Theoretical point of departure
This section will present the theories and previous empirical studies related to factor investing, and in turn, fundamental for this study. It starts with the Efficient Market Hypothesis, and moves on to the Capital Asset Pricing Model, after which the Quality factor and the Value factors are introduced further, which are the factors this study intends to investigate. Furthermore, this section includes an explanation of factor cyclicality, and presents relevant previous research into factor investing models.
Efficient Market Hypothesis
In a market, where all participants are given equal access to information, the efficient market hypothesis suggests that the price of an asset fully reflects all available information (Fama, 1970, p.383). Fama (1970) presents three forms of market efficiency; weak form, semi-strong form and strong form. The weak form is based on the research around the random walk theory of asset prices, and regards information carried by historical prices. The semi-strong form regards the price adjustment to reflect new publicly available information, such as earnings announcements for example, and the strong form suggests that all information, private or public, is accounted for in the asset’s price (Fama, 1970, p.383). There are, however, numerous inconsistencies in the efficient market hypothesis (Jensen, 1978; Banz, 1981; Fama and French, 1992; Ball and Brown, 1968; Jegadeesh and Titman, 1993; Basu, 1975; 1977; 1983). These inconsistencies are called anomalies and points to characteristics giving systematically higher returns than suggested by the efficient market hypothesis (Jensen, 1978, p.4-8). In order to determine if returns are, in fact, abnormal one must consult an asset pricing model, where the Capital Asset Pricing Model is the most commonly used.
The Joint Hypothesis Problem
The following section gives an explanation to the choice and inclusion of CAPM in the practical method as well as the importance of testing the Efficient Market Hypothesis through an asset pricing model. It will answer the question to why CAPM plays such an important role in the practical method as well as in this study.
For a strategy to violate the Efficient Market Hypothesis, it is not simply sufficient that it consequently earns higher returns than the market. Higher risk should, in theory, be rewarded with higher returns. Earning consequently higher returns than the market could therefore be explained by taking on higher risk. A portfolio taking on greater risk than the market may be thus also be rewarded with higher returns. This suggests that in order to examine whether or not a strategy violates the Efficient Market Hypothesis or not, one has to take into account the riskiness of the strategy. This may be accomplished by testing the returns against the Efficient Market Hypothesis through an asset pricing model, such as CAPM for example, as it specifies how much reward an asset should earn due to its inherent risk. However, there are different asset pricing models, such as CAPM, Arbitrage Pricing Theory, and Fama-French Three Factor Model. As the models differ from each other, the specifications of risk and return differ between them as well. This poses a big problem for testing the Efficient Market Hypothesis, since the choice of asset pricing model affects the outcome due to their different specification of risk and return. One will, in other words, be dependent on the choice of asset pricing model. It becomes a joint hypothesis problem, as testing the Efficient Market Hypothesis requires testing a model as well. This implies that a miss-specified asset pricing model may lead to the wrong conclusion when testing the Efficient Market Hypothesis. Thus, two things can go wrong; one or two of the hypotheses is false or the joint hypothesis is false, and you cannot distinguish which one it is. (Fama, 1992; Jensen, 1978)
As explained above, in order to examine whether the quality and value factor violates the Efficient Market Hypothesis, firstly defining risk and return through selecting an appropriate asset pricing model is required, and, secondly, a joint hypothesis test has to be executed. In this study, CAPM has been selected as the asset pricing model founding the test of the Efficient Market Hypothesis. In short, CAPM is the means by which we can test the Efficient Market Hypothesis. The choice of CAPM is based on it being the model used to greatest extent in previous research of factor investing as well as it being greatly recognized within both the practical and theoretical field of economy.
