Position Paper, Values, ethics and sustainability
Journal of Business Ethics (2020) 162:449–472 https://doi.org/10.1007/s10551-018-4005-0 ORIGINAL PAPER Employee Volunteer Programs are Associated with Firm-Level Bene ts and CEO Incentives: Data on the Ethical Dilemma of Corporate Social Responsibility Activities Brian D. Knox 1 Received: 12 February 2018 / Accepted: 18 August 2018 / Published online: 29 August 2018 © Springer Nature B.V. 2018 Abstract Ethical dilemmas arise when one must decide between conflicting ethical imperatives. One potential ethical dilemma is a manager’s decision of whether to engage in corporate social responsibility (CSR) activities. This decision could pit the ethical imperative of honoring unwritten obligations to society against the ethical imperative of honoring con – tractual obligations to the firm. However, CSR activities might only be a minor ethical dilemma or none at all if they simultaneously benefit the firm and society. To examine this I test the association between future-period employee productivity and current-period use of one type of CSR activity: employee volunteer programs. I use a unique sample of 1428 firm-years, hand-collected from sustainability reports of 373 firms. I find evidence that the current-period use of an employee volunteer program has a positive association with future-period employee productivity (moderated by the firm’s current-period employee productivity). I find this result in future periods up to 6 years after a firm uses an employee volunteer program. I also find a positive association between incentives that focus CEOs’ attention on long- term firm outcomes and more extensive employee volunteer programs (also moderated by current-period employee productivity).
Keywords Corporate social responsibility· Employee productivity· Ethical dilemma· Employee volunteer program· Executive compensation Introduction Managers could face ethical dilemmas in deciding whether their rm or department should engage in corporate social responsibility activities (hereafter CSR activities ). Ethical dilemmas are situations in which at least two competing ethical imperatives conict, leaving one to make a subjec – tive decision about how to behave most ethically (given that acting in accordance with all ethical imperatives is impos – sible in such situations; Allen 2012 ; see also McMahon and Good 2016 for a similar discussion). CSR activities are often activities that satisfy unwritten contractual obligations a manager may feel he or she has to society. A manager may feel society has altruistically invested in him or her and feel an obligation to reciprocate altruistically to society’s benet. 1 However, managers are also employees of the rm and have an explicit contractual obligation to act in a way that benets the rm and maximizes shareholder wealth. If CSR activities are unprotable, managers are left with an ethical dilemma in which the ethical imperative of honoring contractual obligations to society conicts with the ethical imperative of honoring contractual obligations to the rm.
The magnitude of this ethical dilemma depends, at least in part, on the degree to which CSR activities are protable. Unfortunately CSR activities’ protability is often di cult to measure. One reason for this is that CSR activities’ ben – ets to the rm are often unknown. If there are benets to the rm from CSR activities, these benets are usually indirect, * Brian D. Knox email@example.com 1 Boise State University, Boise, USA 1 There is some similarity between this explanation of what moti – vates people to engage in CSR activities and what motivates workers to reciprocate when a gift wage is received (e.g. Kuang and Moser 2009 ; Fehr and Gächter 2000 ).
Vol.:(0123456789) 1 3 450 B. D. Knox delayed, or both, making them empirically di cult to measure (see Peloza 2009). Many measures of CSR activities’ benets are incomplete and may miss key trade-o s inherent to CSR activities. Overall market reactions may also be inadequate measures of CSR activities’ protability because markets might ine ciently value rms at more than they are worth when those rms engage in CSR activities (Martin and Moser 2016).
In my study I test the long-term rm-level benets of a specic type of CSR activity: employee volunteer pro – grams. Employee volunteer programs are management-led initiatives to facilitate and encourage employee volunteer – ism in the local community. Prior studies provide evidence that employee volunteer programs improve employee loyalty and skill level (Jones 2010; De Gilder etal. 2005; Peterson 2004). 2 These prior studies primarily focus on individuals.
My study attempts to extend this body of work with rm- level evidence to test whether employee volunteer programs are benecial to the rm. Firm-level data can provide con- rming evidence of the total e ect of employee volunteer programs, net of any costs or disadvantages that might have escaped measurement in prior studies. I treat employee productivity (measured as the natural logarithm of an income statement performance measure per employee) as a proxy for the employee loyalty and skill level that are reportedly improved through employee volunteer programs. Then I use a unique sample of archi- val data (1428 rm-years hand-collected from the Global Reporting Initiative sustainability reports of 373 rms) to measure which rm-years use an employee volunteer pro- gram. I nd support for the hypothesis that current-period use of an employee volunteer program is associated with increased employee productivity in future periods, mod – erated by a rm’s current-period employee productivity. 3 This moderating relationship means that rms with higher current-period employee productivity have a less positive association between current-period use of an employee vol- unteer program and future-period employee productivity.
These results are robust to three di erent proxies of future- period employee productivity and three di erent proxies of current-period use of an employee volunteer program. For some proxies, I nd a signicant result up to 6 years later, which is the latest year I test in this study.
I also examine the association between CEO incentives and employee volunteer programs. This analysis is a cor – roborating test of the long-term e ects of employee volun- teer programs. CEO incentives can motivate managers to shift focus toward or away from long-term rm outcomes. If CEOs truly expect employee volunteer programs to improve long-term employee productivity, then they will most likely encourage more extensive employee volunteer programs (i.e., more volunteer hours) when their incentives motivate them to focus more on long-term rm outcomes. I nd evi- dence of this predicted association between CEOs’ long- term-focused incentives and more extensive employee vol- unteer programs. 4 I also nd evidence that this association, as expected, is moderated by current-period employee pro- ductivity. Firms with higher current-period employee pro – ductivity have a less positive association between theseCEO incentives and employee volunteer program extensiveness. Collectively my ndings suggest that employee volunteer programs are associated with real long-term benets to the rm and that CEOs act in expectation of these benets. My results support the notion that some CSR activities might pose ethical dilemmas of relatively low magnitude or might not pose ethical dilemmas at all. In the case of employee volunteer programs, my study suggests that managers could engage in these CSR activities to satisfy both their contractual obliga- tions to the rm as well as obligations they feel toward society. Hypothesis Development Ethical Dilemmas andManagers’ Choice ofCSR Activities Acting ethically means acting in accordance with the ethi- cal imperatives prescribed by pertinent ethical frameworks.
Sometimes these imperatives are relatively simple, such as the ethical imperative for a doctor to preserve a patient’s life. At other times ethical imperatives conict, making it impossible to meet all ethical imperatives and forcing one to choose between ethically unfullling alternatives. In an emergency, for example, a doctor might have to balance the ethical imperative to preserve the life of a pregnant mother and the ethical imperative to preserve the life of her unborn child. Depending on the resources available and each patient’s likelihood of survival, balancing these ethical 2 I list improved employee skill level as a potential benet from employee volunteer programs. This expectation may be more prac- tical and intuitive than scholarly. Jones (2016) calls into question employee volunteer programs’ e ect on employee skill level, argu- ing that the empirical evidence of this relationship is hardly compel- ling. Given that my hypotheses could be driven entirely by changes in employee loyalty, I expect that my conclusions would be largely unchanged if I were to remove any expectation that employee volun- teer programs improve employee skill level.
3 In my tests of the moderating e ect of current-period employee productivity, I also control for a main e ect of current-period employee productivity on future-period employee productivity. I dis- cuss this further in subsequent sections. 4 As explained in the “Sample and Measures” section and in the “Hypothesis Tests for H2a and H2b”subsection of the “Results” sec- tion, I dene employee volunteer program extensiveness as the natural logarithm of total volunteer hours. I also include several rm-level measures to control for correlations between rm size and employee volunteer program extensiveness.
1 3 451 imperatives can be di cult. Cases where ethical imperatives conict are called ethical dilemmas. 5 CSR activities may present a critical and wide-spread ethical dilemma. For example, Martin (2014) nds that some participants in a laboratory experiment sacrice the rm’s nancial well-being for the benet of society through CSR activities. Some of those participants likely felt that engaging in unprotable CSR activities was the most ethical action they could take because it was in line with an ethical imperative to benet society. The ethical imperative to benet society through CSR activities can be complex and can vary widely from person to person. People can receive many benets from human society without directly paying for them, such as community and religious support, the nurture and direction of parents and mentors, the companionship of family, spouse, and close friends, natural or community resources, the awe-inspiring wonders of nature or art, etc. Often, people feel a desire to reciprocate toward society as a whole out of a feeling of obligation for what they have freely received. Hence the saying that one “wants to give back.” Sometimes legal terms are used to describe this ethical imperative, calling it a social contract (e.g., Thompson and Hart 2006; Dunfee and Don- aldson 1999). Social contracts are almost always unwritten but appear to have strong enough emotional e cacy to inu- ence managers’ decisions (see Martin 2014; see also Martin and Moser 2016). Managers may feel a desire to engage in CSR activities because they benet society at large, and they may feel this is an ethical imperative to fulll their part of a social contract. Managers are also agents of the rm and have generally agreed to a written, legally-binding employment contract to perform certain actions for the benet of the rm. There is an ethical imperative to honestly fulll this contract as well. These employment contracts typically compel the man- ager to engage in activities that increase the protability of the rm. If a CSR activity is not protable for the rm, an ethical dilemma may arise because a social contract and an employment contract suggest opposite actions. If a CSR activity is protable, then managers are less likely to face signicant ethical dilemmas related to that CSR activity, which benets both the rm and society. The degree to which CSR activities present an ethical dilemma hinges on the protability of those CSR activities to the rm.
The above framing, in which managers choose between a contract with society and a contract with the rm, is deliber – ately simple. It is beyond the scope of this paper to explore the multitude of complex variations of this ethical dilemma under di erent rm situations, manager backgrounds and motivations, and CSR activities. However, inthe “Employee Volunteer Programs and Long-term Employee Productiv – ity” subsection,immediately following H1a, I do provide additional discussion of how this ethical dilemma relates to employee volunteer programs specically. Market Reactions toCSR Activities In many cases the beneciaries of a CSR activity, such as the homeless, the environment, local schools, etc., do not directly reciprocate and benet the rm in such a way that would make the CSR activity directly protable to the rm.
Those who argue that CSR activities are protable generally make their case by taking indirect benets of CSR activities into account. These indirect benets, however, are di cult to track. Several studies have used market reactions as a way of tracking overall protability of CSR activities because market reactions are expected to incorporate CSR activi- ties’ indirect benets (e.g., Zahller etal. 2015; Elliott etal.
2013 ; Dhaliwal etal. 2011,2012 ; Kim etal. 2012). Many such studies nd a positive market reaction to CSR activi- ties. But these market reactions may be insu cient to track the protability of CSR activities because nancial markets may ine ciently perceive CSR activities in a positive light. Martin and Moser (2016) conduct an experiment using a laboratory market in which there is clearly no direct or indirect nancial benet from CSR activities. They nd that investors e ectively pay a premium for CSR activities, per se, without any increase to shareholder wealth. This nd- ing raises signicant questions about whether market-based results are e ective proxies for CSR protability, introduc- ing a counter-explanation for why markets might positively react to CSR activities. Subjective feelings toward CSR activities, rather than objectively calculated increases in shareholder wealth, could drive nancial markets’ reactions to CSR activities. From the perspective of data availability, it would be highly lamentable if markets did not e ciently react to CSR activities. Researchers face signicant limitations in measuring CSR activities’ benets, and market reactions could account for a more complete picture of CSR activi- ties’ protability, including hard-to-measure indirect benets from CSR activities. To supplement studies based on market reactions, I focus on the benets of a particular CSR activ – ity, i.e., employee volunteer programs, that prior research 5 The term ethical dilemma is often used to describe ctional sce- narios designed to bring multiple ethical imperatives into conict.
As teaching tools, ctional ethical dilemmas require students to think carefully about the conicting ethical imperatives and to develop strategies for maximizing ethical behavior, which can be useful when actual ethical dilemmas arise in practice. In an ethical dilemma, actual or ctional, managers have no rst-best option that allows them to fully act in accordance with all ethical imperatives. How people go about maximizing ethical behavior in ethical dilemmas is, of course, a source of extensive study in the eld of ethics (e.g.
Ryan and Bisson 2011; Priem and Sha er 2001; Barnett et al. 1994; McCabe etal. 1994; Weber 1994).
Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 452 B. D. Knox suggests will produce indirect benets that I can empirically examine. My empirical measures are separate from market reactions. In this way my study complements prior studies that are based on market reactions by providing independent evidence about at least one CSR activity’s indirect benets to the rm, with little opportunity for subjective evaluation of the worth of that CSR activity.
Employee Volunteer Programs andLong‑term Employee Productivity An employee volunteer program is an initiative sponsored by rm management that provides resources, time, leader – ship, or social encouragement so employees can volunteer in the local community. Peloza etal. (2009) report that the prevalence of employee volunteer programs increased by as much as 150% in the years leading up to their article. Prior research suggests employee volunteer programs can indi- rectly benet the rm by way of signicant improvements in employee loyalty and skill level (Rodell 2013; Jones 2010; Booth et al. 2009; De Gilder et al. 2005; Peterson 2004; Caudron 1994). Although the focus of this paper is rm-level outcomes, it is useful to explain how employees are understood to react to employee volunteer programs and why, specically, employee volunteer programs could improve employee loyalty. Rodell (2013) nds evidence that employee volunteerism improves several types of job per – formance, including task performance, citizenship behavior, and counter-productive behavior (i.e., volunteering is posi- tively associated with the rst two behaviors and negatively associated with the last behavior). She nds evidence that at least part of this e ect is because volunteerism allows employees to mentally and/or emotionally recharge and thus approach their job task with greater attention and e ort (i.e., higher job absorption).
The e ect volunteerism has on citizenship behavior, in particular, is important. Citizenship behavior (sometimes called organizational citizenship behavior) is character – ized by going above and beyond the required behaviors for one’s job. For example, citizenship behavior can relate to how much employees praise their rm outside of work, how much employees make suggestions to improve the overall functioning of the rm or attend non-mandatory but impor – tant meetings, and how much employees are mindful of co-workers, especially new ones (see Rodell 2013). There is some overlap between this concept and the reciprocal behavior theorized by gift wage researchers (e.g., Kuang and Moser 2009; Fehr and Gächter 2000), although it would be incomplete to lump these two terms together as synonyms.
Citizenship behavior has been linked to good rm-level out – comes (e.g., Koys 2001).
In a similar vein Jones (2010) nds that employee vol- unteer programs improve employees’ pride in the firm, which improves their identication with the rm, lead – ing to an improvement in employees’ intention to remain at the rm as well as an improvement in some employees’ citizenship behavior. For measures of citizenship behavior, improvement is moderated by employees’ exchange ideol- ogy, or how strongly they feel that good behavior by the rm should lead to reciprocal good behavior by employees.
Booth etal. (2009) provide evidence that positive responses to volunteerism appear consistent with social exchange theory. Combined, the general thrust of prior studies is that volunteerism has benets related to the e ort and attention employees invest in their work and that employee volun- teer programs can increase employees’ behavior that goes above and beyond minimum work duties, especially if those employees feel a need to reciprocate good treatment by the rm (see also De Roeck and Maon 2016).
Prior studies on employee volunteer programs primar – ily focus on individuals. To examine employee volunteer programs’ e ects at the rm level, I look for a rm-level operational proxy of the expected benets of employee vol- unteer programs. Increased employee loyalty can reduce unwanted employee attrition, which leads to a more expe- rienced and motivated workforce and increases employee productivity (see Kim etal. 2010; De Gilder etal. 2005; Benjamin 2001; Heskett etal. 1997). 6 Greater employee skill level can also lead to employees being more e ective at their work and thus more productive (see Dearden etal. 2006, who document the relationship between training—or skill level—and employee productivity; see also Holzer 1990).
Because employee loyalty and skill level both signicantly impact employee productivity, I use employee productivity as a rm-level proxy for employee loyalty and skill level. 7 It is feasible that while employee volunteer programs improve employee productivity on an individual level, they also lead to costs or disadvantages at the rm level that a ect employee productivity in the opposite direction, possibly o setting some individual-level benets. Potential disadvantages include, for example, that employee volunteer programs might lead to unwanted attrition over time because otherwise skilled and loyal employees nd more fulllment 6 Although I focus on improved employee loyalty and skill level as benets of volunteer programs, volunteerism is also associated with several other types of self-improvement that might a ect long-term employee productivity. For example, in summarizing research on the benets of volunteerism, Brooks (2009) says, “People who volunteer do better nancially [….] People who volunteer are happier [….] Vol- unteering actually gives people a mild sense of euphoria.” 7 In the organizational productivity literature, employee productiv – ity is dened using a logarithmic transformation of an income state- ment performance measure divided by the number of employees (e.g.
Bloom and Van Reenen 2011; Munday and Peel 1997; Huselid 1995).
I discuss and use several such measures for employee productivity in the“Sample and Measures” and “Results”sections.
1 3 453 at a volunteer organization they encounter through an employee volunteer program. 8 Or, the cost of employee volunteer programs might be paid for by a decrease in other human resource activities that impact employee loyalty and skill level. I expect there could be several other potentially o setting disadvantages beyond those listed above. 9 Thus, I believe it useful to the body of knowledge to analyze the e ect of employee volunteer programs at the rm level, where the costs and/or negative e ects of employee volun- teer programs, should they exist, can be balanced against the benets of employee volunteer programs that have beenpre – viously observed at the individual level. I expect that the monetary costs of employee volunteer programs are generally incurred in the short-term, that is, in the current period. 10 However, the indirect benet of employee volunteer programs, i.e., increased employee productivity, is likely to be realized over the long-term, that is, in future periods. Changes in citizenship behavior and unwanted attrition can take considerable time to accrue to a noticeable benet. Employee skills can also take con – siderable time to increase and thus can take time to a ect employees’ productivity (e.g., Jones 2010 used an interval of 6 months before beginning to measure employees’ response to employee volunteer programs; see also Bhatti and Qureshi 2007). Furthermore, skills acquired through volunteer activi- ties, if any, are likely to be soft skills, leadership skills, or teamwork skills that employees might not be able to apply directly to their current work but that may improve employee productivity in future periods when those employees are promoted and receive additional assignments (see Peterson 2004 ). Thus, I expect employee volunteer programs to be associated with improved employee productivity in future periods rather than in the current period.
H1a Firms that use employee volunteer programs in the current period have greater employee productivity in future periods than rmsthat do not use employee volunteer pro- grams in the current period.
Employee volunteer programs might provide a relatively complex ethical dilemma. Managers may feel a desire or an obligation to support employees in things they really care about. These managers may then feel motivated to adopt, support, or maintain an employee volunteer program because it is what employees want. Employee volunteer programs may satisfy an unwritten social contract between managers and employees, rather than between managers and society.
The fact that this social contract may be with employees rather than with society as a whole does not preclude the possibility that sometimes this contract can suggest an action that conicts with the manager’s contract with the rm. It is feasible that some managers, in trying to honor this social contract with employees, might choose an activity that is not in the interest of the rm.
On the other hand this explanation of why managers might use employee volunteer programs leads to a di er – ent reason such programs might be protable to the rm.
Because these programs can be attractive to employees who want to participate in them, they may be useful tools in the war for talent. This could improve employee productivity because more productive workers select into and remain at a rm due to the rm’s employee volunteer program (see Jones etal. 2014). I do not expect this alternative mechanism for employee volunteer program protability to lead to sig- nicantly di erent predictions for H1a and H1b, although it might suggest di erent recommendations for practice. Employee volunteer programs are not likely to be the most e ective means a rm has for improving employee productivity, since this is not the primary purpose of these programs. The primary purpose of the program is to encour – age volunteerism. The benets of employee volunteer pro- grams are indirect, rst, in terms of any skills gained through volunteerism and, second, in terms of employee loyalty.
Skills gained may not be fully transferrable to an employee’s rm-beneting work. Also, employee volunteer programs might not be equally e ective for all members of the rm.
Peterson (2004), for example, provides evidence that females are more likely to perceive employee volunteer programs as building job skills than males. Therefore, the e ects of an employee volunteer program might di er by gender. Simi – larly, De Gilder etal. (2005) nd socio-demographic di er – ences between employees that participate in an employee volunteer program and those who do not. Employee volun- teer program benets may be less direct and more di cult to generalize across the workforce than traditional skill training and loyalty-bolstering activities. Employee loyalty, per prior research, can be a ected by various factors, and employee volunteer programs might only play a small part in determining employees’ inten- tions to stay at the rm or their citizenship behavior (e.g., Jones 2010). Furthermore, if employee volunteer programs are primarily a tool in the war for talent, one would expect the employee productivity benefits of employee volun- teer programs to diminish when current-period employee 8 I am anecdotally aware of instances where this has happened, although I lack empirical evidence of the net e ect these employees’ departure had on their respective rms.
9 For example, it may be that employee volunteer programs tend to coincide with inconsistencies in CSR strategies. Scheidler et al.
(2018) nd that inconsistent CSR strategies tend to lead to higher employee turnover intentions, and higher rm-level employee turno- ver rates.
10 Throughout the paper, I use period and year interchangeably.
I refer to the current year as the current period and future years as future periods.
Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 454 B. D. Knox productivity reects an already-talented set of employees.
Thus, rms that already have high employee productivity in the current period most likely already utilize superior tradi- tional tools for skills training and loyalty-building or already have a large contingent of talented employees. I expect these rms to derive less marginal benet from employee volun- teer programs.
H1b Current-period employee productivity moderates the association between employee volunteer programs and future-period employee productivity such that rms with high current-period employee productivity have a less posi- tive association between current-period use of employee vol- unteer programs and future-period employee productivity than rms with low current-period employee productivity.
H1b predicts moderation. It predicts that current- period employee productivity moderates the effect that current-period use of an employee volunteer program has on future-period employee productivity. I will test this moderation while simultaneously controlling for the main e ect of current-period employee productivity on future- period employee productivity. This is an essential detail of my research design because it helps disentangle my pre- dicted H1a and H1b e ects from many other factors that drive employee productivity. These extraneous (and, in my sample, largely unobserved) factors should drive both current-period employee productivity and future-period employee productivity. I expect that testing the main e ect of current-period employee productivity on future-period employee productivity will provide an e ective control for these unobserved factors, increasing my ability to interpret incremental e ects related to H1a and H1b.
CEO Incentives andEmployee Volunteer Programs As an important second objective of this paper, I include analysis about CEO incentives to corroborate my rm-level tests of the long-term benets of employee volunteer pro- grams. If CEOs expect employee volunteer programs to improve long-term employee productivity, CEOs who are more motivated to achieve long-term protability should cultivate more extensive employee volunteer programs, meaning employee volunteer programs with more volunteer hours. This analysis tests whether CEOs, who have consid- erable knowledge of the rm, expect long-term benets to arise from employee volunteer programs. If CEOs had no belief that these programs would a ect long-term rm out- comes, I generally would not expect incentive-associated di erences in how CEOs manage their rms’ employee vol – unteer programs. 11 Pension compensation is the class of CEO incentive I am most interested in. I expect that pension compensation is positively associated with employee volunteer program extensiveness. Finance research suggests that CEO pension compensation can be a critical driver of executive and rm behavior (Sundaram and Yermack 2007; see also Cassell etal. 2012; Jensen and Meckling 1976). Pension compensa- tion is sometimes called inside debt because pensions are owed to the CEO by the rm. Inside debt, however, may only be collectable by the CEO insofar as the rm has pension assets equal to pension liabilities, and CEOs are likely to lose most or all of their claim to remaining pension com- pensation if the rm goes bankrupt. 12 The risk of losing claim to inside debt motivates CEOs to seek stronger guarantees of the rm’s long-term nancial stability. Sundaram and Yermack (2007) nd a signicant relationship between higher inside debt and decreases in risky rm behavior (see also Cassell etal. 2012). In short, inside debt ties at least a part of the CEO’s compensation to the long-term viability of the rm, shifting CEO atten- tion toward long-term rm outcomes. I expect that CEOs with more pension compensation have higher opportunity cost from passing up the benets of an employee volunteer program. This di erence in opportunity cost leads to the prediction that CEOs’ pension compensation is positively associated with the extensiveness of rms’ employee vol- unteer programs.
H2a CEO pension compensation is positively associated with the extensiveness of employee volunteer programs.
If CEOs anticipate long-term benets from employee vol- unteer programs, as predicted in H2a, it follows that they will be more likely to expect those benets when current- period employee productivity is low (see H1b). When the rm has high current-period employee productivity, exten- sive employee volunteer programs should be less positively associated with long-term employee productivity, meaning less opportunity cost from passing up employee volunteer program benets for CEOs with high pension compensa- tion. I expect an interaction comparable to H1b but with 12 At least, that is true for bankruptcy law in the United States. As described in the “Sample and Measures” section, my sample is restricted to United States rms.
11 Prior research provides some limited survey evidence that employee engagement in employee volunteer programs can be dra- matically a ected by top management’s involvement or lack of involvement in these programs (Benjamin 2001; Seagram and Sons 1989). A secondary reason for including analysis of CEO incentives in this paper is to provide additional evidence in this regard.
1 3 455 current-period employee productivity moderating the asso- ciation between pension compensation and employee volun- teer program extensiveness.
H2b Current-period employee productivity moderates the association between CEO pension compensation and the extensiveness of employee volunteer programs such that CEO pension compensation has a less positive asso – ciation with employee volunteer program extensiveness among rms with high current-period employee productiv – ity compared to rms with low current-period employee productivity.
Sample and Measures Data on a rm’s internal activities, such as their use of employee volunteer programs, can be di cult to acquire because proprietary costs for disclosing internal informa – tion is often quite high. To overcome this I collect a sample of sustainability reports from 2003 to 2013. These reports are archived publicly in the Global Reporting Initiative database (see database.globalreporting.org). The Global Reporting Initiative (hereafter GRI) is an organization that maintains a framework for sustainability reports. When a company uses the GRI framework, GRI tries to include the company’s sustainability report in the GRI database. I included these reports in my sample. In some cases, how – ever, GRI documents that a report was issued but does not have that report in its database. In these cases, GRI gener – ally supplies a link to the company’s website where the report can be found. To ensure the most complete sample possible, I followed links to company websites and col- lected additional sustainability reports not uploaded to the GRI database. 13 For consistency I restrict my sample to publicly-traded rms in the United States. From the sustainability reports examined, I collect data on 1428 firm-years from 373 rms. 14 I match these rm-years to nancial data from Com- pustat to form the nal sample. My sample selection process is summarized in Table 1. 15 The Compustat field names I use to form variables, including control variables, are described below Table 2.
Because the rms in my sample vary signicantly in size, I use the natural logarithm of all rm-level variables. I use three variables to capture employee productivity. The rst variable is SALESPROD, or the natural logarithm of sales per employee. 16 This variable is from the organizational pro- ductivity literature (e.g., Huselid 1995). To this measure, I add two additional less-common measures: GPPROD (the natural logarithm of gross prot per employee) and EARN- PROD (the natural logarithm of earnings per employee).
These two proxies do appear in some prior research but are not as common (e.g., Munday and Peel 1997). I include these measures because employee volunteer programs are not free. Increases in employee productivity, as measured by sales per employee, might be insu cient to detect any o set- ting increases in costs of goods sold or operating expenses.
By including measures of employee productivity based on gross prot and earnings, I can provide more robust tests of employee volunteer programs’ net e ect. From my sample of sustainability reports, I hand- collected any volunteer hours reported. I electronically searched each report le for the word “hours” and recorded any volunteer hours reported. 17 I summed volunteer hours Table 1 Sample selection Reports/rm-years Firms Reports in GRI database for publicly-traded United States rms 3560954 Firm-years (2003–2013) with Compustat data 1428373 Firm-years with the word “volunteer” (VOLMENTION) 1275315 Firm-years with volunteer hours > 0 (VOLPROGRAM) 776168 Firm-years reporting total volunteer hours (VOLSYSTEM) 675113 13 This portion of my search was complicated slightly by the fact that company websites are constantly updated, and I sought sustain- ability reports that might be several years old. On occasion GRI or the company website indicated a report existed, but it was not in the GRI database and the current version of the company website held no active links to it. In these cases, I searched for and occasionally found these sustainability reports on company websites using an internet search engine. 14 A small number of sustainability reports contain data for multiple rm-years.
15 Table 1 includes all three of my main proxies for employee volun- teer programs, two of which are discussed in the “ Alternative Prox- ies for Employee Volunteer Programs” subsection. I include variable denitions below Table 2.Reports in the GRI database were observed in 2016. Some GRI gures referenced in Table 1 may change over time as additional reports are released.
16 That is, ln salesemployees . See, for example, Bloom and Van Reenen (2011). 17 It was more economical to hand-collect employee volunteer pro- gram variables than to use an automated protocol. Sustainability reports contain many di erent references to “hours,” such as total reportable incidence rate (which refers repeatedly to 200,000 work- Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 456 B. D. Knox together by rm-year when multiple gures were given (e.g., I would sum volunteer hours for separate company- wide volunteer projects in the same rm-year). I then took the natural logarithm of rm-year volunteer hours to form the variable LNTOTALHOURS (I use this variable in test- ing H2a and H2b). In H1a I hypothesize that long-term employee productivity is positively associated with rms’ current-period employee volunteer programs. As my prin- cipal variable to test this hypothesis, I use a dummy vari- able, VOLPROGRAM, to capture whether the rm reports volunteer hours in a given rm-year. I code this variable as 1 if the rm-year has volunteer hours reported, and I code it as 0 if the rm-year does not have volunteer hours. This variable assumes that reporting volunteer hours meaning- fully substantiates the existence of an employee volunteer program. Results Descriptive Statistics I provide descriptive statistics in Table 2. Table 3 provides the frequency (%) and means (SDs) of key variables by year. This table includes two variables (VOLMENTION and VOLSYSTEM) that I use in supplemental analysis, as explained in the “ Alternative Proxies for Employee Vol- unteer Programs” subsection, which appears later in the “ Results” section. Table 4 shows correlations (p values) between variables. 18 Table 2 Descriptive statistics Variable definitions: LNASSETS: natural logarithm of total assets (Compustat: AT—assets total); VOLMENTION: dummy variable to indicate that report uses word “volunteer.” 1 indicates “volun- teer” used; VOLPROGRAM: dummy variable to indicate that the firm reports volunteer hours. 1 indicates that hours reported are greater than zero, VOLSYSTEM: dummy variable to indicate that firm reports “total” hours from an employee volunteer program, including total hours from a branch.
1 indicates “total” hours reported; SALESPROD: natural logarithm of SALES per NUMEMP; GPPROD: Natural logarithm of GROSSPROFIT per NUMEMP; EARNPROD: natural logarithm of EARNINGS per NUMEMP; LNSALES: natural logarithm of net sales for the firm-year (Compus- tat: SALE—sales/turnover (net)); LNGROSSPROFIT: natural logarithm of gross profit (Compustat:
GP—gross profit (loss)); LNEARNINGS: net income before extraordinary items (Compustat: IB— income before extraordinary items); NUMEMP: number of employees for the firm-year (Compus- tat: EMP—employees); LNTOTALHOURS: natural logarithm of total volunteer hours reported by the firm (if available) n Mean SDLower quartile MedianUpper quartile LNASSETS 14289.95 1.558.89 9.7810.78 VOLMENTION (indicator) 14280.89 (1275) 0.31– –– VOLPROGRAM (indicator) 14280.54 (776) 0.50– –– VOLSYSTEM (indicator) 14280.47 (675) 0.50– –– SALESPROD 14286.08 0.845.63 5.996.53 GPPROD 14235.00 0.974.47 5.085.61 EARNPROD 13223.52 1.222.75 3.664.29 LNSALES 14289.46 1.278.52 9.4110.39 LNGROSSPROFIT (non- negative) 1423 8.38 1.387.37 8.309.30 LNEARNINGS(non-neg- ative) 1322 6.94 1.485.93 6.907.94 NUMEMP (000s) 142870.44 132.0511.13 31.0073.00 Footnote 17 (continued) hours as the basis for the rate), work hours without a loss-time event, training hours, hotlines (answered “24 hours a day”), kilowatt-hours, megawatt-hours, working hours, and so forth. Error rates from an automated protocol would be high. In “ Alternative Proxies for Employee Volunteer Programs” I perform supplemental analysis that requires a determination of whether rms track total volunteer hours.
This would be di cult to measure using an automated protocol. 18 LNTOTALHOURS has notable correlations with my three prox- ies for employee volunteer programs. As discussed later, VOL – MENTION indicates whether companies’ CSR disclosures use the word “volunteer” and VOLSYSTEM indicates whether companies’ volunteer hours are presented as total hours. As shown in Table4, VOLSYSTEM has a signicant positive correlation with LNTO- TALHOURS. This may mean that implementing a volunteer hour reporting system—allowing one to arrive at “total volunteer hours”— increases the completeness of volunteer hours reported. It could also mean that such a reporting system motivates employees to volunteer more.
1 3 457 Hypothesis Tests forH1a andH1b As described in the“Sample and Measures”section, I test employee productivity using three di erent proxies, (1) the natural logarithm of sales per employee (SALESPROD), (2) the natural logarithm of gross profit per employee (GPPROD), and (3) the natural logarithm of earnings per employee (EARNPROD). In H1a, I predict that the use of an employee volunteer program in the current period is associated with improved employee productivity in future periods. In H1b, I predict that this association is moderated by current-period employee productivity.
Model 1Model 1 describes my regression equation to test H1a and H1b (using the generalized term “PROD” for my three proxies of employee productivity). In Model 1, I predict PROD FUTURE_YEAR using the predictors of PROD YEAR_0 , VOLPROGRAM YEAR_0 (H1a), and the interaction between the two (H1b). “YEAR_0” refers to the current periodfor a given rm-year. “FUTURE_YEAR” refers to a year after whatever year is YEAR_0 in a given rm-year, i.e., year 1 through year 6. By including current-period employee productivity as a predictor of future-period productivity (i.e., B 3PROD YEAR_0 ), I can control for many unobserved var – iables that drive employee productivity. My other two con- trols in the model are important as well. I use LNASSETS to PROD FUTURE_YEAR = B +B B + B 2VOLPROGRAM YEAR_0 ×PROD YEAR_0 + B 3PROD YEAR_0 +B \fLNASSETS YEAR_0 + [Ye\br Fixed E ects ]+ account for the e ect that rm size might have on employee productivity. The variable Year Fixed E ects is a set of indi- cator variables that account for baseline employee productiv – ity levels each calendar year in my sample. In Tables 5, 6, and 7 , I show results for Model 1. 19 In Table 5 I use SALESPROD as a proxy for employee produc- tivity. 20 In Table 6 I use GPPROD as a proxy for employee productivity. In Table 7 I use EARNPROD as a proxy for employee productivity. Some future-period results from each of these measures of employee productivity support H1a and H1b. For SALEPROD (Table 5) I nd support in year 1, year 3, and year 6. For GPPROD (Table 6) I nd support in year 2 and year 3. For EARNPROD (Table 7) I nd support in year 2. These results suggest that rms can experience increased employee productivity in some future periods after maintain- ing an employee volunteer program in the current period. To reiterate, I have controlled for current-period employee pro- ductivity in each of these regressions, making it more likely that employee volunteer programs, instead of an unobserved variable that is merely associated with employee volunteer programs, is responsible for this association. These results suggest that employee volunteer programs can a ect future-period employee productivity for up to six years, depending on which measure of employee produc- tivity is used. The negative coe cients for the interaction Table 3 Frequency (%) and mean (SD), by year, for proxies of employee volunteer programs and employee productivity. Totals are in the last row for rm-years and employee volunteer program frequencies. Averages are in the last row for employee productivity means Year Firm-years Frequency (%) Mean (SD) VOLPROGRAM VOLMENTIONVOLSYSTEMSALESPRODGPPRODEARNPROD 2003 2715 (0.56) 22 (0.81)10 (0.37)6.06 (0.88) 4.92 (0.96)3.41 (1.15) 2004 3320 (0.61) 29 (0.88)13 (0.39)6.01 (0.96) 5.04 (1.00)3.70 (1.09) 2005 4122 (0.54) 35 (0.85)16 (0.39)6.12 (0.86) 5.19 (0.90)3.77 (1.06) 2006 5934 (0.58) 54 (0.92)32 (0.54)6.08 (0.85) 5.09 (0.95)3.76 (1.08) 2007 7745 (0.58) 66 (0.86)38 (0.49)6.11 (0.89) 5.10 (0.98)3.56 (1.27) 2008 9756 (0.58) 80 (0.82)55 (0.57)6.10 (0.83) 5.00 (0.95)3.40 (1.24) 2009 141 82 (0.58) 130 (0.92)70 (0.50)5.97 (0.72) 4.84 (0.94)3.27 (1.21) 2010 192 100 (0.52) 171 (0.89)83 (0.43)6.04 (0.78) 4.99 (0.92)3.51 (1.16) 2011 225 104 (0.46) 203 (0.90)100 (0.44)6.11 (0.86)5.01 (0.99)3.56 (1.25) 2012 263 145 (0.55) 240 (0.91)128 (0.49)6.10 (0.84)5.01 (1.00)3.49 (1.24) 2013 273 153 (0.56) 245 (0.90)130 (0.48)6.13 (0.89)5.03 (0.97)3.57 (1.27) TOTAL/ AVER – AGE 1428 776 (0.54) 1275 (0.89)675 (0.47)6.07 (0.85)5.02 (0.96)3.55 (1.18) 19 I limit FUTURE_YEAR to year 6 to ensure a meaningful number of rms and rm-years are in my analysis.
20 Throughout the “Results” section, I use two-tailed signicance tests at the 90% condence interval. I use this relatively low con- dence interval because many factors a ect employee productivity and because I generally expect hypothesized e ects to be small. I report signicance using stricter condence intervals in the tables. I also report raw p values in parentheses in Table5 through 13. Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 458 B. D. Knox Table 4 Pearson correlations between key variables LNASSETS VOLMENTION VOLPROGRAM VOLSYSTEMSALESPROD GPPRODEARNPROD LNSALESLNGROSSPROFIT LNEARN – INGSNUMEMP VOLMENTION 0.1045 (0.0001) – VOLPROGRAM 0.1314 (< 0.0001) 0.1370 (< 0.0001) – VOLSYSTEM 0.2226 (< 0.0001) 0.0559 (0.0347) 0.6116 (< 0.0001) – SALESPROD 0.2688 (< 0.0001) 0.0513 (0.0528) −0.0728 (0.0059) −0.0096 (0.7170) – GPPROD 0.3399 (< 0.0001) 0.0416 (0.1169) −0.0235 (0.3764) 0.0259 (0.3282) 0.7801 (< 0.0001) – EARNPROD 0.3264 (< 0.0001) 0.0115 (0.6774) −0.0407 (0.1390) 0.0149 (0.5891) 0.7465 (< 0.0001) 0.8412 (< 0.0001) – LNSALES 0.7890 (< 0.0001) 0.0926 (0.0005) 0.0942 (0.0004) 0.1466 (< 0.0001) 0.1507 (< 0.0001) 0.0973 (0.0002) 0.1026 (0.0002) – LNGROSSPROFIT 0.8005 (< 0.0001) 0.0848 (0.0014) 0.1204 (< 0.0001) 0.1577 (< 0.0001) 0.0812 (0.0022) 0.3156 (< 0.0001) 0.2324 (< 0.0001) 0.8971 (< 0.0001) – LNEARNINGS 0.7822 (< 0.0001) 0.0703 (0.0106) 0.0866 (0.0016) 0.1339 (< 0.0001) 0.2130 (< 0.0001) 0.3348 (< 0.0001) 0.4929 (< 0.0001) 0.8346 (< 0.0001) 0.8939 (< 0.0001) – NUMEMP 0.3589 (< 0.0001) 0.0213 (0.4204) 0.1212 (< 0.0001) 0.1033 (0.0001) −0.2721 (< 0.0001) −0.2468 (< 0.0001) −0.2415 (< 0.0001) 0.5181 (< 0.0001) 0.4693 (< 0.0001) 0.3910 (< 0.0001) – LNTOTALHOURS 0.4642 (< 0.0001) −0.0272 (0.4479) 0.0518 (0.1484) 0.4802 (< 0.0001) −0.1156 (0.0012) −0.0755 (0.0354) −0.0653 (0.0783) 0.5173 (< 0.0001) 0.4899 (< 0.0001) 0.4451 (< 0.0001) 0.3650 (< 0.0001) 1 3 459 Table 5 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLPROGRAM YEAR_0 ) and future-period employee productivity (SALESPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 Per Cohen (1988), if is a set of predictors (the H1a and H1b predictors in this case), then f2 =R2with R 2without 1 R2with *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: SALESPROD FUTURE_YEAR SALESPROD YEAR_1 SALESPROD YEAR_2 SALESPROD YEAR_3 SALESPROD YEAR_4 SALESPROD YEAR_5 SALESPROD YEAR_6 (H1a) VOLPROGRAM YEAR_0 0.116* (0.096) 0.15 (0.245)0.470*** (0.006) 0.171 (0.427)0.023 (0.920)0.475** (0.040) (H1b) VOLPROGRAM YEAR_0 × SALESPROD YEAR_0 −0.021* (0.062) −0.026 (0.214)−0.083***(0.002) −0.037 (0.281)−0.01 (0.776)−0.083** (0.019) SALESPROD YEAR_0 1.000*** 0.936**0.934*** 0.823***0.781***0.778*** YEAR FIXED EFFECTS (χ 2) 117.65*** 111.36***73.88*** 61.13***49.46***11.65* LNASSETS YEAR_0 −0.001 −0.0010.005 0.011−0.003−0.008 Intercept 0.129*0.569***0.622*** 1.304***1.721***1.630*** Observations 957767566 419301208 Firms 242227176 14211079 R 2 0.9709 0.94360.9250 0.89860.89150.8831 R 2 without H1a and H1b predictors 0.9708 0.94320.9217 0.89490.88830.8728 f 2 for H1a and H1b predictors 0.00340.00710.0440 0.03650.02950.0909 Table 6 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLPROGRAM YEAR_0 ) and future-period employee productivity (GPPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: GPPROD FUTURE_YEAR GPPROD YEAR_1 GPPROD YEAR_2 GPPROD YEAR_3 GPPROD YEAR_4 GPPROD YEAR_5 GPPROD YEAR_6 (H1a) VOLPROGRAM YEAR_0 0.076 (0.354) 0.144* (0.092) 0.373* (0.054) 0.334 (0.197)0.332 (0.236)0.359 (0.224) (H1b) VOLPROGRAM YEAR_0 × GPPROD YEAR_0 −0.021 (0.192) −0.052* (0.066) −0.081** (0.030) −0.079 (0.112)−0.083 (0.117)−0.09 (0.102) GPPROD YEAR_0 0.979*** 0.934*** 0.942*** −0.696***0.803***0.831*** YEAR FIXED EFFECTS (χ 2) 76.94*** 77.89*** 47.72*** 33.92***17.27***3.52 LNASSETS YEAR_0 0.010 0.013 0.016 0.059*0.015−0.023 Intercept 0.1360.421** 0.416* 1.292***1.196***1.238** Observations 950761 560 415297207 Firms 241227 176 14210978 R 2 0.9440 0.8931 0.8761 0.80490.81820.8250 R 2 without H1a and H1b predic- tors 0.9437 0.8925 0.8733 0.79890.80950.8132 f 2 for H1a and H1b predictors 0.00540.0056 0.0226 0.03080.04790.0674 Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 460 B. D. Knox term in these years suggests that rms with more produc- tive employees in the current period have a less positive association between future-period employee productivity and the current-period use of an employee volunteer pro- gram. 21 While I cannot rule out the possibility that some cost of employee volunteer programs, especially a cost unre- lated to employee productivity, exceeds these programs’ employeeproductivity-related benets, this evidence of a rm-level, long-term benet does run counter to the idea that there are no indirect benets from such programs. Alternative Proxies forEmployee Volunteer Programs My sample comes from sustainability reports, meaning it reects information that rms voluntarily choose to disclose to external parties. There is a possibility that some elements of this reporting decision are correlated with employee pro- ductivity through an omitted variable. For example, it may be that rms only report volunteer hours when the number of hours constitute good news, such as when volunteer hours are higher this year than last year or when volunteer hours exceed industry benchmarks. If so, the underlying causes for this good news (which might not be captured in my above analysis) may actually drive increases in long-term employee productivity.
I attempt to address this possibility by re-running my analysis for H1a and H1b with a new variable, VOLMEN- TION, in place of the variable VOLPROGRAM. This new indicator variable is dened as 1 for any rm-year in which the word “volunteer” appears in the sustainability report and 0 when the word does not appear. It is less likely that the mere use of the word “volunteer” is correlated with omit- ted variables that drive employee productivity. In 1275 out of 1428 rm-years, the word “volunteer” is used at least once, making this variable an especially broad measure of whether the rm uses an employee volunteer program. For this supplemental analysis, I use a regression model that is comparable to Model 1, except that I replace the variable VOLPROGRAM with the variable VOLMENTION. In Tables 8, 9, and 10, I show tests of H1a and H1b using VOLMENTION to proxy for rms’ employee vol- unteer programs. I use the same proxies for employee productivity in Tables 8, 9, and 10 as in Tables 5, 6 and 7 (i.e., SALESPROD, GPPROD, and EARNPROD). I nd signicant results, supporting H1a and H1b, in sev – eral future periods: year 1, year 2, year 5, and year 6 for SALESPROD (Table 8), year 1 and year 4 for GPPROD (Table 9), and year 1 through year 6 for EARNPROD (Table 10). Interestingly, this broader measure of whether rms use an employee volunteer program generally sup- ports these hypotheses in a larger number of years. Another alternative explanation for my results, in an opposite vein, suggests that merely using the word Table 7 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLPROGRAM YEAR_0 ) and future-period employee productivity (EARNPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: EARNPROD FUTURE_YEAR EARNPROD YEAR_1 EARNPROD YEAR_2 EARNPROD YEAR_3 EARNPROD YEAR_4 EARNPROD YEAR_5 EARNPROD YEAR_6 (H1a) VOLPROGRAM YEAR_0 −0.019 (0.903) 0.325* (0.077)−0.316 (0.154)−0.239 (0.358)−0.313 (0.333)−0.315 (0.462) (H1b) VOLPROGRAM YEAR_0 × EARNPROD YEAR_0 −0.01 (0.817) −0.105** (0.028)0.063 (0.280)0.031 (0.644)0.085 (0.297)0.037 (0.732) EARNPROD YEAR_0 0.736*** 0.657***0.582***0.484***0.624***0.671*** YEAR FIXED EFFECTS (χ 2) 29.40*** 37.17***26.72***9.466.312.77 LNASSETS YEAR_0 0.072*** 0.089**0.090**0.115**0.059−0.009 Intercept 0.541*0.657*0.981**0.031*0.959*1.402* Observations 851679499375272195 Firms 22720915912910275 R 2 0.7031 0.6760.64910.59580.63180.5823 R 2 without H1a and H1b predictors 0.70210.67520.64320.58380.62720.5715 f2 for H1a and H1b predictors 0.00340.00250.01680.02970.01250.0259 21 I also re-test each regression in Tables 5, 6 and 7 using two-digit industry indicator variables to control for industry xed e ects.
When I do this, statistical support for H1a and H1b using GPPROD improves (i.e. all six future years produce a signicant e ect) and statistical support for H1a and H1b using SALESPROD and EARN- PROD remains the same.
1 3 461 Table 8 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLMENTION YEAR_0 ) and future-period employee productivity (SALESPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: SALESPROD FUTURE_YEAR SALESPROD YEAR_1 SALESPROD YEAR_2 SALESPROD YEAR_3 SALESPROD YEAR_4 SALESPROD YEAR_5 SALESPROD YEAR_6 (H1a) VOLMENTION YEAR_0 0.174* (0.081) 0.369** (0.018)0.321 (0.124)0.331 (0.169)0.566** (0.030)0.516** (0.047) (H1b) VOLMENTION YEAR_0 × SALESPROD YEAR_0 −0.031* (0.059) −0.071*** (0.005)−0.060* (0.084)−0.067* (0.086)−0.109*** (0.009)−0.083** (0.039) SALESPROD YEAR_0 1.017*** 0.985***0.941***0.868***0.856***0.771*** YEAR FIXED EFFECTS (χ 2) 118.18*** 114.13***72.56***58.29***46.45***10.35* LNASSETS YEAR_0 −0.001 0.0020.0050.011−0.001−0.004 Intercept 0.3880.392*0.592**1.059***1.281***1.604*** Observations 957767566419301208 Firms 24222717614211079 R 2 0.9709 0.94450.92320.89820.89510.8783 R 2 without H1a and H1b predictors 0.9708 0.94320.92170.89490.88830.8728 f 2 for H1a and H1b predictors 0.00340.02340.01950.03240.06480.0452 Table 9 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLMENTION YEAR_0 ) and future-period employee productivity (GPPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: GPPROD FUTURE_YEAR GPPROD YEAR_1 GPPROD YEAR_2 GPPROD YEAR_3 GPPROD YEAR_4 GPPROD YEAR_5 GPPROD YEAR_6 (H1a) VOLMENTION YEAR_0 0.292** (0.014) 0.600*** (0.001)0.502** (0.034)0.495* (0.091)0.483 (0.167)0.607 (0.117) (H1b) VOLMENTION YEAR_0 × GPPROD YEAR_0 −0.061*** (0.009) −0.13*** (< 0.001) −0.104** (0.029) −0.111* (0.055) −0.109 (0.101) −0.105 (0.143) GPPROD YEAR_0 1.021*** 1.018***0.979***0.747***0.834***0.838*** YEAR FIXED EFFECTS (χ 2) 76.96*** 80.35***49.35***32.56***16.56**4.25 LNASSETS YEAR_0 0.009 0.013−0.0160.058*0.018−0.020 Intercept −0.0740.0140.2291.049**1.011*1.054* Observations 950761560415297207 Firms 24122717614210978 R 2 0.9441 0.89440.87490.80360.81430.8137 R 2 without H1a and H1b predic- tors 0.9437 0.89250.87330.79890.80950.8132 f 2 for H1a and H1b predictors 0.00720.01800.01280.02390.02580.0027 Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 462 B. D. Knox “volunteer” (i.e., VOLMENTION) or reporting a number for volunteer hours (i.e., VOLPROGRAM) is cheap talk.
This cheap talk could reach employees and lead them to have a higher opinion of the rm, thereby improving employee productivity even without employees actually participating in the employee volunteer program (see Kim etal. 2010).
Although this alternative explanation is not dramatically dif- ferent from that discussed in the“Employee Volunteer Pro- grams and Long-term Employee Productivity”subsection, it does suggest a di erent mechanism by which employee productivity changes. In practice, this might mean employee volunteer programs can be replaced by anything that makes employees feel good about the rm, without signicant sac – rice to future-period employee productivity. To address this possibility, I again re-run the tests for H1a and H1b, this time using the variable VOLSYSTEM in place of VOL – PROGRAM. VOLSYSTEM is an indicator variable based on whether the volunteer hours reported by the rm are pre- sented as the total hours for the rm or for a branch of the rm (coded 1 for rm-years with volunteer hours reported as total hours and coded 0 for all other rm-years). The VOLSYSTEM variable is intended to measure a costly investment in an employee volunteer program because it involves a monitoring system that is comprehensive enough to report total volunteer hours. Cheap talk, by denition, is free. VOLSYSTEM measures something costly that cannot be construed as cheap talk. The cost of building in-house volunteer monitoring systems or the cost of subscriptions to such systems are signicant enough that a niche indus- try of volunteer program consulting has sprung up to satisfy demand (Jarvis 2011). Christensen and Demski (2003) argue that the value of a monitoring system is determined by “how much […] the decision maker [would] be willing to pay for the information” provided by that system (p. 114). 22 If a rm invests in a volunteer hour monitoring system, a costly endeavor, that signals management’s interest in information about employee’s actual participation in an employee volun- teer program, as opposed to management merely being inter – ested in maintaining the appearance of using an employee volunteer program (see also Basil et al. 2009). 23 If the Table 10 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLMENTION YEAR_0 ) and future-period employee productivity (EARNPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: EARNPROD FUTURE_YEAR EARNPROD YEAR_1 EARNPROD YEAR_2 EARNPROD YEAR_3 EARNPROD YEAR_4 EARNPROD YEAR_5 EARNPRO- D YEAR_6 (H1a) VOLMENTION YEAR_0 0.580** (0.013) 0.677*** (0.006) −0.659** (0.032)0.912** (0.011)0.990** (0.016)1.186* (0.060) (H1b) VOLMENTION YEAR_0 × EARNPROD YEAR_0 −0.162*** (0.008) −0.215*** (0.001) 0.196** (0.021)−0.277*** (0.004)−0.302*** (0.004)−0.287* (0.065) EARNPROD YEAR_0 0.872*** 0.806*** 0.795***0.757***0.948***0.957*** YEAR FIXED EFFECTS (χ 2) 30.09*** 40.21*** 25.81***8.496.113.31 LNASSETS YEAR_0 0.070*** 0.088*** −0.086**0.112**0.055−0.017 Intercept 0.0200.153 0.212−0.001−0.1420.196 Observations 851679 499375272195 Firms 227209 15912910275 R 2 0.7048 0.6831 0.65050.60170.64570.5803 R 2 without H1a and H1b predictors 0.7021 0.6752 0.64320.58380.62720.5715 f 2 for H1a and H1b predic- tors 0.0091 0.0249 0.02090.04490.05220.0210 22 A more extensive treatment of the value and informativeness of monitoring systems (or simply information systems) is given by Mar – schak and Miyasawa (1968). The relevant equation in their analysis is Eq. 4.6 on page 145; that equation is comparable to Christensen and Demski’s (2003) Eq.6.9, presented on page 114.
23 Table 3 shows a general pattern of monotonic increases in employee volunteer programs for each proxy of them and for every year in the sample. However, between 2008 and 2010, the percent- age of rms that reported total volunteer hours (i.e. VOLSYSTEM) decreased precipitously. This is consistent with volunteer hour report- ing systems being costly and with the recession from 2008 to 2009 a ecting how many rms were willing to incur the additional costs of a volunteer hours reporting system.
1 3 463 mechanism for employee productivity improvements is based on employees’ general feelings, I expect that VOLSYSTEM would fail to produce comparable results to VOLPROGRAM and VOLMENTION. That is, VOLSYSTEM labels a number of rm-years as 0 that the other indicator variables label as 1, rm-years that would be expected to have improved employee productivity outcomes if the mechanism is merely employees’ general feelings. 675 out of 1428 rm-years are coded as 1 for VOLSYSTEM.In Tables 11, 12, and 13, I show results from re-running H1a and H1b tests using the variable VOLSYSTEM. I again use a modied version of Model 1, with VOLSYS- TEM in place of VOLPROGRAM. I use the same proxies for employee productivity in these tables as in Tables 5, 6 , and 7 (i.e., SALESPROD, GPPROD, and EARNPROD).
I nd support for H1a and H1b in year 1 through year 4 for SALESPROD (Table 11), in year 1 through year 6 for GPPROD (Table 12), and in year 2 for EARNPROD (Table 13). This alternative proxy provides evidence of robustness to my results in Tables 5through 10. In fact, it is notable that in a few regressions the results using VOLSYSTEM are much stronger than the results using my other two proxies for employee volunteer programs.
This strengthens the argument that these results are not due to employees’ general feelings about employee volunteer programs but is driven by employees’ actual participation in such programs. Supplemental Tests ofH1a andH1b It is possible that the results reported in Tables 5through 13, which suggest a relationship between current-period employee volunteer programs and future-period employee productivity, are inuenced by the omitted variable of future- period employee volunteer programs. Firms that engage in employee volunteer programs in the current period may just be likely to have future-period employee volunteer programs as well, and these future-period employee volunteer programs could a ect future-period employee productivity. In an e ort to test this alternative explanation of my H1a and H1b results, I conduct the two following supplemental tests. First, I re-run the analyses testing H1a and H1b (as shown in Model 1) but add predictor variables for VOLPROGRAM, VOLMENTION, or VOLSYSTEM for every year intervening the current year and the future year in question. For exam- ple, assuming I use VOLPROGRAM, in this supplemental analysis the regression to test employee productivity in year 1 will be the same as shown in Model 1, while the regres- sion to test employee productivity in year 6 will also have a predictor for VOLPROGRAM in year 1, year 2, year 3, year 4, and year 5. I do this for my three main proxies of employee volunteer programs and my three main proxies for employee productivity. In the interest of avoiding multicollinearity, I only interact the current-period employee volunteer program proxy with current-period employee productivity and omit employee productivity observed in intervening years. Table 11 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLSYSTEM YEAR_0 ) and future-period employee productivity (SALESPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: SALESPROD FUTURE_YEAR SALESPROD YEAR_1 SALESPROD YEAR_2 SALESPROD YEAR_3 SALESPROD YEAR_4 SALESPROD YEAR_5 SALESPROD YEAR_6 (H1a) VOLSYSTEM YEAR_0 0.159** (0.021) 0.280** (0.028) 0.477*** (0.004) 0.633*** (0.002)0.322 (0.144)0.379 (0.109) (H1b) VOLSYSTEM YEAR_0 × SALESPROD YEAR_0 −0.027** (0.017) −0.047** (0.024) −0.081*** (0.003) −0.109*** (0.001)−0.056 (0.113)−0.070* (0.063) SALESPROD YEAR_0 1.003*** 0.945*** 0.926*** 0.868***0.799***0.755*** YEAR FIXED EFFECTS (χ 2) 114.86*** 109.26*** 69.72*** 56.86***49.02***10.07* LNASSETS YEAR_0 −0.001 0.001 0.006 0.011−0.002−0.005 Intercept 0.1050.537*** 0.658*** 1.022***1.589***1.740*** Observations 957767 566 419301208 Firms 242227 176 14211079 R 2 0.9710 0.9435 0.9234 0.89960.89170.8811 R 2 without H1a and H1b predictors 0.9708 0.9432 0.9217 0.89490.88830.8728 f 2 for H1a and H1b predictors 0.00690.0053 0.0222 0.04680.03140.0698 Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 464 B. D. Knox If my main tests of H1a and H1b are signicant because of employee volunteer program usage during the intervening years between the current period and the future period, this regression setup should account for it and current-period use of an employee volunteer program should be largely non-signicant. However, when I run this additional analy – sis, I nd that 30 of the 54 regressions support both H1a and H1b (compared to 31 regressions in my main tests of H1a and H1b). In terms of future year and proxies used, 27 of these 30 supporting regressions are the same as my supporting regressions reported in Tables 5 through 13.
Overall this suggests there can be some incremental e ect of current-period use of an employee volunteer program on employee productivity in future years. Of these 30 Table 12 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLSYSTEM YEAR_0 ) and future-period employee productivity (GPPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: GPPROD FUTURE_YEAR GPPROD YEAR_1 GPPROD YEAR_2 GPPROD YEAR_3 GPPROD YEAR_4 GPPROD YEAR_5 GPPROD YEAR_6 (H1a) VOLSYSTEM YEAR_0 0.169** (0.035) 0.410*** (0.004)0.596*** (0.001) 1.25*** (< 0.001) 0.825** (0.002)0.527* (0.050) (H1b) VOLSYSTEM YEAR_0 × GPPROD YEAR_0 −0.037** (0.020) −0.086*** (0.002) −0.124*** (0.001) −0.26*** (< 0.001) −0.18*** (< 0.001)−0.129** (0.012) GPPROD YEAR_0 0.987*** 0.948***0.958*** 0.804*** 0.874***0.864*** YEAR FIXED EFFECTS (χ 2) 75.10*** 77.07***47.50*** 35.43*** 20.66***2.28 LNASSETS YEAR_0 0.010 0.0140.018 0.059* 0.014−0.024 Intercept 0.0800.330*0.320 0.750* 0.829*1.079** Observations 950761560 415 297207 Firms 241227176 142 10978 R 2 0.9440 0.89370.8767 0.8083 0.82510.8306 R 2 without H1a and H1b predictors 0.94370.89250.8733 0.7989 0.80950.8132 f2 for H1a and H1b predictors 0.00540.01130.0276 0.0490 0.08920.1027 Table 13 Regression analysis (see Model 1) of the association between employee volunteer programs (VOLSYSTEM YEAR_0 ) and future-period employee productivity (EARNPROD FUTURE_YEAR ) Variable denitions are the same as in Table2 f2 denition is the same as in Table25 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Dependent variable: EARNPROD FUTURE_YEAR EARNPROD YEAR_1 EARNPROD YEAR_2 EARNPROD YEAR_3 EARNPROD YEAR_4 EARNPROD YEAR_5 EARNPROD YEAR_6 (H1a) VOLSYSTEM YEAR_0 0.172 (0.280) 0.393** (0.032)−0.12 (0.588) 0.302 (0.242)−0.34 (0.282)−0.076 (0.858) (H1b) VOLSYSTEM YEAR_0 × EARNPROD YEAR_0 −0.062 (0.145) −0.134*** (0.006)0.032 (0.582) −0.101 (0.138)0.077 (0.339)−0.011 (0.920) EARNPROD YEAR_0 0.753*** 0.660***0.597*** 0.535***0.624***0.693*** YEAR FIXED EFFECTS (χ 2) 28.18*** 37.22***26.01*** 9.365.532.69 LNASSETS YEAR_0 0.076*** 0.097***0.089** 0.118**0.063−0.004 Intercept 0.4220.578*0.875** 0.7080.919*1.207* Observations 851679499 375272195 Firms 227209159 12910275 R 2 0.7021 0.67380.6441 0.58090.63190.5767 R 2 without H1a and H1b predictors 0.7021 0.67520.6432 0.58380.62720.5715 f 2 for H1a and H1b predictors 0.0000−0.00430.0025 −0.00690.01280.0123 1 3 465 signicant regressions, ve suggest an incremental e ect to an employee productivity proxy in year 5 or year 6, which are regressions where I would especially expect intervening years to render the current-period predictors non-signicant.Second, it is possible that the number of years a rm uses an employee volunteer program has a xed e ect. To test this I introduce a set of indicator variables that measure how many years prior to the current year the rm reports using an employee volunteer program (per whichever proxy the regression is using). This set of indicator variables measures the xed e ects of using an employee volunteer program for one, two, three, etc. prior years. If the previously-omitted variable of prior years using an employee volunteer pro- gram leads to improvements in employee productivity, then the H1a and H1b predictors should lose signicance when I add this set of indicator variables. With these xed e ects indicator variables, however,I nd more rm-years to be signicant than in my main tests of H1a and H1b reported in Tables 5through 13. 24 I also conduct a supplemental test of H1a and H1b using a new proxy for employee volunteer programs called UPLN- TOTALHOURS. This test is important because my main proxies for employee volunteer programs, especially VOL – MENTION, suggest that most rms in the sample had an employee volunteer program of some kind. This potentially limits how I can interpret results supporting H1a and H1b.
This additional test examines a logical extension of H1a and H1b. If employee volunteer programs have positive e ects on future employee productivity, they should have more positive e ects when volunteer hours increase from year to year. I code UPLNTOTALHOURS as 1 for all rm-years in which LNTOTALHOURS is greater in the current year than for the year before. I nd some signicant results, supporting H1a and H1b, using all three proxies for future-period employee productivity. I nd signicant results using SALESPROD for year 1 through year 6, using GPPROD for year 4 and year 5, and using EARNPROD for year 3 through year 5. 25 These results support the underlying logic of H1a and H1b because they suggest incremental improvements in future- period employee productivity are associated with incremental increases in current-period employee use of employee vol – unteer programs. Hypothesis Tests forH2a andH2b To test H2a and H2b, I use a regression model with LNTO- TALHOURS as the dependent variable, di erent forms of CEO compensation as independent variables, as well as sev – eral control variables. Below I present Model 2,which I use to test H2a and H2b. I again use the generic term “PROD” to represent my three proxies for employee productivity. All variables in this model are in the current year for each rm- year observation.
Model 2 Some of the variables in Model 2 are derived from the Excucomp database of CEO compensation (see variable descriptions below Table 14). In Table14 I present descrip- tive statistics for variables new to Model 2 (the remaining variables in Model 2 are dened below Table 2). In Table 15 I show correlations between all variables in Model 2. In Table 16 I present results for Model 2. 26 The coe cient for LNPENSION is signicant and in the hypothesized direction, supporting H2a for all three proxies of current- period employee productivity. Support for H2a suggests that pension incentives, which lead CEOs to value long- term rm outcomes relatively more, are associated with more extensive employee volunteer programs in the current period. Also, for two of my three measures of current-period employee productivity, I nd a signicant negative inter – action between employee productivity and LNPENSION, suggesting that the association predicted in H2a is weaker when current-period employee productivity is higher. This nding supports H2b for those two measures. LNTOTALHOURS =B 0+B B +B 2LNPENSION ×PROD + B 3PROD +B 4LNCURRENT + B \fLNSTOCKOPTIONS + B 6LNTOTALCOMP + B \bLNEARNINGS +B 8LNASSETS +B 9TENURE +B 10AGE + [ ]+ 24 In my initial tests of H1a and H1b, reported in Table 5through 13, 31 of the 54 regressions I run support both H1a and H1b (at the 90% level of condence, two-tailed). When I re-run these regressions with the xed e ects indicator variables described above, seven additional years provide signicant results that support H1a and H1b (again at the 90% level of condence, two-tailed). Using VOLPROGRAM, year 4 and year 5 with the proxy GPPROD are now signicant. Using VOLMENTION, year 5 and year 6 with the proxy GPPROD are now signicant. Using VOLSYSTEM, year 5 and year 6 with the proxy SALESPROD are now signicant, as is year 4 with the proxy EARN- PROD. Generally similar results obtain when I lower the condence level to 95% or 99%.
25 If I raise the condence level to 95%, three of these years are no longer signicant: year 3 for SALESEMPROD and year 4 for GPPROD and EARNPROD. 26 I also re-run my Model 2 tests with added two-digit industry indicator variables to control for industry xed e ects. Results are unchanged in direction and signicance.
Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 466 B. D. Knox These signicant ndings may have little practical sig- nicance, given their relatively small f2 statistics (average for the three regressions is 0.0044). I interpret this to mean that there is an association between CEO incentives and employee volunteer program extensiveness but this asso- ciation may be very limited because CEOs lack direct line authority over many employee volunteer program decisions.
This interpretation supports the general idea that CEOs focused on the long-term rm outcomes act in anticipation of long-term benets from employee volunteer programs, but it does not support the idea that CEO actions generally have dramatic eects on employee volunteerism (see Ben- jamin 2001; Seagram and Sons 1989).
For all three regressions in Table2 16, current compensa- tion (i.e., salary and bonus) has a signicant negative asso- ciation with the extensiveness of an employee volunteer pro – gram. This is consistent with employee volunteer program costs being primarily in the current period with benets not arriving until later periods. It suggests, tentatively, that CEOs who have higher current-period compensation may be less motivated to risk current-period nancial performance for long-term improvements to employee productivity. Discussion Summary ofKey Findings In this study, I examine whether use of an employee vol- unteer program in the current period is associated with rm-level employee productivity in future periods. This association a ects managers’ ethical dilemma of whether to implement such programs. My study complements prior studies that use market reactions as a proxy for indirect CSR benets, and provides rm-level empirical evidence to extend prior individual-level research on employee volun- teer program benets. My study also covers a lot of ground.
In the“Results”section, I present results from more than 200 regressions due to the many proxies I test and due to me testing these proxies several years into the future. It is worthwhile for me to reiterate the totality of these results to emphasize the combined weight of the evidence they provide.
As hypothesized (i.e., H1a), I nd a set of robust results supporting the notion that current-period use of an employee volunteer program can lead to improved future-period employee productivity. This nding persists in at least some future periods for each combination of proxies for current- period employee volunteer programs (i.e., VOLPROGRAM, VOLMENTION, and VOLSYSTEM) and measures of future-period employee productivity (i.e., SALESPROD, GPPROD, and EARNPROD). I also nd similar support when I use two di erent proxies to control for e ects from intervening years’ employee volunteer programs and when I use a proxy based on year-to-year increases in volunteer hours (i.e., UPLNTOTALHOURS). Seemingly regardless of how I dene the constructs, I consistently nd evidence that current-period employee volunteer programs have an incremental e ect on future-period employee productivity. In all of these tests, I control for current-period employee productivity, rm size, and year xed e ects. My proxies for the use of an employee volunteer program provide evi- dence contrary to potential alternative explanations (i.e., that an omitted variable of good news mediates the association or that rms are merely benetting from cheap talk about volunteerism rather than actually being interested in it). In support of H1b, nearly each regression that provides support for H1a also shows that the H1a association is less positive among rms with more current-period productivity, sug- gesting important limits to the improving e ect of employee volunteer programs on employee productivity.
Table 14 Descriptive statistics for Model 2 variables Variable denitions: LNPENSION: natural logarithm of CEO pension compensation (Execucomp: PEN- SION_VALUE_TOT); LNCURRENT: natural logarithm of salary and bonus (Execucomp: TOTAL_ CURR); LNSTOPCKOPTIONS: natural logarithm of CEO stock options compensation reported by the company (Execucomp: OPTIONS_AWARD_FV) (or, if not available for a rm-year, OPTION_ AWARDS_RPT_VALUE); LNTOTALCOMP: natural logarithm of total CEO compensation (Execucomp:
TDC1); LNASSETS: natural logarithm of total assets (Compustat: AT – Assets – Total); SALESPROD: nat- ural logarithm of SALES per NUMEMP; GPPROD: natural logarithm of GROSSPROFIT per NUMEMP; EARNPROD: natural logarithm of EARNINGS per NUMEMP; TENURE: number of years as CEO, including year appointed (derived from start date) (Execucomp: DATE_BECAME_CEO); AGE: age of CEO in the rm-year (Execucomp: AGE) n Mean SDLower quartile MedianUpper quartile LNPENSION 14285.144.420 6.969.31 LNSTOCKOPTIONS 14285.283.740 7.348.16 LNCURRENT 14287.091.306.87 7.057.31 LNTOTALCOMP 14288.991.338.70 9.139.54 TENURE 14266.354.873 58 AGE 142855.78 7.4452 5660 1 3 467 Table 15 Pearson correlations for Model 2 variables Variable denitions are the same as in Tables2 and 14 LNPENSION LNCURRENTLNSTOCKOPTIONS LNTOTALCOMPLNASSETSSALESPROD GPPRODEARNPROD TENUREAGE LNPENSION 0.0725 (0.0061) – LNCUR – RENT 0.0452 (0.0881) 0.1238 (< 0.0001) – LNSTOCK – OPTIONS 0.1196 (< 0.0001) 0.7555 (< 0.0001) 0.2247 (< 0.0001) – LNTOTAL – COMP 0.1597 (< 0.0001) 0.2044 (< 0.0001) −0.0497 (0.0604) 0.2618 (< 0.0001)– LNASSETS 0.0864 (0.0011) 0.0700 (0.0081) −0.0572 (0.0308) 0.0500 (0.0591)0.2688 (< 0.0001) – SALESPROD 0.0532 (0.0446) 0.0533 (0.0445) −0.0315 (0.2356) 0.0630 (0.0175)0.3399 (< 0.0001) 0.7801 (< 0.0001) – GPPROD 0.0393 (0.1532) 0.0596 (0.0302) −0.0420 (0.1267) 0.0614 (0.0256)0.3264 (< 0.0001) 0.7465 (< 0.0001) 0.8412 (< 0.0001) – EARNPROD 0.0723 (0.0063) −0.1993 (< 0.0001) −0.0657 (0.0130) −0.1980 (< 0.0001) −0.0785 (0.0030) −0.0401 (0.1304) −0.0230 (0.3870) −0.0371 (0.1776) – TENURE 0.1534 (< 0.0001) −0.0461 (0.0815) 0.0052 (0.8434) −0.0108 (0.6831)0.0402 (0.1291) 0.0014 (0.9582) −0.0390 (0.1419) 0.0079 (0.7742) 0.2793 (< 0.0001) – AGE 0.1536 (< 0.0001) 0.0752 (0.0357) −0.0555 (0.1212) 0.1370 (0.0001)0.4642 (< 0.0001) −0.1156 (0.0012) −0.0755 (0.0354) −0.0653 (0.0783) −0.0180 (0.6154) 0.0500 (0.1628) Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 468 B. D. Knox Then—as a corroborating test of my ndings related to H1a and H1b—I test CEOs’ actions to see if they are consistent with employee volunteer programs having this type of e ect. As hypothesized (i.e., H2a), I nd that when CEOs have more pension incentives, which focus their attention on long-term rm outcomes, their rms have more extensive employee volunteer programs, which means programs with more volunteer hours. In accord- ance with H2b and as would be expected if CEOs are truly acting in anticipation of long-term benets from employee volunteer programs, I nd evidence from two of my three measures of employee productivity that the H2a incen- tive e ect is also moderated by current-period employee productivity. CEOs whose rms have high current-period employee productivity have a less positive association between CEO pension incentives and volunteer hours. E ect Size and f2 Statistics But are these results practically significant? In Tables 5through 13, I report R 2 both with and without the two predictors that test H1a and H1b. I also report f2 statistics, which are a measure of eect size for multi- ple linear regressions (Cohen 1988; see equation below my Table2 5). 27 Cohen suggests that for small effects f2= 0.02 , for medium eects f2= 0.15 , and for large e5ects f2=0.35 . The average f2 for my H1a and H1b variables across all regressions in Tables2 52through 13 is 0.0263, suggesting the eect size of H1a and H1b is small.
When I exclude regressions in which H1a and H1b are not supported, average f2 is slightly higher, 0.0291, but still obviously a small eect size. The greatest f2 for a regres- sion in these tables is 0.1027, shown in the last column of Table2 12. Although this f2 statistic approaches Cohen’s guideline for a medium eect size, this seems to be largely isolated to that regression and to the pair of proxies reported in that table (i.e., VOLSYSTEM and GPPROD). Table 16 Regression analysis (see Model 2) of the association between CEO compensation and total volunteer hours Variable denitions are the same as in Tables2 and 14 *Signicant at p = 0.10, two-tailed **Signicant at p = 0.05, two-tailed ***Signicant at p = 0.01, two-tailed Predicted sign Proxy for employee productivity SALESPRODGPPRODEARNPROD LNPENSION +0.222**0.186**0.087* LNPENSION × SALESPROD −−0.030*–– LNPENSION × GPPROD −–−0.029*– LNPENSION × EARNPROD −––−0.013 SALESPROD −0.373**–− GPPROD –−0.411***– EARNPROD ––−0.532*** LNCURRENT −0.115**−0.125**−0.106** LNSTOCKOPTIONS −0.0090.011−0.014 LNTOTALCOMP 0.118**0.129**0.101* LNEARNINGS 0.1170.183**0.598*** LNASSETS 0.557***0.534***0.280*** TENURE 0.0150.0170.162 AGE −0.012*−0.012*−0.122* YEAR (xed e ects: χ 2) 54.70***52.44***52.68*** Intercept 4.462***4.048***3.735*** Observations 726726726 Firms 222222222 R 2 0.2519 0.28010.3139 R 2 without H2a and H2b predictors 0.25260.27700.3071 f2 for H2a and H2b predictors −0.00090.00430.0099 27 The negative f 2 statistics in Table 13 and 16 are most likely due to numerical artifacts, such as rounding. Given that the R 2 statistics reported are a composite of both within-rm and between-rm vari- ation (due to my data’s cross-sectional and longitudinal nature). This leads to the possibility of small negative, and erroneous, f 2 statistics due to these composite R 2 statistics.
1 3 469 Only three regressions have f2 statistics greater than 0.07 and two of them are reported in that table. It bears pointing out that many things shape employee productivity. In a study of this kind where I examine rm- level proxies, small eect sizes would be considered the norm and can indicate my control variables are working as intended. A small eect size can be practically signicant when one is considering rms with thousands of employ – ees and billions of dollars in revenue or expenses. In fact, the f2 statistics I nd are better than I might have expected before the fact. To me it is indeed remarkable that this one CSR activity, among a range of factors that could inuence employee productivity, has a discernible eect with an aver – age f2 statistic of 0.0291. Implications fortheEthical Dilemma ofCSR Activities The obvious ethical implication from my ndings is that employee volunteer programs might present an ethical dilemma of relatively little magnitude or none at all: I nd evidence that these programs benet the rm as well as soci- ety. Beyond that, the ethical implications of my ndings are more complex. For one thing, it is unclear how di erent CSR activities compare in benetting society. If employee volun- teer programs do a relatively poor job of benetting society compared to other CSR activities, then growth in employee volunteer programs might suggest managers are less inter – ested in the ethical imperatives of social contracts compared to the ethical imperatives of employment contracts. Ethical dilemmas surrounding CSR activities are made more com- plex with the prospect that CSR activities can help mitigate adverse selection-related ethical dilemmas. Adverse selec- tion describes situations in which buyers and sellers have di erent information, leaving room for dishonest or other – wise unethical representation by sellers so as to maximize their personal gain. Beaudoin etal. (2018) nd evidence that CSR activities can help mitigate at least one type of adverse selection dilemma. In this paper, I generally assume managers’ interests align with the rm’s, and accordingly I nd CEO incentives to have an e ect on employee volunteer program extensiveness. Given that CEO incentives are also tied up in adverse selection, future research could consider employee volunteer programs, specically, as a CSR activity that might uniquely a ect adverse selection-related ethical dilemmas. H1b suggests that firms with high current-period employee productivity have a less positive association between current-period employee volunteer programs and future-period employee productivity. First, this means an ethical dilemma of CSR activities, as it pertains to employee volunteer programs, is more likely among rms with high current-period employee productivity. Second, there is no signicant correlation between the two variables interacting.
I test this using logistical regressions with my employee pro- ductivity proxies predicting my employee volunteer program proxies (all p > 0.20, two-tailed). This suggests that rms with high current-period employee productivity might not signicantly decrease their use of employee volunteer pro- grams, even though my results suggest they derive signi- cantly less benet in future-period employee productivity.
Future research may continue this examination of the ethical dilemma among these rms in particular, to determine if they seek a di erent rm benet from employee volunteer programs or if these managers are simply choosing to do what their social contracts dictate at the expense of their employment contracts. Signaling andReverse Causality Lys etal. (2015) suggest that managers use CSR activities as a signal to those outside the rm. Greater investment in CSR activities in the current period, they suggest, signals a manager’s expectation that the rm has future periods of economic plenty ahead of it. Thus, alternative to my hypotheses, it could be argued that current-period use of an employee volunteer program tracks managers’ knowledge of future operational outcomes, which knowledge is simul- taneously correlated with future-period outcomes such as future-period employee productivity. This omitted variable, i.e., managers’ knowledge of the future, might explain the association observed in my tests of H1a and H1b as well as the association observed in my tests of H2a and H2b. If this is the case, the causal arrows point in the opposite direction compared to what I hypothesized, since future-period results (or managers’ current-period knowledge that leads them to expect these future-period results) would then drive current- period employee volunteer program use. I cannot completely rule out this alternative explanation, in part because I do not directly observe this omitted vari- able. In the case of employee volunteer programs, however, there are multiple reasons why I expect at least some of the causal arrows to ow in the direction I have hypoth- esized in this paper. First, prior research provides strong indications that, on an individual level, signaling to external parties about future-period expectations is not the primary intention of those directly involved with employee volun- teer programs (e.g., Rodell 2013; Jones 2010; Booth etal.
2009; see also De Roeck and Maon 2016). If signaling is occurring, it might be occurring at a level or two higher up the hierarchy of the rm, but I view it as unlikely that this signaling activity completely subsume the intentions of the lower-level and mid-level managers who are otherwise not thinking about signaling. Prior research on individual-level volunteerism strongly supports at least partial causality in the direction of current-period employee volunteer programs Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 470 B. D. Knox causing future-period results. Also, my H2a and H2b tests show very small e ect sizes, which suggests higher levels of the rm may have little direct control over employee volun- teer programs. Even if such top managers wanted to signal external parties through employee volunteer programs, this result suggests they might be prohibitively limited in their ability to do so.Second, the timing of this reverse causality can be dif- cult to work out, depending on the rm. Sustainability reports are not required disclosures in the United States, where rms in my sample come from. The timing of these reports is not dictated by any regulatory body and often lags far behind the current-period use of an employee vol- unteer program. Most reports come out somewhat later than nancial reports and a substantial number are delayed up to two or three years after the period in question. 28 Given this timing, I would expect managers who are trying to signal external shareholders would nd it less worthwhile to signal their expectations of operational outcomes that are likely to have already passed by the time the sustainability report is released. In short, I would expect most managers trying to signal future expectations would focus on signaling their expectations of more than one or two years into the future because these years are those most likely to still be in the future when the report is released. Yet, among my results in Tables 5 through 13, 14 of the 31 regressions thatsup- port H1a and H1b are for year 1 and year 2, which is when I might otherwise expect this signaling to be least useful for the average manager. Third, I perform statistical tests on the association between current-period employee volunteer programs and raw company operational results. I nd this association to be relatively weak compared to my main tests of H1a and H1b, which instead look at the association between employee volunteer programs and employee productivity.
I conduct this supplemental test by modifying Model 1 to predict LNSALES, LNGROSSPROFIT, and LNEARNINGS in place of a future-period employee productivity proxy as the dependent variable. If managers do use employee vol- unteer programs to signal their expectations of improved future operational results, I would expect the proxies for current-period employee volunteer programs to be strongly associated with these future-period operational results. For LNSALES, at the 90% condence level (two-tailed), I nd that 13 of the 54 regressions return signicant results com- parable to my tests of H1a and H1b. For LNGROSSPROFIT the number of signicant regressions is 15 out of the 54 regressions. For LNEARNINGS it is 8 out of the 54 regressions. These results suggest there may be an associa- tion between current-period use of an employee volunteer program and future-period operational outcomes, but this association is notably weaker than the association I examine in my main tests of H1a and H1b. I interpret this as support- ing my view that there is at least some level of causality that runs in the direction of current-period employee volun- teer programs leading to improved future-period employee productivity. Potential Future Data Re nements Although my analysis provides a certain degree of compel- ling evidence supporting my hypotheses and supporting my explanation of these hypotheses, it would be premature to say that I have completely ruled out alternative explana- tions for why employee volunteer programs are associated with employee productivity. Part of this is due to my reli- ance on measures of employee volunteer programs that originate from external sustainability reports and that only coarsely measure rms’ employee volunteer programs. It can be di cult to gather internal rm data, precluding me from more direct measures of rms’ employee volun- teer programs. Through my various analyses, I attempt to overcome the shortcomings of my data and provide robust and informative results. However, a useful future research direction could be to use more rened data.
For example, not all employee volunteer programs are the same. Henning and Jones (2013, pp. 112–113) describe four common approaches to employee volunteer programs (see also Booth etal. 2009). The large-group approach utilizes large-scale, often company-wide, volunteerism projects. The grant-based approach entails the company providing a grant to volunteer organizations at which employees volunteer. The own-time approach encourages or facilitates employees spending time outside of work to volunteer. The paid-volunteerism approach is when companies provide employees with nancial incentives, e.g., paid-time o , to volunteer. Each of these approaches could have di erent e ects inside and outside the rm.
Comparing employee productivity with data on these or other employee volunteer program characteristics could be a fruitful direction for future research that issubstan- tially useful to practitioners. Future research could also overcome data limitations by comparing rm-level results with internal data about the cost of employee volunteer programs. A field experiment could be used to track inter-departmental or inter-site cost di erences upon the adoption of an employee volunteer program. Such a study would provide more direct evidence concerning the causal relationship between employee volunteer programs and employee productivity, including the role that the cost of such programs plays. 28 Data in my sample are coded to the year in which the volunteer – ing activity takes place rather than the year in which the report comes out.
1 3 471 Acknowledgements I am indebted to Chan Li and Dhinu Srinivasan, whose archival accounting research seminar at University of Pittsburgh helped provide impetus and background for this study. I am grateful for many helpful comments, including comments from participants at an October 2014 University of Pittsburgh brownbag, comments from the discussants (Elizabeth Almer, Joleen Kremin, Jim Cannon, Andrew Glover, and Scott Jackson) and participants at the 2015 Brigham Young University Accounting Research Symposium, and comments from the discussant (Nathan Stuart) and participants at the 2016 Manage – ment Accounting Section mid-year meeting. I also acknowledge, with great appreciation, signicant and invaluable e ort from anonymous reviewers.
Funding This article received no funding.
Data Availability Data available from the public sources described in this paper.
Compliance with Ethical Standards Conflict of interest The author declares that he has no conict of inter – est.
Ethical Approval This article does not contain any studies with human participants performed by the author.
References Allen, K. (2012). What is an ethical dilemma. New Social Worker, 19, 4–5.
Barnett, T., Bass, K., & Brown, G. (1994). Ethical ideology and ethical judgment regarding ethical issues in business. Journal of Business Ethics, 13(6), 469–480.
Basil, D. Z., Runte, M. S., Easwaramoorthy, M., & Barr, C. (2009). Company support for employee volunteering: A national sur – vey of companies in Canada. Journal of Business Ethics, 85 (2), 387–398.
Beaudoin, C. A., Cianci, A. M., Hannah, S. T., & Tsakumis, G. T. (2018). Bolstering managers’ resistance to temptation via the rm’s commitment to corporate social responsibility. Journal of Business Ethics. https ://doi.org/10.1007/s1055 1-018-3789-2.
Benjamin, E. J. (2001). A look inside corporate employee volunteer programs. Journal of Volunteer Administration, 19(2), 16–32.
Bhatti, K. K., & Qureshi, T. M. (2007). Impact of employee partici- pation on job satisfaction, employee commitment and employee productivity. International Review of Business Research Papers, 3 (2), 54–68.
Bloom, N., & Van Reenen, J. (2011). Human resource management and productivity. Handbook of Labor Economics, 4, 1697–1767.
Booth, J. E., Park, K. W., & Glomb, T. M. (2009). Employer-sup- ported volunteering benets: Gift exchange among employers, employees, and volunteer organizations. Human Resource Man- agement, 48(2), 227.
Brooks, A. C. (2009). Why Giving Matters. Speech presented at BYU forum address, Provo, Utah. Retrieved August 25, 2017, from https ://speec hes.byu.edu/talks /arthu r-c-brook s_givin g-matte rs-2/.
Cassell, C. A., Huang, S.X., Sanchez, J. M., & Stuart, M. D. (2012). Seeking safety: The relation between CEO inside debt holdings and the riskiness of rm investment and nancial policies. Jour – nal of Financial Economics, 103(3), 588–610.
Caudron, S. (1994). Volunteer e orts o er low-cost training options. Personnel Journal, 73(6), 38–44. Christensen, J. A., & Demski, J. S. (2003). Accounting theory: An informational content perspective. Boston: McGraw-Hill/Irwin.
Cohen, J. (1988). Statistical power analysis for the behavioral sci- ences (2nd edn). Mahwah, NJ: Lawrence Erlbaum Associates.
De Gilder, D., Schuyt, T. N., & Breedijk, M. (2005). E ects of an employee volunteering program on the work force: The ABN- AMRO case. Journal of Business Ethics, 61(2), 143–152.
De Roeck, K., & Maon, F. (2016). Building the theoretical puzzle of employees’ reactions to corporate social responsibility: An integrative conceptual framework and research agenda. Journal of Business Ethics, 1–17.
Dearden, L., Reed, H., & Van Reenen, J. (2006). The impact of train- ing on productivity and wages: Evidence from British panel data.
Oxford Bulletin of Economics and Statistics, 68(4), 397–421.
Dhaliwal, D. S., Li, O. Z., Tsang, A., & Yang, Y. G. (2011). Vol- untary nonnancial disclosure and the cost of equity capital:
The initiation of corporate social responsibility reporting. The Accounting Review, 86(1), 59–100.
Dhaliwal, D. S., Radhakrishnan, S., Tsang, A., & Yang, Y. G. (2012). Nonnancial disclosure and analyst forecast accuracy: Interna- tional evidence on corporate social responsibility disclosure.
The Accounting Review, 87(3), 723–759.
Dunfee, T. W., & Donaldson, T. (1999). Social contract approaches to business ethics: Bridging the “Is-Ought” gap. In R. E. Freder – ick (ed.), A Companion to Business Ethics, (pp. 38–55). Hobo- ken, NJ: Wiley-Blackwell.
Elliott, W. B., Jackson, K. E., Peecher, M. E., & White, B. J. (2013). The unintended e ect of corporate social responsibility per – formance on investors’ estimates of fundamental value. The Accounting Review, 89(1), 275–302.
Fehr, E., & Gächter, S. (2000). Fairness and retaliation: The econom- ics of reciprocity. The Journal of Economic Perspectives, 14(3), 159–181.
Henning, J. B., & Jones, D. A. (2013). Volunteer programs in the corporate world. In Olson-Buchanan, J. B., Bryan, L. L. K., & Thompson, L. F. (Eds.), Using industrial-organizational psychol- ogy for the greater good: Helping those who help others. New York: Routledge, 110–147.
Heskett, J. L., Sasser, W. E., & Schlesinger, L. A. (1997). Service prot chain: How leading companies link prot and growth to loyalty, satisfaction, and value. New York: Free Press.
Holzer, H. J. (1990). The determinants of employee productivity and earnings. Industrial Relations: A Journal of Economy and Society, 29(3), 403–422.
Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate nancial perfor – mance. Academy of Management Journal, 38(3), 635–672.
Jarvis, C. (2011). Hiring a corporate volunteering consultant? Here’s what to expect. Retrieved June 5, 2017, from http://www.reali zedwo rth.com/2011/07/hirin g-corpo rate-volun teeri ng.html.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the rm: Mana- gerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305–360.
Jones, D. A. (2010). Does serving the community also serve the com- pany? Using organizational identication and social exchange theories to understand employee responses to a volunteerism programme. Journal of Occupational and Organizational Psy – chology, 83(4), 857–878.
Jones, D. A. (2016). Widely assumed but thinly tested: Do employee volunteers’ self-reported skill improvements reect the nature of their volunteering experiences? Frontiers in Psychology, 7, 495.
Jones, D. A., Willness, C. R., & Madey, S. (2014). Why are job seekers attracted by corporate social performance? Experimental and eld tests of three signal-based mechanisms. Academy of Management Journal, 57(2), 383–404.
Employee Volunteer Programs are Associated with Firm-Level Benets and CEO Incentives: Dat 1 3 472 B. D. Knox Kim, H. R., Lee, M., Lee, H. T., & Kim, N. M. (2010). Corporate social responsibility and employee–company identication. Journal of Business Ethics, 95(4), 557–569.
Kim, Y., Park, M. S., & Wier, B. (2012). Is earnings quality associ- ated with corporate social responsibility? The Accounting Review, 87(3), 761–796.
Koys, D. J. (2001). The e ects of employee satisfaction, organiza- tional citizenship behavior, and turnover on organizational e ec- tiveness: A unit-level, longitudinal study. Personnel Psychology, 54(1), 101–114.
Kuang, X., & Moser, D. V. (2009). Reciprocity and the e ectiveness of optimal agency contracts. The Accounting Review, 84(5), 1671–1694.
Lys, T., Naughton, J. P., & Wang, C. (2015). Signaling through corpo – rate accountability reporting. Journal of Accounting and Econom- ics, 60(1), 56–72.
Marschak, J., & Miyasawa, K. (1968). Economic comparability of information systems. International Economic Review, 9(2), 137–174.
Martin, P. (2014). Corporate social responsibility: Three experimental studies. Dissertation. University of Pittsburgh, 2014.
Martin, P. R., & Moser, D. V. (2016). Managers’ green investment disclosures and investors’ reaction. Journal of Accounting and Economics, 61(1), 239–254.
McCabe, D. L., Dukerich, J. M., & Dutton, J. E. (1994). The e ects of professional education on values and the resolution of ethical dilemmas: Business school vs. law school students. Journal of Business Ethics, 13(9), 693–700.
McMahon, J. M., & Good, D. J. (2016). The moral metacognition scale: Development and validation. Ethics & Behavior, 26(5), 357–394.
Munday, M., & Peel, M. J. (1997). The Japanese manufacturing sector in the UK: A performance appraisal. Accounting and Business Research, 28(1), 19–39.
Peloza, J. (2009). The challenge of measuring nancial impacts from investments in corporate social performance. Journal of Manage- ment, 35(6), 1518–1541. Peloza, J., Hudson, S., & Hassay, D. N. (2009). The marketing of employee volunteerism. Journal of Business Ethics, 85(2), 371–386.
Peterson, D. K. (2004). Benets of participation in corporate volun- teer programs: Employees’ perceptions. Personnel Review, 33(6), 615–627.
Priem, R. L., & Sha er, M. (2001). Resolving moral dilemmas in busi- ness: A multicountry study. Business & Society, 40(2), 197–219.
Rodell, J. B. (2013). Finding meaning through volunteering: Why do employees volunteer and what does it mean for their jobs? Acad- emy of Management Journal, 56(5), 1274–1294.
Ryan, T. G., & Bisson, J. (2011). Can ethics be taught? International Journal of Business and Social Science, 2(12), 44–52.
Scheidler, S., Edinger-Schons, L. M., Spanjol, J., & Wieseke, J. (2018). Scrooge posing as Mother Teresa: How hypocritical social respon- sibility strategies hurt employees and rms. Journal of Business Ethics. https ://doi.org/10.1007/s1055 1-018-3788-3.
Seagram and Sons. (1989). The Chivas regal report on working Ameri- cans: Emerging values for the 1990s. New York, NY: Research & Forecasts, Inc.
Sundaram, R. K., & Yermack, D. L. (2007). Pay me later: Inside debt and its role in managerial compensation. The Journal of Finance, 62(4), 1551–1588.
Thompson, J. A., & Hart, D. W. (2006). Psychological contracts: A nano-level perspective on social contract theory. Journal of Busi- ness Ethics, 68(3), 229–241.
Weber, J. (1994). Ethical dilemmas in business. Business and Society, 33(3), 329.
Zahller, K., Arnold, V., & Roberts, R. W. (2015). Using CSR disclo- sure quality to develop social resilience to exogenous shocks: A test of investor perceptions. Behavioral Research in Accounting (In-Press).
1 3 Reproduced with permission of copyright owner. Further reproductionprohibited without permission.