The Effect of Industry Competition on
Corporate Social Performance
Master thesis Amsterdam School of Economics Business Economics: Managerial Economics and Strategy
August 2018
Author: Stef Mureau Student Number: 11942851 Supervisor: Dr. A.M. Onderstal
1
Executive Summary
In this paper, we investigate the relationship between the level of market competition and the corporate social performance of firms. We analyse how competition affects overall corporate social responsibility (CSR) ratings, the propensity of firms to engage in CSR-driven differentiation strategies and how competition affects different dimensions of CSR. Using a large sample of U.S. firms, we find that increased levels of market competition improve corporate social performance by significantly reducing the level of negative social actions by firms. These results hold regardless of whether we consider firm or industry-levels of social performance. By studying the within-industry variability in CSR scores, we do not find strong empirical evidence that firms engage more in CSR-driven differentiation strategies in competitive environments. Instead, we observe that social concerns converge to a minimum level as competition intensifies. Additionally, our results suggest that firms prefer to engage in environment-, product-, and employee-related CSR when faced with increased competition. Keywords: market competition, CSR, social performance, competitive advantage
Statement of Originality
This document is written by Stef Mureau, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are 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.
2
Table of Contents
EXECUTIVE SUMMARY 1 TABLE OF CONTENTS 2 1. INTRODUCTION 3 2. THEORETICAL FRAMEWORK 5 2.1 THE CONCEPT OF CORPORATE SOCIAL RESPONSIBILITY 6 2.2 ALTRUISTIC CSR 7 2.3 STRATEGIC CSR 7 2.4 THE RELATIONSHIP BETWEEN MARKET COMPETITION AND CSR 9 3. DATA AND EMPIRICAL STRATEGY 11 3.1 SAMPLE CONSTRUCTION 123.2 MEASURING SOCIAL PERFORMANCE 12
3.3 MEASURING INDUSTRY COMPETITION 15 3.4 CONTROL VARIABLES 17 3.5 EMPIRICAL STRATEGY 18 4. RESULTS 21 4.1 SUMMARY STATISTICS 21 4.2 PRELIMINARY RESULTS 23 4.3 MULTIVARIATE ANALYSIS 25 4.4 ROBUSTNESS TESTS 33 5. CONCLUSION 36 6. DISCUSSION 38 7. REFERENCE LIST 42 8. APPENDIX 46
3
1. Introduction
The effect of market competition on the financial performance of firms is a fundamental topic in the economics literature. However, the question of how competition affects the social performance of firms remains largely unstudied. While it is normally regarded as a desirable market characteristic, helping companies to constantly strive for innovation and efficiency, fierce market competition is also believed to induce a wide range of unethical social behaviour by firms (Shleifer, 2004). As companies have greatly increased their expenditures on corporate social responsibility (CSR) over the years (Barnea & Rubin, 2010) and more than 90% of the G250 companies currently publish annual CSR reports (KPMG, 2015), the question as to why firms engage in socially responsible activities has become more prevalent.
In attempts to answer this question, the majority of empirical studies have focused on analysing the link between CSR and financial performance. After all, investing considerable amounts of resources in CSR would be undesirable if firms did not anticipate some sort of benefit from these investments. While the relationship between firm profitability and CSR has been the subject of many empirical studies, results have remained largely inconclusive (Griffin & Mahon, 1997; Margolis & Walsh, 2003). These mixed findings are most likely the result of inconsistencies in defining and measuring CSR, inconsistencies in measuring financial performance, or both. Even though the results of these studies remain largely inconclusive, two opposing views on the motivation behind CSR have gained popularity; the altruistic view and the strategic view. The altruistic approach to CSR poses that firms are prepared to sacrifice profits for the social interest (Elhauge, 2005), while the strategic approach to CSR views CSR as a profit-maximizing strategy in which firms anticipate a benefit from engaging in socially responsible behaviour (Baron, 2001; Siegel & Vitaliano, 2007).
In this paper, we take a rather different approach to exploring why firms invest in CSR. Instead of directly investigating the relationship between CSR and financial performance, we examine whether the level of market competition is positively associated with firm social performance and explore whether CSR might be used as a differentiation strategy by firms. Consequently, the primary question we aim to answer in this study is:
4 As McWilliams and Siegel (2001) noted, CSR can be an important means of differentiation from competitors, as well as a means to satisfy stakeholders’ and managers’ interests. As the level of competition in a market intensifies, differentiation from competitors can prove to be a valuable business strategy for firms. When considering either an altruistic or strategic approach to CSR, however, we would expect increased market competition to have an opposite effect on these two types of motivations behind CSR. As fierce market competition generally reduces the economic profits of firms, intense market competition would reduce socially responsible behaviour of firms through a lower amount of resources available for CSR. If we would assume CSR to be strategically motivated however, a more competitive environment would lead to higher levels of corporate social activity as firms can use CSR as a method of differentiation from its competitors. As a result, by looking at the effect of market competition on firm social performance, we aim to shed more light on the motivations behind investing in CSR for firms. Investigating how the level of market competition affects the social performance of firms can provide valuable insights into why firms might engage in costly socially responsible behaviour. Since the level of competition in a market is generally overlooked in empirical research on CSR, this can pose limitations to our understanding of corporate social activity. This paper aims to address the lack of empirical research on the role of market competitiveness on firm social performance and seeks to further examine the applicability of CSR as a differentiation mechanism. Also, increasing our understanding of how concentration levels influence both firm- and industry-levels of CSR might provide useful information for governmental policy makers as they reconsider industry competition policies.
In answering our main research question, we not only look at firm-level CSR, but also how market competition affects industry levels of social performance. To study whether competition affects the propensity of firms to differentiate themselves from competitors, we investigate how the level of competition affects the variability in CSR scores across industries. Additionally, we examine how the level of competition affects the seven social dimensions of the KLD database separately in order to increase our understanding of how competition affects different aspects of CSR. A newly developed measure of market concentration, which more accurately estimates the level of market concentration, allows us to analyse the effect of changes in market concentration on social performance more precisely.
5 Overall, the results of our analysis indicate a consistent pattern of improved social performance of firms as a result of increased market competition. Contrary to what might be initially expected, we find that the improved social performance of firms is the result of significantly less negative social actions instead of an increase in the level of positive social action. Additionally, we do not find strong empirical evidence for higher levels of CSR-driven differentiation in competitive industries, instead, the results suggest that the number of social concerns converge to a minimum level in these competitive industries. By analysing the effect of competition on each individual dimension of CSR, we find that not all dimensions are affected by the level of competition in the same way and that firms prefer to reduce employee-related concerns and increase corporate governance-related when faced with increased competition.
