1
Can Online Employee-Submitted Firm Reviews Predict
Firm Performance? A New Perspective Utilizing the
Modern Age of Transparency
Master’s Thesis
Date: May 1st 2016
Program: MSc Business Economics, Finance track Student Name: Jeffrey Bloemen
Student Number: 10381287 Supervisor: Dr. E. Eiling
2
Statement of Originality
This document is written by Student Jeffrey Bloemen who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents
3
Table of contents
1. Introduction 4
2. Theoretical background 6
2.1 Previous research 6
2.2 A new era of human capital valuation 8
2.3 Employee satisfaction 10 3. Methodology 12 3.1 Portfolios 13 3.2 Event study 18 4. Data 21 5. Results 26 6. Conclusion 37 References 39
4
1. Introduction
Employees are the driving force behind a firm’s performance. It is their level of commitment that plays a determinant role in a firm’s success or failure. As ‘insiders’, employees
continuously have a look behind the scenes and are therefore confronted with information that is unobservable to most investors. The internet has provided the (ex) employee with the opportunity to quickly and efficiently spread his opinion regarding the firm in question, allowing outsiders to take these views into consideration when evaluating a firm’s
performance. In this thesis, I test whether these online reviews could be deemed valuable for investors by analyzing the relationship between employee-submitted firm review ratings and firm performance, over the period of 2009 to 2015. If a relationship were to hold, investors could use available review data to realize abnormal returns on their stock portfolio. The annual ‘Best Place to Work’ data that is available on the two prominent business websites Fortune.com and Glassdoor.com is used to test whether portfolios composed of these Best Workplace stocks realize significant excess returns. Each year, the 100 and 50 highest rated companies make it to the Best Workplace list of Fortune® and/or Glassdoor® respectively. It can be argued that the employees of these companies are
relatively satisfied with their employer and the objective of this thesis is to test whether this satisfaction is related to higher firm performance. This thesis attempts to answer the
following research question:
“High employee satisfaction is related to significant excess returns”
There exists a substantial amount of literature in the psychological field that analyzes the relationship between employee satisfaction and employee performance. Where last
century’s literature didn’t value human capital as high as they valued physical assets, recent literature suggests that human capital might be the most valuable asset a firm could own. I mention three recent studies that analyze the relationship between employee satisfaction and firm performance using employee satisfaction data from abovementioned websites, all of which find a position relationship between the two variables. The research for this thesis differs from the research conducted in these studies in the fact that I use employee
5 robustness of the results. In addition, I use a different research model: I opted for Carhart’s four-factor model whereas two of the three recent studies used Tobins’ q as a measure of firm performance. I use this model instead because I consider the four-factor model to be of higher interest for the investor as it compares the returns of my portfolio directly with the returns of the market. As “beating the market” has become an important performance indicator for both individual and institutional investors, I consider this to be the better measure of firm performance for investors. My research is based on the third study that uses the four-factor model as well and it differs in the fact that I construct five different types of portfolios instead of one. In addition, I use both websites’ satisfaction level information rather than just using Fortune’s and the sample period is more recent.
Using the annual Best Workplace data, I compose five different portfolios for both websites. The portfolios rebalance at the publication date of a new list, selling the stock of firms who did not make it to the new list and buying the stocks of firms who did make it to the list. The weight invested in each stock is either equal or depends on either the firm’s rank on that list or its market capitalization at the date of publishing. In addition, I form another portfolio that does not rebalance after a new list publication but rather keeps all stocks until the end of the sample period (12/31/2015). Lastly, I form a portfolio that buys the stocks of firms who made it to both Fortune’s and Glassdoor’s Best Workplace list in a given year. The sample period used is 7 years and therefore the research could be
considered medium term-oriented. I also test for the possibility of realizing short term excess returns, using the event study method where the list publication dates count as the event dates. Significant excess returns for the event study would indicate that buying stocks of firms who made it to a Best Workplace list in a given year and selling the stocks a few days afterwards would be a viable investment strategy.
The outline of the thesis is as follows. Chapter two summarizes the findings of similar
research and gives an overview of existing literature related to human capital and employee satisfaction. Chapter three describes the methodology of the research and chapter four describes the data. Chapter five shows the results and includes and elaborates on the interpretation. Lastly, in chapter six I conclude my research, discuss its implications and remark on what future research could add to the topic.
6
2. Theoretical background
2.1 Previous research
Edmans (2011) found monthly excess returns of 0.29% over the long term period of 1984 to 2009 for a value-weighted portfolio consisting of companies classified as ‘Fortune® Top 100 Best Companies to Work For in America’. He argues that stock markets do not fully
incorporate the value of intangibles, such as R&D expenditures, advertising, software development costs and patent citations, in equity prices. The true value of such intangibles are hard to measure and therefore such intangibles are said to only affect the stock price when it subsequently manifests in tangible outcomes that are valued by the market, with earnings being a primary example of such a tangible outcome. Employee satisfaction, the explanatory variable of interest in this thesis, is another intangible asset of which the true effect on firm performance is unknown. This is because there holds no consensus regarding whether employee satisfaction is actually beneficial to shareholders, or if it is only desirable for stakeholders. If this intangible does indeed have a positive relationship with firm value (through worker performance), then it is very likely to be undervalued on the market, granting the investor the opportunity to realize excess returns. Edmans performed the research using the Carhart four-factor model where he compared the return of companies that were ranked an annual Fortune top 100 list with the returns of the market, reasoning that employees of firms who were placed in such a list experienced high satisfaction at the company they worked for. He concludes that job satisfaction is beneficial for firm value. I use the same method as Edmans to test for excess returns in this thesis. It should be noted that Edmans published a new paper one year later, in which he covered the period of 1984 to 2011 instead and found the excess returns to be 0.25% per month, 4 basis point lower than his 2011 study. This could mean that excess returns that can be realized with the buy-and-hold portfolio strategy have decreased over time.
