Testing The Relation Between Price to Earnings Ratio And Equity Returns As Well Earnings Growth In FTSE 100, United Kingdom

(1)

1

Testing The Relation Between Price to Earnings Ratio And

Equity Returns As Well Earnings Growth In FTSE 100, United

Kingdom

Faculty of Economics and Business

Master Thesis - MSc Finance

Konstantinos Tekonakis

(2)

2

Abstract

This paper investigates the relationship between the price-to-earnings ratio and stock returns as well current/future earnings growth based on the publicly traded firms which are included in FTSE 100 index, United Kingdom for the time period 2009-2015. Three sets of samples namely: All companies, Non-financial firms and Industry level are used to provide a clear picture of explanatory power of price-to-earnings ratio and its interaction with size effect, market momentum and earnings growth factor in order to identify their predictive power for a holding period of one year. The empirical results suggest that P/E effect does not exist for the set of sample includes all the firms listed in FTSE 100. Moreover, accounting measures such as earnings growth and market capitalization play a key role in explaining subsequent equity returns. Testing the P/E ratios linearly with current and/or future earnings growth, we find no strong evidence which support the Extrapolation Theory supported by Lakonishok, et.al. (1994). However, the statistical significance of positive slopes of price to earnings ratios provide some evidence supporting the case of under reaction hypothesis suggested by Jegadeesh and Titman (1993).

Jel Classifications: G11, G14

Keywords: Price-to-Earnings Ratio, Earnings growth, Stock performance Author: Konstantinos Tekonakis* (Student No: S2743523)

Supervisor: Dr. Richard Klijnstra1

Co-assessor: Dr. Peter Smid

1_{I would like to thank Dr. Klijnstra for his helpful advice and useful feedback.}

(3)

3

1. Introduction

Among many challenges for practitioners and academics today is to identify proxies that are able to predict stock returns in the future. One of the most well-known benchmarks in academic literature is the established price-to-earnings ratio (P/E) or its reciprocal (E/P). The typical version of P/E ratio is formed using the stock price which is the typical firm’s equity price recorded in the exchange stock market, divided by earnings (or net income) per share (EPS) that may change in some occasions due to the implementation of accounting practices and the purpose of evaluation. Among researchers, Lakonishok, et. al., (1994), Basu (1997), Basu and O’Shea (2014), Fama and French (1992) and Stefanis (2005) found a predictive power of stocks with low P/E ratios to accumulate higher stock returns than equities with high P/E ratios in different stock exchange markets. The preceding predictive power of price-to-earnings ratio has been subject to market efficiency debate in public writings. To make this clear, a P/E ratio is reported as market anomaly in academic literature when a P/E portfolio formulation generates future abnormal stock returns compared to the peer portfolios construction, when it cannot be explained in terms of risk. On the other hand, the market efficiency hypothesis is split into three cases: Strong efficient market, Semi-Strong efficient market and Weak efficient market hypothesis. Strong efficient market hypothesis supports that all information is completely reflected in stock prices, therefore a portfolio manager cannot predict superior equity returns in the future, as all stock prices are expected to follow random walk movements. Semi-Strong efficient market hypothesis states: new information is quickly incorporated into the equity prices, assuming that all past financial news and information are fully reflected in the stock prices. Hence, Semi-strong market efficient hypothesis would be violated if the P/E ratios have a predictive power. Finally, Weak efficiency market hypothesis says that previous stock returns are not linked with subsequent equity returns, assuming that past information is not incorporated in equity price and they are useless for an investor to forecast future abnormal returns.

(4)

4

because this market index includes the top one hundred companies with the highest market capitalization. It is interesting to concentrate on large capitalization companies as past evidence suggested a link between P/E ratios and subsequent stock movements in large capitalization market. In particular, Becker, Lee, Gup (2012) conducted an empirical research on firms listed in S&P 500 and found a mean reverting pattern of earnings ratios for the period 1871-2003. The concept of mean reverting of price-to-earnings ratios is based on the fact that the P/E ratios will tend to move to the average point over time. Hence, if the mean of P/E ratios is far away from its average point, a subsequent change in equity prices or earnings growth is expected to happen. Regarding the sample period of this study, we believe that the appropriate proxy for time period is from 2009 to 2015. The main reason is that we want to investigate the P/E effect for a more recent period. Furthermore, we observe stock performance in terms of cumulative returns on annual basis. The argument is that testing the predictive power of accounting measures for one-year time period, it could provide insights into the sustainability of business performance of the companies included in the sample. Therefore, this analysis might be fascinating for a portfolio manager who is interested in getting an impression of the next year stock performance with respect to the previous year accounting performance metrics.

Another aspect of price-to-earnings ratio (P/E) is its ability to explain the earnings growth prospects of the firms. The shared factor between price-to-earnings ratio and earnings growth is the earnings. The sense that an increase (decline) of price-to-earnings ratio is related with future increase (decline) in earnings growth is clearly in line with the extrapolation effect supported by Lakonishok, et. al., (1994).

(5)

5

to the public, investors do not push the equity prices up (down) enough. Finally, extrapolation hypothesis supported by Lakonishok, et. al. (1994) assumes that investors believe corporate performance observed in the past will persist at the same level in the future. Therefore, the relation between earnings growth and P/E ratios should be positive. Empirical evidence from London Stock Exchange provides some insights into preceding effects. Specifically, Wu, Li and Hamil, (2012) supported the case of overreaction hypothesis while empirical research by Levis and Liodakis (2001) has evidenced the violation of extrapolation hypothesis. Whether these behavioural effects are subject to the case of FTSE 100, which is part of London Stock Exchange, is an open question of interest.

In this study, we use cross-sectional regressions in an attempt to test the P/E phenomenon (effect) and the predictive power of earnings growth, past market returns and ‘’size effect’’ for a short term period. Similar to the Stefanis analysis (2005), we produce evidence that price-to-earning effect does not hold for the market outlook but only for specific kind of industries, while the firm’s size and earnings growth are key determinants in explaining subsequent stock returns. While we find the market momentum (past market returns) does not drive future stock returns for the overall market, we find some evidence of a positive relation between P/E ratio and following equity returns. In general, the positive slopes of P/E ratios seem to hold for the FTSE 100 suggesting argument of supporting the under-reaction hypothesis suggested by the Jeegadesh and Titman (1993). Furthermore, testing the extrapolation hypothesis introduced by Lakonishok, et. al. (1994) we find little evidence that proves the existence of this behavioural effect.

