How does corporate sexual equality affect firm value? Free cash flow vs cost of capital
Trevor Marwick School of Economics and Finance Curtin University Kent Street, Perth Australia Email:
[email protected]
Mostafa Monzur Hasan* School of Economics and Finance Curtin University Kent Street, Perth Australia Telephone: +618 92663414 Email:
[email protected]
& Adrian (Wai-Kong) Cheung Flinders Business, College of Business, Government and Law Flinders University, Bedford Park Campus, Adelaide, SA 5042, Australia. E-mail address:
[email protected].
*Contact author
0
How does corporate sexual equality affect firm value? Free cash flow vs cost of capital
Abstract An often overlooked component of corporate social responsibility (CSR) is a firm’s stance on sexual orientation equality; more specifically, its view on the lesbian, gay, bisexual, and transgender (LGBT) society. Using the discounted cash flow (DCF) model, we identify two finance channels through which firms’ adoption of sexual orientation equality may affect firm value: free cash flow (FCF) and the cost of capital (COC). Basic finance principles predict that any increase in firm value is attributed to either an increase in cash flows (or earnings), or a decrease in the discount rate (or expected return), or both. Firms that have a high corporate sexual equality index (CEI) score are expected to generate higher positive FCFs via an increase in labour productivity. CEI firms are also perceived to be riskier than similar firms due to stakeholders’ divided views regarding sexual orientation equality. Using 555 U.S. firms with 3,329 firm-year observations from 2002 to 2014, we establish that CEI affects firm value negatively through the COC channel. Importantly, a positive direct effect of CEI on firm value persists to alternative estimation methods and different specifications of firm value, FCF, and CEI. Keywords: LGBT, corporate sexual equality index, discounted cash flow model, free cash flow, cost of capital JEL classification: G11; G12; J70; M14
1
1. Introduction Sexual orientation equality is a new area of research that has received growing attention over the past decade (Anteby & Anderson, 2014; Theodorakopolous & Budhwar, 2015), particularly with interest to its effects on firm performance/value (Cunningham, 2011; Pichler et al., 2016; Shan et al., 2017; Wang & Schwarz, 2010). Sexual orientation equality is an integral part of a firm’s diversity management as it may act as a signal of future favourable employee outcomes; such as decreased levels of job anxiety and higher levels of career fulfilment, subsequently attracting talented LGBT workers (see Tejeda, 2006; Wang & Schwarz, 2010). One definition of sexual orientation equality is presented by Shan et al. (2017) as “how a firm treats their lesbian, gay, bisexual, and transgender (hereafter LGBT) employees” (p. 1813). However, the Oxford dictionary expresses sexual orientation equality as “the state in which access to rights or opportunities is unaffected by gender”, which is a slightly different (yet subtle) definition. Nevertheless, it is easy to see that embracing sexual orientation equality can be met with positive outcomes for both employees and employers. Within business management literature, organisational diversity has been linked to value-enhancing benefits for firms (Richard, 2000; Theodorakopoulos & Budhwar, 2015). This stream of studies provides support to the potential benefits stemming from adoption of diversity-management practices. Richard (2000) finds that cultural (racial) diversity adds firm value and contributes to firm competitive advantage. Day and Greene (2008) also suggest that firms must consider diversity-supportive policies to stay competitive in the battle for talented employees. Likewise, LGBT-supportive workplace policies and practices should provide performance-enhancing benefits to firms. Empirical research studying the effects of corporate sexual equality on firm-level outcomes mainly suggest that a positive relation exists (Cunningham, 2011; Shan et al., 2017; Wang & Schwarz, 2010). Adoption of LGBT-supportive policies enhance the recruitment process and a firm’s ability to attract talented candidates (Wang & Schwarz, 2010), reduces
2
employee turnover (Metcalf & Rolfe, 2011), and creates an open, tolerant, and comfortable workplace that promotes greater labour productivity (Cunningham, 2011; Shan et al., 2017). Unlike the myriad of conclusions drawn from papers within corporate social responsibility (hereafter CSR) literature (Husted & Allen, 2009; Luo & Bhattacharya, 2006; McWilliams et al., 2006; Waddock & Graves, 1997; Wright & Ferris, 1997), the effects of embracing sexual orientation equality predominantly suggest that it has a positive impact on firm outcomes. Our paper builds on the findings of Shan et al. (2017) who examine the direct effect of corporate sexual equality on firm performance, along with the indirect effects via the labour market channel. Using corporate equality index (hereafter CEI) scores obtained from the Human Rights Campaign (hereafter HRC) from 2002 to 2006, they find evidence that U.S. firms that embrace sexual orientation equality experience, on average, higher stock returns and firm performance/value. These effects are also partially mediated through the labour market channel, i.e., sexual orientation equality gives rise to an increase in employee productivity, subsequently driving higher stock returns and firm performance/value. The authors acknowledge that the indirect effect amounts to only 10 percent of the total effect, and the labour market channel is just one of many plausible channels through which LGBT-supportive policies influence firm performance/value. Our paper differs and extends upon Shan et al. (2017)’s findings in the following ways. First, instead of identifying from real sector (or operational side of firms), we look at the problem from a finance perspective. In particular, according to the discounted cash flow (DCF) valuation model, a firm value is determined by its future cash flows and expected return (or discount rate). Thus, the effects of sexual orientation equality on firm value, if they exist, must be captured by two finance channels. The first one is free cash flow (hereafter FCF) and the other is cost of capital (hereafter COC). Shan et al. (2017) show that sexual orientation equality can increase firm value. However, the DCF model predicts that an increase in firm value can
