Avoiding Value Traps:
Corporate Governance and Value Investing
Ádám Both*
January, 2017
ABSTRACT
This paper investigates the relationship between corporate governance and value investing, using a sample based on the S&P 500 index – for the period of 2005-2014. A two-stage analysis is carried out. Firstly, the relationship between corporate governance and valuation is analysed with the help of a panel regression of corporate governance scores on valuation ratios. Secondly, portfolios are constructed using a two-way sort. The findings fail to detect a positive relationship between corporate governance and future stock returns. Governance metrics do not affect valuation in an economically significant manner. The main portfolios of interest (High MTB and Strong Governance) in the two-way sort do not outperform a purely value based investment strategy. The exposure patterns of some two-way sort portfolios to the Quality factor could be interpreted as limited evidence of corporate governance being a proxy for various attributes of Quality, including superior operating performance.
Keywords: Corporate Governance, Value Investing, Cash Flow, Investment Strategy
1
1. Introduction
The aim of this paper is to evaluate the effect of corporate governance on company valuation, and attempt to implement the findings as an investment strategy. This paper uses a method of linking corporate governance to value investing. There is evidence of corporate governance being a proxy for superior performance in terms of profitability and other operating metrics. Therefore, it could be reasonably assumed that a strategy that combines a valuation (market-to-book) and a quality (corporate governance) criterion may prove to be more profitable, than a strategy that buys stocks irrespective of profitability (value investing) or irrespective of valuation (growth or quality investing).
Prior research is inconclusive about the effect of corporate governance –and stronger shareholder rights - on equity returns. A landmark paper exploring the relationship between governance and returns is Gompers, Ishii and Metrick (2003), where the authors find an approximately 8% abnormal return per annum on a portfolio that holds long positions in strong governance companies and shorts weak governance ones. Subsequent studies failed to find superior stock market performance by strong governance companies – e.g. Core, Guay and Rusticus (2006), Bhagat and Bolton (2008). Although Bhagat et al. (2008) found superior operating performance by strong governance firms. Others could only find abnormal returns only when additional criteria were considered such as Cremers and Nair (2006). The authors only found stronger performance by strong governance companies, when institutional block-holders were present. Another strand of empirical literature employed an event study approach – e.g. Cohn, Gillan and Hartzell (2016). Unfortunately, investors are unable to reap governance related returns in those cases without access to insider information. This serves as a main motivation behind this current piece of research – examining corporate governance in a way that could be usable by investors.
2
The sample used in the paper is based on the S&P 500 equity index, for the time period spanning from 2005 to 2014. Thus, the sample covers a portion of the bull market before the Global Financial Crisis, the crisis and a large part of the recovery afterwards. Due to the characteristics of the S&P 500 index, the sample is inherently biased towards large market capitalisation. Moreover, it consists exclusively of United States based companies, resulting in strong geographical concentration and developed market bias. Data is obtained from the Thomson Reuters Datastream service, with corporate governance metrics originating from the Asset4 sub-database. These inherent characteristics do not necessarily need to be described as limitations. The index is a widely-used benchmark, any finding related to the characteristics and return patterns of its constituents are important to international investors. Although the importance of broadening the sample, when passing final judgement on the strategy outlined in the paper cannot be denied.
Findings of the paper are varied, though largely in line with prior research. The results of the regression show that corporate governance scores only have an economically meagre effect on valuation. When disaggregating the aggregate score into category scores, none of them appears to drive the relationship. Coefficient values are low consistently. Considering the portfolio sorts, somewhat counter intuitively the portfolio of stocks from the highest valuation and weakest governance quartile shows a statistically highly significant monthly alpha of about 1.3%. The main portfolio of interest – low valuation and strong governance – on the other hand does not show a statistically significant alpha. It should be pointed out, though that it outperforms the market portfolio – both if defined as the overall S&P 500 index or as the Market factor of Fama and French and the ‘Quality minus Junk’ portfolio - in terms of Sharpe ratios and expected returns. Some portfolios from the double sort show factor exposure patterns to the Quality factor that could be interpreted as a form of evidence towards strong governance being a proxy for superior operating performance – this finding is robust to a number of alternative specifications. On the other hand, findings pertaining to return patterns are not.
3
2. Prior Research
A: The Concept of Corporate Governance
In contemporary financial literature, the issue of corporate governance emerges in two very different contexts. One area considers corporate governance a sub-section of ESG, research falling in this category understands governance issues to be in conflict with shareholder value maximization, as the viability of the inclusion of non-financial criteria is analysed and evaluated – such as El Ghoul, Guedhami, Kwok and Mishra (2011). The second school of thought takes a strictly financial view of corporate governance, inspecting the effect of corporate governance as primarily but not exclusively an issue of shareholder rights – and its effect on firm performance, Mergers & Acquisitions activities, management risk taking, management entrenchment and stock performance – e.g. Gompers et al. (2003), Core et al. (2006) and Becht et al. (2016), among others. The latter strand of corporate governance research relates to the extensive literature on optimal control rights allocation and shareholder activism, or even research pertaining to the ‘Law and Economics’ literature about ensuring optimal levels of investor protection and market transparency. Work by La Porta, Lopez de Silanes, Shleifer and Vishny (2002) falls into this category, where the authors find a positive relationship between valuation and investor protection. In a more recent publication McLean, Zhang and Zhao (2012) find a positive relationship between investor protection and efficiency in company-level resource allocation. This current piece of research considers corporate governance a strictly shareholder value enhancing tool. Either as an indirect proxy for financial risks associated with a certain group of companies, or as a catalyst of returns on its own.
B: Shareholder Activism and the Event Study Approach
4
Vishny (2003) – according to their research, overvalued companies can use their stock as a currency to acquire competitors.
The research conducted by Becht et al. (2016) is similar to that of Cunat, Gine and Guadalupe (2012|. The authors analyse CARs around the passage of shareholder-friendly proposals on general meetings. The sample is limited to proposals that fail or pass only narrowly, as to avoid the effect of the outcome being already priced in – the passage of a governance proposal increases shareholder value by 2.8 percent. Another event study by Larcker, Ormazabal and Taylor (2011) on the other hand, finds that easing proxy access and curbing executive compensation by federal law is shareholder value destroying. The authors argue that some of these regulations represent unnecessary state meddling in the optimal allocation of control rights. Moreover, arguably some proxy access changes may benefit a certain (smaller) group of shareholders over the majority. The findings of Larcker et al. (2011) are statistically significant only in case the events relating to the passage of the law are disaggregated (e.g. Introduction, Senate hearing, first appearance on the news, etc.). When aggregated into larger categories and conducting a pooled multi-event analysis, the abnormal returns are no longer statistically significant.
As one of the aims of the current paper is to link corporate governance characteristics to value investing, an important piece of prior research is that of Cohn, Gillan and Hartzell (2016). The authors also analyse returns around the US Congressional votes on Dodd-Frank proxy access regulations. There is an overlap between the events studied by Cohn et al. (2016) and Larcker et al (2011). The main point of difference being that Cohn et al. (2016) introduce size and holding period caveats and employ a stricter event identification methodology. Due to the inclusion of these restrictions and a more thorough selection process of ‘activist’ investors the findings of Cohn et al. (2016) can very well be considered more robust. Similarly, to the current piece of research, Cohn et al. (2016) use book value of equity when determining the value of leverage.
The key weakness of the event study approach in general is that the findings are of limited applicability if one aims to implement corporate governance criteria in an investment strategy – without access to insider information, the returns are not achievable to investors. One of the main contributions of this current piece of research is its investor centric approach.
