“Impact of CEO pensions on value relevance of R&D expenditures”

Since CEO pension is unsecured and unfunded liabilities of the firm, it induces CEOs to have long-term incentives towards minimizing their firms’ default risk. Motivated by the unique characteristics of CEO pension, this study investigates the impact of CEO pension on the value relevance of R&D expenditures. Using Tobin’s Q ratio to measure firm value, the empirical results show that CEO pension intensifies the relation between R&D expenditures and Tobin’s Q ratio. The results remain robust in two-stage least square and propensity score matching regression analysis to address the endogeneity issues in the relation between CEO pension and the value relevance of R&D expenditures. In addition, the regression results with ROA and F-score as the alternative dependent variables also confirm that CEO pension intensifies the relation between R&D expenditures and firm value. This study further investigates whether CEOs awarded with larger CEO pensions are more likely to engage in R&D activities with long-term perspective. Although CEOs awarded with larger CEO pensions may be reluctant to engage in additional R&D activities, because firms with high R&D costs are par-ticularly vulnerable to liquidity shortages in the short term, CEOs are expected to carefully choose and perform less risky, but more valuable, R&D activities to increase their firm’s sustainability. Using a sample of U.S. firms listed in ExecuComp from 2006 to 2015, it was discovered that CEOs awarded with larger CEO pensions were more likely to pay R&D expenditures positively related to Tobin (1969) (Tobin’s Q ratio), a proxy for long-term firm value. These results remain robust by using two approaches, two-stage least square (2SLS) regression and propensity score matching (PSM), to alleviate the endogeneity concerns that may exist in the hypothesized relationship. Furthermore, the results with return on asset (ROA) and Piotroski’s (2000) F -score as the dependent variables, instead of Tobin’s Q ratio, also confirm that CEOs awarded with larger CEO pensions intensify the relationship between R&D expenditures and long-term firm value.


INTRODUCTION
Unlike the ordinary workers' pension plans, CEO pensions, also known as Supplementary Executive Retire Plans (SERP), are a type of defined benefits retirement plans that require a firm to make lumpsum payments, or periodic pension payments, after a CEO retires.These CEO pensions are widely used by U.S. firms as extra compensation.For example, Sundaram and Yermack (2007) found that 78% of Fortune 500 firms provide the extra pension plans to their CEOs.Wei and Yermack (2011) assessed that 84% of their sample firms provide the extra pension plans to CEOs.The size of CEO pensions also far exceeds that of ordinary pension plans.For example, Bebchuk and Jackson Jr. (2005) found that Standard and Poor (S&P) 500 firms use CEO pensions to make up on average 34% of total CEO compensation.Wei and Yermack (2011) documented from their sample firms that CEO pensions account for approximately $10 million.
CEO pensions have particular characteristics that differ from other types of CEO compensation, such as salary, bonus, stock, and option grants.As a firm is required every fiscal year-end to record future payments for retirement benefits as liabilities, CEO pensions give the CEO the same status as that of a creditor against the firm.Additionally, this status provides the right to receive debt claims with other unsecured debt holders if the firm goes bankrupt.Therefore, the unsecured and unfunded features of CEO pensions are known to align a CEO's interests with those of outside debt holders in an effort to lower the bankruptcy risk.For example, Cassell et al. (2012) find that firms with CEO pension plans are reluctant to engage in additional R&D activities, because high R&D costs increase liquidity shortages in the short term.
This study further investigates whether CEOs awarded with larger CEO pensions are more likely to engage in R&D activities with long-term perspective.Although CEOs awarded with larger CEO pensions may be reluctant to engage in additional R&D activities, because firms with high R&D costs are particularly vulnerable to liquidity shortages in the short term, CEOs are expected to carefully choose and perform less risky, but more valuable, R&D activities to increase their firm's sustainability.
Using a sample of U.S. firms listed in ExecuComp from 2006 to 2015, it was discovered that CEOs awarded with larger CEO pensions were more likely to pay R&D expenditures positively related to Tobin (1969) (Tobin's Q ratio), a proxy for long-term firm value.These results remain robust by using two approaches, two-stage least square (2SLS) regression and propensity score matching (PSM), to alleviate the endogeneity concerns that may exist in the hypothesized relationship.Furthermore, the results with return on asset (ROA) and Piotroski's (2000) F-score as the dependent variables, instead of Tobin's Q ratio, also confirm that CEOs awarded with larger CEO pensions intensify the relationship between R&D expenditures and long-term firm value.