A recent study conducted by Guthrie et al (2015) using Glassdoor.com data over the period of 2008 to 2012 concludes that publicly traded firms with actively involved founders have a company culture that is different from non-family firms and scion
7 company. The terms ‘company culture’ and ‘employee satisfaction’ are used
interchangeably in the paper as the authors define company culture as the average employee satisfaction level, as measured by a firm’s score on Glassdoor’s website. The research shows that family firms are related to higher company scores on Glassdoor’s website: the average family firm scores 0.24 points higher (ratings range from 1 to 5) on the website than a scion firm. This implies that family-firms have a company culture that leads to higher employee satisfaction, leading to better firm performance. They also find that a one-standard-deviation increase in company rating on Glassdoor’s website is related to an increase of Tobin’s q of 0.069. This increase in Tobin’s q is significant at the 1% level for all different methods used: OLS, dynamic OLS, 2SLS and dynamic GMM model. Their reasoning behind this finding is that family firms have a human-capital-enhancing culture and that it is this culture that has a positive effect on firm value through Tobin’s q, a finding that is “consistent with family firms’ long-term focus on human capital”.
Guiso et al (2015) also researched the effect of corporate culture on firm
performance. They find that the extent to which managers self-advertise the firm’s integrity-based core values has no effect on firm performance. However, when analyzing the
employee’s perspective, they find a significant relationship between the level of managerial integrity that they perceive and Tobin’s q: a one-standard-deviation increase in perceived managerial integrity is related to an increase of the firm’s Tobin’s q by 0.19 to 0.47 standard deviations. These findings contribute to the findings of Guthrie et al (2015) in concluding that individual employee satisfaction, aggregated as overall company culture, is positively related to firm performance as measured by Tobin’s q. However, they do not theorize a causal relationship between enhanced human capital (as a result of higher satisfaction) and firm performance in the way that the paper of Guthrie et al did. All in all, the literature seems to agree that the firm ratings reported at Fortune.com and Glassdoor.com are a solid proxy for employee satisfaction and that this holds a positive relationship with a firm’s performance.
8 2.2 A new era of human capital valuation
The clash of views regarding employee satisfaction’s value could be attributed to changes in both the firm and human capital over time. The twentieth-century firm relied much on its tangible, physical assets so that it could grow big and gain a competitive advantage over its competitors through economies of scale (Chandler, 1990). However, where in the 20th century most labor was unskilled and firms put large emphasis on cost efficiency, the millennial generation of workers will be judged upon quality and innovation. As Zingales (2000) describes it: “Employees are not merely automata in charge of operating valuable assets but are valuable assets themselves, operating with commodity-like physical assets.” In his paper, he makes a distinction between explicit and implicit contracts and states that this distinction has a massive influence on how a firm is defined. An explicit contract states the terms and requirements that a firm and an employee agreed upon establishing the partner relationship. It functions as a form of protection for both parties as their basic rights and responsibilities towards each other are defined. However, excellence does not thrive through a basic mindset which means that in order for both employee and firm to
outperform others, more is needed than what could be described in a formal contract. Zingales marks a firm’s reputation as the distinctive factor that motivates employees to do more than what they are required to do. A firm that has a reputation of treating their employees well and rewarding them more than what the firm is required to do, incentivizes the employee to do perform well and to do more than what is expected from him. This automatically leads to the formation of implicit contracts, where the reputation of the firm functions as an important asset in order to have this relationship established. On the other hand, if a firm does not reward employees’ unasked investments, the reputation of the firm might turn into a liability. Theoretically, this means that the firm has a value different from the sum of its individual parts, the difference to which Zingales refers using the term ‘organizational capital’.
Another theory that links this organizational capital to firm performance uses the employee’s potential future access to the firm’s resources as the most important incentive for an employee to make firm-specific investments (Rjana et Zingales, 1998). In this theory, the employer wants to acquire power over the highly valued human capital so that it can combine this with another critical resource of the firm, such as his own human capital. The
9 possible employee knows that the employer will have power over him and that he has partial ownership of the firm as well. I order for the employee to commit to this seemingly unattractive labor agreement, he has to know that making firm-specific investments will reward him enough to justify the agreement. The firm does this by rewarding important employees with partial ownership of the firm (stock-based compensation). This leads to a whole network of firm-specific investments which competing firms are unable to imitate as it takes time to build the reputation for a firm to do this. As a result, the organizational capital, based on the reputation of the firm, has allowed the firm in question to develop a competitive advantage.
World trade and the internet have had an enormous impact on labor mobility and it only makes sense that in a time where human capital is valued highly, the rewards have to be appropriate in order to retain the highest level of talent. Ramasamy and Yeung (2007, p. 323) quote a Fortune.com survey of the 3,500 largest companies in the US which found that a company’s market value is, on average, for 72% determined by its intangible assets, of which between 40 to 75 percent is brand. They mention that although brand value is mostly consumer-oriented, the reputation of a firm towards employees plays a part in brand value determination as well. Firms with a valuable brand should be able to attract more talented employees to increase the value of both the firm’s intangible and physical assets as a result of higher profits. However, it takes time to build a strong brand, both towards employees and consumers, and the results might only become apparent in the long term. The increased appreciation for employees is in line with the perspective of the Resource-Based-View (RBV) of the firm, where firm value is determined by its resources rather than its products
(Wernerfelt, 1984). Human capital is considered a resource and it would be extremely valuable for research if its true value could be determined. However, the high amount of individual variety of employees and firms makes it nigh impossible to make a conclusive valuation (Weisbrod, 1959). As each firm is different, human capital might be more valuable for certain firms than it is for others, but there appears to be increasing consensus that human capital is indeed an important resource that makes up a significant part of a firm’s value.
10 2.3 Employee satisfaction
The relationship between employee performance and employee satisfaction has been analyzed extensively in the psychological field. In their paper, Iaffaldamo and Muchkinsy (1985) report a correlation between job satisfaction and job performance ranging from 6 to 29 percent. They refer to the mean value of 17% correlation and conclude that job
satisfaction and job performance have little relationship with each other, a conclusion in line with the research findings up until that date. Chapman and Chapman (1969) even went as far as saying that the relationship between job satisfaction and job performance is illusory, although they agree that intuitively they should be related with each other. However, it should be noted that this literature dates from the past century. As mentioned in section 2.2, this was an era where the focus was on physical assets and where human capital was not valued the way it appears to be valued to date.
However, recent literature appears to have reached consensus that there is a significant causal relationship between employee satisfaction and job performance. Hayes et al (2002) found that there exists a causally positive relationship between employee satisfaction or employee engagement at the business-unit level (per department) and a firm’s financial performance. Baysinger et al (2012) confirmed that employee engagement and job satisfaction were the best indicators of overall employee performance.
