Printer Friendly

The effects of liquidity shocks on corporate investments and cash holdings: evidence from actuarial pension gains/losses.

This paper investigates how anticipated liquidity shocks affect corporate investment and cash holdings by examining the impacts of actuarial pension gains/losses that do not reduce current internal resources but will reduce those available in the future. Using a sample from Japanese manufacturing firms in which pension deficits had a huge impact on the internal resources of sponsoring firms, I show that pension losses significantly decrease the capital expenditures of sponsoring firms. Pension losses also increase corporate cash holdings, suggesting precautionary demands for cash prepared for future pension contributions. Overall, the results indicate that managers consider anticipated liquidity shocks in determining current investment and cash-saving policies.

**********

In recent years, pension liabilities have had a large impact on corporate liquidity management among firms in developed economies. (1) Rauh (2006) investigates the impact of mandatory contributions (MCs) made by firms with heavily underfunded plans and finds that mandatory pension contributions significantly reduce the capital expenditures of sponsoring firms and that this relationship is more prominent for firms that appear to face financing constraints. (2) More recently, after demonstrating that MCs depress the liquidity and investment of sponsoring firms, Phan and Hegde (2013) find that freezing defined benefit (DB) pension plans reduce the negative impact of MCs on liquidity. However, Bakke and Whited (2012) argue that the managers of DB-sponsoring firms can anticipate the outcome of underfunded pension plans and optimize their funding policy subject to the constraints imposed by threshold rules and endogenously choose whether they want to be close to the threshold. (3-4)

I contribute to the literature by examining the impacts of anticipated liquidity shocks associated with realized pension losses on current investment. In other words, I focus on the timing of the emergence of the deficits rather than on the timing of the required contributions that are enforced as a consequence of pension deficits. Thus, this paper complements the findings of Rauh (2006) by demonstrating how anticipated liquidity shocks affect current investment of sponsoring firms. This paper also contributes to the growing literature that investigates the motives for cash holdings (Opler et al., 1999; Almeida, Campello, and Weisbach, 2004; Bates, Kahle, and Stulz, 2009; Iskandar-Datta and Jia, 2012) by examining the impacts of anticipated liquidity shocks associated with pension losses on cash holdings.

Although the emergence of pension deficits does not necessarily reduce the amount of internal resources immediately, it will reduce the amount of internal resources available in the future because the sponsoring firms should fund the deficits to recover funding status according to the funding rules regulated by the government. If firms rationally anticipate such future required contributions, sponsoring firms may reduce investments before additional contributions are required because firms may optimize their investments subject to the constraints imposed by the funding rules (Bakke and Whited, 2012). Furthermore, the development of pension deficits may increase the precautionary demand for cash to prepare for possible future cash flow shortfalls associated with future contributions. I verify these hypotheses using a sample of publicly traded Japanese manufacturing firms during the 2001-2011 fiscal years in which the majority of Japanese sponsoring firms have heavily underfunded pension plans. This is important because Rauh's (2006) findings may stem from a small number of heavily underfunded firms whose attributes are significantly different from those of other firms (Bakke and Whited, 2012).

After controlling for time-varying investment opportunities and cash flows as well as firm fixed effects, I find that pension gains/losses that do not immediately result in required contributions significantly affect the investment behavior of sponsoring firms. I find an asymmetric effect of pension gains/losses on the investments of sponsoring firms such that while pension gains do not necessarily increase the capital expenditures of plan sponsors, pension losses significantly decrease the capital expenditures of the plan sponsors. This is probably because while pension losses will increase future contributions, pension gains do not reduce future contributions if the pension plans are underfunded.

I also find that the relationship between pension deficits and investment is more prominent among large manufacturing firms than small manufacturing firms. This finding is most likely because large manufacturing firms have their own pension plans and promise their employees larger retirement benefits. I argue that the fluctuations in capital expenditures associated with pension gains/losses may contribute to the loss of competitiveness in representative Japanese manufacturing firms in the 2000s. I show that the impact of pension losses on investment is more prominent among financially constrained large firms, suggesting that financially constrained sponsoring firms adjust their investment with consideration for the future contributions associated with recognized pension losses.

Finally, I document that pension losses significantly increase cash holdings of sponsoring firms. This result indicates that firms save internal resources to prepare for future contributions and severance payments. My results suggest that the precautionary demands for cash prepared for anticipated liquidity shocks associated with pension losses may partly explain the increases in cash holdings among companies in developed economies.

Overall, anticipated liquidity shocks appear to affect current investment and cash-saving behavior of sponsoring firms. This is consistent with the argument of Bakke and Whited (2012) that firms anticipate and actively manage future MCs and may endogenously choose the timing and amount of the contributions. Because other theories cannot consistently explain the findings, I argue that capital market imperfections make sponsoring firms consider the impacts of anticipated liquidity shocks in determining current investment policies. Hence, this paper contributes to the literature that examines the impacts of internal resources on investment (for a survey, see Hubbard, 1998 and Stein, 2003) by showing that the amount of anticipated liquidity in the future impacts current investment. My findings also relate to the debt overhang literature (Myers, 1977) because the development of pension deficits can be regarded as a debt overhang problem from the perspective of new investors. Finally, my results suggest that precautionary demands for cash associated with pension losses may partly explain increases in corporate cash holdings.

The paper is structured as follows. I first briefly review the literature and develop the hypotheses in Section I. Next, I describe the data and methodology in Section II. Section III presents the empirical results, and Section IV presents the paper's conclusions.

I. Institutional Background and Hypotheses

A. Institutional Background

As discussed in McLellan (2005), the Japanese corporate postretirement system has a long history and almost all large Japanese firms have at least one DB postretirement plan. While severance payments began to spread between 1910 and 1920, the origins of the system date back to the traditional lump-sum retirement allowance in the Edo period (1603-1867) that was a token of appreciation for long-term hard work. According to the survey by the Ministry of Health, Labor, and Welfare of Japan, the proportion of plan sponsors among Japanese firms with more than 1,000 employees is 78.6% for severance payment amounts and 82.0% for DB pension plans at the beginning of 2003.5

Mainly because of the long-term stock market stagnation in the past 20 years, the funding levels of Japanese sponsoring firms are among the lowest of developed economies, as indicated in Yermo and Severinson (2010). By my estimations, the average pension deficit to market value of equity ratios were at their highest level at 20% in fiscal year 2002. Thus, as severe underfunding is common among Japanese companies, analyzing Japanese corporate pensions presents a unique opportunity to understand the impact of pension deficits on investment.

Furthermore, it is also worthwhile to note that funding regulations for DB pension plans in Japan are not as rigid as those in the United States (Standard & Poor's, 2006). Specifically, whereas US firms are required to make MCs to their pension plans in the case of severe underfunding, Japanese firms have a moratorium on recovering the funding status of their pension plans. For example, with respect to employee pension funds (EPFs), which had been the most popular DB pension plan among Japanese firms, sponsoring firms with underfunded plans are required to recover the funding status of underfunded pension plans for 10 years. (6) Thus, an analysis of a Japanese sample would present evidence regarding the reaction of firms to pension losses that do not immediately reduce the internal resources of sponsoring firms.

B. Prior Literature and Hypotheses

Raising capital from capital markets may impose on the firm a higher cost for capital than internal resources mainly due to asymmetric information (Myers and Majluf, 1984), agency costs, and incomplete contracting of contracts when issuing new debt or equity securities. Rational investors expect that only the worst types of firms would raise external funds and thus require an additional cost of capital for firms raising capital externally, thus resulting in the firm preference for internal resources. In other words, capital markets' imperfections make firms rely on internal resources in financing projects (Fazzari, Hubbard, and Petersen, 1988; Kaplan and Zingales, 1997; Lamont, 1997; Rauh, 2006). On the other hand, sponsoring a DB pension plan reduces such internal resources when pension deficits result in additional contributions to compensate the deficits. In developed economies, DB sponsoring firms must compensate the deficits according to the funding rule specified by the government. Rauh (2006) makes the first contribution to understanding how pension deficits affect corporate investment decisions. Concretely, he focuses on mandatory pension contributions as a natural experimental setting because MCs can be regarded as exogenous shocks to internal resources. (7) He finds the negative association between mandatory pension contributions and firm investment levels, an association that is especially evident among firms that are likely to face financing constraints, which is measured by credit rating, dividend ratios and other measures. More recently, Phan and Hegde (2013) investigate the impact of freezing DB pension plans and replacing them with defined contribution (DC) plans on the corporate behavior of sponsoring firms. They first find that MCs depress the liquidity and investment of sponsoring firms. Furthermore, they find that freezing DB pension plans attenuate the negative impact of MCs on liquidity. Relatedly, Franzoni (2009) finds a negative association between mandatory pension contributions and stock returns over the subsequent 12 months and further finds that this association is stronger for firms that are more financially constrained. On the other hand, Campbell, Dhaliwal, and Schwartz (2010) examine the relationship between capital expenditures and abnormal returns surrounding key dates in the legislative process that led to the adoption of the Pension Protection Act of 2006 (PPA 2006), which accelerated near-term cash outflows to DB pension plans by sponsoring firms. They find that firms with greater investment requirements are associated with more negative abnormal returns than are firms with relatively smaller ones. Relatedly, Campbell, Dhaliwal, and Schwartz (2012) find that an increase in mandatory pension contributions increases the cost of capital for firms facing greater external financing constraints. They argue that the firm cost of capital is an intervening variable that can explain Rauh's finding.

