Fundamental Indexation versus Achievers Indexation
A study on Payout Policies of 2000-2013 on the European STOXX 600
Jan Thomas Kalsbeek1 Master thesis: Finance University of Groningen
Abstract
A lot of investors move from market capitalization based indices to indices that have an alternative weighing scheme. This study focuses on indices based on payout policies which cover cash dividend, share buyback and the combination of these two. The construction of the indices is based on two approaches; fundamental indexation which emphasizes on the firm’s absolute payout and achievers indexation which focuses on the firm’s payout ratio.
This study makes use of the Stoxx Europe 600 ranging from January 2000 to September 2013. Each of the constructed portfolios generates excess returns compared to the market cap weighted benchmark. Moreover, the achievers indexation approach performs substantially better than the fundamental indexation method. The four factor model regression shows a value bias for all portfolios and a small cap bias for the achievers indices. Furthermore, only the dividend and total payout indices of the achievers indexation approach significant alpha.
JEL classification: G14; G32; G35
Keywords: Share buyback; Cash dividend; Fundamental Indexation; Stock price performance
1 Student at the University of Groningen; Faculty of Economics and Business.
Email: jt.kalsbeek@gmail.com. Student number: s2222094.
2
1. Introduction
Investing in dividend paying stocks is immensely popular. The effect of a dividend payment is easy to understand and quantify. Investors receive a regular income from its investment, which they might reinvest or spend otherwise. The dividend may give a shareholder certain comfort. Even when the stock price performs poorly an investor still receives income from the dividend. When a company temporarily performs poorly, a company usually continues to pay dividend, as was seen in the 2008 financial crisis. However, the drawback of dividend is that it is taxed for certain investors in most countries. One solution for evading dividend taxes is for a firm to repurchase their outstanding shares. By buying back shares the tax is effectively avoided. One of the consequences is that share repurchases have greatly increased in the last two to three decades.
In Europe the popularity of share buybacks started years later than in the United States (US). This was due to new regulations with regard to share buybacks. Since 1998, Germany allows listed firms to buy back their shares. In the US, the regulation constraints were already alleviated in 1983.
Hackethal and Zdantchouk (2006) found in the period 1998 to 2003 that more than 180 German firms had already repurchased shares of their company. Since the beginning of the millennium the repurchases have gained substantially both in number of companies as well as invested amounts.
Share buybacks are gaining more popularity with firms relative to cash dividends, however investing in dividend themed stocks is more popular than ever. The number of active investment funds and passive Exchange Traded Funds (ETF’s) that focus on stocks with a high dividend payout or dividend ratio has never been larger. It seems investors like the value bias of a dividend fund; hence, the current number of actively managed funds and ETF’s that focus on dividend involves hundreds worldwide. Interestingly, a similar explosion of funds has not yet happened to investment funds and ETF’s that focus on share repurchases as a main theme. In this study only two actively managed funds with a buyback theme were found; these funds focus on the European and US stock market, respectively. Additionally, this study found only one buyback themed ETF, which is focused on the US market. Interestingly, there are no active or passive funds that specialize on the total payout policy. The total payout policy is constituted by dividends and share repurchases; both are funded by retained earnings. Investing with a focus on total payout policy would make sense; as mentioned before, share buybacks tend to partially replace dividends.
Share repurchases can be interpreted both as positive or negative signal to the market: positive when the buyback announces a prosperous future for the firm and its share price, negative if it is a signal to the investor that the company does not have suitable investment opportunities. Berkshire
3
Hathaway is a perfect example of not finding enough investment opportunities. It repurchases shares until its market-to-book ratio of 1.2 is reached.2 It would be fascinating to investigate the long term price movement in case of a share buyback. Interesting is also the fact that repurchases are used as an incidental form of payout, where dividends are considered a structural payout. This indicates why a buyback announcement might be regarded as a negative sign.
One of the best known strategies for capturing stocks with good fundamental values is fundamental indexation (FI), first proposed by Arnott, Hsu and Moore (2005). FI is a passive investment strategy and creates portfolios based on several fundamental values: book value of assets, cash flows, revenues, sales, dividends, employees and a composite version of all proxies. In the period 1962 to 2004 Arnott, Hsu and Moore (2005) showed an annual excess return of 2.12 percent on the US stock market using the composite fundamental index. However, FI tends to invest in shares with a large market capitalization.
To investigate the payout policy more thoroughly the ‘achievers indexation’ (AI) is created. The AI is based on a company’s payout ratio. The ratio of a company is calculated by dividing its payout by its market capitalization. Hence, three ratios are calculated: dividend ratio, buyback ratio and total payout ratio. This concept of the AI is already performed on the US stock market. However, this is the first time the approach is applied to the buyback and total payout on the European stock market.
This study investigates the difference between a portfolio created on FI or AI. The following research question derives from this:
Does Fundamental Indexation show a larger excess return to the benchmark than Achievers Indexation with respect to the firm’s payout policy?
This study focuses on portfolios based on the payout structure of companies. The total payout of a company includes cash dividends and share buybacks. Current literature on long term buyback returns is scarce. This scarcity provides an opportunity to shed light onto this matter. The results can provide us with valuable information on how an investor can use the payout policy of a company for profitable investment opportunities. Thus this study will not focus on the announcement of payouts, but on the long term effects on the stock price.
The study is performed on the broad STOXX Europe 600 index, in the period from January 2000 until September 2013. The developed European market was chosen because this area has not yet been investigated with respect to share buybacks. There is however literature of buybacks on a national
2 http://www.berkshirehathaway.com/qtrly/3rdqtr13.pdf
4
level. Because multiple large European countries alleviated the buyback regulation just before 2000, we chose this period as a starting date. Adding more years to the study would have made the returns biased for the few countries where share repurchases have been allowed for a longer time.
The remaining part of this paper is divided in four sections. Section 2 reviews the existing literature that is relevant to this study. In the third section the data requirements, methodology and the descriptive statistics are displayed. Section 4 displays the results and comments of the statistical tests. In the last section the conclusion is presented together with a discussion on limitations and recommendations for future research.
