University of Amsterdam, Amsterdam Business School
MSc Business Economics, Finance track
Master Thesis
Does abusing employees pay off?
Adam Olejniczak
July 2016
Thesis supervisor: dr. R. Perez Ribas
Abstract
Some of the big corporates mistreat labor to reduce production costs. I investigate the influence of detected cases of sweatshops on companies’ financial performance. I answer the question if companies lose or save money on abusing employees. I use a large set of news items, related to Corporate Social Responsibility, and financial data about involved companies. I regress a number of items describing violated human rights on companies’ financial performance measures. I find that sweatshops allow saving very small amounts of money in the context of the entire corporate, while losses beard by owners when scandals come to light are very painful. I show that at average companies face a decrease of value close to 1% in few years after the scandal is disclosed, and even over 2% if the company is suited. I also show the mechanism that leads companies to violate employees’ rights and what happens after disclosures. This research adds a new evidence that companies are losing money if they abuse workforce.
Page 1 of 46
Statement of originality
This document is written by Student Adam Olejniczak 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.
Page 2 of 46
Table of Contents
Statement of originality ... 1 1. Introduction ... 3 2. Literature review ... 6 3. Data ... 10 4. Empirical Method ... 244.1 Testing the effect of disclosed scandals ... 24
4.2 Testing the effect before scandals are disclosed ... 26
5. Results ... 28 5.1. Main results ... 28 5.2. Additional results ... 32 6. Discussion... 36 7. Conclusion ... 39 8. References ... 40 List of Tables ... 45 Appendix ... 46
Page 3 of 46
1. Introduction
This thesis investigates the reasons why some companies abuse their employees, even if the whole world seems to be increasing the social awareness. Ethics in business and especially in finance is an important topic for increasing number of investors. Multiple studies, such as Becchetti and Ciciretti (2009), show that investments in the companies that operate with high ethical standards are less risky and provide higher returns. However, not many studies have been done on the other aspect of the companies’ ethical performance: human rights violations- and the influence on their financial dimension.
In this thesis, I attempt to show the incentives behind sweatshop practices and the effect of disclosing them. In theory, companies that violate the rights of their employees first benefit from reducing costs. However, once these incidents are discovered, they lose value. In the long run, the imbalance between costs and benefits makes ethical companies to perform better.
I test this hypothesis using a large sample of news articles related to human rights. In particular, I study the influence of several types of published stories which are directly related to the abuse of employees in developing countries. I estimate both effects, before and after the story was published. I show what happens when the sweatshops take place, so possible benefits for the companies. I also present markets’ reaction after they are disclosed.
Recent studies on ethics in finance involve the investment perspective of the Corporate Social Responsibility (CSR). Arx and Ziegler (2008) and Deng, Kang and Low (2013) show that companies with high CSR standards give investors higher returns. Empirical research by Kasimati (2007)
Page 4 of 46 support these findings. All these studies are in line with the Capital Assets Pricing Model (CAPM) described in Fama and French (2006) and the theory of better long-run performance of low-risk companies developed by Hong and Sraer (2012). Studies by Kempf and Osthoff (2007), Ammann, Oesh and Schmin (2011), and Roberts (2011) show that ethical companies benefit from cheaper capital and often beat the benchmarks. A significant loss noted by investors after the sweatshops are disclosed are shown in Rock (2001).
Previous research concludes that ethical companies perform better. Not many studies, however, investigate the effect of disclosed sweatshops. There is no explanation how companies’ financial performance is affected by these scandals. This research fills the gap. I use unique dataset, containing information about many various scandals from across the world. I show how published articles that describe sweatshops affect companies’ market value, sales, profitability and investments. I repeat the same study looking for the effect of lawsuits and actions companies take after the scandals. The main contribution of this study is showing that companies actually lose money on abusing employees. I explain what leads them to do so, and what is the mistake they make.
Findings reveal a negative impact of disclosed sweatshops, which is in line with all of the described studies. On the other hand, I do not find any benefits for the companies before the scandals come to light. I argue that the reason is that sweatshops happen on single factories’ level. At this level savings seems to be big. However, disclosed scandals do not affect only responsible factories, but also the rest of the corporate. At the end loses are much higher than savings.
Page 5 of 46 This paper illustrates the mechanism and aims to explain why companies decide on unethical practices. I argue that corporates make a mistake of overestimating small, current savings and underestimating future losses. My findings suggest that reduced costs are not sufficiently balanced by the benefits from sweatshops.
In the next sections, I investigate an impact of discovered cases of abusing employees, related to them lawsuits and actions took by companies to recompense workers on various aspects of their financial performance. First, I test the hypothesis that companies abusing their workforce first benefit from higher net incomes (thanks to reduced costs), but once a sweatshop gets discovered, they face the decrease of profitability. Second, I investigate the influence of disclosed scandals on the sales level. I check whether customers boycott products made by unethical companies. I also investigate the influence of discovered illegal practices on investments level. The hypothesis behind this is that companies that were proved to violate human rights are not welcomed in many countries with their new investments. That would mean that they are constrained against developing their enterprise, as potential partners do not trust them. Finally, I argue that these companies are paying more for the new investments. They have to introduce high control standards, provide better work conditions to new employees and accept worse investment conditions than other companies- due to ruined reputation.
I do not find any evidence for the decreased sales levels and investment costs after scandals are disclosed. However, I indicate increased sales levels before the lawsuits and companies’ responses, as well as increased costs afterwards. I reject all of the listed auxiliary hypotheses. However, these results show the process that happens inside the companies, after the sweatshops are discovered.
Page 6 of 46 All of the findings are in line with most of the recent studies. Shown in previous research better long-term performance of highly ethical companies strengthen my finding that companies abusing employees do not benefit from that. The fact, that scandals’ disclosures have a negative impact on the companies’ value is also harmonious with mentioned studies. My finding, that sales levels are not affected by the scandals can be explained with the paper by Prasad, et al. (2004). They show, that most of the consumers are always choosing cheaper goods, regardless the working conditions of labor producing them. They also point, that a group of the clients that are actually aware of the ethical performance of the companies is too small to impact general sales level.
2. Literature review
Corporate Social Responsibility (‘CSR’) means all of the actions taken by companies to increase their ethical standards (Williams, 2000). Over the last years, it became a very important issue for the governments, companies, and investors. A research has been done in each of the aspects of these policies. A social impact is important for each of the companies’ performance dimensions. However, despite the trend to keep on increasing social responsibility standards, there are still some companies that not only do not imply them but sometimes also violate employees’ rights and environmental norms. A lot of past research examines these two aspects of the companies’ performance to address the questions about the implications for all of the stakeholders’ groups. More and more investors turn into high ethical standards, limiting their capital engagement in companies with poor social performance. Brammer, Brooks and Pavelin (2006) basing on the
Page 7 of 46 sample of companies operating in the United Kingdom investigated the relationship between many CSR attributes and stocks performance. They find that companies with high ethical standards provide lower stock returns. This finding, if set together with the Capital Assets Pricing Model (Fama and French, 2006) shows that investments in the highly ethical companies are less risky. Becchetti and Ciciretti (2009) using a much bigger sample, find the same results. They indicate that socially responsible companies’ stocks are characterized by significantly lower returns. They also provide strong evidence that the variance of returns is much lower for these companies. Using the Corporate Environmental Performance measures and a sample of over 500 companies listed in the USA, de Haan, et al. (2012) show similar findings, pointing that ethical companies perform worse in the short-run.