3
happen only if there is an increase in FCF, an decrease in expected return (or cost of capital), or both. In our view, it is important to disentangle these two channels because they provide totally different mechanisms through which sexual orientation equality can affect firm value. Second, it is noteworthy that the positive impact of sexual orientation equality on firm value, as Shan et al. (2017) and others have showed, can only capture the net effect of sexual orientation equality, which may mask the complementary or substitutive nature of indirect effect of sexual orientation equality via the FCF channel and the COC channel. Thus, we also examine the relative importance of these two channels and check if they are complementary or substitutive. Third, we extend Shan et al. (2017)’s sample that originally covered the period from 2002-2006 to 2002-2014 and check if their results are sensitive to the sampling period. COC Corporate sexual orientation equality may strengthen FCF because the LGBT community remains loyal to high CEI firms (Oakenfull, 2013). From the perspective of consumers who classify themselves as LGBT or supportive of the LGBT community, they may be more able and willing to purchase goods and services from businesses that are strong advocates of sexual orientation equality, while being prepared to pay a premium over low CEI firms (Colgan et al.,
2007). CEI firms also offer employees an open and comfortable
environment that promotes higher labour productivity. As such, one may view investment in sexual orientation equality as providing a competitive advantage over low CEI firms, which successively leads to an increase in sales revenue, and therefore, an increase in FCF. The impact of corporate sexual orientation equality on COC is ambiguous. From an investor’s standpoint, however, a firm’s inability to embrace sexual orientation equality may either increase or decrease firm risk. On the one hand, growing pressures from society for companies to support the LGBT community may cause concerns for low CEI firms. For example, businesses that have a pro-LGBT agenda may be less willing to cooperate with low
4
CEI firms opposed to high CEI firms. If shareholders take this view, they would expect higher (lower) returns for low (high) CEI firms, indicative of greater (smaller) firm risk; consistent with CSR literature (Kotchen & Moon, 2011; Godfrey et al., 2009). On the other hand, opponents of LGBT workplace equality warn of the adverse effects of embracing sexual orientation equality such as public and/or consumer backlash. Investors who share this view would demand higher (lower) returns for high (low) CEI firms. Using CEI scores obtained from the HRC of 555 U.S. firms with 3,329 firm-year observations between 2002 and 2014, we examine the importance of two mediating finance channels (i.e., FCF and COC) through which corporate sexual equality affects firm value. We find that only the COC channel holds significance in explaining the effects of CEI on firm value, which is robust to several different estimation methods and alternative specifications. The COC channel shows that sexual orientation equality has a negative mediation effect on firm value, whereas the FCF channel and total indirect effect are insignificant at the 5% level. We also find evidence that the direct effect of adopting sexual orientation equality persists once controlling for the potential issue of reverse causality, i.e., whether firm value drives investment in LGBT-supportive policies and practices or vice versa. More specifically, these findings suggest that incorporating LGBT-supportive workplace policies has a positive direct impact on firm value and that the effects are strengthened in the future; viz. through future (rather than current) cash flows. Our study contributes to the literature in the following ways. First, we address some issues of the findings presented by Shan et al. (2017). Specifically, the authors find that CEI firms experience greater stock returns (or COC) and higher firm performance/value, which contradicts the DCF valuation model; i.e., any increase in stock returns (or COC) is associated with a decrease in firm value, ceteris paribus. Second, unlike other studies estimating the direct effect of CEI on firm-level outcomes (Cunningham, 2011; Pichler et al., 2016; Shan et al.,
5
2017; Wang & Schwarz, 2010), we incorporate a system of equations that describe the total effect through the DCF model. According to the DCF model, firm value is captured through future cash flows and the relevant discount rate (COC). We provide empirical evidence of the effectiveness of corporate sexual equality in generating firm value, mediated through the FCF and COC channels. Third, we address several endogeneity issues in the form of omitted variable bias and reverse causality by using the two-stage least squares (2SLS) estimation method, and replacing contemporaneous CEI by its lagged variant, respectively. The remainder of the paper is structured as follows. Section 2 discusses recent literature regarding corporate sexual equality, FCF, and the COC, along with our hypotheses. Section 3 details the sample selection, key variables, and methodology. Section 4 reports our results, and Section 5 provides a discussion of our findings along with our conclusions.