Capital Asset Pricing Model
The Capital Asset Pricing Model was developed by Treynor (1961), Sharpe (1964), Lintner (1965) and Moussin (1966). The model proposed that the return of an investment was to be relative to the market risk associated with that same investment. It showed on a linear relationship between the two factors (1) expected return and (2) market risk, where an investment with high market sensitivity would require a high expected return, and vice versa. Black (1972) further explored the model, by constraining the hypothetical investors access to a risk-free asset. First by removing it entirely, then by only allowing long positions in the risk-free asset (Black, 1972, p. 455). Black (1972, p. 455) also found that the expected return on a risky asset must be a linear function of its beta. If the borrowing is restricted, the slope on the expected return is lower than in the case where the investor has access to borrowing (Black, 1972, p. 455). CAPM as a formula is usually written as: E(Ri) = Rf + Bi[E(Rm) – Rf], where E(R) is the expected return, B is the beta, E(Rm) is the expected market return, and Rf is the return on a risk-free asset. The Beta is calculated as the covariance of the market return and the return on the risky asset, divided by the variance in the market return. CAPM uses the assumptions (1) that all investors hold the same opinions about the values for all assets, (2) the probability distribution of the possible returns are joint normal, (3) investors are wealth maximizing and risk averse, and (4) an investor can take both long and short positions in any asset, meaning that the investor can both borrow and lend at the risk-free rate (Black, 1972, p. 444). This is known as the Sharpe-Lintner-Black model. A study by Fama and French (2004, p.25) clarifies that the CAPM became very popular ever since, and has for over four decades been widely used in both practice and for educational purpose; the model’s simplistic properties, yet powerful prediction power of the risk and return relationship, has made it a very popular tool for quite some time. This, despite criticism from, among others, Roll (1977), who proposed the Market Proxy Problem, which criticizes the capital asset pricing model due to the difficulties of defining the essential component market portfolio.
Challenging the CAPM, Fama and French (1992), Statman (1980), Rosenberg, et al. (1985, referenced in Fama and French, 1992) and Chan et al. (1991) sought out to investigate risk and return relationships, studying the relationship between value and stock performance, concluding that there was, indeed, a tendency that cheaper stocks gave higher risk adjusted returns. Further research was done by Piotroski (2000), who studied the relationship between quality and stock performance, finding that more profitable firms tend to have higher returns. More specifically, evidence showed that among high Book-to-Market firms, the healthiest firms appeared to generate the strongest returns (Piotroski, 2000, p.28). This led to different models being developed in order to capitalize on such anomalies. There is evidently a theme of contradictions to the CAPM model, where Beta repeatedly throughout studies lack to explain certain returns (Fama and French, 2004, p.35-36).
Further critique to CAPM was presented by Ross (1976) who proposed the Arbitrage Pricing Theory. The Arbitrage Pricing Theory was an asset pricing model, just as CAPM, but estimated the return of a single asset through a linear combination of many independent macroeconomic variables (Ross, 1976, p. 341-343). The development of the Arbitrage Pricing Theory came to be a continuation of the theoretical foundation for factor investing as it allowed for many different sources of systematic risk in contrast to the CAPM, which only allowed one type of systematic risk.
This section will present theory and empirical evidence regarding the quality factor. This will serve as a foundation for the choice of including the quality factor in this study, and will show that it does in fact have power to improve the performance of a value factor portfolio, and thus outperform the market index. The quality factor is somewhat difficult to define, but common definitions include earnings stability, dividend payouts, investment levels, profitability, cash flow, level of accruals, etc.
Piotroski (2000) finds that the distribution of the return of a broad book-to-market portfolio can be shifted by applying a simple accounting based fundamental analysis. Piotroski (2000, p. 28) found that among high Book-to-Market firms, the healthiest firms appeared to generate the strongest returns. Piotroski (2000, p. 37) therefore suggests that a generic Book-to-Market portfolio can be improved by using relevant historical information in order to eliminate firms with weak future prospects. Furthermore, Piotroski (2000, p. 37) found that the benefit of financial statement analysis of the high Book-to-market firms was concentrated to small and medium sized firms, with low share turnover, and no analysts following. Piotroski (200, p. 38) also found that there is a relationship between the historical information, and to both the subsequent quarterly earnings announcements and the future performance. This implies that using fundamental analysis in order to assess firms based on quality, can improve the performance of the portfolio.