C: Governance and Firm Performance
5
expenditure and higher return on assets, among others. Bearing in mind that the information used to construct the index was publicly available, the return shown raises concerns about the efficient market hypothesis. Subsequent studies were not able to replicate the results showing superior stock market performance by strong governance firms outside the sample period tested in Gompers et al. (2003). Anderson and Reeb (2003) when attempting to replicate the results exclude technology firms from the sample, and were unable find superior stronger performance by strong governance stocks. Moreover Core, Guay and Rusticus (2006) find that the exact same portfolio building technique that is used by Gompers et al. (2003) does not lead to superior stock market returns, outside the sample period – in fact, companies with weaker shareholder rights perform better, the excess return falls to 4.8% from 8.3% in the case of the strong governance portfolio. Therefore, the authors conclude that the returns found by Gompers, Ishii and Metrick (2003) fit into a wider set of pricing anomalies present on the stock market in the 1990s. In addition to analysing the relationship between stock market performance and corporate governance Core et al. (2006) also inspect whether the under or outperformance by companies is expected by the market or not. They do so by testing analyst forecast errors and announcement day returns with respect to operating profitability metrics. The authors find that strong governance companies show superior operating characteristics, although the market prices in this superior performance. Therefore, stronger corporate governance does not result in higher returns.
6
financial metrics. The authors also suggest that excessive leverage may suggest a breakdown of external governance mechanisms, thus the papers also provides justification for controlling for leverage in the regression.
In order to identify the type of investors who care about corporate governance, the work by McCahery, Sautner and Starks (2016) is of great importance. The authors find that long-term investors and the ones who care less about liquidity intervene more intensely when it comes to governance issues. The research of McCahery et al. (2016) can also be interpreted as support for the holding period of at least one year, employed in the current piece of research.
As evidenced by the papers reviewed in the prior paragraphs, the empirical research on the effect of corporate governance on shareholder value is not conclusive. Moreover, a large amount of literature investigated the relationship in a manner that findings are not applicable as an investment strategy – e.g. they employ an event study approach.
D: Modifying Value Investing
7
current piece of research, because there is evidence of outperformance by strong corporate governance stocks based on accounting or operating metrics. If the findings hold in the current sample, a two-way sort based on value and corporate governance metrics may result in a strategy that could be described as ‘Quality at a Reasonable Price’ or ‘Growth at a Reasonable Price’. Double-sort portfolios are expected to show a simultaneous positive and statistically significant exposure to the Value and ‘Quality minus Junk’ factor to confirm this assumption, as the QMJ factor takes into account profitability.
Negative correlation has been showed between the value effect and the quality factor by Asness, Frazzini and Pedersen (2013), too. The authors refer to the potential of a quality and value based investment hybrid strategy to outperform a pure-value or pure-quality based investment strategies. The reason being – similarly to Novy-Marx (2013) – that both valuation and profitability are taken into account, rather than only one characteristic. Moreover, Asness et al. (2013) show that Quality stocks in general have lower betas, as investors appear to purchase these stocks in times of turmoil (‘flight to quality’ effect). In summary, the research by Asness et al. (2013) may suggest that a corporate governance (‘Quality’) and market-to-book (‘Value’) based sort may lead to an investor being able to avoid stocks commonly referred to as value traps. The strategy presented in the current piece of research is also evaluated against the
Quality minus Junk portfolio of Asness et al. (2013).
8
an investment based factor is incorporated into the well-known 3-factor model. The newer five factor model on the other hand ignores a number of well-documented factors that affect returns, such as momentum, low-beta and the low volatility – as pointed out by Blitz, Hanauer, Vidojevic and van Vliet (2016). Therefore, rather than using the five-factor model to analyse portfolio performance the ‘Quality
minus Junk’ factor of Asness et al. (2013) is preferred, as it aggregates a larger number of quality
attributes, including the ones previously mentioned.
E: Contribution to Existing Literature
In summary, this paper contributes to the existing literature both about the effect of corporate governance and about the modifications of value investing. There is evidence in prior literature that there is a positive relationship between firm performance and corporate governance – even if the superior performance does not result in superior stock returns. Secondly, empirical research suggests that investment strategies that combine valuation and profitability criteria have the potential to outperform the market. If strong corporate governance can indeed be considered a proxy for higher profitability, the investment strategy presented in this paper should have the potential to show improved performance over a purely profitability (corporate governance) and purely valuation (market-to-book) based strategies.
3. Data
A: Sample Selection
9
degree of clarity about the effects of corporate governance on stock returns. The detailed description of the
Quality (or Quality minus Junk, QMJ) factor is provided in the Methodology section of this paper.
The main limitation of the dataset is that it clearly has a large market capitalisation bias, as the constituents of the underlying index fall into that category of stocks. Moreover, due to the developed nature and strong regulatory and legal framework of North American capital markets, the possible returns premiums associated with corporate governance may be of limited applicability in developing markets. One has to bear this in mind if the findings are to be applied on other samples of companies. There is a relatively large amount of literature devoted to the analysis of the effect of financial development and legal systems on corporate valuation – such as La Porta et al. (2002) and McLean et al. (2012). As all the companies present in the sample used in this piece of research are from the same (United States) jurisdiction, this places a limit on the variability of the data used in the analysis. Moreover, as Value stocks tend to comprise smaller companies – see correlation between the Value (HML) and Size (SMB) factors – a large market capitalisation index may not be the ideal sample to test a value investing strategy or its modification.
B: Variable Definitions
The financial variables comprise share price, book value per share, free cash flow per share, total assets per share and common shareholders’ equity. In order to smoothen the effect of outliers, rather than taking total assets directly, their natural logarithms are taken – represented by the Size control variable. The value of common equity is used to compute the financial leverage ratio of the companies in the sample, using
Equation (1). The inclusion of the financial leverage ratio is vital, as previous research has showed – such
as Jensen (1986) – that increased leverage may lead to increased agency costs. Ashbaugh-Skaife et al. (2006) show that CEOs in firms with speculative credit ratings are more frequently overcompensated, or that strong governance structures have a positive impact on a firm’s credit ratings. Prior research also suggests that high leverage may suggest a breakdown of governance, therefore controlling for leverage is vital.
𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑡=
𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡
𝐶𝑜𝑚𝑚𝑜𝑛 𝐸𝑞𝑢𝑖𝑡𝑦𝑖,𝑡
(1)
10
their relationships with one another are evaluated. CGBS (Corporate Governance – Board Structure) represents the aggregate score achieved based on 58 sub-variables, where the diversity of the board of directors and the professional requirements to fulfil given positions is assessed. As opposed to CGBF, a higher emphasis is placed upon the qualifications of each board member, rather than the functioning of given corporate committees – such as: whether the CEO has held the position of Chairman before, disclosure about the mandates of each board member, among others. The CGCP (Corporate Governance – Compensation Policy) score consists of 46 sub-variables. It evaluates how and to what extent a number of financial and non-financial criteria are embedded into the remuneration guidelines of board members. CGVS (Corporate Governance – Vision and Strategy), aggregate score based on a further 39 variables, reflects upon the company’s ability to clearly communicate its ability to integrate financial and non-financial criteria into its decision-making processes. The final variable (CGSR) assesses the rights of shareholders of the company, and whether it ensures the equal treatment of minority shareholders. It consists of 77 sub-variables ranging from the analysis of anti-takeover provisions to shareholder rights related controversies covered by the media at the company. Prior research has found anti-takeover provisions to be of particular importance when investigating the effect of corporate governance on shareholder value. A full list of variables is provided in the Appendix.