Meta-analysis performed by Bono et al (2011) found a correlation between overall job satisfaction and job performance of 30%. Organ (1988) argued that the reason for the low correlation between employee satisfaction and job performance found in last century’s literature is because job performance is narrowly measured by only looking at specific job tasks. He reasons that if the definition of job performance would include citizenship behaviors, the relationship would increase and give a better idea of the effect of employee satisfaction on job performance. Citizenship behavior is an employee’s commitment to tasks that he is not obliged to do: a volunteer’s mindset that arises as a result of the employee identifying with the entity it is part of. Seven years later, Organ and Ryan (1995) find evidence to confirm the theory with a mean correlation of 28% between job satisfaction and citizenship behavior. The large difference in results between large century’s literature and recent literature could be attributed to the shift from blue to white collar workers: from physical to intellectual. White collar jobs are focused more on intellectual capability than blue collar jobs do and job
11 satisfaction might be more required for a white collar employee in order for him to perform at his maximum level.
Hausknecht and Trevor (2011) find that higher employee satisfaction is related to a lower turnover rate. They argue that collective turnover hurts the firm’s financial
performance through loss of firm-specific human capital, disrupted operations and increased costs of recruitment and training. In addition, it saddles remaining employees with “newcomer socialization” which lowers productivity and might affect morale, lowering the level of collective functioning. Erez et al (2001) found that turnover could be predicted by an employee’s level of “job embeddedness”, which is defined as how well they see themselves fit for the job, their link with colleagues and what they would have to sacrifice in case they left their job. They find that including job embeddedness in standard employee satisfaction models increases their accuracy in predicting both voluntary turnover and intention to leave, suggesting that employees who like their workplace in addition to their job have a lower chance of leaving the company they work for. These findings are all in line with citizenship behavior theory, which suggests that employees are unlikely to leave a company if they identify with its culture.
There appear to be large differences between the cultures of companies, cultures that are both hard to describe and hard to change but which are, according to recent literature, certain to affect a firm’s financial performance as a result of the varying degrees of employee satisfaction. The recent literature suggests that this relationship is positive, because of both higher employee performance and lower turnover-related costs for the firm. In this thesis, I test whether this relationship is indeed significantly positive by using the annual Best Workplace rankings at Glassdoor.com and Fortune.com as a proxy for employee satisfaction, similar to the recent studies of Edmans (2011, Guthrie et al (2015) and Guiso et al (2015):
H1: High employee satisfaction is related to significant excess returns
12
3. Methodology
The objective of this thesis is to test whether there exists a relationship between firm reviews submitted by (ex) employees, the proxy for employee satisfaction, and the performance of the stock of the firms in question. For the research, I use the annual firm ranking lists (together referred to as Best Workplace lists) on the websites Fortune.com and Glassdoor.com. The websites use different methods of determining which companies are selected for their Best Workplace lists and therefore a large amount of firms included in Fortune’s lists might not be part of Glassdoor’s lists and vice versa. In order to candidate for a placement in its annual Fortune 100 Best Companies to Work© list, firms have to apply first. Fortune charges its customers a fee of at least €995 per year in order to be eligible for placement on its Best Companies to Work list.1 This could lead to selection bias where only those firms who value their reputation highly, apply. Usually the deadline to apply is half a year before the planned publication date of the list. The firm’s definitive score will then be based on two different surveys which will be sent to a random sample of employees of each firm and for which they have an allowed response time of one month. Two-thirds of a firm’s final score is based on the results of a 57-question Likert scale survey that involves questions regarding the employees’ attitudes about their workplace experience. The other third is based on a second type of survey that asks open-ended questions about methods of internal communication, recognition programs, diversity efforts, etc. Unlike the Likert scale survey, Fortune rates the answers to the second type of survey themselves which means that
external subjectivity is involved. Fortune then combines the ratings of both surveys to assign a final score to each firm. The 100 firms with the highest score make it to the Best
Companies to Work list which is published somewhere in the first three months of the following year, roughly half a year after the deadline of answering the surveys (e.g. the Fortune 100 Best Companies to Work list 2017 will be published on March 2nd 2017). The scores are not reported on the ranking lists.
Unlike Fortune’s list, the annual Glassdoor 50 Best Places to Work© list is said to be “solely determined by employee feedback” (Glassdoor does not function as an intermediary to rate the submissions) and the feedback is conducted over an entire-year period. In
13 addition, a firm does not have to apply first in order to be eligible for a possible ranking on the list. Every individual is able to make an account at Glassdoor.com and review a company for which they work or for which they have worked previously. Individuals can rate (Likert scale grading with a range of 1 until 5) the company on the following characteristics: Culture and Values, Work/Life Balance, Senior Management, Compensation and Benefits and Career Opportunities. In addition, they are also asked whether they would recommend working there to a friend, whether they approve of the CEO and whether they hold a positive business outlook regarding the firm’s future, using the same grading system. The current ratings for each firm are always visible on Glassdoor.com, which means that an investor could theoretically determine a given year’s Best Places to Work list days before it is actually published. A graph also shows the historical rating values of firms but it only goes back to January 2014, which leaves the data unusable for this research. Guthrie et al (2015) and Guiso et al (2015) were allowed access to the historical database for their research but sadly my request for access was denied by a Glassdoor representative. In November of each year, Glassdoor applies a proprietary algorithm to determine each of its award categories. As part of the algorithm, “Glassdoor considers the quality, quantity and consistency of reviews”, although the company does not go into more detail. In December, they publish the Best Places to Work list for the year that follows (e.g. the Best Place to Work list 2017 will be published in December 2016) For this research, I make use of the U.S. large companies version of the Best Places to Work list, for which a firm has to have a minimum of 1,000 employees in order to be eligible for placement. The small and medium Best Places to Work lists consist of too little publicly traded companies to be considered usable for research.
3.1 Portfolios
Monthly stock prices of public companies over the period of December 2008 until December 2015 is used for the research, which is retrieved from the CRSP database. I create different portfolios using the annual Best Workplace lists from Glassdoor and Fortune and compare the performance of these portfolios with the performance of the S&P500 index.