Manescu (2011) examined the effect of many social, environmental and governance attributes on the stocks returns, finding that information about companies’ ethical performance is not incorporated in the stocks prices. The author argues, that all of the differences in stocks’ returns are due to mispricing and not due to the certain CSR attributes. This finding contrasts with previously mentioned studies as well as with the paper by Arx and Ziegler (2008). They show a significant link between the corporate social performance and stocks prices using a large sample of companies operating both in the USA and in Europe.
The existence of the CSR performance causality on stocks returns is also shown in the paper by Kasimati (2007), which examines this relationship basing on a sample of companies listed on the Athens Stock Exchange. This author shows that the long-run returns from investments in the companies applying high CSR standards are higher than average. Kempf and Osthoff (2007)
Page 8 of 46 support this finding. They show that the portfolios consisting of highly ethical companies perform better than benchmarks.
All of the above results are in line with the findings of Hong and Sraer (2012), showing the better long-run performance of low-risk companies. Described papers show that companies operating socially responsible are characterized with the lower risk and short-run returns but better performance in the long term. Thus, there are many reasons to believe that ethical companies give investors better investment returns.
Not only normal stock returns are important to understand the CSR impact on the companies’ performance. One of the most convenient opportunities for investors to quit, are mergers and acquisitions. Research on the CSR impact on the stock returns after M&As announcement, performed by Deng, Kang and Low (2013) shows a strong, positive effect. It means that investments in socially responsible companies guarantees above average returns in case of the companies’ acquisition. Thus, high ethical standards implemented by the companies, increase shareholders’ value.
There is also a strong link between the corporate social responsibility and the governance indicators. A research on these measures and firms’ value was also performed by Ammann, Oesh, and Schmin (2011). It shows a positive causality of high governance standards on companies’ valuations, showing that solid ethical standards have a positive impact on the financial performance. Similar findings are shown in Gompers, et. al. (2001), who point on higher returns on investment in companies with high corporate governance standards.
Page 9 of 46 All of the above studies are done from the investors perspective. To better understand the process it is also important to examine the impact of companies’ ethics on the other dimensions of the financial performance- ones directly impacting companies through managerial decisions. Ghoul, et al. (2011) test the influence of the CSR on the cost of equity. They find a significantly negative relationship, so companies that operate with high ethical standards benefit from cheaper equity. Similar research performed by Goss and Roberts (2011) shows that also the cost of debt decreases when companies improve their ethical standards. Both papers indicate a positive causality of CSR on the companies’ cost of capital. Kamatra and Kartikaningdyah (2015) use a sample of companies listed on the Indonesia Stock Exchange to examine the effect of CSR on the list of financial performance indicators. They show some significant correlations, proving the better financial performance of the socially responsible companies.
All of the described studies provide evidence that companies benefit from increasing the social responsibility standards. However, not many researches were done on the other aspect of ethics in finance- the impact of human rights violation cases on the financial performance. Rock (2001) performs an event study on a sample of companies that in the 1990s disclosed the sweatshop practices. He shows a significant decrease of the stock prices after the disclosures. He also demonstrates a positive effect of the actions company takes to prevent the next sweatshops on their stock prices. Prasad, et al. (2004) shows that most of the consumers prefer cheaper goods, regardless the working conditions in the producing company. However, they show a narrow niche in the market, consisting of the clients willing to pay more for the goods produced in fair working conditions. However, this niche is too small to impact entire company. Thus, conclusions from this study are that violating human rights do not impact the sales.
Page 10 of 46
3. Data
Part of the original data comes from the Business and Human Rights Resource Center. It covers articles and reports on 3,204 companies from April 1996 to October 2014. Most of the content is related to human rights violations and environmental issues. The dataset contains multiple types of items describing mentioned events. Most of them are press articles, but there is also some information about lawsuits or company responses to the related accusations. The total number of stories in the original data is 18,364. However, many of these have multiple threads, the total is 45,555. This dataset contains variables listed in Table 1.
Table 1: List of the variables in the original dataset
Variable Description
CompanyID A unique number that identifies each company within the dataset.
Company A name of the company.
ItemTitle A title of a corresponding story.
ItemAuthor A name of the author and the source of the story.
ItemType A type of an item.
ItemDate A date when the story was published.
Over 90 Item Categories Tags describing the content of the story.
The table shows a complete list of variables included in the original data set.
Each company is characterized by the unique identifying number (‘CompanyID’). Additionally, the full names of the companies are included in the data. The next four variables describe each of the items, giving its title, author, naming the type of an item and the first date of publishing, respectively. The set contains over 90 further variables, which are the tags that describe the content of an item.
Page 11 of 46 The financial data come from the CRSP/Compustat Merged database. The sample contains all of the companies that were listed on the stock exchanges across the USA between 1998 and 2014. The number of companies is 8,872. After downloading the entire database, I have deleted the observations for which data about a number of shares outstanding, closing price, sales, net income or total assets is missing.
To match these data with companies’ financials, a number of adjustments have to be made. The only information that can be used in this matching is the company name, which sometimes is not in the same format. However, it is possible to merge both datasets. To do so, I first decapitalize all names and delete all special signs. Afterward, I am performing the merger on the 95% probability level. It means that the observations are being matched, even though some small differences in the companies’ names appear (i.e. ‘CompanyX’ and ‘CompanyX Ltd.’). As the original dataset also contains some private companies, subsidiaries, and some companies listed outside the USA, not all of the observations are matched. Not matched observations are excluded from the dataset, as there is no information about corresponding companies’ financials in the available databases. Thus, examining the relationships is not possible.
After the matching procedure, the remaining number of observations in the dataset is 36,178 with 8,872 companies. Merging adds some companies to the set, that are never mentioned in any of the stories. They are used as a control group. Almost all of the stories are in English, but there are some exceptions. Stories in other languages, however, cannot have a direct impact on the stocks listed in the USA. If relevant, they are also repeated in the American media with the same corresponding story ID (it is always the same for all of the items describing one issue). Thus,
Page 12 of 46 deleting non-English items does not lose any important information. This action limits the dataset to 35,301 observations. All the steps and number of observations after performing each of them are in Table 2.
Table 2: Operations performed on the original dataset.
Step
number Step description companies after Number of performing the step
Number of observations after performing the step
1 Importing an original dataset 3,205 45,555
2 Adding to the dataset tickers and dropping observations with missing tickers.
1,713 26,770
3 Deleting stories in languages other
than English. 1,713 25,890
4 Generating variables 1,713 25,890
5 Dropping duplicates in terms of
year and a ticker simultaneously 1,713 5,670
6 Merging with annual data about the companies (table 5), rationing the data
8,872 55,527
The table shows all of the steps that I take to prepare the data set for performing the research and the number of companies and observations in the data set after performing described steps. An observation i,t is an ith company in a tth year.