C: Descriptive Statistics
Overall sample statistics are provided in Table 1. The total number of observations is 51,156 (42,305 in the case of Price-to-Free Cash Flow) over the 10-year period, with the amount of observations steadily increasing over the years of the sample period. Companies that do not have data available for a given variable have been excluded from the sample, thus the number of unique companies in the sample is lower than the number of index components. Moreover, companies that had negative shareholders’ equity values for a given time period have also been removed, likewise companies with negative Free Cash Flow – for the time periods the negative values were present. In order to mitigate the effect of outliers the top and bottom 1% percentiles of values in both Market-to-Book and Price-to-Free Cash Flow have been removed from the sample. Table 1 also lists descriptive statistics. Mean (median) values of the variables of interest range from 49.16 (42.64) for the variable Vision and Strategy and 80.02 (82.43) for Board Structure. The
Vision and Strategy score also shows the highest variability, as it has the highest standard deviation among
11
Table 2 provides insight into the correlation structures between the variables. Overall the correlation coefficient values between the corporate governance variables seem to be relatively low. Presumably, due to their similarities in nature, the correlation between the Board Functions and Board Structure variables in comparison with the other corporate governance related variables is high. The Vision and Strategy variable appears to have the strongest effect on the aggregated corporate governance score, with a coefficient value of 0.69. The relationship between Compensation Policy and the overall score seems to be the weakest (coefficient value of 0.35). When taking into consideration the control variables of Size and
Leverage, a negative relationship between leverage and corporate governance surfaces, albeit low
coefficient values, suggesting only a weak association. The relationship between leverage and firm performance or leverage and corporate governance has also been explored in prior research. Eisdorfer et al. (2013) find that the capital structure of executive compensation should track that of the entire firm, in order to ensure the optimization of decision-making. George et al. (2010) propose that the relationship between leverage and returns is negative, although low leverage can result in higher exposure to systematic risk. Moreover, they posit that operating performance deteriorates more significantly in the case of low leverage firms, in the event of financial distress. Company size appears to be positively correlated with corporate governance scores, on an aggregate basis, with the exception of Compensation
Policy. When considering the valuation ratios and corporate governance metrics, no clear pattern emerges.
Coefficient values alternate between positive and negative values and are low. The exception to this finding is the relationship between size and leverage, where a comparatively strong positive relationship presents itself (correlation coefficient value of 0.32).
Table 3 presents the descriptive statistics of factors used in this paper, to evaluate performance. For the period of interest the Market factor shows the highest premium, while Size being the lowest. The latter’s decreasing value or in some cases reversal is a relatively well-researched phenomenon - e.g. Dimson et al. (2002). Considering correlation coefficients, the high negative coefficient values of the ‘Quality minus
Junk’ factor with the Value, Market and Size factors are in line with its inherent characteristics as
12
Table 1 – Governance & Financial Characteristics
Mean Median Standard Deviation Max Min N
CGBF 79.67 82.41 10.72 93.30 5.06 51 156 CGBS 80.02 82.43 12.73 94.70 8.64 51 156 CGCP 69.02 73.68 17.08 93.65 3.87 51 156 CGVS 49.25 42.64 32.11 98.64 8.43 51 156 CGSR 65.23 68.23 25.39 98.92 1.02 51 156 CG Total 68.64 68.94 10.36 92.29 11.91 51 156 Size 16.54 16.42 1.37 21.67 12.49 51 156 Leverage 3.93 2.51 6.12 212.67 0.86 51 156 Market to Book 3.33 2.53 2.76 22.86 0.51 51 156 Price to FCF 28.48 9.53 33.63 347.05 2.41 42 305
The sample is based upon the S&P 500 equity index for the period beginning in 2005 to 2014, companies with negative equity values or free cash flow have been removed from the sample, for the periods with negative values. Leaving 488 unique companies in the sample. Values for Market-to-Book and Price-to-Free Cash Flow have been winsorised at 1%, in order to reduce the effect of outliers. The table also presents descriptive statistics of the variables used in the study. CGBF stands for Board Functions, CGBS for Board Structure, CGCP for Compensation
Policy, CGVS for Vision and Strategy and CGSR for Shareholder Rights. CG Total stands for the combined score
achieved on all previously described variables. Size stands for the natural logarithm value of total assets, leverage has been calculated as Total Assets divided by common equity. Price to FCF stands for Share Price divided by Free Cash Flow per share.
Table 2: Pairwise correlations – Governance & Financial Characteristics
CGBF CGBS CGCP CGVS CGSR CG Total Size Lev MTB P/FCF
CGBF 1 CGBS 0.30 1 CGCP 0.03 0.06 1 CGVS 0.11 0.09 -0.01 1 CGSR 0.06 0.06 0.01 0.07 1 CG Total 0.39 0.41 0.35 0.69 0.56 1 Size 0.18 0.06 -0.19 0.37 -0.05 0.19 1 Lev 0.06 -0.02 -0.06 0.01 -0.07 -0.05 0.32 1 MTB -0.10 0.01 0.01 -0.06 0.06 -0.03 -0.32 0.10 1 P/FCF -0.08 -0.02 0.03 0.02 -0.01 -0.01 -0.08 -0.01 -0.01 1
Table 2 presents the pairwise correlation matrix between of the variables. Lev and MTB stand for Leverage and
Market to Book, P/FCF stands for Price-to-Free Cash Flow. Due to sample restrictions, the number of observations
13
Table 3: Factor Characteristics
Mean St. Dev. Min Max Sharpe Ratio MKT-𝑅𝑓 HML SMB QMJ
MKT- 𝑅𝑓 0.63 4.34 -17.23 11.35 0.14 1
HML 0.20 2.54 -7.27 11.25 0.08 0.20 1
SMB 0.10 2.21 -4.29 6.11 0.05 0.41 0.12 1
QMJ 0.27 2.71 -7.30 9.08 0.10 -0.74 -0.44 -0.39 1
Table 3 presents the descriptive statistics used in the study. The values in the first four columns are in percentages. Pairwise correlation coefficients are presented in the second four columns. MKT-𝑅𝑓, HML and SMB refer to the standard Fama & French factors, QMJ stands for the ‘Quality minus Junk’ factor of Asness et al. (2013).
4. Methodology
A: Outline of the Study
The goals of this study are two-fold. Firstly, the presumed outperformance of stocks that represent companies with high corporate governance scores is tested - determining whether a portfolio consisting of stocks with superior corporate governance structures is able to outperform the portfolio of stocks with poor corporate governance, essentially aiming to replicate the findings of Gompers et al. (2003). Secondly, corporate governance metrics areconsidered in a two-way sort of the sample, applying them as an additional screen when constructing a portfolio of Value stocks. Here, corporate governance is considered as a means to achieve incremental improvement over a well-documented strategy.
Prior literature documents adjustments to Value investing in a number of ways. Ball et al. (2015) build a model based upon free cash flow rather than market-to-book. This method is going to be used as a robustness check on the findings. Novy-Marx (2013) employs gross profitability as a selection criterion, in conjunction with a value investing strategy. However, Ball et al. (2015) find that a strategy’s Sharpe ratio can be increased by with the addition of a cash flow based operating profitability factor, rather than an accruals-based profitability factor. As certain forms of shareholder value destroying behaviour, such as proneness to engage in overly ambitious corporate acquisitions, as in Fu et al. (2013), it would be reasonable to assume that an investing strategy that excludes companies performing poorly based on corporate governance from a set of already under-priced assets would enable an investor to achieve superior risk adjusted returns.
B: Cross-sectional Tests of Valuation and Governance
14
added into the model. For the sample period between 2005 and 2014. Monthly Market-to-Book values are calculated (2) using the stock prices and per share book values from Datastream.
𝑀𝑇𝐵𝑖,𝑡 = 𝛼𝑖+ 𝛽1𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒 + 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜀𝑖,𝑡 (2)
Where MTB stands for Market-to-book ratio, 𝛽1 represents the coefficient value attributable to an
aggregate score achieved on all 5 corporate governance variables. 𝛽2 refers to the control variables of the
regression.
As a robustness check, an equation testing the relationship between corporate governance and Price to Free Cash Flow is estimated, the free cash flow measure for the purposes of this study is different from the one used in Ball et al. (2015). The authors of that paper emphasised the difference between accruals as an accounting concept and cash flow – in order to explore the viability of the operating profitability factor of Novy-Marx (2013) - thus their method for determining free cash flow differed. For the purposes of this paper a simplified version of Free Cash Flow is used.
