Term Equity Risk Premium

+1.5 +1.8 +0.8 +1.0Global Market Sentiment

+0.1 +0.1 +0.0 +0.5Typical Risk/Return Relationship

(3.4) +0.6 (0.9) +1.6Equity Factors

(0.3) (0.3) (1.7) (1.2)Credit Factors

(1.0) +0.5 (0.3) +0.5Government Bond Factors

2.1 7.9 3.2 7.6Total Excess Return

15.8% 21.8% 17.0% 19.6%Volatility Forecast

February 2000 March 2003 August 2008 August 2009

TMT Collapse Credit Crisis

As of September 30, 2009Source: Barclays Capital, Global Financial Data, MSCI and AllianceBernstein


24/36


Models that rely exclusively on past market patterns often fail. Our forecasts take account of historical market events, stress-

tested across a wide array of global markets and periods, but also give weight to fundamental insights and market theory. In

our portfolio construction, we limit the aggressiveness of our forecasts given the difficulty of predicting market returns with

great certainty. We also ensure consistency by making sure that return forecasts for highly correlated assets are linked. These

features help us to express our insights without making overly aggressive changes to our asset-allocation recommendations. n

Risk and Return Forecasts Must Be IntegratedSo far, we have argued that asset allocation based solely on the

return opportunity ignores crucial information about risk.

Similarly, focusing solely on volatility ignores important signals

embedded in return forecasts. We believe the key is to frame

the question in terms of return per unit of risk. In other words,

when the risks change, are we adequately compensated?

Sometimes the answer to this question is obvious. When risk is

low and expected returns are high, it makes sense to add

exposure to an asset. We found that in most cases where riskswere below average and expected returns were above average,

our framework would have called for an overweight in equities

(Display 14). In high-risk, low-return situations, the decision is

often equally clear-cut: reduce equity exposure.9

But, more commonly, the risk/return trade-off is a complex one.

Expected returns may be rising, but may not be high enough to

justify increasing exposure to an asset. This would have been

the case in late 2008, when our return expectations would have

been above average but the risks would have outweighed the

return opportunity. In high-risk, high-return environments, we

would have been overweight only 47% of the time.

Similarly, low expected returns do not necessarily call for a

reduction in equity exposure. There are times when our equity

return forecasts are below average, but so are the risks, so that

an overweight in equities might provide the best trade-off. One

such period was 20032005, when bargains were increasingly

scarce as equities recovered from the TMT bubble, but equity-

market volatility was also very low by historical standards. Our

tool set would have called for an overweight 37% of the time

when both risks and expected returns were below average.

Display 14

Opportunities Must Be Weighed Against Prevailing Risks

Monthly Global Equity Positioning (19702009)

Expected Volatility

ExpectedReturn

Low Risk, High Return

Low Risk, Low Return

High Risk, High Return

High Risk, Low Return

Overweight Underweight

10%

90%

47%53%

37%

63%

18%

82%

Through September 30, 2009Source: MSCI and AllianceBernstein

n Its not safe to assume that times of exceptional volatility offer exceptional return opportunities. Historically, the highest-

volatility periods have generally not yielded the best returns.

n On their own, there is only so much that forecasts of risks or returns can tell us. The key is to integrate the two perspectives,

framing the opportunity in terms of return per unit of risk.

Chapter Highlights


25/36


Achieving More Consistent OutcomesThe Portfolio Impact of Dynamic Asset Allocation

We know that a well-designed long-term asset allocation is

crucial to the success of any investment program. We also know

that market risk structures and return opportunities are

constantly changing. So how can we use this insight to enhance

a given long-term asset-allocation strategy?

Smoothing VolatilityOur research suggests that dynamic asset allocation may be

helpful in reducing portfolio volatility. The simulation results in

Display 15 show the volatility of a dynamically managed

portfolio that could invest in global stocks, REITs, bonds and

cash, and the volatility of a static portfolio that had a long-run

average of 55% stocks, 35% bonds, 10% REITs and 0% cash,

rebalanced monthly. At times of moderate volatility, the dynamic

strategy behaved in much the same way as the rebalanced static

allocation, but at times of high volatility there were significant

differences. The static portfolio was susceptible to large

fluctuations during each of the bear markets of the last 40

years, whereas the dynamic approach was more stable.

