Epoch Investment Partners, Inc.

april 18, 2011

Free-Cash-Flow Investing: A Value Strategy

by kenneth n. hightower, analyst, quantitative research & risk management david pearl, co-cio and portfolio manager william w. priest, co-cio and portfolio manager michael welhoelter, portfolio manager and head of quantitative research & risk management

executive summary
Free-cash-ow investors follow the same process as value investors. Using a valuation metric (free-cash-ow yield) to narrow the investable universe down to a manageable list of candidates, free-cash-ow investors uncover value by determining which companies are ecient allocators of capital. This focus on the capital allocation decisions of the rm instead of on traditional accounting-based nancial metrics can be interpreted as what we term passive style drift under unusual and highly volatile market conditions. We believe an alternative systematic risk framework, grounded in the capital budgeting decisions of the rm, better encapsulates the systematic risks faced by the free-cash-ow investor.

proposition of a company lies within its sources and uses of free cash ow. In this way, we approach the investment problem in much the same way as the managers of the very rms in which we invest. It is akin to a capital budgeting decision. There is, however, a downside to our dierentiation. Although asset allocators want their managers to have an edge, they become nervous when this edge results in a signicant deviation from factor characteristics of the assigned benchmark. The job of an asset allocator is to manage their own or their clients exposure to systematic risk factors, which means that they feel most comfortable when beta and other benchmark characteristics trend with the benchmark while alpha exceeds it. This goal is easiest to achieve when the investment manager retained by the asset allocator ts nicely into a particular style box. This focus on investment styles has its roots in Fama and Frenchs series of papers1 from the early 1990s, which separate systematic risk along three dimensions: market exposure, size, and value/growth (where value and growth are dened by accounting metrics). At Epoch, we eschew the accounting metrics that underpin the traditional notions of value and growth. Our style can, therefore, under extraordinary market conditions, appear to be dierent from our stated value style: a phenomenon that, over time, manifests itself as style drift. As previously noted, style drift is anathema to the asset allocator. When managers game their style box by actively changing their systematic risk exposures, it confounds the style expectations at the asset allocation level. There is another, more benign type of style drift which we term passive style drift. Passive style drift occurs when the systematic biases of an investment process combine with extreme events to cause the appearance of style drift when viewed through the classic accounting-based nine-box model. An example is the drift from value to growth and back again experienced by the Epoch U.S. Large Cap Value portfolio during the nancial crisis of 2007/8 through the market recovery of

introduction
What dierentiates you from the competition? What is your edge? These are two of the most important questions asked of any money management rm. After all, the investment business is an odd one: there are really only a few styles of investing, and dierentiating yourself often comes down to just outperformance or, as it is commonly referred to in investment circles, active return. Active return results from being dierent from the benchmark. It comprises both alpha and beta. In most cases, beta is expected to be 1.0, with the majority of active return coming from alpha. This edge is usually attributed to the experience and stock-picking prowess of portfolio managers and analysts, or the sophistication of a quantitative model. The quandary is that without decades of historical performance data, it is very dicult to prove these assertions either true or false. More often than not, one is left with faith: faith in the people who manage a strategy and faith in the process they follow. At Epoch Investment Partners, we have experienced people, sophisticated models and a demonstrated record of outperformance; but what sets us apart from the rest of the investing world is our focus on the generation of free cash ow and the allocation of capital. While most traditional value and growth managers use accounting measures like earnings or book value to underpin their process, we believe that the true value

1 Fama, Eugene F., and Kenneth R. French, 1992, The Cross-Section of Expected Stock Returns, Journal of Finance 47, 427-465, Fama, Eugene F., and Kenneth R. French, 1996, Multifactor Explanations of Asset-Pricing Anomalies, Journal of Finance 51, 55-84, and others.

epoch perspectives free-cash-ow investing: a value strategy

2009/10. Our process of investing in companies with robust free cash ow, ecient allocation of capital, and transparent business models gives us a natural bias against owning the equities of banks and certain other nancial entities that largely drove returns in the value benchmarks during the crisis and recovery. These value stocks became value traps, ensnaring many a manager and making those who did not own them look less like value and more like growth. Interestingly, an alternative systematic risk framework2 has recently been proposed that decomposes risk more along the lines of the capital allocation framework that we use at Epoch. This production or supply-side model posits the use of market, investment, and return-on-assets factors as an attractive alternative to the classic Fama-French model. In this framework, this is akin to changing coordinate systems in risk space, our U.S. Large Cap Value portfolio as an example has consistent systematic risk exposures that very closely match our investment philosophy.