𝑃𝐹𝐶𝐹𝑖,𝑡 = 𝛼𝑖+ 𝛽1𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝐺𝑜𝑣𝑒𝑟𝑛𝑎𝑛𝑐𝑒 + 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜀𝑖,𝑡 (3)
C: Governance and Valuation – Investment Considerations
two-15
way sort portfolios are constructed, based on quartile scores. In addition, quartile portfolios based solely on Market-to-Book, Price-to-Free Cash Flow and Corporate Governance are constructed to provide context for the characteristics of the two-way sort. The performance attribution analysis of the portfolios is carried out with the help of the following factor equations – the standard 3-factor model of Fama & French and with the addition of the Quality factor of Asness et al. (2013):
𝑅𝑝− 𝑅𝑓 = 𝛼 + 𝛽1𝑀𝐾𝑇 + 𝛽2𝐻𝑀𝐿 + 𝛽3𝑆𝑀𝐵 + 𝜀𝑖 (4)
And with the addition of the ‘Quality minus Junk’ factor:
𝑅𝑝− 𝑅𝑓 = 𝛼 + 𝛽1𝑀𝐾𝑇 + 𝛽2𝐻𝑀𝐿 + 𝛽3𝑆𝑀𝐵 + 𝛽4𝑄𝑀𝐽 + 𝜀𝑖 (5)
The excess returns on the portfolios is evaluated against the returns on the Quality minus Junk factor portfolio of Asness et al. (2013). This factor is constructed in the following manner: the aggregate quality score is the combination of the sub-scores in Profitability, Growth, Safety and Payout. Due to its comprehensive nature, this comparison allows for the better identification of characteristics of companies in each portfolio, without obtaining detailed accounting data for each company in the sample
In turn these sub-scores represent an average score attained consisting of:
- Gross profits over assets, return on equity, returns on assets, cash flow over assets and gross margin for Profitability.
- Five-year prior growth of the previous metrics for Growth.
- Low beta (BAB), low idiosyncratic volatility, low leverage, low bankruptcy risk (determined by O and Z-scores) and low ROE volatility for Safety.
- Net equity and debt issuance and total net payout above profits for Payout.
16
5.
Findings
A: Effect of Corporate Governance on Valuation
In order to avoid concerns of multicollinearity in the table presented in the main paper, the corporate governance sub-scores are regressed on the dependent variable – market-to-book and price-to-free cash flow, respectively – separately. A model including all variables at once and various other specifications is presented in the Appendix.
Table 4: Valuation and Corporate Governance
Dependent Variable: Market-to-Book
Intercept 15.322*** 16.512*** 15.572*** 17.038*** 15.497*** 16.841*** (102.816) (107.074) (100.330) (100.630) (106.209) (114.178) CG-Total 0.016*** (14.685) CGBF -0.010*** (-8.925) CGBS 0.006*** (6.613) CGCP -0.009*** (-14.462) CGSR 0.006*** (14.608) CGVS 0.008*** (21.134) Size -0.819*** -0.775*** -0.793*** -0.813*** -0.786*** -0.867*** (-91.748) (-87.624) (-90.830) (-91.854) (-90.339) (-92.030) Leverage 0.107*** 0.103*** 0.104*** 0.103*** 0.105*** 0.109*** (54.529) (53.069) (53.291) (53.180) (53.953) (55.615) R-squared 0.151 0.149 0.148 0.151 0.151 0.155 N 51,156 51,156 51,156 51,156 51,156 51,156
Table 4 presents the OLS regression results of Corporate Governance variables on Market-to-Book. SUMCG refers to the aggregate score, CGBF to Board Functions, CGBS to Board Structure, CGCP to Compensation Policy, CGSR to Shareholder Rights, CGVS to Vision and Strategy. Size refers to the natural logarithm of total assets. The variables are described in more detail in the Sample section of this paper. *, **, *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
17
is no evidence of one particular aspect of corporate governance driving the effect on valuation, thus, in the two-way sort the best method moving forward appears to be the usage of the aggregate score. The control variables are highly statistically significant, in the case of Size (log total assets), the effect is highly significant economically too – values ranging from -0.78 to -0.87, with mean (median) values of 16.54 (16.42) for this variable. When inspecting R-squared, values are consistent over different specifications at about 15%. Interestingly, Shareholder Rights do not have a substantial effect on valuation, although previous research often finds a strong relationship between valuation and shareholder rights (especially takeover restrictions). Cunat et al. (2012) in their research the abolition of takeover provisions shows significant compounding effect, as more anti-takeover provisions are eliminated, arguing that in essence this excess return is a manifestation of the takeover premium.
Table 5: Valuation and Corporate Governance (cont.)
Dependent Variable: Price-to-Free Cash Flow
Intercept 60.338*** 73.453*** 64.533*** 59.768*** 62.816*** 65.925*** (29.504) (35.444) (30.636) (27.491) (31.772) (33.895) SUMCG 0.025 (1.590) CGBF -0.214*** (-14.150) CGBS -0.042*** (-3.268) CGCP -0.016* (1.680) CGSR -0.014** (-2.294) CGVS 0.054*** (10.099) Size -2.033*** -1.689*** -1.979*** -1.962*** -2.020*** -2.429*** (-17.484) (-14.498) (-17.234) (-16.757) (-17.580) (-19.902) Leverage 0.005 0.006 0.004 0.005 0.004 0.006 (0.670) (0.813) (0.562) (0.641) (0.580) (0.789) R-squared 0.007 0.012 0.007 0.007 0.007 0.009 N 42,305 42,305 42,305 42,305 42,305 42,305
Table 5 presents the OLS regression results of Corporate Governance scores on Price-to-Free Cash Flow, with the addition of Size – natural logarithm of total assets and Leverage as control variables. SUMCG refers to the aggregate score, CGBF to Board Functions, CGBS to Board Structure, CGCP to Compensation Policy, CGSR to Shareholder
Rights, CGVS to Vision and Strategy. T-statistics are in parentheses. *, **, *** denote statistical significance at the
18
The effect of corporate governance on valuation when regressing the scores on Price-to-Free Cash Flow (presented in Table 4) appears to be more varied when comparing the results with that of Market-to-Book. The aggregate corporate governance score does not have a statistically significant effect on Price-to-Free Cash Flow on any conventional level of significance. Board Functions shows the highest coefficient value, -0.214, from the sub-scores, its effect is highly statistically significant and based on the coefficient value it exerts a negative effect on Price-to-Free Cash Flow based valuations. Considering the mean (median) values of P/FCF at 28.48 (9.53) - the economic effect is still meagre. Overall, higher corporate governance scores do not appear to translate into higher Price-to-Free Cash Flow based valuations. The statistically significant corporate governance scores are mostly negative. It is worth pointing out, that in the work of Cunat et al. (2012) the proposals relating to corporate board functions and structures were showed to have the most profound effect on abnormal returns. In this current piece of research the relationship may be considered to be present, in the context of valuation, rather than returns – albeit in a different direction. In another study Bhagat et al. (2008) find that focus on board independence may be misleading, as it has a negative effect on subsequent operating performance. From the control variables,
Size is statistically and economically highly significant in every specification – it has a strong negative
effect on free cash flow based valuations, as was the case with market-to-book based ones.
19
B: Portfolio Performance
Table 6: Portfolio Returns and Sharpe Ratios
Panel A: One-way sort Portfolios
High / Strong 2 3 Low / Weak
MTB 0.88 (0.12) 0.59 (0.09) 0.89 (0.18) 0.73 (0.13)
PFCF 1.58 (0.19) 0.97 (0.17) 0.83 (0.16) 0.44 (0.09)
CG 0.56 (0.12) 0.67 (0.12) 0.86 (0.12) 0.92 (0.15)
Panel B: Corporate Governance and Value
High 2 3 Low
Strong 1.2 (0.14) 0.69 (0.12) 0.96 (0.20) 0.55 (0.11)
2 0.8 (0.10) 0.63 (0.11) 0.79 (0.13) 0.56 (0.10)
3 0.83 (0.10) 0.64 (0.06) 0.96 (0.17) 0.96 (0.15)
Weak 0.94 (0.14) 0.67 (0.11) 0.9 (0.16) 1.79 (0.29)
Panel C: Corporate Governance and Price-to-Free Cash Flow
High 2 3 Low
Strong 1.1 (0.20) 1.01 (0.17) 0.75 (0.15) 0.81 (0.16)
2 1.16 (0.14) 0.07 (0.01) 0.94 (0.16) 0.26 (0.05)
3 1.38 (0.14) 1 (0.13) 0.93 (0.17) 0.39 (0.07)
Weak 1.27 (0.18) 1.04 (0.18) 0.95 (0.16) 0.49 (0.08)
Table 6 shows monthly average excess returns (in percentages) of quartile portfolios, with the holding period of 12 months, with non-annualised Sharpe ratios in parentheses. Panel A contains one-way sort portfolios formed on Market-to-Book, to-Free Cash Flow and Corporate Governance, from High to Low Market-to-Book and Price-to-Free Cash Flow, and from Strong to Weak Corporate Governance. Panel B shows the monthly average excess returns on four by four two-way sort portfolios, formed on Market-to-Book (from High to Low) and the aggregate Corporate Governance score (from Strong to Weak). Panel C contains monthly average excess returns of portfolios formed on Price-to-Free Cash Flow (from High to Low) and the aggregate Corporate Governance score.