2. Literature Review and Hypotheses Development 2.1 The discounted cash flow model The discounted cash flow model (DCF) indicates that the market value of a firm is represented by the present value of all future cash flows, discounted at the appropriate discount factor (such as the COC), such that the value of the firm is given by:
t=∞
V0 = ∑ t=1
FCFt FCFt = , t (1 + COC) COC
(1)
where V0 is the market value in year 0, FCFt is the excess cash flow expected in year t and COC is the required rate of return. The second equality applies if FCFt is a perpetual cash flow. This equation clearly shows the two factors affecting firm (or market) value are FCFt and the COC. Any increase in present value (stock price or firm value) is attributed to either an increase in 6
cash flow (or earnings), or a decrease in the discount rate (or expected return), or both. In addition, any change in the discount rate can be due to either a change in the risk-free rate or a change in the risk premium or both. In other words, there are only two channels that may affect firm value: the FCF channel and the COC channel. In the context of this study, any change in firm value due to the corporate sexual equality policies must work via the FCF channel, or the COC channel, or both.
2.1.1 Free cash flow Richardson (2006, p. 160) defines FCF as “cash flow beyond what is necessary to maintain assets in place and to finance expected new investments.” Managers face increasing pressures from stakeholders to allocate resources that maximise shareholder value. However, agency theory (Jensen, 1991, 1993; Jensen & Meckling, 1976) suggests that FCF provides managerial opportunities to pursue projects that maximize their wealth at the expense of shareholders (see Jensen, 1993, p. 841). Consequently, investment in such projects would deteriorate firm value, or equivalently, shareholder value. Studies have also documented an inverse relationship between FCF and future firm performance (Dechow et al., 2008). These findings are consistent with agency theory, whereby a reduction in FCF decreases agency costs, which subsequently increases firm value. While researchers have uncovered reasons as to why FCF may be detrimental to future firm performance, there exists a number of reasons as to why an increase in FCFs would lead to an increase in firm value. First, one would predict that having access to more capital should capable the firms to avail investment opportunities that were previously inaccessible due to financing constraints. Second, with a greater opportunity set of positive NPV projects, firm’s managers have a greater selection of investments that will yield positive returns, subsequently creating shareholder value (firm value). In theory, this statement makes perfect sense, however,
7
in practice, this depends on the manager’s ability and willingness to act in the best interests of shareholders. Corporate governance is a mechanism that addresses the agency problems induced by the disjoint of control and ownership amongst managers and shareholders within the modern corporation (Gompers et al., 2003). Firms that employ good corporate governance practices align managers’ decisions with shareholders’ best interests, leading to optimal outcomes for the company. Therefore, with appropriate measures in place, excess cash flows can lead to better and wider project selection with an increased future firm performance/value.
2.1.2 Cost of capital The COC (or a firm’s expected stock return) has been an area of interest, primarily because of its use in equity valuations and capital budgeting. As the ex-ante COC is unobservable, scholars used ex-post stock returns estimated via asset pricing models as a suitable proxy for COC. Over time, however, researchers have uncovered a number of shortcomings outlined in the assetpricing literature (e.g. Elton, 1999; Fama & French, 1997, 2004). For example, Fama and French (1997, 2004) empirically find that the CAPM is an “imprecise” and “noisy” proxy of a firm’s COC. Contemporary literature uses accounting-based valuation models to obtain the implied rate of return, which equates the present value of future cash flows with the current market price (Echterling et al., 2015). While asset pricing models such as the CAPM rely on ex-post data for ex-ante valuation, the implied approach utilises ex-ante forecasted cash flows for estimation. Three popular approaches to the implied cost of capital (hereafter ICC) are established in the literature, and they include the residual income valuation model, the dividend discount model, and the abnormal earnings growth model (Echterling et al., 2015).