investors process is exactly analogous to the traditional value managers process. First, nd stocks that are attractive based on a valuation metric FCF yield vs. P/B or P/E then add alpha by discriminating among inexpensively priced stocks that are likely to be valuable rather than those that are merely cheap. Value benchmarks used for style analysis can more accurately be described as inexpensive benchmarks. Like the bargain bin at a discount store, they contain a grab bag of gems and garbage mixed together. Just because something is cheap doesnt mean it is a value. Many traditional value managers learned this lesson the hard way in 2008 when banks and other nancials looked enticingly cheap. Our free-cash-ow discipline, accompanied by the transparency inherent within nance data relative to accounting assumptions, provide the insight and discipline that have allowed us to avoid these companies and their equity securities.3 Our application of free-cash-ow investing focuses on our ability to follow a dollar of revenue, or an incremental dollar of new capital, through the business. What are the claims on these dollars as time passes and products are made or services rendered? In terms of understanding how a business actually works, the lens of free-cash-ow analysis is superior to the lens of the accountant with its myriad accruals and assumptions of useful asset lives.4 Our process allows us to more fully understand the drivers of the business and to invest as if we intended to purchase the entire rm. This process makes us inherently biased against opaque and arcane business models, and it leaves us with a consistent aversion to owning complex companies that are characterized by extensive accruals, o-balance sheet items, and policies that obfuscate the measurement of cash ows (e.g. nancial entities with many lines of business and immeasurable elements of leverage.) It also means we do not own companies that can blow up nancially such as what happened to many rms in 2008-2009, particularly in the nancial sector.

anatomy of a value investing process

Although there are many styles of investing, all share two characteristics. First, an investor must decide upon which systematic risk factors they wish to emphasize and second, they must earn their keep by adding alpha or returns in addition to the systematic factor bets they have assumed. The Epoch style of investing is to invest in companies that are ecient allocators of capital. We rene the process by seeking those superior allocators of capital who have large free cash ows relative to their market capitalizations. Such companies are commonly said to have a high Free-Cash-Flow Yield. The more opportunities good allocators have to allocate capital, the more valuable they are to those who invest in them. The traditional value investor generates value by his or her ability to separate good productive book assets from garbage book assets. Similarly, the free-cash-ow investor must be able to discriminate between managements that are good stewards/ allocators of capital and those that misallocate capital. Value is created by investing in the business via internal projects or acquisitions if and only if this results in a return in excess of the cost of capital. When this is not the case, value is created by returning free cash ow to shareholders. In the same way a traditional high B/M (Book Value to Market Value) investor wants to buy a dollar of book value as cheaply as possible, the free-cash-ow investor looks to buy a dollar of free cash ow as cheaply as possible. Thus, the free-cash-ow

passive vs. active style drift

At Epoch, we have a consistently applied, value-based investment process that does not rely on the accounting metrics that have come to dene many style indices. As a result, our strategy may appear to drift from time to time simply because of extraordinary dynamics within the index. For example, we consider the emergence of a massive weight in nancials as an extraordinary event in the history of value indices. This event induced what appeared to be a style drift in one of our portfolio strategies when viewed through the prism of

2 Chen, Long, Novy-Marx, Robert and Zhang, Lu, An Alternative Three-Factor Model (April 1, 2010). Available at SSRN: http://ssrn.com/abstract=1418117. 3 See our series of white papers: Financial Services Stocks: The Wheels are Coming O (August 2005), Financial Services Stocks: The Storm Clouds Gather (November 2006), The Canary in the Coal Mine: Subprime Mortgages, MortgageBacked Securities, and the US Housing Bust (April 2007), A Roller Coaster Called Credit (September 2007). http://www.eipny.com/php/white_papers.php 4 See our white paper Mixing Financial Principles with Accounting Standards A Slippery Slope (August 2005), http://www.eipny.com/pdf/AccountingAnd8082005.pdf.

epoch perspectives free-cash-ow investing: a value strategy

a classic nine-box style analysis. Consider the trailing 12-month exposures of our U.S. Large Cap Value portfolio for the period 2001-2010. Figure 1 shows that, using a traditional rolling manager-style analysis, the portfolios exposure drifted from value to growth and back towards value.
2 1.5 Current

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4

R1000V
1

R1000G

-0.6 2005

2006

2007
R1000 Value Financials

2008

2009
R1000 Growth ex. Financials

2010

2011
R1000 Growth Financials

Small - Large

0.5 0 -0.5

R1000 Value ex. Financials

R2000V
-1 -1.5 -2 -2 -1.5 -1 -0.5 0 0.5

R2000G

gure 2: epoch large cap value rolling 24-month exposures to the russell 1000 value and growth nancials and ex-nancial indices

