DEPARTMENT OF BUSINESS STUDIES Uppsala University
Bachelor thesis
Authors: Rickard Bergman, Philip Gunnarsson Thesis Tutor: PhD Jiri Novak
Autumn 2010
Economic Value Added
®applied on the American
Stock Market
Can the EVA® fundamental analysis increase the returns to a hedge-portfolio strategy with stocks sorted after book-to-market valuation and size?
ABSTRACT
In this paper, the popular fundamental analysis model Economic Value Added is tested for any ability to generate returns above that explained by book-to-market effects on American large cap stocks. A zero net-investment hedge portfolio-test was undertaken where the Economic Value Added®
fundamental analysis was applied on a sample of large cap stocks, sorted into quintiles after book to market valuation. The portfolio investing in the extreme quintiles gained positive returns between the years 1999 – 2010 equal to an average yearly total return of 7,32 %. During the test-period, the benchmark portfolio constituent of stocks sorted in the same way but without the Economic Value Added® analysis only managed to score returns equaling 2,3 %, adding evidence in favor of the Economic Value Added® analysis. The Economic Value Added also showed a better risk-profile than the benchmark portfolio, measured as the Modigliani Risk-Adjusted Performance over the entire period, further acknowledging the abnormal returns. However, the Economic Value Added® sample portfolios where unevenly distributed regarding number of stocks, foremost in the short-sold part for some years, mitigating the test as strong evidence in favor of the Economic Value Added® analysis.
An independent samples t-test also did not reject the null hypothesis. Despite the mixed results of the test, the strength in the specification of sample and choice of method leads us to conclude that that the Economic Value Added® seems like a moderately effective tool for identifying mispriced stocks.
Table of Contents
1. INTRODUCTION ...3
1.1 Background ...3
1.2 Research question ...5
1.3 Purpose ...5
2. THEORY ...5
2.1 Economic Value Added - EVA® ...6
2.2 Capital Asset Pricing Model and Weighted Cost of Capital ...7
2.3 The efficient market hypothesis – EMH ...9
2.4 The size and book-to-market effect ... 10
2.5 The summing and comparison of returns... 11
2.6 Performance metrics and portfolio evaluation ... 13
3. METHOD ... 15
3.1 Choice of method ... 15
3.2 Method specifications ... 15
3.2 Dataset ... 17
3.3 Basic assumptions ... 17
3.4 Evaluating the results ... 18
3.5 Method critique ... 19
4. RESULTS ... 20
5. ANALYSIS ... 25
6. DISCUSSION ... 26
6.1 Conclusion ... 28
6.2 Suggestions for further research ... 28
7. SOURCES ... 29
APPENDIX 1 - Terminology ... 32
APPENDIX 2 - Portfolios ... 33
3
1. INTRODUCTION
1.1 Background
The activity of fundamental analysis has drawn a lot of attention from the investor- and research community. Fundamental analysis, in its essence, is the activity of searching for under- or overpriced stocks with the potential for superior performance on the stock market.
This activity does not always follow the same pattern, but often includes evaluating a company’s financial statement, management efficiency or other factors such as the macroeconomic environment when searching for these stocks. A field of interest where some approaches has become more popular than others, but where ambiguity still is present amongst investors regarding which techniques are useful in the search for good stocks. In academia, the connection of fundamental analysis with stock returns was first presented by researchers (Ou and Penman, 1989). They combined several different items from the financial statement into a single measure of the company´s performance, creating a signal for buying or selling stocks. Through this method they were able to outperform the market index return, showing the use of fundamental analysis of stocks. Their work was followed by (Lev and Thiagarajan, 1993) who performed analysis of companies financial statements chosen through evaluating both accepted theory and professional working opinion to arrive at the signals.
They also applied contextual procedures to condition the signals chosen, gaining returns in excess of the market. More recent work of (Abarbanell and Bushee, 1998), (Piotroski, 2000) and (Mohanram, 2005) has also further refined the concept of fundamental analysis, giving even stronger evidence in favor of the activity through hedge-portfolio tests over long time- periods under different market conditions. Also showing a high degree of abnormal returns i.e. returns above the expected.
However, many of these analysis´s mentioned above is time consuming and rather intricate to perform for the ordinary investor, supporting the idea of finding and exploring more basic methods. Alternatives also supported by research do exist, leaning against standard economic concepts combining accounting and market data that are more easily available. Economic Value Added® hereafter named “EVA” (Abate, Grant & Stewart, 2004) is one of these methods that can be used as a fundamental analysis for picking stocks. The EVA is a straightforward fundamental analysis which gained a lot of attention under the early eighties.
The EVA is marketed by the Stern Stewart Co, which also is the registered trademark owner
4 of EVA. The method is presented as an alternative to standard accounting based metrics for performance analysis such as ROE return on equity, EPS earnings per share or ROC return on capital. The stated superiority of the EVA compared to these measures is also often stressed by their users and developers: “EVA style of investing (…) emphasizes the fundamental ability of a company to create wealth through the generation of economic earnings rather than accounting earnings” (Abate et al. 2004 p. 61). The concept of EVA is that any residual left after deducting the cost of the company´s total debt and equity capital from a certain profit measure is the EVA, suggesting that the company is creating additional shareholder wealth. Simply put, this approach compensates the investor for her opportunity cost through including the cost of equity capital to the cost of debt capital in the calculation.
This makes sense because it accounts for the return she could have earned elsewhere if she invested in a company with similar risk (Sharma, 2010). If the company still after deducting this cost is creating wealth, it is according to the model suitable as a candidate for investment, potentially leading to superior returns.
In disfavor for the EVA method, (Sharma, 2010) shows that previous research about EVA as an investment strategy is to some degree inconclusive. Other research though has shown that the use of residual income type of analysis, a method similar to EVA, can earn abnormal returns unexplained by common risk factors (Ali, Hwang, Trombley, 2003). This paired with further evidence from (Worthington and West, 2004) that Economic Value Added® is superior to the residual income method, leads us to assume that the EVA analysis makes sense as an investment tool in line with the thought of a simple metric for finding over- or undervalued stocks.