20
the authors document that the cash-flow based profitability factor subsumes the effect of the conventional
Value factor.
Considering the two-way sort, the results are to a large extent in contradiction with expectations. The portfolio from the bottom right corner – highest valuation as defined by market-to-book and weakest governance, simultaneously – shows the best performance. Both in terms of excess returns, 1.79% per month, and a Sharpe ratio of 0.29. When comparing the results of the market-to-book sort with the two-way sort, it can be observed that the main portfolio of interest in the two-two-way sort (Strong Governance & High Market-to-Book) does perform better than the High Market-to-Book portfolio. The portfolio from the two-way sort shows a monthly average excess return of 1.2, with a non-annualised Sharpe ratio of 0.14 – compared with 0.88 and 0.12 for the High Market-to-Book portfolio. The difference between Sharpe ratios of these two portfolios is not statistically significant though. When defining the returns attributable to value investing as the HML portfolio of Fama & French, the outperformance is not significant either. On the other hand, the Strong Governance & High Market-to-Book outperforms the market portfolio in a statistically highly significant manner. The market portfolio being defined as the Fama & French Market factor portfolio and the S&P 500 index for the period of interest. The main portfolio of interest from the two-way sort also outperforms the Quality minus Junk portfolio of Asness et al. (2013) for the period of interest. This is in line with the findings of Novy-Marx (2013), where the author finds that a strategy combining valuation and quality metrics outperforms a purely quality or profitability-based strategy.
Panel C of Table 6 presents the returns and Sharpe ratios of the Governance and Price-to-Free Cash Flow based two-way sort. The main portfolio of interest – Strong Governance & High Price-to-Free Cash Flow – shows a monthly average excess return of 1.1 percent, with a Sharpe ratio of 0.20. This portfolio has the highest Sharpe ratio from the entire price-to-free cash flow and governance based two-way sort. Hinting that the inclusion of the governance criterion may be feasible when paired with price-to-free cash flow, rather than market-to-book. The Sharpe ratio of the Strong Governance & High Price-to-Free Cash Flow is statistically significantly higher than that of the Strong Governance and & High Market-to-Book portfolio. The Strong Governance & High Price-to-Free Cash Flow portfolio outperforms both market portfolios – Market factor & S&P 500 index – and the Quality minus Junk portfolio in a statistically significant way. Although, it fails to outperform purely value based strategies, namely the High Market-to-Book portfolio, High Price-to-Free Cash Flow and the HML factor portfolio in statistically significant manner. As Price-to-Free Cash Flow based portfolio is able to outperform the market and the Quality
minus Junk portfolio, the viability of the strategy that combines valuation with quality or profitability
21
C: Alphas and Factor Exposures
Table 7 presents factor exposures of one-way sort portfolios – formed on Market-to-Book and the aggregate Corporate Governance score - based on the Fama & French 3-factor model and with the addition of the ‘Quality minus Junk’ factor, as defined by Asness et al. (2013). Alphas are indistinguishable from zero in every specification. The exposure to the Market factor is highly statistically significant in the case of every portfolio. Although, the exposure to the market factor does not automatically translate into higher expected return on the portfolio – e.g. in the case of the book-to-market sort, the highest expected return is achieved by the 3rd quartile portfolio, in the case of the Corporate Governance sort, the highest expected return is showed by the ‘Weak’ governance portfolio. The insignificance of the SMB factor can either be caused by the large market capitalization basis of the sample or another well-researched phenomenon – e.g. Cochrane (1999) - the disappearance or the reversal of the Size premium. The exposures to the Quality factor of the Corporate Governance portfolios, lend support to the notion put forward in prior literature that strong corporate governance leads to superior operating performance, but not correlated with future stock market performance – see Core et al. (2006) and Bhagat et al. (2008). The overall expected return of the strong corporate governance portfolio is lower than that of the weak corporate governance portfolio. Although the coefficient values on the Quality factor – positive and statistically significant at the 5% level in the case of Strong governance, negative and statistically significant at the 5% level in the 3rd quartile – point towards a strong and positive relationship between Quality and strong Corporate Governance.
22
market-to-book. Unfortunately, the QMJ factor does not allow for a closer identification of which aspect of Quality is driving the relationship. The lower exposure to the market factor of ‘Strong’ governance portfolios, hints that Beta may exert a strong influence. Intuitively, ‘High’ market-to-book portfolios show positive and highly statistically significant exposure to the HML factor. Exposure to the Size factor is insignificant apart from two cases when it is negative and statistically significant - with coefficient values of -0.574 for the Strong Governance & High Market-to-Book portfolio,10 percent level of significance, and in the 3rd Governance quartile of the 2nd Valuation quartile with the value of -0.755, significant at the 5 percent level.
The absence of outperformance based on corporate governance as the only selection criterion is in line with recent prior literature – Core et al (2003), Cremers et al. (2005). The returns documented by Gompers et al. (2003) have largely shown to be, relating to a larger set of valuation anomalies present during the sample period. The results show that unlike accounting information as shown by Frankel, Lee (1998) or Piotroski (2000), corporate governance may not be very effective in improving value investing, as the strong governance portfolios – from the two-way sorts – fail to generate alpha. However, the Strong Governance & Low Valuation portfolios even with the absence of alpha are able to outperform the QMJ portfolio and the market in a statistically significant way.