(6)

6

2. Literature Review

2.1 Macro-level Foresight

Past evidence (Campbell and Shiller (1998, 2001)) on U.S. stock market shows that long run stock market movements can be explained by proxies like price-to-earnings ratio. The foundation of this argument was based on the notion that the tendency of P/E ratio revert to some past benchmark levels after a follow-up time period. Representative evidence provided by Campbell and Shiller (1998) illustrates the predictive power of price-to-earnings with the phrase: ‘’ratios are extraordinarily bearish’’, mentioning the mean revert pattern of stock prices for the next years after assessing the historical financial ratios (price-to-earnings and dividend-to-price) based on a linear pattern of equity prices movements. Few years later, Campbell and Shiller (2001) updated their past work by confirming their past concerns that the smoothed P/E and dividend to price ratio based on the historical average stock prices are able to predict future equity performance, but while assessing its predictability power of explaining future earnings growth found a scarce of indicative power.

(7)

7

In both research analysis, Zorn, et. al. (2009) and Becker, et. al. (2012) have a common set of companies listed in S&P500. Zorn, et. al. (2009) used a smaller time period from 1953 to 2003 instead of a sample period of 1871-2003 investigated by Becker, et. al. (2012). For specified period 1871-1953, the mean reverting pattern of P/E ratios could predict stock or earnings growth changes while macroeconomic factors could predict changes of P/E ratios. This might concern further investigation between the price-to-earnings ratios effect and macro level factors. Of course, to our case, this scope of investigation is left to future research with a focus on the behavior of price-to-earnings ratios for a long-term period.

2.2 Investment Strategy Predictability

Price to earnings ratios and value investment strategies are ideas which were first introduced by Graham and Dodd (1934), as a forefinger of stock performance. The link between the earnings-to-price ratio and fundamental value of equity is subject to different types of investments strategies. Two famous investment strategies usually show up in academic studies: Value investing and Contrarian investing. These types of investment strategies are based on the concept of taking position that will produce remarkable excessive risk-adjusted returns or abnormal returns due to the market anomaly pattern, across different selection of peer portfolio of stocks in the future. Sometimes, these strategies look the same but it might be different with regard to the psychological attitude of a portfolio manager. From a financial point of view, a value investor will invest in low equity prices with outstanding accounting performance characteristics such as low book value or P/E ratios, expecting isolated stock returns in a follow-up period. However, the contrarian investing strategies are also related to the investor’s ‘sentiments’. That means, a contrarian investor could use accounting measures as a value investor does, but also pays attention on ‘’sentiments’’ concerning equities among investors such as sell-side analyst coverage and earnings predictions which are related to the firm’s prospects. Therefore, a contrarian investor may sell (buy) a stock with high (low) P/E ratio because the firm did (not) realize considerable earnings accumulation in previous years.

(8)

8

respect to the announcement of firm’s news. In particular, investors act based on release of good (bad) corporate information and push the stock prices up (down) in first two years. Subsequently, the equity prices will go back to equilibrium and a decline (increase) in stock prices will take place in a following period of two-three years. On the other side, there is also the under-reaction hypothesis suggested by Jeegadesh and Titman (1993). This theory is based on the concept that the past winners outperform the past losers, so a portfolio manager could take a long position on a portfolio formed by previous years winning firms minus the previous years’ losing firms. This portfolio formulation is known as momentum strategy. The association between momentum strategy and under-reaction hypothesis is that investors under-react to the announcement to financial news and information, so stock prices are undervalued and there is an opportunity to gain abnormal returns.

Empirical research on London Stock Exchange (Wu, Li and Hamil, (2012)) supported the argument of the overreaction hypothesis. The authors performed an empirical research on London Stock Exchange using a sample of 1,745 equities for a time period 1970-2009 in order to access the contrarian performance of stock prices. They sorted stock prices by using the Book-to-Market value as a proxy suggested by Fama and French (1992) in order to split stock into three categories: Low, Middle, High. Then, Wu, Li and Hamil, (2012) formed six Loser- minus-Winners portfolios from the interaction of three classifications in order to test two hypotheses: First, losing firms outperform the winning firms for a long term period of 5 years. Second, losing enterprises could generate positive rate of returns. Finally, their results derived from Fama and MacBeth model (1973) model and extension of CAPM model (suggested by Liou 2006) with regard to liquidity factors, they concluded that that first hypothesis is valid for both models, while the second is only for Fama and MacBeth model as after they control for liquidity premium, past losers do not generate positive stock returns anymore. Importantly, they concluded that their empirical findings are in line with the overreaction market hypot hesis.

(9)

9

except those with negative earnings in order to rank portfolios sorted on market capitalization, dividend to price ratio and price to earnings ratio which are determined at the end of the calendar year. Then, the accumulated stock returns were computed from April until the March, and their results suggested, at least for the period 1961-1985, a superior ability of P/E ratios and dividend to price ratios to capture stock returns compared to firm’s size benchmark. Testing the efficient market hypothesis and the relationship between P/E ratio and future stocks returns, Basu (1997) found the low P/E stocks generate superior risk-adjusted returns relative to high P/E stocks listed in NYSE market. His findings, however, conclude that the price-to-earnings effect is a market anomaly rather than a fundamental element by identifying no relation with the ‘’earnings information effects’’. Similar results are documented by Fama and French (1992) who found no statistically significant results between average stock returns and price-to-earnings, whilst they documented a great outperformance of high earnings-to-price with regard to low earnings-to-price ratio, using a sample of 2267 stocks from July 1963 to December 1990.

Stefanis (2005) tested the relationship between the price-to-earnings ratios and stock returns and found evidence that the price-to-earnings phenomenon holds for the Athens Stock Exchange market and is consistent with the framework of ‘’value strategies’’ supported by Graham and Dodd (1934). In particular, Stefanis (2005) documented a negative relationship between equity returns and price-to-earnings ratio, supporting the view of overreaction hypothesis of De Bondt and Thaler (1985,1987) as indicator for chance of capturing isolated gains for a short-term period. Basu and O’Shea (2014) investigated the price-to-earnings ratio as market efficient proxy, using 651 stocks included in New Zealand Stock Exchange and Australia Securities Exchange. Their findings illustrate that the negative trade-off between price-to-earnings ratio and equity returns is explained by market efficiency theory when they regressed quantiles of portfolios from Low P/E to High P/E ratios (also Low-minus-High Portfolios) against Market, SMB, HML, Momentum factor. The price-to-earnings phenomenon was captured by these preceding premiums2. Given the empirical findings that the P/E proxy is related with superior future stock returns, in line with Semi-Strong efficient market theory, we form the following hypothesis:

2_{The factors are defined as follows:SMB is the small cap stocks minus big cap stocks, Market is the market}

(10)

10

(H1): Consistent with the concept of value premium, we assume that the

price-to-earnings ratios should be negatively related with future stock returns. Therefore, we

expect that the investors overreact to news for increased (decreased) earnings by bidding the firm’s equity price too high (low) in the short term and when a correction takes place, stock prices decline (increase) accordingly.