When conducting research on investment strategies like EVA, the methods for measuring performance have been refined throughout the course and history of the fundamental analysis field of research. It has been found that large abnormal returns often are attributed to common risk factors leading to the mitigation of the explanatory value of some analysis´s (Barber &
Lyon, 1997). (Fama & French, 1992) were some of the first to pinpoint the size- and book-to- market effects as two of the most important factors for explaining returns above the expected by the common models. The size effect is concerned with the observed effect that stocks with very small market capitalization earns returns above the expected. The book-to-market effect is the phenomenon that companies with high book value of equity in comparison to the market value of equity observes abnormal returns. This obviously could be a problem when evaluating a fundamental analysis method. However, it is when researchers have utilized the
5 fundamental analysis on stocks subject to one or more of these risk factors that the highest returns have been achieved (Abarbanell & Bushee, 1998) and (Piotroski, 2000). Telling us that if a method of selecting stocks from this group of the most “risky” securities is gaining returns above that of the group as a whole, it can be said to have been “controlled for risk”
and thus conveying some information of value to the investor. Following this thought we therefore look at the possibility that signs of under- or over performance amongst stocks with extreme valuations, ex ante, could be obtained by sorting stocks subject to the book-to-market effect through the Economic Value Added fundamental analysis model. Leading us to the research question of the paper:
1.2 Research question
Can the EVA® fundamental analysis increase the returns to a hedge-portfolio strategy with stocks sorted after book-to-market valuation and size?
1.3 Purpose
The purpose of this paper is to, through a quantitative study; investigate if the widespread EVA fundamental analysis can serve as a heuristic to increase the returns to an investment strategy utilizing size and book-to-market sorted stocks to create hedge-portfolios. The previous research about the EVA analysis as an investment strategy is inconclusive; we therefore intend to raise evidence on its potential abilities or disabilities through this portfolio test.
2. THEORY
The first step in this paper is providing the theoretical background which we build our method around. In the first section the concept known as EVA is more thoroughly introduced followed by additional theory sections covering key theoretical concepts affecting the EVA and method as a whole.
6
2.1 Economic Value Added - EVA
®In this section the concept of Economic Value Added is introduced with proper explaining of the different components briefly mentioned in the background.
EVA formula:
NOPAT
–
WACC * KWhere:
NOPAT is the net operating profit after tax
WACC is the weighted average cost of capital
K is capital employed, equity capital + debt capital - cash
Economic Value Added, or EVA®, was spread commercially in the early 1980s by Joel M.
Stern and G. Bennett Stewart III who also founded the Stern Stewart & Co and registered the trademark EVA. The method quickly gained attention within corporate finance and management accounting. Significant with the Economic Value Added method is that the cost of all capital within the firm is compared with a profit measurement, NOPAT standing for Net operating Profit after Tax. EVA has because of its simplicity of the calculation and logical approach been very popular when developing incentive programs within firms, but it has also been used when evaluating potential investment opportunities as well as implementing or evaluating whole investing strategies. It should also be mentioned that the concept of EVA is basically the same approach as “Economic Profit” and similar to other residual income based models, however, for consistency, the EVA label will be used throughout the paper. (Abate et al, 2004).
EVA includes calculating the cost of capital which, can be explained as the required rate of return demanded by investors. It reflects the return the investors could have received if they had invested their money elsewhere. The cost of capital is the opportunity cost of the investor, and is what a firm must, theoretically, at least generate to keep its investors. If a particular company’s return is lower than the cost of equity, the company is according to the model, destroying shareholder wealth. “Only by earning more than the cost of equity can a company create wealth. The cost of equity is a critical cutoff rate, an invisible but profound dividing
7 line between superior and inferior corporate performance” (Stewart, 2003). However, as can be understood by looking at the formula in the beginning of the section, the EVA is a positive or negative dollar amount. Therefore it can be hard to evaluate and compare with other companies when just looking at its face value. The proper method for dealing with this issue is therefore to utilize the EVA-spread.
EVA-spread formula
Where:
is the return on capital or = NOPAT / K is the weighted average cost of capital
The EVA can also be expressed as a percentage spread, which is called the EVA-spread, where the residual is a positive, neutral or negative expression in percentage. Since WACC includes the cost of all capital within the firm, a ROC higher than the WACC indicates that the firm in question is value creating and the opposite that the company is value destroying.
The key feature of the spread is that since it is scaled as a percentage value instead of the simple face value, it is possible to compare in-between companies. (Abate et al, 2004). This feature makes it usable across the full sample and is therefore selected as our key component in deciding which stocks to buy or sell.
2.2 Capital Asset Pricing Model and Weighted Cost of Capital
Since the EVA analysis contains the task of defining the cost of equity capital, the background to the original method is introduced in this section, covering the CAPM and the WACC which then is followed by further research leading us towards the final method.
The Nobel Prize winner Henry Markowitz provided the foundation of our modern theories in his article portfolio selection (Markowitz, 1952). The concept of the risk-averse investor which utility function is negative towards mean expected return variance was introduced, leading to a theory of investor preferences, taking the importance of diversification into account when constructing portfolios constituent of individual securities. This work was
8 followed, about a decade later, by the Capital asset pricing model developed by (Sharpe, 1964, Lintner, 1965, Mossin, 1966) building on the work of Markowitz. One of the major ideas of CAPM is that the residual risk of a diversified portfolio is just non-diversifiable or systematic. The CAPM model is utilized to determine the expected return of a security or portfolio through the β beta, the covariance with the market return, the risk free rate and the expected market return. The β in CAPM captures the earlier mentioned systematic component, or the risk that is non-diversifiable. The Beta is estimated through a regression, a method to describe the function between to variables, commonly by measuring the historical return of the security in comparison to the market. The market beta is according to the theory 1, which means that a security with a Beta higher than 1 has a higher expected return since a rational investor wants to be compensated when facing a higher systematic risk. The expected return is commonly used as the cost of capital which is included in the calculation of the WACC which forms a key part of the Economic Value Added formula. (Reilly, Frank, Brown, & Keith, 2006).
CAPM-formula
is an individual security
the expected return of the security
is the risk-free rate
(beta) is the sensitivity of the expected return of the asset,
dependent on the expected market return, or:
is the expected market return equity risk premium
Worth mentioning is that the expected market return actually, according to the model, consist of a portfolio where all investable risky-assets in the economy are held, like stocks, art or real estate. In reality, this portfolio is not possible to invest in or observe through any known
9 method and common broad stock indices often serve as substitutes, per example the S&P 500, Eurostoxx 50 or OMXSPI Benchmark. (Reilly et al, 2006)
As a consequence of CAPM the weighted average cost of capital can be calculated as follows below, which later on will have consequences for the calculation of the EVA-spread.