This paper contributes to the relatively scarce literature on the market competition-CSR relationship by using a recently developed measure of market concentration to investigate how the level of market competition affects the social performance of firms. In analysing this relationship, we not only look at aggregate CSR levels, social concerns, social strengths or the different CSR dimensions, but also at how competition affects the propensity for firms to use CSR as a differentiation strategy. By investigating the relationship between competition and CSR, we aim to indirectly answer whether CSR is mainly strategically or altruistically motivated.
Our study is organized as follows. In section 2, we discuss and evaluate the existing body of literature and prior studies on our topic. In section 3, we describe the sample, variables and empirical method used in our analysis. In section 4, we present the results of our univariate and multivariate models, followed by several robustness tests. In section 5, we summarize our main conclusions, followed by a discussion of potential limitations and possible steps for future research in section 6.
2. Theoretical Framework
In this section, we examine the existing body of literature on our topic and evaluate prior studies on the relationship between competition and CSR. In section 2.1 we first discuss and define the broad concept of CSR. In section 2.2 and 2.3, we examine two opposing views on the motivation for CSR; the altruistic view and the strategic view. In section 2.4, we look at the relationship between market competition and the social performance of firms in more detail.
6
2.1 The concept of corporate social responsibility
Before developing the main arguments linking competition to corporate social responsibility, it is convenient to first define what is meant by the term corporate social responsibility. Even though scholars and researchers have devoted much attention to CSR over the last few decades, there is still no universally accepted definition of the term (Dahlsrud, 2008; Sheehy, 2015). Prior literature has often defined corporate social responsibility as actions that appear to further some social good, beyond the interests of the firm or that which is required by law (McWilliams & Siegel, 2001). Others have highlighted the valuable role of CSR in the implicit social contract between business and society, with possible mutual gains for both sides (Baron, 2001; Davis, 2005). The European Commission (2001), however, defined CSR as ‘a concept whereby companies integrate social and environmental concerns in their business operations and in their interaction with their stakeholders on a voluntary basis’. As the concept of corporate social responsibility remains incredibly broad, scholars are yet to come to a consensus on the most appropriate definition of the term. However, in order to have a clear understanding of what we consider CSR to be, we adopt the definition set out by the European Commission for the remainder of our study.
Despite the vast literature available on CSR, explanations as to why firms engage in socially responsible activities are still widely debated. While some firms might engage in socially responsible behaviour as a result of new regulations or in anticipation of social pressure (Kotchen and Moon, 2012), there is also a large and growing literature on the relationship between corporate social performance and financial performance of firms (Waddock & Graves, 1997; McWilliams & Siegel, 2000; Margolish & Walsh, 2003; Orlitzky et al., 2003). While the majority of studies find a positive correlation, this view is not supported by many other researchers (Waddock & Graves, 1997; Wright & Ferris, 1997; McWilliams & Siegel, 2000). Since researchers seem to find contradictory or ambiguous results, questions remain as to why companies engage in CSR. In recent literature, however, two competing views on the motivations behind CSR have gained popularity; the altruistic view and the strategic view.
7 2.2 Altruistic CSR According to the altruistic approach to CSR, firms are willing to sacrifice profits for the sake of social interest and therefore, CSR is used in a purely moral perspective (Elhauge, 2005). Firms that engage in altruistically motivated CSR can potentially incur a competitive disadvantage as this generally leads them to have higher costs compared to firms that do not engage in altruistic CSR (Waddock & Graves, 1997). Additionally, when a firm’s higher management is altruistically motivated, it is more likely to refrain from unethical business practices, even when it might be profitable to engage in these business practices (Rosen et al., 1991). Bénabou & Tirole (2009) argue that stakeholders can have demand for corporations to engage in philanthropy on the behalf of the stakeholders, as transaction costs are likely to be lower if this philanthropic behaviour goes through the firm. With respect to our previously adopted definition of CSR, this implies that companies integrate social and environmental concerns in their business operations as they are willing to sacrifice profits for the sake of the social interest.
In modern day firms, an improved firm social performance could be the result of altruistic behaviour by managers or other owner-shareholders, or by managers who want to extract private reputational benefits from increased social performance. This latter view is based on agency theory, which implies that CSR is essentially a misuse of company resources and it would be better to allocate these resources either to projects that add value to the firm or return them to shareholders. Barnea and Rubin (2010) find evidence for higher investment in CSR by firm insiders (managers, directors and large blockholders) when they bear little of the costs of doing so. Following this line of reasoning, overinvestment in CSR can be viewed as an additional managerial perk and another symptom of the agency problem between insiders and other shareholders.
2.3 Strategic CSR
On the contrary, the strategic view of CSR asserts that firms engage in socially responsible behaviour to gain a competitive advantage and therefore engage in profit maximizing CSR. Companies are assumed to be socially responsible because they anticipate some benefit from these socially responsible actions (Siegel & Vitaliano, 2007). Following this line of reasoning, CSR is considered to be an additional business strategy, similar to any other business strategy or action directed at maximizing a firm’s
8 profits. When we consider our above-mentioned definition of CSR, this implies that firms integrate social and environmental concerns in their business operations and in their interaction with their stakeholders in order to maximize profits.
Many theoretical explanations linking social performance to improved financial performance have been presented in existing literature. The underlying belief is that some competitive advantage can be achieved by distinguishing oneself from other companies through engaging in CSR activities. For instance, an improved social performance of firms can increase a consumers’ willingness to pay for a product, enabling firms to charge a price premium for its products (Waddock & Graves, 1997; Reinhardt, 1998; McWiliams & Siegel, 2006; Bagnoli & Watts, 2003), grant access to new markets or attract more socially responsible customers and investors (Siegel & Vitaliano, 2007). It can also increase the willingness of people to work for a firm while simultaneously improving current employee morale and retention (Turban & Greening, 1997; Backhaus, 2002). If a product or service has important attributes that cannot be observed after purchase (i.e. credence goods), ethical behaviour of firms sometimes be used to signal the unobservable high quality of products or services to potential customers (Feddersen & Gilligan, 2001; Fisman et al., 2006). Additionally, since consumers tend to not only base their purchasing decisions on the utility derived from a product or service, but also on the reputation of a firm, enhancing firm reputation through an improved social performance can increase the demand for a product. Folkes and Kamins (1999) show that superior product attributes have a higher positive impact on consumer attitudes towards ethically behaving firms than towards unethically behaving firms, further stressing the importance of corporate social performance.