To perform the merger, each observation has to be uniquely identified with the entity and time variables. This is not the case in this database (due to multiple items for some companies that published within one year). Thus, I cannot perform the merger without dropping the duplicates. To make sure that no information is lost during this action, I generate variables listed in Table 3. A variable ‘mainstream’ is generated by classifying the source of each item. If the item was described in the mainstream source, the variable is equal to 1, and 0 otherwise. It is meant to
Page 13 of 46 allow distinguishing these stories that may influence the company’s market value, from these that cannot have any impact. If a story was published only in a small, local newspaper, its influence could not be such significant as if it was published in one of the bestselling global magazines. However, most of the stories have multiple threads. In some cases, a story was first published in a smaller magazine and then broadcasted by the global media. In that case, the first story could also have a significant impact on a company’s performance. I generate a variable ‘mainstream2’ to reflect it. It counts a story as mentioned in the mainstream source if the author of any of the items describing the given story was categorized as one of the 100 most important global media sources, following the “4 International Media and Newspapers”. Nearly 1,500 items were classified as mentioned in the mainstream sources using this methodology.
As the dataset contains almost only string variables, the need to obtain information from the text is a priority. It can be done by generating numerical variables that count the amount of certain substrings within the text variables. Each story is labeled with up to 90 different tags, precisely describing the content of the item. As the number of unique tags within the dataset is limited, I am able to read all of them. I choose these related to human rights violations. Using these tags (as listed in table A.1 in the Appendix). I classify each story as related to this or not. I create a dummy variable called ‘HumanRights’. It is equal to 1 if any of the words included in the categories that describe the story covers with at least one of the relevant tags. The value of this variable is constant for each story and does not vary across the time. If there are a few items classified as the parts of the same story, and only some of them contain relevant tags, this variable takes the value 1 for each of the items- it is classified by the story, not by the separate items. The last frequent value of this variable is ‘response’, which means that an item is a
Page 14 of 46 statement issued by the company in response to a certain case, or an action took by the company after the story happened. I also create the variables ‘article’, ‘lawsuit’ and ‘response’. Each of them is equal to 1, if an item’s type is consistent with its name, and 0 otherwise.
The next step is to select these stories of each type, that are strictly related to the violated human rights. To do so, I integrate all of the variables generated in the previous step, with the variable ‘HumanRights’. As the outcome, three dummy variables (‘HRarticle’, ‘HRlawsuit’ and ‘HRresponse’) are generated. Each equal to 1 if the corresponding item was respectively an article, a piece of information about a lawsuit or a company response, related to the human rights issue, and 0 otherwise.
One of the categories used to generate the variable ‘HumanRights’ contains the name of the country in which did the story occurs. To obtain this information, I am using the identical method as in the case of that variable. Basing on that I generate three dummies. The first, ‘africa’ is equal to 1 if the corresponding story took place in any of the African countries. The next, ‘asia’ is equal to 1 if the story was based in any Asian country, excluding Japan. The last ‘americas’ takes the value 1 if the story happened in any country located in the Latin America (on both American continents, excluding the USA and Canada). Based on these three variables, I create another one, called ‘developing’. It is equal to the sum of the three mentioned above. Thus, it is equal to 1 if it happened in any of the countries, except for Canada, the USA, Japan, Europe, Australia, or Oceania, for which it is 0. These are the countries in which production costs are relatively high and the control procedures work correctly. Thus, this variable allows distinguishing the cases of violated human rights for which limiting the costs was the background.
Page 15 of 46
Table 3: The list of dummy variables
Variable Variable equal to 1 Number of observations
if =1 mainstream A story was mentioned at least once in a mainstream media source. 592 mainstream2 A company was mentioned at least once in a mainstream source. 11,138 HumanRights A story describes issues related to a violation of human rights 2,099
article An item is an article. 25,891
HRarticle An item is an article and it describes a violation of human rights. 1,929
lawsuit An item is a lawsuit. 97
HRlawsuit An item is a lawsuit related to the human rights violation. 32
response An item is a company’s response to any issue. 479
HRresponse An item is a company’s response to a human rights violation related issue. 82
africa A story took place in Africa. 3,955
americas A story took place in Latin America, South America or in Mexico. 1,307
asia A story took place in Asia. 331
developing A story took place in Asia, Africa, Latin America, South America or Mexico. 4,515
Total Number of observations (company-year) 25,891
Total number of companies 1,713
The table shows the dummy variables created in the dataset. I base some of the variables on anothers. I use them to construct independent variables for the research.
All of the dummies collect necessary information about each of the items. However, as mentioned before, to perform the databases merge, duplicates in terms of both a year and an entity have to be dropped. This action would lose information about the total number of relevant cases that happened in each year, leading to equal treatment of small incidents and large scandals. To avoid this problem- I create the set of counter variables, presented in Table 4.
Page 16 of 46 The first variable – ‘counterHRa’ counts the total number of the human rights related articles, that were published for a corresponding company in a given year. The highest value of this variable is equal to 14. That means that there was a maximum of 14 articles that described the same story in one year. Variables ‘counterHRl’ and ‘counterHRr’ are built analogically- they use respectively ‘lawsuit’ and ‘response’, instead of ‘article’ dummy. The maximum number of human rights-related lawsuits for one company within one year was equal to 1. One company took a maximum of 8 actions in response to human rights issues over one year.
Table 4: The list of counter variables
Variable Description Minimum Maximum N of observations if bigger than 0 counterHRai,t
Number of Human Rights Violation articles involving the ith company in the tth
year
0 14 721
counterHRli,t
Number of Human Rights Violation lawsuits involving the ith company in the tth
year
0 1 32
counterHRri,t
Number of Human Rights Violation related ith
company’s responses in the tth year
0 8 54
Total number of observations (company-year) 5,657
This table shows the variables counting each of the types of the stories that were published for each company in each year. Values are created basing on the dummies presented in table 2. Duplicates in term of both a year and an entity are dropped right after generating above variables. The last column shows values after that action is taken.
Once all of the described variables are created, I prepare the dataset to be merged with companies’ financials, which is the last step described in Table 2. The first step is to organize the
Page 17 of 46 dataset in such a way that each observation can be uniquely identified. The design of the research requires the use of panel data. Thus, the set has to be converted into this form. To do so, I sort all of the observations by ticker and by year. Afterward, I drop all of the duplicates in terms of both variables simultaneously. The construction of previously described variables allows performing this step without losing any relevant information. At this stage, the dataset has a panel form and it can be merged with companies’ financials, that come from the CRSP database. Once the merge is done, variables required to test hypotheses are generated, as described in Table 5.