As a result of their different behavior in high-volatility periods,

the dynamic strategys long-term annualized volatility was

considerably lower than that of the static portfolio7.8%

compared with 9.2%.

Display 15

Dynamic Asset Allocation Can Result in a Smoother Ride over Time

12-Month Rolling Volatility

Static Rebalanced Dynamic Allocation

Long-Term Average

7.8%9.2%

Static Dynamic0

5

10

15

20

70 73 76 79 82 85 88 91 94 97 00 03 06 09

Percent

Through September 30, 2009The performance depicted above is hypothetical and is derived from a back-tested simulation. Please read Note on Simulation Results on page 30 for importantadditional information.Static portfolio results are based on a portfolio that is 55% MSCI World Index, 35% Barclays Global Aggregate Index (as adjusted to reflect duration only) and 10% FTSENAREIT, rebalanced halfway back to target when weights become +/5% from their long-term target. For physical securi ty positions, we assume one- way transaction costs of 0.6% for equities and bonds and 1.0% for REITs. For equity and bond derivatives, we assume total one-way transaction costs and cost of financing of 0.5%.Source: Barclays Capital, MSCI and AllianceBernstein


26/36


Since the dynamic approach tends to scale back exposure

during periods of elevated volatility, investors can experience a

smoother, more consistent pattern of returns. The bumpier ride

provided by the static allocation reflects more frequent discon-

nects from the investors desired returns and risk profile because

current market forces are not taken into account.

Trimming the TailsWe argued at the beginning of this paper that long-term

investors are justified in worrying about short-term risks because

of the damage done by extreme (tail) events. Display 16shows simulated total returns generated by the dynamic asset-

allocation approach during bear-market periods and in the

recoveries that followed.

During the bear markets in our study, the dynamic approach

would have significantly outperformed the static allocation,

mainly by reducing the portfolios exposure to high volatility.

Dynamic allocation would have outperformed in all five of the

major bear markets of the past 40 years.

For example, in the simulation the dynamic approach lost 23%

during the credit crisis from 2007 to early 2009, compared with

34% for the static allocation. And during the TMT collapse, the

dynamically managed portfolio lost 11%, compared with 18%

for the static allocation. On average, when the markets fell, the

loss suffered by the dynamic strategy was 20% less severe than

that of the static allocation.

Of course, there is no such thing as a free lunch. In the 12-

month periods following bear-market troughs, the dynamic

approach underperformed the static allocation, typically because

still-elevated volatility and correlation forecasts were calling for

more cautious exposure to the markets. This meant that not asmuch of the upside was captured. For example, as the markets

rallied between the beginning of March and the end of

September 2009, the simulated dynamic portfolio returned

21% compared with the static allocations 30%. And in the

recovery following the Black Monday crash in 1987, the

dynamic approach lagged the static portfolio by about 3%.

But while the dynamic strategy tended to lag initially during

recoveries, over the entire simulation history we found that it

performed nearly as well in rising markets, capturing more than

90% of the gains that a static mix would have achieved. In

other words, the dynamic portfolios ability to keep more of its

Display 16

Dynamic Approach Can Help Outperform in Bear Markets

(23)

(11)(18)Jan 00Sep 02

(12)(18)

(7)(15)

(34)

14

22

19

30

13

19

16

21Oct 07Feb 09

Jan 90Sep 90

Sep 87Nov 87

DynamicAllocation Relative

StaticRebalanced

7

8

11

6

Dec 87Nov 88

Oct 02Sep 03

Mar 09Sep 09

Oct 90Sep 91

DynamicAllocation Relative

StaticRebalanced

(1)

(23)(27)Jan 73Sep 74 22 214 Oct 74Sep 75 (1)

(3)

(9)

(3)

Simulated Performance During Bear MarketsPercent

Simulated Performance During Recoveries (Year After Decline)Percent

Through September 30, 2009The performance depicted above is hypothetical and is derived from a back-tested simulation. Please read Note on Simulation Results on page 30 for importantadditional information.Static portfolio results are based on a portfolio that is 55% MSCI World Index, 35% Barclays Global Aggregate Index (as adjusted to reflect duration only) and 10% FTSENAREIT, rebalanced halfway back to target when weights become +/5% from their long-term target. For physical securi ty positions, we assume one- way transaction costs of 0.6% for equities and bonds and 1.0% for REITs. For equity and bond derivatives, we assume total one-way transaction costs and cost of financing of 0.5%.Source: Barclays Capital, FTSE NAREIT, Global Financial Data, MSCI and AllianceBernstein


27/36


capital intact during downturns, and its greater flexibility in

consistently exploiting day-to-day return opportunities, largely

compensated for its modest underperformance in recoveries.