As support for the strategic view on CSR has grown over the years, more and more researchers have devoted significant effort to empirically studying whether there is indeed a positive relation between social performance and financial performance. A positive relationship between CSR and financial performance would, after all, lend support to the idea that firms engage in profit maximizing CSR. While the majority of studies indicate a positive correlation between social performance and several indicators of financial performance, some do not find any correlation and a few studies even find a negative correlation (see Margolis & Walsh, 2003, for an extensive literature overview). However, even if a majority of studies find a positive relationship, this cannot be considered as direct proof for the profit maximizing use of CSR. While it does
9 indicate CSR could potentially be used by companies to obtain a strategic advantage of some sort, it could also be an indicator that profits are being used to improve a firms’ social performance. This issue of reverse causality has been pointed out in prior literature: do increased levels of CSR improve financial performance, or are higher levels of CSR the result of improved financial performance (Siegel & Vitaliano, 2007; Benabou & Tirole, 2010; Kotchen & Moon, 2012)? In an attempt to study the direction of the relationship between social performance and financial performance, Waddock and Graves (1997) empirically investigate the relationship using corporate social performance as both a dependent and independent variable in their models. As they find a positive relationship for both models, they argue that, in line with the slack resources theory, firms benefit from having more slack resources available due to stronger financial performance and, consequently, also have greater freedom to invest in CSR.
2.4 The relationship between market competition and CSR
Market competition and the competitive pressure it produces is a fundamental determinant in the way companies run their businesses. It is often considered to be one of the main driving forces behind the provision of quality goods and services and helps companies to remain innovative. However, as the level of competition becomes so intense that firms struggle to survive, managers may feel inclined to turn to more unethical business practices in order cut costs and stay in business (Fraedrich, 1992). Shleifer (2004) presents five examples of unethical business practices as a result of fierce market competition: child labour, corruption, executive pay, earnings manipulation and even commercial activities by universities. While Shleifer’s evidence on unethical firm behaviour is mostly anecdotal, there have been some empirical studies that investigate the relationship between market competition and (un)ethical behaviour.
Cai and Liu (2009) use data on Chinese industrial firms to examine the relationship between market competition and the propensity to avoid corporate income tax and show that competition is positively associated to the level of corporate tax avoidance. Using extensive datasets on general practitioners (GPs) and patients in Norway, Markussen and Røed (2017) and Brekke et al. (2017) find that GPs are more likely to issue sickness certificates in competitive environments than in
non-10 competitive environments, suggesting that the decision to issue sickness certificates may be affected by private economic interests. Additionally, Zhang et al. (2010) find evidence for higher levels of corporate philanthropic giving by firms in competitive industries after the 2008 Sichuan earthquake compared to firms in less competitive industries. This suggests firms might be using corporate philanthropic giving strategically in order to differentiate themselves from competitors or improve firm reputation.
The examples of studies presented above generally focussed on certain aspects of CSR, such as corporate philanthropic giving or corporate tax evasion or avoidance. When we consider empirical studies that examine the relationship between competition and corporate social performance by using datasets on the CSR scores of firms as dependent variables, the results remain rather contradictory. Graafland (2016) hypothesizes that intense price competition lowers the environmental performance by inducing short-termism in companies. By analysing a large dataset on more than 3000 firms from twelve different European countries, he finds evidence for a lower environmental performance of firms in competitive industries resulting from a shorter time horizon that companies in these competitive industries apply in their strategic decision making. Lee et al. (2018) find a similar negative relationship between competition and CSR using a sample of Korean firms listed on the Korean Exchange. In contrast, Fernandez and Santaló (2010) show that U.S. firms in more competitive environments have higher social ratings as a result of more positive and less negative social activity. Using a different approach, Flammer (2015) examines whether and how competition affects CSR by testing if foreign competition (measured by tariff reductions) affects CSR investments of domestic U.S. firms. The results indicate an increase in the social performance of domestic U.S. firms when these firms face greater foreign competition, suggesting CSR is being used by domestic U.S. firms as a competitive strategy and to differentiate themselves from foreign competitors.
As there is no clear consistent result from prior empirical research on the link between competition and CSR, we do not formulate hypotheses in our research, instead, we aim to explore the relationship between competition and CSR in more detail. By studying this relationship, however, we also attempt to answer whether CSR is either strategically or altruistically driven by using a different approach. As competition has contradictory effects on strategic and altruistic CSR, we aim to indirectly answer
11 whether CSR is mainly strategically or altruistically driven by investigating the relationship between competition and CSR.
If CSR were to be driven by altruistic preferences of either managers or other corporate directors, we would expect intense market competition to lead to decreasing levels of corporate social performance. In classic economic theory, intense market competition is believed to reduce the price mark-ups of products and consequently, the strength of a firm’s economic position (Nickell et al., 1997; Aghion et al., 1999). Therefore, more competition in a market would decrease the rents available for CSR activity under the altruistic view and as a result, decrease CSR levels (Fernández-Kranz & Santaló, 2010). Even if, as argued under the agency view of CSR, managers divert economic rents of the company to socially responsible activities purely to extract private reputational benefits and advance their own careers, intense competition would still diminish the amount of resources available for socially responsible behaviour and therefore decrease overall CSR levels. From this point of view, a negative relationship between market competition and CSR levels holds, regardless of whether altruism is the true source of CSR or CSR is used to extract private reputational benefits.
Conversely, when we consider the effect of competition on the use of strategic CSR, we would not expect the same negative relationship between product market competition and CSR levels. If CSR activities are primarily undertaken because of their possible competitive advantage for the firm, a more competitive environment could lead to an increase in corporate social behaviour as firms might use CSR as means of differentiation or as a way to obtain a competitive advantage. It is important to note that while CSR might be completely altruistically motivated, this does eliminate any possible competitive advantage or other strategic advantage as a result of this socially responsible action. CSR could actually be altruistically motivated but still be strategically chosen as well.
3. Data and empirical strategy
In the following section, we describe the sample and variables used in our empirical analysis and outline the method required for determining the effect of market competition on corporate social performance. First, we describe the sample used in our study. Second, we explain why we use KLD data to measure social performance and how
12 we construct several variables from this dataset. Third, we clarify how industry competition is measured and discuss several advantages to the measure of industry concentration that we use in our analysis. Fourth, we illustrate what control variables we include in our models and in the fifth and final paragraph, we outline the empirical strategy we apply in our research.