In an attempt to make sure that outperformance of the portfolios is not attributable to other factors, I make use of the Carhart four-factor model:
14
Rp – Rf = α + βrm (RM-Rf) + βhmlHML + βsmbSMB + βumdUMD + ε
(1)
In this model, a significant alpha α would imply significant excess returns of the portfolio(s) in question and is therefore the variable of interest in this equation. Rp represents the return of the Best Workplace portfolio in question, Rf represents the risk free rate and RM- Rf, HML, SMB and MOM are the four control factors. RM- Rf represents the return of the market minus the risk free rate, SMB represents the performance of small stocks compared to big stocks, HML the performance of highly-valued stocks compared to growth stocks and UMD the return of high-prior-return (12 months period) firms relative to low-prior-return firms. The monthly values of these four factors are retrieved from Ken French’s website. French calculates the monthly values of the SMB, HML and UMD factors as follows. For the SMB factor, he creates portfolios of the smallest firms on the market of which he subtracts the returns of the biggest firms (based on market capitalization) on the market, controlling for the book-to-market values of these stocks:
SMB = 1/3 (Small Value + Small Neutral + Small Growth) – 1/3 (Big Value + Big Neutral + Big Growth) (2)
For the HML factor he repeats the process but this time he computes the returns of the portfolios with large market-to-book ratio stocks of which he subtracts the returns of the portfolios with the small market-to-book ratio stocks, controlling for the market
capitalization of the stocks:
HML = ½ (Small Value + Big Value) – ½ (Small Growth + Big Growth)
(3) Lastly, for the UMD factor he computes the return of two high prior return (12 month
15 UMD = ½ (Small High + Big High) – ½ (Small Low – Big Low)
(4)
To construct the portfolios, French and his team look at all the stocks that trade on the NYSE, AMEX and NASDAQ and create the portfolios accordingly.
The research question and main hypothesis of this thesis is:
H1: High employee satisfaction is related to significant excess returns
I compose five different portfolios to test for significant excess returns: α. The first portfolio is the Best Workplace Standard Rebalancing portfolio (SRp). It invests an equal amount in stocks of the firms that appear on a Best Workplace list and holds these stocks until at least the next annual list is published. If the firm is republished on the next annual list, its stock will remain in the portfolio, whereas if it loses its placement on the list, the stock is sold and hence dropped from the portfolio until and if it re-enters at a later list publication. The rebalancing takes place on every last trading day of December for the Glassdoor portfolios and every last trading day of March for the Fortune portfolios. As I use firm publications on these lists as a proxy for employee satisfaction of the firms in question, I expect this
portfolio to realize significant excess returns:
H1: A Best Workplace Standard Rebalancing portfolio (SRp) realizes significant excess returns
A different form of a Rebalancing Portfolio invests an amount in a firm’s stock that is based on the market capitalization of the firm in question. Instead of investing an equal amount in the stock of each firm on the list, the portfolio invests more in firms with higher market capitalization and less in firms with lower market capitalization, so that the relative amount invested in each stock is equal to the relative market capitalization of that firm compared to the total market capitalization of all Best Workplace firms in a given year. The average
16 monthly value of the SMB factor in my sample period is 0.14% with a median value of
0.24%. This implies that, over my sample period, small market capitalization firms realized higher returns on their stock than high market capitalization firms on overall the market, in line with French’s theory. However, it is possible that investing more in high market
capitalization firms proves to be a better strategy in these portfolios for reasons not
explained by the four-factor model. The large firms in the Best Workplace lists are firms that can be classified as enormous instead: firms with a market capitalization at the 90th
percentile. I theorize that firms with an enormous market capitalization realize different returns than firms with a high market capitalization. For this reason, I also create a Best Workplace Market Capitalization-Weighted Rebalancing Portfolio (MCWRp) to test whether such portfolio realizes significant excess returns:
H2: A Best Workplace Market Capitalization-Weighted Rebalancing portfolio (MCWRp) realizes
significant excess returns
It seems plausible that employees of companies that are ranked higher in the Best
Workplace lists are more satisfied with their company than employees of companies that are ranked lower in the Best Workplace lists. For example, the average grade of the number one company in Glassdoor’s Best Place to Work list is 4.7 whereas the average grade of the number 50 firm is ‘only’ 3.5. It shows that, despite the fact that both companies have satisfied employees, the employees of the company ranked first can be considered much more satisfied than the employees of the company ranked fiftieth. For this reason, I create portfolios that invest an amount in the stocks of each Best Workplace company depending on the ranking of the firms in question. This Best Workplace Rank-Weighted Rebalancing portfolio (RWRp) is then further divided into two different portfolios, to make sure the results are robust. The first (median) portfolio divides the Best Workplace list in two halves. The portfolio invests double the amount in the stock of firms that appear in the upper half (top 25 for Glassdoor, top 50 for Fortune) compared to firms that appear in the bottom half of the list: the investment weights are 2 and 1. The other portfolio type (quintile) sorts the firm rankings in quintiles instead of in two halves. The investment weights for this portfolio are as follows: 2; 1,75; 1.5; 1.25 and 1. This means that twice as much is invested in firms
17 that are ranked in the top 20% of the Best Workplace lists compared to the firms that are ranked in the bottom 20% of these lists and that 75% more (1.75) is invested in firms that are ranked in the second quintile (top 20%-top 40%) compared to firms that are ranked in the bottom 20% of the lists, etcetera.
H3: A Best Workplace Rank-Weighted Rebalancing portfolio (RWRp) realizes significant excess
returns
The fourth unique portfolio type only invests in stocks of firms that appear in both Glassdoor’s and Fortune’s Best Workplace list in a given year. This means that only one regression will be run for this portfolio, compared to the previous three portfolio types where Glassdoor and Fortune’s data each allowed for a unique regression. However, like the other portfolios, this portfolio will still rebalance at the end of each year by selling the stocks of firms who do not make it to both next annual Best Workplace lists, and buying the stocks of firms who have (re-)entered both new Best Workplace lists. As a firm has to enter both websites’ Best Workplace lists in a given year to have its stock added to this portfolio, its employees are more likely to indeed be satisfied with their employer compared to the other portfolios. The portfolio therefore functions as a solid robustness test for employee
satisfaction:
H4: A Best Workplace Double Placement Rebalancing portfolio (DPRp) realizes significant excess
returns
As mentioned, all of the abovementioned portfolios rebalance after every new list publication. The rebalancing takes place on every last trading day of December for the Glassdoor portfolios and every last trading day of March for the Fortune portfolios. The Double Placement Rebalancing portfolio rebalances at every last trading day of March as well.
Lastly, a Hold-Until-End portfolio (HUEp) is formed for both Glassdoor and Fortune which, unlike the previously mentioned portfolios, does not rebalance at the end of every year.