Table 7: One-way Sort Portfolio Performance Attribution
α MKT- 𝑅𝑓 HML SMB QMJ R-squared
Panel A: Fama & French 3-factor Model
High -0.001 1.302*** 0.664*** -0.101 0.738 2 -0.002 1.225*** 0.277** -0.279 0.722 3 0.003 0.952*** 0.102 0.02 0.737 Low 0.001 0.961*** -0.107 0.047 0.561 Strong 0.000 0.961*** 0.046 -0.192 0.788 2 0.000 1.110*** 0.107 -0.038 0.749 3 0.000 1.303*** 0.411*** -0.071 0.688 Weak 0.002 1.158*** 0.172 0.081 0.734
Panel B: Quality minus Junk Factor
High 0.002 1.152*** 0.545** -0.142 -0.374* 0.743 2 -0.001 1.149*** 0.217 -0.300 -0.189 0.722 3 0.002 1.016*** 0.153* 0.037 0.159 0.737 Low 0.000 1.054*** -0.033 0.072 0.233 0.562 Strong -0.002 1.04*** 0.109 -0.170 0.198** 0.791 2 0.000 1.110*** 0.107 -0.038 0.000 0.747 3 0.001 1.179*** 0.312** -0.104 -0.308** 0.691 Weak 0.002 1.113*** 0.136 0.069 -0.113 0.733
23
Table 8: Two-way Sort Portfolio Performance Attribution
MTB CG α MKT- 𝑅𝑓 HML SMB QMJ R-squared
Panel A: Fama & French 3-factor Model
High Strong 0.002 1.433*** 0.578** -0.574* 0.543 2 -0.002 1.288*** 0.691*** 0.07 0.663 3 -0.002 1.501*** 0.793** -0.29 0.663 Weak 0.000 1.229*** 0.672*** -0.072 0.739 2 Strong 0.000 1.085*** 0.335*** -0.186 0.718 2 0.000 1.103*** 0.04 -0.246 0.601 3 -0.005 1.831*** 0.353 -0.662* 0.56 Weak -0.001 1.175*** 0.244** 0.047 0.696 3 Strong 0.004 0.867*** 0.079 -0.035 0.583 2 0.001 1.079*** 0.151 0.118 0.66 3 0.002 0.993*** 0.279* 0.198 0.723 Weak 0.002 1.014*** 0.074 0.174 0.635 Low Strong 0.000 0.836*** -0.055 -0.163 0.521 2 -0.001 1.031*** -0.171 -0.007 0.647 3 0.003 0.971*** -0.054 0.416 0.499 Weak 0.013*** 0.663*** -0.003 0.368 0.276
Panel B: Quality minus Junk Factor
High Strong 0.002 1.434*** 0.578** -0.574* 0.001 0.539 2 0.0001 1.136*** 0.570** 0.029 -0.378 0.667 3 -0.001 1.131*** 0.637 -0.343 -0.485 0.669 Weak 0.003 1.067*** 0.543 -0.116 -0.404* 0.746 2 Strong -0.001 1.109*** 0.354*** -0.179 0.061 0.716 2 0.001 1.018*** -0.028 -0.269 -0.212 0.602 3 0.000 1.490*** 0.081 -0.755** -0.849*** 0.575 Weak 0.000 1.074*** 0.163 0.02 -0.25 0.697 3 Strong 0.002 0.998*** 0.183* 0.001 0.327*** 0.591 2 0.001 1.09*** 0.159 0.121 0.027 0.657 3 0.003 0.939*** 0.236 0.183 -0.138 0.722 Weak 0.003 0.993*** 0.057 0.168 -0.053 0.632 Low Strong -0.002 0.984*** 0.063 -0.123 0.368*** 0.533 2 -0.002 1.121*** -0.099 0.0175 0.224 0.648 3 0.002 1.031*** -0.006 0.432 0.150 0.496 Weak 0.013** 0.702*** 0.028 0.379 0.098 0.27
24
6. Alternative Specifications
The following section presents the factor exposure analysis of the portfolios formed on Price-to-Free Cash Flow and Corporate Governance. Moreover, the portfolios characteristics are analysed using a 24-month holding period, instead of the 12-month one used in the base model. The adjustment to the portfolio formation, whereby the top (bottom) quartile portfolios will take a short position in the bottom (top) quartile of the sample is also examined.
A: Price-to-Free Cash Flow
Table 9 shows the portfolios factor exposures of the Price-to-Free Cash Flow based one-way sort portfolios. As with the portfolios in the base model variant of the one-way sort, alphas estimated are overwhelmingly statistically insignificant - except for the ‘High P/FCF’ portfolio, where the alpha is statistically marginally significant, translating into a monthly abnormal return of 0.9%. Exposure to the
Market factor is positive and highly statistically significant in every case. In the case of the 3-factor model
there is no disconnect between the Market factor exposure and expected return, as the ‘High P/FCF’ portfolio shows the highest expected return in that category of the one-way sort. Although, the misalignment between the beta and expected return emerges when adding in the ‘Quality minus Junk’ factor. Interestingly the portfolio consisting of stocks in the 3rd quartile based on Price-to-Free Cash Flow shows positive and statistically significant exposure to the Value and Quality factors – as Asness et al. (2013) posit that these two factors ‘pull’ in a different direction, the results point towards a unique characteristic of the Price-to-Free Cash Flow based sort.
Table 9: Performance Attribution - Price-to-Free Cash Flow
P/FCF α MKT- 𝑅𝑓 HML SMB QMJ R-squared
Panel A: Fama & French 3-factor Model
High 0.006 1.319*** 0.800** -0.224 0.552
2 0.003 0.982*** 0.175 0.283 0.627
3 0.002 1.049*** 0.087 -0.082 0.785
Low -0.001 0.973*** -0.126 -0.095 0.729
Panel B: Quality Factor
High 0.009* 1.154*** 0.669** -0.269 -0.41 0.555
2 0.002 1.018*** 0.203* 0.293 0.089 0.624
3 0.000 1.165*** 0.180** -0.051 0.289*** 0.792
Low -0.002 1.022*** -0.087 -0.082 0.122 0.729
25
Table 10 shows the alphas and factor exposure estimates of the Price-to-Free Cash Flow and Corporate Governance based two-way sort. Alphas are insignificant in every specification. The exposure to the
Market factor of the Fama & French model is positive ad highly statistically significant in every
specification. When considering coefficient values in most cases ‘Strong’ corporate governance – especially on the two extreme ends of the Price-to-Free Cash Flow & Corporate Governance two-way sort- portfolios show lower exposure to this factor. Supporting the propositions that the returns achieved by these portfolios –may be related to the low Beta anomaly- as defined by Frazzini et al. (2014) and Auer et al. (2015) -, similarly to the findings in the main model. This connection would warrant further analysis, as the construction of the ‘Quality minus Junk’ factor considers this anomaly – the positive and highly significant exposure to this factor of portfolios on the stronger end of corporate governance may give support to this claim, but to reach a definitive conclusion a different model would have to be constructed, isolating the effect of low Betas more specifically. The results shown in Table 10 are also in line with prior findings concerning superior operating performance by strong corporate governance companies, as coefficient estimates of the ‘Quality minus Junk’ factor is positive and highly statistically significant in the case of ‘Strong’ governance portfolios.
26
Table 10: Performance Attribution – Two-way Sort P/FCF & CG
P/FCF CG α MKT- 𝑅𝑓 HML SMB QMJ R-squared
Panel A: Fama & French 3-factor Model
High Strong 0.005 0.967*** 0.211 -0.062 0.573 2 0.002 1.317*** 0.734*** 0.108 0.614 3 0.003 1.541*** 1.029*** -0.463 0.523 Weak 0.004 1.236*** 0.469* -0.024 0.651 2 Strong 0.003 0.967*** 0.19 0.188 0.581 2 -0.006 1.036*** 0.111 0.167 0.644 3 0.003 0.891*** 0.199 0.649** 0.387 Weak 0.004 1.022*** 0.132 0.137 0.622 3 Strong 0.001 1.042*** 0.12 -0.217 0.737 2 0.003 0.993*** 0.143 0.208 0.637 3 0.003 1.026*** 0.092 0.082 0.702 Weak 0.002 1.149*** -0.031 0.073 0.673 Low Strong 0.003 0.783*** 0.011 -0.065 0.428 2 -0.003 0.998*** -0.293** -0.147 0.672 3 -0.003 1.161*** -0.105 0.042 0.715 Weak -0.002 1.165*** -0.141 0.139 0.670
Panel B: Quality Factor
High Strong 0.003 1.098*** 0.316** -0.026 0.327** 0.579 2 0.004 1.134*** 0.588** 0.058 -0.456 0.619 3 0.006 1.278*** 0.82* -0.534 -0.655** 0.531 Weak 0.005 1.146*** 0.398 -0.048 -0.225 0.650 2 Strong 0.002 1.043*** 0.251** 0.209 0.19 0.580 2 -0.005 0.968*** 0.057 0.149 -0.168 0.643 3 0.003 0.936*** 0.236 0.662** 0.114 0.382 Weak 0.004 1.005*** 0.119 0.132 -0.041 0.619 3 Strong -0.001 1.162*** 0.215** -0.184 0.298 0.744 2 0.001 1.121*** 0.245 0.243 0.320** 0.642 3 0.001 1.118*** 0.165 0.107 0.23 0.704 Weak 0.001 1.212*** 0.019 0.09 0.156 0.672 Low Strong 0.002 0.896*** 0.100 -0.035 0.280** 0.432 2 -0.004 1.075*** -0.231 -0.126 0.193 0.673 3 -0.003 1.156*** -0.11 0.04 -0.014 0.712 Weak -0.002 1.133*** -0.167 0.13 -0.08 0.668
27
B: Two-year Holding Period
Table 11: 𝑬𝑹 & Sharpe Ratios for 24 months Panel A: One-way sort
High/Strong 2 3 Low/Weak
MTB 0.71 (0.10) 0.64 (0.10) 0.89 (0.18) 0.66 (0.14)
CG 0.72 (0.18) 0.55 (0.10) 1.12 (0.18) 1.24 (0.20)
Panel B: Two-way sort
High 2 3 Low
Strong 0.40 (0.04) 0.39 (0.07) 0.78 (0.18) 0.55 (0.12)
2 0.99 (0.13) 0.70 (0.11) 0.82 (0.16) 0.45 (0.09)
3 0.76 (0.08) 0.98 (0.11) 1.10 (0.17) 0.80 (0.13)
Weak 1.33 (0.18) 1.03 (0.15) 1.08 (0.15) 0.82 (0.08)
Table 11 shows average monthly excess returns on the 24-month holding period portfolios, with non-annualised Sharpe ratios in parentheses. Panel A presents returns on one-way sort portfolios sorted on Market-to-Book (from High to Low) and Corporate Governance (from Strong to Weak). Panel B presents the average monthly returns of two-way sort portfolios.