Weighted Average Cost of Capital-formula
Where:
is the expected return of the security or the cost of equity capital
E is the market value of equity
D is the market value of debt
is the risk-free rate is the tax shield
The Weighted Average Cost of Capital, as can be derived from the name, weights the amount of equity and debt capital, expensing each part according to the cost of the financing source, making it useful as a tool to establish the total cost of capital. (Reilly et. al, 2006)
2.3 The efficient market hypothesis – EMH
Anyone who intends to make a test or research about issues regarding the behavior on the financial markets will have to relate the findings to the efficient market hypothesis. Although a body of critique has arisen under its academic existence, the basic assumption amongst many practitioners, researchers and economists is that of an information efficient market and it is therefore presented in this section.
The key implication of the efficient market hypothesis model “EMH” is that of information efficiency, it originally states three different versions of information efficiency, weak, semi- strong and strong. The difference in between them lies foremost in what amount of
10 information that is reflected in the price of a security according to each of the suggested states. (Haugen, 2001 p. 575)
The weak form represents that all available public information of value is reflected in the current price at any moment. Effectively leading to the assumption that no trading strategy that is based only on historical price information can generate excess return; making technical analysis useless for trading purposes. (Haugen, 2001 p. 575)
The semi-strong version states that prices reflect all publicly available information and that price adjusts instantly to new information, making fundamental analysis useless as a method according to this version. (Haugen, 2001 p. 575)
The strong version states that all information, public as well as insider- and hidden information, is reflected in the price of the security. Evidence for and against the weak and the semi-strong is mixed but evidence for the strong version is scarce. (Haugen, 2001 p. 575) These suggestions will have influence over the analysis and discussion, if abnormal returns can be realized through applying an analysis, one has to consider the above stated different versions of the EMH. In this case a positive outcome would stipulate that the market probably only is weakly efficient rejecting the other options.
2.4 The size and book-to-market effect
This section covers the underlying theory for choosing to test the EVA analysis on large cap stocks with extreme valuations. Since research shows that some factors are amongst the most important for explaining return above that of CAPM expected return, we introduce this theory to be able to make assumptions about the equity cost of capital and to test EVA on firms subject to one of these factors to see if it can reveal potentially mispriced stocks.
Following the introduction of the CAPM model, the research examining its predictive abilities has found mixed results (McGoun, 1993). When the CAPM-beta component fails to capture the variation in stock returns it is defined as an anomaly. Although the scope of this paper does not include covering all of the different anomalies that has been found, two of the most important is included to get more significance in any results.
In the early eighties, (Banz, 1981) first documented the size effect. Banz found that small stocks often yielded high returns exceeding the returns predicted by conventional CAPM beta.
11 Banz also found that the size effect was not linear in its nature and the returns where strongest amongst NYSE stocks with small market capitalization, while stock returns among average and large capitalizations stocks showed little difference. Banz concludes that this anomaly has been present for at least 40 years, at the time, and that it remains unknown if size itself is a risk-factor or if it serves as a proxy for some other risk.
In (Fama & French, 1992) the work of Banz is more thoroughly tested and the size effect is found to be one of two major explanations for common variations in stock return. The other concept is that of the relative market valuation of a stock, the book-to-market effect. Fama and French find that stocks with high book-to-market have positive returns unexplained by CAPM Beta. They also find, just as Banz, that the size effect is clustered around smaller market capitalization stocks. This leads Fama and French to conclude that the combination of size and book-to-market ratio has the best performance in explaining the cross sectional variation in stock returns, and when these two factors are taken into account CAPM beta becomes insignificant.
Building additional strength to this above stated body of knowledge, (Davis, 1994) also test for the book-to-market anomaly and find similar evidence on a previously untested data sample, free from survivorship bias from 1940-1962 before the more commonly used Compustat-file, utilized in Fama & French (1992) and many other articles. (Barber and Lyon, 1997) also provides further proof of abnormal returns and extended mispricing unexplained by earnings surprises effects in a hedge portfolio strategy based on book-to-market. This was concluded since the returns to the strategy were not clustered mainly around earnings announcements. However, they attribute that a significant proportion of the abnormal returns are due to the size effect described earlier, adding to the notion that the combination of the two is the major explanatory variables for variations in returns.
2.5 The summing and comparison of returns
The research surrounding the methods in tests aimed at detecting abnormal returns has found some major implications to take into account to improve accuracy of the tests, and is therefore addressed in this section for later improvements of the final method.
When evaluating an analysis method the concept of “Abnormal return” is often utilized as a part of the methodology. Abnormal return is the difference between the expected return of a
12 market model as CAPM and the actual return of the tested stocks. It can also be measured as the ex-post return of a stock minus the ex post return of the market (Ross, Westerfield &
Jaffe, 2005 p. 360). If the difference is substantial, some information of value in the analysis is present that can be considered to have been foregone by the model used to estimate the expected return considering the previous section about EMH. However, the specification of the way one measures the results will have impact on what conclusions that can be drawn from the findings.
When measuring abnormal returns a common approach is measuring the abnormal return over several periods such as a days, weeks, months or years. Thereafter the returns are summed horizontally, a method usually known as CAR or cumulative abnormal return. (Ross et al., 2005 p. 360) However, the combined use of these two approaches might yield miss-specified results according to (Barber & Lyon, 1997)
When dealing with the issue of measuring abnormal return, (Barber and Lyon, 1997) advocates two adjustments to the common methods of measuring abnormal returns that increase the accuracy of the tests. The first adjustment stems from the discussion about the choice between two common approaches of summing the portfolio returns. The two approaches are the buy and hold abnormal return hence referred to as “BHAR” and cumulative abnormal return mentioned earlier as “CAR”. The BHAR-method includes calculating the compounded yields of the chosen periods return awarded through the tested strategy, measured against the compounded returns of the benchmark portfolio or index. The second approach, CAR, is the same as horizontally summing the returns for the chosen period for the tested portfolio and the benchmark portfolio or index. However (Barber and Lyon, 1997) shows that the results begin to diverge over time and the accuracy of the BHAR is greater when measuring over longer time-periods. The major conclusion is that when measuring the long run abnormal return one should use the BHAR when measuring over extended time-periods since it is more accurate than CAR.