18 Instead, a firm that enters a Best Workplace list has its stock added to the portfolio in that specific year and it will be held until the final data point at December 31th, 2015. This means that, unlike the other portfolios, the amount of firms that are included in these portfolios continues to increase with every new annual Best Workplace list publication:
H5: A Best Workplace Hold-Until-End portfolio (HUEp) realizes significant excess returns
All of the portfolios are expected to realize significant, positive excess returns. For this reason, a one sided t-test will be used to test the significance of the results. For the research I assume that no transaction costs are involved upon buying or selling stocks. However, in reality the individual investor is very likely to experience these costs and hence his realizable alpha would suffer.
3.2 Event study
The annual Best Companies lists provide the investor with information that is not observed by many other investors. In the previous section, portfolios were formed with the intention to realize excess stock returns by creating an investment portfolio with a medium-length time horizon, where the excess returns could be attributed to possible undervaluation that corrects itself on the market with time. I formulate a method to realize possible short term excess returns related to the Best Workplace list publications:
H6: The Best Workplace list publications allow the investor to realize significant short term excess returns
Glassdoor and Fortune are websites with millions of unique visitors each month and the publication of a Best Workplace list might incentivize readers to buy stocks of the firms that are listed on those lists. I test whether the list publications lead to a short term price jump of the firms in question by using a cross-sectional event study test:
19 CAARt CAARt tCAAR (5)
Here, tCAAR is the t-statistic of this Cumulative Average Abnormal Return test, CAARt represents the cumulative average abnormal returns at time t and σCAARt represents the standard deviation of cumulative abnormal returns across the sample at time t. Time t represents the end of the estimation window. This is done for all firms that appear in a Best Workplace list in a given year. To calculate a firm’s abnormal return, I first regress the return of the company’s stock on the return of the market for the 90 trading days prior to the Best Workplace list publication, using a standard CAPM regression:
(RM Rf) R (6)
Again using the CAPM with the α and β that follow from this regression, the expected returns for each company’s stock can be calculated for the five days following the event date of list publication and the event day itself. The expected return is then subtracted from the realized return for each of these days to calculate the abnormal returns. Summing up these abnormal returns over the specified time period yields the cumulative abnormal return for a specific firm’s stock. The average of the cumulative abnormal return of each firm is then used to calculate the t-statistic of the cumulative average abnormal return per year using equation (5).
To calculate the cumulative abnormal returns, two different time windows are tested. The first window does not assume so-called ‘leaking’ of information before the event and therefore starts calculating the abnormal returns from the event date until five days after this event: [t, 5]. The second window does control for information leaking before the event date and therefore serves as a robustness test by calculating the abnormal returns
20 five days before this event as well: an event study time window is [-5, 5]. A significant t-statistic would imply the ability to realize significant short term excess returns: an investor could, on average, realize above-expected returns by buying the stocks of firms that are listed on a given Best Workplace list and selling the stocks 5 days afterwards. The excess returns realizable with this strategy are expected to be higher than zero. For this reason, a one-sided t-test will be also be used for this event study to test the significance of the results, again using the assumption that no transaction costs are involved.
21
4. Data
The research uses historical stock price data from the CRSP database, retrieved from the Wharton Research Data Services website2. In addition to historical stock prices, I also retrieve yearly market capitalization and book equity data from the database. For the
portfolios regressions, monthly stock data is used over the period from December 2008 until December 2015. The historical values of the four factors RM-Rf, SMB, HML and UMD are retrieved from Kenneth French’s website.3 As explained in section 3.1, RM-R
f summarizes the return on the market minus the risk free interest rate, SMB the performance of small stocks relative to big stocks, HML the performance of highly-valued stocks compared to growth stocks and UMD the return of high-prior-return firms relative to low-prior-return firms.
Every annual Glassdoor Best Place to Work For list consists of 50 firms, whereas Fortune’s Best Companies to Work list consists of 100 firms. The lists from 2009 until 2015 are used for the research because 2009 is the first year for which Glassdoor published a Best Workplace award and having data for the 2 websites available to the robustness of my research. Previous studies used data of only one of the two websites’ Best Workplace lists. Both websites’ Best Workplace lists contain a substantial amount of private and non-profit organizations, which are omitted from the sample as they do not have stock price data available. In addition, subsidiaries are also omitted as it is impossible to determine their contribution to the stock price movement of the mother firm. An exception to this would be when a firm was taken over while it was part of a portfolio, but this did not occur in my sample. This brings the amount of unique firms in the sample to 95 for Glassdoor and 70 for Fortune, which is interesting as Fortune’s annual Best Workplace list contains 100
companies compared to Glassdoor’s 50. This can be largely explained by the fact that although Fortune’s Best Workplace list contains many different organizations each year, the majority of those aren’t usable for the research as they are either private or non-profit, whereas Glassdoor only lists public companies and their subsidiaries. In addition, of Glassdoor’s 95 firms, only 3 appeared in every consecutive list whereas for Fortune this number is much higher at 15. This could be partially explained by the selection bias that
2 https://wrds-web.wharton.upenn.edu/
22 occurs due to the fact that Fortune charges firm a fee in order to be eligible for its Best Workplace list, whereas Glassdoor does not. Google, Apple and Chevron are the companies that made it to every Glassdoor Best Workplace list but both Apple and Chevron were not listed even once at Fortune’s Best Workplace list. Google held the second ranking on Fortune’s list in 2009 and has been the number one ranked company ever since the 2010 Best Workplace list. However, at Glassdoor, it still achieved a top 10 placement in 6 of the lists 7 but only in 2015 did it manage to be ranked first there. This further shows that the different selection methods of the websites lead to different Best Workplace list
compositions. The amount of firms used for the research per Best Workplace list range from 23 to 35 for Glassdoor, 30.57 firms on average. For Fortune, this ranges from 36 to 43, 39.86 firms on average.
23 Table 1 presents the summary statistics for the independent variables used in the portfolio regression model and show the values of the market- and book equity for the firms in both websites’ portfolios. Both the market and the book valuations of equity are much higher for Glassdoor’s portfolios than for Fortune’s portfolios. The average market capitalization of a public firm in Glassdoor’s portfolio is 57.9 billion compared to Fortune’s 30.8 billion. The median market capitalization value is 30.1 billion for Glassdoor compared to Fortune’s 13.3 billion. The two outlier companies in Glassdoor’s portfolio are Apple (500.1 billion market capitalization in 2014) and Google (340 billion market capitalization in 2015), for Fortune this is only Google as Apple is not included in any of its Best Workplace lists. However, they are not omitted from the sample as I have no reason to believe that the firms’ stock returns are unsuited for the research. The average and median book values of equity are also noticeably higher for Glassdoor than for Fortune. These findings all suggest that smaller companies have a higher chance of entering Fortune’s Best Workplace list than Glassdoor’s, possibly due to the fact that firms have to apply first if they want to be eligible for a
placement on Fortune’s Best Workplace list and small companies might value placement on the list more highly (for publicity reasons) than large companies do.