THEORETICAL BACKGROUND AND HYPOTHESIS
Traditional agency theory provides that the optimal CEO compensation structure should have the ratio of equity-based to debt-based compensation equal to the firm's equity-to-debt ratio ( The dependent variable TobinQ is the Tobin's Q ratio calculated as the market value of assets divided by the book value of assets, & RD is the dollar amounts of R&D expenditures scaled by assets, FSIZE is the natural log of a firm's asset size, LEV is the ratio of book equity to debt, MTB is the ratio of market value to book value of equity, INT is the inventory turnover ratio as calculat- ed by dividing total asset by sales, CAPEX is the capital expenditure scaled by assets, CEO tenure is the number of years since the CEO started her/ his current position, CEO age is the CEOs cur- rent age, and CEO gender is set to 1 for male CEOs and 0 for female CEOs.
In this model, if CEO pensions intensify the relationship between R&D expenditures and firm value as hypothesized, the coefficient on β is expected to have a positive value.
The other variables are selected to control for factors that may affect the hypothesized relationship.
First, FSIZE is included, because larger firms tend to engage in more aggressive R&D activities that have positive firm value (Shehata, 1991).
LEV controls for the degree of a firm's financial distress (Bhagat & Welch, 1995).
, CEO tenure , CEO age and CEO gender are included to control for CEO-specific characteristics.Finally, the effects regarding the industry and the fiscal year are also controlled.All variables are winsorized at the top and bottom 1% to eliminate the impact of outliers.

Summary statistics
Table 1 presents the summary statistics for dependent and independent variables in the final sample of 8,707 firm-year observations.The dependent variable TobinQ shows a mean value of 1.566, implying that the sample firms consist of relatively large firms whose market value of equity is 1.566 times greater than its book value of assets.The implication is also confirmed by the large mean value of R&D (0.054).The mean value of INSDEBT is 0.28, indicating that 28% of CEOs in the sample are classified as those awarded with larger CEO pensions.However, it is low compared to the ratios reported by both Sundaram and Yermack (2007) and Bebchuk and Jackson Jr. (2005), possibly because the sample firms taken from the ExecuComp database include S&P 500 companies, which are smaller than those used by Sundaram and Yermack (2007) or Bebchuk and Jackson Jr. (2005).The summary statistics of other variables are generally similar to those in the prior literature mentioned above.

Correlation
Table 2 presents the Pearson correlations between the dependent and independent variables in the final sample.The coefficients and p-values show the general expectations on the hypothesized relationships based on the univariate analysis results.Specifically, it shows that the dependent variable TobinQ is positively related to R&D expenditures with the p-values of 0.000, indicating that R&D expenditures generally increase the firm value.INSDEBT is negatively correlated with TobinQ with the p-values of 0.000, implying that CEO pensions negatively contribute to firm value.However, the main independent variable,

&, R D INSDEBT ⋅
is positively correlated with TobinQ with the p-value of 0.007.These results are consistent with the hypothesis.

Endogeneity issue
Although the OLS regression using Equation ( 1) controls for various factors that might affect the hypothesized relationship, an endogeneity issue may arise if a selection bias exists due to observed firm characteristics that were inadvertently excluded from the controls.To alleviate this endogeneity concern, 2SLS and PSM regression analysis was also performed.The regression results are presented in Table 4.