The other suggestion by (Barber & Lyon, 1997) is that, when measuring long run abnormal return as the ex post return of portfolio minus the index return, one should instead use carefully constructed portfolios as benchmarks because indices are due to their calculation, subject to three different biases that are discussed more thoroughly in the article. The benchmark suggested uses control firms sorted after size and book-to-market ratios similar to those in the test-portfolio and measure the buy-and-hold difference between the test portfolio
13 and the benchmark portfolio. This method increases the accuracy of the test since it is more likely that a carefully constructed benchmark portfolio will lead to an unexaggerated more fair view of the abnormal return.
This method is therefore utilized through the choice of a benchmark portfolio approach instead of standard index-comparison.
2.6 Performance metrics and portfolio evaluation
When evaluating the performance of two or more sets of portfolios, the conclusions drawn from abnormal return and statistic methods can be reinforced through including common risk measures. These adds important value through adjusting the returns to the risk taken, making it more comparable and is therefore introduced in this section
The discussion surrounding abnormal return and the summing of abnormal returns earlier in the theory chapter have been presented along with its implications for our research. Still, any result cannot be evaluated solely on the grounds of a return in excess of benchmark performance. If an analysis or heuristic conveys valuable investor information it should also show better “performance metrics” conveying the notion of lesser riskiness in relation to its benchmark.
The Sharpe-Ratio is a common approach to evaluating portfolio returns, it scales the average excess returns with the standard deviation of the excess return, thus measuring the compensation for risk that the investor receives, where the higher the ratio the better the compensation. (Sharpe, 1966).
Sharpe ratio-formula
Where:
is the average of excess returns* over a period is the standard deviation of those excess returns
* The excess return is return for the period minus the risk free rate
14 The Sharpe-Ratio of the portfolio can then be compared with the Sharpe Ratio of the benchmark portfolio. However, the Sharpe-Ratio is rather one-dimensional and suffers from some known drawbacks when the sample is not normally distributed. Simply put, samples that diverge from the normal distribution may have a high standard deviation and suffer from a very low Sharpe-Ratio making it hard to interpret. (Modigliani & Modigliani, 1997)
(Modigliani & Modigliani, 1997) suggest a solution to the problem of the comparison related issues of the Sharpe-ratio, through the use of their modified version of the Sharpe-Ratio, the Modigliani Risk-Adjusted Performance or M2.
Modigliani Risk-Adjusted Performance-formula
Where:
is the average excess return over a defined period
is the standard deviation of the excess returns is the average risk free rate for the same period is the standard deviation of the benchmark portfolio excess returns
The M2 measures the test portfolios excess return standard deviation in relation to the benchmark excess return standard deviation, thus, making the M2 dependent on the relative riskiness of the test-portfolio related to the benchmark. For example, if the test-portfolio shows the same excess returns but only half the standard deviation, the M2 expression will be twice as big as the benchmark since it only induced half the risk to receive those returns.
(Modigliani & Modigliani, 1997). This expression is also in line with the thought that a rational investor has a negative utility of expected return variation (Markovitz, 1952).
15
3. METHOD
3.1 Choice of method
To be able to fully capture any effectiveness in the EVA fundamental analysis, we have chosen a quantitative method utilizing publicly available secondary data to form the portfolios.
3.2 Method specifications
To take into account the anomaly of the book-to-market and reduce the size effect, the full unsorted sample is at the starting year sorted after market value, reducing the concern for size effects through only including stocks with a large market capitalization. A large market capitalization is defined as the inflation adjusted current definition of a large cap stocks (Yahoo, 2010) which is a market capitalization greater than or equal to $5 billion. The deflating is done back to the starting date for the test, to capture the past constituent list of large cap stocks in order to avoid survivorship bias. Thereafter the sample is sorted after book-to-market ratio dividing the stocks into quintiles from high to low value accordingly.
The initial list of stocks formed through market value is used throughout the test; however, for the book-to market ratio this procedure is repeated for each subsequent year forming new portfolios since valuations change from year to year.
Funds are then invested in the middle of March each year, to make sure the financial statement information has been available for a while, in the highest and lowest quintiles according to a hedge-portfolio method similar to the one used by (Fama and French, 2008). In this hedge-portfolio the net-investment is equal to zero through buying the stocks with high book to market (low valuation) and short selling an equal amount worth of stocks with low book-to-market (high valuation). Any returns, positive or negative, are realized in the end of the year and the portfolios are rebalanced each year to maintain equal weights. This portfolio will serve as the benchmark portfolio to our EVA-portfolio.
The EVA-portfolio is equal to the benchmark portfolio except for one major difference. After sorting the sample into book-to-market quintiles, the stocks in each of the extreme quintiles are also sorted through the EVA-spread value from high to low. The stocks with the highest (lowest) book-to-market ratio are sorted after positive (negative) EVA spread. The high (low)
16 book-to-market stocks with positive (negative) EVA spread are bought (short sold) creating the desired Economic Value Added portfolio to be tested, this procedure is then repeated for each year just as the benchmark portfolio.
We believe that this hedge-portfolio approach singles out any desired difference better than a simple long-only test of the Economic Value Added. This since the impact of any medium- to long-term market behavior is more likely to be equaled out, and any valuable information conveyed through the Economic Value Added analysis is realized as a residual of long and short returns.
The concept of abnormal returns defined earlier in theory section is utilized in assessing any positive outcome along with a standard comparison against the risk free rate. Following the recommendations of (Barber and Lyon, 1997), the abnormal return is defined as the difference between the buy and hold return of the equally weighted quintile portfolios of large cap stocks sorted after book-to-market valuation with EVA analysis applied, and the buy and hold return of the equally weighted quintile portfolios of large cap stocks sorted through book-to-market without the EVA analysis.
When forming the EVA-portfolios the calculation of the EVA-spread value for each individual stock is performed. This procedure is done as described in the theory chapter.
However, one major difference to the CAPM formula leading to the calculation of the WACC is present and important for the results.
