*4.1. Baseline Model*

The primary purpose of this research is to investigate the influence of executive pension incentive on firms' active management of risk. Utilizing the final sample of multinational firms that are exposed to foreign currency risk, we first implement the baseline model regressions of hedging activity on a series of regressors, which contain the measures of pension incentives as the main independent variables, equity incentive variables, the proxies for other hedging drivers discussed in Section 3.4, and control for heterogeneity across industries and time. The baseline model is specified as follows:

$$\begin{aligned} \text{Hodge}\_{l,t} &= \begin{aligned} \alpha + \beta \star \text{PersonUnenceive}\_{l,t-1} + \gamma\_1 \text{EquityIncomeive}\_{l,t-1} + \gamma\_2 \text{Leverage}\_{l,t-1} + \gamma\_3 \text{Indirect Sex}\_{l,t-1} \\ &+ \gamma\_4 \text{Tangible}\_{l,t-1} + \gamma\_5 \text{Market} / \text{book}\_{l,t-1} + \gamma\_6 \text{Corr}(\text{CF}, \text{Investment})\_{l,t-1} + \gamma\_7 \text{TxDummy}\_{l,t-1} \\ &+ \gamma \text{y3TaxCorxity}\_{l,t-1} + \gamma \text{cCsshHdyling}\_{l,t-1} + \gamma\_{10} \text{CoveribleBond}\_{l,t-1} + \gamma\_{11} \text{Ln}(\text{TotalAssets})\_{l,t-1} \\ &+ \sum\_{k=1}^{47} \text{Industry}\_{k} + \sum\_{l=1}^{10} \text{Year}\_{l} + \varepsilon\_{l,t} \end{aligned} \tag{1}$$

where hedging, the dependent variable, is measured by the dummy variable of hedging or the continuous variable of hedging intensity. The measures of pension incentives include the logarithm of CEO pension benefits in dollar amount, the ratio of CEO pension benefits to CEO total compensation, the ratio of CEO pension benefits to CEO equity compensation. Additionally, following the literature, the continuous variable of CEO pension relative leverage and the dummy variable of CEO pension relative leverage greater than one are constructed.<sup>2</sup> As to CEO equity-based compensation, the incentives of stock and option compensation are gauged by Delta incentive, Vega incentive, and a ratio of Vega incentive to Delta incentive.

To analyze the impact of pension incentives on the hedging adoption, we first perform a Probit model analysis by regressing the dummy variable of hedging on the explanatory variables specified in Equation (1). The results are reported in Panel A of Table 3. To facilitate the interpretation of regression results, we report the incremental effects on the probability of implementing hedging strategy for a one standard deviation change in continuous explanatory variables. However, the incremental effects for the logarithmic variables (i.e., the logarithm of CEO pension relative leverage) or the incremental effects for the dichotomous variables (i.e., the dummy variable of CEO pension relative leverage greater than 1 or the dummy variable of positive tax credit) follow the traditional pattern.<sup>3</sup> Z-values based on robust standard errors are reported in parentheses to test whether the coefficients are equal to zero. Consistent with the theoretical predictions provided on the sign column, we first observe strong evidence across models that CEO pension incentives have a significantly positive influence on the decision of currency hedging. From Model 1 through Model 3, we look at the effect of pension incentives by directly examining the dollar amount of CEO pension incentive, the proportion of pension benefits to total compensation, and the proportion of pension benefits to equity-based compensation. The coefficients of pension incentive in these models all show significant and positive relations with corporate hedging decision. For example, a one-standard-deviation increase in the overall pension benefits in dollar amount leads to an increase of 2.2% in hedging probability. Similarly, a one-standard-deviation increase in pension proportion in CEO total compensation (equity

<sup>2</sup> We also test the alternative measures, such as the CEO pension relative leverage and the squared CEO pension relative leverage, and we find general consistent results. However, considering the potential skewness in the variable of CEO pension relative leverage, we adopt the logarithm form of CEO pension relative leverage or use the conversion to a dummy indicator of CEO pension relative leverage.