Apart from the predictive power of price-to-earnings ratio to explain future equity returns, there is evidence that supports that P/E ratios are useful indicator of future growth performance. Gordon model is an equity valuation model which can be used as a measure to predict future earnings growth. The basic assumption of the Gordon’s model is that the future equity is positively related with earnings growth. One of the adopters of this view, Lakonishok, et. al. (1994) investigated fundamental metrics such as price-to-earnings, book-to-market, cash flow-to-price, sales growth to assess the assumption: portfolio managers are willing to buy stocks which are characterized by a great past or current corporate performance, so they push the stock prices up regarding the fundamental benchmark by causing a drop in a follow up time period. His empirical evidence shows that investors believed ‘’glamour stocks’’ (equities with exceptional past corporate performance) will continue to persist against value stocks in the future, mentioning the ability of P/E ratios to provide insights into corporate performance. Basu and O’Shea (2014) also documented the P/E indicator as proxy to identify earnings growth performance at least for the next year period. The predictive power of the price-to-earnings ratio to verify future price-to-earnings growth found not to be the case of Athens Stock Exchange market. In particular, Stefanis (2005) found no significant relation between price-to-earnings ratio and future/current growth. Similar empirical evidence fou nd by Levis and Liodakis (2001) who dismissed the Lakonishok’s hypothesis using a dataset of 3,868 publicly listed firms in London Exchange Market for period 1968-1997. Theirs study focuses mainly on framework of extrapolation effects and the association between contrarian strategies, earnings per share predictions, earnings unexpected news, and error-extrapolation effect as well as the effect of earnings unexpected news on contrari an strategies. Given the empirical findings, we formulate the following hypothesis:

(11)

11

3 Data and Research Design

This section describes the models, methodology and data that we use in our analysis. This section is split into four parts: 3.1) Stefanis Model, 3.2) Lakonishok model, 3.3) Data Analysis, 3.4) Descriptive Statistics.

3.1 Stefanis Model

In this stage, we present the model and methodology that are going to be implemented. Similar to Stefanis (2005) methodology, we test the predictive power of price to earnings ratio with the following equity returns. The first econometric model (M1) is described as follows:

Model 1: CRit=αο+βοDummyit-1+γοln(P/E)it-1+εt

Model 2: CRit=αo+βοDummyit-1+γοln(P/E)it-1+δοln(MC)it-1 + εt

Model 3: CRit=αo+βοDummyit-1+γοln(P/E)it-1+δοln(MC)it-1+λοEGit-1 + εt

Model 4: CRit=αο+βοDummyit-1+γοln(P/E)it-1+δοln(MC)it-1+λοEGit-1+θοINDEXit-1 + εt

Model 5: CRit=αο+λοEGit-1 + εt

The dependent variable CRit is the difference of cumulative stock returns calculated

between the calendar year t and t+1. The independent variable, ln(P/E)it-1, is the logarithm

form of price to earnings ratio calculated with three months lag, Dummyit-1 denotes a dum

my variable when the earnings are negative equal to 1, otherwise the dummy is equal to 0. In addition, INDEXit-1 defines the difference of cumulative market returns calculated

between the fiscal year t and t-1 and ln(MC)it-1 is the logarithm form of market

capitalization of each corporation.

The intuition of this model is to test the predictive power of different explanatory variables using the latest public available information to future equity returns for a short term period (1 year holding period). The dummy variable (Dummyit-1)is applied to equate

with the linear pattern between dependent and independent variables, as well as identify if the negative earnings influences the analysis. The market (momentum) factor (INDEXit-1)

(12)

12

3.2 Lakonishok Model

The second model tests the ‘’extrapolation effect’’ hypothesis supported by Lakonishok, et al. (1994). The basic assumption of extrapolation theory is that investors expect that past earnings growth performance will sustain for a short-term period after the financial reports of firms becomes available to the public. Following the methodology of Stefanis (2005) we construct two models, (M2) and (M3), which we use for the case of FTSE 100 in order to test the extrapolation theory hypothesis.

(M2): EG

it

=ζ

ο

+ζ

1

(P/E)

it

+ε

t

(M3): EG

it+1

=η

ο

+η

1

(P/E)

it

+ε

t

In the models above, EGit denotes the earnings growth computed for holding period

of 1 year t, EGit+1 is next year’s earnings growth for the same longevity. The variable

(P/E)it is the price to earnings ratio but it is different with P/Et-1 that

we use for the model (M1).In particular, the (P/E)it is calculated at the end of the year t-1,

when the annual earnings growth of each firm is realized3.

All in all, all models (M1, M2, M3) are going to be tested in line with semi-strong market efficiency hypothesis using the same denominator (earnings per share) to form the price-to-earnings ratio. The predictive aspects of P/E ratio are tested with interaction of candidate variables which are selected based on empirical evidence documented in the literature. The interaction of models (Stefanis and Lakonishok model) is going to provide insights for investors conduct and corporate performance using a specific set of samples for a particular period of time.

3.3 Data Analysis

This sub-section represents the data we use in this study, in order to perform Ordinary Least Squares analysis based on a cross-sectional pattern among common stocks. Firstly, we use data for corporations listed in FTSE 100 for a period of 2009-2015. FTSE 100 is the index that includes the top one hundred enterprises that are publicly listed in

(13)

13

London Exchange market within the highest market capitalization. In the samples, only 92 firms are included because 8 firms dropped out or joined the index later on4_{. The main}

reason why 8 companies are excluded is to control for past market effects (returns). The samples are categorized as follows: 1) a sample that includes all corporations, 2) a sample that excludes financial firms and 3) a set of samples that includes data per industry level. The reason that we separate the sample in many sub-groups is to provide a clear picture of candidate fundamental metrics that we use in research analysis and control for financial industry as well as industrial groups (such as basis metals, non-metallic mineral and concrete) that are highly correlated in term of risk and earnings profile. Based on past academic research, financial and industrial industry follow different accounting practices which affects the behavior of price-to-earnings ratio.5 In our case, we control for each occa sion in order to identify supportive evidence. All accounting data and information are collected from DataStream (Thomson Reuters) database. The stock prices are adjusted for dividend distributions, stock splits, new issues of shares and capitalization of reserves in order to calculate daily returns. Specifically, the analysis of the variables formation that are used in each dataset is described as follows:

Cumulative Returns of Equity (CR) and FTSE 100 (INDEX)

We calculate daily returns from the 1st_{of April of the current calendar year 2010 until the}

31st_{March of next calendar year. Thereafter, we continuously compound the daily returns}

and then we subtract the difference of the last cumulative stock return minus the first stock returns for every year. The formula that is used to calculate the cumulative returns is the following:

CR=(1+R

i1

)(1+R

i2

)(1+R

3

)...(1+R

it

) -1

where

R

i1

,R

i2

,R

i3

...R

it are the returns of stock

i

at day

‘’t’’

.

Using this formula we explicitly take into consideration the daily volatility of stocks.