As presented in Table 11 the 24-month holding period average returns do not differ in pattern in comparison with the base model in the case of the one-way sorts. Value performs poorly during the sample period, and firms in the bottom quartile based on the aggregate Corporate Governance score outperform. A major difference between the findings of the base specification and the 24-month holding period is that in the two-way sort portfolios in the ‘Strong’ Corporate Governance quartile underperform the one-way sort based on Market-to-Book. Therefore, corporate governance is not shown to be an appropriate means to improve a Value investing strategy on a longer time frame. Overall, in the two-way sort the ‘High Book-to-Market & Weak Governance’ portfolio offers the best reward for risk – monthly average expected return of 1.33% and a Sharpe ratio of 0.18.
28
Table 12: Two-way sort 24-month Holding Period
MTB CG α MKT- 𝑅𝑓 HML SMB QMJ R-squared
Panel A: Fama & French 3-factor Model
High Strong -0.007 1.813*** 0.449 -0.873*** 0.634 2 0.000 1.389*** 0.366 0.136 0.708 3 -0.004 1.558*** 0.992** -0.248 0.670 Weak 0.005 1.026*** 0.556*** 0.282 0.476 2 Strong -0.004* 1.186*** 0.380** -0.303*** 0.783 2 0.000 1.103*** 0.335* -0.324* 0.608 3 0.000 1.445*** 0.629*** -0.332 0.516 Weak 0.002 1.160*** 0.266* 0.226 0.579 3 Strong 0.003* 0.827*** 0.088 -0.255** 0.641 2 0.002 0.985*** 0.163 0.091 0.706 3 0.004 1.053*** 0.236 0.382 0.610 Weak 0.004 0.928*** 0.259 0.355 0.405 Low Strong 0.000 0.878*** -0.125 -0.25 0.582 2 -0.001 0.962*** -0.098 -0.086 0.686 3 0.001 1.128*** -0.103 0.105 0.692 Weak -0.002 1.537*** 0.193 0.330 0.474
Panel B: Quality Factor
High Strong -0.006 1.737*** 0.388 -0.893*** -0.190 0.632 2 0.003 1.177*** 0.198 0.079 -0.526 0.719 3 0.000 1.305*** 0.790* -0.316 -0.629* 0.680 Weak 0.009 0.791*** 0.369 0.219 -0.585 0.489 2 Strong -0.004* 1.198*** 0.390*** -0.300*** 0.029 0.781 2 0.001 1.028*** 0.275 -0.344* -0.187 0.607 3 0.003 1.254*** 0.477 -0.384* -0.476* 0.520 Weak 0.004 1.014*** 0.15 0.186 -0.363 0.583 3 Strong 0.002 0.850*** 0.107 -0.249** 0.059 0.638 2 0.001 0.994*** 0.171* 0.093 0.023 0.703 3 0.003 1.086*** 0.263* 0.391 0.084 0.607 Weak 0.004 0.958*** 0.283 0.363 0.072 0.400 Low Strong -0.002 1.045*** 0.008 -0.205 0.416*** 0.600 2 -0.002 1.017*** -0.054 -0.071 0.138 0.685 3 0.000 1.166*** -0.073 0.116 0.094 0.690 Weak 0.000 1.409*** 0.092 0.295 -0.318 0.472
29
C: Long-Short Portfolios
Table 13: Long-Short Portfolios
α MKT - 𝑅𝑓 HML SMB QMJ R-squared 𝐸𝑟 Sharpe-ratio
Panel A: Fama & French 3-factor Model
MTB -0.002 0.341* 0.771*** -0.147 0.276 0.15 0.03
CG -0.002 -0.197*** -0.126 -0.273* 0.199 0.37 -0.13
PFCF 0.008* 0.346 0.926*** -0.129 0.244 1.14 0.20
CGMTB -0.011 0.771** 0.581** -0.942*** 0.212 -0.59 -0.08
CGPFCF 0.007* -0.198* 0.352* -0.200 0.074 0.61 0.15
Panel B: Quality factor
MTB 0.002 0.098 0.577* -0.213 -0.606*** 0.313
CG -0.004 -0.072 -0.026 -0.239 0.31** 0.224
PFCF 0.011** 0.132 0.756** -0.187 -0.532** 0.260 CGMTB -0.01 0.732** 0.550** -0.953*** -0.097 0.205
CGPFCF 0.004 -0.035 0.483*** -0.156 0.407* 0.093
Table 13 presents the characteristics of one and two-way sort portfolios, formed by holding a long position in stocks with the highest valuation ratio (and/or strongest governance) and holding a short position on the most overvalued (and/or weakest governance) stocks, for a holding period of 12 months. Panel A presents factor exposures based on the Fama & French 3-factor model and portfolio expected monthly average expected excess returns and Sharpe ratios. Panel B adds the ‘Quality minus Junk’ of Asness, Frazzini (2013) to the model. For the purposes of the regression Newey-West heteroscedasticity and autocorrelation robust standard errors have been used. *, **, *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
30
7. Conclusion
The paper analyses the relationship between corporate governance, valuation and equity returns. Moreover, an investment strategy is introduced combining valuation criteria and corporate governance metrics. To investigate the nature of the relationship between corporate governance and valuation firstly, a panel regression is conducted. The results show that governance metrics exert a statistically highly significant effect on valuation, but the economic significance of the effect is meagre. The portfolio sorts provide mixed evidence about the viability of the corporate governance criterion.
To test the investment strategy, the sample is divided into quartiles and a two-way sort based on market-to-book and corporate governance is carried out. In the main model the portfolios are constructed on a long only basis, stocks are held for a period of 12 months, with rebalancing taking place on the 1st of January each year. This results in 120 return observations per portfolio. As robustness checks, sorts based on Price-to-Free Cash Flow, long-short portfolio formation and a 24-month holding period are tested. The performance of the portfolios is evaluated using the Fama & French 3-factor model, and the same model with the addition of the Quality minus Junk factor of Asness et al. (2013).
Contrary to expectations, the two-way sort portfolio of stocks in the highest valuation quartile and with the weakest governance scores achieves a statistically and economically highly significant alpha of approximately 1.3%. The factor analysis of the portfolios also shows that in line with prior literature, corporate governance can indeed be considered a proxy for superior operating characteristics, as portfolios that consist of strong governance stocks show positive and statistically significant exposure to the ‘Quality
minus Junk’ factor in several instances. As the construction of the QMJ factor takes into account a number
of risk proxies – e.g. beta and volatility – the findings support the notion that corporate governance is also a useful proxy for risk. For this reason, some portfolios of strong governance stocks show a lower exposure to the Market factor, providing further evidence about the risk proxy aspect of governance. It should be stressed, that these findings only hold in the case of the main model, as in most alternative specifications the picture becomes more blurred.