Display 17shows how the distribution of returns of a static

portfolio compared with the returns of a dynamic allocation

approach over the past four decades. We found that more of

the dynamic strategys returns would have fallen within the

(10)%20% range, reflecting lower average portfolio volatility.

Tail events were also less frequent. The incidence of extreme

gains was lower, but so was the frequency of extreme losses,

with annual losses of more than 20% reduced from eight

occurrences to just one.

Enhancing Risk-Adjusted ReturnsAlthough boosting returns is not a primary aim of our dynamic

asset-allocation tool set, simulations showed that a dynamic

approach would have generated slightly higher average total

returns than a conventional balanced strategy since 1970. Much

of the time, these extra returns were picked up during risk-

reduction periods, but at other times they were generated by

exploiting return opportunities during periods of normal

volatility.

Compared with the rebalanced 55% equity, 35% bond,10%

REIT portfolio discussed above, we found that a dynamic

approach significantly enhanced risk-adjusted total returns,

resulting in a Sharpe ratio of 0.46 compared with 0.36 for the

static allocation (Display 18, page 24). We found that including

more asset classes in the asset-allocation decision could improve

risk-adjusted returns even further.

Its worth noting that these results were achieved without

dramatic changes in portfolio weights. For example, the most

turbulent phase of the credit crisis in late 2008 would have

called for a weighting of about 61% in bonds, while the lowest

bond allocation, in 1978, was about 15%. We found that the

bond allocation would have been within 20% of the long-term

target for roughly 90% of the time.

Display 17

Dynamic Allocation May Reduce the Frequency of Extreme Losses

Frequency of Rolling 12-Month Returns19702009 (Simulated)

73

64

8

19

1

15

366

386

Below (20)% (20)%(10)% (10)%20% Above 20%

Static Rebalanced

Dynamic Allocation

Through September 30, 2009The performance depicted above is hypothetical and is derived from a back-tested simulation. Please read Note on Simulation Results on page 30 for importantadditional information.Static portfolio results are based on a portfolio that is 55% MSCI World Index, 35% Barclays Global Aggregate Index (as adjusted to reflect duration only) and 10% FTSE

NAREIT, rebalanced halfway back to target when weights become +/5% from their long-term target. For physical securi ty positions, we assume one- way transaction costs of 0.6% for equities and bonds and 1.0% for REITs. For equity and bond derivatives, we assume total one-way transaction costs and cost of financing of 0.5%.Source: Barclays Capital, FTSE NAREIT, Global Financial Data, MSCI and AllianceBernstein


28/36


ConclusionThe long-term asset-allocation decision is one of the most

important decisions an investor is ever likely to make, but we

believe that complementing the long-term allocation with a

dynamic asset-allocation strategy can add further value by

making portfolio positioning more sensitive to short- and

medium-term fluctuations in forecast risk and return.

While much of the existing body of research on dynamic asset

allocation focuses on boosting investment returns, we believe

that the strategy has more to offer from a risk-management

perspective. We believe that a dynamic approach can create a

more consistent fit between investor objectives and portfolio

outcomes, smoothing volatility and reducing the incidence of

outsize losses, without necessarily sacrificing return potential.

Display 18Dynamic Allocation Can Enhance the Risk/Return Trade-Off

Total Return

Volatility

Sharpe Ratio

9.1%

9.2%

0.36

Static Rebalanced

9.5%

7.8%

0.46

Dynamic Allocation

+0.4%

(1.4)%

+0.1

Change

Historical Simulation: Asset Allocations

0

25

50

75

100

70 73 76 79 82 85 88 91 94 97 00 03 06 09

Cash Equities REITs Bonds

Percent

Through September 30, 2009The performance depicted above is hypothetical and is derived from a back-tested simulation. Please read Note on Simulation Results on page 30 for importantadditional information.