Both cumulative stock returns of firms (CR) and share index FTSE 100 (INDEX)

4_{Eight firms were excluded: TUI , Worldplay, Coca-Cola, Merlin Entertainment,Royal Mail, Int.Cons.Airline}

Group, Glencore, Direct Line In. Group. FTSE 100 includes 101 stocks as the Dutch royal Shell Group offers two types of stock (A and B) that differ with regard the way of taxation and dividend policy mechanism. We decide to continue with both as their stock prices are not exactly the same, so therefore it might affect the behavior of P/E ratios.

5_{Stefanis (2005) excluded the financial related firms by arguing these companies apply different accounting}

(14)

14 are calculated based on the way we described above6_.

Price to Earnings Ratio (P/E)

Following Basu (1997), Basu and O’shea (2014), Fama and French (1992) and Stefanis (2005) P/E ratio is used as proxy for future stock returns. The formation of price-to-earnings is calculated by using the closing share price at 31th of March of the current fiscal year as the numerator and the earning per share (EPS) of 31 of December of previous calendar year as the denominator7. In essence, we calculated the price-to-earnings ratio with a three months lag in order to link price and earnings with the semi-strong market hypothesis which states that all public financial information and historical price data are quickly incorporated in the stock prices. Furthermore, this formation of P/E leads to a robust identification of subsequent equity movements as its definition is not suffered from ‘’look ahead’’ bias. Finally, logarithms are applied to smooth the value of observations.

Market Capitalization (MC)

The variable market capitalization (MC) is calculated every day by using last closing stock price, after controlling for new stock issues or changes in capital. Afterwards, we multiply the stock price times the relevant amount of common stock outstanding. The figures of market capitalization of each firm is observed at the end of 31 March. In each regression analysis, we use logarithms formula on market value of firms in order to smooth the figures and reduce the divergence of data points in our sample.8 Market value is used to due to its strong explanatory power of equity returns (Fama and French (1992), Stefanis (2005), Basu and O’Shea (2014)).

Earnings Growth (EG)

The earnings growth of year t is calculated from the difference between EPS for the fiscal year t and EPS for the fiscal year t-1, both divided by the EPS for the calendar year t-1

6_{In essence, we observe cumulative market returns with one year lag, therefore we start calculate daily market}

returns from 1st_{April 2009 ultil 31}st_{March 2010. Past cumulative market returns drive future stock returns}

(Stefanis(2005)).

7_{The EPS is directly downloaded using Datastream database. The relevelant describion states that the Earnings is}

calculated by substracting taxes and preffered dividend from net income, but before extraordinary items. Thereafter, the EPS are calculated based on the earnings divided by the weighted average of common stock outstanding.

8_{The application of logarithms is in line with Stefanis (2005) methodology who supported his framework based on}

(15)

15

for each stock. Earnings growth is chosen due to its predictive ability to capture future equity returns (Stefanis (2005)).

3.4 Descriptive Statistics

Table 1 indicates the descriptive statistics of dependent and explanatory variables for 92 firms listed in FTSE 100 for the period 2009-2015. The cumulative returns reach a mean of 0.138 with standard deviation remains on 0.257 for a period of 5 years. FTSE 100 includes only the 100 companies in London Exchange Market with the highest market capitalization. The magnitude of market value of corporations is apparent as the average and median is around 16.923 and 6.827 billion British pounds respectively. The mean and median of P/E ratios is 24.553 and 15.228 respectively. A close look at average of earnings growth is low (0.578) for time period of 5 years. The reason is that we start computing the change of earnings per share from 2008 until 2009 in order to formulate the first earnings growth variable. This year, many companies realized negative earnings growth, therefore the mean of earnings growth looks unusual. Moreover, financial firms have influence on the sample, as they reported unstable earnings growth during the time period 2008-2015. If we exclude financial sector, the average of the next year’s earnings growth (EGt+1) climbs to 0.999. In addition, we have excluded eight companies, so

(16)

16

frequency: 31 March – 1 April) are realized by Mining Group (-0.040), while the highest cumulative returns are realized by Industrial Group (0.236). Finally, Industrial industry has also the second highest earnings growth (on average) which is around 1.393.

Table 1: Descriptive Statistics – Model (M1),(M2),(M3)

Statistics CR Dummy P/E MC EG EGt+1 INDEX Mean 0.138 0.092 24.553 16923.720 0.578 0.769 0.110 Median 0.112 0.000 15.528 6827.680 0.113 0.138 0.029 Maximum 1.402 1.000 1292.000 130414.900 88.000 88.000 0.439 Minimum -0.472 0.000 0.000 495.290 -20.500 -20.500 -0.041 Std. Dev. 0.257 0.290 69.472 22648.330 4.757 5.222 0.170 Observations 465 465 465 465 465 372 465 Table 1 indicates the descriptive statistics of both dependent and independent variables that were used in models (M1), (M2), (M3) in regression analysis. The time period for our dataset is between 2009 and 2015. No logarithms are applied to any of each variable above. CR represents the cumulative returns from a holding period of 1 year, P/E is the trailing price-to-earnings ratio, Dummy is a dummy variable that is equal to 1 when earnings are negative, otherwise it is zero, MC is the market capitalization of listed firms, INDEX is the market momentum, EG is the earning growth and finally EG+1 is earnings growth of the next year. The figures are reported in term of million British pounds.

Furthermore, table 2 illustrates the correlation matrix of model (M1). It is apparent that correlation matrix represents a positive correlation coefficient between annual cumulative returns and logarithm forms of P/E ratios (p-value= 0.070), while cumulative returns have a highly negative relation with ‘’size effect’’ (ln(MC)) (p-value= -0.407). Consistent with Stefanis (2005), we find positive correlation between earnings growth and cumulative returns while past cumulative market returns found to have a negative trade-off with regard to stock returns.

Table 2: Correlation Matrix (M1)

(17)

17

Moreover, we apply the Augmented Dicky-Fuller (ADF) unit test to test unit root in the series for sample set which includes all companies. Past studies related to P/E behaviors have examined the stationary pattern of P/E ratios and the other performance variables (which are included in their research design) using Augmented Dicky-Fuller(ADF) unit test9.The Dicky-Fuller (ADF) is implemented before and after the application of logarithms on MC and P/E ratios. We checked every variable and we find no evidence of non-stationary pattern at any level of statistical significance (1%,5%,10%). The p-values are zero for each variable included in the sample.

4 Empirical Results

4.1 Relation Between P/E Ratios and Equity Returns

In this section, we discuss the empirical findings for the econometric models (M1),(M2),(M3). Table 3 illustrates the results derived from regression model (M1), providing some evidence that the price-to-earnings ratio is positively related with following stock returns in model 1 and 2. White heteroscedasticity consistent standard error estimation method was implemented in cases that there was evidence of heteroscedasticity in order correct the standard errors. We implement five different set of regression models in order to shed light on the additivity power of each variable to enhance the robustness of the econometric model. Overall, we decide to continue with model three and four, as they provide the Best Linear Unbiased Estimator (BLUE) after applying white heteroscedasticity consistent standard errors. Moreover, we control for negative P/E ratios by adding a dummy variable.