Robustness test
As robustness tests, instead of using Tobin's Q ratio, ROA and F-score were used as the dependent variables to measure firm value.Since Tobin's Q ratio measures a firm's value based on stock market participants' expectation, it is known to have some negative reflection as a result of wrong expectations by stock market participants.To alleviate this concern, accounting performance measure ROA is often used as an alternative measure of firm value.Additionally, F-score is pervasively used to measure firm value.F-score is known to have several advantages over existing measures, such as Tobin's Q and ROA, because it does not depend on any corporate event, whereas Tobin's Q and ROA could be the result of the heterogeneity of corporate events (Chung et al., 2015).Specifically, F-score is calculated by combining three aspects of a firm's financial status: (1) profitability (net income, operating cash flow, ROA, and earnings quality); ( 2) liquidity (leverage, liquidity, and the absence of dilution); and (3) operating efficiency (gross margin and asset turnover).
The results with ROA and F-score as alternative dependent variables are presented in Table 5.Consistent with the hypothesis, it shows that the coefficients on & R D INSDEBT ⋅ are all significantly positive in both [Model 1] and [Model 2], supporting the argument that CEO pensions intensify the relationship between R&D expenditures and firm value.

Additional tests
As additional tests, we examine the effect of CEO turnovers on our hypothesis relationship.Since CEO turnover often initiates CEO pension plans, small amounts of CEO pensions in the year of CEO turnovers may not have enough power to induce CEOs to engage in more R&D activities with long-term perspective.
To test this conjecture, we search and find that 793 CEO turnover cases from our sample.Then, we perform subsample analysis by dividing our sample with and without CEO turnover.The results are presented in Table 6.Consistent with our conjecture, it shows that our hypothesized relationships are only significant for the sample without CEO turnover.In addition, we examine the effect of CEO gender on our hypothesis relationship.The literature finds that women have more conservative and long-term perspective than men.Therefore, the effect of CEO pensions to induce managers to engage in more conservative but long-term R&D investment can be offset among woman managers.We examine this conjecture.The results are presented in

CONCLUSION
This study proposed how debt-based CEO compensation affects CEO decisions about investment.Many researchers suggest that CEOs who receive CEO pensions make more conservative investments in order to protect their insider debt from bankruptcy and their aligned interests with creditors.
If CEOs are entitled to CEO pensions, their status and decision-making motives become similar to that of creditors, as both have a shared interest in reducing default risk.According to Cassell et al. (2012), firms that provide CEO pensions have a higher tendency to spend less on R&D expenditures; however, we considered value relevance of R&D expenditures and long-term performance.Investments, such as R&D expenditures, are necessary for long-term sustainability; therefore, we examined how to choose such investments to reduce the default risk in relation to long-term performance.Our results support our hypothesis that CEO pensions are more likely to concern R&D activities in an effort to increase long-term firm value.This implies that the CEO who receives the CEO pension is more likely to invest in the portion of the firm that is related to sustainability than the CEO who does not receive such R&D investment.In other words, CEOs appear to be more careful when choosing to invest in the necessary R&D expenditures and considering long-term performance in investment decision-making.Even when we look at the indicators ROA and F-score, which show shortterm performance, investment in R&D, which can reduce default risk and increase sustainability, can have a positive affect similar to that of long-term performance.In that regard, CEOs attempt to select activities that sustain their firms in order to ensure they receive their own pension fund.
We did not explore other meaningful ways of investment, but instead focused on R&D activities.To further examine long-term performance, we can consider that future decisions on capital investments, such as facility investment in addition to R&D investment, may affect larger or longer-term outcomes in future research.
If the CEO pensions reduce the default risk of firms and induce decision-making to improve sustainability, there may be many future discussions about the CEO compensation system that solve the agency problem between stock-based compensation and debt-based compensation.