In contrast to Piotroski (2000), this is also explored by Mohanram (2005), who instead studied the effect of financial statement analysis on low book-to-market portfolios, known as portfolios of growth stocks. Mohanram (2005) points to the empirical phenomenon that low book-to-market portfolios tend to underperform relative to the market, which is also shown in the previous section, but also points to the fact that the variation in stock return is considerable. The author is thus going to apply financial statement analysis in order to try and separate the winners from the losers in a low book-to-market portfolio (Mohanram, 2005, p. 137). Low book-to-market firms are defined as ratios below the 20th percentile for the entire market, and Mohanram (2005) uses three categories of signals when doing the financial statement analysis (Mohanram, 2005, p. 138). These groups of signals are defined as; Category 1: Signals based on Earnings and Cash Flow Profitability, Category 2: Signals Related to Naive Extrapolation, and Category 3: Signals Related to Accounting Conservatism (Mohanram, 2005, p. 138). Category one includes measures such as Return on Assets using Net Income before extraordinary items divided by total assets (Mohanram, 2005, p. 138). It also includes a binary measure, which checks if the cash flow return on assets exceed the the median of all low book-to-market firms in the same industry (Mohanram, 2005, p. 139). The final measure in the first category concerns whether or not the cash flow exceeds the net income of the firm (Mohanram, 2005, p. 139). In the second category Mohanram (2005,
pp. 139-140) uses a measure which concerns the variability of earnings relatively to all low book-to-market firms in the same industry, and a measure the variability in sales growth also relatively to all low book-to-market firms in the same industry. The final category includes three measures, which checks if a firm’s research and development, capital expenditure, and its advertising spending, are higher than other low book-to-market firms in the same industry (Mohanram, 2005, p. 140). These measures are aggregated into what Mohanram (2005) calls GSCORE. Mohanram (2005) employs a strategy that buys firms with a high GSCORE and sells firms with low GSCORE. The findings indicate that this strategy is able to differentiate the winning stocks from the losing ones, which is shown by the fact that high GSCORE firms earned substantially higher returns than the los GSCORE firms (Mohanram, 2005, p. 166). However, Mohanram (2005, p. 166) notes that much of this performance comes from the gains on the short positions in the low GSCORE firms, and thus the strengths of this strategy lies in its ability to find the stocks to avoid. Furthermore, when controlling for factors such as book-to-market, accruals, momentum, etc, GSCORE is positively correlated with future returns over a long time period (Mohanram, 2005, p. 166). Mohanram (2005, p. 167) argue that these results must be because of the misinterpretation by investors regarding the financial information given by firms, which is contrasted to the findings by Piotroski (2000) which point to ignorance of certain information. These findings by Mohanram (2005) show, like the findings by Piotroski (2000), that fundamental analysis of the firm’s financial information, in order to discover its inherent quality, can lead to excess return and improves a portfolio created using book-to-market.
Fama and French (2006) explored the book-to-market equity, together with the expected profitability, and the level of expected investments. Fama and French (2006, p. 495) create a proxy for profitability on investments, using cross-section regressions. These fitted values from the regressions are then used in the cross-section return regression, which tests the effect of profitability and investments on the stock return (Fama and French, 2006, p. 495). These tests exclude financial companies, and is mostly during 1963 – 2003 (Fama and French, 2006, p. 496). For example, Fama and French (2006) found that firms with high dividends, high levels of accruals, and high book-to-market ratio, grew their assets more slowly. On the other hand, smaller firms, firms with lower dividends, low book-to-market, and low accruals, grew faster (Fama and French, 2006, p. 495). Regarding profitability, they found that profitability is both persistent and mean reverting (Fama and French, 2006, p. 501). Furthermore, they found that high book-to-market firms were less profitable, and non-dividend paying firms are also less profitable (Fama and French, 2006, p. 501). In order to test how these factors impact the expected return, Fama and French employ three steps: (1) using simple cross-section regressions to examine how they add to the explanation of stock returns provided by size and book-to-market equity, (2) using more complicated proxies gathered from fitted values from the regressions, and (3) using portfolios to examine if the profitability and investment effects persist (Fama and French, 2006, pp. 502-503). Fama and French (2006, p. 514) find that high book-to-market equity had higher expected returns when controlling for expected profitability and investments. When given the book-to-market and the expected profitability, higher investments correlated with lower return (Fama and French, 2006, p. 514. Fama and French (2006, p. 514) find that these results are in line with the the existing evidence, both that book-to-market equity has good descriptive power, and that more profitable firms have higher expected returns, while more investments or higher level of accruals are associated with lower returns. Thus, this shows evidence that examining the firm’s quality, based on profitability, or level of investments, can improve the explanatory power of the book-to-market equity measure.