Static portfolio results are based on a portfolio that is 55% MSCI World Index, 35% Barclays Global Aggregate Index (as adjusted to reflect duration only) and 10% FTSENAREIT, rebalanced halfway back to target when weights become +/5% from their long-term target. For physical security positions, we assume one- way transaction costs of 0.6% for equities and bonds and 1.0% for REITs. For equity and bond derivatives, we assume total one-way transaction costs and cost of financing of 0.5%.Source: Barclays Capital, FTSE NAREIT, Global Financial Data, MSCI and AllianceBernstein

Chapter Highlights

n Dynamic asset allocation can make the investment experience less turbulent in times of market upheaval by smoothing out

volatility and reducing extreme outcomes.

n In back-testing, the dynamic approach tended to outperform in bear markets, while lagging somewhat in recoveries.


29/36


Notes

1 Global equity returns refer to the MSCI All Country World Index after November 2000 and a market-weighted combination of the MSCI

World and MSCI Emerging Markets indices before that.

2 Investment-grade corporate bond outperformance refers to the excess returns relative to government bonds of the Barclays Global AggregateCorporate Index, hedged into US dollars.

3 In this example, if volatility fell to 5%, that would reduce both the upside and downside risk, to a gain of 17% and a loss of 3%. This assumes a

95% level of confidence. In statistical terms, assuming returns occur in a normal distribution, if an investment has an annual expected return of

7% and a volatility of 9%, there is a 68% probability of generating returns in a range of 7% plus or minus 9%, in other words (2)%16% (a one-

standard-deviation event). There is a 95% probability of generating returns in a range of 7% plus or minus 2 9%, in other words (11)%25%

(a two-standard-deviation event).

4 As a measure of past volatility, we used an exponentially weighted average using daily data with a three-week half-life (5% decay per day).

5 While correlations vary significantly over time, we will tend to forecast somewhat slower shifts in correlations than volatilities. This is because

trends in correlation are more difficult to discern from very recent events. Consider a time when volatility rises for stocks and bonds. This

could mean one of two things for correlations: either they will increase (typically when inflation is a concern) or they will decrease (typically

when deflation or economic growth is a concern). Until it is clear which source is driving the volatility, we cant draw any conclusions about

future correlations. Short-term measures can be noisy, so we rely more heavily on medium-term measures, which we find to be more

effective in determining which factor is likely to be driving correlations and their likely value in the future.

6 This assumes an unconstrained investor with a portfolio consisting of one risky asset class and cash.

7 For example, S&P 500 excess returns following low volatility averaged 7.6% after one year, 7.4% after three years and 7.1% after five years.

After high volatility, the average was 5.3% after one year, 6.2% after three years and 6.6% after five years. For global equities, the results were

6.5%, 4.2% and 5.3% after high volatility and 4.2%, 3.4% and 3.0% after low volatility. Excess returns for fixed income were 1.0%, 0.7% and

1.0% versus 2.4%, 2.8% and 2.9%; for currencies 2.0%, 1.9% and 1.6% versus 1.5%, 1.4% and 1.5%; and for commodities 4.6%, 6.9% and 5.2%

versus 3.7%, 4.1% and 4.6%.

8 The display does not show the contribution of economic factors, because during the two periods under discussion they were not a material

driver of expected returns in our models.

9 As shown in the display, in isolated cases our framework would have been overweight when risks were above average and returns were below

average, or underweight in the opposite scenario. The most likely cause for this is correlation considerations; for example, when the diversifica-

tion opportunity offered by bonds was unusually attractive or unattractive.

Asset class definitions are as follows: global equitiesMSCI All Country World Index or MSCI Developed World Index; emerging-

market equitiesMSCI Emerging Markets Index; Non-US equitiesMSCI Europe, Australasia and Far East (EAFE) Index; REITs

FTSE NAREIT Global Real Estate Index; high-yield creditBarclays Capital US High Yield Index; investment-grade creditBarclays

Capital US Investment Grade Index; global government bondsBarclays Capital Global Aggregate Treasury Index; US Treasury

bondsBarclays Capital US Treasury Index; commodity futuresproprietary composite; foreign currencyGDP-weighted basket of

currency returns relative to the US dollar.


30/36


Glossary

BenchmarkA standard barometer against which investments can be

measured in terms of performance, characteristics, construction

and similar criteria. Sometimes widely recognized instruments

(e.g., US Treasuries) or interest rates (e.g., the US fed funds rate

or LIBOR) serve as benchmarks. More commonly, a benchmark is

composed of an unmanaged group of securities with the same

general characteristics as the portfolio being measured against

it. Stock indices such as the S&P 500, the FTSE 100 and the

Nikkei 225 are commonly used for equities, while indices like

the Barclays Global Aggregate are often used in fixed income.

BondA security that pays interest. The issuer agrees to pay the

bondholder a regular set sum based on the amount borrowed

and the bonds coupon, and to repay the principal amount of

the loan at a future date. Many variations exist on this basic

format, including bonds with no coupon and with variable

coupons. The price of a bond is quoted assuming a par value of

100; thus, if a bond price is quoted as $90 and the principal

value of the actual holding is $1,000, that holding is valued at

$900. A bond selling above 100 is said to be trading at a

premium; at 100, at par; and below 100, at a discount. The

price varies over the life of the bond as interest rates, perceived

credit quality and other factors fluctuate, and as the bond

approaches maturity. A bonds price is inversely related to its

yield: it rises when the bonds yield falls and declines when the

yield rises.

CorrelationA statistical measure of the relationship between two variables.

Possible correlations range from +1 to (1). A zero correlation

indicates that there is no relationship between the variables; in

other words, a change in one variable will be matched by a

totally random change in the other. A correlation of (1) indicates

a perfect negative correlation, meaning that if one variable rises

relative to its own average, the other always falls relative to its

own average. A correlation of +1 indicates a perfect positive

correlation, meaning that if one variable rises relative to its

average, the other variable does the same.

DurationA measure of a bonds price sensitivity to changes in interest

rates, expressed in years. Duration approximates how much a

bonds price will change if interest rates change by a given

amount. For each year of duration, a bonds price will fall (or

rise) roughly one percentage point for each one-percentage-

point increase (or decrease) in yield. Thus, a bond with a longer

duration will perform worse when rates rise than a bond with a

shorter duration; conversely, it will perform better when rates

fall. Technically, duration is the weighted average term to

maturity of the bonds cash flows. Thus, it is shorter than

maturity for coupon-bearing bonds, which make annual or

semiannual payments throughout the life of the bond. Duration

is a good approximation of price sensitivity when interest-rate

changes are small, but less so when interest-rate changes are

large.

EquityOwnership of a company in the form of shares that represent a

claim on the corporations earnings and assets. Common-stock

holders have the right to vote on directors and other key

matters. Preferred-stock holders do not have voting rights, but

have priority when it comes to dividend payments. A firm can

authorize additional classes of stock, each with its own set of

contractual rights.


31/36


Equity Risk PremiumA forward-looking estimate of how much equities are likely to

outperform bonds. Equity investors typically demand a higher

return due to their greater risk of not receiving cash flows for

their investment.

Excess ReturnThe difference between returns, which may be applied to

managers or sectors. When referring to a manager or portfolio,

the excess return is typically the same as the active returnthe

difference between the managers or portfolios return and thatof the benchmark. A fixed-income sectors excess return is the

difference between its return and that of a comparable-duration

government bond: if short-term corporate debt returns 6% and

a short-term government security returns 4%, the excess return

is 2%. (See Risk-Free Rate.)

Information RatioThe ratio of a portfolios excess return, or premium, to its

tracking error, or the standard deviation of the premium over

the period being measured. It is designed to measure how much

excess return a manager delivers for each unit of risk. A higher

number indicates a more favorable balance of reward to risk

than a lower number. A positive information ratio indicates that

the portfolio outperformed, and a negative number indicates

that the portfolio underperformed. (See Excess Return and

Sharpe Ratio.)

Market ValueThe current price of a security in the market, as reflected by the

last reported price on an exchange, or the current bid-ask

spread if the security is traded over the counter.

Normal DistributionThe frequency distribution of a set of data that follows a bell-

shaped curve. The most frequent values are clustered around

the mean and fall off smoothly on either side of it. Extremely

large values and extremely small values are rare and occur near

the tail ends. In a normal distribution, 68% of observations fall

one standard deviation above or below the mean, while 95% of

observations fall two standard deviations above or below the

mean and 99.8% fall within three standard deviations. There

are other kinds of distribution. For example, in a fat tailed

distribution, the extremities are larger than those of a normal

distribution, implying a higher probability of experiencing

extreme values. (See Standard Deviation and Tail Event.)

Return on Equity (ROE)A measure of how much profit a company is able to generate

with the capital provided by shareholders. This measure is

calculated by dividing after-tax income for a specified time

period (e.g., trailing 12 months, trailing five years, forward 12

months) by the book value. Return on equity is expressed as a

percentage.

RiskIn common parlance, the chance of loss or of something bad

occurring. In financial parlance, it usually means the uncertainty

of outcomes due to one or many causes; it can be positive as

well as negative. Risk is usually measured by the standard

deviation of returnsin other words, the extent to which

returns may vary from the norm. Volatile assets tend to have a

wider range of possible returns and thus are said to be higher-

risk.

Risk-Free Rate

An investment with a predictable rate of return. An example is a

short-term government bond. A short-term government bond

has the explicit backing of a government, and the time period

before the bond matures is short enough to minimize the risks

of inflation and market interest-rate changes. Its yield is

therefore considered risk-free.

Sharpe RatioA measure of the risk-adjusted return of a financial security (or

asset or portfolio). It compares the excess return of an asset

against the return of a risk-free asset such as cash or govern-

ment bonds and divides that by the volatility of the excess

return. (See Information Ratio.)

SpreadThe difference between two variables, such as a securitys bid

and ask prices (bid-ask spread). In the corporate bond market,

the yield spread is the difference in yield between two bonds,

most often between the yield of a corporate bond and a

benchmark, such as a government bond, of comparable

maturity. Valuation spreads measure the difference between

expensive and cheap segments of the market.


32/36


Standard DeviationA statistical measure of risk that shows how aligned or at

variance the returns of an asset, industry or fund are relative to

their historical performance.

Tail EventExtremely large values and extremely small values, which are

rare and occur near the tail ends of a frequency distribution.

(See Normal Distribution.)

Transaction CostsThe costs incurred when buying or selling an asset security, such

as commission, fees and any indirect taxes.

VolatilityThe extent to which the price of a financial asset or market

fluctuates, measured by the standard deviation of its returns.

Volatility is a commonly cited risk measure.

YieldA component of the return on an investment. A shares dividend

yield is its annual dividend payment as a percentage of its

market price. A bonds yield is its annual interest payment as a

percentage of its market price. Measures of yield include current

yield, which considers only coupon interest, and yield to

maturity, which is the rate that equates the present value of the

bonds expected cash flows with its market price.

Yield Curve

A line connecting the yields of bonds from one end of thematurity spectrum to the other. Because yields typically rise

sharply at the short end of the spectrum and rise more gradually

at longer maturities, the plotted line usually forms a curve.


33/36



34/36


Note on Simulation Results

The asset-allocation framework discussed in this paper is

a new strategy for which actual data are not yet avail-

able. The portfolios and their performance are hypotheti-

cal and do not represent the investment performance or

the actual accounts of any investors. The securities in

these hypothetical portfolios were selected with the full

benefit of hindsight, after their performance over the

period shown was known. The results achieved in our

simulations do not guarantee future investment results.

The model performance information in this presentation is based

on the back-tested performance of hypothetical investments over

the time periods indicated. Back-testing is a process of

objectively simulating historical investment returns by applying a

set of rules for buying and selling securities, and other assets,

backward in time, testing those rules, and hypothetically

investing in the securities and other assets that are chosen.

Back-testing is designed to allow investors to understand and

evaluate certain strategies by seeing how they would have

performed hypothetically during certain time periods.

It is possible that the markets will perform better or worse than

shown in the projections; that the actual results of an investor

who invests in the manner these projections suggest will be

better or worse than the projections; and that an investor may

lose money by investing in the manner the projections suggest.

The projections assume the reinvestment of dividends and

include transaction costs of 0.6% for purchases and sales of

equities and bonds and 1.0% for real estate investment trusts

(REITs). For equity and bond derivatives, we assume total

one-way transaction costs and cost of financing of 0.5%. We

assume no deduction for advisory fees, and that assets are

allocated in the manner the projections suggest for nearly 40

years and are rebalanced monthly.

Although the information contained herein has been obtained

from sources believed to be reliable, its accuracy and complete-

ness cannot be guaranteed. While back-testing results reflect

the rigorous application of the investment strategy selected,

back-tested results have certain limitations and should not be

considered indicative of future results. In particular, they do not

reflect actual trading in an account, so there is no guarantee

that an actual account would have achieved the results shown.

Back-tested results also assume that asset allocations would not

have changed over time and in response t

designing a smoother ride

Documents