Finally, in the
second stage, Equation (1) is re-estimated with the matched firms to have the closest predicted value between the two scores of firms with 1 INSDEBT = or 0. The first and second stage regression results are presented in [Model 3] and [Model 4], respectively.Fortunately, the regression results in [Model 4] remain qualitatively the same with the main regression results in Table 3.That is, the coefficient on & R D INSDEBT ⋅ is significantly positive, confirming that the main regression results are not contaminated by the endogeneity issue mentioned above.
compensation was greater than firm leverage.Specifically, CEO leverage is measured as the CEO pension amount divided by the CEO equity-based compensation.The CEO pension amount is the sum of the present values of the CEO pension and deferred compensation as reported in ExecuComp.The CEO equity-based compensation is calculated by adding the total values of option-based and stock-based compensation.Then, the ratio of CEO leverage to firm leverage is calculated as dividing the value of CEO leverage by firm leverage (total debt-to-equity ratio).Additionally, a dummy variable was created, INSDEBT, which is set to 1 when the ratio of CEO leverage to firm leverage is greater than 1, and 0 otherwise.Thus, CEOs where INSDEBT is equal to 1 are those awarded with larger CEO pensions.
Phan (2014)al.(2012)(2007)mans & Liu, 2011).It reasons that, unlike equity based compensation, which aligns the CEO's interests with those of shareholders by reducing the CEO's risk averseness, debtbased compensation matches the CEO's interests with those of outside creditors to minimize the firm's default risk.However, despite the importance of debt-based compensation to the design of CEO compensation structures, prior literature has paid less attention to the debt-based compensation, as compared to equity-based CEO pensions attract many researchers in examining whether CEOs who receive pensions make more conservative investment, financing, and accounting decisions to protect their insider debt from firm bankruptcy.For example,Sundaram and Yermack (2007)found that firms providing larger CEO pensions have lower bankruptcy risk than those without such pension plans.Wei and Yermack (2011) and Anantharaman, Fang, and Cong (2013) documented that firms with larger CEO pension plans are more likely to face lower cost of debt, but higher cost of equity, than those without such pension plans.Cassell et al. (2012)found that CEO pensions are negatively related to the volatility of the firm's future stock returns, R&D expenditures, and financial leverage ratio, but are positively related to the diversification and asset liquidity.Phan (2014)addressed that CEOs awarded with larger CEO pensions are less likely to perform mergers and acquisitions that may increase the firm's default risk.With respect to With INSDEBT as the main independent variable of interest, the following ordinary least square (OLS) regression model is developed after controlling for various firm-and CEO-related characteristics:

Table 1 .
Summary statistics 19)e: This table summarizes the descriptive statistics of the dependent/independent variables in the final sample of 8,707 firm-year observations.It presents the mean, median, standard deviation, 25th percentile, and 75th percentile values.The two databases used in this study are COMPUSTAT and ExecuComp over the period 2006-2015.http://dx.doi.org/10.21511/ppm.16(4).2018.19

Table 3
1 Since we use ordinary least square regression model which assumes residuals' normality, we examine the residuals' normality in our main regression model.The results show that D statistics from Kolmogorov-Smirnov, W2 statistics from Cramer-von Mises, and A2 from Anderson-Darling are all indicating that our residuals are normally distributed.term between R&D and INSDEBT in [Model 3], which is significantly positive at 1% significance level.The results remain robust in [Model 4] of Fama-MacBeth regression analysis to alleviate time-series dependence on the hypothesized relationship.Overall, these results support the hypothesis, namely that CEO pensions intensify the positive relationship between R&D expenditures and firm value.

Table 3 .
Regression of Tobin's (1969) Q ratio on R&D expenditures and CEO pensions

Table 4 .
Addressing endogeneity This table presents the results using two-stage least square (2SLS) and propensity score matching (PSM) regression analysis.[Model 1] and [Model 3] present the first-stage regression results of 2SLS and PSM, respectively.[Model 2] and [Model 4] present the second-stage regression results of 2SLS and PSM, respectively.The t-statistics are reported in parentheses.***, **, and * indicate the statistical significance at the 1%, 5%, and 10% level, respectively.

Table 5 .
Robustness test Note: This table presents the regression results with return on asset (ROA) and Piotroski's (2000) F-score as the dependent variables.The t-statistics are reported in parentheses.***, **, and * indicate the statistical significance at the 1%, 5%, and 10% level, respectively.

Table 6 . CEO turnover Independent variables Dependent variable: TobinQ Subsample with CEO turnover = 1 Subsample with CEO turnover = 0
Note: This table examines the effect of CEO turnover on our hypothesis relationships.The t-statistics are reported in parentheses.***, **, and * indicate the statistical significance at the 1%, 5%, and 10% level, respectively.

Table 7 .
CEO genderNote: This table examines the effect of CEO gender on our hypothesis relationships.The t-statistics are reported in parentheses.***, **, and * indicate the statistical significance at the 1%, 5%, and 10% level, respectively.