31
Prior to passing judgement on the governance and valuation based strategy’s inability to outperform conventional value investing, sample limitations have to be considered. Value investing is mostly associated with smaller stocks, and the S&P 500 is a large market capitalisation index. Therefore, it is not the ideal sample to test strategies that are derived from value investing.
32
Appendix
Table A1: Return Distribution
Strong CG & High MTB Strong CG & High P/FCF HML QMJ S&P 500 MKT-𝑅𝑓
Skewness -0.17 -0.21 0.89 0.30 -0.72 -0.77
Kurtosis 14.32 1.71 3.22 1.71 2.31 2.01
Table A1 presents the characteristics of the distribution of the returns series of the main portfolios of interest. ‘Strong CG & High MTB’ and ‘Strong CG & High P/PFCF’ refer to the portfolios with the 12-month holding period, falling to the strongest governance quartile and the lowest valuation quartile simultaneously. Characteristics of selected factors are also included. HML refers to the Value factor of the Fama & French model, QMJ denotes the ‘Quality
minus Junk’ Asness and Frazzini (2013). S&P 500 refers to the characteristics of the ‘Total Return Index’ of the
eponymous equity benchmark for the period of 2005-2014 and MKT-𝑅𝑓refers to the Fama & French ‘Market’ factor. The statistical significance of the Sharpe ratios is tested with the help of Welch’s T-test, using the following formula:
𝑡 =
𝑋̅̅̅̅−𝑋1 ̅̅̅̅2 √𝑠12 𝑁1+ 𝑠22 𝑁2 (A1)Where 𝑋̅̅̅, 𝑠1 12 and 𝑁1 denote the first sample mean, variance and the number of observations, respectively.
When testing the Sharpe ratios of the main portfolios of interest – low valuation and strong governance – it can be concluded that both the ‘Strong Governance and High Market-to-Book’ and the ‘Strong Governance and High Price-to-Free Cash Flow’ portfolios outperform the market in a statistically significant manner. The market being defined as the Fama & French Market factor or the S&P 500 index. Moreover, both portfolios of interest outperform the Quality factor portfolio in a statistically significant way. On the other hand, no statistically significant outperformance can be detected when testing the Sharpe ratios of the main portfolios of interest against the ‘High Market-to-Book’, ‘High Price-to-Free Cash Flow’ and the Fama & French Value portfolios.
Bailey and de Prado (2014) propose a test of Sharpe ratios to correct for back test overfitting, selection bias and non-normality. To further investigate the findings, their test is applied to the various specifications of the ‘Strong Governance and Low Valuation’ portfolios. Bailey et al. (2014) propose the following equation to test the validity of the findings:
𝑆𝑅̂ = √𝑉[{𝑆𝑅0 ̂ }] ((1 − 𝛾)𝑍𝑛 −1[1 − 1 𝑁] + 𝛾𝑍 −1[1 −1 𝑁𝑒 −1]) (A2)
33
the number of independent trials and e refers to Euler’s number. The significance of the estimated Sharpe ratio is then determined by the following formula:
𝐷𝑆𝑅 ̂ ≡ 𝑃𝑆𝑅̂ (𝑆𝑅̂ ) = 𝑍 [0 (𝑆𝑅
̂ − 𝑆𝑅̂ )√𝑇−10 √1− 𝛾̂3𝑆𝑅̂ +𝛾4̂ −14 𝑆𝑅̂2
] (A3)
Where 𝑆𝑅̂ is the selected strategy’s estimated Sharpe ratio, T is the sample length, 𝛾̂3 is the skewness of
returns, 𝛾̂ is the kurtosis of the returns of the selected strategy. For the purposes of the test, the Strong 4
Governance & High Market-to-Book, long-only and 12-month holding period portfolio is the main specification – Governance & Price-to-Free Cash Flow, Long-Short portfolio formation and the 2-year holding period being alternative configurations. Carrying out the test, the results show that the ‘true’ Sharpe ratio of the strategy is reliably over zero, at the 5 percent level of significance.
Table A2: Portfolio Characteristics
N CG - Score MTB N CG - Score P/FCF High Strong 23 81.44 1.21 18 81.15 9.08 2 26 72.18 1.19 22 72.07 8.6 3 29 65.49 1.17 23 65.5 8.12 Weak 33 55.15 1.12 26 55.62 7.55 2 Strong 31 81.33 1.98 23 81.42 16.04 2 26 73.29 1.96 24 72.3 15.64 3 25 65.57 1.99 21 65.24 15.72 Weak 26 55.93 2 21 55.69 15.38 3 Strong 30 81.33 3.06 25 81.5 23.97 2 27 72.43 3.03 20 72.09 23.24 3 25 65.46 3.08 24 65.71 24.02 Weak 26 56 3.05 20 54.59 23.38 Low Strong 24 81.67 8.18 23 81.8 104.61 2 26 72.32 14.07 23 72.36 97.98 3 28 65.64 10.2 21 65.52 104.07 Weak 28 54.35 8.92 22 53.78 114.51
34
Table A3: Valuation and Corporate Governance
Panel A: Corporate Governance & Market-to-Book
Intercept 4.593*** 17.735*** 4.514*** 18.466*** 4.528*** 18.111*** (39.682) (97.028) (39.076) (102.188) (37.881) (100.551) CGBF -0.026*** -0.013*** -0.025*** -0.010*** -0.025*** -0.012*** (-22.272) (-11.945) (-20.858) (-9.124) (-21.064) (-10.611) CGBS 0.007*** 0.008*** 0.008*** 0.008*** 0.008*** 0.008*** (7.053) (8.835) (7.733) (9.183) (7.645) (9.106) CGCP 0.001 -0.011*** -0.001 -0.014*** 0.000 -0.012*** (0.846) (-15.826) (-0.868) (-20.35) (-0.670) (-18.355) CGSR 0.008*** 0.006*** 0.007*** 0.005*** 0.007*** 0.006*** (15.776) (13.008) (13.905) (11.908) (14.197) (12.612) CGVS -0.005*** 0.008*** -0.005*** 0.009*** -0.005*** 0.008*** (-13.611) (21.11) (-13.223) (22.737) (-13.318) (22.175) Size -0.875*** -0.924*** -0.900*** (-90.318) (-96.747) (-94.592) Leverage 0.111*** 0.114*** 0.112*** (56.905) (59.531) (58.866)
Period Fixed Effects No No Yes Yes No No
Period Random Effects No No No No Yes Yes
Adjusted R-squared 0.017 0.164 0.040 0.199 0.016 0.172
Observations 51,156 51,156 51,156 51,156 51,156 51,156
Panel B: Corporate Governance & Price-to-Free Cash Flow
Intercept 44.473*** 77.401*** 42.354*** 79.926*** 43.004*** 78.360*** (29.310) (31.001) (27.861) (31.977) (27.818) (31.520) CGBF -0.273*** -0.228*** -0.263*** -0.210*** -0.265*** -0.222*** (-17.237) (-14.217) (-16.623) (-13.169) (-16.781) (-13.935) CGBS 0.007 0.005 0.031** 0.029** 0.025* 0.012 (0.506) (0.375) (2.260) (2.131) (1.814) (0.901) CGCP 0.057*** 0.021** 0.062*** 0.020** 0.060*** 0.020** (5.937) (2.193) (6.412) (1.969) (6.261) (2.065) CGSR -0.001 -0.012* -0.025*** -0.038*** -0.019*** -0.020** (-0.198) (-1.889) (-3.674) (-5.522) (-2.799) (-3.003) CGVS 0.030*** 0.060*** 0.040*** 0.075*** 0.037*** 0.065*** (5.792) (11.067) (7.737) (13.626) (7.247) (11.921) Size -2.104*** -2.394*** -2.201*** (-16.572) (-18.888) (-17.446) Leverage 0.008 0.009*** 0.008 (0.979) (1.166) (1.035)
Period Fixed Effects No No Yes Yes No No
Period Random Effects No No No No Yes Yes
Adjusted R-squared 0.008 0.015 0.023 0.031 0.008 0.015
Observations 42,305 42,305 42,305 42,305 42,305 42,305