Of the four factors RM, SMB, HML and UMD, two show counterintuitive data. According to French’s theory, value stocks should, on average, outperform growth stocks and high prior-return stocks should outperform low-prior-return stocks. For the HML factor in my sample both the mean and median value are negative, which is not in line with French’s theory. The average UMD factor is negative, but the median value is positive and therefore this factor’s value might be in line with his theory. The fact that the mean value is negative has to do with the largely negative factor values in 2009 following the 2008 stock market crash. The minimum UMD factor of -34.58% fell in this period. As a result, I consider the median value of the UMD factor to be more representative of the UMD factor over the course of the sample period.
24 Table 2: Portfolio Returns
Table 2 reports the returns of the different portfolios over the sample period. Every
portfolio realized a higher average return than the market’s 1.26% average monthly return, except for both Glassdoor’s and Fortune’s Market Capitalization-Weighted portfolios. These portfolios put more investment weight into stocks of high-market-capitalization stocks. Their lower returns could be partially explained by French’s theory that the stocks of big firms realize lower returns than the stocks of small firms, controlling for their market-to-book ratio. This is explained further in the results section. The standard deviation of all portfolios’ returns are higher than the standard deviation of the returns of the market. This could be attributed to the higher amount of stocks that are part of a market portfolio, reducing the impact of a few volatile stocks and therefore reducing overall volatility. The few volatile stocks in my portfolios have much greater impact on the portfolios’ overall
25 levels of volatility. This can also be seen from the fact that the maximum monthly return was 11.35% for the market, much lower than the maximum return of any of the portfolios. Lastly, it should be mentioned that the lowest return of -29.73% for the Double Placement Rebalancing portfolio is much more negative than what is the case for the other portfolios. Again, this could be attributed to the fact that this portfolio consist of much less stocks than any of the other portfolios.
For the event study with list publication as the event, daily stock price data is retrieved from the CRSP database over the period from August 2008 until March 2015. The stock price data starting August 2008 is used to cover the ninety days period before the first Best Workplace list publication and the last list publication is in March 2015. I test for the possibility of realizing excess returns for the period of list publication until five days after, [t, +5], and also test a window of five days before the publication until five days after as a test for
robustness: [-5, +5]. The data for this study is provided in the regression output in section five: Results.
26
5. Results
Each portfolio realizes different returns, although the common finding is that alpha α is not significant in any of the regressions and that the returns of the portfolios are highly
explained by the returns on the market: the βrm coefficient is close to 1 and therefore the portfolios contain clear systematic risk. This is due to the fact that all companies in the portfolio samples are part of the S&P500. The Simple Rebalancing portfolio indicates an alpha of 0.9% per year although this value is far from significant for both websites’ regressions.
Table 3: Simple Rebalancing portfolio results
For these Simple Rebalancing portfolios, the coefficient on SMB is significantly positive for both Glassdoor and Fortune, although substantially more so for Fortune’s portfolio. This could be partially explained by the fact that the average market capitalization of the companies in Fortune’s portfolios is roughly 60% that of Glassdoor’s portfolio: Fortune’s
27 portfolios contain, on average, smaller companies. As a result, the return of the portfolio can be significantly explained by the SMB factor which controls for the fact that, on average, smaller firms realize higher returns than bigger firms. The values of 0.202 and 0.289 indicate that there holds a relationship between the overall returns of small firms compared to large firms on the market, with the returns of the Simple Rebalancing portfolios. A 100 basis points return of small over large firms (SMB factor of 0.1 in a given month) is related to a 20 and 29 basis points higher return for the portfolios of Glassdoor and Fortune respectively.
The HML coefficient of -0.212 is significant at the 1% level only for Glassdoor, which implies a negative relationship between the returns of the portfolio on one side, and the returns of high-value firms over growth firms on the market on the other side. French found that stocks with a high market-to-book ratio realized higher returns than firms with a low market-to-book ratio and uses the HML factor to control for this finding. However, over the sample period used in this research, both the median and mean HML values are -0.2% which does not correspond to the positive value mentioned by French. The negative relationship indicates that for every 100 basis points increase of returns of high-value firms over low-value firms in a given month (HML factor of 0.1), the returns of Glassdoor’s Simple
Rebalancing portfolio are expected to decrease by 21 basis points. Looking at Table 1, the market-to-book ratio for Glassdoor’s portfolio is higher than Fortune’s portfolio. Given the negative mean HML factor value of -0.002 as opposed to an expected positive value, I argue that that the Glassdoor portfolios do better partially because they contain relatively high market-to-book ratio stocks whose variation in returns can be significantly explained by the HML factor, but the unintuitive negative mean HML factor value over the sample period has resulted in a negative relationship instead. In other words: despite an overall negative return in the sample period for high-value compared to growth stocks on the market, a significant part of Glassdoor’s portfolio can be explained by the high market-to-book ratio composition of the portfolio, which, according to French’s theory, realize higher returns. The average and median market-to-book value for Glassdoor’s portfolios are 4.35 and 3.88 respectively, which is higher than the values of 4.05 and 2.99 for Fortune. This implies that the returns of the Glassdoor portfolio can be at least partially explained by the market-to-book value of its stocks. The market-to-market-to-book ratio of the Fortune portfolio is between 10% to 30% lower each year and this, assuming French’s theory is valid, is possibly why the coefficient for the HML coefficient is not significant for the Fortune regression. Over a
28 longer period of research, French’s HML factor is expected to be positive and I would expect the coefficient of HML to be positive (rather than negative) as well for Glassdoor’s relatively high market-to-book value portfolio.