When calculating the equity cost of capital part of WACC through the CAPM, one major implication from the theory chapter will have impact on our test. (Fama and French, 1992) finding that when controlling for the common risk-factors size and book to market, CAPM beta becomes insignificant This leads to some adjustments to our calculations. When taking these findings into consideration it leads to us assuming that beta coefficient is insignificant and equal to one for the sampled stocks, meaning that for this sample the expected return is equal to the equity risk premium plus the risk free rate. We therefore in other words, harmonize the cost of equity capital in the WACC calculation for our sample. This can be interpreted in the way that; since our sample, as a consequence of the sorting, are collected from a rather homogenous population; the systematic risk is more likely to bear the same characteristics, effectively making the need to compensate the investor for systematic risk through beta less pronounced or unnecessary.
17 Furthermore, when choosing which figure that best represents the equity risk premium Damodaran (2010) recommends that tests aimed towards detecting abnormal returns under the assumption of a market that can be mispriced over an extended period of time, one should preferably use the historical average return of the stock market as the equity risk premium.
And finally, all methods are selected so that the information utilized in the test also had to be available ex-ante to the investor to make the investment decision, otherwise hindsight biases could distort the test-results.
3.2 Dataset
The collected sample for the test is stocks listed on the New York Stock Exchange and Nasdaq Stock Market through the years 1998-2010. The sample is collected from the Thompson Reuters Datastream database and the equity risk premium needed for the calculation of the WACC is collected from (Damodaran, 2010) where we have chosen the 80- year historical geometric average of the American stock market return less the risk free rate as the equity risk premium. Out of this sample any financial stocks and stocks missing either market- or accounting-data at the starting year will be removed. We only include stocks with a market cap larger than three point seven billion dollars, or as defined earlier, the 1998 inflation adjusted value of the 5 billion dollar worth of market capital. The calculation is made through the online inflation calculator (Bureau of Labor Statistics, 2010)
3.3 Basic assumptions
When performing these kinds of tests, some basic assumptions are necessary to be able to draw realistic conclusion about the outcome. Our main assumptions regarding the investment portfolio is:
1. No transaction costs are incurred
2. All stocks are available to loan for short selling purposes without any lender interest being charged
3. Potential outliers distorting the test is limited through data sample winsorising
Regarding these assumptions, some have more impact than others, for instance. The absence of transaction cost is not a realistic assumption since all transactions on a financial market
18 incur costs. On the other hand, when comparing performance of the EVA versus the benchmark portfolio, if one induces transaction costs on the two, one can assume that since the invested capital is equal, it would have a minor effect when interpreting any difference.
The assumption that stocks are available for short selling is a realistic assumption since stocks with large market capitalization often are liquid and short selling opportunity are available.
The final assumptions can have small or big impact on the results depending on the distribution of the sample. Winsorising (Wilcox & Keselman, 2003) is a common statistical procedure limiting the possibility for outliers to distort the sample, through transforming outliers to a percentile value instead of just removing them through ordinary trimming. We will on our test conduct a 95 % winsorising, which will set the bottom 2,5 % of the returns to the 2,5th percentile and the top 2,5 % returns to the 97,5th percentile.
3.4 Evaluating the results
When evaluating the results of the portfolio test, we start with examining if there are any realized abnormal returns that support the Economic Value Added analysis method. We also follow the recommendations of (Barber & Lyon, 1997) for the construction of the benchmark in the method and therefore should be able to draw better conclusions regarding the effectiveness of the analysis. As mentioned in the performance metric and portfolio evaluation section we also add the M2 for added value in the analysis. Lastly, we will perform a student’s independent samples t-test, where we test the null hypothesis that there is no difference in the mean return between the two populations. The two samples will include the individual returns for all of the stocks included in the portfolios for the full test period.
H0:μ1 = μ2
where:
H0 = the null hypothesis
μ1 = the mean of population 1, (the EVA portfolio) and μ2 = the mean of population 2. (the Benchmark portfolio)
Stock market returns however do not always follow this distribution, and samples may be subject to skewness or kurtosis, hence, being vulnerable from this drawdown. Without further elaborating on the issue, simple descriptive statistics solves the problem through performing
19 the Shapiro-Wilk test. The Shapiro Wilk test, tests the null-hypothesis that the sample came from a normally distributed population. If the null hypothesis of the Shapiro Wilk test is rejected the value of the independent samples t-test is mitigated, but not completely useless since the student´s t-test is similar but not the same as the normal distribution. (Shapiro &
Wilk, 1966).
3.5 Method critique
The chosen methods for sampling, testing and evaluating is not without inherent issues which need addressing, this to be able to draw conclusions about the results from the test with more complete understanding.
When sampling the stocks, the completeness of the chosen database, in this case, Thompson Reuters Datastream, will set limitations to the sample. Any stock that fall out of the sample because data, accounting or market generated, are missing could skew the sample.
Furthermore, the market value cutoff line limits the number of observations. And since, the full sample also is divided into quintiles after book-to-market; the number of observations in each sub-sample will be even smaller, potentially limiting the possibility to draw conclusions when narrowing it down further through the investment criteria.
The chosen method to sample a fixed list based on the market values of 1998 do have the advantage of not inducing hindsight- or survivorship bias except for any database related ones. However, because of this fixed list of stocks, we do not include or exclude stocks that might have increased or decreased in market value for subsequent years. Thus, the test potentially might forego long or short investment opportunities.
Regarding the EVA analysis, the formula is rather straightforward where the residual amount, the spread, is utilized as the only signal for the investment strategy. This leads to that in cases where companies have a very minor positive (negative) spread, small differences in the estimation of the cost of capital will have a large impact on whether one should invest or not.
The estimation of the cost of capital is made through the common WACC, this as such is a rather intuitive method since it weights the sources of financing. The potential problem however, lies with the assumption in the CAPM estimation of the equity cost of capital, where we have disregarded CAPM beta as a risk factor through our sample sorting. Although we
20 have strong arguments that this technique is valid according to previous research, the test result will, to a great degree, be dependent on this assumption.
Also, the estimation of the market risk premium is in this paper made through the use of the historical return of the American Stock Market. Here, we do have support in research, and the technique of harmonizing the cost of capital also makes some logical sense, since large companies make out the backbone of the stock market, the difference expected in security return should be minor. Still, other techniques do exist with valid arguments for their unique approach, for example, one could use the implied market risk-premium (Damodaran, 2010).
The key point is that choices has to be, and has been made, that will impact the test results.
Regarding the assumptions, brief arguments have been given in the description of each assumption. This though, does not mean that they not will have importance for the outcome.