<sup>3</sup> Specifically, for each of continuous explanatory variables we multiply the coefficients of the regressions by the standard deviations of each independent variable, while for the logarithmic variables and the dichotomous variables we directly report the coefficient of estimation.

compensation) increases the hedging probability by 6.5% (4.8%). More importantly, the effect of CEO pension incentives on hedging probability is economically meaningful. In Model 4 the continuous measure of CEO pension relative leverage has a positive and significant coefficient (0.006), suggesting that a percentage increase in CEO pension relative leverage leads to about 1.6% increase in the likelihood of hedging. In Model 5, we find when the CEO pension relative leverage is greater than one, that is, when CEO's wealth "leverage" is greater than firm's capital "leverage", the probability of corporate hedging is higher by 20.9%.


**Table 3.** Baseline Models of Impact of Pension Incentive on Hedging.


**Table 3.** *Cont*.

This table shows the results of baseline model regressions of hedging activity for the sample of 3492 firm-years from 2006 through 2010. Panel A reports the effects on the probability of hedging from Probit model for a one standard deviation change in continuous explanatory variables (or for a change from zero to one in dummy explanatory variable). In Panel B we apply Tobit model to regress the total notional amount of hedging positions scaled by book value of total assets on the set of explanatory variables. We control for industry effects by using the Fama-French 48-industry classification and control for time effects by using year dummies. The robust standard errors are used to calculate Z-Statistics (Probit model) or T-Statistics (Tobit model) that are reported in parentheses below estimates. \*, \*\* and \*\*\* denote significance at the 10%, 5% and 1% levels, respectively.

For firm-related variables, the results are generally consistent with the literature. Leverage has the positive impact on hedging, and both interest coverage and tangible assets are negatively related with hedging. We also find a significantly negative association between hedging and the correlation of cash flow and investments, suggesting that firms with a lower degree of matching between cash flows and investment needs tend to hedge more. Furthermore, we find evidence to support the potential tax benefits of hedging. For example, the dummy variable of positive tax credit has positive and significant effect on hedging. Lastly, we find that cash holding and convertible bonds are both negatively associated with hedging, indicating that they may serve as possible substitutes for hedging.

Liability-based incentive urges executives' inclination to take risk-reduced investment, i.e., more likely to hedge and increase the magnitude of hedging activity. To detect how pension incentive affects the intensity of hedging, we turn to an investigation by applying the continuous measure of hedging activity. Specifically, we replace hedging dummy variable in Equation (1) with the notional amount of hedging positions scaled by the book value of total assets. We keep the same set of explanatory variables. Given the nature of non-negative notional amount of hedging position, we adopt Tobit regressions for our baseline models. The results are shown in Panel B of Table 3. Across the models, we find a significantly positive relation between the notional amount of hedging and all measures of pension incentives. The positive relations between pension incentive and hedging notional amount to total assets ratio are detected from Models 1 through 4. For example, we observe from Model 1 that

firms increase about USD 81 million (=0.009 \*USD 8987 million) when pension compensation increases one-standard-derivation or equivalently USD 7.59 million. In addition, the coefficient (0.004) in Model 4 shows that for a one percent increase in CEO pension relative leverage, the hedging position rises by 40 basis points of total assets (0.4%), and in other words, firms on average increase their hedging positions by USD 35.95 million (=0.04%\*USD 8987 million). Finally, this significant and positive relation also holds in Models 5, when CEOs' wealth leverage is greater than firms' leverage, the ratio of hedging position to total assets is higher by 1.8%. Considering that the average hedging notional amount to total assets ratio is 2.9% (=USD 260.70/USD 8987), the change of the CEO pension relative leverage from less than or equal to one to greater than one boosts the hedging intensity by more than 60% (1.8%/2.9% = 62%). Overall, the results of baseline models provide supportive evidence to our first hypothesis.