(18)

18

Table 3: Regression Model (M1)

Model Alpha Dummy Ln(P/E) Ln(MC) EG INDEX 1 Coefficient -0.027 0.203 0.057 t-statistic (-0.408) (2.337)** (2.284)** 2 Coefficient 0.803 0.125 0.038 -0.085 t-statistic (6.967)*** (1.619) (1.694)** (-8.867)*** 3 Coefficient 0.822 0.072 0.022 -0.083 0.009 t-statistic (7.508)*** (1.409) (1.230) (-8.624)*** (6.778)*** 4 Coefficient 0.832 0.087 0.023 -0.084 0.009 -0.080 t-statistic (7.633)*** (0.243) (1.286) (-8.780)*** (6.439)*** (-1.148) 5 Coefficient 0.131 0.012 t-statistic (9.616)*** (8.100)*** Model No. 1 2 3 4 5 Observations 455 455 455 455 455 F-statistic 6.055 33.007 31.175 25.283 29.765 Prob.(F-statistic) 0.002 0.000 0.000 0.000 0.000 R-squared 0.027 0.176 0.213 0.215 0.060 Adj. R-squared 0.023 0.171 0.206 0.207 0.058 Autocorrelation 5% Robust Robust Robust 5%

Heteroscedasticity 5% 1% 1% 1% Robust

Linearity 10% 1% Robust Robust Robust

The tables 3 indicates a specific set of explanatory variables that included in cross-sectional analysis with regard to the cumulative returns. The dependent variable is CR: cumulative returns while the explanatory variables are as follows: P/E: the price-to-earnings ratio, Dummy: dummy variable is one when the price-to-earnings ratio is negative, otherwise it is zero. MC: the market capitalization, EG: earnings growth, INDEX: cumulative returns of index with one-year lag. The time period is from 2009 to 2015. In each case below we perform Serial Correlation test LM, White Heteroscedasticity test and Ramsey test to assess the robustness of the results. The t-statistic results are reported after applying a White Heteroscedasticity-consistent standard errors & covariance regression when White Heteroscedasticity test is not robust at 10% confidence level. *Confidence at 90% level, **Confidence at 95% level, ***Confidence at 90% level. Regarding the robustness test, we report the level of significance for p-value when the tests are not robust at 10% level otherwise we declare ‘’Robust’’. In particular: 10% means: 5%<p-value<10%, 5% means: 1%<p-value<5%, 1% means: 1%<p-value and Robust means p-value>10%.

(19)

19

(EG) variable the model becomes BLUE (Best Linear Unbiased Estimators). The EG is 0.009 and statistically significant at 1% level. Interestingly, EG absorbs the significance power of P/E ratio while the MC remains statistically significant at level of 1%. The slope of EG shows that each change of one unit in earnings growth, cumulative returns (CR) changes nine units in model 3. Moreover, it apparent that when we add market momentum the Adjusted R-squared increases from 0.206 to 0.207 but the factor is insignificant. Finally, we test the effects of cumulative returns directly to EG variable in order to test the statistical power of the factor. The econometric results show that the model 5 suffers from serial correlation. After applying Newey-West Approach the EG factor is 0.012 and statistical significant at 1%. In general, EG factor is an important explanatory variable that satisfies Gauss Markov assumptions after controlling for heteroscedasticity or serial correlation violation. Furthermore, earnings growth variable kills the P/E ratio and past market effects do not influence future equity returns.

Following the Stefanis (2005) methodology, we implement the same econometric models (1-5) per industry level. We also use a separate sample which excludes financial firms. The main reason is to control for potential influence of financial reporting rules and business profiles. The empirical findings are the following:

Price to Earnings ratio: An examination of model 1-4 per industry level illustrates that

there is some evidence of P/E phenomenon. For three out of ten industries, we find that

P/E ratio is statistically significant with following stock returns across the model 1-4. For

two out of three industries the P/E ratio is positive while only for Oil & Gas industry we find a negative coefficient of P/E ratios. It is clear that the P/E phenomenon does not hold for the most of industry levels but a reason that we report some evidence of statistical significance in our initial sample is due to the Industrial Group. In particular, all P/E ratios are positive and statistical significance at least on the level of 5% as well as all models 1-4 are robust. A first glance at Non-financial results, it is obvious that the coefficient of P/E ratios is 0.060 and statistically significant at 5% significance level for only model 1. Turning to Financial group, we find the P/E ratios to be not statistic al significant at any confidence level.

Market Capitalization: In most cases, we find evidence that MC is negatively and

(20)

20

Industrial and Financial group which include 61 companies out of 93 of our initial sample. Their MC variable is negative and statistically significant at level of 5% (at least), as well as model 3 and 4 are BLUE (for the case of financial sector we controlled for heteroscedasticity). A first look at Non-financial firms, the coefficient of MC is negative and statistically significant at 1% level for all relevant models. Despite the exclusion of Financial firms model 4 and 5 are still Best Linear Unbiased Estimators. Interestingly, both R-squared and adjusted R-squared are increased when we drop out financial sector from my initial group. Surprisingly, we find that the MC variable is positively related with future stock returns for the case of Utilities. Specifically, the market value is 0.270 and statistical significant at 5% level after controlling for the heteroscedasticity. No relation between market capitalization and stock returns can be verified for Consumer Goods, Telecommunication, Pharmaceutical and Technology industry.

Earnings Growth: Removing the financial companies from the original sample, the strong

pattern of earnings growth remains positive and statistically significant at 1% level. In Appendix, the coefficient of earnings growth for financial sector is positive and statistically significant at least on level of 10% level across model 3-5. When we add the cumulative market returns it becomes positive and statistical significance at 1% level, as well as the Adjusted R-squared increases from to 0.137 to 0.138. Four out of ten industries do provide evidence of a positive relation of previous earnings growth and cumulative stoc k returns. Three of these industries represent the majority of observations in our initial sample that provide an argument why earning growth factor is strong for the whole market pattern.

Cumulative market returns: While INDEX is not statistically significant on our initial

(21)

21

4.2 Relation Between P/E Ratio and Earnings Growth

In this step, we test the second hypothesis using models (M2) and (M3). The concept is based on the extrapolation hypothesis suggested by Lakonishok, et. al. (1994) and assumes that investors over-extrapolate past earnings growth into the future. His main hypothesis is also in line with Gordon Model which an increase in earnings growth should not cause a decrease in the fraction of P/E (at least in long term) as the price is expected to increase in the near future. In other words, a positive earnings growth is supposed to persist and generates superior equity returns in the future. Past evidence shows that extrapolation effect does not hold for the case of London Stock Exchange, but we need to verify if the pattern does change for the case of FTSE 100.