The assumption of no transaction costs and short selling availability probably are not the key issue, since it is a fairly common practice in similar testing situations. The winsorising however do have the potential of distorting the tests real-world connection since it sometimes do exist extreme data points that now is foregone. A test without this limitation could yield very different figures, especially if one or more short positions yield extreme negative returns.
It is though important to recognize the objective of the test of evaluating the EVA analysis and if it is useful as a method of picking stock. The bulk of the observations are still expected to be in middle of the distribution and it is the nature of the majority of the observations that we try to capture through the means. This is a choice of harmonizing the potential negative returns for long and short positions, and the following result must be interpreted through that scope.
4. RESULTS
Following the method described earlier in the paper, the list of stocks needed to perform the test was collected through the Thompson Reuters Datastream. Out of the original sample containing all stocks that could be retrieved through the database, we removed the stocks that were missing either accounting or market generated data for the base year 1998. Stocks that had less than 3.75 billion worth of market value on the 20th March 1998 were also removed, resulting in a list of 210 stocks. The remaining sample was then for each of the years sorted from high to low book-to-market i.e. the book value of equity divided by the market value of equity. Following this step the sorted list then was divided into quintiles resulting in 42 stocks
21 per quintile. The quintile with the highest book-to-market and the quintile with the lowest book-to-market constitute the benchmark portfolio for each year.
Each year’s list of stocks in the highest and lowest quintile then had their EVA-spread values calculated utilizing the method described earlier with the expected return of equity being equal to the equity risk premium plus the risk free rate when calculating the WACC. After calculating the EVA spread for each year and quintile, we sort from high to low EVA-spread within the quintiles. Then we form equally weighted portfolios according to the rule that all high (low) book-to-market stocks with positive (negative) EVA-spread are included in the portfolio with equal weight in each stock within the portfolio. This procedure resulted in a benchmark portfolio including 82 stocks per year and book-to-market and EVA-spread sorted portfolios with a different number of stocks for each year as described in table 1. Since we calculate the EVA on the previous year’s data we invested in the following year meaning that the 20th of March 1999 is the starting date of the portfolio and the portfolios was terminated the 20th March 2010.
Table 1 - Number of stocks in EVA portfolio total long short
1999 – 2000 32 29 3
2000 – 2001 30 29 1
2001 – 2002 31 26 5
2002 – 2003 37 24 13
2003 – 2004 36 27 9
2004 – 2005 28 23 5
2005 – 2006 28 27 1
2006 – 2007 32 28 4
2007 – 2008 35 29 6
2008 – 2009 33 29 4
2009 – 2010 36 27 9
22 After formation, each portfolio with high (low) book-to-market and positive (negative) was bought (sold) in the beginning of each year and the positions were closed in the end of the year realizing the results. The remaining funds were then invested the following year according to the buy-and-hold method to take compounding into account. The Benchmark portfolio were also bought and sold in the same manner. This procedure resulted in the following annual- and compounded returns described in table 2 - 4 below.
Table 2 – Summary of EVA- and Benchmark portfolio
EVA Portfolio Benchmark (B/M)
Portfolio
Year Return Year Return
1999 6,07 % 1999 -9,32 %
2000 24,23 % 2000 6,94 %
2001 31,77 % 2001 12,23 %
2002 9,09 % 2002 2,06 %
2003 -10,67 % 2003 5,48 %
2004 6,31 % 2004 3,47 %
2005 25,73 % 2005 12,24 %
2006 -19,09 % 2006 -2,38 %
2007 18,83 % 2007 2,73 %
2008 1,71 % 2008 -6,20 %
2009 -13,43 % 2009 -1,90 %
Total Return
(buy and hold) 91,47 %
Total Return
(buy and hold) 25,55 %
23
Table 3 – Annual return
Table 4 - Buy and hold return
The initial results from the portfolio tests indicate a positive outlook in favor of the EVA strategy; however before any analysis can take place, risk-metrics and statistics calculations
-30%
-20%
-10%
0%
10%
20%
30%
40%
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Annual Return
EVA Portfolio Benchmark Portfolio (B/M)
-20%
0%
20%
40%
60%
80%
100%
120%
140%
99 00 01 02 03 04 05 06 07 08 09 10
Buy and Hold Return
EVA Portfolio
Risk Free Rate
Benchmark Portfolio (B/M)
24 are conducted. The associated risk metrics specified in the method section and the results from the t-test are presented in table 5-7.
Table 5 - Risk-measures
Sharpe ratio EVA portfolio 0,176
Sharpe ratio Benchmark portfolio -0,324
EVA portfolio vs. Benchmark, M2 5,62%
Table 6 – Shapiro Wilk test
Group N Mean Std. Deviation Std. Error Mean Data 1,00000 358 ,0647827 ,41746364 ,02206364
2,00000 924 ,0185199 ,39901281 ,01312656
Group
Kolmogorov-Smirnova Shapiro-Wilk
Statistic Df Sig. Statistic df Sig.
Data 1,00000 ,078 358 ,000 ,968 358 ,000
2,00000 ,045 924 ,000 ,978 924 ,000
25
Table 7 – Independent samples t-test
5. ANALYSIS
Following the results of the test, some ambiguity is present although the returns at first sight seem quite positive for the EVA analysis. If we only consider the returns, the EVA portfolio gained a buy-and-hold return of 91,47 % while the benchmark portfolio scored a lot worse with a buy-and-hold return of 25,55 % over the same period. These figures indicate that the EVA analysis do contain some useful information potentially foregone by the investor community with a BHAR of 65,92 % over the period.
If we take a closer look at the definition of abnormal return, which is the difference between the expected return and the actual return, we should expect the difference to equal zero if no information of value was conveyed through the EVA analysis under the assumption of a weak efficient market. Since all publicly available information such as accounting data should be represented in the stock price according to weak EMH, the abnormal returns add a positive piece of evidence in favor of the EVA analysis.
Along with the abnormal return the Shapiro-Wilk test shows that the samples probably do not come from normally distributed population and the M2 risk measure is therefore applicable.
Levene's Test for Equality of
Variances t-test for Equality of Means
95% Confidence Interval of the Difference
F Sig. t Df
Sig.