#### *4.2. Alternative Models to Control for Equity-Based Incentives*

While the focus of this research is not about equity-based incentive, it may be arguable that equity-based incentive is designed differently for the sample firms that are more/less likely to do currency risk management. From our baseline models, we find that Vega incentive is negatively associated with hedging decision, but an insignificant impact of Delta incentive on hedging decision is also observed. Although the results of equity incentives seem mixed from the theoretical predictions, these discordant results are also reported in the literature. For instance, Guay (1999) illustrates that shareholders should consider the slope and convexity of the relation between executives' compensation and firm performance, where slope (Delta) refers to the sensitivity of executives' compensation to stock price and convexity (Vega) refers to the sensitivity of executives' compensation to stock return volatility. As a result, granting equity compensation to executives should, ceteris paribus, generate incentives to take more risks, or equivalently hedge less. Knopf et al. (2002) provide the evidence to support the positive association between hedging and the sensitivity of a manager's equity wealth to stock price and the negative association between hedging and the sensitivity of a manager's equity wealth to stock return volatility. Rajgopal and Shevlin (2002) report a significantly negative relation between risk-taking incentives (Vega) and commodity hedging for a sample of gas and oil firms.

But in a similar setting, Rogers (2002) only find weak evidence of Delta and Vega incentives which affect the hedging decision. Accordingly, Rogers (2002) proposes economic interpretation of the ratio of Vega to delta should be more intuitive because it measures the CEO risk-taking incentive per dollar of value-increasing incentives from option and stock holdings that delta and Vega measure managerial motivation from "value-creating" and "risk-taking" incentives, respectively. On the other hand, Lewellen (2006) demonstrates that the "moneyness" of stock options has different impacts on managerial risk attitude. In particular, those in-the-money options discourage risk-taking. Meanwhile, as emphasized by Carpenter(2000) equity compensation does not necessarily result in more risk-taking behavior because it makes executives' wealth more vulnerable to stock price fluctuation. Hirshleifer and Suh (1992) also discuss the side effect of equity compensation, which encourages executives to work hard but also affects their attitude toward project risks, ending up with less risk-taking. Gerakos (2010) provides some evidence that executives granted with more pension benefits are paid less on other dimensions of compensation.

To ensure that the results of the baseline model are not vulnerable to the potential inconsistent impact of equity-based incentive, we perform the alternative models to recognize the evidence from Carpenter (2000). The results of the alternative models are shown in Table 4 and report the regressions by adopting the ratio of Vega incentive to Delta incentive. As expected, we find the significant and negative coefficients of this variable across the models. For example, the ratio of Vega incentive to Delta incentive are negatively associated with both hedging propensity in Model 3 of Panel A (−0.226) and hedging intensity in Model 3 of Panel B (−0.028). With this new control variable in Table 4, the coefficients of pension incentive measures keep positive. Notably, the effect of pension incentive on hedging propensity and on hedging intensity are also statistically significant, consistent with the

finding in Table 3. To this end, the alternative models help confirm that hedging activity is plausibly affected by the executive pension incentive.


**Table 4.** Alternative Models of Impact of Pension Incentive on Hedging.

This table shows the results of the alternative models of pension incentive on hedging activity by recognizing a different measure of equity incentives (the ratio of Vega incentive relative to Delta incentive). Panel A reports the effects on the probability of hedging from Probit model for a one standard deviation change in continuous explanatory variables (or for a change from zero to one in dummy explanatory variable). In Panel B we apply Tobit model to regress the total notional amount of hedging positions scaled by book value of total assets on the set of explanatory variables. We control for industry effects by using the Fama-French 48-industry classification and control for time effects by using year dummies. The robust standard errors are used to calculate Z-Statistics (Probit model) or T-Statistics (Tobit model) that are reported in parentheses below estimates. \*, \*\* and \*\*\* denote significance at the 10%, 5% and 1% levels, respectively.