8
Researchers have revealed a number of firm- and non-firm-specific factors that affect the COC (or expected return). These include financial leverage, firm size, liquidity, firm age, corporate disclosure, economic growth, and industry-specific elements.1
2.2 Corporate sexual equality A firm’s stance on corporate sexual equality may be viewed as how a company treats their lesbian, gay, bisexual, and transgender employees (Shan et al., 2017). Historically, social acceptance of homosexuality has been negative. In 1965, 70% of survey respondents in a public opinion poll specified that homosexuals were harmful to American life (Herek, 2002). Today, however, approximately 54% of Americans consider gay or lesbian relations morally acceptable (Gallup, 2012). Similarly, 63% of Americans feel that discrimination against gay men or lesbians is a severe problem (Gallup, 2014). Furthermore, almost two-thirds of the population know someone who is either gay or lesbian, and 64% of those people believe that their behaviour is not a moral issue; whereas 53% of those who do not know a gay or lesbian person feel the same way (Pew Research Center, 2016). Comparable findings are presented in other parts of the world such as in Australia, where 61.6% of the population voted in favour of allowing same-sex couples to marry (ABS, 2017). On 26th September, 2017, the U.N. High Commissioner for Human Rights released new standards of conduct to eliminate discrimination against LGBT people in the workplace and to provide support to LGBT employees.2 Given the recent media and political attention surrounding LGBT orientation equality, many firms have been proactive and increased their efforts to support LGBT employees in recent years; up 27% from 2002 to 2013 (Human Rights Campaign, 2014). With these developments, researchers have begun to investigate the effects of sexual orientation
1
See discussion in Section 3.3.5 for a detailed discussion. http://www.csrandthelaw.com/2017/10/10/new-standards-of-conduct-for-companies-seeking-to-operate-withrespect-for-the-rights-of-lgbti-people/ 2
9
equality policies on firm performance (Cunningham, 2011; Pichler et al., 2016; Shan et al., 2017). Firms that adopt non-discriminatory practices within their workplace are perceived to have an open, diverse, and supportive environment, which can stimulate greater productivity. Studies argue that LGBT-friendly firm policies reflect an improvement in organisational culture that endorses efficiency and productivity within the workplace, indirectly leading to an increase in firm performance (Cunningham, 2011; Shan et al., 2017; Wang & Schwarz, 2010). Wang and Schwarz (2010) claim that sexual orientation equality practices attract skilled and gifted individuals who would otherwise be unnoticed. Similarly, Shan et al. (2017) show that adoption of LGBT-supportive policies lead to higher stock return and market valuation. They also show that the relation between corporate sexual equality and firm performance is mediated by the labour market channel. Other empirical studies, however, have found that adoption of LGBT-friendly policies can lead to value-neutral outcomes. For example, Johnston and Malina (2008) observe a significant and positive abnormal return on the day a company announces the adoption of LGBT-supportive policies. However, extending the event window to three days results in insignificant abnormal returns, signalling a neutral reaction from shareholders. This findings provide evidence that shareholders do not react unfavourably to firms allocating resources to sexual orientation equality policies. While the impact of embracing sexual orientation equality within the workplace is far from established, it coincides with the idea that shareholders are more tolerant to the LGBT community than previously thought.
2.3 Hypotheses development In a recent paper, Shan et al. (2017) document that firms with a higher degree of corporate sexual equality have higher stock returns and market value. Standard finance literature suggests
10
that this increase in stock returns is likely due to an increase in risk premium (assuming that the risk-free rate is constant). However, an increase in stock returns (COC), according to the DCF model, may result in a drop in stock price or firm value, which is inconsistent with Shan et al. (2017)’s result that Tobin’s q is higher when employing LGBT-friendly firm policies. One possible explanation is that LGBT policies give rise to an increase in both cash flows and risk premium and that the former is greater than the latter, resulting in an overall increase in stock returns and market valuation. Traditional economic theory suggests that managers make decisions to maximize shareholders’ wealth or firm value (Friedman, 1970). CSR Literature indicates that socially responsible activities can aid in the differentiation of a firm’s products from the rest of the market (McWilliams & Siegel, 2000), can help companies’ avoid paying government-imposed fines (Brown et al, 2006; Freedman & Stagliano, 1991), and can alleviate a firm’s financial and market risks (Godfrey, 2005; Lee & Faff, 2009); all of which improve future cash flows. Sexual orientation equality is considered by some authors (Johnston & Malina, 2008; Shan et al., 2017) as a specific-type of CSR. Therefore, we expect these effects to persist when firms invest in LGBT-related firm policies. We argue that, corporate sexual orientation equality generate higher FCFs through several additional ways. First, studies have shown that LGBT stakeholders are more willing to purchase goods and services from firms that have embraced sexual orientation equality (Oakenfull, 2013). The estimated net worth of the LGBT consumer market is in excess of $835 billion in USA (Witeck & Combs, 2006). In addition, global LGBT spending power is estimated to be $3.7 trillion.3 Thus, if firms can position themselves to capture this market, they will generate considerable cash flows through a rise in sales revenue. Second, LGBTsupportive firm policies and practices improve employee productivity. Firms that embrace
3
https://www.gaystarnews.com/article/global-lgbt-spending-power-estimated-to-be-3-7trillion/#gs.jGGeIVc
11
sexual orientation equality tend to offer a working environment that promotes collaboration, teamwork, and other positive workplace attitudes (Badgett et al., 2013), which all benefit the firm in the form of higher cash flows. This leads us to our first hypothesis.
Hypothesis 1: Firms with higher (lower) CEI scores are expected to generate more (less) free cash flows, which increases (decreases) firm value.
Scholars argue that CSR provides insurance-like protection and lowers idiosyncratic risk as firms that invest in social capital are less susceptible to shocks within the economy (Kang et al., 2016; Luo & Bhattacharya, 2006). As another form of CSR, sexual orientation equality has received growing attention and has become widely accepted across the globe in recent years (Pew Research Center, 2016). Approximately 85% of Americans oppose job discrimination based on a person’s sexual orientation (Burns, 2012). Moreover, society now urges the firms to embrace sexual orientation equality and offer protection against discriminatory actions based on a person’s sexual identification. Therefore, firms that fail to adopt LGBT-supportive policies and practices within the workplace may face adverse outcomes. For example, suppliers that have a pro-LGBT agenda may be hesitant to conduct business if discrimination against sexual orientation equality is witnessed inside the firm. Additionally, investment opportunities from potential pro-LGBT firms will cease to exist. From the consumers’ perspective, a firm’s inability to acknowledge and embrace sexual orientation equality could lead to a decline in demand for their goods and services. Moreover, anti-LGBT practices may be met with public backlash from “gay” consumers, and possibly end in an absolute boycott of the firm's products. Empirical evidence also suggests that firms are penalised for not conducting their business in ways that coincide with social norms and values (Kotchen & Moon, 2011). In sum, firms that
12
fail to adopt LGBT-supportive policies and practices may be subject to adverse outcome and higher level of risk, which increase COC and lower firm value.
Hypothesis 2a: Firms with lower (higher) CEI scores tend to have a higher (lower) level of risk and, thus, a higher (lower) cost of capital, which will then translate into lower (higher) firm value.
While sexual orientation equality has become increasingly accepted in countries across the globe (Pew Research Center, 2016), a large proportion of society remain less tolerant to the idea. Investors must consider the impacts of employing corporate sexual equality practices within the workplace; particularly the influence on labour productivity and investment opportunities. There remains a high level of uncertainty as to how anti-LGBT consumers, suppliers and workers will act if firms choose to invest in this type of social capital. Moreover, investment opportunities may become scarce if anti-LGBT companies choose to not associate or cooperate with the company. This directs us to the following hypothesis:
Hypothesis 2b: Firms with higher (lower) CEI scores tend to have a higher (lower) level of risk and, thus, a higher (lower) cost of capital, which will then translate into lower (higher) firm value.
3. Research Method 3.1 Data and Sample We collect corporate sexual orientation equality data manually from the HRC’s Corporate Sexual Equality Index annual report, financial data from Compustat merged database, stock price data from the Center for Research in Security Prices (CRSP), and CSR scores from the 13
MSCI database. We achieve our sample from the intersection of companies listed on the HRC’s annual reports which are covered by CRSP/Compustat merged dataset for the period 2002 – 2014. We start our sample in 2002 since data on CEI score is available from this period. The initial dataset contains 4,959 firm-year observations. We exclude firms from the financial sectors as the characteristics of these companies differ substantially from the rest of the sample regarding risk, complexity, and accounting practices. We also eliminate 1,630 firm-year observations with missing data in the computation of the variables used in the study. The final sample consists of 555 U.S. firms with 3,329 firm-year observations. Sample selection process is explained in Table 1.
[Insert Table 1 here]
Table 2 provides a breakdown of the industries for the full sample of firm-year observations. The manufacturing, wholesale and retail trade, and utility sectors represent 42.99%, 19.74%, and 16.28% of the sample, respectively. To control for industry effects, we include two-digit standard industrial classification (SIC) dummies in the regressions.
[Insert Table 2 here]
3.2 Empirical Model To estimate the direct, indirect (through ex-ante COC and FCF), and total effects of CEI on firm value, we employ the following simultaneous equation model:
qi,t = β0 + β1 CEIi,t + β2 COCi,t + β3 FCFi,t + β4 BTMi,t + β5 SIZEi,t + β6 LEVi,t + β7 R&Di,t + β8 SP500i,t + β8 AGEi,t + Σt yeart + Σt industryi,t + εi,t (2) 14
COCi,t = α0 + α1 CEIi,t + α2 BTMi,t + α3 SIZEi,t + α4 BETAi,t + α5 LOSSi,t−1 + α6 LEVi,t + α7 ZSCOREi,t + α8 ROEi,t + Σt yeart + Σt industryi,t + vi,t
(3)
FCFi,t = γ0 + γ1 CEIi,t + γ2 ∆STDEBTi,t + γ3 ∆NWCi,t + γ4 SIZEi,t + γ5 ROEi,t + γ6 CFi,t + γ7 ACQ i,t + γ8 LEVi,t + Σt yeart + Σt industryi,t + ωi,t
(4)
Our simultaneous equation model consists of three equations. We follow Faff et al. (2016) and specify Eqn. (4) which estimates the effects of CEI on FCF. Based on the view that LGBTsupportive policies improve sales and employee productivity, we expect CEI to increase FCF. We follow prior studies (Dhaliwal et al., 2014; Ding et al., 2015; Hann et al., 2013) and specify Eqn. (3) to estimate the effects of CEI on firm risk (or expected return). We argue that LGBTsupportive policies may increase or decrease firm risk and COC. Finally, to test whether firms with higher CEI scores have greater firm value, we follow Shan et al. (2017) and specify Eqn. (2) where the dependent variable is the Tobin’s q ratio. Building on from Shan et al. (2017)’s methodology, we include FCF, and the COC as key channels through which CEI affects firm value indirectly. Based on the DCF valuation model, we expect that FCF or COC) or both will affect firm value (q). We predict that higher FCF (COC) has a positive (negative) effect on Tobin’s q. Also, note that we include CEI as a key explanatory variable in all three equations for a number of reasons. First, we expect that CEI has a linear relationship with FCF, COC, and firm value (q). Second, the impact of FCF and COC on Tobin’s q may be nonlinear, while sexual orientation equality (CEI) is linear. Third, the DCF model estimates a firm’s value based on future FCFs, therefore, in theory, we should include FCFs at later dates (i.e., t + j where j = {1,2,3,4,5,….∞)) in Eqn. (4); however, we argue that FCF at time t is strongly correlated with future FCFs. In addition, we predict that CEI can explain the missing effects from omitting future cash flows from the model (i.e., Eqn. (2)). 15
3.2.1 The Mediation Process We employ two econometric approach to estimate the indirect (via FCF and the COC), direct, and total effects of CEI on firm value. First, we use structural equation modelling (SEM), which comprises of a system of linear simultaneous equations that represents the relationships between our key variables. In particular, we are interested on the effects of the main explanatory variable (CEI) on the channel variables (FCF and COC) which, in turn, affects the dependent variable (q) (see Figure 1). One advantage of SEM over 2SLS is that the maximum likelihood (ML) estimator is more efficient in sufficiently large samples, given that the model specification is correct and all assumptions are valid. Additionally, the ML method allows for correlations between the error terms, whereas the two-stage least squares (2SLS) method does not. To address these concerns and check the robustness of our findings, we also use the 2SLS method to control for potential endogeneity. There are a number of advantages of using the 2SLS over the ML method for SEM. Evidence from econometric literature suggest that 2SLS may perform better in smaller samples (Bollen, 1996) and it allows the testing procedures for issues such as heteroscedasticity and specification errors (Pesaran & Taylor, 1999), and it does not require numerical optimization algorithms.
[Insert Figure 1 here]
The direct effect of CEI on firm value is captured by 𝛽1 in Eqn. (2), while the indirect effects are captured by 𝛽2 𝛼1 for the COC channel and 𝛽3 𝛾1 for the FCF channel, respectively. For example, consider Hypothesis (1) from Section 2.3, in which we test whether the indirect effect of CEI through the FCF channel is significant. We construct the null hypothesis as follows: 16
𝐻𝑦𝑝𝑜𝑡ℎ𝑒𝑠𝑖𝑠 1: 𝛽3 𝛾1 = 0.
(5)
We posit that the indirect effect from the FCF channel is significant; i.e., the effect of CEI on FCF (𝛾1 ) through which it affects firm value (𝛽3 ), jointly (𝛽3 𝛾1 ), is statistically different from zero. This is achieved by taking the product of the coefficient of CEI from the FCF channel and the coefficient of FCF in Eqn. (3) (𝛽3 𝛾1 ). We repeat the same process for the COC channel. This method allows us to observe whether sexual orientation equality, when tested independently and jointly, are significant (or insignificant) predictors of firm value. Since this procedure requires us to take the product of two random variables, we must also estimate the standard error of the indirect effects. As proposed by Goodman (1960), the standard error of the indirect effect under the assumption of normality is given as follows:
𝑠 2 (𝛽3 𝛾1 ) = 𝑠𝛽23 𝛾12 + 𝑠𝛾21 𝛽32 − 𝑠𝛽23 𝑠𝛾21 .
(6)
However, the multivariate delta method (Sobel, 1982, 1986) is more commonly used in literature to obtain the estimator of the variance, as it is deemed to provide the least unbiased estimates (MacKinnon & Dwyer, 1993; MacKinnon et al., 1995; MacKinnon et al., 2002). The first order Taylor series solution estimator (delta method) for the variance of the indirect effect is provided as follows:
𝑠 2 (𝛽3 𝛾1 ) = 𝑠𝛽23 𝛾12 + 𝑠𝛾21 𝛽32 .