The main hypothesis states that companies that violate their employees’ rights perform worse in the long run. The most complex measure of the company’s long term performance is its market value (Banz, 1981). To test this hypothesis, I build a regression model that uses the logarithm of the market value (‘value’) as a dependent variable.
My auxiliary hypotheses state what the reasons are for the slower market growth after a scandal is disclosed. A set of variables are created to test each of the reasons:
the decrease in profitability- I test it using a variable called ‘netin’, which is the ratio of net income over lagged total assets. Net income measures company’s profit. I divide it by the total assets, to make the value comparable between different companies, as well as between different years for the same company.
the decrease in sales- I examine the corresponding hypothesis using the log of sales (‘sales’), which represents the growth rate of sales level.
Page 18 of 46 lower investment rate- I check the causality regressing independent variables on the Research and Development expenses standardized by the lagged value of total assets (variable ‘rd’).
the higher price of new investments- I test this hypothesis using regressions that involve ‘capex’ as an independent variable. It is the ratio of Capital Expenditures over a year divided by the value of Property, Plant and Equipment (fixed assets) at the beginning of the year
higher operating costs- I examine this relationship with the use of ‘opex’ variable, which is the ratio of operating expenses to the total assets.
Basing on all of the variables created in the previous steps, I create independent variables that I include in the regressions. The first of them- ‘dcH’ measures the total number of a human rights related items, associated with each company in each year, under the condition that the story happened in one of the developing countries. Thus, it counts only the scandals, for which costs limitation was the background. Even though testing an overall influence of all stories on the dependent variables could give an idea of the overall relationship, I split all types of items to more accurately measure the impact of each of them. Variables ‘dca’, ‘dcl’ and ‘dcr’ show the number of the articles, lawsuits and company responses, respectively, related to each company in each year. Additionally, some control variables are going to be included in the regressions. ‘size’ is a logarithm of the total assets, which allows controlling for the company’s size dynamics. ‘debt’ -is the ratio of the total debt over total assets, which allows controlling for different effects for the companies with different financing structure. ‘FP’ is a variable based on the Standard Industry
Page 19 of 46 Classification Code. It is constructed the way to take value 1 if the company operates in the industry operating directly with final customers- in wholesale trade, retail trade, and services. For some of the regressions, I use average Sales, Capital Expenditures, and Operational Expenditures levels in the industry in a given year, which are represented by the variables ‘meansale’, ‘meancapex’, and ‘meanopex’ respectively. I calculate each of these variables as the mean of corresponding variable’s (‘sales’, ‘capex’, and ‘opex’) level in the company’s industry.
Page 20 of 46
Table 5: Definitions of the variables
variable item Description
Dependent Variables
valuei,t Log(PRCC_C×CSHO)
Log of market value, calculated as the product of the year’s closing price and the number of shares outstanding.
netini,t NI/lagged AT Net Income over lagged Total Assets
salesi,t Log(sale) Log of sales
R&Di,t XRD/lagged AT Research and Development expenditures over lagged Total Assets
capexi,t CAPX/lagged PPENT Capex over lagged value of Property, Plant, and Equipment
opexi,t XOPR/laggedAT Operating expenses over lagged Total Assets
Independent variables
lawsuiti,t developing ×counterHRli,t
A number of lawsuits against company ‘i’ in year ‘t’ regarding violating human rights in developing countries
articlesi,t developing ×counterHRai,t
A number of press articles describing the cases of violating human rights that name company ‘i’ published in year ‘t’
responsei,t developing ×counterHRri,t
A number of ‘ith‘ company responses to the human
rights violation related items in year ‘t’ in developing countries.
Control variables
sizei,t Log(AT) Log of Assets Total
debti,t DT/AT Total Debt to Total Assets Ratio
FPi,t SIC A company works directly with final customers.
meansalei,t mean(sales) by(ticker year) A mean of sales’ change rate in a given year in the industry.
meancapexj,t mean(capex) by(ticker year) Mean of capital expenditures over lagged fixed assets ratio in a given industry in a given year
meanopexj,t mean(opex) by(ticker year) Mean of operational expenses over lagged total assets ratio in a given industry in a given year
The main independent variable- ‘value’ represents annual changes of companies’ market capitalization. For nearly half of the observations this variable’s value is smaller than the mean (very close average and middle values). Small standard deviation shows a strong concentration around the mean. The biggest decrease of the market value a single company noted in one year
Page 21 of 46 was equal to -6.82%, while the largest increase to 13.38%. By deleting observations with missing market value, I exclude those that become bankrupt over the examined period. The biggest decrease of the market value might be bigger otherwise. The average annual growth rate is 6.22%.
On average, companies achieve the net income close to 8% of their total assets’ value. High standard deviation observed for this variable is caused by the large outliers present in the sample- small companies achieving high incomes and companies that noted loses.
The average annual growth of sales in the sample was equal to 5.99%. The biggest decrease in the sales level noticed by one company over a single year was equal to 6.21% while the biggest increase- over 13%.
The middling value of the ratio between research and development investment and the total assets in the sample is equal to 7%. The biggest value of this ratio is slightly smaller than 40%. Half of the companies are investing in the research and development less than 2% of their total assets value. This ratio is very concentrated within this small interval (0-2%), and strongly dispersed for bigger percentages.
Annual capital expenditures for half of the companies account for between 1% and 19% of their fixed assets, but because of the existence of many large outliers, mean of this ratio was equal to 34%. Operational expenditures account for between 0 and 40 times the value of total assets, with the middle value of 0.33.
The construction of independent variables used in the research does not allow presentation of many descriptive statistics. For most of the observations each of them was equal to 0- it is mainly
Page 22 of 46 due to a very large control group, which consists of all of the companies listed in the USA that was never proven to abuse employees.
The average rate of total assets’ value growth was close to 6.5% and the mean of the financial leverage was equal to 0.21. Some of the companies were financed fully by the equity, but some of them also had a really high debt share in the financing structure- even nearing 100%. On average, companies’ sales were growing by 6.05% a year. In the worst year, in the worst performing industry, sales did grow by approximately 1%. The best year for a single industry brought the companies average growth of sales exceeding 13%.
The last two variables, being industries’ averages, are distributed almost identically that the variables ‘capex’ and ‘opex’. Their distributions vary for each of the industries, but not in the general sample.
Page 23 of 46
Table 6: Descriptive statistics
variable Min max mean rd p50 N
Dependent Variables value -6.82 13.38 6.22 2.28 6.26 55,529 netin -6.91 61.63 0.08 1.59 0.03 55,528 sales -6.21 13.09 6.04 2.28 6.11 55,528 rd 0 0.39 0.07 0.10 0.02 31,144 capex 0.01 1.82 0.34 0.43 0.19 51,778 opex 0.00 39.48 4.40 9.93 0.33 17,783 Independent variables lawsuiti,t 0 1 - - - 55,527 articlesi,t 0 14 - - - 55,527 responsei,t 0 3 - - - 55,527 Control variables sizei,t 2.72 10.15 6.46 2.05 6.49 55,527 debti,t 0 1 0.25 0.21 0.21 55,527 FPi,t 0 1 - - - 55,529 meansalei,t 1.36 10.34 6.05 1.05 5.93 55,529 meancapexj,t 0.01 1.82 0.34 0.15 0.32 55,528 meanopexj,t 0.03 41.43 4.55 2.64 4.56 55,242
Page 24 of 46
4. Empirical Method
To investigate the effect of detected human rights violation incidents on corporate financial performance, I construct regression models that use the information included in the dataset. I format the set as panel data, where a year is the time variable and a ticker is the panel variable. Regressions use an Ordinary Least Squares method.