These studies indicate that DB sponsoring changes the investment behaviors in response to required contributions. However, sponsoring firms may also change their investment behavior before additional contributions are required. Although newly developed pension deficits themselves may not immediately affect the amount of internal resources available today, they will affect those available in the future because these deficits must be compensated by the sponsoring firm in the following years. Managers of these firms may optimize their investments subject to the constraints imposed by the funding rules (Bakke and Whited, 2012). Furthermore, sponsoring firms may voluntarily make contributions before required contributions to pursue tax-favorable treatments associated with pension plans. Thus, I expect that the development of pension deficits decreases investment, making it possible to prepare for future contributions. On the other hand, the development of pension deficits can be regarded as a debt overhang problem (Myers, 1977) from the perspective of new investors. In firms with pension deficits, benefits from new investments will go to existing creditors of pension benefits, that is, to employees rather than to new investors. New investors will anticipate this and require higher expected returns for firms with pension deficits. Thus, I expect that the debt overhang effects associated with pension deficits also contribute to the negative relationship between pension deficits and investments. Based on these reasons, I hypothesize that pension losses that do not immediately result in required contributions nevertheless deter managers of DB sponsoring firms from financing investment opportunities that otherwise would be exploited. In addition, sponsoring firms with pension deficits may prefer to hold cash to prepare for possible future cash flow shortfalls associated with future contributions. Thus, I also hypothesize that pension losses increase corporate cash holdings in preparation of future cash flow shortfalls associated with future contributions.

II. Research Design and Sample Construction

A. Research Design

Following the investment-cash flow literature, I estimate the following augmented investment-cash flow model that includes pension gains/losses as an explanatory variable. Thus, I have:

[Cape.sup.j,t] = [[alpha].sub.j] + [[beta].sub.1][Q.sub.j,t-1] + [[beta].sub.2] [Cashflow.sub.j,t] + [[beta].sub.3][PenGL.sub.j,t-1] + [v.sub.j,t], (1)

where [Capex.sub.j,t] is capital expenditure normalized by total assets at the beginning of the year for firm j in period t. I also submit regressions in which I use total investment (Inwall) consisting of capital expenditure, research and development (R&D), and advertisement expenditure as the dependent variable rather than Capex. I also estimate regressions in which changes in cash holdings (d_Cash) is used as the dependent variable to directly examine precautionary demands for cash holdings to prepare for future pension contributions. [Q.sub.j, t-1] captures the time-varying values of investment opportunities and is defined as the sum of the market value of shareholder equity and the book value of debt divided by total assets for firm j at the beginning of the year in period t. [Cashflow.sub.j, t] denotes the gross cash flows, which is defined as the sum of net income and depreciation and amortization adding back R&D and also scaled by total assets at the beginning of the year. I add back R&D expenditures because R&D expenditures are not necessarily determined before capital expenditures. [PenGL.sub.j, t-1] is the variable of interest and constructed as the difference between the amount of accumulated pension gains/losses before recognition at t - 1 and the amount of accumulated pension gains/losses at t - 2. Concretely, this is calculated as the changes in the amount of accumulated gains/losses adding back the recognition of gains/losses at t - 1. I take a one-year lag for pension gains/losses because the management of sponsoring firms can measure the amount of pension gains/losses after the complex actuarial calculation of the status of the pension plans based on the market value of plan assets and interest rates to evaluate the pension benefit obligation (PBO) at the end of the year. (8) Variations in PenGL are mainly due to the changes in the market value of dedicated pension assets and occasionally due to changes in the discount rate to evaluate the present value of pension obligations, which reflect the interest rate changes of straight bonds with high rating grades. Thus, variations in PenGL are considered as exogenous shocks to the net pension deficits and thus to the amount of internal resources available in the future.

I further estimate the specification in which I decompose PenGL based on the sign of PenGL to allow for pension gains and pension losses to have different coefficients. Specifically, PenGL_Pos equals PenGL for positive PenGL observations and zero for negative PenGL observations. PenGL_Neg is defined in a similar way (it equals PenGL for negative PenGL observations and zero for others).

[Capex.sub.j,t] = [[alpha].sub.j] + [[beta].sub.i] [Q.sub.j,t-1] + [[beta].sub.2] [Cashflow.sub.j,t] + [[beta].sub.3] NewGL-[PoS.sub.j,t-1] + [[beta].sub.4]NewGL_[Neg.sub.j,t-1] + [v.sub.j,t]. (2)

In estimating Equations (1) and (2), I do not include time dummies because variations of PenGL mainly comes from fluctuations of share prices and inclusions of time dummies will substantially underestimate the effect of pension deficits on corporate investments. (9)

B. Sample

The sample for this study consists of manufacturing companies listed on the Tokyo Stock Exchange. I collect data from the Nikkei Financial Quest database for the fiscal years 2001-2011 and require the firms to have nonmissing values for total assets, cash flows, and Tobin's q to be included in the sample. The above procedures result in 6,741 firm-years (657 firms). To examine the impact of pension deficits, I exclude firms that do not have a DB postretirement plan during the sample period. To accomplish this, I exclude those firms for which the PBOs have been zero during the sample period. This results in the exclusion of only 14 observations (two firms), indicating that almost all listed Japanese manufacturing companies have DB postretirement plans. I now have 655 firms and 6,727 observations. I further confine the sample to those firms having at least three positive capital expenditure years. As a result of this, the sample is reduced to 6,698 firm-year observations (637 firms). I winsorize all variables included in the empirical analyses at the top and bottom 1% to protect the sample from the influence of outliers.

Table I reports the descriptive statistics for the variables used in the regressions and related variables. The mean of Capex is 0.043, with a relatively large standard deviation of 0.031. The average of total investment (Invall) is 0.078, suggesting that investments other than capital expenditures account for a significant portion of the corporate investments among the sample. On the other hand, the average of djCash is 0.004, lower than that of Capex, while the standard deviation of d_Cash is 0.042, which is larger than that of Capex.

The mean of PenGL is -0.004, suggesting that pension losses are more frequent than pension gains during the sample period. The standard deviation of PenGL is 0.016, which is approximately half that of Capex, while the average of PBO is 0.132, indicating that pension liabilities have a substantial impact on sponsoring firm capital structure. On the other hand, the average of dedicated pension assets (PASS) is 0.072, which is far smaller than that of PBO, thus indicating severe underfunding of Japanese corporate DB plans.

Table II reports the descriptive statistics for the subsamples classified by firm size. Specifically, I classify firms with total assets above the 70th percentile as large and those below the 30th percentile as small. The average of Capex for large firms is 0.051, which is substantially larger than the corresponding figure of small firms, 0.037. A similar argument also applied for Invall, indicating that large manufacturing firms are more aggressive in investment than small manufacturing firms in Japan. The average PBO is larger for large firms than small firms as expected, implying that large firms on average provide their employees with more benefits than small firms and hence tend to experience liquidity concerns induced by DB pension plans. In fact, the average of the PBOs divided by the number of employees is 6.2 million yen (approximately $50,000) for large firms and 4.7 million yen for small firms, indicating that employees of large companies receive higher pensions. On the other hand, the difference in the average of PASS (pension assets) is also substantial, suggesting that small firms tend to have a tax-unfavorable lump-sum payment, for which the assets are funded within the sponsoring firm and are not contained in the pension assets. Thus, I expect that large firms are more likely to face financial management problems associated with pension deficits.