The last coefficient is that of the momentum factor UMD. The coefficient is negative at the 1% level for both Glassdoor and Fortune’s portfolio, which indicates a negative relationship between the previous returns gained by positive momentum stocks over negative momentum stocks, and the returns gained on these websites’ portfolios. The values of -0.167 and -0.172 for respectively Glassdoor’s and Fortune’s portfolios indicate that for every 100 basis points increase of return for positive momentum stocks over negative momentum stocks in a given month (UMD factor of 0.1), is related to a 17 basis points lower return on the websites’ portfolios. The UMD factor has a monthly mean value of -0.3% whereas its median value is a positive 0.4%. This can be explained by the aftermath of the 2008 financial crisis. Over the period January 2009 until December 2009 the average UMD factor value was -5.4% with a median value of -3.5%. The interpretation of this finding is that in 2009, firms whose stock price had dropped in the past 12 months (start of the crisis), on average experienced higher returns than firms whose stock price had risen in the past year. The economic crisis of 2008 became a widespread phenomenon which resulted in many firms seeing the price of their stock plummet. Emotions have likely played a larger-than-usual role on the stock market during the crisis which results in stock prices dropping beyond their reasonable fundamental value. The market at least partially corrects this in 2009, hence the unintuitive negative (following French’s reasoning) UMD factor. The standard deviation of the UMD factor was 11% in in 2009 as compared to only 3% in the following years, which further emphasizes the chaos that reigned over the stock market during and after the crisis period.
As the coefficients and t-values of the alpha and UMD variables are similar for both websites’ portfolio regressions, it can be concluded that for a Simple Rebalancing portfolio no significant alpha is to be realized when controlling for the returns on the market, the returns on small over large market capitalization stocks, the returns of high-value over growth stocks and the returns of positive momentum over negative momentum stocks. The only variable whose contribution value is uncertain is that of HML in the Fortune regression and this can at least partially be explained by the relatively low market-to-book ratio of
29 Fortune’s portfolio as compared to Glassdoor’s portfolio and a market portfolio.
For the MCW portfolio, again no significant alpha is to be realized although the alpha coefficient on Fortune’s portfolio becomes much closer to significant at the 10% level. However, the coefficient indicates a negative alpha of 3.9% per year, which is much lower and the lowest of all regressions.
30 In addition, the Market Capitalization-Weighted portfolios are the only portfolios to realize lower average returns than the market, as shown in Table 2. This hints at the portfolio being an unviable investment strategy, although the insignificance of the coefficient prevents me from drawing a conclusion. The SMB coefficient is much smaller for Glassdoor and now far from significant for both portfolios. This is in line with expectations: the returns of a portfolio that invests more in stocks of firms with high market capitalization should show little relationship with the returns of small stocks on the market. The value of the HML coefficient is no longer significant either, which could be partially explained by the low market-to-book value/”growth-status” of the high market capitalization stocks in the
portfolio. The returns of these portfolios still hold a negative relationship with the returns of momentum stocks on the market. The Four-Factor model is still a good fit, although it explains less of the variations of the portfolio’s returns than the Simple Rebalancing portfolio.
The results of the RWRp are similar to the results of the Rebalancing portfolio. No significant alpha is to be realized and all other factors in the model can significantly explain the
variation in returns of the portfolios, apart from the HML factor for Fortune (like was the case for the SRp). In addition, the difference between the median portfolios and the quintile portfolios is small. I conclude that the Simple Rebalancing portfolio and the Rank-Weighted portfolio have similar characteristics and that neither of them is able to realize excess returns. This suggests that the rankings on the lists do not matter for composing an
investment strategy: the well-performing firms’ stocks are somewhat randomly distributed among the Best Workplace lists and not clustered at the higher rankings despite the large differences in grades between the top and bottom ranked firms. This further suggests that employee satisfaction, as proxied for using the Best Workplace lists, is not related to significant excess returns for the stocks of the firms in question.
31 Table 5: Rank-Weighted Rebalancing portfolio results
32 The results of the Double Placement Rebalancing portfolio differ from the previous
portfolios in the fact that the HML and UMD coefficients are now not significant for both websites and that the explanatory power of the model has decreased substantially: the R² is only 42.1%.
33 The factors do a much worse job at explaining the variation in returns of the portfolio, which could be attributed to the small amount of firms of which each year’s DPRp is composed (as the firms have to appear in both Glassdoor’s and Fortune’s Best Workplace list), leading to a high standard deviation of returns. The standard deviation of returns for this portfolio is the highest of all portfolios with a value of 6.45% and this is likely why the regression shows less confidence in the effect of the HML and UMD factors being non-zero: the relationship between the HML and UMD factors and the portfolio’s variation of returns is not apparent enough to draw a valid conclusion at even a 10% significance level. Alpha is positive but still far from significant. The relationship of the portfolio’s returns with the returns of the
market, the RM-Rf factor, is still significant and close to 1, although the beta’s coefficient has decreased to 0.857. The positive SMB coefficient suggests that the portfolio is composed of a substantial amount of small companies as to explain the relationship with the return of small companies on the market, suggesting that consensus regarding which firms are the best to work for is highest among small companies (most firms that are part of this portfolio have a market capitalization that French would identify as small: below the 30th percentile).
Lastly, there appear to be no significant differences between the HUEp and the rebalancing portfolio. The regression coefficients have values and significance levels similar to those from the Simple Rebalancing portfolio regression: the relationship with small stocks’ returns is more positive whereas the relationship with high-value stocks has become less negative. However, there is still no significant alpha to be realized as the other factors explain most of the variation in returns. This suggests that for an investor there is little difference between holding a stock for an indefinite time and filtering out the stocks that are dropped from a new year’s Best Workplace list. A company is expected to have less satisfied employees and perform worse when it is dropped from a Best Workplace list series, and therefore a HUEp is expected to do worse than a Rebalancing portfolio. The findings show that this is not the case which further suggests that the effect of employee satisfaction on a firm’s financial performance is insignificant.
34 Table 7: Hold-Until-End portfolio results
35 Event Study
Table 8 shows the results of the event study. For each year’s Best Workplace list publication I test whether the publication of the list leads to the possibility of realizing excess returns on the stocks that are listed on that year’s list, over a five day period.
36 The list publication does not lead to the possibility to realize excess returns on a portfolio of stocks that are bought on the day of list publication. The test results are far from
insignificant for both websites’ portfolios in each of the seven years. The similar results for Glassdoor and Fortune are surprising. Glassdoor has the grades of each company viewable on its website over the entire year whereas Fortune does not. Each year’s Glassdoor Best Place to Work For list is simply a summary of the highest ranked companies at that time. For this reason, Glassdoor’s Best Workplace publishing do not provide the investor with new information. Individuals could look up the ratings of all public companies and they would have the same information as what is listed on the Best Workplace list. However, this process would be time intensive without any sort of algorithm and it is likely that the Best Workplace lists are something that only a small group of investors take time to look at when making their investment decisions, as in practice markets are not fully efficient. The effect of a Best Workplace list publishing on the stock prices of the firms published in these lists appears to be insignificant: no excess returns can be realized by investing in Best Workplace firms the day of a Best Workplace list publication and selling the stocks a few days after.