I identify an impact of scandals before and after they are disclosed, by using both forwarded and lagged values of independent variables. However, they are included in separate regressions- otherwise too many observations are omitted (for example, if I use 5 lags of a variable and there is financial information only about 7 years for a specific company, the model would have to be limited to the use of only 2 values of financial variables).
4.1 Testing the effect of disclosed scandals
To test the influence of scandals on companies’ financial performance, after they are disclosed, I use regression:
= + , + , + , + , + , + , + +
+ + , (1)
It contains values of variables measuring the amount of relevant items published for each company in a given year and a few years before. ‘dca’ is a number of relevant press articles, ‘dcl’ a number of lawsuits and ‘dcr’ is a number of company responses. I include the following lags as long as they are significant. In case, if coefficients are equal to zero, only the first lags are included. This regression shows the impact of disclosed scandals on the dependent variables (‘X’).
Page 25 of 46 is the entity fixed effect, and the time fixed effect. Logarithm of total assets and financial leverage ratio are included as control variables, to improve the model and avoid the omitted variables’ bias.
The main hypothesis states that scandals negatively influence companies market value. To investigate this, I use ‘value’ as a dependent variable. In case, if coefficients corresponding to the explaining variables are significantly smaller than zero, it means that published items negatively impact companies’ growth. The interpretation is that companies grow slower after scandals are published. If these coefficients are equal to zero, it means that scandals do not affect companies’ rate of growth. I interpret an impact of each type of items separately.
I use ‘netin’ as a dependent variable to estimate the influence of bad press on companies’ profitability. Negative coefficients mean that companies violating human rights are facing significant costs after the scandals are published. The ‘netin’ variable can be read as a sum of profits (in Euros) made of each Euro of total assets’ value. Thus, nonsignificant coefficients mean that scandals did not change the assets’ profitability.
To understand the role of consumers’ behavior in the whole process I use ‘sales’ as a dependent variable. It represents a growth of sales level. The model tests whether companies proved to abuse employees face a sales decrease. As an additional control I use average industry’s sales change in a given year. Significantly negative coefficients corresponding to the human rights variables show that consumers buy less from the companies that are involved in scandals. These coefficients statistically equal to zero mean that clients do not take the social performance of companies into consideration while making their decisions.
Page 26 of 46 To test whether companies previously involved in scandals are constrained against some investments I use ‘rd’ as a dependent variable. If coefficients corresponding to the explaining variables are significantly smaller than zero, hypothesis is confirmed. Non-significant coefficients mean that investments in development are not impacted by the scandals.
To examine if cost of new investment is affected after scandals, I run the same regression on the ‘capex’ variable. If coefficients - are significantly bigger than 0, a hypothesis is confirmed. These coefficients not statistically different from 0 mean that the scandals do not affect investment costs.
Regression on the ‘opex’ variable shows if the operational costs change after the scandals are published. Non-significant coefficients mean that operations are not harmed by the scandals.
4.2 Testing the effect before scandals are disclosed
To test the effect of sweatshops, before scandals are published I use similar regressions replacing lagged with forwarded values of explanatory variables. The general equation has a form of:
= + , + , + , + , + , + , +
+ + + , (2)
The main hypothesis states that companies benefit from cheaper labor before scandals are disclosed. Coefficients of the above regression on ‘value’ variable, significantly bigger than zero confirms the hypothesis. It means that companies that abuse employees grow faster than others. In case, if these coefficients are not significant, interpretation is that there are no benefits before scandals disclosure.
Page 27 of 46 I use a dependent variable ‘netin’ to identify an effect of sweatshops on companies’ profitability. Positive coefficients mean that companies that abuse employees are more profitable than others. I also use variable ‘sales’ to determine how abusing workforce influence sales level. The last of the auxiliary hypotheses states that the sweatshops’ background is cost limitation. The regression that uses ‘opex’ as dependent variable measures the operational costs before the scandals are disclosed. Reduced costs obtained due to abusing employees are reflected in significantly negative coefficients corresponding to the human rights variables. That means that the operational costs of companies harming the workforce are lower. Thus, this practice does make sense from the costs perspective- it helps with limiting them. Coefficients equal to 0 mean that sweatshops are not effective in reducing costs.
The described set of regressions provides a broad image of the whole mechanism leading companies to violate human rights. It allows understanding all aspects of financial performance of the companies that abuse their employees compared to these that are not proven to be doing so. This research allows not only testing the correlation between the financial results and social performance but also includes an explanation of each aspect of the described incidents. The results answer all of the questions asked in the hypotheses and in the title. They deliver an empirical evidence whether the companies benefit or lose on violating employees’ rights.
Page 28 of 46
5. Results
5.1. Main results
Table 7 presents the regressions that answer the hypothesis that there is a positive effect before, and negative after scandals’ disclosure. The first two models are the regressions measuring the effect of already disclosed scandals while two others show the effect prior to the scandals’ disclosure. The second and the fourth models correspond to the first and the third one, respectively, but they use standard errors adjusted for the clusters in the panel variable.
An impact of the articles that describe cases of violated human rights is negative and significant. It shows that these publications slow down companies value growth. However, there are no reasons to believe that this effect is different from 0 more than a year after the story is first published. Thus, the conclusion is that the companies that are proven to abuse employees grow slower than the others in the year when the scandal was disclosed and a year after. The non-significant coefficient for ‘dca’ values lagged more than by one year, means that the average speed of growth is the same for the companies that never violated employees’ rights and for these that did not do that last two years.
Does it mean that, in the long run, scandals do not affect companies value? Well, it does not. These results show that the punishment for the companies not operating ethically is a year or two of slower growth, which means in the future company value is smaller than it could be. However, that is not enough to address the question asked in the hypothesis. To find the long-term effect on the companies’ market value, the performance prior to the scandals’ disclosure is examined. To do so, I run a regression on a log of Market Value, which involves the forwarded
Page 29 of 46 values of the ‘dca’, ‘dcl’ and ‘dcr’ variables. As can be seen from the model 1.2 included in table 7, there are no statistical reasons to believe that the companies did benefit from abusing the workforce before the scandals were disclosed. There is no evidence that ‘unethical’ companies grow faster than the average.
In some cases, companies issue press statements, or take other actions after the publication of unfavorable articles. Most of them aim to explain the situation and to warm up the company’s picture or to reward employees. Taking such actions is actually beneficial for the company. Positive and significant coefficients suggest that it is effective for the companies to try warming up their image after they were accused for violating the work force rights. The effect before response were published, which is equal to 0, confirm that companies do not benefit from abusing employees.