III. Results

A. Pension Gains/Losses and Capital Expenditures

Table III displays the basic estimation results for the capital expenditure equation for the full sample. Columns 1 and 2 report the estimation results using [PenGL.sub.t-1], and columns 3 and 4 use [PenGL.sub.t] to compare the impacts of the realized pension gains/losses in the year t - 1 and the concurrent pension gains losses in the year t. In columns 2 and 4, I include PenGL__Pos and PenGL_Neg instead of PenGL to allow pension gains/losses to have different coefficients depending on their signs. I first check the coefficients of the basic determinants of investment. The coefficients of Q and Cashflow are positive in the expected sign and are statistically significant at the 1% level in all specifications.

In the regression in which I insert PenGL of the year t - 1 (column 1), the coefficient of PenGL is positive and statistically significant at the 1% level, suggesting that realized pension gains/losses in the year t - 1 affect the investment behavior of sponsoring firms in the year t. The point estimate indicates that a $1 increase in pension deficits results in a $0,053 decrease in capital expenditures. Although smaller than the coefficient for MCs reported in Rauh (2006), this indicates that sponsoring firms also decrease their investments with the emergence of pension deficits.

In the specification for which I allow pension gains/losses to have different coefficients depending on the sign of pension gains/losses (column 2), the coefficient for PenGL_Neg is statistically significant, whereas that for PenGL_Pos is insignificant, indicating that the effect of pension gains/losses on investment is found only for pension losses. In other words, sponsoring firms decrease capital expenditures in response to shocks that exacerbate the funding status of pension plans, but they do not necessarily increase capital expenditures in response to shocks that improve funding status. This is possibly because sponsoring firms, in response to pension gains, cannot reduce their future contributions if their pension plans are underfunded. In contrast, the incidence of pension losses among underfunded pension plans will cause an increase in future contributions to recover funding status. Because almost all of the sample firms had underfunded pension plans during the sample period as shown in Table II, such asymmetric impacts between pension gains and losses on future contributions created the differentiated results for PenGL_Pos and PenGL_Neg. It should also be noted that sponsoring firms have limited access to overfunded assets dedicated for their DB plans, whereas they have to compensate for the deficits in cases of underfunding. The point estimates in column 2 indicate that a $1 loss results in a $0.07 decrease in capital expenditures. Thus, I have demonstrated that pension deficits that do not necessarily result in immediate reductions in internal resources also deter the management of sponsoring firms from financing projects.

Next, I check the results for the specifications for which I insert pension gains/losses at t rather than t - 1 (columns 3 and 4). If the association between PenGL and investment behavior reflects the sponsoring firm financial performance or economic conditions, I expect [PenGL.sub.t] rather than [PenGL.sub.t-1] to have a larger positive impact on investment. However, if concerns for future contributions in response to the realization of pension losses deter management from exploiting investment opportunities, [PenGL.sub.t-1] will have a stronger positive effect on investment than [PenGL.sub.t]. The results are inconsistent with the former argument. (10) On the other hand, the coefficients of PenGL_[Neg.sub.t] are not statistically significant (column 4), which is quite different from the statistically significant coefficients of PenGL_[Neg.sub.t-1] reported in column 2. This result is also inconsistent with the former argument. Overall, the results in Table III indicate that the observed relationship between pension gains/losses and investment is unlikely to be the result of the financial performance of sponsoring firms or economic conditions because financial performance and economic conditions have more of a positive relationship with current pension gains/losses than lagged pension gains/losses. I argue that the positive and statistically significant coefficients of [PenGL.sub.t-1] suggest that anticipated liquidity concerns attributed to realized pension losses decrease current investment. Considering these results, I use pension variables in the year t - 1 rather than t in the following tables.

B. Firm Size and Pension Gains/Losses Investment Sensitivity

In this subsection, I divide the sample into subsamples depending on firm size, and reestimate Equations (1) and (2) for each subsample. Specifically, I classify firms with total assets above the 70th percentile as large and those below the 30th percentile as small. I mainly focus on large firms because large firms in general have their own pension plans and thus determine funding policies by themselves. Furthermore, large firms promise their employees large amount of pension plans and tend to have large PBO, as shown in Table II. Contrarily, small firms generally have cooperative pension plans with firms in the same industries and thus do not have discretion over funding policies. Relatedly, the amount of pension benefits is generally small among small firms. Therefore, large firms are a more suitable sample to verify the hypotheses.

Table IV presents the results. Columns 1 and 2 report the estimation results for small firms and columns 3 and 4, large firms. Interestingly, the results indicate that large manufacturing firms suffer severely from the impact of pension deficits regardless of their relatively good access to external capital markets. In fact, whereas the coefficients on PenGL are not statistically significant for small firms, those for large firms are statistically significant and more economically significant than those for the full sample reported in Table III. I argue that anticipated liquidity shocks associated with realized pension losses depress investment of large firms.

On the other hand, the observed impacts of pension losses on investment might be brought by the extent of overinvestment rather than liquidity constraints. In fact, the average of Capex is larger for large firms than for small firms and thus there might be the case that pension losses reduce overinvesting. To verify whether overinvesting drives the results, I compare the attributes possibly associated with overinvesting. (11) I first check average director ownership ratios (Dir). In the sample, Dir of small firms is 4.0%, whereas that of large firms is 0.8%, suggesting that both small and large firms in the sample do not have concentrated ownership by board of directors. In contrast, the average market value of equity owned by each director (Own) is larger at the large firms: 91 million yen at small firms and 301 million yen at large firms. I find a similar difference for the market value of equity owned by chief executive officers (CEOs) (CEO_Own): 483 million yen at small firms and 1,603 million yen at large firms. Furthermore, the percentage of firms with a managerial stock option plan (SO) is 21.8% for small firms and 38.8% for large firms. Thus, it is unlikely that managers of large firms have fewer incentives to pursue shareholder value than managers of small firms. Furthermore, large firms on average have higher institutional investor ownership ratios (Inst) and foreign investor ownership ratios (Frgn). Specifically, the average of Frgn is 23.7% for large firms and 6.9% for small firms. As for institutional ownership ratios, including foreign investors, the mean for large firms is 38.3% and 14.7% for small firms. Relatedly, the percentage of outside directors on the board (Idrto) is higher at large firms: 11.5% for large firms and 7.7% for small firms. (12) Thus, I argue that managers of large firms are more likely to be monitored by shareholders than managers of small firms. Consistently, the average Q ratio (0 is also higher for large firms (1.22 for large firms and 1.04 for small firms). (13) Overall, the additional evidence indicates that liquidity constraints associated with pension losses rather than overinvestment drive the results for large firms. (14)

With respect to the regression results for which I separate PenGL depending on its sign (Table IV, columns 2 and 4), the coefficients of PenGL_Neg are positive and statistically significant only for large firms. On the other hand, the coefficients of PenGL_Pos are not statistically significant as shown in Table III. The point estimate in column 4 suggests that a $1 increase in pension losses results in a $0.13 decrease in capital expenditures among large firms. These results are consistent with articles in Japanese newspapers that report financial difficulties for representative Japanese firms induced by underfunded DB pension plans. I argue that investment constraints resulting from pension deficits may be an important factor that explains the decreases in competitiveness among representative Japanese manufacturing firms in the 2000s.

C. Pension Gains/Losses and Cash Savings

Thus far, I have demonstrated that pension deficits that do not immediately result in required pension contributions reduce investments. In other words, anticipated liquidity shocks reduce current investment. One possible explanation for this is that managers of sponsoring firms anticipate future additional contributions and prefer to save cash to prepare for possible cash flow shortfalls associated with future pension contributions. To test this possibility, I submit regressions in which the dependent variable is changes in cash holdings (d_Cash) rather than capital expenditures. In this analysis, it should be noted that increases in cash holdings may partly reflect increases in internal funding prepared for severance payments without tax-preferable treatments. In this severance payment scheme, firms save dedicated assets within the firm and directly pay lump-sum payments to their employees upon retirement. On the other hand, such internal funding is not incorporated in pension assets, and thus, such postretirement schemes affect pension gains/losses mainly through occasional changes in the PBO corresponding to these plans. Therefore, the estimated coefficient of PenGL will chiefly reflect corporate savings behavior in response to pension gains/losses that are mainly attributed to changes in the market value of DB pension plans assets.