Table 4: Regression model (M2) – Current Growth

Sample F- Adj. Number of Category Alpha P/E Statistic R-squared Observations

All companies Coefficient 0.608 -0.001

t-statistic (2.854)** (0.008) 0.137 -0.001 465

Non-Financial Coefficient 0.789 -0.002 t-statistic (2.660)** (-0.517) 0.267 -0.002 350

Consumers Goods Coefficient 0.439 -0.024

t-statistic (1.943)* (-2.229)** 4.970 0.092 40 Consumer Services Coefficient 0.879 -0.088

t-statistic (3.280)*** (-2.500)** 6.254 0.088 55 Financials Coefficient -0.216 0.024 t-statistic (-0.634) (1.234) 1.545 0.004 115 Industrials Coefficient 1.471 -0.002 t-statistic (1.829)* (-0.310) 0.096 -0.007 125 Technology Coefficient 0.742 -0.016 t-statistic (2.137)* (-1.924)* 3.701 0.230 10 Mining Coefficient 1.017 -0.009 t-statistic (1.552) (-0.355) 0.126 -0.026 35 Oil & Gas Coefficient 0.612 -0.017

t-statistic (1.598) (-1.053) 1.109 0.003 35 Pharmaceutical Coefficient -0.154. -0.025 t-statistic (0.356) (1.275) 1.627 0.025 25 Telecommunication Coefficient 0.304 -0.008 t-statistic (1.821) (-0.224) 5.021 0.223 15 Utilities Coefficient 0.021 0.031 t-statistic (0.013) (0.299) 0.089 -0.115 10

(22)

22

Table 4 illustrates the empirical results of explanatory power of P/E to predict current earnings growth. Based on the results on table 4, we find that all the models are BLUE (Best Linear Unbiased Estimators). The econometric results indicate that there is no relationship between P/E and current growth when we perform a cross-sectional analysis for the sample which includes all the corporations. However, there is some evidence of the predictive power of P/E ratio. For instance, the coefficient of P/E ratio for Consumer Goods and Consumer Services is -0.024 and -0.088 as well as statistical significant at 5%. In addition, the coefficient of Technology Group is -0.016 while adjusted R-squared is 0.230, showing a weak but considerable fit in the regression model compared to the rest of the industries.

(23)

23

Table 5: Regression Model (M3) – Future Growth

Sample F- Adj. Number of Category Alpha P/E Statistic R-squared Observations

All companies Coefficient 0.736 0.001

t-statistic (2.566)** (0.355) 0.126 -0.002 372 Non-Financials Coefficient 0.854 0.005

t-statistic (3.657) (0.684) 0.231 0.001 284 Consumers Goods Coefficient -0.307 0.018

t-statistic (-1.367) (1.717)* 2.949 0.059 32 Consumer Services Coefficient 0.833 -0.009

t-statistic (2.735)* (-0.997) 0.994 -0.000 44 Financials Coefficient 0.019 0.040 t-statistic (0.048) (0.207) 0.043 -0.010 92 Industrials Coefficient 1.021 0.021 t-statistic (1.688)* (0.808) 0.004 0.070 100 Technology Coefficient -0.052 0.003 t-statistic (-0.085) (0.144) 0.021 -0.166 8 Mining Coefficient 1.735 -0.035 t-statistic (2.275) (-1.259) 1.069 0.039 28 Oil & Gas Coefficient 0.075 0.007

t-statistic (0.189) (0.257) 1.587 0.021 28 Pharmaceutical Coefficient 0.157 0.003 t-statistic (0.367) (0.064) 0.026 -0.053 20 Telecommunication Coefficient 0.353 -0.011 t-statistic (0.859) (-0.776) 0.056 -0.037 12 Utilities Coefficient 3.395 -0.196 t-statistic (2.040) (-1.877) 0.054 0.399 8

The table above illustrates the cross-sectional regression results separated into different categories: All companies, Nonfinancial, ten Industries. The time period is from 2009 until 2015. *Confidence at 10%, **Confidence at 5% level and ***Confidence at 1%. EGt+1 is earnings growth for year t+1 as dependent variable while P/E is the price-to-earnings ratio in year t.

4.3 Empirical Discussion

In this section, we discuss the empirical findings of this study compared to previous empirical findings. Before we start to explore the empirical results, we remind again on this point that our case is FTSE 100 and not the London Stock Exchange. Therefore, the results should be examined with caution as FTSE is known as an index which includes big stocks (high market value). Therefore, any argument of value or contrarian investing is considered with respect to peer firms listed in FTSE 100 index.

(24)

24

returns are well documented in international markets. However, evidence regarding the use of P/E as a market efficiency proxy is controversial. Our survey provides some evidence that price-to-earnings phenomenon exists for specific firm sectors and not for the whole market outlook. In general, hypothesis (H1) seems to be violated as our evidence is not strong. Our empirical results illustrate that there is a strong positive relation between price-to-earnings ratios and future stock returns for at least two out of ten industrial sectors which includes a large number of corporations listed in FTSE 100. These evidence is contrary to previous empirical results. In particular, Stefanis (2005) and Basu and O’Shea (2015) found a negative relationship between P/E ratios and subsequent year stock returns. Therefore, their findings propose that investing in low P/E ratios could generate superior rate of returns compared to high P/E ratios. Obviously, their results are clearly in line with value framework suggested by Graham and Dodd (1934). Our empirical evidence regarding the P/E ratio is inconsistent with these findings because the P/E ratio found to be positively related with future equity returns. In other words, investing in growth stock portfolios (high P/E ratios) appears to be more profitable with regard to the value portfolios (low P/E ratios).

Based on the above evidence, we reject (H1), as the P/E ratio is positively related with future stock returns and a value investing strategy in large cap type index is not a profitable option. Put it differently, if an investor has tried to capture value premium by forming portfolios on FTSE 100, the investor would have generated losses for the holding period 2009-2015.

(25)

25

gh portfolios. Theirs findings show a positive and statistical significant SMB (small cap stocks minus large cap stocks) factor for each year, to prove the ‘size effect’ for Australian and New Zealand Stock Exchange.

The EG variable plays a key role in our cross-sectional analysis. When we add earnings growth to the regression the statistical power increases and the regression model

(M1) becomes Best Linear Unbiased Estimators (BLUE). The results are consistent with

findings of Stefanis (2005) who found also a strong positive earnings growth relation with following equity returns. The positive sign of EG denotes an argument against contrarian investment. As earnings growth are translated in terms of change in earnings per share (EPS), an implementation of strategy against the expected EPS would have resulted in negative rate of returns for the time period 2009-2015.