(2- tailed)
Mean Difference
Std. Error
Difference Lower Upper Data Equal
variances assumed
0,632 0,427 1,838 1280 0,066 0,04626287 0,02516573 -0,00310775 0,09563348
Equal variances not assumed
1,802 624,202 0,072 0,04626287 0,02567315 -0,00415334 0,09667907
26 The risk measure with a M2 of 5,62 % also supports the thought of abnormal return in favor of the EVA analysis. It is clear through the M2 that the investor holding the EVA sorted portfolio had been compensated for the slightly higher standard deviation through the higher returns.
The M2 thus shows that the EVA portfolios has a better risk adjusted return where the reward is a lot greater for the strategy, when measuring against the benchmark, adding a second piece of evidence to the table.
The t-test does not reject the null-hypothesis adding a negative piece against the two positive signs from the BHAR and M2. However, the different statistical and theoretical concepts points towards different directions, increasing ambiguity on how to interpret the intact null- hypothesis. The Shapiro-Wilk test for instance shows that methods for evaluating normally distributed samples might be of lesser importance. On the other hand, one also ought to remember the conclusions from (Barber and Lyon, 1997) that since both samples is constituent of closely matched stocks; it is more likely that the test is accurate. Still, in a case where we would have significant results it still would be hard to give that too much weight all on itself since the sample size is not that large.
Finally, regarding the formation and quality of the portfolios, it is evident that the number of stocks with high book-to-market and positive EVA-spread dominates the stocks with low book-to-market and negative EVA-spread. This is a problem when evaluating the results but since the test was a zero net investment any single portfolio only represents half of a year´s return as its whole, which still does not clear the problem but mitigates its importance to some extent.
6. DISCUSSION
If we look at the evidence presented in the analysis, there is uncertainty about how to interpret the figures that have been brought to the table.
If we see past the contradictory evidence, we must put weight on the fact that the enhanced risk measure M2 and that the EVA sort on average beat the benchmark portfolio, with a sample that although it was relatively small not in any way was miniscule. This indicates strength in the EVA analysis. But, if we break down the figures, the total return equals a yearly yield of 7,32 % giving the investor only an average compensation over the 11 year
27 period, not that much higher than the average risk free rate for the period of 5,58 %. How can these average results then be interpreted, what does the economic value really add?
Since the EVA portfolio seemingly outperforms the benchmark, it is relevant to reflect over what kind of information the Economic Value Added might signal that leads to this relative over performance. The EVA analysis carries the thought of economic surplus. Economic surplus in the EVA sense means that the company has created shareholder value, compensating the shareholders through earning more than the sources of financing. But, when this analysis is applied to a stock market context with extreme valuations in the analyzed stocks, the time lag of a year makes it more difficult to connect the past tense performance conveyed through the accounting numbers. This concern is stemming from that where the earlier research surrounding fundamental analysis has been able to gain higher return it often has come from more extensive and thorough methods of analysis. Of course, fundamental analysis in its nature is both time lagging and dependent on extended mispricing. The methods used in other articles have often been concerned with signals pointing towards companies that
“does the right things” and thus can be able to sort out more “mispriced” stocks even if they are not making economic profit in the EVA sense. It is therefore from that point of view possible that the Economic Value Added analysis points out some but not all stocks worth buying (selling) since it is dependent on accounting profit (or loss). This makes it on one hand less risky in the way that a company need, due to the nature of the analysis and the method, economic profit to be bought (sold), but at the same time making it less attractive as a contrarian signal. One other drawdown discussed in the method critique also is obvious with hindsight, that the effect of the estimation of the cost of equity capital is, because of the nature of the analysis, quite big on the EVA-spread calculation. This issue is troublesome since it is easy to include or exclude stocks on the verge of making it into the portfolio through the equity cost of capital, possibly leading to including mediocre performers. On the other hand could a method that weights the stocks after their relative EVA spread combined with a minimum threshold spread lessen these effects if the test is repeated in the future.
Furthermore, when reflecting over if the EVA analysis return is due to more risk or to mispricing, at least here the evidence is more in favor of mispricing out of line with the stronger versions of the EMH. If markets could not be mispriced for prolonged periods of time the EVA analysis would be worthless since it utilizes accounting data, which by its nature lagging the current information of the company. Since it in this test has been applied on stocks with high (low) relative valuation, it is possible that the EVA conveys a strong signal
28 to investors that valuation is stretched too far by emphasizing the economic efficiency of a company. And, this signal may, because of its more implicit nature, take more time to be absorbed by the market where investors earlier have overemphasized some other factor. This resulting in a revaluating that might take more time than the efficient market stipulates, thus making this ex-post analysis style useful to at least some degree in contrarian investing. It is also possible that the EVA is a predictor of some other important aspects affecting the company valuation. But since the performance still not in any way is stellar, the issue of cause and effect is hard to separate which might be a good topic for further research where firm characteristics and different estimation techniques could be a good way of extending this paper.
6.1 Conclusion
Although the discussion of the actual use and value of EVA is motivated, the purpose of the paper was to test whether the EVA analysis could increase the returns to a hedge portfolio sorted through size and book-to-market.
Despite the mixed evidence, our methods allow us to put different weight on the findings enabling us to draw some conclusions. The EVA analysis did in fact increase the return of the hedge portfolio with a better risk-profile, which leads us to conclude that it does serve as a heuristic that increases the returns to a contrarian investment strategy of this kind utilizing size and book-to-market sorting. Especially in the light of the benchmark portfolios unexpectedly meager performance, the EVA seems to have been adding value in this case.
However, due to the intact null hypothesis and the relatively average overall performance it is difficult to assess whether the EVA analysis alone adds the edge to the strategy, or if it serves as a proxy for some other unknown explanatory factor discussed previously. We can also conclude that since valuation in the sample is relatively high (low) the abnormal returns is probably rooted in mispricing as it is a more plausible explanation for the superior performance. This since it is more unlikely that the sampled stocks where subject to more risk than suggested by its sorting.
6.2 Suggestions for further research
The possible extensions of this type of research are widespread, especially since the results in this paper were dependent on different methodological and theoretical important choices to
29 arrive at its destination. One example would be to further examine the obvious problem with the estimation of the equity cost of capital in the EVA analysis. How does the result differ depending on what method of estimation is used? Additional research of this and other methods across multiple stock markets could also help to shed more light over the analysis efficiency. To study the behavior of EVA portfolios performance under different contexts as times of high and low-inflation, bull and bear market conditions or over different sectors of stocks could also potentially bring more clarity in the matter.