Table V presents the regression results of increases in cash holdings for pension gains/losses for small firms (columns 1 and 2) and large firms (columns 3 and 4), respectively. The coefficients of PenGL and PenGL_Neg are negative and statistically significant at the 1% level in all specifications for small firms, suggesting that small firms prefer to increase cash holdings in response to pension losses. This effect is not only statistically significant but also economically significant. The coefficients of PenGL_Neg indicate that small sponsoring firms save more than $0.30 per dollar for every $ 1 in pension losses. On the other hand, the coefficients of PenGL_Neg are also negative and statistically significant for large firms, with relatively small point estimates. The possible explanation for relatively small PenGL coefficients for the large subsample is that large firms in general have their own pension plans and have discretion over funding policy. (15) Thus, large firms can voluntarily make contributions to their pension plans after the recognition of pension deficits to pursue tax-favorable treatments associated with pension plans.

Overall, the results here indicate that pension deficits that do not reduce the current internal resources of sponsoring firms but will reduce future resources significantly increase the cash holdings of sponsoring firms. In other words, anticipated liquidity shocks increase current cash holdings. Combined with the results for capital expenditures reported in Table IV, the results suggest that sponsoring firms with pension losses increase cash holdings to prepare for possible cash flow shortfalls associated with future contributions and, hence, forgo the investment opportunities that otherwise would be exploited.

D. Robustness Tests

1. Other Types of Investment

In this subsection, I consider other types of investment to confirm the robustness of the results. As the recent literature argues, R&D investments are gaining in importance among firms in developed economies (Brown, Fazzari, and Petersen, 2009). Thus, I should also pay attention not only to capital expenditures but also to overall investments, which includes R&D investments. To accomplish this, I apply total investment defined as the sum of capital expenditures, R&D expenditures, and advertisement expenditures as a dependent variable (Invall). Whereas temporal changes in internal resources are unlikely to affect R&D investments because of substantial adjustment costs, firms with pension losses may hesitate to initiate a new R&D project that requires additional resources in the years following the initiation of the project.

The results are reported in Table VI. Although the results are quite similar to those reported in Table IV, the coefficients of PenGL are slightly larger than those reported in Table IV. The results indicate that the incorporation of other types of investments does not alter the results.

2. Interaction between Capital Expenditures and Cash Savings

In this subsection, I examine the interaction between capital expenditures and cash savings. As I have discussed, the results indicate that pension losses affect both capital expenditures and changes in cash holdings. It might be the case that managers of sponsoring firms simultaneously determine investment and cash savings after the recognition of pension gains/losses. To consider the simultaneity of Capex and d_Cash, I run simultaneous regressions for which capital expenditures and changes in cash holdings are the dependent variables and compare the coefficients of Capex (d_Cash) in the d_Cash (Capex) regression of three-stage least squares estimates with that of the ordinary least squares (OLS) estimates. In other words, I compare the coefficients of the exogenous components of Capex (d_Cash) with those containing endogenous components.

In the untabulated results, the coefficients of Capex (d_Cash) in the d_Cash (Capex) regression in the three-stage least squares estimates are not statistically significant for both subsamples, suggesting that the exogenous components of Capex (d_Cash) do not affect d_Cash (Capex). However, Capex (d_Cash) significantly affects d_Cash (Capex) in the OLS estimates, suggesting that the endogenous components of Capex (d_Cash) do affect d_Cash (Capex). These results indicate that management endogenously determines the capital expenditures and cash savings, and these two variables are substitutional in resource allocations.

3. Adjustment Costs Associated with Investment

In this subsection, I consider the adjustment costs associated with investment. Specifically, I estimate the augmented regression for which I insert a one-year lag of Capex and the square term of Capex ([Capex.sup.2]) to account for the possible adjustment costs associated with investment.

To estimate this dynamic specification, I apply the first-difference Generalized Methods of Moments (GMM) procedure developed by Arellano and Bond (1991) for dynamic panel models with lagged dependent variables. Recent studies use lagged levels of endogenous variables dated at least t - 3 as instruments to estimate first-difference GMM (Brown et al., 2009; Brown and Petersen, 2009). The rationale behind this is that the exclusion restriction is satisfied by taking lagged values of at least t - 3 even in the case of an error term that follows the AR(1) process. Following recent studies, I use lagged levels of endogenous variables dated t - 3 to t - 5 as instruments.

In the untabulated results, the coefficient of lagged Capex and [Capex.sup.2] are statistically significant in the expected sign for large firms, although the null that all instruments are valid is rejected in some specifications. These results suggest substantial adjustment costs associated with investment for large firms. On the other hand, coefficients of these variables are not statistically significant for small firms. Observing the coefficients of pension variables for large firms, I find that the coefficients of PenGL and PenGL_Neg are positive as expected and that the latter is statistically significant. Thus, I argue that the results for the effect of pension losses on investment are robust to the consideration of adjustment costs associated with investment as well as the possible endogeneity of control variables. (16)

4. Access to Capital Markets and the Pension Deficits-Investment Relationship

Next, I consider the relationship between the extent of financing constraints and the effect of pension losses on corporate investments. If the argument that sponsoring firms prepare for future cash flow shortfalls associated with future required contributions by reducing investments is true, the observed relationship between pension losses and investments will be pronounced among those firms facing difficulties in raising capital from the capital markets. This is because these firms have to pay more attention to internal resource management than firms that have access to external financing. It should be noted that in Japanese firms, standard measures of financing constraints, such as firm size and firm age, are closely related to the depth of DB pension plans. Thus, I have to search for measures of financing constraint other than firm size and age. I apply two measures of financing constraint in the context of Japanese firms. The first measure is the extent of dependence on bank loans. Even after the deregulation of Japanese capital markets, corporate bond issues in Japan are practically limited to firms with high bond ratings. Firms facing less financing friction prefer to raise funds by issuing bonds rather than bank loans, whereas the financing of firms with higher financing friction are limited to bank loans. Because equity financing had been very limited during the sample period, whether firms have access to the bond markets is crucial from the financing constraint of Japanese firms. I expect that firms that do not have access to the bond markets carefully consider the impacts of recognized pension losses on future liquidity in deciding current investment. I measure the dependence on bank loans by the ratio of bank loans to financial debt and classify those firms for which the bank loan ratio has been zero during the sample period as bank-dependent firms. I insert the interaction term between BankDum and PenGL (PenGL_Neg) in the regression for which BankDum assumes the value of one for bank-dependent firms that do not have access to the bond markets and zero for others. I also consider the ability to cover financial debts and use IntRto instead of BankDum as a proxy for the extent of financing constraint, where IntRto is defined as interest payments divided by operational earnings including financial revenues. Firms having lower levels of this variable ratio can be recognized as having capacity for debt financing. I insert the interaction terms between these measures of financing constraint and PenGL (PenGL_Neg) to verify the extent to which the financing constraint affects the impacts of pension losses on investment.

Table VII reports the regression results. In the results for the large subsample (panel B), the coefficients of the interaction terms are statistically significant in the expected sign in all specifications. For example, when I insert the interaction term between PenGL and BankDum (column 1), the coefficient of PenGL is 0.056, and that of the interaction term is 0.133, indicating that the impact of pension gains/losses on investment is 0.189 (0.056 + 0.133) for bank-dependent firms, which is far larger than the figure for firms that have access to the bond markets, 0.056.

On the other hand, in the specification where I insert the interaction term between PenGL_Neg and BankDum, the coefficient of PenGL_Neg and the interaction terms are 0.103 and 0.180, respectively. This indicates that a $ 1 increase in pension losses will decrease investment by $0.283 for bank-dependent firms, whereas the figure for firms with access to the bond markets is $0.103. This result suggests that liquidity concerns under capital market imperfections drive the pension losses-investment relationships. Similar arguments can also be applied to the specifications in which I use the ability to cover financial debt (IntRto) instead of the bank-dependence dummy. The coefficients of the interaction terms are positive and statistically significant, indicating that the impacts of pension gains/losses, especially pension losses, are larger for firms with higher interest payments to earnings ratios than firms with lower ratios.

On the other hand, the coefficients of the interaction terms are also positive for small firms (panel A) and statistically significant in the specification with the interaction between PenGL_Neg and IntRto (column 4). Although I could not find statistically significant effects of pension gains/losses on investment for small firms (Table IV), the result here indicates that even small firms may decrease their investment upon the emergence of pension deficits if they are financially constrained. Overall, the results of Table VII indicate that firms with limited access to capital markets face more severe difficulties financing investment opportunities when pension deficits increase. This suggests that the relationship between pension losses and investment is the result of future liquidity concerns associated with realized pension losses.