Past market returns are not related with future stock returns in our initial sample. However, a cross-sectional analysis per industry level shows statistical significant results for 3 out of 10 sectors. Our results are inconsistent with Stefanis (2005) who reported a past market effect to drive a negative future stock performance in Athens Stock Exchange. Testing the Lakonishok, et. al. (1994), extrapolation hypothesis, we find no strong results for the market outlook. Therefore, we cannot reject (H2), at least for the whole market outlook. However, testing the industry separately, we find the P/E ratios of three out of ten industries are negatively related with current earnings growth. This evidence is contrary to Lakonishok, et. al. (1994) and Gordon’s model which states that seeking in past earnings growth performance will persist in the future. However, we find a reverse pattern of Consumer of Goods sector to be consistent with the preceding theories the Basu and O’Shea findings. As we regress the future earnings growth against P/E ratio, we find a positive relation that is in line with the two previous models.

(26)

26

proportion of P/E ratios up. In essence, this relevance behavior of price-to-earnings ratio could be verified if investors do not correct the equity prices with respect to fundamentals when the release of earnings growth news takes place, as their psychological attitude indicates a more ‘’conservative’’ reaction that results in an opportunity for abnormal returns because stock prices are already undervalued.

5. Conclusion

This paper adds value to the current literature by revisiting the predictive aspects of P/E ratio for large capitalization companies. This study empirically tests the relation between price-to-earnings ratio and future equity returns, as well as earnings growth based on a sample of firms that are listed in FTSE 100 United Kingdom for the time period 2009-2015. In general, (H1) is violated because we find no statistically significant (positive) relation between P/E ratios and equity returns. Moreover, (H2) cannot be rejected, as the coefficients of P/E ratios are statistically insignificant.

In addition, the ‘’size effect’’ holds for the companies listed on FTSE 100 and the additivity of EG plays a key role in explaining future equity returns. Controlling for previous year’s cumulative market returns, we find not strong significance influences for the following equity returns.

Excluding the financial firms from our sample generates robust results. Stefanis (2005) mentioned that the application of different accounting practices influences the analysis, as it is true in our case, as the model 3 and 4 result in greater results, when we exclude financials from the sample. Therefore, exclusion of financial industry might be wise choice when we test the P/E effect.

Finally, the positive (and statistically significant) coefficient of P/E ratios of some of industry sectors supports the under-reaction theory supported by Jeegadesh and Titman (1993) rather than overreaction theory (De Bondt and Thaler (1985,1987)). These evidence might suggest that investors do not react quickly to corporate news as a result ‘’good news’’ (bad news) for enterprise’s earnings could lead to abnormal returns by taking a long (short) position immediately after the announcement of financial statements.

(27)

27

the techniques used in the current study with more advanced econometric models to increase the reliability between P/E ratio and equity returns and/or earnings growth. Finally, a more diversified time-period could be used to test these hypotheses as the samples are restricted for the period after the financial crisis (2009-2015).

Open Questions

(28)

28

Appendix

Part 1A: Stefanis Model

This part of appendix includes all information related to industry levels. In particular, this part is split into sub-parts: Subpart 1 and Subpart 2. In Subpart 1, every variable is reported without any application of logarithmic forms in order to provide clear insights of firm’s characteristics. In Subpart 2, we apply logarithm forms to P/E ratios and MC as we implement in regression model (M1). CR is the cumulative returns for one-year period, P/E is the price to earnings ratio, Dummy is a dummy variable that is employed when the P/E ratios are negative otherwise it is zero. MC is the market value of corporation, Index is the past cumulative market returns, EG is the earnings growth of the current year and EGt+1 is the earnings growth of the next year.

Subpart 1A

Table 1A: Non Financials

CR P/E Dummy MC INDEX EG EGt+1

Mean 0.141 27.77 0.037 17411.030 0.109 0.744 0.999 Median 0.106 16.93 0.000 7592.410 0.029 0.099 0.128 Maximum 1.402 1292.000 1.000 118570.800 0.439 88.000 88.000 Minimum -0.377 0.000 0.000 495.290 -0.041 -12.016 -1.534 Std. Dev. 0.261 7.866 0.189 21901.480 0.170 5.258 5.770 Observations 350 350 350 350 350 350 248

Table 2A: Consumer Goods

(29)

29 Table 3A: Consumer Services

CR P/E MC INDEX EG EGt+1

Mean 0.159 20.03 8346.526 0.109 0.524 0.651 Median 0.130 15.51 5909.505 0.029 _0.128 _0.142 Maximum 0.999 196.93 35047.73 0.439 7.500 7.500 Minimum -0.333 6.87 634.100 -0.041 -4.572 -0.957 Std. Dev. 0.281 2.6193 7899.298 0.171 1.765 1.616 Observations 50 50 50 50 ₅₀ ₄₄

Table 4A: Financial Sector

Mean 0.126 15.95 0.130 15440.590 0.109 _0.074 _0.072 Median 0.120 13.40 0.000 5203.280 0.029 _0.179 _0.186 Maximum 0.856 116.43 1.000000 130414.900 0.439 _13.091 _13.091 Minimum -0.472 0.000 0.000000 1178.840 -0.041 _-20.500 _-20.500 Std. Dev. 0.243 1.559 0.338255 24826.540 0.170 2.671 2.910 Observations 115 115 115 115 115 115 92

Table 5A: Industrial Sector

Mean 0.231 43.12 0.040 6342.616 0.109 1.393 1.921 Median 0.232 18.68 0.000 4162.640 0.029 0.136 0.185 Maximum 1.402 1292.00 1.000 52079.430 0.439 _88.000 _88.000 Minimum -0.284 0.000 0.000 495.290 -0.041 _-12.016 _-0.995 Std. Dev. 0.302 12.809 0.196 7974.598 0.170 _8.502 _9.339 Observations 125 125 125 125 125 ₁₂₅ ₁₀₀

Table 6A: Technology Sector

Mean 0.126 32.47 6617.480 0.109 0.232 0.029 Median 0.145 24.74 5639.600 0.029 _0.166 _0.030 Maximum 0.255 104.52 11515.080 0.439 _1.953 _0.923 Minimum -0.052 14.10 3174.140 -0.041 _-0.835 _-0.835 Std. Dev. 0.091 2.683 3214.442 0.179 0.809 0.608 Observations 10 10 10 10 ₁₀ ₈

Table 7A: Mining Sector

(30)

30 Table 8A: Oil & Gas Sector

Mean 0.001 14.54 0.085 47031.190 0.109 _0.339 _0.164 Median -0.010 11.57 0.000 39427.020 0.029 _0.082 _0.059 Maximum 0.337 71.69 1.000 118570.80 0.439 _9.512 _1.813 Minimum -0.284 0.000 0.000 10181.980 -0.040 _-0.874 _-0.937 Std. Dev. 0.136 1.399 0.284 30619.660 0.172 _1.668 _0.656 Observations 35 35 35 35 35 ₃₅ ₂₈

Table 9A: Pharmaceutical Sector

Mean 0.174 23.89 26502.590 0.109 _0.356 _0.222 Median 0.164 23.07 11158.250 0.029 0.202 0.109 Maximum 0.815 77.88 77432.690 0.439 _2.705 _2.612 Minimum -0.107 6.10 1222.350 -0.041 _-0.590 _-0.590 Std. Dev. 0.225 1.417 26840.870 0.173 0.847 0.758 Observations 25 25 25 25 ₂₅ ₂₀