7. SOURCES
Articles
Abarbanell, J. & B. Bushee (1998) “Abnormal returns to a fundamental analysis strategy”.
Accounting Review, 73, 19-45.
Abate J., Grant J. L., Bennet Stewart III (2004) “The EVA style of investing – Emphasizing the fundamentals of wealth creation”, The Journal of Portfolio Management Summer2004, Vol. 30 Issue 4, p 61-72.
Ali, A., L. Hwang and M. A. Trombley. (2003). “Residual-income-based valuation predicts future stock returns: Evidence on mispricing vs. risk explanations”. The Accounting Review (April): 377-396.
Banz, Rolf W. (1981), The Relationship between Return and Market Value of Common Stocks, Journal of Financial Economics 9, 3.
Barber, Brad M., and John D. Lyon. 1997. Detecting long-run abnormal stock returns: The empirical power and specification of test statistics. Journal of Financial Economics 43, no. 3:
341-372.
Bernard, Victor, Jacob Thomas, and James Wahlen. (1997). Accounting-Based Stock Price Anomalies: Separating Market Inefficiencies from Risk. Contemporary Accounting Research 14, no. 2: 89-136.
James L. Davis. (1994). “The Cross-Section of Realized Stock Returns: The Pre- COMPUSTAT Evidence”. Journal of Finance, 49(5), p 1579-1593.
Fama, Eugene F. (1970). "Efficient Capital Markets: A Review of Theory and Empirical Work". Journal of Finance 25 (2): 383–417.
30 Fama, Eugene F., and Kenneth R. French, (1992), “The Cross-Section of Expected Stock Returns”, Journal of Finance 47, 427.
Fama, Eugene F. and French, Kenneth R, (2008), “Dissecting Anomalies” Journal of finance 63, Issue 4 (August)
Lakonishok, J., Shleifer, A. & Vishny, R.W., (1994). “Contrarian Investment, Extrapolation, and Risk”. Journal of Finance, 49(5), 1541-1578.
Lev, B. & S. Thiagarajan (1993) Fundamental information analysis. Journal of Accounting research, 31, 190-215.
Lintner, John (1965). “The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets”, Review of Economics and Statistics, 47 (1), 13-37.
Markowitz, H.M. (1952). ”Portfolio Selection”. Journal of Finance 7 March (1): 77–91 Modigliani, Franco and Leah (1997). "Risk-Adjusted Performance". Journal of Portfolio Management, Vol. 23, No 2. (Winter) p: 45-54
Mohanram, P. (2005) Separating Winners from Losers among Low Book-to-Market Stocks using Financial Statement Analysis. Review of Accounting Studies, 10, p 133-170.
Mossin, Jan. (1966). “Equilibrium in a Capital Asset Market”, Econometrica, Vol. 34 No. 4, p 768–783.
McGoun, E.G. (1993). “The CAPM: A Nobel Failure”. Critical Perspectives on Accounting 4 (2) (June): p 155-177.
Ou, J.A. & Penman, S.H., 1989. ”Financial statement analysis and the prediction of stock returns”. Journal of Accounting and Economics, 11(4), p 295-329.
Piotroski, J. D. (2000). “Value investing: The use of historical financial statement information to separate winner from losers”. Journal of Accounting Research. 38: p 1-41
Sharma, A. K. (2010). “Economic Value Added (EVA) - Literature Review and Relevant Issues”. International Journal of Economics and Finance. Vol 2, No 2
Sharpe, William F. (1964). “Capital asset prices: A theory of market equilibrium under conditions of risk”, Journal of Finance, 19 (3), p 425-442
31 Sharpe, William. Forsyth. (1966). "Mutual Fund Performance". Journal of Business 39 (1): p 119–138
Shapiro, S. S. Wilk, M. B. (1965). "An analysis of variance test for normality (complete samples)". Biometrika 52 (3-4): 591–611.
Stewart, G. B., III. (2003, June 2). “How to fix accounting – Measure and report economic profit”. Journal of Applied Corporate Finance, 15 (3), p 63-82.
Wilcox, R.R.; Keselman, H.J. (2003), "Modern robust data analysis methods: Measures of central tendency", Psychological Methods 8 (3): 254–274
Worthington, A. & West, T. (2004). “Australian Evidence Concerning the Information Content of Economic Value-Added”. Australian Journal of Management, 29(2), p 201-224.
Booksources
Haugen, Robert. A (2001) Modern Investment Theory, 5th ed. Prentice Hall
Reilly, Frank K. och Brown, Keith C. (2006) Investment analysis and portfolio management.
9. Canadian edition. Mason, Ohio: Thomson/South-Western
Ross, Stephen A, Westerfield, Randolph W. Jaffe, Jeffrey (2005) Corporate finance, 7th ed.
McGraw-Hill WebSources
Damodaran, 2010. “The Equity Risk Premium: Determinants, Estimation and Implications:
The 2010 Edition”, Retrieved November 20th, 2010 from:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1556382
U.S Bureau of Labor Statistics, 2010. Inflation calculator. Retrieved November 20th, from:
http://data.bls.gov/cgi-bin/cpicalc.pl
Yahoo!® Finance,2010. ”Stock Screener”. Retrieved November 20th, 2010 from:
http://screener.finance.yahoo.com/newscreener.html
32
APPENDIX 1 - Terminology
In the course of this essay, some expressions that might not be known for the reader have been used, therefore we briefly define the most important below:
“Abnormal return” actual return – expected return
“Book-to-market” market valuation measurement, (book value of equity / market value of equity)
“CAPM” capital asset pricing model
“Excess returns” returns over a period minus the risk free rate
“Ex-ante return” expected return of portfolio investment
“Ex-post return” actual return of portfolio investment
“EVA®” economic value added (ROC – WACC*K)
“EVA®-spread” (ROC – WACC)
“Hedge-portfolio” a zero net invested portfolio
“Large Cap” a market value (number of outstanding shares times the market price) larger than three point seventy-five billion dollars as of 1 January 1998
“NOPAT” net operating profit after tax
“ROC” returns on capital
“Short selling” borrowing shares and selling at market value in speculation of buying it back at a cheaper price in the future, thus, returning the loan and collecting any difference in price as a profit.
“The American stock market” stocks listed on the New York stock exchange and the Nasdaq OMX marketplace.
“WACC” weighted average cost of capital