5. Corporate Governance and the Pension Deficits-Investment Relationship

I next consider the possible impacts of corporate governance characteristics on the relationship between PenGL and cash holdings. For managers to be engaged in liquidity management associated with a pension plan, managers should be well monitored and have incentives to pursue long-term shareholder value. This is because the consequences of pension losses will not appear immediately after the occurrence of the deficits. Under short-termism, in deciding on current investment, managers do not pay attention to liquidity concerns in the future associated with recognized pension losses. Furthermore, pension-funding rules are complex and require advanced knowledge of pension plans and pension plan regulations. Meanwhile, managers at firms with higher foreign ownership are likely to be well monitored by investors and hence they would have incentives to pursue long-term shareholder value. In fact, the foreign ownership ratio is positively and statistically correlated with firm Q ratio ([rho] = 0.360, p-value = 0.000), suggesting that well-monitored managers pursue shareholder value. Thus, I expect that managers of firms with substantial foreign ownership will properly address anticipated liquidity shocks in deciding investment. I collected data of ownership by institutional investors from Nikkei Cges database. Since these data are available only from fiscal year 2003 and are fairly stable during the period, I created an institutional ownership dummy (InstH) that takes a value of one for firms for which the average institutional ownership ratio during 2003-2011 falls above the median of each subsample and zero for others and insert the interaction term between InstH and PenGL (PenGL_Neg) in the regression. I also create a similar dummy variable for the foreign ownership ratio (FrgnH) and insert its interaction term with PenGL (PenGL_Neg) in the regression.

Table VIII presents the results. The coefficients of the interaction terms are positive as expected in all specifications and statistically significant in some specifications. Specifically, the coefficient of the interaction between PenGL and InstH is statistically significant at the 1% level for the small subsample (panel A), and the interaction between PenGL and FrgnH has a statistically significant coefficient in the results for the large subsample (panel B). Although not statistically significant, the interaction terms with FrgnH have coefficients with t-values of approximately 1.6 in both specifications in the small subsample. Although the results are not robust, these results suggest the possibility that managers of well-monitored firms exert more effort in liquidity management to consider the possible impacts of pension losses. (17)

6. Possibility of Benefit Reductions

In Japan, sponsoring firms facing financial distress can reduce the benefits to employees if the specified conditions are satisfied to avoid the possibility of bankruptcy. Specifically, sponsoring firms in financial distress can reduce the benefits to employees if the majority of the employees approve of the benefit reductions. Thus, managers of sponsoring firms in financial distress anticipate possible future benefit reductions and may not prepare for future contributions to recover the funding status of their plans.

To cope with this possibility, I exclude those observations for which the shareholder equity-to-total assets ratios are lower than 0.2. Although I do not reproduce the results, I find quite similar results with those reported in Table IV, while the coefficients of PenGL_Neg are slightly larger than those reported in Table IV for the large firm subsample. This suggests that some financially distressed firms in the sample do not prepare for future required contributions, and accordingly, excluding them presents more definitive evidence of the relationship between pension losses and corporate investments.

7. Endogeneity of Discount Rates

As discussed in Section I, changes in PenGL are mainly the result of the changes in the market value of plan assets and occasionally the changes in the discount rates to evaluate the present value of pension obligations. Although changes in discount rates should reflect the interest rates of the highly graded bonds in the market, managers may use their discretion when determining discount rates to decrease the amount of the PBO.

To cope with this possibility, I recalculate the PBO using the yearly averages of discount rates instead of the reported discount rates, and I use this adjusted PBO to measure pension gains/losses (I denote this as PenGL2). (18) In the untabulated results for which I use PenGL2 instead of PenGL as a pension gain / loss variable, the coefficients of PenGL2 are positive for both subsamples and statistically significant at the 1% level for the large firm subsample. Similar results are also found when I allow PenGL2 to have different coefficients depending on the sign of the pension gains/losses. Thus, the results are robust with respect to the possible endogeneity of the discount rates to evaluate the PBO.

IV. Conclusions

I investigate how anticipated liquidity shocks affect firm financial management behavior by examining the impacts of actuarial pension gains/losses on capital expenditures and changes in cash holdings. Using a new sample of Japanese firms for which pension deficits have a large impact on the majority of the sponsoring firms, I show that the pension losses that do not immediately result in required contributions significantly reduce the investment of sponsoring firms. These results suggest that sponsoring firms optimize their investments with consideration for anticipated liquidity concerns associated with current pension losses, consistent with the argument of Bakke and Whited (2012). In addition, I document that pension losses significantly increase cash holdings, suggesting that sponsoring firms with pension deficits have higher precautionary demands for cash to prepare for the possible cash flow shortfalls associated with future contributions. Finally, the extent of financing constraints increases the impact of pension losses on investment. I argue that liquidity concerns under capital market imperfections drive pension losses-investment relationships. Overall, anticipated liquidity shocks appear to affect current financial management behavior.

References

Agca, S. and A. Mozumdar, 2008, "The Impact of Capital Market Imperfections on Investment-Cash Flow Sensitivity," Journal of Banking & Finance 32, 207-216.

Almeida, H., M. Campello, and M.S. Weisbach, 2004, "The Cash Flow Sensitivity of Cash," Journal of Finance 59, 1777-1804.

Arellano, M. and S. Bond, 1991, "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations," Review of Economic Studies 58, 211-291.

Bakke, T.-E. and T.M. Whited, 2012, "Threshold Events and Identification: A Study of Cash Shortfalls," Journal of Finance 67, 1083-1 111.

Bates, T.W., K.M. Kahle, and R.M. Stulz, 2009, "Why Do US Firms Hold So Much More Cash than They Used To?" Journal of Finance 64, 1985-2021.

Bergstresser, D., M. Desai, and J. Rauh, 2006, "Earnings Manipulation, Pension Assumptions, and Managerial Investment Decisions," Quarterly Journal of Economics 121, 157-195.

Brown, J.R., S.M. Fazzari, and B.C. Petersen, 2009, "Financing Innovation and Growth: Cash Flow, External Equity, and the 1990s R&D Boom," Journal of Finance 64, 151-185.

Brown. J.R. and B.C. Petersen, 2009, "Why Has the Investment-Cash Flow Sensitivity Declined So Sharply? Rising R&D and Equity Market Developments," Journal of Banking & Finance 33, 971-984.

Campbell, J.L., D.S. Dhaliwal, and W.C. Schwartz, 2010, "Equity Valuation Effects of the Pension Protection Act of 2006," Contemporary Accounting Research 27, 469-536.

Campbell, J.L., D.S. Dhaliwal, and W.C. Schwartz, 2012, "Financing Constraints and the Cost of Capital: Evidence from the Funding of Corporate Pension Plans," Review of Financial Studies 25, 868-912.

Chang, X., S. Dasgupta, and G. Hilary, 2006, "Analyst Coverage and Financing Decisions," Journal of Finance 61, 3009-3048.

Fazzari, S., G.R. Hubbard, and B.C. Petersen, 1988, "Financing Constraints and Corporate Investment," Brookings Papers on Economic Activity 1, 141-195.

Franzoni, K, 2009, "Underinvestment vs. Overinvestment: Evidence from Price Reactions to Pension Contributions," Journal of Financial Economics 92, 491-518.

Hubbard, G.R., 1998, "Capital-Market Imperfections and Investment," Review of Economics and Statistics 36, 193-225.

Iskandar-Datta, M.E. and Y. Jia, 2012, "Cross-Country Analysis of Secular Cash Trends," Journal of Banking & Finance 36, 898-912.

Kaplan, S.N. and L. Zingales, 1997, "Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints?" Quarterly Journal of Economics 112, 169-215.

Lamont, O., 1997, "Cash Flow and Investment: Evidence from Internal Capital Markets," Journal of Finance 52, 83-109.

McLellan, S., 2005, "Corporate Pension Reform in Japan: Big Bang or Big Bust," Benefits Quarterly 21, 31-41.

Mercer, 2012, "US Pension Plans' Funded Status Erodes in First Half of 2012," Available at: http://www.mercer.com/press-releases/pension-fimd-status-first-half-2012 (accessed June 30, 2013).

Myers, S.C., 1977, "Determinants of Corporate Borrowing," Journal of Financial Economics 5, 147-175.

Myers, S.C. and N.S. Majluf, 1984, "Corporate Financing and Investment Decisions When Firms Have Information that Investors Do Not Have," Journal of Financial Economics 13, 187-221.

Opler, T, L. Pinkowitz, R. Stulz, and R. Williamson, 1999, "The Determinants and Implications of Corporate Cash Holdings," Journal of Financial Economics 52, 3-46.