3.1 Sample construction
In order to properly analyse the effect of market competition on firm social performance, we combine three different sets of data; the KLD Social ratings dataset on the CSR behaviour of firms, Compustat data on yearly company balance sheet data and Keil’s (2017) dataset on industry concentration levels. The KLD dataset was merged with Compustat using eight-digit firm CUSIP numbers and Keil’s dataset was merged with Compustat by six-digit North American Industry Classification System (NAICS) codes. After combining these datasets, we end up with an unbalanced sample of 14,326 firm-year observations for 3,304 unique firms and a total of 3,085 industry-year observations for 525 industries. The average number of years per firm in our dataset is 4.3 years, while the average number of years per industry in our sample is 5.9 years.
Even though the sample consists of firms from virtually every industry, the manufacturing industry (NAICS code 31-33) is the largest group in our sample (35.83%), followed by firms in the Finance and Insurance industry (NAICS code 52, 20.86%) and firms in the information sector (NAICS code 51, 9.31%). The sample covers the period from 2002 to 2009. Over the years, the distribution of industries in our sample does not change significantly. 3.2 Measuring Social Performance In order to measure the level of corporate social responsibility of firms, we use data available from the KLD Social Ratings Database. This database was created in 1991 by Kinder, Lydenberg, Domini Research & Analytics and has been the most widely used dataset on the corporate social performance of firms in recent academic literature (Kotchen and Moon, 2012). The ratings are based on annual evaluations of firms using several sources, including public records and media reports, and on assessments of experts outside the respective firms. As a result, KLD ratings are generally more objective than self-reported CSR activities.
13 KLD data considers seven major qualitative issue areas: community, diversity, corporate governance, employee relations, environment, human rights and product quality and safety. Within each of these seven issue areas, the social performance is analysed on various aspects or activities, differentiating between socially responsible behaviour (strengths), and socially irresponsible behaviour (concerns). Every indicator is assigned a binary value of either “1”, indicating the presence of the corresponding strength or concern, or “0” if the company did not have a strength or concern in that particular indicator. The coverage of KLD data has grown substantially over the years, originally starting with 650 companies and now containing over 3000 of the largest US companies by market capitalization (Kinder et al., 2006).
Using KLD scores provides several advantages. One advantage of using KLD scores is that it allows for differentiation between negative and positive social actions. Contrary to several other measures of corporate social performance, this makes it possible to differentiate between companies only engaging in socially responsible activity, and companies that engage both in socially responsible and irresponsible activity. As intense market competition is sometimes believed to encourage unethical corporate behaviour, the measurement of social concerns in KLD data allows us to analyse whether this is true. Another possibility is that, in response to intense market competition, firms reduce their social concerns instead of increasing their social strengths, which would not be detectable when only using a single aggregate CSR score in empirical analysis. Additionally, since KLD data covers many different dimensions of CSR, it is possible to examine how competition affects different dimensions of corporate social behaviour or whether some dimensions of CSR are more likely to be used as a competitive strategy by firms.
Using the available KLD data, we construct several measures of social performance to use in our analysis. The first measure is the aggregate CSR score for each firm in each year which is formed by computing the difference between a firm’s total strengths and total concerns across all seven different social dimensions. This is a measure that is frequently used in empirical research. However, as pointed out in previous literature, there are several drawbacks to the use of an aggregated score of firm social performance. First, using only aggregated CSR scores underweights certain KLD dimensions as not all the number of strengths and concerns are equal across all dimensions (Fernández-Kranz & Santaló, 2010). Second, positive and negative social
14 actions should be considered separately in empirical research, as independent variables (in this case, the level of market competition) can have contradictory effects on both positive and negative social actions. For example, firms in very competitive markets might choose to engage more in positive social behaviour as opposed to reducing negative social behaviour. Third, Creyer and Ross (1996) show that consumers value positive and negative social actions differently when considering their perceived value of a firm’s products. Unethical firm behaviour is more likely to be punished than ethical firm behaviour is rewarded, indicating that it might be more beneficial for firms to reduce their social concerns as opposed to increasing their level of positive social activity. To account for these potential drawbacks and in order to increase our understanding of how market competition affects both positive and negative social behaviour of firms, we also consider social strengths and social concerns separately in our research.
We add several other measures of CSR to our models to extend our analysis. In their research, Mattingly and Berman (2006) combine the different KLD social dimensions into four different factors; institutional weaknesses, institutional strengths, technical weaknesses and technical strengths (see appendix A1 on how these factors are computed). Following the literature of Siegel and Vitaliano (2007), we include dummies for public and non-public CSR activities. The dummy variable for public CSR equals ‘1’ if a firm has more strengths than concerns in community, environment, human rights and diversity, and ‘0’ if it has more concerns than strengths in these CSR areas. Consequently, we also include a dummy variable for non-public CSR which equals ‘1’ if a firm has more strengths than concerns in corporate governance, employee relations and product, or ‘0’ otherwise. The combined results for our specifications with the four different CSR factors as defined by Mattingly and Berman (2006) and the dummies for public and non-public CSR are included in the appendix (table A2).
Additionally, we also examine the effect of competition on industry levels of CSR by constructing separate dependent variables for strengths, concerns and aggregate CSR scores based on average industry levels. These variables were created by computing the average value of these three variables separately for every respective industry in our sample.
15
3.3 Measuring industry competition
As our main proxy for market competitiveness, we use the Herfindahl-Hirschman index (HHI) which is a commonly accepted measure of market concentration. The HHI is used by the US Department of Justice and the Federal Reserve in order to examine the (anti-)competitive effects of possible mergers or acquisition in a market. It is constructed by adding the squared market share of every firm in an industry for a given year: 𝐻𝐻𝐼 = %(𝑀𝑆))+ , ) - .
Where 𝑀𝑆) represents the market share of firm 𝑖, expressed as a percentage, and 𝑛 denotes the number of firms in the market. The higher the concentration in a market, the lower the level of competition is, and vice versa. The Herfindahl-Hirschman index can range from 0 to 10,000, a low value indicating a low level of concentration (and thus a high level of competition). On the contrary, as the value of the index gets closer to the maximum value of 10,000, market concentration increases and the level of competition decreases. For example, in a monopolistic market consisting of one firm with a 100% market share, the HHI would equal the maximum value of 100+ = 10,000. If, however, there are a total of 100 firms in a market with a market share of 1% each, the index would equal only 100, implying that this particular market is very competitive.
By squaring the market shares of each firm, the HHI allocates more weight to firms with a higher market share than firms with only small market shares. This is in line with the theoretical concept that, when there are only a small number of firms in a market with relatively high market shares, there is a high probability that the level of competition in this market will be low, holding all else constant (Rhoades, 1993). Also, by squaring the market share of each firm, information regarding both the number of firms in the market as well as the distribution of market shares within the industry is taken into account. When analysing the competitiveness of markets, the U.S. Department of Justice considers the following values as industry concentration boundaries; an index below 1,000 is competitive, between 1,000 and 1,800 is considered to be moderately concentrated and markets with HHI-values greater than 1,800 are regarded as highly concentrated markets, and therefore not very competitive.
16 In later sections, we employ these boundaries in our univariate tests to compare mean levels of our variables of interest.
In earlier studies, the Herfindahl-Hirschman index is usually computed using Compustat data on total sale volumes of firms in a market. As Ali et al. (2009) point out in their research, using Compustat data to compute this index gives poor values compared to the actual industry concentration levels. As Compustat only covers publicly listed companies and not private firms, there can be severe measurement errors when computing levels of industry concentration. If all large companies in an industry were publicly listed, only active in one single industry and only within the United States, only then would Compustat data provide accurate estimates of concentration levels. As this is generally not the case, Compustat-based industry measures can provide rather inaccurate levels of industry concentration.
The U.S. Census Bureau, which is part of the U.S. Department of Commerce, publishes their own HHI and concentration ratios based on information reported by firms themselves. As firms are required to report this information correctly to the U.S. Census Bureau (under the threat of possible legal penalties or fines), the Census HHI is considered to be the most accurate measure of industry concentration. However, as Census data only covers manufacturing industries and is only computed in five-year intervals, it is fairly limited in its capability to be used as appropriate measure in empirical research.
In order to account for the drawbacks of both Compustat-based concentration measures and the Census concentration measures, we use Hoberg and Phillips’ fitted HHI as main independent variable in our research. Hoberg and Phillips (2010) combine Census and Compustat data to compute a measure of industry concentration which covers both private and public firms in all industries. This ‘fitted HHI’ (denoted as fHHI in our tables) is computed by regressing the Census HHI against the Compustat HHI, the average number of employees in Compustat, the average number of employees in the Bureau of Labor Statistics database and interaction values for each of these variables with the Compustat HHI. As this measure covers all industries and years and has a much higher correlation with the Census HHI than Compustat-based concentration ratios, this measure is more suitable for use in empirical research. However, the original values of the fitted HHI are computed at the relatively high 3-digit Standard Industrial Classification (SIC) level for only 255 industries. For this reason, we use data recently
17 compiled by Keil (2017), who computed fitted HHI-values at the more accurate 6-digit NAICS level for a higher number of industries. This dataset also contains values for the Census HHI and the four-firm concentration ratio (CR4) initially provided by the U.S. Census Bureau, which we will use in our robustness tests.
3.4 Control variables
The degree in which firms behave socially responsible can be affected by many different factors. In order to better isolate the effect of our independent variable and decrease the possibility of omitted variable bias, we include various firm characteristics as control variables to our regression models. Following the recommendations of McWilliams and Siegel (2001), we include both advertising intensity (ADVi) and R&D intensity (RDi) in our models, defined as advertising expenditures over sales and R&D expenditures over sales. Consumers value positive social action of firms and as a result, more and more firms are incorporating CSR attributes into their marketing strategies. Also, CSR may be an instrument for firms to signal their reliability, quality or concern for certain environmental or social issues (Siegel & Vitaliano, 2007). Consequently, advertising might be used to provide consumers with information regarding these socially responsible activities, especially if these attributes are fairly difficult for consumers to identify by themselves. Additionally, firms are more likely to invest in advertising under high competitive pressure (Fernández-Kranz & Santaló, 2010). Investment in R&D may result in both CSR-related innovations and product-related innovations and, as a result, investment in R&D can potentially lead to differentiation through the use of CSR (McWilliams & Siegel, 2001). Consequently, firms might increase their R&D and advertising expenditures when subject to a more competitive environment and omitting these expenditures as controls could lead to an overestimation of the effect of market competition on corporate social performance. Unfortunately, reporting advertising and R&D expenditures to the US Securities and Exchange Commission is not mandatory and as such, the majority of firms have missing values for at least one of these variables. In order to not lose most of our observations, we assign a value of ‘0’ to any missing values for these two variables and create two additional dummy variables (dADVi and dRDi) that equals ‘1’ if a firm reports these kinds of expenditures or ‘0’ if this is not the case.
18 As the size of a firm in an industry is likely to be related to both the level of competition in a market and the social performance of that particular firm, we include firm size as a control in our model. There are several arguments for this claim; in more concentrated industries, the number of firms is generally lower and the size of firms larger. While smaller firms may be more financially constrained against investing in socially responsible activities, there might be more incentives for these firms to increase growth through such activities. Also, as firms grow larger, their public visibility and operational impact increases and it might be expected from these firms that they actively participate in socially responsible behaviour. More specifically, as firms continue to grow, the probability of qualifying for any of the strengths and concerns under scrutiny by KLD increases. As a result, we include firm size as a control in our model, expressed as the logarithm of the total value of assets of a firm.
To account for any effect of competition on the social performance of firms through firm operating profits, we include firm earnings before interest and taxes (EBIT) and return on assets (ROA) as controls in all of our specifications. When we consider the altruistic view on CSR, the reduced amount of resources available for CSR as a result of intense market competition should have a negative effect on the social performance of firms. Therefore, if the altruistic view of CSR holds, market competition should have non-positive or negative effect on firm social performance. Additionally, as GDP might have a positive effect on the social performance of firms, we include the GDP per industry as a control variable in our regressions.
3.5 Empirical strategy
In this section, we establish the method required to determine the effect of competition in a market on the social performance of firms in that respective market. In our analysis, we perform several different regressions in order to examine this effect. As noted above, we employ different measures of corporate social performance and regress our independent variable, the fitted HHI (fHHI), separately on these social performance measures. Since socially responsible behaviour can be used as an instrument by firms to intensify (or introduce) competition in an industry, competition might not just be a determinant of social actions but for some part also a result of it. Additionally, CSR strategies might be implemented by incumbent firms in order to create additional entry barriers for new firms. To account for this potential endogeneity
19 problem, we lag our independent variable one year and use this lagged value (lfHHI) in all of our specifications.
After merging multiple datasets, our sample consists of yearly data on many firms for over a total period of eight years. As we are dealing with panel data, we need to account for the possibility that observations are not independent and standard errors might be biased. In order to mitigate these potential problems, we cluster standard errors at the firm level in all our firm-year specifications and cluster standard errors at the industry level for all our industry-year specifications. Additionally, we introduce year dummies to every specification to eliminate any potential time biases in our estimations and, depending on the specification, include firm or industry fixed effects to take into account any unobserved firm or industry characteristics. For our firm-year level specifications, we run separate regressions using both firm fixed effects and industry dummies in order to examine if there are any notable disparities between these two models.As both our Strengths and Concerns dependent variables are left-censored, we include Tobit regressions for estimating the effect of our independent variable on these variables. When using Tobit specifications, firm fixed effects are substituted by industry dummies in order to correct for unobserved industry characteristics that could simultaneously be correlated to both CSR levels and market competition.
For our industry-level specifications, we first exclude firms for which less than four years of data is available in order to have enough variation in both industry competition levels and CSR values. By excluding these firms, we lose a total of 2,720 firm-year observations, which amounts to 18.99% of the number of original sample observations. After excluding these firms, we end up with a sample of 422 industries and an average of 6.9 years of data per industry. A total of 182 industries (43.13%) cover the maximum 8 years of data and another 128 industries (30.33%) cover 7 years of data.
In order to examine the effect of competition on industry-levels of CSR, we compute the values of our main dependent variables (aCSR, Strengths and Concerns) at the industry level by calculating the average of CSR ratings across all firms in an industry for a given year. Subsequently, we regress these new variables (indCSR, indStrengths and indConcerns) on levels of competition and include both year and industry dummies. When adding industry dummies to our regressions, we include the
20 six-digit NAICS codes as dummies to account for any unobserved industry characteristics that might affect our estimations. As the omitted and unobservable factors are more likely to be relatively stable within industries over time but not between industries, it could be informative to include industry dummies and use the variability in concentration levels over time rather than using a cross-sectional approach (Nickell, 1996). An additional argument for running regressions with industry dummies is that there might be important differences in CSR levels across industries. As Siegel and Vitaliano (2007) point out, firms selling experience goods have on average higher CSR scores than firms selling search goods.
To examine whether firms engage more in CSR-driven differentiation strategies in competitive markets, we investigate how the level of competition affects the variability in CSR scores for our three main CSR variables. We construct measures of within industry variability for our three main dependent variables by computing the standard deviation of CSR scores within an industry per year. We do this for the aggregate CSR score, the number of strengths and the number of concerns for each industry and each year.
As a final step in our analysis, we run regressions on the strengths and concerns separately for every individual social dimension covered by the KLD database in order to examine how the level of competition affects these different CSR dimensions. By isolating the strengths and concerns for every CSR dimension, we can detect whether some of these dimensions considerably differ in their relation to competitive intensity.
In order to evaluate the robustness of the results of our univariate and multivariate tests and the inferences made from these results, we perform several robustness checks. As described above, the Hoberg and Phillips’ fitted HHI is not without measurement error, and as such, we include additional regressions using the HHI and four-firm concentration ratios computed by the U.S. Census Bureau in order to test the robustness of earlier results. As these concentration ratios are only computed by the Census Bureau every five years, we were only able to use this measure for the years 2002 and 2007.
21
4. Results
The following section examines the effect of market competition on the corporate social performance of firms in an empirical setting. In section 4.1, we start off by providing summary statistics for the variables used in our research. In section 4.2, some preliminary results are presented in which we compare differences in variable means by the level of competition in an industry. In section 4.3, we perform various multivariate regressions in order to examine the effect of competition on CSR, CSR variability and disaggregate CSR dimensions. This is followed by some robustness checks in section 4.4 in which we to attempt to validate earlier results and inferences made from those results.
4.1 Summary statistics
We start off by showing summary statistics and correlation coefficients of variables used in our research. Table 1 provides descriptive statistics for all variables used in our analysis. The table is divided into our dependent variables of interest, our independent variables and control variables. Missing values for certain (control) variables are the result of missing values in the Compustat database. The mean of our aggregate measure of CSR, aCSR, is -0.471, indicating that in our sample, firms have on average more social concerns than social strengths. The maximum number of concerns and strengths for individual firms in our sample is 18 and 22, respectively. When looking at our dummies for public and non-public CSR, we find that the mean of public CSR is significantly higher than the mean of non-public CSR, suggesting that firms prefer to engage in public CSR. This is consistent with the idea that social actions can be viewed as a way to improve a firms’ public image. The average level of competition for industries in our sample is 0.060 (equivalent to a Herfindahl-Hirschman index of 600), indicating that the majority of industries in our sample are competitive when considering the boundaries set out by the U.S. department of Justice.
22
Table 1: Summary statistics
Variable Mean Std. Dev. Min Max Obs
Panel A: Social performance proxies aCSR -0.471 2.008 -11 15 14,326 Concerns 1.686 1.781 0 18 14,326 Strengths 1.214 1.846 0 22 14,326 dPublic 0.298 0.457 0 1 14,326 dnonPublic 0.170 0.375 0 1 14,326 INSTw 0.245 0.672 0 6.98 14,326 INSTs 0.509 0.942 0 8.40 14,326 TECHw 0.845 0.804 0 7.18 14,326 TECHs 0.278 0.457 0 4.66 14,326 Panel B: Market competition proxies fHHI 0.060 0.025 0.04 0.50 14,254 lfHHI 0.060 0.025 0.04 0.50 10,764 HHI 0.080 0.060 0 0.30 966 CR4 37.343 18.00 1.80 99.20 2,528 Panel C: Control variables ADVi 0.027 0.075 0 3.324 5,951 dADVi 0.415 0.493 0 1 14,326 RDi 0.838 11.292 -2.263 609.762 7,566 dRDi 0.528 0.499 0 1 14,326 Assets 7.216 1.712 1.50 14.61 14,323 EBIT 528.314 2691.947 45026.00 66290.00 14,243 ROA 0.060 0.163 -6.82 1.95 14,243 GDP 13002.74 33983.64 0.00 279121 14,326 Table 1 provides summary statistics at firm-year level for the variables in our sample. Panel A presents statistics for our social performance proxies, panel B for our market competition proxies and panel C for the control variables in our sample. The variables dADVi and dRDi are the corresponding dummy variables for ADVi and RDi and indicate whether or not firms report advertising and R&D expenditures.
Table 2 presents the correlation coefficients between the main dependent variables, the competition measure fHHI and control variables. Similar to the findings of Kotchen and Moon (2012), we find a significant positive correlation (0.3872) between the amount of social strengths and the amount of social concerns. This indicates that firms might try to offset negative social behaviour by engaging in more socially responsible behaviour. Additionally, it is also consistent with the findings of Mattingly and Berman (2006) who argued that positive social action is not necessarily associated with less negative social action. We also see that our measure of product market
23 competition, the fitted HHI, is more heavily correlated with social concerns than with social strengths, suggesting that competition has greater effect on social concerns than on the social strengths of firms.
Table 2: Correlation coefficients
aCSR Concerns Strengths ADVi RDi Assets EBIT ROA GDP fHHI aCSR 1 Concerns -0.531*** 1 Strengths 0.576*** 0.387*** 1 ADVi 0.038*** -0.002 0.039*** 1 RDi -0.0004 -0.021** -0.021** 0.006 1 Assets 0.035*** 0.425*** 0.448*** -0.041*** -0.062*** 1 EBIT 0.030*** 0.409*** 0.427*** 0.005 -0.012 0.425*** 1 ROA 0.008 0.067*** 0.073*** -0.049*** -0.164*** 0.142*** 0.073*** 1 GDP 0.027*** 0.065*** 0.092*** 0.045*** 0.056*** -0.142*** 0.002 -0.123*** 1 fHHI -0.095*** 0.213*** 0.102*** 0.018*** -0.013 0.101*** 0.070*** 0.072*** 0.037*** 1 Table 2 displays the correlation coefficients between our main dependent variables, control variables and our proxy for market concentration (fHHI). * Significant at 10%, ** Significant at 5%, *** Significant at 1% 4.2 Preliminary results First, we employ a univariate analysis by examining mean levels of our variables of interest depending on different levels of market competition. To identify the differences in these averages, we first divide the sample into three subsamples based on the boundaries set out by the U.S. Department of Justice. A HHI index below 1,000 is considered to be a competitive market, between 1,000 and 1,800 is considered moderately competitive and markets with an HHI-index greater than 1,800 are considered to be not very competitive. Table 3 shows these average levels of our variables of interest by HHI level; column (1) presents the averages for competitive markets, column (2) for moderately competitive markets and the column (3) for the least competitive markets. Column (4) displays the difference between the mean levels of the most competitive and least competitive markets and the significance levels of these differences.
24 Table 3: Variable means by HHI-levels Competitive Concentrated Variable <0.10 0.10-0.18 >0.18 Difference (1)-(3) (1) (2) (3) (4) Social performance proxies aCSR -0.449 -0.689 -2.026 1.577*** Strengths 1.176 1.902 1.491 -0.315* Concerns 1.625 2.591 3.518 -1.892*** Public 0.060 0.400 -0.113 0.172** nonPublic -0.509 -1.088 -1.226 0.717*** dPublic 0.296 0.355 0.228 0.068 dnonPublic 0.173 0.109 0.079 0.094*** INSTw 0.232 0.416 0.619 -0.386*** INSTs 0.491 0.872 0.562 -0.071 TECHw 0.817 1.267 1.428 -0.611*** TECHs 0.273 0.344 0.386 -0.113 Market competition proxies fHHI 0.055 0.129 0.230 -0.175*** HHI 0.077 0.169 0.185 -0.108*** CR4 36.056 59.509 52.216 -16.160*** Control variables ADVi 0.011 0.018 0.008 0.002 RDi 0.468 0.015 0.007 0.461 Assets 7.183 7.795 7.670 -0.487*** EBIT 500.571 1005.456 931.105 -430.534*** ROA 0.058 0.099 0.096 -0.038* GDP 12736.67 20674.48 10838.17 1898.500 Employees 8.209 63.043 79.128 -70.919*** Table 3 presents the mean levels of our variables of interest, grouped by level of industry competition. Column (4) displays the difference between competitive industries and least competitive industries, as defined by the U.S. Department of Justice. The estimated mean differences are the differences in variable means when moving from least competitive industries to most competitive industries. * Significant at 10%, ** Significant at 5%, *** Significant at 1%
When moving from concentrated to competitive markets, we observe a statistically significant increase in aggregate CSR level. If we look at strengths and concerns separately, we find that an increase in the level of competition in a market slightly decreases overall strengths but is also associated with a greater decrease in the average amount of concerns. This seems to suggest that the improved social performance of firms in more competitive industries is not necessarily due to an increase in positive social behaviour, but more likely the result of a decrease in negative social behaviour. We observe similar patterns for the average level of institutional and
25 technical social factors, where increased competition seems to be associated with a significant decrease in both institutional and technical weaknesses, but also a decrease in institutional and technical strengths, though not statistically significant. In competitive industries, a larger proportion of firms seems to have more strengths than concerns in both public and non-public CSR areas. Surprisingly, the difference appears to be smaller for public CSR and not statistically significant. However, if we look at the original values for public and non-public CSR, we see that levels of non-public CSR are still systematically lower (and negative) for non-public CSR compared to public CSR. With respect to our control variables, we observe that firms in more competitive markets have on average higher levels of R&D and advertising intensity, are smaller in size (using Assets as proxy), have lower levels of profit (using EBIT as proxy) and lower return on assets. As the number of observations across the three groups of competition in table 3 were not equal and some groups might have been more subject to outliers, we also divided the sample into equal terciles and compared the resulting means as a robustness test. Quantitatively, however, the results remained the same as in the above table.
4.3 Multivariate analysis
4.3.1 Firm and Industry-level OLS
In the following section, we explore the relationship between competition and CSR using multivariate OLS regressions. We analyse the effect of our competition variable on both firm and industry-levels of CSR and examine this effect on social strengths and concerns separately. Table 4 and 5 show the results of these regressions with both firm and industry-levels of CSR as dependent variables. The first three columns of table 4 show the estimated coefficients of our independent variable and control variables in a model with firm-year observations. In these specifications, we include firm fixed effects and year dummies to account for unobserved firm heterogeneity and control for serial correlation between observations. Column (4) displays the estimated coefficients when firm fixed effects are substituted by industry dummies, where an industry is defined as the six-digit NAICS code of each respective industry. The last two columns present the regression results for our Tobit-specifications assuming both our Strengths and Concern variables are left-censored, in which we also include year and industry dummies.
26 When looking at the effect of competition on the aggregate level of CSR, we find that an increase in the HHI-value of an industry is associated with a decrease in overall CSR levels. This is the case for both firm level OLS with firm fixed effects as well as for firm level OLS regressions including industry dummies. Consistent with the results of our univariate tests, we observe a significant increase in concerns as markets become less competitive (as the value of our fitted HHI increases), while this result is less pronounced for our measure of social strengths. This confirms our initial suspicions that the improved social performance of firms in more competitive markets is the result of less social concerns, and not necessarily a result of more social strengths. In both models of table 4, we observe statistically significant estimates for our lagged competition measure on both our aggregate CSR measure and our measure for the number of social concerns. Table 4: Firm-level OLS and Tobit models OLS Firm level with firm FE OLS/Tobit Firm level with Industry dummies aCSR Concerns Strengths aCSR Concerns Strengths
(1) (2) (3) (4) (5) (6) lfHHI -10.257 *** 10.005 ** -0.252 -9.293 *** 12.130 *** 2.850 (3.902) (4.320) (2.999) (1.848) (2.045) (2.164) ADVi -0.683 0.254 -0.430 -0.233 0.946 ** 2.099 (0.649) (0.359) (0.478) (0.356) (0.470) (1.515) RDi -0.001 0.000 -0.001 * -0.001 -0.006 *** -0.005 (0.002) (0.001) (0.001) (0.002) (0.002) (0.004) Assets -0.186 ** 0.117 * -0.069 0.003 0.412 *** 0.608 *** (0.073) (0.062) (0.061) (0.028) (0.033) (0.045) EBIT 0.000 0.000 ** 0.000 *** 0.000 0.000 *** 0.000 *** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) ROA 0.377 ** -0.395 *** -0.018 0.269 ** 0.345 ** 0.595 * (0.147) (0.131) (0.078) (0.118) (0.189) (0.328) GDP 0.000 * -0.000 0.000 * 0.000 *** 0.000 *** 0.000 *** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Observations 10,715 10,715 10,715 10,715 10,715 10,715 R2 0.014 0.108 0.069 0.010 0.089 0.074 All regressions include year dummies and dummies for missing observations on R&D intensity and advertising intensity. Robust standard errors (in parentheses) are used assuming that observations are independent across firms but not within firms and across time. The measure aCSR is the aggregate CSR score per firm computed by subtracting a firms' strengths by its concerns. The independent regressor lfHHI is the lagged value of our fitted HHI variable. Columns (1) – (3) include firm fixed effects and columns (4) – (6) include industry dummies, added at the six-digit NAICS code level. Column (1) – (4) are OLS regressions, column (5) and (6) are Tobit regressions. * Significant at 10%, ** Significant at 5%, *** Significant at 1%
27 When considering the estimates of our control variables included in the models of table 4, we find less consistent results when comparing the different specifications. The estimates on our proxy for firm size, the logarithm of the total value of assets for each respective firm, are somewhat mixed but seem to point to a considerable increase in social concerns as firms grow. A possible explanation for this could be that larger firms have a higher visibility and are therefore more susceptible to receiving a higher number of strengths or concerns in the KLD dataset, when compared to smaller firms with less substantial operational impact. While we find both a negative and positive coefficient for ROA on concerns in the specification with firm fixed effects and the specification with industry dummies respectively, return on assets is unambiguously associated with a significant increase in average CSR scores.
Table 5 presents regression results when considering average industry levels for our three dependent measures of social performance. In these specifications, we include year dummies in order to account for serial correlation between observations and introduce industry fixed effects as there might be unobserved industry heterogeneity (Kotchen and Moon, 2012). The relatively low number of observations in this specification compared to the number of observations in previous firm-year specifications is the result of now only having 422 unique industries. Furthermore, industries with only one year of observations are excluded as there is no lagged value of our independent variable fHHI for these industries.
28 Table 5: Industry-level OLS models OLS Industry level with year dummies Dependent indCSR indConcerns indStrengths (1) (2) (3) lfHHI -6.131 *** 7.641 * 1.606 (1.916) (4.270) (2.679) ADVi 2.855 *** -2.030 ** 0.216 (1.043) (0.961) (0.696) RDi -0.020 ** 0.017 ** -0.003 * (0.009) (0.008) (0.001) Assets 0.005 0.003 0.021 (0.027) (0.018) (0.019) EBIT 0.000 ** 0.000 *** 0.000 ** (0.000) (0.000) (0.000) ROA -0.027 -0.253 * -0.251 ** (0.178) (0.142) (0.124) GDP 0.000 0.000 * 0.000 * (0.000) (0.000) (0.000) Observations 2,319 2,319 2,319 R2 0.034 0.192 0.106 All regressions include year dummies and dummies for missing observations on R&D intensity and advertising intensity. Robust standard errors (in parentheses) are used assuming that observations are independent across industries but not within industries and across time. Column (1) includes random effects, column (2) and (3) include fixed effects. The dependent variables are CSR measures at the industry level, computed by calculating the average of CSR ratings across all firms in an industry for a given year. The independent regressor lfHHI is the lagged value of our fitted HHI variable. * Significant at 10%, ** Significant at 5%, *** Significant at 1% The results presented in table 5 are consistent with our previous firm-year level estimations. Again, we observe a significant increase in (industry) concerns causing a decline in CSR scores when the level of competition decreases (and level of market concentration increases). A positive but not significant effect is observed for the level of market competition on industry strengths. For our control variable ‘Assets’ we observe that an increase in firm size is associated with both an increase in the number of concerns and strengths, similar to what we witnessed in the firm-year level Tobit specifications and in line with the explanation that larger firms are more receptive to receiving higher numbers of concerns and strengths in the KLD dataset. Contrary to our firm-level results, we now find a statistically significant and unambiguous estimates of advertising intensity on social performance. Advertising intensive industries seem to be associated with significantly less industry concerns and an improved overall social
29 performance, as well as a higher level of social strengths. As McWilliams and Siegel (2001) argue, advertising can play an important role in signalling socially responsible firm or product characteristics to consumers, especially if some of these characteristics might not be immediately visible. Overall, the reported results in table 4 and table 5 seem to indicate a consistent pattern; more market competition is associated with an improved social performance of firms due to a considerable decline in social concerns. This relationship holds whether we consider either firm level CSR or industry level CSR. The effect of increased competition on the reduction in social concerns of firms seems to suggest that firms do not use CSR purely as a method of differentiation from competitors. If firms engage in CSR to differentiate themselves from competitors, we would expect a more notable increase in social strengths as a result of intensified competition in a market. While reducing negative social action can still be used as a method of differentiation, it would seem more straightforward to advertise positive social action rather than communicating non-negative social action. However, as this is not what we observe in our analysis, we assume that, while CSR is still being used to improve a firms’ (social) image, it is not necessarily used to create some competitive advantage by way of differentiation. The presented results are in line with the findings of Creyer and Ross (1996), who find that consumers are more likely to punish unethical firm behaviour than reward positive social action. Consequently, reducing the level social concerns can be interpreted as a strategic CSR.
4.3.2 CSR variability
In order to increase our understanding of how competition affects the propensity of firms to differentiate themselves from competitors using CSR, we examine how the level of market competition affects the variability in CSR scores of firms within an industry. By definition, not all firms can differentiate by using the same strategy. As a result, if firms in more competitive industries engage in socially responsible behaviour in order to differentiate themselves from competitors, we should expect a higher within industry variation of CSR levels in more competitive environments. To test whether this holds true, we regress the within-industry standard deviations of our three main dependent CSR variables per industry per year on our lagged value of the measure of market concentration.