Lawsuits, that were related to the violations of human rights have a strong, negative impact on companies’ growth rate a year after they begin. It means, that the growth in the year when the company is suited is not charmed, but it is much slower a year after. However, companies are growing again with rate similar to the average in the following years. No effect on the companies’ value growth before the lawsuits confirms that there are no benefits coming from taking advantage of employees.
These results show that companies indeed face negative consequences of abusing employees- they do not grow any faster before these scandals are disclosed, but there is an evidence that the decrease of the growth speed after publishing relevant articles is negative. The effect is especially strong for the lawsuits, but it can be reduced if company take actions in response to
Page 30 of 46 the scandals. Thus, the punishment, which is a year or two of the slower growth can be actually really substantial- even if afterward the growth rate is again similar to the industries’ average, they will be always less developed than they could be (a shift in the absolute market value).
Page 31 of 46
Table 7: Influence of human rights violation related scandals on the market value growth rate.
Model 1.1 Model 1.2 dca -0.03* (-0.02) L.dca -0.021** (-0.01) L2.dca 0.001 (-0.02) F.dca 0.005 (-0.01) F2.dca 0.013 (-0.01) dcr 0.168*** (-0.05) L.dcr 0.205** (-0.06) L2.dcr 0.085* (-0.04) F.dcr 0.038 (-0.06) F2.dcr 0.067 (-0.07) dcl 0.037 (-0.11) L.dcl -0.894*** (-0.07) L2.dcl 0 (.) F.dcl 0.087 (-0.07) F2.dcl 0.177 (-0.15)
Controls yes yes
Year fixed effects yes yes
Entity fixed effect yes yes
Clustered standard errors (by
entity) yes yes
Number of observations 37383 37383
Number of companies 5995 5995
The table provides results the Human Rights Violations related variables regressed on the log of Market Value. L.X and the L2.X mean the first and the second lag of X, while F.X and F2.X mean its first and second forward. For both regressions, a set of control variables, year fixed effect and entity fixed effects were used. All of the standard errors are corrected for the clusters in entities. The last two rows show a number of observations companies used in regressions.
Page 32 of 46
5.2. Additional results
But what is the mechanism behind these results? Are companies punished for not respecting their employees’ rights by the customers boycotting their merchandises and services? Or is the growth speed decrease caused only by the financial costs they face (indemnities and fines)? Or maybe the companies are constrained against performing new investments, which limits their growth possibilities? To answer these questions, and understand the mechanism, I run regressions presented in Table 8.
There is no statistical evidence that companies’ assets profitability was impacted by the scandals publications. This finding means that the profitability is not a medium by which market value is slowed down after scandals disclosure. There is no evidence, that companies benefited from increased profitability before the scandals were disclosed. However, significantly positive coefficient for the first forward of ‘dcl’ values, means that the net income to total assets ratio was higher for the companies a year before the human rights-related lawsuits. Does it mean that companies benefited before the lawsuits? Well, it does not. The non-significant coefficient for all of the following forwards of this variable, and for other variables, do not allow accepting this explanation. The most likely, companies that already know that the lawsuit is coming, are preparing themselves for that. They may be selling some assets and inventory, or cutting unnecessary expenses. It allows companies to cover the costs of the lawsuit.
Above statement finds a confirmation in the regressions on the sales level. A year before the lawsuit, the log of sales is much bigger than in the other years. This and the previous finding set together shows that companies try increasing cash levels right before the lawsuit starts- increasing sales (for example by selling stock on reduced price) can be one of the ways of doing
Page 33 of 46 that. The regressions also show a significant increase in sales a year after the lawsuit starts. These can be also interpreted as selling off the inventory due to the need of covering lawsuits related costs. The other coefficients in the regressions on the log of sales mean that there is no effect of disclosed scandals on sales dynamics. It means that customers do not make choices basing on the companies’ social performance.
Non-significant coefficients in the regressions on the research and development expenses mean that the scandals do not impact this category. Social responsibility does not influence the level of the investments related to the development.
There are statistical evidences that capital expenditures of the companies that were proved to abuse employees are higher than in the case of the others. The effect is the strongest in lawsuits’ year and next years. However, non-significant coefficients corresponding to the variables, which count the number of relevant press articles, mean that increased capex is not caused by the scandals itself. The strongest effect is observed for the lawsuits, but also actions that companies take in response to the scandals significantly increase these expenses. It means, that these are not the scandals to be responsible for increased expenditures, but the costs connected to the lawsuits and to the actions warming up companies’ image.
There is a very strong empirical evidence that operational expenses are much higher two years before companies take the response to the human rights related scandals, during the year when they do so and a year after. This effect is especially strong in a year before the action is taken. It means that warming up the public relations is very costly for the companies. This finding, set together with the fact, that there is no evidence that operational expenses are any lower before
Page 34 of 46 the scandals are disclosed, give a good idea what is the answer for the question whether abusing employees pay off.
All of the described findings show that companies have no interest in abusing employees. There are no reasons to believe that they benefit doing so, but there is a strong evidence that they pay a high price after the scandals are disclosed. Taking responses to scandals is very costly. They have to sell out the inventory, often losing a part of potential incomes. Also, all of the actions taken increase operational expenses. And finally, increased expenses related to lawsuits and actions taken, as well as weakening capital position- due to investors closing their investments- lead to the decrease of the market value’s dynamic. Thus, it shows that companies that abuse the workforce, lose on that. A clear answer to the question answered in the title shows up- abusing employees does not pay off.
Page 35 of 46
Table 8 Influence of human rights related scandals on companies financials.
netin netin Sales Sales R&D Capex Opex Opex
F2.dca -0.038 0.019* 0.108 -(0.02) -(0.01) (0.08) F.dca -0.072** 0.016 0.036 -(0.03) -(0.01) (0.03) dca -0.001 0.01 0 -0.002 0.028 (0.00) -(0.01) (0.00) (0.00) -(0.03) L.dca -0.001 0.003 0 -0.002 0.012 (0.00) -(0.01) (0.00) (0.00) -(0.02) F2.dcl 0.211 0.104 -0.682 -(0.11) -(0.08) (0.69) F.dcl 0.777*** 0.445*** 0.00 -(0.18) -(0.06) (.) dcl 0.022 0.071 0 0.080* 0.743 -(0.02) -(0.07) (0.00) -(0.03) -(0.58) L.dcl 0.005 0.106* 0 0.047** 0 -(0.01) -(0.04) (.) -(0.02) (.) F2.dcr -0.267 0.027 0.186*** -(0.28) -(0.08) (0.04) F.dcr -0.287 0.078 0.675*** -(0.37) -(0.11) (0.05) dcr -0.006** 0.024 -0.004 -0.005 0.269*** (0.00) -(0.05) (0.00) -(0.02) -(0.04) L.dcr 0.002 0.113** -0.004 0.025** 0.288*** (0.00) -(0.04) (0.00) -(0.01) -(0.05)
Controls yes yes yes yes yes yes yes yes Year fixed effects yes yes yes yes yes yes yes yes Entity fixed effect yes yes yes yes yes yes yes yes Clustered s. errors yes yes yes yes yes yes yes yes No. observations 37383 37383 37383 37383 37383 37383 37383 37383
No. companies 5995 5995 5995 5995 5995 5995 5995 5995 * p<0.05, ** p<0.01, *** p<0.001
The table provides results the Human Rights Violations related variables regressed on the Net Income over lagged Total Assets; Log of sales; Research and Development expenditures over lagged Total Assets; Capex over lagged value of Property, Plant and Equipment; and Operating expenses over lagged Total Assets ratios. L.X and the L2.X mean the first and the second lag of X, while F.X and F2.X mean its first and second forward. For all regressions, a set of control variables, year fixed effect and entity fixed effects were used. All of the standard errors are corrected for the clusters in entities. The last two rows show a number of observations companies used in regressions.
Page 36 of 46
6. Discussion
The thesis shows that companies that abuse employees lose money. The most likely reason for the companies to underpay the labor or not to provide sufficient working conditions is savings. However, even if it does save some money, these amounts are so small, that they do not matter for the entire company. Thus, corporates do not benefit from sweatshop practices. They do not reduce the level of operational expenditures, nor development and investment costs. Obviously, they also do not improve the quality of their services or products.
Back in the days, before the flying became cheap and internet broadly available, hiding sweatshop practices was way much easier. Now, the world seems to be much smaller than ever before. It also influences the approach of the public opinion. Western societies started caring about what is happening in the developing countries, and what are the working conditions for labor there. Also, many journalists dedicated their work to discovering these stories. Now it is much harder to hide sweatshops, which means that companies have to be aware of that what they are doing will come to light.
And it does come to light, causing serious implications for the responsible companies. Very often such a disclosures are followed by the lawsuits. Also, some companies afterward take actions meant to improve working conditions, but also to warm up their image. And both- lawsuits and these actions- cost. That means that a disclosure is already a strong signal that the company will have to cover many costs in a close future.
Page 37 of 46 A position of the company on financial markets decreases. Some investors quit the investment, because of the costs they forecast for the close future. However, there is also the other group of investors. Many institutions made a strong turn into high ethical standards in their investment policies. They simply do not want their capital to be involved in any company, which ethical performance may be questioned. Thus, they are withdrawing the money they invested after scandals are disclosed. What does it mean? It means that the market value of the company is charmed. Even if it is still increasing after the scandal was disclosed, it grows much slower than it could be.
So how much in the absolute terms do companies lose? As an example, let’s consider the company that was proved to abuse its employees in 2007, when it’s market value was equal to EUR 1Bn., and the scandal was described in 10 different sources. The average market value growth rate in 2008 and 2009 was almost equal to 6.3%. Thus, if it did not violate the workforce rights, it would be worth EUR 1.63Bn. in 2008 and EUR 1.13Bn. in 2009. Because of the scandal in 2008, it is worth EUR 3M. less and in 2009 its market value is EUR 5.41M. lower than it could be. And that is still not the end of the story. Even if after 2 years the company grows again with a normal rate each of the growths is from the smaller base. Thus, in 2015 this company is worth EUR 1.526Bn. instead of EUR 1.534Bn. As can be seen, just within the 7 years, its market value decreased by almost 7,500,000 euros. And what if the company was suited in that year? The decrease would be even stronger. In 2008 the market value would be EUR 12M. lower, and in 2014 over EUR 20M. lower than if the company respected employees’ rights. It means that the sweatshop practice would have to allow the company saving this amount to be profitable.
Page 38 of 46 Lawsuit cause much stronger consequences for the companies than just the scandals disclosures. Firstly, it means a huge level of the expenses company has to cover. All of the legal costs, possible fines, travel expenses etc. boost the costs. These costs have to be covered somehow, and it is never easy. Budgets are prepared well in advance, so paying for that from the normal operational activity is hard. Also, it is not possible to raise financing to do so- right before the lawsuit is probably the worst time to issue new shares. What do companies do then? They sell the stock they have, sometimes even some of the assets. It forces them to reduce prices. If the company is in a rush to sell the inventory, it receives less money than it would have received in normal conditions. Thus, also profitability is charmed.
If the company was proved to violate workforce rights, and especially if it was suited afterward, it may be forced to take some actions to address the accusations. This action may be a new social program for the employees, factory renovation, salaries raise, etc. It may be also some PR actions organized to warm up the company’s image. These actions are very costly, which can be seen in the very high capital and operational expenditures when they are taken. To finance them, companies are forced to increase sales just as in the case of lawsuits. Once again, it is a loss of money company could avoid if they just operated ethically.
Given that, companies have to consider why do they mistreat labor. The most likely, the whole process starts at the bottom. Management of separate factories see great savings in these practices. Indeed, the saving of underpaying the staff in a single factory may strongly cut the costs this single factory bears. However, these savings are nothing from the financial perspective of the whole global corporate. Thus, the money saved become actually invisible. But when the sweatshop is discovered, it charms the entire corporate. As shown before, the loss in the market
Page 39 of 46 value even close to 1.3% within the first 7 years from the disclosure and even later is huge. And it cannot be addressed by small savings of a single factor.
7. Conclusion
In the time of a strong turn into high ethical standards in finance, there are some still non-solved issues. Ethical investments give investors higher returns- as shown in Kempf and Osthoff (2007). More and more investors decide not to support companies with poor ethical performance. It means, that companies which social performance may be questioned, find it harder to raise capital. This finds confirmation in Ghoul, et al. (2011), that shows that the higher the ethical standards, the cheaper the equity. Stocks of unethical companies are also characterized with higher variance, so with higher risk, as shown in Becchetti and Ciciretti (2009).
Even though all of the facts are showing that is does not pay off to abuse employees, some companies still keep on doing that. It is because people tend to see the short-term benefits as certain but they believe that they can avoid long-term consequences. Companies decide on the sweatshop practices even though the benefits are very low and a potential risk is very high. Possible gains from the sweatshops may seem high on the local levels, but does not matter at the level of the entire corporate. Consequences of discovered scandals are very strong and affect all of the stakeholders. The most important loss is the market value decrease, that will be never made up. No matter what actions company takes, its market value is always lower than it could be if it never abused labor.
Page 40 of 46 The change has to be started at corporates’ top level managements. The companies that do not turn into high ethical standards may face serious problems in the future. This thesis shows that it is in the best interest of companies’ owners and top level managements to increase control and prevent unethical incidents to happen at the bottom levels. I show an evidence, that even though sweatshops may seem profitable- they are not. Companies lose money on abusing employees.
8. References
1. Ammann, M., Oesch, D., & Schmid, M. M. (2011). Corporate governance and firm value: International evidence. Journal of Empirical Finance, 18(1), 36-55.
2. Banz, R. W. (1981). The relationship between return and market value of common stocks. Journal of financial economics, 9(1), 3-18.