Asness et al. (2013) examined if investors were willing to pay more for quality. They claim that high-quality stocks have earned high risk-adjusted returns, and low-quality stocks have had negative risk-adjusted returns (Asness, et al., 2013, p. 2). They found that a Quality-minus-Junk factor (QML) going long high-quality stocks, and shorting low-quality stocks earned high risk adjusted returns (Asness, et al., 2013, p. 25). Asness et al. (2013, p. 3) uses Gordon’s growth model to express the price-to-book as P/B = (Profitability * pay-out ratio) / (Required return – growth). This leads them to identify four factors of quality, and thus factors an investor would want to pay an increased price for (Asness, et al., 2013, p. 2-4). The four parts of the quality assessment are profitability, growth, safety, and pay-out ratio (Asness, et al., 2013, p. 3-4). Profitability is defined as profit per unit of book value, and all else equal, a more profitable firm deserves a higher stock price (Asness, et al., 2013, p. 3). Asness et al. (2013, p. 3) also suggest that an investor would want to pay more for a firm that is growing its profits, and that an investor should pay a higher price for a safer stock. Safety is defined by Asness et al. (2013, p. 3) as a stock with lower volatility, and thus lower beta, and a lower required rate of return. Pay-out ratio is defined by Asness et al. (2013, p. 4) as the fraction of profits that gets paid out to shareholders. This can be seen as a measure of shareholder friendliness, and is decided by the managers (Asness, et al., 2013, p. 4). Asness et al. (2013, p. 4) point out that if a higher pay-out ratio is achieved at the cost of growth or lower future profitability, an investor should not pay more for it, but if all else is equal, higher pay-out should be positive. According to Asness et al. (2013, p. 4) the companies that are profitable, growing, stable, and has high pay-out ratio, continue to show the same characteristics on average over the next five or ten years. Furthermore, Asness et al. (2013, p. 5) found that the Quality minus Junk factor performed well, and during market downturns it did not crash, but instead showed mild positive convexity, therefore benefitting from investors moving toward quality in market turmoil. Asness et al. (2013, p. 6) also combines the Quality minus Junk idea with the High Minus Low (HML) idea from Fama and French (1992), into what Asness et al. call Quality at reasonable price (QARP). While the QML buys and sells based on quality measures, irrespective of price, HML does the reverse, it buys solely on price, and ignores quality (Asness et al., 2013, p. 6). Combining the two into QARP improves the value investing, which is consistent with what Graham and Dodd (1934, referenced in Asness et al., 2013, p. 6) writes, and with what Piotroski (2000) found. Finally, Asness et al. (2013, p. 25-26) found that the price investors are willing to pay for quality vary over time, and that low price of quality was a predictor of high future return. These findings implicate that higher quality firms outperform lower quality firms, and thus gives reason to believe that selecting stocks based on their quality can lead to abnormal returns.