2. Literature review
In the US share buybacks started to arise from the 1980’s. Most research on share buybacks is therefore focused on the US stock market. In most European countries the share buybacks started later, this is partly due to legislation. For example in Germany and France, the regulations on share buybacks were reduced in 1998. Germany alleviated regulation on share repurchases with the
‘Corporation Control and Transparency Act” (KonTraG)’. In the same year France legalized buybacks with the ‘2 July 1998 Act’. This led to a significant increase of stock repurchases in these countries.
In the United Kingdom (UK) the repurchasing of shares is allowed since 1981 with the ‘Companies Act’ law. Rau and Vermaelen (2002) discussed that due to tax and regulatory environment buybacks are less popular in the UK than in the US. However, in the UK the regulations on dividends changed in 1997, making pension funds indifferent to the different forms of payout. As a consequence buybacks gained significantly in popularity.
Fama and French (2001) showed that in the US the number of companies that pay cash dividends dropped from 66.5% to 20.8% in the period 1978 to 1999. In approximately the same period, from 1980 to 2000, Grullon and Michaely (2004) found that share buyback expenses of the total earnings increase from 4.8% to 41.8% in the US. Grullon and Michaely also noticed a substantial increase in buybacks and argued that buybacks instead of dividends are increasingly popular for spending excess capital. Skinner (2008) confirmed the findings of Grullon and Michaely. Skinner concluded that share buybacks replaced dividends as the main form of payout in the US in the past three decades. This replacement is not remarkable since dividends and share buybacks are both part of the payout policy of a firm to the investor; both are funded by retained earnings. However, when an investor receives cash dividend, the investor pays dividend tax in most countries, whereas share buybacks are free of tax and drive up an investors ownership in a firm and indirectly the stock price.
5
The value of dividend payments is not only in the payout amount. Miller and Rock (1985) suggested that dividend announcements convey information, i.e. the dividend signaling hypothesis. With dividend signaling the company’s management can signal future company results to investors. This works in both ways; an increasing dividend signals positive company prospects, which lead to higher stock prices and vice versa. Arnott and Asness (2003) found a relationship between a high dividend ratio and high growth figures in future earnings. However, Grullon, Michaely, Benartzi and Thaler (2005) contradicted the theory of dividend signaling in the period 1963 to 1997 on the US market. Grullon et al. found no correlation between the dividend changes and future earnings. The theory of dividend signaling can also be applied to buyback announcements, where buyback announcements convey information on the future prospects of a company. Chan, Ikenberry, Lee and Wang (2010) discussed that a company can also use a buyback announcement to give a false signal to the market and alter market expectations. False signaling is a cheap form of manipulating investors, and can be used to increase the earnings per share to fulfill the expectations of investors.
The suspected firms do not benefit from the false signaling in the long-run and there are only limited cases of false signaling according to Chan et al.
In a study on the German market, Hackethal and Zdantchouk (2006) pointed out an abnormal stock return of 12% around the announcement date of a share repurchase program. McNally and Smith (2007) studied the Canadian market and focused their research on profitable investment strategies after the buyback announcement. McNally and Smith failed to found abnormal profits when investing following a buyback announcement, when taking into account trading costs and price impacts.
Although results of the stock performance on the announcement day are interesting, they cannot be used for profitable trading in most cases because one cannot anticipate on them beforehand.
Examination of the lagged stock performance can be applied to make profitable trades for investors.
However, the number of studies on long-term buyback results is scarce. Lakonishok and Vermaelen (1990) performed a study on the lagged effect of stock repurchase in the US and found a significant long term result in the first two years following the buyback announcement. The long-term return is usually calculated by taking the two to four year buy-and-hold return after the buyback announcement. In the period 1980-1990, Ikenberry, Lakonishok and Vermaelen (1995) found a 12.1% outperformance of the benchmark on the US market. Several years later, Ikenberry, Lakonishok and Vermaelen (2000) also found long-term buyback outperformance on the Canadian stock market. With an excess return of 7%, the Canadian abnormal return is smaller than on the US market. The Asian stock markets have also been subject to long-term buyback performance studies.
6
The Korean market does not show superior returns in the period 1994-2000, according to Lee, Yung and Thornton (2005). The long-term buyback performance over the whole Hong Kong stock market does not show abnormal returns either. Zhang (2005) found excess returns at small and high book- to-market firms on the Hong Kong market. The positive result with respect to the high book-to- market ratio, better known as value firms, is interesting. A book-to-market ratio above one implies the stock is undervalued, and the firms repurchase their shares at a discount. This discount should reinforce the impact of the buyback and result in a better stock performance. The study of Lakonishok and Vermaelen (1990) was replicated by Peyer and Vermaelen (2009); the study argued there was an overreaction on the share price following on bad news. Furthermore Peyer and Vermaelen proved significant long-term returns in the first two years following the repurchase announcement on the US market between 1991 and 2001.
Ninety percent of financial executives agree or strongly agree the fact that they buy back stock when they think the share price is undervalued. This was investigated by Brav, Graham, Harvey and Michaely (2005) who have done an extensive in-depth study among CFO’s. If a firm repurchases its undervalued shares, a subsequent better performance can be expected. Furthermore, Brav et al.
proved that dividend is regarded as a steady factor in the payout policy, where share buybacks are more commonly used for incidental payouts to investors. Even though share buybacks quickly catch up on dividends, dividends will remain important. The investigation of Brav et al. also showed that financial executives do not necessarily see a link between the dividend payout and earnings.
Companies try to keep dividend payments relatively steady. When a firm makes an above average profit, it is more common to distribute the excess cash in the form of share buybacks than dividends.
Repurchases are also used to improve the earnings per share. An increase of the earnings per share can be used to satisfy the desires of investors and analysts. The last item Brav et al. pointed out is that dividend tax is not the only reason to switch from cash dividend to share buybacks.
Dittmar (2000) argued that firms execute more buybacks to distribute excess cash or to benefit from a potential undervaluation. Stephens and Weinbach (1998) found a negative correlation between share repurchases and preceding stock price performance. Consequently, when the stock shows a temporary undervaluation according to the firm’s managers, a buyback gives them an opportunity to reduce the number of outstanding shares for a relative low-cost price.
2.1 Fundamental Indexation
Many studies were performed that rejected the efficiency of market cap indices (see e.g. Ross (1978), Kandel and Stambaugh (1987), Gibbons, Ross, and Shanken (1989), Zhou (1991)). This implies there
7
are anomalies in the stock markets. Arnott, Hsu and Moore (2005) found a way to benefit from a market anomaly and created FI. FI is a fundamental based index and weighs companies on the absolute values of their book value of assets, cash flows, revenues, sales, dividends, employees.
Arnott, Hsu and Moore performed a study on the US stock market in the period 1962 to 2004: FI outperformed its benchmark with 2.12 percent per year. This result was confirmed by Hemminki and Puttonen (2007) for the European market. In the period 1996 to 2006 Hemminki and Puttonen showed an excess return of 2.66 percent per year. FI does not include the proxy of share buybacks.
However, currently the share repurchases are the main form of payout in the US, making it interesting to investigate a fundamental share buyback index.
As mentioned before, FI tends to have the biggest part of the index invested in the largest market cap companies. The biggest firms usually have the highest absolute dividend and/or buybacks. This means that a firm with a low payout ratio but a high absolute payout can have a higher weight than a company with a high payout ratio but a low absolute payout. AI allocates most weight to firms with the best payout ratio. Thus a smaller company more easily receives a bigger weight in an index.
Literature on the buyback ratio is scarce; however there are a vast number of studies on the dividend yield. Studies show a positive correlation between dividend yield and future stock price performance. Fama and French (1988) concluded that dividend yield explains a large portion of the 2-4 year returns on the US market. Fama and French also found high autocorrelation in the expected returns and showed that the variance grew faster than the future returns.
The current scarcity of buyback related research makes it worth investigating. A buyback can be perceived positive or negative by investors: positive since money is spent to increase the stock price, negative because a buyback can be perceived as a lack of good investment opportunities, which might result in a lower future growth. Based on the current literature, it is known that FI outperformed market cap based indices. Additionally it is interesting to know if the FI approach is the best way to capture a firm’s fundamental values and therefore this study compares FI to AI.
3. Data and methodology
3.1 Data description
This study focuses on the developed European stock market, making use of the Stoxx Europe 600.
This market capitalization index represents the 600 largest European stocks that are listed in one of the 18 developed European countries (see Appendix A.1 for the current number of listings per
8
country). During the investigated period 1,156 unique stocks were used. It is possible that a part of these stocks are no longer listed at this moment, because of factors like rebalancing, merging, bankruptcy or acquisitions. There is no case of survival bias, since all constructed indices are rebalanced annually.
As mentioned before, multiple large European countries have allowed to execute share buybacks since 1998. This is the main reason why this study starts with data from 1999. The other reason is the lack of information on the constituency of the benchmark. The benchmark index is the only index with coverage of developed Europe and constituency lists dating back to 1999. It is possible that a company has a double listing in the Stoxx Europe 600. This can happen when a company is listed in more than one country, like the Dutch Unilever N.V. and the British Unilever Plc. If this is the case the double listing is removed, so only one listing is eligible for the newly composed index. Based on the data, portfolio returns are created from 2000 up to the third quarter of 2013; this resulted in 165 months of data. The data includes two market crashes: the bursting of the dot-com bubble in 2001- 2002, and the global financial crisis with its market collapse in 2008. Therefore we can also observe how the indices performed in the distress years.
Most data is collected from Datastream. This includes the yearly amount of cash dividends, share repurchases, market capitalization and the constituencies of the Stoxx Europe 600. The research is focused on returns in the Euro currency. When stocks are not listed in Euros, the values are converted into Euros on a daily basis. This prevents currency effects in this study.
Table 3.1 – Number of payouts
The numbers noted in the table represent the total number of firms that distributed earnings with buybacks, dividends or both.
Year Buyback Dividend Total payout
1999 88 495 497
2000 143 470 474
2001 154 504 512
2002 142 495 505
2003 161 506 516
2004 188 517 529
2005 242 534 547
2006 293 540 555
2007 346 537 560
2008 367 548 565
2009 215 512 536
2010 261 524 554
2011 317 549 570
2012 282 551 572
9
Table 3.1 shows the number of firms listed in the Stoxx 600 which repurchased shares, paid out cash dividend or did both in the concerned year. As can be seen in the table, the number of buyback firms substantially increased. The number of dividend paying companies also increased, subsequently the total number of corporations using either repurchases or dividends increased as well. There was a clear fall in the number of payout companies after the financial crisis of 2008. The number of repurchasing firms has not yet recovered to pre-2008 level. Of the whole sample, United Textiles SA was the firm with the smallest market cap. In 1999 United Textiles had a market cap of 77 million euro. A small market cap might cause liquidity problems when a portfolio invests a large portion, these consequences are not taken into account in this study. However, the smallest companies of the Stoxx 600 grew significantly; by 2013 the smallest firms had a market cap of more than 500 million.
Graph 3.1 - Payout amounts. The graph displays the buyback and dividend payout amounts. The combined amount forms the total payout in the respective year.
Graph 3.1 demonstrates the absolute payout amount. Both the buyback and dividend amounts are displayed; the two combined result in the total payout. In the graph it is clearly observable that the payout amounts substantially decrease in 2009. Especially the amounts spent on share repurchases decreased considerably and have not recovered in recent years. This slow recovery of buybacks is also exposed in table 3.1. The paid out dividend amounts recovered after the crisis, which is consistent with the increasing number of firms that pay dividends.
0 100 200 300 400 500
Payout value in millions
Year
Graph 1 - Payout amounts
Dividend Buyback
10
Data collected on the four factor model is collected from Kenneth French’s data library3. These factor loadings are revised if necessary; the market is replaced for the Stoxx Europe 600 and the risk-free rate is replaced for the 3-month Euribor.
This study makes use of cash dividends and not stock dividends. There are several reasons for this:
- Data on stock dividends is poorly accessible, whereas the data for cash dividends is widely available;
- The majority of dividend payments are done in cash, not in stock payments;
- Most existing studies on dividends usually only use cash dividends.
Each form of share repurchases is used: open market, tender offer, Dutch auction and targeted repurchases. It is possible a firm issues new shares; this can be the case with employee stock options. Kayka (2002) argued that a repurchase announcement has less effect when the firm’s management has a high level of options. However, the impact of employee option is not taken into account with share buybacks. The matter would deviate too much from the main purpose of this investigation and the information on employee stock options is typically not readily available for investors.
An important consideration when investing in indices are trading costs. Yet, this study does not specifically focus on the trading costs. An index tracker can have an expense ratio as low as 0.1% per year4. An index based on AI can have an expense ratio of 0.2% per annum, excluding the 0.5%
management fee5. The difference of the expense ratios is negligible and therefore not taken into account in this study.
3.2 Methodology of the index construction
In this study six portfolios are created. These indices were created for the buyback, dividend and total payout. The total payout consists of the weighted average of the buyback and dividend. The three versions will each be tested based on two investment philosophies; AI and FI.
- Achievers Indexation: This form of indexation is based on the ‘NASDAQ Buyback Achievers Index’. The original version of this index is compiled by shares listed on the NASDAQ, NYSE or NYSE MKT. Firms are incorporated to the index when they have repurchased at least 5 percent of their outstanding shares in the twelve months before. It should be noted that the
3http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
4https://www.spdrs.com/product/fund.seam?ticker=spy
5http://www.invescopowershares.com/pdf/P-PS-PRO-1.pdf
11
AI of the Stoxx 600 in this study includes all repurchasing firms regardless of the buyback percentage. Abiding the 5 percent rule would result in too few constituents in the index. The higher the buyback ratio the more weight a stock receives in the index. The index is constructed annually in January and rebalanced in April, July and October. This approach is also applied to the cash dividend and total payout amounts.
- Fundamental Indexation: The concept of FI was first introduced by Arnott, Hsu and Moore (2005). FI constructs indices based one several proxies, one of these proxies is the dividend payout. In contrast to the AI, the fundamental dividend index does not focus on the dividend ratio, but instead focuses on the absolute dividend payout. The FI portfolios are constructed annually in January and are not rebalanced during the year. On top of the fundamental dividend index, a buyback and total payout version are also created.
The new indices use the constituency list of December of the prior year. Thereafter, the indices are constructed in January according to their investment philosophy. For all indices the daily returns are calculated and subsequently the monthly and yearly returns can be derived. The virtual investment will take place on the first trading day of the year. All firms are required to disclose their buyback amounts quarterly, however most companies publish it weekly or at the end of the month. Because of this, it is realistic to invest in a new index on the first trading day of the year.
At the annual reweighing of the indices in January, firms are capped at 10 percent. The maximum weight of 10 percent is based on the Stoxx Europe 506, although more indices use this maximum.
This prevents firms with a large payout to be too dominant in the index.7 3.2.1 Achievers Indexation
The weight of a stock in the AI is based on the payout ratio. As shown before, the ratio is computed by taking the dividend, buyback or total payout of the company and dividing this value by the firm’s market cap of the beginning of the year. So a company with a relatively high payout will receive a high weight in the index, whereas a firm with no payout will not be part of the index. The weight of a stock in the index is calculated as follows:
(3.1)
6 http://www.stoxx.com/download/indices/factsheets/sx5e_fs.pdf
7 Results showed that a stock could get an index weight considerably more than 10%, when no maximum weight was considered.
12
where WAi,t stands for the weight in the AI portfolio of stock i on date t, PRi,t is the payout ratio of stock i on date t and SPRi,t is the sum of the payout percentages of all stocks in the index on date t.
The daily total return of a constituent is computed as:
( ) (3.2)
where is the return of stock i on date t, is the price of stock i on date t and is the price of stock i on date t-1. When combining the weight and the return, the index can be calculated as follows:
(3.3)
where is the return of relevant fund on date t. The three created AI portfolios are rebalanced quarterly in April, July and October to their January ratios. Rebalancing takes place on the first trading day of the quarter and restores the weight of every constituent back to the original weight that is determined in January.
3.2.2 Fundamental Indexation
The three indices based on FI are weighed on their absolute payout amount. However, in contrast with the approach of Arnott (2005), this research does not use the trailing five-year average payout, but it is performed on the trailing one-year payout. This choice was made because AI also makes use of the one year payout. As a result the comparison between FI and AI is better fitting. The weight of a single constituent is calculated as follows:
(3.4)
where WFi,t stands for the weight in the FI portfolio of stock i on date t, VPi,t is the absolute value of the payout amount of stock i on date t and TVPi,t is the total sum of payouts of all stocks in the index on date t. The return equations of the various FI portfolios are the same as equations 3.2 and 3.3. The FI portfolios are constructed in January. Unlike the AI approach, the FI method is not rebalanced quarterly.
3.3 Methodology of the excess returns
Excess returns are measured by comparing the mean return with the Stoxx 600 with each of the six constructed indices. The excess return is calculated by subtracting the benchmark return from the index return for the concerning time period. To investigate if the difference in returns between the index and the benchmark are significant, a t-test was applied to the monthly returns. This study makes use of the two-sided paired samples t-test. Yet, the paired samples t-test assumes the returns
13
to be normally distributed, which is not the case in this study. For that reason a different test is performed that assumes the returns to be non-parametric: the Wilcoxon signed rank test. The Wilcoxon test assigns each outcome a rank, orders these ranks and calculates a Z-score on the positive ranks.
In addition a binomial test was also performed. The binomial test can be used to express how many times an index out- or underperforms the benchmark. The outcome is expected to be 50 percent, meaning the index and the benchmark have the same number of out- or underperformance. The test helps to determine if the outperformance of the constructed indices is significantly different from 50 percent. It should be noted that the binomial test can only be used to determine out- or underperforms and not the magnitude of return.
Each of the three tests performed on the excess return is used with a two-tailed basis; because it is obviously unknown beforehand if the portfolios will perform better or worse than the Stoxx 600.
3.4 Methodology of the risk characteristics
As a final aspect, the influence of risk characteristics on the return was investigated. The model’s factors attempt to explain where the excess return originates from. The outcome shows if an active portfolio manager, or in our case the constructed index, is relying on one of the included factors and/or can create alpha. The return that cannot be explained by the risk factors is better known as alpha. If an index significantly generates alpha, it can be concluded it performs better than the benchmark after correcting for the factor loadings and consequently ‘adds value’.
Fama and French (1993) were the first to introduce the three factor model for the research on risk characteristics. Although the three factor regression was also performed, this study will focus on the results of the four factor model. The four factor model was first described by Carhart (1997) and includes the additional momentum factor.
The first risk factor is the market factor, comparable with the classical risk factor known from the capital asset pricing model (CAPM). It is calculated by taking the Stoxx 600 return and subtracting the risk-free rate. When the market factor is below one, it implies the portfolio is shows a lower beta compared to the benchmark. The second element is the small-minus-big market capitalization (SMB) factor, also known as the size effect. This factor looks if the returns come from large or small firms, a positive SMB coefficient represents a small cap bias and a negative SMB coefficient indicates a bias towards large cap stocks. The third aspect is the high-minus-low book-to-market (HML) factor, also known as the value premium. When the HML coefficient is positive this indicates undervaluation and
14
implies there is a value bias. A positive HML coefficient indicates overvaluation and a focus towards growth stocks. The final added feature is the winners-minus-losers (WML) factor, better known as the momentum factor. When the WML coefficient is positive the portfolio makes use of momentum and invests in outperforming stocks. If the WML coefficient is negative the portfolios is invest in stocks that show no momentum.
The four factor model was tested on the monthly returns. The main test was performed on the whole investigated period. Additionally, extra tests were performed on two sub periods; from January 2000 to December 2007 and from January 2008 to September 2013. January 2008 was chosen because the financial crisis started to unfold at that time. This allowed us to see if the risk characteristics changed in recent years. As mentioned before, the monthly factor loadings were retrieved from Kenneth French’s website and were adapted for the risk-free rate and the Stoxx 600. The equation for the four factor model is as follows:
( ) (3.5) where, is the risk-free rate in month t, is the alpha, , , , are the coefficients of the concerning factor, is the market return in month t and , , depict the monthly value for the SMB, HML, WML factors, respectively. The equation for the three factor model is similar except it excludes the WML factor and its coefficient.
3.5 Data analysis
In this section the first data is analyzed. First, the descriptive statistics are displayed. Thereafter information on the number and amount of payouts is presented. In the last part of this section the absolute, excess and investment returns are displayed. In table 3.2 the descriptive statistics are presented.
The first part of table 3.2 displays the returns of the Stoxx 600 and the constructed indices. The results clearly show an outperformance of each index compared to the benchmark. Additionally, all AI portfolios have higher returns than FI portfolios. In contrast with one would expect based on the risk/return tradeoff; the portfolios with the highest returns do not display higher volatility. All of the indices have a high correlation with the Stoxx 600. However, both buyback portfolios display the lowest correlations and highest tracking error, meaning that these indices deviate most from the Stoxx 600. The closely related information ratio is positive for each of the portfolios. An information ratio between 0 and 0.5 indicates that the portfolios have a ‘skill’ between good and above average according to Grinold and Kahn (1995). This result is partly backed up by the Sharpe ratio, which
15
measures the risk/return of a fund. Although it is observed that the Sharpe ratio is more positive for AI portfolios, the value is far from reaching the value of 1.0, which is considered a good risk/reward return. The result of the Sharpe ratio is somewhat distorted, since it does not account for the poor stock performance of the developed world’s equity market of the past 13 years. As an improvement to the Sharpe ratio, the Sortino ratio only looks at the downside risk. Positive volatility is the feature an investor wants, so this ‘risk’ does not need to be taken into account. For the minimum acceptable return the average monthly return of the benchmark was used. The Sortino ratio is above zero for each created portfolio and each of the AI portfolios accomplishes a better result than the FI portfolios. The indices do a better job by having less downside deviation than the Stoxx 600. All Jarque Bera values are significant on 5 percent, implying each index is non-normally distributed.
Table 3.2 - Descriptive statistics
This table represents the descriptive statistics from the Stoxx 600 and the constructed AI and FI portfolios. Data ranges from January 2000 to September 2013. ** denotes statistical significance on a 5% level.
Benchmark Achievers Indexation Fundamental Indexation
Stoxx 600 Buyback Dividend Total payout Buyback Dividend Total payout
Annual geometric mean 1.97% 8.06% 8.07% 8.10% 5.06% 4.32% 4.07%
Monthly geometric mean 0.15% 0.64% 0.64% 0.64% 0.40% 0.35% 0.33%
Monthly minimum -14.06% -19.82% -18.80% -19.20% -15.37% -14.45% -15.57%
Monthly maximum 14.52% 20.26% 21.09% 20.83% 21.40% 17.89% 19.41%
Daily standard deviation 1.30% 1.30% 1.26% 1.14% 1.16% 1.40% 1.23%
Annual volatility 20.70% 20.61% 19.94% 18.07% 18.48% 22.28% 19.59%
Correlation 85.85% 93.17% 92.61% 87.23% 95.51% 94.14%
Tracking error 2.75% 1.81% 1.90% 2.58% 1.41% 1.67%
Information ratio 0.19 0.28 0.27 0.11 0.15 0.12
Sharpe ratio 0.34 0.37 0.36 0.18 0.15 0.14
Sortino ratio 0.30 0.34 0.33 0.16 0.14 0.12
Kurtosis 2.06 2.72 2.67 1.67 1.74 1.82
Skewness -0.56 -0.52 -0.58 -0.32 -0.49 -0.47
Jarque-Bera 14.67** 7.99** 10.12** 15.03** 17.65** 15.63**
Panel A of table 3.3 shows the annual results of the benchmarks and the six created indices. For the performance during market crashes, the years of the dotcom bubble and the financial crisis were specifically investigated. In 2002 the created indices performed poorly, however five out of the six portfolios performed better than the Stoxx 600. Especially in 2003 all of the constructed indices showed a much better recovery. In 2008, during the beginning of the financial crisis the Stoxx 600 and the portfolios all performed very poor and lost about 40 percent of their value. However, like the dotcom bubble, in the 2009 recovery year the created indices showed a much larger increase of value than the benchmark. Panel B of table 3.3 displays the excess returns of each portfolio compared to the Stoxx 600. The AI portfolios show their largest outperformance of the benchmark in
16
the first years of the investigated period. The FI portfolios show a similar picture, however the excess returns are smaller in most years.
Table 3.3 – Annual returns and excess returns
This table displays the annual returns of the Stoxx 600 and the constructed AI and FI portfolios. Panel A represents the absolute annual return. Panel B shows the excess returns of the corresponding created portfolio in comparison with the Stoxx 600.
Panel A: Yearly returns Achievers Indexation Fundamental Indexation
Year Stoxx 600 Buyback Dividend Total payout Buyback Dividend Total payout
2000 -1.09% 16.80% 10.89% 11.79% 18.91% 7.34% 8.96%
2001 -15.44% 5.95% -3.20% -0.86% -2.55% -9.69% -8.13%
2002 -26.96% -19.92% -16.31% -17.64% -27.81% -23.44% -25.93%
2003 14.13% 26.93% 29.88% 29.77% 20.55% 22.83% 24.66%
2004 12.13% 22.83% 23.02% 22.49% 14.93% 17.10% 15.52%
2005 26.98% 31.36% 27.86% 28.75% 29.11% 24.99% 26.61%
2006 22.38% 31.48% 29.93% 30.40% 19.83% 20.93% 21.00%
2007 0.27% 3.49% -0.53% 0.74% 9.48% 5.94% 6.22%
2008 -40.93% -43.85% -42.24% -42.88% -40.83% -39.95% -41.57%
2009 31.27% 49.31% 43.96% 45.68% 47.53% 33.97% 39.51%
2010 11.50% -1.72% 16.61% 12.60% -1.94% 7.02% 3.22%
2011 -7.82% -13.27% -12.18% -12.48% -14.69% -8.01% -11.12%
2012 20.11% 15.74% 19.48% 18.49% 11.98% 12.90% 12.61%
2013 Q3 12.21% 27.18% 20.41% 22.31% 22.75% 16.99% 18.32%
Panel B: Excess returns Achievers Indexation Fundamental Indexation
Year Buyback Dividend Total payout Buyback Dividend Total payout
2000 17.89% 11.98% 12.88% 20.00% 8.43% 10.05%
2001 21.40% 12.24% 14.58% 12.89% 5.75% 7.31%
2002 7.04% 10.65% 9.32% -0.85% 3.52% 1.03%
2003 12.80% 15.75% 15.63% 6.42% 8.70% 10.53%
2004 10.70% 10.89% 10.37% 2.81% 4.97% 3.39%
2005 4.38% 0.88% 1.77% 2.13% -1.99% -0.37%
2006 9.10% 7.55% 8.02% -2.56% -1.45% -1.39%
2007 3.21% -0.80% 0.47% 9.20% 5.67% 5.95%
2008 -2.92% -1.31% -1.95% 0.10% 0.98% -0.64%
2009 18.04% 12.70% 14.41% 16.26% 2.70% 8.24%
2010 -13.22% 5.11% 1.10% -13.44% -4.48% -8.28%
2011 -5.45% -4.36% -4.66% -6.87% -0.19% -3.30%
2012 -4.37% -0.63% -1.62% -8.13% -7.21% -7.50%
2013 Q3 14.98% 8.21% 10.10% 10.54% 4.79% 6.11%
To better visualize the effect of the returns, table 3.4 represents a virtual investment of €1,000 in each of the indices. The 2013 Q3 result denotes the end value when one would have invested €1,000 since the beginning of 2000. As can be seen in the results, all constructed indices outperform the benchmark. Especially the indices based on the AI philosophy beat the Stoxx 600 substantially, almost all of the AI portfolios triple in value. Remarkable is the return of the achievers total payout portfolio. This index has the highest returns, meaning it is more heavily invested in buyback firms when repurchased stocks perform well and more invested in dividend stocks when dividend shares
17
perform well. FI portfolios outperform the benchmark, but perform worse than the AI method. The fundamental buyback index does achieve the highest FI return.
Table 3.4 - Results of investing 1,000
This table displays the cummulative return of virtually investing €1,000 in the Stoxx 600 and each of the created AI and FI portfolios. The period ranges from January 2000 up to the third quarter in 2013.
Benchmark Achievers Indexation Fundamental Indexation
Year Stoxx 600 Buyback Dividend Total payout Buyback Dividend Total payout Year 1,000 1,000 1,000 1,000 1,000 1,000 1,000 2000 989 1,168 1,109 1,118 1,189 1,073 1,090 2001 836 1,238 1,073 1,108 1,159 969 1,001 2002 611 991 898 913 836 742 741 2003 697 1,258 1,167 1,185 1,008 912 924 2004 782 1,545 1,435 1,451 1,159 1,067 1,068 2005 993 2,030 1,835 1,868 1,496 1,334 1,352 2006 1,215 2,669 2,384 2,436 1,793 1,614 1,636 2007 1,218 2,762 2,372 2,454 1,963 1,709 1,738 2008 720 1,551 1,370 1,402 1,161 1,027 1,015 2009 945 2,315 1,972 2,042 1,713 1,375 1,416 2010 1,053 2,275 2,300 2,300 1,680 1,472 1,462 2011 971 1,973 2,019 2,013 1,433 1,354 1,299 2012 1,166 2,284 2,413 2,385 1,605 1,529 1,463 2013 Q3 1,308 2,905 2,906 2,917 1,970 1,788 1,731
4. Results
In this section the outcomes of the statistical test are discussed. The first will be on the results of the excess returns. Next, this study shall emphasize on the results of the four factor regression.
4.1 Results of performed tests on the excess return
Table 5 shows the results of the three tests, i.e. paired samples, Wilcoxon signed rank and binomial test, performed on the excess returns. The excess returns were calculated by subtracting the benchmark return from the index return. The three different panels each represent one of the tests.
As panel A of table 4.1 shows, the t-tests of the AI portfolios indicate a considerable outperformance of more than half a percent per month. These results are significant on a one percent level. Thus, it can be concluded the mean returns are significantly higher than the Stoxx 600. FI shows an outperformance as well, ranging from 0.2 to 0.3 percent. The fundamental buyback and total payout indices show no significant difference compared to the benchmark. However, the fundamental dividend index is significant on a 10 percent level, making its mean positively different from the benchmark.
18
Table 4.1 – Test results excess return
This table shows the test results on the monthly excess returns ranging from January 2000 to September 2013. The excess returns are the differences between the created AI and FI portfolios compared with the Stoxx 600. Each test is performed on a two-tailed distribution. Panel A displays the average monthly excess returns using the paired samples test. Panel B shows the results of the Wilcoxon signed rank test: it displays the number of positive and negative monthly returns. Panel C denotes the results of the binomial test, the results display how many times the new index beats the benchmark. *,** and ***
denote statistical significance on a 10%, 5% or 1%, respectively.
Achievers Indexation Fundamental Indexation Panel A: Paired samples test Buyback Dividend Total payout Buyback Dividend Total payout
Excess return 0.53% 0.52% 0.52% 0.29% 0.21% 0.20%
T-statistic 2.64 3.65 3.52 1.43 1.83 1.50
Probablility 0.90%*** 0.04%*** 0.06%*** 15.49% 6.95%* 13.52%
Panel B: Wilcoxon signed rank test Buyback Dividend Total payout Buyback Dividend Total payout
Positive ranks 103 102 104 93 83 90
Negative ranks 62 63 61 72 82 75
Z-score -2.97 -3.56 -3.78 -1.57 -1.66 -1.58
Probability 0.29%*** 0.04%*** 0.02%*** 11.58% 9.72%* 11.43%
Panel C: Binomial test Buyback Dividend Total payout Buyback Dividend Total payout
Positive proportion 61.82% 61.82% 63.03% 56.36% 51.52% 53.33%
Probability 0.30%*** 0.30%*** 0.10%*** 11.92% 75.56% 43.64%
As mentioned in the data section, the distributions of the monthly returns are not normally distributed. To deal with the non-parametric data, the Wilcoxon signed rank test is performed. The results of the Wilcoxon test can be seen in panel B of table 4.1. The results show that each of the six indices has more positive than negative ranks. A positive rank is determined by a outperformance of one of the portfolios compared to the Stoxx 600. Similar to the t-test, the results of the AI portfolios are all significant on one percent, meaning the excess returns of the constructed indices are higher than the benchmark’s. Of the three FI portfolios only the dividend index showed a significant outcome on a ten percent level. Therefore, the returns of the FI portfolios do not differ significantly from the Stoxx 600.
Panel C of table 4.1 shows the results of the binomial test. The outcomes of five of the six indices are similar to previous tests. Only the fundamental dividend index does no longer show significance on a ten percent level; this index does not display more out- than underperformance.
The outcomes of each index are almost all similar. Each AI portfolio displayed highly significant excess returns, meaning the indices add additional return in comparison with the Stoxx 600.
Regarding FI, only the dividend index showed significant excess returns on two of the three tests.
19
The fundamental buyback and total payout indices displayed no significant excess returns at all, consequently it cannot be concluded the excess returns differ from zero and add additional return compared to the Stoxx 600.
4.2 Results of the four factor regression
Table 4.2 shows the results acquired from the four factor regression. The table displays each of the six newly constructed portfolios. The results include the alpha and the coefficients of each risk factor with their appropriate standard error and t-statistic. The adjusted r-squares are displayed to explain the movement of the index by the Stoxx 600. The results are based on the monthly returns. First, the results of the AI method will be discussed, followed by those of the FI approach.
Table 4.2 - Results four factor regression
This table contains the results of the four factor regression. The results are based on the monthly returns of the created AI and FI portfolios, ranging from January 2000 to September 2013. The coefficient, standard deviation and t-statistic is displayed for each factor.
Furthermore, the adjusted r-squared and Durbin-Watson statistic is showed for each portfolio. Panel A contains the three indices based on the Achievers Indexation investment philosophy, whereas panel B covers the indices based on Fundamental Indexation. *,** and ***
denote statistical significance on a 10%, 5% or 1%, respectively.
Panel A: Achievers Indexation
Buyback Dividend Total Payout
Coefficient Std. Error T-statistic Coefficient Std. Error T-statistic Coefficient Std. Error T-statistic
Alpha 0.19% 0.17% 1.16
0.30% 0.10% 2.88 *** 0.27% 0.11% 2.61 ***
Market 0.91 0.04 23.61 *** 0.96 0.02 38.92 *** 0.95 0.02 38.70 ***
SMB 0.26 0.07 3.73 *** 0.32 0.04 7.14 *** 0.31 0.04 6.89 ***
HML 0.61 0.06 10.48 *** 0.37 0.04 10.07 *** 0.43 0.04 11.43 ***
WML -0.13 0.04 -3.39 *** -0.10 0.02 -4.17 *** -0.10 0.02 -4.24 ***
Adj. r-squared 86.21% 93.63%
93.69%
DW statistic 2.16 *** 2.13 *** 2.23 ***
Panel B: Fundamental Indexation
Buyback Dividend Total Payout
Coefficient Std. Error T-statistic Coefficient Std. Error T-statistic Coefficient Std. Error T-statistic
Alpha 0.03% 0.15% 0.22 0.10% 0.08% 1.24 0.05% 0.09% 0.53
Market 0.89 0.04 24.65 *** 0.91 0.02 46.21 *** 0.93 0.02 41.82 ***
SMB 0.01 0.07 0.09 -0.03 0.04 -0.71 -0.01 0.04 -0.26
HML 0.56 0.05 10.34 *** 0.28 0.03 9.52 *** 0.37 0.03 10.93 ***
WML -0.15 0.04 -4.05 *** -0.10 0.02 -4.97 *** -0.12 0.02 -5.24 ***
Adj. r-squared 87.69% 95.59% 94.88%
DW statistic 2.14 *** 2.23 *** 2.28 ***
In panel A of table 4.2, the results of the AI portfolios are displayed. Each of the three portfolios exhibit positive alphas on the monthly returns. The monthly alpha reaches from 0.19 to 0.3 percent, which results in 2.33 to 3.68 percent alpha per year. Not all alphas of the AI portfolios are significant,
20
the dividend and total payout indices are significant on a one percent level, the buyback index does not show significant alpha. Concerning the dividend and total payout results, it can be inferred that these two indices beat the market after correcting for the market, size, value and momentum premiums. The market factor loading is the value that is most closely linked to the beta of the CAPM.
For each of the three portfolios the factor loadings lie between 0.91 and 0.96 and all are significant on a one percent level. Interestingly, the beta of the portfolios is a little below the market’s beta, suggesting the portfolios move somewhat less than the benchmark. The SMB and HML factor loadings are both positive for each AI portfolio and are significant on a one percent level. Based on the positive coefficients, there is a small cap and value bias for the AI portfolios. It is noteworthy to mention that the buyback portfolio has the biggest focus on undervalued stocks. The WML factor loadings show slightly negative but highly significant values. The AI portfolios show a negative momentum bias, although the bias is small. The adjusted r-square is also considered in the four factor regression, and is high for each of the three indices, ranging from 86 percent for the buyback index to 94 percent for the dividend and total payout index. Based on the high adjusted r-squares, it can be inferred that the four factors explain most of the portfolios movement. The last value, the Durbin Watson statistic, is used to test for autocorrelation and is tested on a one percent significance level. All three DW statistics are close to two and do not fall within the limits of autocorrelation, therefore the null-hypothesis is rejected meaning there is no evidence for autocorrelation in the sample.
Panel B of table 4.2 displays the same statistics as panel A, but concerns the three FI portfolios. A part of the results are similar to the AI: all three FI portfolios each show small alpha, however none of these alphas are significant. Based on the regression it cannot be inferred the portfolios will structurally create positive alpha. The SMB coefficients are all close to zero and highly insignificant.
Abovementioned means there is no size bias. In contrast to the AI approach, FI portfolios do not show a size bias. This can be explained by the fact that an AI tends to invest in stocks with a smaller market cap compared to a FI, according to what one would expect. The market factor loading of the FI portfolios is around 0.9, implying the FI portfolios exhibit less risk than the Stoxx 600. The HML factor loadings are highly significant and have positive values. Hence, the FI portfolios have a significant bias towards value stocks. Similar to the AI approach, the fundamental buyback portfolio has the biggest focus on undervalued stocks compared to the dividend and total payout portfolios.
The WML factor loading is slightly negative and significant on a one percent level. The indices have a small negative momentum, which is comparable to the AI method. The adjusted r-squares are very high and on average larger than the adjusted r-squares of the AI portfolios. This means the four factor model explains the performance of the FI portfolios very well. The results of the Durbin
21
Watson statistic are similar to the AI portfolios and fail to show evidence for autocorrelation. This results in contrast with Fama and French (1988) who found a high autocorrelation in the expected returns.
In addition to the analysis of the whole period 2000 to 2013, the four factor model was also performed on two sub periods. The two sub periods last from January 2000 to December 2007 and from January 2008 to September 2013. The results are depicted in Appendix A.2. Interestingly, with respect to the AI method there is no longer significant alpha in the first period, in the second period the dividend and total payout show significant alpha on a five percent level. The SMB factor of the buyback AI is no longer significant, meaning there is no longer a small cap bias in the second period.
For the FI portfolios the results of the split periods are similar to the whole period. Except for the fundamental total payout portfolio the SMB factor is negative and significant on a ten percent level, meaning there is a large cap bias.
In this study a regression on the three factor model was also performed, which leaves out the momentum factor; the results are displayed in Appendix A.3. All factors of AI and FI show results and significance levels similar to the four factor regression. However, the alpha of the achievers total payout portfolio is no longer significant. The alpha of the dividend AI portfolio has decreased and is significant on a ten percent level.
5. Conclusion
The rise of smart beta investing has been enormous in the last decennium. Investors are looking for other ways to invest than the traditional market cap weighted portfolios. The smart beta investors look for low cost ways to invest in portfolios with alternative weighing schemes. This study investigated passive investment opportunities using two investment philosophies: fundamental indexation (FI) and achievers indexation (AI). The main difference between the two approaches is that FI looks at the company’s absolute payout values, while AI looks at the firm’s payout ratios relative to its market capitalization.
The payout policy covers cash dividend, buyback and the combination of the two: total payout.
Accordingly six portfolios have been created. This study was performed on the European stock market using the Stoxx Europe 600. The period of investigation was almost fourteen years, ranging from January 2000 to September 2013. To review the findings of this study, the research question is brought back into mind:
