1
The Impact of Community Development Initiatives on Firm Productivity: The Role of Slack Resources and
Capital Investment
Master’s Thesis SCM
MSc Supply Chain Management Faculty of Economics and Business
University of Groningen The Netherlands
First Supervisor: dr. Xun (Bruce) Tong Second Supervisor: dr. Kirstin Scholten
Student: Yingying Ku (S2507390) Word Count: 10058
Acknowledgements: I would like to express my special thanks to dr. Xun Tong for his constructive guidance throughout the planning and development of this master thesis project. I would also like to thank dr. Kirstin Scholten for her valuable time for assessing and ensuring the quality of my research.
2
ABSTRACT
Integrating the welfare of the communities into a firm’s core business models is a voluntary corporate social-responsibility referred to as community involvement and development (CID).
It has become a trend nowadays favored by the stakeholders. However, prior research has limited exploration on the effect of such philanthropy on firms’ labor productivity. This study investigates their relationship using the long-horizontal event study method, and the empirical evidence disclosed a negative and significant relationship of the two. Moreover, regression analyses were used to test the effects of slack resources and capital investment on the abnormal performance of labor productivity following a CID event. The results defend the theory that slack resources attenuate the negative effect of CID whilst capital investment strengthens it.
The practical insights are that improving productivity and providing welfare to the communities are not mutually reinforcing, and that firms should carefully control all investments and use slack resources to measure their resilience on supporting CID implementations.
Key words: Longitudinal Event Study; Firm-level Labor Productivity; External Corporate Social Responsibility; Community Involvement and Development; Abnormal Performance;
Manufacturing Sector
3
Contents
1. INTRODUCTION ... 5
2. LITERATURE REVIEW ... 7
2.1 CID under CSR ... 7
2.2 CID (external CSR) and productivity ... 8
2.3 Factors that influence the impact of CID on productivity ... 9
3. HYPOTHESES DEVELOPMENT ... 11
3.1 Resources devoted to CID and labor productivity ... 11
3.2 The level of slack resources affects CID’s impact on productivity ... 12
3.3 The amount spent on capital investment affects CID’s impact on productivity ... 13
4. METHODOLOGY ... 15
4.1 Data collection and sample selection ... 15
4.2 Selection of the control firms ... 17
4.3 Variable definitions and measurement ... 18
4.3.1 Dependent variable ... 18
4.3.2 Variables that affect the impact of CID event on productivity ... 18
4.3.3 Control Variables ... 19
4.4 Research design ... 19
4.4.1 Event study approach for longitudinal analysis ... 19
4.4.2 Cross-sectional regression analysis of contextual factors ... 20
5. RESULTS ... 21
5.1 Event Study Results (H1) ... 21
5.2 Regression Analysis (H2 & H3)... 22
4
6. DISCUSSION AND CONCLUSION ... 26
6.1 Major Findings ... 26
6.2 Theoretical Contributions ... 27
6.3 Managerial Implications ... 28
6.4 Limitations and future research ... 29
REFERENCES ... 31
APPENDICES ... 38
Appendix A: Regression Analysis with Unstandardized co-efficient of the 96 samples .... 38
5
1. INTRODUCTION
Community involvement and development (CID) is one of the seven core subjects of corporate social responsibility (CSR) in the guidance published by International Organization for Standardization, known as ISO 26000. The specific practices of CID include job creation, skill development and provision of health, welfare, education and other services that benefit the communities (Banks, Scheyvens, McLennan, & Bebbington, 2016; ISO26000, 2011). Although the idea of integrating economic performance with socially desirable goals has become a trend, the discussions about the influence of CSR on the economic performance/ competitiveness of firms have never reached a consensus outcome (Lu, Chau, Wang, & Pan, 2014). While some studies provide support to the positive relationship between the two, there is one dominant critique of CSR originates from the principal-agent paradigm. It suggests that the purpose of a firm is mainly for the profit of its principals, which are the shareholders (Wang, Dou, & Jia, 2016). From the agency theory perspective, firm managers are the agents on behalf of the shareholders, and their decisions of redistributing the shareholders’ wealth involuntarily to those in society who have no rightful claim, is against the value-maximizing interest of their principals (McWilliams, Siegel, & Wright, 2006; Friedman, 1970). Those resources should have instead been used for internal value-adding programs or returned to the shareholders (McGuire, Sundgren, & Schneeweis, 1988; Barnett, 2007).
The empirical studies and theoretical discussions on the relationship between CSR and corporate financial performance (CFP) are substantial in amount (Lu et al., 2014). However, studies that explore the impact of CSR on other firm competitiveness dimensions are relatively limited (Sun
& Stuebs, 2013). Those dimensions include productivity, output quality, innovation and reputation (Vilanova, Lozano, & Arenas, 2009). To discover the impact of CID in the operations management domain, this study focuses on the relationship between CID and productivity. Achieving higher productivity is a sign of good operational efficiency in operations management as it implies higher production output with lower use of resources (Banker & Mashruwala, 2007; Stuebs & Sun, 2010).
CID initiatives are external CSR as the efforts direct toward external stakeholders, i.e., community.
In contrast, internal CSR are practices related to worker benefits and the responsibilities in production processes (Hameed, Riaz, Arain, & Farooq, 2016). In terms of CSR’s effect on productivity, extant studies provide well explanation on how internal CSR leads to employee motivation and job satisfaction that help to improve productivity (Hasan, Kobeissi, Liu, & Wang, 2018; Malik, 2015; Sanchez & Benito-Hernândez, 2015; Becchetti, Pinnacchio, & Di Giacomo, 2005). However, prior research has not paid much attention on the relationship between external CSR and productivity. Although Sun & Stuebs (2013) argue external CSR lead to improved productivity, they use the CSR rating data provided by Kinder, Lydenberg and Domini (KLD) as a measurement to test its relationship with future firm productivity. Since KLD evaluates both
6 internal and external CSR contributions of the firms, their study leaves a question of whether external CSR alone would benefit the firm’s productivity. Sanchez & Benito-Hernândez (2015) found no positive correlation between labor productivity and firm’s relationship with the community and other external stakeholders. However, instead of using control firms to compare the productivity performance, they used the samples’ past performance, i.e., before CSR implementation as the performance benchmark. This approach neglects the differences in economy and the industry between the comparing years (Hendricks & Singhal, 2008). In addition, it does not help clarify a continual debate whether firms that allocate limited resources to CID (i.e., external CSR), which they found non-value-adding to productivity, would be at an economic disadvantage compared to others that do not (McGuire et al., 1988).
The resource trade-offs between internal programs and external CSR leads to productivity differentials among firms (Boring, 2019; Shah & Ward, 2003). However, some researchers consider firms with higher slack resources have more freedom to devote to CSR without losing the ability to initiate and execute strategic changes and innovative programs (Wang et al., 2016;
Marlin, 2014). Since those changes for process improvements usually require updated fixed assets, e.g., equipment and machinery (Mo, 2009; Hendricks & Singhal, 1997), those that incurred high capital investment while implementing CID had not neglected their productivity performance at their manufacturing facility. Yet, the relationship has not been empirically examined yet.
Consequently, this research focuses on the following two research questions:
1. What impact does the implementation of CID have on a firm’s labor productivity?
2. Whether and to what extent do the level of a firm’s available slack resources and capital investment influence the relationship between CID and labor productivity?
This study aims to provide three main contributions. Firstly, to relate CID participating firms to lower labor productivity. Previous empirical studies mostly do not distinguish the internal and external dimensions of CSR, and the impact of external CSR alone on productivity is often neglected (Sun & Stuebs, 2013; Weber, 2008; Becchetti et al., 2005). By empirically examining the productivity of the sampled firms and control firms, this contribution will be confirmatory.
Secondly, to defend the idea that the level of slack resources negatively influences the proposed negative relationship between CID and productivity. Several researchers proposed firms with higher slack tend to invest more on CSR (Wang et al., 2016), but so far, the discussion and linkage on the resources allocation trade-off between external CSR and productivity has not been established. Lastly, to uncover the linkage between capital expenditure and labor productivity when additional resources are put into CID. The practical implication of this study is that managers should treat decisions regarding CID implementation carefully as they treat all investment decisions, and they should not overlook the consequences of their decisions made under moral imperatives on the economic performance of the firm.
7 This study is structured as follows: firstly, extant literature is reviewed to provide better understanding of the variables and the mechanism that supports their links. In section three, the hypotheses of the proposed relationships between the variables will be presented. Subsequently, the research design and data collection method will be discussed. The empirical results will be presented and discussed afterwards. Finally, in the concluding section, the limitations, managerial and theoretical implications of this study will be provided, and the suggestions for further research will be given.
2. LITERATURE REVIEW
2.1 CID under CSR
CSR aims to contribute to sustainable economic development through transparent and ethical behavior that includes the right of all stakeholders (ISO 26000, 2016; Freeman & Velamuri, 2006).
CSR is recognized as “the voluntary integration of social and environmental concerns into business operations and into their interaction with stakeholders” (The European Commission, 2002). It guides organizations to proactively adapting their behaviors and practices to meet the social goals desired by the public (Arena, Azzone, & Mapelli, 2018; Campbell, 2007; Barnett, 2007). CSR has gained interest in operations domain since getting the support of society and the various stakeholder are recognized as the prerequisite for the companies to operate with greater freedom and guarantees of survival (Martínez, Fernández, & Fernández, 2016).
Corporate involvement and development (CID) focuses on integrating the welfare of the communities served by the companies into their core business models (ISO26000, 2011). It is normally achieved by implementing discretionary initiatives and corporate resources (Kotler &
Lee, 2008). It is worth clarifying in this research the difference between the CID under CSR and the community-based capacity building projects. The latter cope with community oppositions against undesired local business activities, such as oil drilling (Orubu, Odusola, & Ehwarieme, 2004), and facilities, such as wind turbines (Pasqualetti, 2011). It can be achieved by recognizing the local needs and providing decision-making (Barker, 2005; Wang et al., 2016). These projects are used as a rational strategy to minimize conflicts arising from negative events (Becchetti et al., 2008), which do not meet the definition of CSR as “initiatives that advance the promotion of some social good, beyond the interests of the firm, shareholders and legal requirements” (McWilliams
& Siegel, 2001, p.117). Therefore, these projects are not voluntary CSR and will not be included in the scope of this research.
8 2.2 CID (external CSR) and productivity
Productivity defines the level of aggregate output per unit of input (Sickles & Zelenyuk, 2019). It is a common economic-performance-indicator that measures how efficiently resources are utilized by firms to achieve desired outcome (Melitz, 2000; Cainelli, Evangelista, & Savona, 2005). Shop- floor operations is one of the most important management components that could lead to differences in productivity between organizations in the same industry (Bloom & Van Reenen, 2007). For industrial firms, human resource is one of the key inputs in production processes (Hannula, 2002). Therefore, labor productivity is a common productivity measure used to improve internal efficiency (Stuebs & Sun, 2010). It measures the ratio of production output over total employment or person-hours (Samuelson & Nordhaus, 1989). Improving labor productivity is a focus in operations management as it can affect a firm’s overall operational performance and gain competitiveness (Malik, 2015; Stuebs & Sun, 2010). For example, it is one of the most frequently cited benefits associated with lean practices (Shah & Ward, 2003).
The positive link between CSR and productivity are mostly associated with practices relating to the workforce (i.e. internal CSR) (Edmans, 2013; Banker & Mashruwala, 2007). Improving internal CSR was found to increase worker-operating performance (McGuire & Schneeweis, 1988;
Banker & Mashruwala, 2007). Better working conditions, increased worker involvement and ethical labor practices lead to better job satisfaction and employee motivation, which were found associated with productivity improvements (Tsoutsoura, 2004; Sun & Yu, 2015). A rather small group of researchers link external CSR (i.e. CID initiatives) to productivity benefits. They proposed that external CSR assists firms in building good reputations, and reputable firm attracts and retains better talent and motivates personnel, which could eventually turn into productivity benefits (Roberts & Dowling, 2002; Barnett, 2007). In Stuebs & Sun’s (2010) empirical study, the relationship between reputation and labor productivity were found positively associated. However, they measure reputation by using Fortune's list of America's most admired companies. The key attributes of this score are not limited to external CSR; therefore, the positive relationship of the two does not represent the direct link between external CSR and labor productivity.
In Sun & Stuebs (2013), they further investigated the relationship between CSR and firm productivity by examining 170 chemical companies’ KLD in year t and their productivity in year (t+1), (t+2) and (t+3). Their regression analysis disclosed that CSR is positively related to firms’
productivity in year (t+1) and (t+2) at a significant level. Their research design clearly serves the purpose of their study, and the result provides empirical evidence on the suggestion that participating in CSR initiatives can lead to better future productivity. However, their research scope is limited to one industry (chemical industry), which lacks generalization in the overall secondary sector. Secondly, they did not test separately the effects of internal and external CSR on productivity. The KLD index provides mixed estimation of the firms’ contributions to both
9 internal and external CSR. Their study provides no direct empirical evidence on the impact of external CSR on productivity.
Finally, in Sánchez & Benito-Hernández (2015), the effects of internal and external CSR on productivity were investigated separately, using 929 small-business samples. They provide empirical evidence on the increased productivity following only internal CSR actions. They discovered no correlation between labor productivity and relationships with the community and other external stakeholders. Consequently, they emphasized the increased labor productivity are predominantly contributed by the internal CSR. A limitation in their study is that instead of using control firms that have not implemented external CSR to compare and calculate the abnormal performance of productivity, the sample firms’ past performance was used as the performance benchmark before CSR implementation. This method does not control the differences in the economy and industry between the comparing years (Hendricks & Singhal, 2008). Moreover, some researchers contend that firms that incur costs from CSR actions put themselves at an economic disadvantage comparing to other more economic-oriented and less socially responsible firms (McGuire et al., 1988; Malik, 2015). The view remains the subject of continual debate as some other researchers advocate the opposite (Wang et al., 2016; Stuebs & Sun, 2010). To clarify this doubt, my study follows Lo, Pagell, Fan, Wiengarten, & Yeung (2014), uses firms that have not implemented CID and are comparable to the sample firms as the control firms. Their productivity performances are applied as the performance benchmark.
2.3 Factors that influence the impact of CID on productivity Slack Resources
Slack are defined as the resources “available to an organization that are in excess of the minimum necessary to produce a given level of organizational output” (Marlin, 2014, p. 23). Wang et al., (2016) in their meta-analysis, identify a strand of literature proposing that the number of firms’
slack resources determine their CSR level. Those studies use the slack-resources theory to provide support to their view that firms with relatively abundant resources may have more degrees of freedom in terms of investments to CSR (Godfrey, Merrill, & Hansen, 2009). Consider the fact that CSR practices may take significant cost, slack-resource theory suggests that having slack resources may increase firms’ willingness to absorb the costs in the future (McGuire et al., 1988).
This view explains one of the drivers of CSR behaviors. Other researchers, however, highlight that slack resources provide firms with the opportunity to foster innovation and change (Marlin, 2014;
Bourgeois, 1981). Productivity enhancement are often achieved with the presence of innovation and change (Cainelli et al., 2005; Al-Shuaibi, 2016). In the operational level, lean manufacturing practices and other process improvement programs are part of the innovation and change (Shah &
Ward, 2003). While innovation and change drive productivity, they are strongly connected to the
10 provision of financial resources (Dabla-Norris, Kersting, & Verdier, 2012). Thangavelu & Findlay (2012) investigated the impact of financial constraints on productivity growth of Vietnamese firms, and they reported a positive effect of liquidity on firm productivity. The resources needed to attain a company’s goal on changes and improvements are slack in nature (Fauzi & Idris, 2009).
From the two research streams that provide different viewpoints on the usage/benefits of slack resources, a research gap is identified. Extant studies have limited exploration on the trade-off between CSR and productivity in terms of resource allocation. Since firms have limited resources (Barnett, 2007), managers when making strategic decisions may face a choice of expending resources on either CID or productivity enhancements. Yet, from the slack resource theory perspective, firms with higher slack are expected to be more resilient to the negative effect of decreased internal finance on their productivity-enhancing activities (Chen & Guariglia, 2011).
Their development, innovation and overall investment decisions in terms of productivity may therefore, be less likely affected by their contribution to CID. However, the researchers have not yet empirically tested this proposition.
Capital Investment
Process improvement programs and innovative capital equipment that help to increase productivity are linked to investments in capital (Mo, 2009). While slack resources provide firms with the opportunity to invest on internal value-adding programs, capital investment includes the spending on fixed assets that help to improve production processes (Watson, 1998). A few examples of those fixed assets are lean manufacturing cells (Sullivan, McDonald, & Van Aken, 2002), new machinery for business process re-engineering program (McAdam & McCarron, 2002), IT-based manufacturing systems for agility strategy (Mo, 2009) and other innovative IT hardware (Cainelli et al., 2005). Furthermore, Hendricks & Singhal (1997) discover the capital expenditure to assets ratio is higher for firms that implemented TQM programs than those that did not. They, therefore, relate investments in process control systems and better/new equipment with implementing TQM programs. The examples mentioned above deliver a message that increased capital investment could serve the purpose of improving productivity. Although CID competes for the limited corporate resources required for capital investment (Lu et al, 2014), there may be a right balance on the resource trade-off between the two that would not significantly affect the CID implementing firms’ productivity competitiveness (Barnett, 2007). However, this proposition has not been empirically tested before. To better understand its impact on the proposed change in productivity competitiveness attributable to CID implementation, this research investigates the level of capital investment of the sample firms.
11
3. HYPOTHESES DEVELOPMENT
3.1 Resources devoted to CID and labor productivity
In the face of the often-fierce competition among do-for-profit business firms, increasing productivity to gain competitiveness is often a goal in operations management (Mefford, 2009).
To increase productivity, firms should push the frontier of knowledge via innovation or converge toward it by adopting changes (Dabla-Norris et al., 2012). Cainelli et al. (2005) found that innovation activities lead to higher productivity. For manufacturing firms, process-improvement programs, e.g., lean and TQM are considered as process innovation (McWilliams & Siegel, 2001).
Moreover, the meta-analysis of Appelbaum, St-Pierre, & Glavas (1998) suggests that in most cases, firm productivity was enhanced following organizational changes. The changes could be small, such as new technology installation, or could be major, such as implementation of business process re-engineering program that aims to restructure the design of workflows (Cawsey, Deszca, &
Ingols, 2011). The likelihood of change adoption and innovation are, however, depending on the availability of financing. Dabla-Norris et al. (2012) state that “the level of productivity is constrained in the absence of finance” (p.426). Effective strategic organizational changes require updates on both the software and hardware components of firms. Human resources are the software that requires rigorous training and educational initiatives for changes, and capital assets are the hardware to be renewed (Appelbaum et al., 1998; Zwick, 2004). Updating the two requires financial resources (Leung et al., 2004).
In such context, the moral imperative that compels firm managers to allocate budgets for CID philanthropy, from the shareholders’ perspective, could mean a wrong use of a firm’s limited resources (Friedman, 1962; McWilliams et al., 2006). External CSR like CID lacks economic justification. It was found in Sánchez & Benito-Hernández (2015) not value adding to productivity.
Balabanis, Phillips, & Lyall (1998) also contend that external CSR has negative effects to a firm’s subsequent financial performance as it generates significant cost. In short run, the potential gains of external CSR were found to be offset by the cost burden (Boring, 2011). Although a small number of researchers link external CSR to productivity benefits, who contend that the good reputation generated increases attractiveness for recruitment and leads to better employee motivation and retention that are beneficial to productivity (Stuebs & Sun, 2010; Weber, 2008).
These researchers have not provided empirical evidence on the direct positive relationship of the two (also not focus on doing so). Instead, the majority of CSR researchers link productivity enhancements to internal CSR (Lo, Pagell, Wiengarten, Fan, & Yeung, 2014; Malik, 2015, Becchetti et al., 2005). Most studies have reached a consensus outcome: the positive effects of internal CSR, i.e., increased employee morale and satisfaction lead to enhanced productivity (Boring, 2018; Sun & Yu, 2015; Hasan et al., 2018).
12 The fact that CID initiatives compete for the limited corporate resources should not be overlooked (Lu et al., 2014). Firms that provide external financial support on CID may spend less on internal CSR or other internal value-enhancing change practices that are directly associated with productivity improvements (Gallego-Álvarez, Prado-Lorenzo, & García-Sánchez, 2011). Also, the need to maintain current resources after expending on philanthropies reduces the firm’s opportunity for innovation and other self-sufficiency investments (Fisher, Geenen, Jurcevic, McClintock, & Davis, 2009). Operations managers that foremostly take care of the shareholders’
profit-maximization interest may alternatively invest corporate resources on improving productivity by adopting organizational changes and innovative production strategies/systems (Lo et al., 2014; Barnett, 2007), or by improving job satisfaction and working conditions (Bhatti &
Qureshi, 2007; Phipps, Prieto, & Ndinguri, 2013). Some internal programs aim directly to improve production efficiency, such as lean, while some were empirically proven to improve employee self-development and participation that lead to increased productivity, such as the Quality circle program (Marks, Mirvis, Hackett, & Grady, 1986).
External CSR, internal CSR and other internal productivity-enhancing programs are all competing for the same limited corporate resources (Fisher et al., 2009). However, unlike the latter two that were proven by empirical studies to improve productivity and competitiveness (Sun & Yu, 2015;
Edmans, 2013; Bhatti & Qureshi, 2007), external CSR like CID initiatives bring no short-term economic benefits in operations management, but only result in decreased corporate resources (Lu et al., 2014). As firms’ access to resources was found to have strong effect on firm productivity (Gatti & Love, 2008), it is likely that there would be productivity differentials among firms that devote extra resources to CID and those that did not. The hypothesis is proposed as follows:
Hypothesis 1: Firms implementing CID practices have lower labor productivity than those that do not.
3.2 The level of slack resources affects CID’s impact on productivity
The slack resources theory is developed based on the view that a firm is able to carry out its activities due to the resources it owns (Fauzi & Idris, 2009). Some CSR researchers contend that firms with abundant slack resources have more degrees of freedom on devoting additional resources to commit to the society (Godfrey et al. 2009; Wang et al., 2016). Corporate resources can be in different forms, but most studies recognize financial performance as a proxy for slack resources (Arora & Dharwadkar, 2011). Fisher et al. (2009) suggest that firms with additional financial resources have more advantages to apply asset-based CID. Having sufficient resources to build social capital while keeping the responsible behaviors among their employees, firms with more slack resources have higher ability to maintain a certain level of productivity and efficiency
13 (Fisher et al., 2009). Moreover, Chen and Guariglia (2011) enclose that the firm-level productivity is positively influenced by the availability of the internal finance of the firms. They further found that for higher slack-resources firms, their productivity performance is less affected by the changes in short-term internal financial resources. Therefore, they stress that a firm’s slack resources help to smooth the negative impacts of its losses in internal finance (e.g., costs of CID implementation) on their productivity-enhancing activities.
Furthermore, slack resources allow firms to initiate and execute changes in strategy and finance productivity-enhancing activities (Buchholtz et al., 1999; Chen & Guariglia, 2011). If a firm has high level of slack resources available, their ability to invest for other programs, e.g., internal CSR, organizational changes and innovation for productivity improvements is less likely be affected by CID implementation (Arora & Dharwadkar, 2011). Consequently, this study expects firms with higher level of slack resources can attenuate the negative effect of CID implementation on productivity. It is hypothesized as follows:
Hypothesis 2: The more the slack resources are possessed by a firm, the less likely their labor productivity will be negatively affected due to CID implementation.
3.3 The amount spent on capital investment affects CID’s impact on productivity
While slack resources provide the opportunities for firms to invest in productivity enhancing activities, capital investment is the expenditure actually spent on the internal fixed and tangible assets (Aktas, Croci, & Petmezas, 2015). As discussed earlier, researchers have provided empirical evidence on the positive link between productivity and the level of financial resources committed to innovation and changes (Cainelli et al., 2005; Dabla-Norris et al., 2012). Researchers also relate capital expenditure to innovation and organizational changes because updated equipment, machinery and other hardware are required to support change programs (Appelbaum et al., 1998;
Hendricks & Singhal, 1997; Mo, 2009). Moreover, for firms with low slack resources, the increase in capital expenditures was found associated with increasing firm performance (Aktas et al., 2015).
They suggested channeling cash-release towards internal fixed assets are effective for manufacturing improvements.
In Al-Shuaibi (2016), the hypothesis: “There is a significant positive relationship between CSR and productivity” (p.142) was rejected. However, the researchers further discovered that socially responsible acts actually foster innovative programs inside firms, which in turn enhances productivity. However, whether CID alone leads to innovation has never been addressed or proven in prior study. I have discussed earlier an opposite view that CID takes corporate resources, which could reduce the opportunity for innovation. Yet, some firms are able to find the right balance to commit resources to CID without affecting their ability to adapt to the internal and external forces
14 for change and innovation to stay competitive (Fauzi & Idris, 2009; Barnett, 2007). In either case, i.e., investing in capital is facilitated by CID or is simply the managers’ strategic decision to stay competitive, it does not change the fact discovered in prior studies that investing more on adopting compatible capital assets is associated with productivity improvement (Cainelli et al., 2005; Chen
& Guariglia, 2011). Thus, I hypothesize capital investment decreases the negative impact of CID practices on a firm’s labor productivity as follows:
Hypothesis 3: The higher the capital investment was made by a firm, the less likely their labor productivity will be negatively affected due to CID implementation.
15
4. METHODOLOGY
Figure 4.1 displays the conceptual model of this research. To test the first hypothesis (H1), this research adopts a long-horizon event study. This method studies the abnormal performance (AP) of firms that are associated with a specific event, i.e., CID implementation by comparing their accounting-based performance against a group of benchmark firms (Hendricks & Singhal, 2015).
In long-horizon event studies, the year when the firm has not yet been impacted by the event is usually defined as the base year and used to calculate the AP (Lo et al., 2014). For my study, to observe for a relatively shorter period, quarter is used as the time horizon instead. To ensure time- specific data, this study denotes the particular event calendar quarter as quarter t and uses quarter (t-1) as the base time. To explore the long-term impact of CID, this research investigates the AP of labor productivity of the event quarter and the nearest quarters after the event, namely quarter t, (t+1) and (t+2).
For the second and third hypotheses (H2 and H3), this study uses regression analysis to test whether the level of slack resources and capital investment in quarter t affect the AP of productivity after the event in quarter (t+1).
Figure 4.1: Conceptual Model
4.1 Data collection and sample selection
This research focuses on the manufacturing industry of the United States. The unit of analysis is public firm that has devoted to CID events. The service firms are excluded because the service productivity concept is different from that of physical products due to their intangibility and perishability characteristics (Gronroos & Ojasalo, 2004). This research performs the retrospective news event detection (RED). It is a method to discover events from historical news corpus (Li, Wang, & Ma, 2005). This approach is appropriate for this study as the news articles disclose not only the name of the firms that involve in CID practices/events, but also the time information of each event. The news announcements were searched in Academic LexisNexis and Factiva. These two online databases record global business news from reliable major publishers, which include
16 The Wall Street Journal (WSJ), Dow Jones News Service (DJNS), Business wire and P.R.
Newswire. The news searching process of this research follows the keyword development strategy of the previous events studies in the production and operations management domain (Jacobs, 2014;
Jacobs & Singhal, 2014). They sampled the events by, firstly established a set of preliminary keywords, and used them to search the headlines and lead paragraphs of news announcements of a fixed period. They further derived additional keywords that were frequently used for the event from those announcements he found, such that more announcements were collected. For my study, the keywords derived from the definition of CID in ISO 26000 standards (2011) were used as the preliminary search strings. They are shown in the panel A of Table 4.1 below. The scope of the CID news announcement is set from 2007 to 2017, and the additional set of keywords spotted are presented in panel B of Table 4.1.
Table 4.1: Keywords Used in Search for CID Announcements
Panel A: Preliminary Set of Keywords Used for Searching
Announc* or Communit* or district* or neighbor* or people or public* or soci* or state* or local* or popula* resident* or territor* or ISO 26000 or standard* or guid* or CSR or responsi* or ISO Educat* or cultur* or disciplin* or learn* or literacy or scholar* or school*
or scien* or stud* or teach* or train* or prepar* or pedagogy* or tuition* or tutor* or upbring* or enlighte* or student* or Health* or hygien* or health care or health protection or medic* or wellness program or primary care or energy* or fitness or strength* or well- being or condition* or nutrition* or safe* or Job creation оr employ* or work* occupation*
or labor* or ventur* or hir* or select* or appoint* or empower* or internship* or skill* or staff* or open* or Social investment or social investing or social contribution or social financing or social investing or social funds or social funding or social initiative or social initiatives or social project or social projects or wealth* or incom* or technolog*
Panel B: Additional Set of Keywords Spotted (Near 5)
Aid* or build* or benefit* or contribut* or creat* or effort* or endevour* or giv* or inovat*
or invest* or launch* or particip* or protect* or promo* or recogn* or respect* or right* or solve* or support* or sustain* or sponsor* or volunteer* or Develop* or advanc* or improv*
or increas* or progress* or boost or betterment or grow* or enhance* or enrich* or Involv*
or engag* or concent* or encourage* or relat* or cooperat* or partner* or access*
The symbol “*” is used as a truncation operator which guides the news databases to search for the specified stem word with all valid endings (Jacobs & Singhal, 2014).
After deleting the privately owned firms and non-manufacturing firms, the total sample size is 104.
However, some firms have multiple CID events in the same event quarter (quarter t). This study
17 aggregates the quarterly events for each firm because those events will be examined with the same value of AP in the same period. 7 samples were aggregated, leaving a total of 97 samples.
The quarterly financial data and other company information of the sampled firms (e.g., employee number) required for measuring the other variables were acquired from the online COMPUSTAT database from Wharton Research Data Services (WRDS) (Jiang & Prater, 2006; Lo et al., 2014;
Lu et al., 2014). To ensure consistency in data time when comparing among firms, this study follows Hendricks & Singhal (2015) and uses the calendar data year and quarter from the finance report instead of the fiscal data year and quarter.
4.2 Selection of the control firms
To estimate the sampled firm’s expected performance if the CID event had not occurred, it is common in the long-horizon event study to match each sampled firm with one portfolio of comparable control firms to provide a performance benchmark for the samples (Lo et al., 2014). I created each portfolio based on not only the labor productivity performance, but also industry and firm size as Barber & Lyon (1996) assumed operating performance varies by industry and firm size. I matched these factors following the methods in Lo et al. (2014). I first matched the first two-digit of the standard industrial classification (SIC) code between sample and control firms to compare between firms within the same industry. When the matching criteria are more in numbers, the chances of finding the matched control firms become lower (Barber & Lyon, 1996). Therefore, I chose the two-digit matching instead of four-digit matching to include more potential control firms in the first stage for further selection.
After the first selection, those that fall within 50-200% of the sample firm’s total asset in quarter (t-1) were picked. This range (a factor of 2) helps remove firms that are extremely poor matches on size when comparing with the samples (Hendricks & Singhal, 2008) while keeping more firms for further matching in labor productivity of the control group.
Lastly, only those that fall within 90-110% labor productivity of the sampled firm in quarter (t-1) were chosen. This range is tighter because performance is the key observation, and the sample firms could have historically performed well or poorly (Barber & Lyon, 1996). However, there were 20 samples with no firms remain in the above comparison group. I then loosen their matching range of labor productivity to 85-115%. Nevertheless, there were still 14 samples (out of the 20) with no matching firm found, I then followed Hendricks & Singhal (2008) and Jiang & Prater (2006) to relax the requirement of the two-digit SIC code to one-digit, and looking for control firms with labor productivity between 90-110% of the samples. I had 7 samples’ (out of the 14) selection range on total assets narrowed down to 80-120% of that of the sample firms to reduce
18 the large number of their control firms while the remaining 7 samples keeps the original range because their control firms were only a few.
4.3 Variable definitions and measurement 4.3.1 Dependent variable
Labor productivity in this study will be measured by each firm’s total production (in millions of USD) divided by the number of employees. This study does not follow Cainelli et al. (2005) and Sun & Yu (2015) that use sales per employee as measurement because it does not consider inventory in the output. Since the sales data can only represent the number of outputs that have been successfully sold, those finished goods inventories that are ready for sale or resale should also be taken into consideration for measuring the actual production output (Sullivan et al., 2002).
In Zeile (1998) and Sanchez & Benito-Hernândez (2015), total production is measured as the monetary value of production output, e.g., in millions of USD, calculated as total sales plus the change in stock. The change in stock is the value of ending finished good inventory minus that of the beginning finished good inventory.
4.3.2 Variables that affect the impact of CID event on productivity
To test H2 and H3, abnormal productivity is used as dependent variable, and slack resources and capital investment are used as independent variables.
Slack can be in many forms (Jensen, 1986). Resources that provides high discretion to management are categorized to potential slack and available slack (Arora & Dharwadkar, 2011). Potential slack, such as the debt-to-equity ratio measures a firm’s ability to raise cash shortly (Navarro, 1988). I do not use it to measure slack because the equity’s ability to cover debt is more important in the event of a business downturn, which is not the case for the CID event. Available slack is defined as “the extent to which firms have resources that are untapped, but readily available” (Marlin, 2014, p. 24). Cash and account receivable are often used as the measurement of available slack (Seifert et al., 2004; Arora & Dharwadkar, 2011). However, these variables are short-term assets, and they alone cannot represent a firm’s short-term financial health. A proper measurement of a firm’s liquidity, i.e., available slack, takes into consideration its short-term obligation (Chen &
Guariglia, 2011; Ek & Guerin, 2011). Therefore, my study follows Marlin (2014), operationalizes available slack using working capital as the measure. It is calculated by current assets minus current liabilities, normalized by total assets. For operations management, having sufficient working capital is the prerequisite of improving labor productivity as it helps maintaining a regular supply of raw materials for smooth production processes and provides timely payment of wages (Becchetti et al., 2005; Zeile, 1998; Aktas et al., 2015).
19 Following Hendricks & Singhal (1997), the Capital investment is measured using the capital expenditure to assets ratio. The ratio helps to clarify the level of capital invested by firms with different sizes and financial conditions. Capital expenditure in this study is the property, plant, and equipment expenditures, excluding the fixed assets acquired through merger or acquisition.
4.3.3 Control Variables
To establish the rigorousness of this study, industry, firm size, return on assets (ROA), capital structure and cash dividend in quarter (t-1) are controlled. The former four factors are the most frequently used control variables in explaining the CSR-CFP relationship (Lu et al., 2014).
According to Hendricks & Singhal (2008), industry-specific effects lead to variability in accounting data. The industry-type is controlled based on the four-digit SIC code.
Second, following Lo et al. (2014), firm size is measured as the natural algorithm of the number of employees. It should be noted that this measurement is different from the measurement using total assets for selecting the control firms with similar firm size. This difference is also in their long-horizon event study.
Third, ROA is calculated by net income divided by total assets. It is controlled because firms that are more profitable could have more resources to achieve higher performance in the future (Lu et al., 2014).
Moreover, following Lu et al. (2014), capital structure is operationalized as leverage ratio. It can be calculated as total liabilities divided by total equity (total assets minus total liability). Leverage is controlled since the access to external financial resources and being under financial constraints can have crucial effects on the firm’s ability to improve its productivity performance, and high leverage captures the credit constraints faced by a firm (Thangavelu & Findlay, 2012).
Lastly, considering firms returning free cash as dividends to the shareholders would reduce liquid resources for productivity-enhancing activities, this study follows Arora & Dharwadkar (2011), controls cash dividend. However, they also indicate that cash dividends reduce agency loss (if any) caused by CSR.
4.4 Research design
4.4.1 Event study approach for longitudinal analysis
To be consistent to prior research using the long-horizontal event study method, e.g., Hendricks &
Singhal (2008) and Lo et al. (2014), this research firstly tests H1 with longitudinal analysis for the
20 long-term impact of CID event on labor productivity. This approach calculates the abnormal performance of the sample firms with the proposition that the effects of CID event are reflected in the financial performance reported by the firms. In Lo et al. (2014) the calculation is provided as follows:
AP(t+j) = PS(t+j) − EP(t+j)
where AP is sample firm’s abnormal performance, t is the event time, j is the ending time of comparison, PS is sample firm’s actual performance and EP is sample firm’s expected performance.
Expected performance is the sampled firm’s performance if the CID event had not happened;
however, it is unobservable, consequently it is estimated using the performance of the benchmark firms (i.e. control firm portfolio) that are similar to the samples. Hendricks & Singhal (2008) estimates EP with the performance of the sample firm in the base time plus the change in median performance of the control firms over the period of interest. The calculation formula is as follows:
EP(t+j) = PS(t+i) + (PCk(t+j) − PCk(t+i))
where EP is sample firm’s expected performance, PS is sample firm’s actual performance, i is the base time, PCk represents the median performance of the control firms of the kth sample firm, and j is the ending time of comparison.
4.4.2 Cross-sectional regression analysis of contextual factors
H2 and H3 explore the impact of contextual factors (i.e. slack resources and capital investment) on the causal relationship proposed in H1. To test both hypotheses, AP of labor productivity is used as the dependent variable while slack resources and capital investment being the independent variables in a cross-sectional analysis. Following Lo et al. (2014) and Hendricks & Singhal (2008) that use the similar methodology, this study adopts ordinary least squares (OLS) regression.
However, this study tests not only the significance of the independent variables. Three models are set: one that includes slack resources (IV1) and all control variables, one that includes capital investment (IV2) and all control variables, and the last one with only the control variables. By comparing these models, the effects of the IVs could then be observed. However, in section 5.2 industry type (CV1) and slack resources (IV1) were found to have multi-collinearity. Consequently, I removed the control variable industry type from the analyses. The OLS regression models is therefore specified as follows:
Model 1: APk= α+ β1 × FirmSize(k) + B2 × ROA(k) + B3 × CapitalStructure(k) + B4 × CashDividend(k) + B5 × SlackResources(k) + ε
21 Model 2: APk= α+ β1 × FirmSize(k) + B2 × ROA(k) + B3 × CapitalStructure(k) + B4 × CashDividend(k) + B5 × CapitalInvestment(k) + ε
Model 3: APk= α+ β1 × FirmSize(k) + B2 × ROA(k) + B3 × CapitalStructure(k) + B4 × CashDividend(k) + ε
where APk is abnormal productivity performance, α is the intercept (constant), βx are the coefficient of variable x, and ε is the error term. All IVs and CVs are the data in quarter (t-1).
5. RESULTS
5.1 Event Study Results (H1)
The statistic results of the quarterly cumulative abnormal performance (AP) of labor productivity are presented in Table 5.1. Based on the consistent results of the (parametric) paired-sample-t-test, the (non-parametric) Wilcoxon signed-rank (WSR) and sign tests, the sample firms’ AP was proven negatively significant at the 5% level for the comparison period quarter (t-1) to quarter t.
The mean change of this period was -6.365 thousand US dollar per worker and the median change was -3.872 thousand. 62.89% of the sampled firms have negative AP on labor productivity. This observation provides support to H1. Furthermore, for the cumulative period quarter (t-1) to quarter (t+1), only the sign test result indicates that the sample firms’ actual performance was significantly worse than their expected performance at the 10% level as 58.76% of the sample firms underperformed. Moreover, all three tests indicate that the cumulative AP for quarter (t-1) to quarter (t+2) does not significantly differ from zero. The results imply that CID initiatives have relatively faster and shorter negative effects on the sample firms’ labor productivity. I therefore discarded the comparison period quarter (t-1) to quarter (t+2) for regression analysis in the following section.
Table 5.1: Results of the 97 sample firms’ cumulative AP of labor productivity over the 10-year study period (2007-2017)
Q(t-1) to Qt Q(t-1) to Q(t+1) Q(t-1) to Q(t+2)
Mean ($ in thousands) -6.365 -4.316 -7.957
t-statistic -2.043** -1.236 -1.552
Median ($ in thousands) -3.872 -2.729 -1.912
Z-statistic (1) -2.143** -1.128 -1.420
% negative AP 62.89% 58.76% 56.70%
Z-statistic (2) -2.437** -1.625* -1.218
Note: t-statistic is resulted from the t-test (mean); Z-statistic (1) is resulted from the WSR test (median); Z- statistic (2) is resulted from the sign test (percentage).
All tests are two-tailed: *p<=0.10; **p<=0.05; ***p<=0.01.
22 5.2 Regression Analysis (H2 & H3)
Before conducting the analysis, the independent variables and control variables were checked whether there is multi-collinearity. Table 5.2 displays the correlation matrix of the variables. It shows that the correlation between CV1 Industry type and IV1 Slack resources is higher than 0.7 and this suggests the problem of multi-collinearity between the two (Sartal Rodríguez, & Vázquez, 2019).
Furthermore, table 5.3 shows the tolerance and variance inflation factors (VIF) between the IVs and CVs. The result shows that when both the Industry type and Slack resources are included in the linear regression, the tolerance values between the two and other variables become very low while their VIF value gets very high (the underlined values). Since the cut-off value of the collinearly tolerance is 0.1, and 2.0 for the VIF value (Neter et al., 1990), the highlighted (underlined) values in table 5.3 have further indicated the high chance of multi-collinearity between Industry type and Slack resources.
Table 5.2: Correlation of the IVs and CVs in regression analysis
Correlation Matrix
1 2 3 4 5 6 7
1. Industry Type 1
2. Firm Size 0.259 *** 1
3. ROA -0.120 -0.060 1
4. Capital
Structure 0.116 0.003 -0.050 1
5. Cash Dividends 0.504 *** 0.521 *** -0.055 -0.013 1
6. Slack Resources 0.923 *** 0.297 *** -0.126 0.115 0.523 *** 1 7. Capital
Investment -0.279 *** 0.024 0.148 * 0.037 0.094 -0.240 *** 1
All tests are two-tailed: *p<=0.10; **p<=0.05; ***p<=0.01.
Table 5.3: Collinearity statistics of the IVs and CVs in regression analysis
Collinearly Tolerance (VIF)
1 2 3 4 5 6 7
1. Industry Type 0.143 0.142 0.142 0.145 0.618 0.147
(6.996) (7.043) (7.026) (6.903) (1.617) 6.896
2. Firm Size 0.727 0.723 0.722 0.899 0.727 0.723
(1.375) (1.384) (1.385) (1.112) (1.375) (1.383)
23
3. ROA 0.966 0.967 0.968 0.966 0.967 0.982
(1.035) (1.034) (1.033) (1.035) (1.034) (1.019)
4. Capital Structure
0.968 0.966 0.968 0.977 0.967 0.977
(1.033) (1.035) (1.033) (1.023) (1.034) (1.023)
5. Cash Dividends 0.539 0.658 0.528 0.534 0.535 0.570
(1.855) (1.520) (1.893) (1.871) (1.868) (1.753)
6. Slack Resources 0.622 0.144 0.143 0.143 0.145 0.143
(1.608) (6.956) (6.994) (6.996) (6.913) (7.004)
7. Capital Investment
0.855 0.826 0.839 0.835 0.891 0.826
(1.170) (1.210) (1.192) (1.197) (1.122) (1.211)
To eliminate multi-collinearity problem from my analyses, I removed the control variable Industry type from the models. Table 5.4 displays the correlation matrix of the variables without the control variable Industry type. It shows that none of the correlations presents high values (higher than 0.7), and this suggests that the effect of multi-collinearity is limited (Sartal et al., 2019). In addition, table 5.5 shows the tolerance and VIF between the IVs and all other CVs. The numbers tell that the variables should cause no problem to the estimates of regression coefficients as none of the tolerance value is close to 0.1, and the largest VIF value is 1.585, which is still below the strictest cut-off value 2.0 (Neter, Wasserman, & Kutner, 1990).
Table 5.4: Correlation of the IVs and CVs in regression analysis without the CV Industry type
Correlation Matrix
1 2 3 4 5 6
1. Firm Size 1
2. ROA -0.060 1
3. Capital Structure 0.003 -0.050 1
4. Cash Dividends 0.521 *** -0.055 -0.013 1
5. Slack Resources 0.297 *** -0.126 0.115 0.523 *** 1
6. Capital Investment 0.024 0.148 * 0.037 0.094 -0.240 ** 1
All tests are two-tailed: *p<=0.10; **p<=0.05; ***p<=0.01.