(7)
The basic idea behind this method is to use a Taylor series expansion to derive a linear model that approximates the more complicated function (i.e., 𝛽3 𝛾1 in this example). However, for this to work, the sample must be large, and the assumption of normality must hold. To 17
address these concerns, we also use the bootstrapping method suggested by MacKinnon et al. (2004) to improve the accuracy of small sample sizes. The fundamental idea is to generate simulated bootstrap samples from the original data sample and use them to calculate the bootstrap test statistic. In our case, we resample with replacement, and for each simulated bootstrap sample, we estimate the indirect effect of CEI on firm value. In particular, we resample 1000 times and construct a confidence interval at the predetermined significance level (i.e. 100(1 − 𝛼)). In practice, however, the distribution tends to be skewed and has excess kurtosis when the product of the means of two random variables is nonzero. MacKinnon et al. (2004) show that the bootstrap confidence intervals can be further improved by using the accelerated bias correction bootstrap method; which adjusts for the skewness in the bootstrap distribution. If the bias-corrected confidence interval does not contain zero, we reject the null hypothesis in favour of the alternative.
3.3 Measurement of Variables 3.3.1 Corporate sexual equality (CEI) Following Shan et al. (2017), we chose the firm’s CEI score as a proxy for how extensively firms administer sexual orientation diversity within the workplace. In August 2002, the HRC, the largest U.S. organisation that is at the forefront of fighting for the U.S. Equality Act 2010, commenced an annual survey to rate U.S. companies on their level of corporate sexual equality; 0 being the lowest and 100 being the highest. From 2010, however, the scale changed to -15 to 100 to reflect additional criterion in their computation of each firm’s CEI score. To control for this, we transformed the scores back to the original 0 to 100 scale for those years affected. In particular, we construct the following scaling formula:
CEIscaled =
(y − x)(CEI − min) + x, max − min 18
(8)
where x and y represent the range of the CEI score (in this case x = 0 and y = 100), min is the minimum CEI score observed (i.e., -15), max represents the maximum CEI score observed (i.e., 100), and CEI is the firm-year score. CEIscaled denotes the CEI score in the new range (i.e., 0 to 100). Each year, the HRC sends a survey to large U.S. corporations listed on the Fortune 500, Forbes’ list of the 200 largest privately held firms, and firms with greater than 500 employees. While the rating criteria have evolved over time, the survey generally measure firms equal employment opportunities; whether employers provide equivalent partner benefits; the provision of transgender-inclusive health insurance; nondiscriminatory policies with regards to LGBT training, resources, and accountability measures; LGBT employment efforts; and if a firm demonstrates public support for LGBT equality outside of the workplace. For example, companies who employ LGBT recruitment efforts, whether that includes workers or suppliers, will net 15 points towards their CEI score; on the other hand, employers will have 25 points deducted if they are associated with a public anti-LGBT blemish on their records within survey specific period. These scores are then tallied across all categories. Over the course of the sample period, the number of sub-categories for each criterion has changed, and we control for this by employing year fixed effects within the model.
3.3.2 Estimation of cost of capital Following the literature, we use the implied approach to estimate the COC (Claus & Thomas, 2001; Easton, 2004; Gebhardt et al., 2001; Gode & Mohanram, 2003; Gordon & Gordon, 1997; Ohlson & Juettner-Nauroth, 2005). The key advantage is that implied COC models do not rely on noisy realised returns or any asset pricing models. Lee and Faff (2009) also suggest that the
19
use of ICC estimates offers greater transparency regarding economic trends that would be less apparent in realised returns. Following Hou et al. (2012), we use earnings forecasted from a cross-sectional model to proxy for cash flow expectations, instead of using analysts’ forecasts. The authors empirically demonstrate the superiority of model-based forecasted compared to analysts’ forecast on a number of criteria including coverage, forecast bias, and earnings response coefficient. Rather than focusing on one particular model, we incorporate four (4) different models with six (6) alternative COC measures and take the average to mitigate the effect of measurement errors associated with one particular method (Chen et al., 2009; Hail & Leuz, 2006; Hou et al., 2012). These models include Gordon and Gordon (1997); Claus and Thomas (2001) (CT); Gebhardt et al. (2001) (GLS); Easton (2004)’s PEG and MPEG; and Ohlson and Juettner-Nauroth (2005) (OJ) modified by Gode and Mohanram (2003).4
3.3.3 Free cash flow We follow Unsal et al. (2016) and measure FCF as operating income before depreciation minus interest and related expenses, income taxes, and capital expenditures. We then scale FCF by total assets, multiplied by 100, at the end of the fiscal year.
3.3.4 Tobin’s q Tobin’s q is extensively used in financial literature as a proxy for future operating performance, and firm value (Hong & Kacperczyk, 2009; Shan et al., 2017). We measure Tobin’s q as the ratio of the market value of a company’s assets (measured by the market value of outstanding stock and debt) divided by the book value of assets. We then take the natural log of Tobin’s q as our proxy for firm value.