Phan, H.V and S.P. Hegde, 2013, "Pension Contributions and Firm Performance: Evidence from Frozen Defined Benefit Plans," Financial Management 42, 373-411.

Rauh, J.D., 2006, "Investment and Financing Constraints: Evidence from the Funding of Corporate Pension Plans," Journal of Finance 61, 33-71.

Standard & Poor's, 2006, Corporate Ratings Criteria, New York, S&P.

Stein, I, 2003. "Agency, Information and Corporate Investment," in G. Constantinides, M. Harris, and R. Stulz, Eds. Flandbook of the Economics of Finance, Elsevier, Amsterdam, 111-165.

Towers/Watson, 2012, "Global Pension Study 2012," Available at: http://www.towerswatson.com/ assets/pdf/6267/Global-Pensions-Asset-Study-2012.pdf (accessed June 30, 2013).

Yermo, J. and C. Severinson, 2010, "The Impact of the Financial Crisis on Defined Benefit Plans and the Need for Counter-Cyclical Funding Regulations," OECD Working Papers on Finance, Insurance, and Private Pensions, No. 3.

Takafumi Sasaki, is an Associate Professor of Finance at Tokyo University of Science in Tokyo, Japan.

I appreciate the helpful comments from Raghavendra Rau (Editor), an anonymous referee, Hideki Hanaeda, Takato Hiraki, as well as seminar participants at NLI Research Institute and Waseda University. This work was supported by Japan Society for the Promotion of Science (JSPS) Grant-in-Aidfor Scientific Research (B) Grant Number 24330124 and Grant-in-Aid for Scientific Research (C) Grant Number 22530370. All errors remain the sole responsibility of the author.

(1) The Towers/Watson survey (2012) indicates that the liabilities of DB pension plans have grown more rapidly than dedicated plan assets in recent years in 13 of the major pension markets. The survey includes Australia, Brazil, Canada, France, Germany, Hong Kong, Ireland, Japan, the Netherlands, South Africa, Switzerland, the United Kingdom, and the United States. Mercer (2012) reports that the aggregated pension deficits, which are defined as the difference between the present value of pension liabilities and the fair-value of dedicated plan assets, of S&P 1500 firms had grown by $59 billion in the first half of 2012, amounting to $543 billion.

(2) In addition, Franzoni (2009) and Campbell and Dhaliwal (2010) present evidence indicating that an increase in mandatory pension contributions is associated with negative abnormal stock returns for severely financially constrained firms.

(3) In developed economies, the government prescribes the funding rule for defined benefit plans, and sponsoring firms have to recover pension plan-funding status according to the formula specified by the funding rules. Specifically, sponsoring firms have to make additional contributions in the following years if the funding status of their pension plans fall below a certain point (threshold point). Although the funding rule specified by the Japanese government is not as restrictive as that specified by the US government, sponsoring firms in Japan have to make additional contributions to their pension plans to recover funding status if their plans are severely underfunded.

(4) Bakke and Whited (2012) use observations near funding thresholds to alleviate this self-selection issue and find causal effects of mandatory contributions on receivables, R&D, and hiring, but not on capital expenditures.

(5) http://www.mhlw.go.jp/toukei/itiran/roudou/jikan/syurou/03/index.html (in Japanese).

(6) Other popular DB pension plans (fund-type DB corporate pensions and contract-type DB corporate pensions) follow the same funding rule as EPFs.

(7) MCs can be a good instrument for internal cash flows that do not correlate with the private information regarding investment opportunities held by management, hence making it possible to circumvent the well-known specification problem that cash flows may contain private information regarding investment opportunities (Kaplan and Zingales, 1997).

(8) I compare the impact of PenGL, with that of [PenG.L.sub.t-1] on investment in the basic regression.

(9) In fact, 95.8% of the variance of PenGL is derived from within-variance.

(10) A little bit surprisingly, the coefficient of PenGL, is negative in Table III (column 3). A possible explanation for this is that firms with poor financial performance may be more aggressive in earnings management while cutting investment. Specifically, firms may choose higher rates of assumed rate of return on plan assets to boost reported earnings. Relatedly, Bergstressor, Desai, and Rauh (2006) find that firms choose higher expected returns for earnings management purpose.

(11) In this analysis, the sample shrinks because of the availability of ownership data.

(12) As for independent director ratio, it is important to acknowledge that the percentage of independent directors might be optimized with regard to the extent of monitoring by large investors. If this is the case, I expect to observe a positive correlation between the percentage of independent directors on the board and the ownership of blockholders. However, the correlation between Idrto and Blockholder is positive (p = 0.004, p-value = 0.005). Furthermore, the correlation between the independent director ratio and the institutional ownership ratio is 0.141 (p-value = 0.000). Considering that the average independent director ratio is very low in the sample, these figures suggest that the percentage of independent directors is not optimized with regard to the extent of monitoring by large investors.

(13) It should be noted that average Q ratios for small firms may be low because of lack of liquidity, and private benefits accruing to controlling shareholders.

(14) I also check the possibility that entrenched managers prefer overinvesting by examining the impact of reciprocal shareholdings on investment. However, I do not find the evidence consistent with this entrenchment story.

(15) On the other hand, many small firms have multiemployer-type pension plans and have narrow discretions over funding policy.

(16) In some specifications, I treat cash flow as an endogenous variable and also instrument this by lagged levels dated t - 3 and t - 5. Thus, this approach also enables us to cope with the possible endogeneity of the control variables.

(17) One concern for this interpretation is that higher pension loss sensitivities of investment among firms with higher foreign ownership may be brought by substantial financing constraints. However, it is worthwhile to note that the majority of foreign investors are institutional investors. As a result, the mean of the institutional ownership ratios including foreign investors is substantially higher for large firms (38.3%) than for small firms (14.7%). Meanwhile, the number of financial analysts following a company is, in general, greater for firms with higher institutional ownership ratios. Related studies have shown that analyst-following or institutional investor ownership mitigates financial frictions (Chang, Dasgupta, and Hilary, 2006; Agca and Mozumdar, 2008). Thus, I argue that firms with a higher foreign ownership ratio are unlikely to face severe financial frictions.

(18) I assume the duration of pension obligations as 15 years, but the results are robust to changing this assumption.
Table I. Descriptive Statistics

This table provides the distributional statistics for the variables
used in the empirical analyses and related variables. The sample
consists of manufacturing firms listed on the Tokyo Stock Exchange
that have nonmissing values for total assets and cash flows for the
fiscal years 2001 /2011.1 further confine the sample to those firms
having at least three positive capital expenditure years and have at
least one positive PBO year. I collect the financial data from the
Nikkei Financial Quest database. Capex is capital expenditures, and
Invall is constructed as the sum of capital expenditures, R&D, and
advertisement expenditure. Q is defined as the sum of the market
value of shareholder equity and the book value of debt divided by
total assets. Cashflow denotes the gross cash flows, which is defined
as the sum of net income and depreciation and amortization adding
back R&D. PenGL is calculated as the changes in the amount of
unrecognized gains/losses adding back the recognition of gains/
losses. PBO is the projected benefit obligation, and PASS represents
dedicated pension assets. All variables are normalized by total
assets at the beginning of the year. I winsorize all variables
included in the empirical analyses at the top and bottom 1% to
protect the sample from the influence of outliers.

Variable   Obs.     Mean    Std. Dev.     Min     Max

Capex      6,698    0.043     0.031      0.001   0.158
Invall     6,698    0.078     0.046      0.000   0.498
d_Cash     6,698    0.004     0.042     -0.374   0.433
PenGL      6,698   -0.004     0.016     -0.070   0.051
PBO        6,698    0.132     0.093      0.003   0.482
PASS       6,698    0.072     0.059      0.000   0.279
Q          6,698    1.091     0.374      0.541   2.775
Cashflow   6,698    0.089     0.052     -0.054   0.228

Table II. Descriptive Statistics for the Subsamples Classified
by Firm Size

This table provides the distributional statistics for the variables
used in the empirical analyses and related variables for the
subsamples classified by firms size. I assign those firms for which
the average total assets in the sample period are larger (smaller)
than the 70th percentile (30th percentile) as large firms (small
firms). Capex is capital expenditures, and Invall is constructed as
the sum of capital expenditures, R&D, and advertisement expenditure.
Q is defined as the sum of the market value of shareholder equity and
the book value of debt divided by total assets. Cashflow denotes the
gross cash flows, which is defined as the sum of net income and
depreciation and amortization adding back R&D. PenGL is calculated as
the changes in the amount of unrecognized gains/losses adding back
the recognition of gains/losses. PBO is the projected benefit
obligation, and PASS represents the dedicated pension assets. All
variables are normalized by total assets at the beginning of the
year. I winsorize all variables included in the empirical analyses at
the top and bottom 1% to protect the sample from the influence of
outliers.