Regarding the quality of a firm’s accounting, Perotti and Wagenhofer (2014) examined how common measures of earnings quality improved the decision quality of a company’s reporting for investors. They hypothesized that firms with higher earnings quality will be less mispriced than firms with lower earnings quality (Perotti & Wagenhofer, 2014, p. 545). The mispricing was measured as the difference if the mean absolute excess return of portfolios of high values of the quality measures, versus portfolios of low measures in quality of earnings (Perotti & Wagenhofer, 2014, p. 545). Perotti and Wagenhofer (2014, p. 567) constructed the portfolios based on eight different measures of earnings quality, namely persistence, predictability, abnormal accruals, accrual quality, earnings response coefficient, and value relevance. They found that the earnings smoothness, measured as the persistence and as predictability, had a positive relationship with absolute excess return (Perotti and Wagenhofer, 2014, p. 567). Furthermore, they found that accruals measures were the most useful earnings quality measures (Perotti and Wagenhofer, 2014, p. 567). This is somewhat related to a study whether or not stock prices reflect the information carried in accrual and cash flow parts of the firm’s earnings, Sloan (1996) found that investors tend to fixate on earnings, thus not making the stock price reflect the values generated in accruals or cash flows. Sloan (1996, p. 290) found that the earnings related to accruals showed less persistence than earnings related to cash flow. Thus, Sloan (1996, p. 290) concludes that firms with high levels of accruals have lower future abnormal stock returns, and those with less accruals have higher levels of abnormal return. This abnormal return is concentrated around future earnings announcements (Sloan, 1996, p. 290). Sloan (1996, p. 306) found that and investor could earn abnormal returns by abusing the fact that stocks were not priced according to the information carried in the accrual and cash flow parts of earnings. Even though not directly linked to this study, this shows evidence that assessing companies based on their quality can pay off.
Even though there are several definitions on what makes a company higher quality, for example better earnings stability, quality of accounting, and level of profitability, these studies show that it might be possible to achieve abnormal returns by thoroughly researching a company’s financials, and that there might be abnormal returns from finding higher quality companies to invest in.
The theory and empirical evidence presented in this section will serve as a foundation for the choice of the value factor in this study. The theories and empirical evidence shows that the value factor can outperform the market index on a risk adjusted basis, and gives us reason to believe it is worthwhile to include it in this model. The value factor refers to the finding that stocks that are cheaper have outperformed stocks that are more expensive. Cheapness is usually measured by book-to-market equity (Book value of equity/Market value of equity), earnings yield (Earnings/Price), Price-to-earnings (Price/earnings), or in some cases as dividend yield.
According to Fama and French (1992, p. 450) size, and book-to-market equity captures the cross-sectional variation in stock returns associated with size, earnings yield, book-to-market equity, and leverage, for the 1963 to 1990 period. Furthermore, Stattman (1980) and Rosenberg, et al. (1985, referenced in Fama and French, 1992) finds a positive relationship between average return and Book-to-Market equity on the US market. These findings showed evidence of outperformance by cheap stocks over more expensive stocks.
Table of contents :
1.1 Problem Background
1.1.1 Factor Investing
1.1.2 Common Factors
1.2 Problem Discussion
1.2.1 The Importance of Factor Investing
1.2.2 Choice of Factors
1.3 Purpose and Research Question
1.4 Theoretical and Practical Contribution
1.5 Choice of Topic
2. Theoretical point of departure
2.1 Theoretical Framework
2.1.1 Efficient Market Hypothesis
2.1.2 The Joint Hypothesis Problem
2.1.3 Capital Asset Pricing Model
2.1.4 Quality Factor
2.1.5 Value Factor
2.1.6 Factor Cyclicality
2.2 Previous Research
2.2.1 Fama and French
2.2.2 The Magic Formula
3.2 Literature Search
3.3 Scientific Approach
3.3.1 Ontological Assumptions
3.3.2 Epistemological assumptions
3.4 Research Approach and Methodological Choice
3.5 Practical Method
3.5.2 Factor Model and Portfolio Creation
3.5.3 Calculations and comparisons
3.5.5 Statistical Method
3.5.6 Method Discussion
5.1.1 Portfolio 1
5.1.2 Portfolio 2
5.1.3 Portfolio 3
5.1.4 Portfolio 4
5.1.5 Portfolio 5
5.1.6 Summary of Hypotheses
5.2 Additional Data
5.2.2 Data from Regression Analysis
6. Analysis of Results
8.1 Applicability of Results
8.2 Recommendations for Further Research
8.3 Ethical and Societal Considerations
9. Truth Criteria