4
For detailed discussion about the implied cost of equity models, see Echerling et al. (2015)
20
3.3.5 Control Variables We follow prior studies (Chen et al., 2009; Dhaliwal et al., 2014; Gompers et al., 2003; Hong & Kacperczyk, 2009; Shan et al., 2017) and include several risk factors and firm characteristics as controls in our simultaneous equation model. Book-to-market (BTM) is commonly used as a proxy for growth and risk (Fama & French, 1992). A smaller book-to-market ratio signals higher firm growth opportunities and vice versa. Firms that have a high BTM ratio carry greater uncertainty and risk relative to similar firms with lower BTM ratios. We measure the BTM ratio as the natural log of the ratio of a company’s book value (or historical accounting value) to its market value. Large firms are associated with a lower default risk, information risk and valuation uncertainty (Berger and Udell, 1995; Fama & French, 1992). We use the natural log of total assets to control for size effect (SIZE). Financial leverage (LEV) is a common proxy for firm-level risk. Companies that issue more debt are perceived to have a greater risk of default than similar firms with smaller debt obligations. We measure leverage as total debt (short-term debt + longterm debt) divided by shareholders’ equity. Firms with higher systematic risk (BETA) are perceived to have greater overall risk than equivalent firms with less systematic risk (Lintner, 1965; Sharpe, 1964). We estimate the market model beta using the prior 36 monthly valueweighted CRSP index returns and only use data that is available for at least 12 months before estimation. Firms are unable to sustain ongoing losses in earnings over a long-term period. Collins et al. (1999) find that loss firms have a higher propensity to abandon their resources (i.e., undergo liquidation). As another proxy for risk, we include a dummy variable (LOSS) equal to 1 if the firm has experienced a loss in the prior year, 0 otherwise. We also include Altman’s (1968) Z score as an additional control variable to capture bankruptcy risk. We expect a negative association between the COC and Z-Score. We include return on equity (ROE) as a
21
control for profitability and efficiency. A higher ROE signals that the company can generate more profits without the need of much capital. We compute ROE as net income divided by total shareholders’ equity. Older firms have a propensity to engage in rent-seeking behaviour and organisational rigidities, leading to a decline in profitability (Loderer & Waelchli, 2010). Consistent with Shan et al. (2017), firm age (AGE) is measured as the natural log of the number of years since a company is listed in the CRSP database. R&D intensity is considered as a form of investment in “technical” capital. Researchers have outlined the importance of including R&D expenditure as a determinant of firm performance (McWilliams & Siegel, 2000; Waddock & Graves, 1997). We measure R&D intensity as R&D expenses divided by total sales. We treat missing R&D expenses as equal to 0. The Standard and Poor’s 500 is widely recognised as a gauge of large-cap U.S. firms. Following Shan et al. (2017), we include a dummy variable equal to 1 if a firm is listed on the S&P500, 0 otherwise. Any increase in short-term debt will have a direct effect on a firm’s FCF as excess cash is used to pay off debt within the current year. Therefore, we expect a negative association between the change in short-term debt (∆STDEBT) and a firm’s FCF. We follow Faff et al. (2016) and include the change in net working capital (∆NWC) as a control for FCF. We measure net working capital as current assets minus current liabilities, scaled by lagged total assets. To control for the effects of omitted variable bias, cash flow is used as a measure to isolate the effects of the primary variables of interest on FCF. We measure cash flow (CF) as operating income before depreciation minus interest and related expenses and income taxes, scaled by total assets and multiplied by 100, at the end of the fiscal year. Consistent with Faff et al. (2016), we use acquisitions (ACQ) as a control variable for changes in cash balances as well as its direct effect on a company’s FCF.
22
4. Empirical Results and Discussion 4.1 Summary statistics and correlation Table 3 reports the summary statistics and Pearson correlation matrix of the COC estimates. As seen in Panel A of Table 3, the average COC has a mean (median) of 7.13% (6.48%) which is lower than what is reported by many other papers (Dhaliwal et al., 2014; Ding et al., 2015; Hann et al., 2013). There are three reasons as to why this is the case. First, prior studies utilise analyst based earnings forecasts in their estimation of the COC, whereas we employ a modelbased approach to forecast earnings (see Hou et al., 2012). Second, the firms that the HRC choose to survey are larger and more closely followed by investors. Therefore, we expect lower COC (risk) for these firms. Third, the sample mainly consists of industries (e.g., manufacturing, retail, and utility) that have lower level of risk and lower excess premiums (see Gebhardt et al., 2000, p. 152). Panel A shows that the COC is highest under Easton’s (MPEG) model, with a mean (median) of 9.53% (8.17%). Ohlson and Juettner-Naroth (OJ) method produce the lowest average estimated COC with a mean (median) of 6.07% (5.60%). Panel B of Table 3 reports the Pearson correlations between the different COC estimates. We find that COC measures are highly positively correlated (at p