1.5

Value - Grow th
US Value (Gross)

systematic risk and the capital budgeting decision: a new paradigm

Systematic or style risk is encapsulated nicely in a theoretical sense by the Arbitrage Pricing Theory (APT)5. At an abstract level, APT says that a limited and perhaps small number of systematic factors can explain and account for the majority of cross-sectional equity risk. The shortcoming of APT is that it doesnt provide much insight into what these common factors might be. A large body of research has been generated in an attempt to address this issue by proposing both fundamental and purely statistical factors. The Fama-French three-factor model that uses the dierence in the returns of small-cap stocks vs. large-cap ones, high bookto-market (value) vs. low book-to-market (growth) stocks, and the market return (beta) is one of the more famous and useful iterations of APT. The attraction of the Fama-French model is that the factors it uses to describe/explain equity returns Value/Growth and Small/Large operate in alignment with the investing universes traditional segmentations. This made it a very popular model and helped pave the way for the nine-stylebox model, which segments the market into large, mid, and small cap on the one hand, and value, core, and growth on the other. Allocation among boxes in the nine-box model covers an investors systematic risks (beta), while the choice of the manager within each box provides the alpha. While the Fama-French model has been satisfactory in some respects, there are many anomalies it is unable to explain. And since APT does not specify which factors explain the market, it leaves the door open for other factor models to explain returns. Examples include outcomes associated with short-term prior returns, failure probability, earnings surprises, accruals, net stock issues, and stock valuation ratios.

Russell 4 style indices. Rolling 12 month windows.

gure 1: rolling manager style 8/2001-9/2010

Source: Factset

Style drift poses a concern for many investors because it implies active drift, in which an institutional money manager switches investment styles to capture returns from whatever style happens to be working at the moment. This type of active drift obfuscates systematic risk exposures at the asset allocation/ plan sponsor level. By contrast, passive drift as evidenced in the recent performance of our U.S. Large Cap Value portfolio is much more benign and does not reect a change to the managers systematic risk exposures. To further elaborate the distinction between active and passive drift, lets take a look at Figure 2. This graph allows for a ner decomposition of style exposures by separating the Russell 1000 Value and Growth indices into nancials and non-nancials. In this gure it can be seen that our portfolios exposures have been fairly stable over time with only modest uctuations. The dierence between what we see in Figure 1 versus Figure 2 is the result of dierences in market capitalization and volatility of the sub-components. At the beginning of 2007, nancials made up approximately 40% of the market value of the Russell 1000 Value index. Our very low exposure to nancials combined with the extraordinary volatility of the nancial sector, therefore, made us appear non-value (i.e. growth) in the eyes of any historical analysis that aggregated the value and growth indices instead of separating them.

5 Ross, Stephen A., 1976, The Arbitrage Theory of Capital Asset Pricing, Journal of Economic Theory 13, 341-360.

epoch perspectives free-cash-ow investing: a value strategy

At Epoch we believe there is an equally enlightening way of looking at systematic risk: one that frames the problem in the same way rms face their own capital allocation decisions. This investment-based model was proposed by Chen, Novy-Marx, and Zhang in a 2010 working paper titled An Alternative ThreeFactor Model. 6 Their alternative workhorse model uses the returns from the market portfolio (Beta) and factors based on the returns from a portfolio of low-investment stocks minus the return of a portfolio of high-investment stocks (INV) and the returns from a portfolio of high-ROA stocks minus the return on a portfolio of low-ROA stocks (ROA). Their INV and ROA factors earn signicant average returns of 0.28% and 0.76% per month, which persist after adjusting for their exposures to the Fama-French factors. They show that their new model outperforms traditional asset pricing models in explaining anomalies such as earnings surprises, total accruals, net stock issues, and asset growth. One interpretation of their success in explaining anomalies is that their model captures more of the systematic risk in the equity market than the Fama-French model. In the elaboration of their model, Chen, Novy-Marx, and Zhang (CNZ) note that investment should predict returns because given expected cash ows, a high cost of capital implies a low net present value of new capital and therefore lower investment, whereas a low cost of capital implies a high net present value of new capital and therefore higher investment. Similarly, return on assets (ROA) predicts returns because high (low) expected ROA relative to low (high) investment implies high (low) discount rates. This is because high (low) discount rates are necessary to oset the high (low) expected ROA and induce low (high) net present values of new capital and therefore low (high) investment. If the discount rate was not high (low) enough to oset the high (low) expected ROA, rms would observe high (low) net present values of new capital and invest more (less). To demonstrate this, CNZ consider a simple model in which a rm makes investment choices over two periods in order to maximize its present value. In this simple two-period model, the rm starts with initial assets, A0, and makes a choice of how much to invest, I0. The rm then produces during the two periods and exits the market at the end of the second period by liquidating all of its assets. The terminal value of the rm is TV=A1*(1-d) where d is the rate of depreciation. The rms free cash ow in the initial period is given by FCF0=(ROA0*A0)I0-C0, where C0 represents the costs and frictions associated with deploying the investment. The rms level of assets in the second period is given by A1=I0+ A0*(1-d) and the free cash ow in that period is FCF1=ROA1*A1. To nd the market value