Panel A. Small Firms

Variable    Obs.     Mean    Std. Dev.    Min      Max

Capex       2,007    0.037     0.030      0.001   0.158
Invall      2,007    0.067     0.044      0.000   0.498
djCash      2,007    0.004     0.047     -0.200   0.433
PenGL       2,007   -0.003     0.012     -0.061   0.052
PBO         2,007    0.123     0.099      0.003   0.482
PASS        2,007    0.056     0.058      0.000   0.279
Q           2,007    1.044     0.382      0.541   2.775
Cashflow    2,007    0.080     0.052     -0.054   0.228

Panel B. Large Firms

Capex       2,006    0.051     0.029      0.001   0.158
Invall      2,006    0.092     0.046      0.000   0.363
d_Cash      2,006    0.004     0.039     -0.325   0.387
PenGL       2,006   -0.004     0.017     -0.061   0.052
PBO         2,006    0.142     0.091      0.003   0.482
PASS        2,006    0.086     0.059      0.000   0.279
Q           2,006    1.216     0.399      0.541   2.775
Cashflow    2,006    0.102     0.051     -0.054   0.228

Table III. The Effect of Pension Gains/Losses on Capital Expenditure

This table reports the regression estimates examining the effect of
pension gains-losses on capital expenditures for the full sample.
Capex is capital expenditures. Q is defined as the sum of the market
value of shareholder equity and the book value of debt divided by
total assets. Cashflow denotes the gross cash flows, which is defined
as the sum of net income and depreciation and amortization adding
back R&D. PenGL is calculated as the changes in the amount of
unrecognized gains-losses adding back the recognition of gains-
losses. PenGL_Pos equals PenGL for positive PenGL observations and 0
for negative PenGL observations. PenGLJdeg is defined in a similar
way (it equals PenGL for negative PenGL observations and 0 for
others). Lag represents one-year lagged variables. All variables are
normalized by total assets at the beginning of the year. I winsorize
all variables included in the empirical analyses at the top and
bottom 1% to protect the sample from the influence of outliers.
Clustered standard errors within a firm are reported in parentheses.

Dependent Variable = Capex

                                       Full Sample

0                  0.019         0.019         0.020         0.020
                  (0.002) ***   (0.002) ***   (0.002) ***   (0.002) ***
Cashflow           0.055         0.055         0.061         0.061
                  (0.011) ***   (0.011) ***   (0.011) ***   (0.011) ***
PenGL (Lag)        0 053
                  (0.015) ***
PenGL_Pos (Lag)                  0.030
                                (0.031)
PenGL_Neg (Lag)                  0.066
                                (0.026) **
PenGL                                         -0.027
                                              (0.014) *
PenGL Pos                                                   -0.040
                                                            (0.031)
PenGL_Neg                                                   -0.018
                                                            (0.027)
Firm fixed        yes           yes           yes           yes
  effects
[R.sup.2]          0.08          0.08          0.08          0.08
N                    6,698         6,698         6,698         6,698

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table IV. Firm Size and the Effect of Pension Gains/Losses
on Capital Expenditure

This table reports the regression estimates examining the effect of
pension gains-losses on capital expenditures for the subsamples based
on firm size. I assign those firms for which the average total assets
in the sample period are larger (smaller) than the 70th percentile
(30th percentile) as large firms (small firms). Capex is capital
expenditures. Q is defined as the sum of the market value of
shareholder equity and the book value of debt divided by total
assets. Cashflow denotes the gross cash flows, which is defined as
the sum of net income and depreciation and amortization adding back
R&D. PenGL is calculated as the changes in the amount of unrecognized
gains-losses adding back the recognition of gains-losses. PenGL_Pos
equals PenGL for positive PenGL observations and 0 for negative PenGL
observations. PenGL_Neg is defined in a similar way (it equals PenGL
for negative PenGL observations and zero for others). Lag represents
one-year lagged variables. All variables are normalized by total
assets at the beginning of the year. I winsorize all variables
included in the empirical analyses at the top and bottom 1% to
protect the sample from the influence of outliers. Clustered standard
errors within a firm are reported in parentheses.

Dependent Variable = Capex

                         Small Firms                 Large Firms

Q                  0.016         0.016         0.020         0.020
                  (0.003) ***   (0.003) ***   (0.003) ***   (0.003) ***
Cashflow           0.057         0.057         0.057         0.057
                  (0.018) ***   (0.018) ***   (0.020) ***   (0.020) ***
PenGL (Lag)        0.021                       0.073
                  (0.039)                     (0.021) ***
PenGL_Pos (Lag)                 -0.005                      -0.027
                                (0.077)                     (0.046)
PenGLJdeg (Lag)                  0.035                       0.129
                                (0.072)                     (0.036) ***
Firm fixed            yes           yes           yes           yes
  effects
[R.sup.2]          0.06          0.06          0.11          0.11
N                    2,007         2,007         2,006         2,006

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table V. Firm Size and the Effect of Pension Gains/Losses
on Saving Cash

This table reports the regression estimates examining the effect of
pension gains-losses on increases in cash holdings for the subsamples
based on firm size. I assign those firms for which the average total
assets in the sample period are larger (smaller) than the 70th
percentile (30th percentile) as large firms (small firms). d_Cash
represents the changes in cash holdings. Q is defined as the sum of
the market value of shareholder equity and the book value of debt
divided by total assets. Cashflow denotes the gross cash flows, which
is defined as the sum of net income and depreciation and amortization
adding back R&D. PenGL is calculated as the changes in the amount of
unrecognized gains-losses adding back the recognition of gains/
losses. PenGL_Pos equals PenGL for positive PenGL observations and
zero for negative PenGL observations. PenGL_Neg is defined in a
similar way (it equals PenGL for negative PenGL observations and zero
for others). Lag represents one-year lagged variables. All variables
are normalized by total assets at the beginning of the year. I
winsorize all variables included in the empirical analyses at the top
and bottom 1% to protect the sample from the influence of outliers.
Clustered standard errors within a firm are reported in parentheses.

Dependent Variable = d_Cash

                        Small Firms                 Large Firms

Q                 -0.017        -0.017        -0.005        -0.005
                  (0.005)***    (0.005)***    (0.006)       (0.006)
Cashflow           0.305         0.308         0.173         0.173
                  (0.043)***    (0.043)***    (0.055)***    (0.055)***
PenGL (Lag)       -0.200                      -0.053
                  (0.083)**                   (0.045)
PenGL_Pos (Lag)                  0.005                       0.073
                                (0.165)                     (0.084)
PenGL_Neg (Lag)                 -0.311                      -0.125
                                (0.115)***                  (0.060)**
Firm fixed            yes           yes           yes           yes
  effects
[R.sup.2]          0.05          0.05          0.02          0.02
N                    2,007         2,007         2,006         2,006

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table VI. Firm Size and the Effect of Pension Gains/Losses
on Total Investment

This table reports the regression estimates examining the effects of
pension gains-losses on total investments for the subsamples based on
firm size. I assign those firms for which the average total assets in
the sample period are larger (smaller) than the 70th percentile (30th
percentile) as large firms (small firms). Invall is constructed as
the sum of capital expenditures, R&D, and advertisement expenditures.
Q is defined as the sum of the market value of shareholder equity and
the book value of debt divided by total assets. Cashflow denotes the
gross cash flows, which is defined as the sum of net income and
depreciation and amortization adding back R&D. PenGL is calculated as
the changes in the amount of unrecognized gains-losses adding back
the recognition of gains-losses. PenGL_Pos equals PenGL for positive
PenGL observations and zero for negative PenGL observations.
PenGL_Neg is defined in a similar way (it equals PenGL for negative
PenGL observations and zero for others). Lag represents one-year
lagged variables. All variables are normalized by total assets at the
beginning of the year. I winsorize all variables included in the
empirical analyses at the top and bottom 1% to protect the sample
from the influence of outliers. Clustered standard errors within a
firm are reported in parentheses.