3. Barber, B.M. and Lyon, J.D., 1997. Detecting long-run abnormal stock returns: The empirical power and specification of test statistics. Journal of financial economics, 43(3), pp.341-372. 4. Basu, K., & Van, P. H. (1998). The economics of child labor. American economic review,
412-427.
5. Baye, M.R. and Beil, R.O., 2006. Managerial economics and business strategy (Vol. 5). New York, NY: McGraw-Hill.
6. Becchetti, L., & Ciciretti, R. (2009). Corporate social responsibility and stock market performance. Applied financial economics, 19(16), 1283-1293.
7. Brammer, S., Brooks, C. and Pavelin, S., 2006. Corporate social performance and stock returns: UK evidence from disaggregate measures. Financial management, 35(3), pp.97-116.
Page 41 of 46 8. Carhart, M.M., 1997. On persistence in mutual fund performance. The Journal of
finance, 52(1), pp.57-82.
9. Chesterman, S., 2008. The Turn to Ethics: Disinvestment from Multinational Corporations for Human Rights Violations-The Case of Norway's Sovereign Wealth Fund. American University International Law Review, 23, pp.577-615.
10. de Haan, M., Dam, L., & Scholtens, B. (2012). The drivers of the relationship between corporate environmental performance and stock market returns.Journal of Sustainable Finance & Investment, 2(3-4), 338-375.
11. Deng, X., Kang, J. K., & Low, B. S. (2013). Corporate social responsibility and stakeholder value maximization: Evidence from mergers. Journal of Financial Economics, 110(1), 87-109. 12. Ek, V.M.C., 2015. Ruggie, John Gerard, Just Business, Multinational Corporations and Human
Rights, Norton, Amnesty International Global Ethics Series, 2013. Anuario Mexicano de Derecho Internacional, 15, pp.881-886.
13. El Ghoul, S., Guedhami, O., Kwok, C. C., & Mishra, D. R. (2011). Does corporate social responsibility affect the cost of capital?. Journal of Banking & Finance, 35(9), 2388-2406. 14. Fama, E. F., & French, K. R. (2004). The capital asset pricing model: Theory and
evidence. Journal of Economic Perspectives, 18, 25-46.
15. Fama, E.F. and French, K.R., 1992. The cross-section of expected stock returns. the Journal of Finance, 47(2), pp.427-465.
16. Feldman, S.J., Soyka, P.A. and Ameer, P.G., 1997. Does improving a firm's environmental management system and environmental performance result in a higher stock price? The Journal of Investing, 6(4), pp.87-97.
Page 42 of 46 17. Filbeck, G. and Gorman, R.F., 2004. The relationship between the environmental and financial performance of public utilities. Environmental and Resource Economics, 29(2), pp.137-157. 18. Gompers, P. A., Ishii, J. L., & Metrick, A. (2001). Corporate governance and equity prices (No.
w8449). National bureau of economic research.
19. Goss, A., & Roberts, G. S. (2011). The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance, 35(7), 1794-1810.
20. Hong, H., & Sraer, D. (2012). Speculative betas (No. w18548). National Bureau of Economic Research.
21. Hopper, T., Northcott, D. and Scapens, R.W., 2007. Issues in management accounting. Pearson education.p. 25
22. Kamatra, N., & Kartikaningdyah, E. (2015). Effect Corporate Social Responsibility on Financial Performance. International Journal of Economics and Financial Issues, 5.
23. Kasimati, E. (2007). Corporate governance and stock returns: evidence from Greece. Journal of Interdisciplinary Mathematics, 10(1), 149-158.
24. Kempf, A. and Osthoff, P., 2007. The effect of socially responsible investing on portfolio performance. European Financial Management,13(5), pp.908-922.
25. Kohler, W., 2004. International outsourcing and factor prices with multistage production*. The Economic Journal, 114(494), pp.C166-C185.
26. MacKinlay, A.C., 1997. Event studies in economics and finance. Journal of economic literature, 35(1), pp.13-39.
27. Manescu, C., 2009. Is corporate social responsibility viewed as a risk factor? Evidence from an asset pricing analysis.
Page 43 of 46 28. Mǎnescu, C., 2011. Stock returns in relation to environmental, social and governance performance: Mispricing or compensation for risk? Sustainable development, 19(2), pp.95-118.
29. Margolis, J.D. and Walsh, J.P., 2001. People and profits?: The search for a link between a company's social and financial performance. Psychology Press.
30. McWilliams, A. (2000). Corporate Social Responsibility. Wiley Encyclopedia of Management. 31. Naseem, I. (2012). Developing World is Not the Only Custodian of Child Labour.Journal of
Research in Economics and International Finance, 1(7), 179-186.
32. Peiris, D., and Evans, J., 2010. The relationship between environmental social governance factors and US stock performance. The Journal of Investing, 19(3), pp.104-112.
33. Prasad, M., Kimeldorf, H., Meyer, R., & Robinson, I. (2004). Consumers of the World Unite- A Market-based Response to Sweatshops. Labor Studies Journal,29(3), 57-79.
34. Ramasastry, A. (2002). Corporate Complicity: From Nuremberg to Rangoon-An Examination of Forced Labor Cases and Their Impact on the Liability of Multinational Corporations. Berkeley J. Int'l L., 20, 91.
35. Rock, M.T., 2003. Public disclosure of the sweatshop practices of American multinational garment/shoe makers/retailers: Impacts on their stock prices. Competition & Change, 7(1), pp.23-38.
36. Rosen, E. (2002).Making sweatshops: The globalization of the US apparel industry. University of California Press.
37. Scholtens, B., 2008. A note on the interaction between corporate social responsibility and financial performance. Ecological economics,68(1), pp.46-55.
Page 44 of 46 38. Sievänen, R. (2013). The non-response of pension funds to climate change and human
rights. Journal of Sustainable Finance & Investment, 3(3), 204-222.
39. Starr, M. A. (2008). Socially responsible investment and pro-social change. Journal of Economic Issues, 42(1), 51-73.
40. Sufian, M. A., & Zahan, M. (2013). Ownership structure and corporate social responsibility disclosure in Bangladesh. International Journal of Economics and Financial Issues, 3(4), 901. 41. Von Arx, U. and Ziegler, A., 2008. The effect of CSR on stock performance: new evidence for
the USA and Europe. CER-ETH-Center of Economic Research at ETH Zurich, Working Paper, (08/85).
42. Weissbrodt, D. and Kruger, M., 2003. Norms on the responsibilities of transnational corporations and other business enterprises with regard to human rights. The American Journal of International Law, 97(4), pp.901-922.
Page 45 of 46
List of Tables
Table 1: List of the variables in the original dataset ... 10
Table 2: Operations performed on the original dataset. ... 12
Table 3: The list of dummy variables ... 15
Table 4: The list of counter variables ... 16
Table 5: Definitions of the variables ... 20
Table 6: Descriptive statistics ... 23
Table 7: Influence of human rights violation related scandals on the market value growth rate. ... 31