Table 10A: Telecommunication Sector

Mean 0.141 23.53 34525.620 0.109 0.096 0.164 Median 0.153 13.95 17796.200 0.029 _0.043 _0.059 Maximum 0.526 137.68 92375.250 0.439 1.813 1.813 Minimum -0.276 4.61 2066.930 -0.041 _-0.937 _-0.937 Std. Dev. 0.225 3.342 36262.290 0.175 0.610 0.656 Observations 15 15 15 15 15 12

Table 11A: Utilities Sector

Mean 0.101 14.54 4102.910 0.109 _0.459 _0.781 Median 0.101 13.61 4080.090 0.029 _-0.090 _0.002 Maximum 0.211 23.61 5461.920 0.439 _5.293 _5.293 Minimum 0.005 8.60 2857.950 -0.041 _-1.275 _-0.371 Std. Dev. 0.064 0.475 715.308 0.179 _1.812 _1.889 Observations 10 10 10 10 ₄₀ ₃₂

Subpart 2A

Table 1AA: Non-Financial Group

CR Dummy Ln(P/E) Ln(MC) EG INDEX

(31)

31 Table AA2: Consumer Goods Industry

CR 1.000 Dummy 0.040 1.000 Ln(P/E) 0.014 -0.974 1.000 Ln(MC) -0.140 -0.199 0.194 1.000 EG -0.002 -0.443 0.388 -0.043 1.000 INDEX -0.443 -0.016 -0.049 -0.112 0.039 1.000

Table AA3: Consumer Services Industry

CR 1.000 Dummy 0.259 1.000 Ln(P/E) -0.172 -0.563 1.000 Ln(MC) -0.511 -0.404 0.268 1.000 EG 0.068 0.405 0.414 -0.135 1.000 INDEX -0.083 0.264 0.112 -0.138 0.393 1.000

Table AA4: Financial Industry

CR 1.000 Dummy -0.046 1.000 Ln(P/E) 0.084 -0.895 1.000 Ln(MC) -0.361 0.106 -0.141 1.000 EG 0.222 0.132 -0.119 -0.147 1.000 INDEX -0.029 0.272 -0.208 -0.066 0.211 1.000

Table AA5: Industrial Industry

CR 1.000 Dummy -0.065 1.000 Ln(P/E) 0.185 -0.579 1.000 Ln(MC) -0.396 -0.036 0.046 1.000 EG 0.381 -0.049 -0.007 -0.171 1.000 INDEX -0.029 0.083 -0.081 -0.189 0.036 1.000

Table AA6: Mining Group

CR 1.000 Dummy -0.104 1.000 Ln(P/E) 0.090 -0.555 1.000 Ln(MC) -0.328 0.200 -0.140 1.000 EG 0.501 0.072 0.205 0.008 1.000 INDEX 0.514 -0.069 0.203 -0.015 0.415 1.000

Table AA7: Oil & Gas Industry

(32)

32 Table AA8: Pharmaceutical Industry

CR Ln(P/E) Ln(MC) EG INDEX CR 1.000 Ln(P/E) 0.558 1.000 Ln(MC) -0.321 -0.440 1.000 EG -0.112 -0.097 -0.095 1.000 INDEX -0.165 -0.077 -0.052 0.321 1.000

Table AA9: Technology Industry

CR Ln(P/E) Ln(MC) EG INDEX CR 1.000 Ln(P/E) -0.321 1.000 Ln(MC) -0.384 0.326 1.000 EG -0.277 0.811 -0.066 1.000 INDEX 0.184 -0.250 -0.250 -0.128 1.000

Table AA10: Telecommunication Industry

CR Dummy Ln(P/E) Lm(MC) EG INDEX

CR 1.000 Dummy 0.429 1.000 Ln(P/E) -0.408 -0.664 1.000 Lm(MC) -0.054 -0.098 -0.187 1.000 EG 0.028 -0.059 -0.130 0.298 1.000 INDEX -0.016 0.518 -0.229 -0.037 -0.008 1.000

Table AA11: Utilities Industry

CR Dummy Ln(P/E) Lm(MC) EG INDEX

CR 1.000 Dummy 0.607 1.000 Ln(P/E) -0.550 -0.953 1.000 Lm(MC) -0.035 -0.690 0.712 1.000 EG -0.053 0.002 0.025 0.029 1.000 INDEX 0.164 0.646 -0.490 -0.512 0.068 1.000

Part 1B: Lakonishok Analysis

(33)

33

Subpart 1B

Table 1B: All companies

EG EGt+1 P/E Mean 0.578 0.769 22.999 Median 0.113 0.138 14.644 Maximum 88.000 88.000 1130.000 Minimum -20.500 -20.500 0.000 Std. Dev. 4.757 5.222 61.915 Observations 465 372 465

Table 2B: Consumer Goods

Table 3B: Consumer Services

Table 4B: Telecommunication Sector

Table 5B: Non financials

(34)

34 Table 6B: Industrial Sector

Table 7B: Financial Sector

Table 8B: Mining Sector

Table 9B: Utilities Sector

EG EGt+1 P/E Mean 0.459 0.781 13.752 Median -0.090 0.002 14.367 Maximum 5.293 5.293 21.710 Minimum -1.275 -0.371 0.000 Std. Dev. 1.812 1.889 5.993 Observations 40 32 40

Table 10B: Technology Sector

(35)

35 Table 11B: Pharmaceutical Sector

Table 12B: Oil & Gas Sector

Subpart 2B:

Consumer good Consumer Services

EG EGt+1 P/E EG EGt+1 P/E

EG 1 EG 1

EGt+1 -0.073 1 EGt+1 0.004 1

P/E -0.092 0.299 1 P/E -0.381 -0.15 1

Non-Financial Financial

EG EGt+1 P/E EG EGt+1 P/E

EG 1 EG 1

EGt+1 -0.018 1 EGt+1 0.011 1

P/E -0.026 0.072 1 P/E 0.112 0.022 1

Technology Pharmaceutical

EG 1 EG 1

EGt+1 0.639 1 EGt+1 -0.467 1

P/E -0.387 0.059 1 P/E 0.261 0.039 1

All companies Industrials

EG EGt+1 P/E EG EGt+1 P/E

EG 1 EG 1

EGt+1 0.004 1 EGt+1 -0.034 1

P/E -0.014 0.018 1 P/E -0.037 0.283 1

Telecommunication Mining

EG 1 EG 1

EGt+1 -0.245 1 EGt+1 0.497 1

P/E -0.369 -0.238 1 P/E -0.094 -0.240 1

Oil & Gas Utilities

EG 1 EG 1

EGt+1 -0.001 1 EGt+1 -0.372 1