37
6. Conclusion
In this thesis, I analyzed the relationship between employee satisfaction and firm
performance. I constructed five different types of portfolios in which I included only stocks of firms who were placed in an annual Best Workplace list of either Glassdoor.com or
Fortune.com and tested for excess returns using Carhart’s four-factor model over the period 2009-2015. As the Best Workplace lists are determined by employee-submitted review ratings, I consider a firm’s placement in such a list as a solid proxy for high employee
satisfaction, in line with the research methods of recent literature. I found no evidence that there exists a positive relationship between employee satisfaction and firm performance for any of the portfolios. The large differences in portfolio compositions adds to the robustness of my research as most of the variation in returns can be explained by the four control factors in the model.
This conclusion contradicts the recent findings of Edmans (2011), Guiso et al (2015) and Guthrie et al (2015) who all found a positive relationship between employee
satisfaction and firm performance. I argue that the difference lies in the fact that two of these three papers used Tobin’s q as a measure of firm performance instead. While I consider this measure appropriate, I consider the four-factor model to be more suited to answer the research question from an investor’s point of view, which was the goal of this thesis. The four-factor model compares the return of the portfolio relative to the market which is a common performance benchmark for investors. Tobin’s q, however, does not measure the firm’s increased performance relative to the performance of other firms. My findings also contradict the findings of Edmans, who used the four-factor model as well. I reason that this is attributable to the fact that my research is medium term-oriented with a recent sample period, whereas his research was long term-oriented with an older sample period (1984-2009), which means the different results might stem from differences in firm characteristics over time. In addition, I found that the Best Workplace list publications themselves do not offer to the opportunity to realize short term excess returns either. I tested this using the cross-sectional event study method and found no excess returns in any of the seven years.
My findings contribute to the ongoing debate regarding the true value of employee satisfaction for shareholders. There is consensus that employee satisfaction is beneficial for
38 employee performance, although to this day it is uncertain whether the benefits of
increasing employee satisfaction weigh up to the costs a firm would have to pay to realize such goal. I conclude that this is not the case, although admittedly my research is lacking at certain important criteria. First and foremost, my sample size is low for a buy-and-hold strategy to be effective and it is likely that the insignificant alpha is partially related to this. However, as Glassdoor’s Best Workplace data only dates back to 2009, I was unable to extend the length of the sample period without hindering the robustness of my results. Secondly, it is unclear whether the annual Best Workplace lists are a valid proxy for employee satisfaction. Ideally, this would be tested by comparing turnover data for Best Workplace firms with turnover data for firms that are not published on such list. I found such data to be unavailable as most firms only report the number of employees in their income statements. Lastly, my research is lacking in the fact that I do not control for differences between Best Workplace firms. It is likely that human capital is more valuable for firms in certain industries than it is for others, in which case the insignificance of my results could be explained by my failure to control for important firm characteristics: omitted variable bias.
All in all, there is much research to be contributed to the question of whether employee satisfaction is related to firm performance. It might not be so much the
employee’s level of satisfaction that is related to a firm’s performance, but rather a firm’s overall culture. Previous studies used these terms interchangeably but I feel that a
company’s culture is much more likely to have a direct relationship with a firm’s
performance. For these reasons, I recommend future researchers to learn from where my research is lacking and to take a broader definition of employee satisfaction, measuring a firm’s culture instead. I am certain that future human capital-related research will be able to uncover its relationships with financial topics. Hopefully, this thesis contributes towards having that vision realized.
39
References
Baysinger, M., Brummel, B. J., Dalal, R. S., Lebreton, J. M., 2012. The Relative Importance of Employee Engagement, Other Job Attributes, and Trait Affect as Predictors of Job
Performance. Journal of Applied Social Psychology 42, p. 295-325
Bono, J. E., Judge, T. A., Patton, G. K., Thoresen, C. J., 2001. The job satisfaction-job
performance relationship: a qualitative and quantitative review. Psychological Bulletin 17, p.
376-407
Chandler, A., 1990. Scale and Scope. Belknap Press
Edmans, A., 2011. Does the stock market fully value intangibles? Employee satisfaction and equity prices. Journal of Financial Economics 101, p. 621-640
Edmans, A., 2012. The link between job satisfaction and firm value, with implications for corporate social responsibility. Academy of Management Perspectives 26, p. 1-19
Erez, M., Holtom, B. C., Lee, T. W., Mitchell, T. R., 2001. Why People Stay: Using Job
Embeddedness to Predict Voluntary Turnover. The Academy of Management Journal 44, p. 1102-1121
Guiso, L., Sapienza, P., Zingales, L., 2015. The value of corporate culture. Journal of Financial
Economics 117, p. 60-76
Guthrie, J. P., Huang, M., Li, P., Meschke, F., 2015. Family firms, employee satisfaction, and corporate performance. Journal of Corporate Finance 34, p. 108-127
Harter, J. K., Hayes, T. L., Schmidt, F. L., 2002. Business-Unit Level Relationship Between Employee Satisfaction, Employee Engagement, and Business Outcomes: A Meta-Analysis.
Journal of Applied Psychology 87, p. 268-279
Hausknecht, J. P., Trevor, C. O., 2011. Collective turnover at the group, unit and
organizational levels: Evidence, issues, and implications. Journal of Management 37, p. 352-388
Iaffaldano, Muchinsky, 1985. Job Satisfaction and Job Performance: A Meta-Analysis.
Psychological Bulletin 1985, p. 251-273
Organ, D. W., 1988. A restatement of the satisfaction-performance hypothesis. Journal of
40 Organ, D. W., Ryan, K., 1995. A meta-analytic review of attitudinal and dispositional
predictors of organizational citizenship behavior. Personnel Psychology 48, p. 775-803 Raghuram, R., Zingales, L., 1998. Power in a theory of the firm. Quarterly Journal of
Economics 108, p. 387-432
Ramasamy, B., Yeung, M., 2007. Brand value and firm performance nexus: Further empirical evidence. Journal of Brand Management 15, p. 322-335
Weisbrod, B., 1959. The Value of Human Capital. Econometrica 27, p. 425-436
Wernerfelt, B., 1984. A resource-based view of the firm. Strategic Management Journal 5, p. 171-180