at the beginning of the initial period, the free cash ows and the terminal value of the rm in the second period are discounted back to the initial period at a rm specic discount rate r. Thus, the present value of the rm can be written as: PV = FCF0 + FCF1 /r + TV/r. Following Cochrane7, CNZ show that in maximizing the rms initial market value, the tradeo for the rm is that of forgoing the initial free cash ow in exchange for higher free cash ow in the second period. Solving the maximization problem with respect to A1 yields the optimality condition: r = (Expected protability +1) / (Marginal cost of investment). This says that the investment return, dened as the ratio of the marginal benet of investment in date one divided by the marginal cost of investment in date zero, should equal the rms discount rate. The implications for systematic returns across stocks are that, given an expected ROA, rms that invest more should have lower expected returns. Additionally, given a level of investment normalized by assets, rms with higher expected ROAs should have higher expected returns. The Fama-French and nine-box models approach stock returns from the point of view of an investor allocating risk capital. CNZs investment-based model approaches stock returns from the perspective of the rm itself in the form of its capital allocation problem. To elucidate this dierence, Figure 3 shows the risk factor exposures to both the Fama-French and CMZ 3-factor models for the Epoch U.S. Large Cap Value portfolio, the Russell 1000 Value ex-nancials, and the Russell 1000, core, value, and growth benchmarks. In the traditional Fama-French framework, our portfolio looks very similar to the core Russell 1000. However, when viewed through the lens of the CMZ investment based 3-factor model that more closely resembles our investment process, our portfolio looks much more like the Russell 1000 Value ex-Financials.

6 Chen, Long, Novy-Marx, Robert and Zhang, Lu, An Alternative Three-Factor Model (April 1, 2010). Available at SSRN: http://ssrn.com/abstract=1418117. 7 Cochrane, John H., 1991, Production-based asset pricing and the link between stock returns and economic uctuations, Journal of Finance 46, 572-621.

epoch perspectives free-cash-ow investing: a value strategy

risk factor exposures: 10/2002-11/2010

Epoch U.S. Large Cap Value R1000V .xFin R1000 R1000V R1000G

Coecients Intercept Mkt-RF SMB HML Adjusted R Square 0.21 0.90 -0.09 -0.01 0.93

P-value 0.09 0.00 0.11 0.86

Coecients 0.25 0.90 -0.13 0.09 0.93

P-value 0.04 0.00 0.01 0.07

Coecients -0.06 0.98 -0.09 0.00 0.99

P-value 0.10 0.00 0.00 0.91

Coecients -0.05 0.96 -0.16 0.29 0.98

P-value 0.49 0.00 0.00 0.00

Coecients -0.07 1.00 -0.01 -0.29 0.98

P-value 0.30 0.00 0.71 0.00

The p-value is the probability that the associated coecient is zero. Bold numbers represent p-values that are less than 10%.

gure 3a: fama-french 3-factor model

Epoch U.S. Large Cap Value R1000V .xFin R1000 R1000V R1000G

Coecients Intercept Mkt-RF I/A LMH ROA HML Adjusted R Square 0.12 0.95 -0.10 0.18 0.94

P-value 0.25 0.00 0.10 0.00

Coecients 0.18 0.94 -0.17 0.06 0.93

P-value 0.13 0.00 0.01 0.24

Coecients -0.08 0.96 -0.06 -0.04 0.99

P-value 0.04 0.00 0.01 0.02

Coecients -0.04 0.91 -0.10 -0.25 0.96

P-value 0.68 0.00 0.04 0.00

Coecients -0.12 1.01 -0.01 0.17 0.97

P-value 0.17 0.00 0.91 0.00

gure 3b: investment-roa 3-factor model

Figures 4 and 5 expand on this theme and show the similarity of the exposure dynamics of our U.S. Large Cap Value portfolio and the Russell 1000 Value ex-Financials through the recent nancial crisis.
1.5 1.4 1.3 1.2
0.5 0.4 0.3 0.2

Exposure

Exposure
2006 2007
Mkt.RF (EpLgVal)

1.1 1.0 0.9 0.8 0.7 0.6 0.5 2005 2008 2009 2010 2011

0.1 0.0 -0.1 -0.2 -0.3 -0.4 2005 2006 2007

SMB (EpLgVal)

2008

2009

2010

2011

Mkt.RF (R1000V.xFin)

SMB (R1000V.xFin)

gure 4a: fama-french rolling 24-month market exposure

1.0 0.8 0.6
Exposure

gure 4b: fama-french rolling 24-month size (SMB) exposure

1.5 1.4 1.3 1.2

0.2 0.0 -0.2 -0.4 -0.6 2005 2006 2007 2008 2009 2010 2011

Exposure

0.4

1.1 1.0 0.9 0.8 0.7 0.6 0.5 2005 2006 2007
Mkt.RF (EpLgVal)

2008

2009

2010

2011

HML (EpLgVal)

HML (R1000V.xFin)

Mkt.RF (R1000V.xFin)

gure 4c: fama-french rolling 24-month book/market (hml) exposure

gure 5a: chen, novy-marx and zhang rolling 24-month market exposure

epoch perspectives free-cash-ow investing: a value strategy

0.5 0.4 0.3 0.2

0.5 0.4 0.3

Exposure
2006 2007
IA.LMH (EpLgVal)

Exposure

0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 2005 2008 2009 2010 2011

0.2 0.1 0.0 -0.1 -0.2 -0.3 2005

2006

2007

2008

2009

2010

2011

IA.LMH (R1000V.xFin)

ROA.HML (EpLgVal)

ROA.HML (R1000V.xFin)

gure 5b: chen, novy-marx and zhang rolling 24-month inv exposure

gure 5c: chen, novy-marx and zhang rolling 24-month roa exposure

conclusion
Unlike most investors, we deploy investment capital in the same way companies conduct capital budgeting exercises. Because of this, there are times we may appear to drift between the traditional style designations of value and growth despite our consistent investment approach. However, it has been shown that, because of our focus on free cash ow, we have been able to ride out the market uctuations that put so many other money managers in dire straits. A recent study by BofA Merrill Lynch reinforces this point and is reected in Figure 6. Their study identied three explanatory variables that dominated equity market returns over a 25 year period (free cash ow yield, quality as reected in high ROE rms, and Beta). In examining the relative contribution of these three variables, we see (1) the dominance of cash ow over time relative to the other two variables and (2) that every speculative rally lead by high Beta stocks has been followed by a period where strategies based on Free-Cash-Flow yield ruled the day. The lesson is that true value is found by investing in rms with strong free cash ows and ecient capital allocation strategies.
Cumulative Performance in Period

Free-cash-ow investors follow the same process as value investors. Using a valuation metric (free-cash-ow yield) to narrow the investable universe down to a manageable list of candidates, value is uncovered in companies that generate signicant free cash ow and are ecient allocators of capital. This focus on the capital allocation decisions of the rm instead of traditional accounting-based nancial metrics can lead to what we term passive style drift under unusual and highly volatile market conditions but not the dreaded active style drift so properly feared by consultants. We have also tried to show that an alternative systematic risk framework, grounded in the capital budgeting decision process of the rm, better encapsulates the systematic risks faced by the free-cash-ow investor than the traditional Fama-French Returns based methodology. Sunlight is the best disinfectant said Oliver Wendell Holmes, and more sunlight is found through the prism of cash-ow analysis than traditional accounting metrics.

% 240

QUALITY

QUALITY

QUALITY

VALUE

VALUE

VALUE

VALUE

BETA

BETA

BETA

160 120 80 40 0 -40 -80

86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10

P/FCF

ROE

BETA

gure 6: after beta its cash ow

Source: BofA Merrill Lynch U.S. Quantitative Strategy, December 2011
The information contained herein reects, as of the date hereof, the views of Epoch Investment Partners, Inc. and sources believed by Epoch Investment Partners, Inc. to be reliable. No representation or warranty is made concerning the accuracy of any data compiled herein. In addition, there can be no guarantee that any projection, forecast or opinion in these materials will be realized. The views expressed herein may change at any time subsequent to the date of issue hereof. These materials are provided for informational purposes only. The reference to the security listed herein does not constitute a recommendation to buy, sell or hold such security.

epoch perspectives free-cash-ow investing: a value strategy

BETA

200