Dependent Variable = Invall

                         Small Firms                 Large Firms

Q                  0.021         0.021         0.019         0.019
                  (0.005) ***   (0.005) ***   (0.004) ***   (0.004) ***
Cashflow           0.110         0.111         0.127         0.127
                  (0.022) ***   (0.023) ***   (0.030) ***   (0.030) ***
PenGL (Lag)       -0.041                       0.091
                  (0.065)                     (0.025) ***
PenGL Pos (Las)                  0.061                       0.001
                                (0.097)                     (0.056)
PenGL_Neg (Lag)                 -0.096                       0.143
                                (0.121)                     (0.044) ***
Firm fixed            yes           yes           yes           yes
  effects
[R.sup.2]          0.08          0.08          0.10          0.10
N                    2,007         2,007         2,006         2,006

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table VII. Financing Constraints and the Effect of Pension
Gains/Losses on Capital Expenditure

This table reports the regression estimates examining the effect of
pension gains-losses on capital expenditures considering the
interaction between financing constraints and pension gains-losses.
Specifically, I insert the interaction term between BankDum and PenGL
(PenGL_Neg) in the regression for which BankDum takes a value of one
for firms that do not have access to the bond markets and zero for
others. I also use IntRto instead of BankDum as a proxy for the
extent of the financing constraint, whereby IntRto is defined as
interest payments divided by operating earnings. I assign those firms
for which the average total assets in the sample period are larger
(smaller) than the 70th percentile (30th percentile) as large firms
(small firms). Capex is capital expenditures. Q is defined as the sum
of the market value of shareholder equity and the book value of debt
divided by total assets. Cashflow denotes the gross cash flows, which
is defined as the sum of net income and depreciation and amortization
adding back R&D. PenGL is calculated as the changes in the amount of
unrecognized gains-losses adding back the recognition of gains-
losses. PenGL_Pos equals PenGL for positive PenGL observations and
zero for negative PenGL observations. PenGL_Neg is defined in a
similar way (it equals PenGL for negative PenGL observations and zero
for others). Lag represents one-year lagged variables. All variables
are normalized by total assets at the beginning of the year. I
winsorize all variables included in the empirical analyses at the top
and bottom 1% to protect the sample from the influence of outliers.
Clustered standard errors within a firm are reported in parentheses.

                                   Panel A. Small Firms

                   Dependent Variable = Capex

                       Bank-Dependence            Interests Earnings
                            Dummy                       Ratio

Q                  0.016         0.016         0.016         0.016
                  (0.003) ***   (0.003) ***   (0.003) ***   (0.003) ***
Cashflow           0.058         0.058         0.056         0.054
                  (0.018) ***   (0.018) ***   (0.018) ***   (0.018) ***
PenGL (Lag)       -0.023                      -0.003
                  (0.049)                     (0.054)
PenGL (Lag) *      0.124
  BankDum         (0.077)
PenGL_Pos (Lag)                 -0.007                       0.002
                                (0.076)                     (0.078)
PenGL_Neg (Lag)                 -0.018                      -0.043
                                (0.093)                     (0.101)
PenGL_Neg (Lag)                  0.156
  * BankDum                     (0.114)
PenGL (Lag) *                                  0.068
  IntRto                                      (0.075)
PenGL_Neg (Lag)                                              0.189
  * IntRto                                                  (0.107) *
Firm fixed          yes           yes           yes           yes
  effects
[R.sup.2]          0.06          0.06          0.06          0.06
N                 2,007         2,007         2,007         2,007

                                 Panel B. Large Firms

Q                  0.020         0.020         0.020         0.020
                  (0.003) ***   (0.003) ***   (0.003) ***   (0.003) ***
Cashflow           0.057         0.057         0.057         0.057
                  (0.020) ***   (0.020) ***   (0.020) ***   (0.020) ***

PenGL (Lag)        0.056                       0.039
                  (0.021) ***                 (0.025)
PenGL (Lag) *      0.133
  BankDum         (0.050) ***
PenGL_Pos (Lag)                 -0.025                      -0.027
                                (0.046)                     (0.045)
PenGL_Neg (Lag)                  0.103                       0.050
                                (0.035) ***                 (0.043)
PenGL_Neg (Lag)                  0.180
  * BankDum                     (0.087) **
PenGL (Lag) *                                  0.165
  IntRto                                      (0.053) ***
PenGL_Neg (Lag)                                              0.334
  * IntRto                                                  (0.077) ***
Firm fixed          yes           yes           yes           yes
  effects
[R.sup.2]          0.11          0.11          0.11          0.12
N                 2,006         2,006         2,006         2,006

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table VIII. Ownership by Foreign Investors and the Effect
of Pension Gains/Losses on Capital Expenditure

This table reports the regression estimates examining the effect of
pension gains-losses on capital expenditures considering the
interaction between corporate governance and pension gains-losses.
Specifically, I insert the interaction term between InstH and PenGL
(PenGL_Neg) in the regression for which InstH is a dummy variable
that takes the value of one for firms of which institutional
ownership ratios greater than the median and zero otherwise. I also
use a dummy variable based on the ownership ratio of foreign
investors (FrgnH) instead of InstH as a corporate governance
variable. I assign those firms for which the average total assets in
the sample period are larger (smaller) than the 70th percentile (30th
percentile) as large firms (small firms). Capex is capital
expenditures. Q is defined as the sum of the market value of
shareholder equity and the book value of debt divided by total
assets. Cashflow denotes the gross cash flows, which is defined as
the sum of net income and depreciation and amortization adding back
R&D. PenGL is calculated as the changes in the amount of unrecognized
gains-losses adding back the recognition of gains-losses. PenGL_Pos
equals PenGL for positive PenGL observations and zero for negative
PenGL observations. PenGL_Neg is defined in a similar way (it equals
PenGL for negative PenGL observations and zero for others). Lag
represents one-year lagged variables. All variables are normalized by
total assets at the beginning of the year. I winsorize all variables
included in the empirical analyses at the top and bottom 1% to
protect the sample from the influence of outliers. Clustered standard
errors within a firm are reported in parentheses.

                     Panel A. Small Firms

                  Dependent Variable = Capex

                    Institutional Investor         Foreign Investor
                        Ownership Ratio             Ownership Ratio

Q                  0.015         0.016         0.016         0.016
                  (0.003) ***   (0.003) ***   (0.003) ***   (0.003) ***
Cashflow           0.058         0.058         0.058         0.058
                  (0.018) ***   (0.018) ***   (0.018) ***   (0.018) ***
PenGL (Lag)        0.005                       0.005
                  (0.039)                     (0.040)
PenGL (Lag) *      0.384
  InstH           (0.183) **
PenGL_Pos (Lag)                 -0.002                      -0.002
                                (0.077)                     (0.076)
PenGL_Neg (Lag)                  0.016                       0.007
                                (0.073)                     (0.071)
PenGL_Neg (Lag)                  0.371
  * InstH                       (0.281)
PenGL (Lag) *                                  0.201
  FrgnH                                       (0.126)
PenGL_Neg (Lag)                                              0.325
  * FrgnH                                                   (0.206)
Firm fixed            yes           yes           yes           yes
  effects
[R.sup.2]          0.06          0.06          0.06          0.06
N                    2,007         2,007         2,007         2,007

                                  Panel B. Large Firms

Q                  0.020         0.020         0.020         0.020
                  (0.003) ***   (0.003) ***   (0.003) ***   (0.003) ***
Cashflow           0.056         0.057         0.058         0.058
                  (0.020) ***   (0.020) ***   (0.020) ***   (0.020) ***
PenGL (Lag)        0.002                      -0.010
                  (0.056)                     (0.053)
PenGL (Lag) *      0.082
  InstH           (0.060)
PenGL_Pos (Lag)                 -0.027                      -0.030
                                (0.046)                     (0.046)

PenGL_Neg (Lag)                  0.098                       0.074
                                (0.085)                     (0.082)
PenGL_Neg (Lag)                  0.035
  * InstH                       (0.089)
PenGL (Lag) *                                  0.097
  FrgnH                                       (0.057) *
PenGL_Neg (Lag)                                              0.067
  * FrgnH                                                   (0.087)
Firm fixed            yes           yes           yes           yes
  effects
[R.sup.2]          0.11          0.11          0.11          0.11
N                    2,006         2,006         1,995         1,995

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.
COPYRIGHT 2015 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Sasaki, Takafumi
Publication:Financial Management
Article Type:Report
Date:Sep 22, 2015
Words:12451
Previous Article:Why do insiders hedge their ownership? An empirical examination.
Next Article:Extrapolation errors in IPOs.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters