Document not found! Please try again

Determinants of Actuarial Choices for Defined-Benefit Pension Plans ...

2 downloads 11672 Views 498KB Size Report
Sep 2, 2014 - Determinants of Actuarial Choices for Defined-Benefit Pension ..... Oxley Act (SOX), the expected rate of return and the discount rate are ...
International Review of Business Research Papers Vol. 10. No. 2. September 2014 Issue. Pp. 192 – 207

Determinants of Actuarial Choices for Defined-Benefit Pension Plans: Canadian Evidence Ines Ben Salah* and Houcem Smaoui** This study examines the factors driving the choice of the actuarial assumptions, such as the discount rate, the salary growth rate, and the expected rate of return on asset, under pension accounting disclosures required for defined-benefit pension plans (DBPP, hereafter) for a sample of 190 Canadian firms over the period 2000-2006. We use a system of panel data simultaneous equations of the actuarial assumptions. We tackle the endogeneity problem of the actuarial choices using a system GMM approach. The evidence shows that the choice of the actuarial rates is mainly driven by the magnitude of leverage and the size of the sponsoring firms. Unlike previous studies, we find no significant relationship between the level of pension funding and the choice of actuarial assumptions.

Filed of Research: Actuarial Assumptions; Managerial Discretion; Defined-Benefit Pension Plans

1. Introduction During the 90s, the exceptional returns gained by pension funds allowed several plans to realize surplus assets that were used either to improve benefits or to halt contributions. However, the stock market downturn that followed the burst of the Dotcom bubble in the early 2000s has resulted in substantial losses for the majority of DBPPs compromising the creditworthiness of several of them. This situation worries firms sponsoring the DBPP, regulatory and monitoring pensions, future retirees, accounting standard setters and investors. Several U.S. researches have focused on investor perception of pension accounting data and managerial discretion in the choice of actuarial assumptions for the computation of pension accounting data. These researches conclude that assets and pension liabilities are considered by the market as assets and liabilities of the firm, and that managers of the sponsoring firms use their managerial discretion in the choice of the actuarial assumptions to manage pension data published in the financial statements. Unlike the American context, very few Canadian empirical studies have examined the determinants of pension funding reporting data, despite the publication of the new Section 3461 "Employee Future Benefits" which resembles the American Standards SFAS 87. To our knowledge, only two empirical studies focused on the perception of Canadian investors of pension data disclosed in the financial statements. First, Wiedman and Wier (2004) found that the average Canadian investors consider the funding deficit as a liability *

Dr. Ines Ben Salah, Department of Finance and Accounting, Institut Superieur de Gestion, Tunis, Tunisia, Email : [email protected] ** Dr. Houcem Smaoui, Department of Finance, Tunis Business School, Tunis, Tunisia. Email: [email protected]

Salah & Smaoui of the firms sponsoring the DBPP, but does not consider the funding surplus as a firm‟s asset. Second, Ben Salah et al. (2013) find that the Canadian market considered funding surplus and deficit as assets and liabilities of the firm. Moreover, managers of the sponsoring firms use their managerial discretion in the choice of actuarial assumptions in the reporting of pension data. Further, investors are aware of this managerial discretion and readjust to the decrease (increase) the market value of the firm when managers choose liberal (conservative) actuarial assumptions. However, Ben Salah et al. (2013) focused on the impact of the actuarial choices on firm‟s valuation and failed to correct for the endogeneity problem in their modeling of the actuarial choices, which could bias their results. The main objective of our study is to examine the managerial discretion in the choice of the actuarial assumptions used in the computation of the pension data disclosed in the financial statements for a sample of Canadian firms. We study the determinants of three actuarial choices: interest rates, salary growth rates, and expected rates of return on plan assets. The choice of actuarial assumptions about the actuarial rates has an impact on: (1) the reporting of pension data in the financial statements and notes to the financial statements, (2) the liquidity of the firm to the extent that the decrease in contributions to these plans can save cash for use in the financing of investment projects, and (3) the tax base since contributions are deductible and returns on pension assets are not taxable. We believe our study improves on prior research in many ways. First, the literature on the managerial discretion in the choice of actuarial assumptions is essentially based on the results of studies conducted in the U.S. context. Our research aims to enrich the literature by analyzing the determinants of the actuarial choices in the Canadian context. Indeed, our study focuses on the determinants of the actuarial assumptions used in the calculation of pension accounting data disclosed in the financial statements and notes to the financial statements according to Canadian standards (Canadian Institute of Chartered Accountants, CICA, Chapter 3461). Second, Ben Salah et al. (2013) studied the managerial discretion in the actuarial choices. However, the main objective of this study was to examine the impact of the actuarial assumptions on firm‟s valuation. Wiedman and Wier (2004) is the only available Canadian study that focuses on CICA chapter 3461, issued in 2000, which resembles SFAS 87. Their results contend a different valuation depending on the funded status of the regime. Unfortunately, their study covered a period of declining stock market (2000-2001) which could cast some doubt on their results. We believe a Canadian study over a wider time period (2000-2006) is needed to shed some light on the behavior of Canadian managers in the choice of actuarial assumptions. Third, Bodie et al. (1987) studied interest rates only, while Asthana (1999) and Brown (2006) used two assumptions: interest rates and salary growth rates. We complement the existing literature by using three actuarial assumptions: interest rates, salary growth rates, and expected rate of return on asset actuarial assumption. Finally, Asthana (1999) contends that the interest rate assumption cannot be separated from the joint process of selecting other actuarial rates such as the salary growth rates and the cost method. Consequently, actuarial rates are jointly distributed and a system of simultaneous equations in which the actuarial rates should be treated as endogenous must be employed, otherwise standard OLS regression estimates would be inconsistent 193

Salah & Smaoui and biased (Maddala, 1988). Our paper minimizes the limitation of previous studies by estimating a system of panel data simultaneous equations using interest rates, salary growth rates, and expected rate of return on asset actuarial assumption as jointly determined endogenous variables. We resolve the endogeneity problem of the actuarial choices using a system GMM approach. The evidence shows that the higher the firm‟s leverage, the more likely the firm is to choose liberal actuarial rates to understate the degree of their unfunded pension liabilities. Moreover, larger firms are more likely to choose conservative actuarial choices to avoid visibility costs. However, contrary to previous studies, we find no significant relationship between the level of pension funding and the choice of actuarial assumptions. The remainder of the paper is organized as follows. Section 2 presents the relevant literature and our hypotheses. Section 3 describes our empirical models and reports our results, and Section 4 consists of concluding remarks.

2. Literature Review For the DBPP meeting the requirements of the Employee Retirement Income Security Act of 1974 (ERISA 1974), earnings on pension plan assets are tax-free and contributions to the pension plan are tax deductible. Tepper (1981) argues that since the deferral on pension plan contributions and the tax deferral on pension plan earnings have specific values, firms have incentives to make the largest possible contributions to the pension plan. In the same vain, Tepper (1983), Bodie et al. (1987), Thomas (1988), and Thomas and Tung (1992) have examined the role of actuarial choices in funding management using tax considerations. Bodie et al. (1987) document an inverse relationship between profitability and discount rates and positive relationship between profitability and tax liabilities. Thomas (1988) finds that corporate taxes affect pension funding behavior. As tax status (defined as a function of current marginal tax rates and expected future taxable income) declines, firms tend to choose less conservative cost methods and interest rates to contract contributions. Thomas and Tung (1992) report intra-firm variations in actuarial cost methods and interest rates that are consistent with defense contractors managing pension costs to maximize their reimbursements. More recently, Asthana (1999) studies the effects of sponsoring firm‟s financial and pension characteristics on their actuarial choices and funding decisions. He examines empirically the hypothesized management of pension funding by testing for systematic biases of actuarial choices: interest rates, cost methods, and salary growth rates. The results support the hypothesized management of funding and actuarial choices, and show that as firms become underfunded, they make liberal actuarial choices to avoid visibility costs, and that as firms become overfunded, they tend to make conservative actuarial choices in order to avoid visibility costs. Further, as the annual pension contributions raise relative to the authorized contribution ranges, firms take conservative actuarial choices in order to reduce penalties and increase tax benefits. As the annual contributions decrease

194

Salah & Smaoui relative to the authorized contribution ranges, firms take liberal actuarial choices to reduce penalties and increase tax benefits. Ashtana (1999) shows that the larger (lower) the cash flow from operations, profitability, and tax liabilities, and the lower (larger) the debt of firm, the higher the probability that the firm‟s managers will make conservative (liberal) actuarial choices to maximize contributions. Healy and Palepu (1989) examine firms‟ accounting and dividend responses to an increase in the tightness of dividend constraints. They show that a decrease in pension expense, related to changing pension assumptions, not only increases reported earnings but also decreases cash outflows. They conclude that the pension changes are motivated by cash management concerns, rather than tightness of the dividend covenant. Mittelstaedt (1987) examines empirically possible motivational factors leading to reductions in pension plan overfunding. His results indicate that firms that are in severe financial strain terminate their pension plans. Firms with less severe financial strain tend to change actuarial assumptions to reduce required cash contributions to pension plans. Ghicas (1990) notes that, prior to the adoption of SFAS 87 (Employers‟ Accounting for Pensions), firms tend to use the same actuarial cost method for both funding and accounting purposes. The choice of the appropriate actuarial cost method was an important factor considered in the process of the promulgation of the SFAS 87. The actuarial cost methods have been classified by actuaries into two categories: (1) cost allocation methods, and (2) benefit allocation methods. Usually, actuaries encourage the use of cost allocation methods since they foster the security of the benefits of the participants in the pension plan by accumulating more assets in the pension plan than the actuarial present value of promised benefits. Ghicas finds that the decrease in pension funding is driven by higher debt, lower working capital, and lower rate of undertaking new investments for cost methods switching firms as compared to non-switching firms. Godwin et al. (1996) investigate factors motivating managers to adjust pension expense through changes in actuarial-rate assumptions. Their study shows that managers tend to employ one specific pension technique and manipulate actuarial-interest-rate assumptions to manage earnings. Godwin et al. provide evidence that managers are more likely to respond to tighter dividend constraints, more restrictive debt covenants, and reductions in earnings by increasing interest rate assumptions. Gopalakrishnan and Sugrue (1995) study the determinants of the choice of the actuarial rates under pension reporting guidelines required by SFAS 87. They find that the choice of the salary progression rate is motivated by the level of pension funding, whereas the choice of the discount rate is mainly driven by the level of funding and the magnitude of leverage. They contend that higher leverage along with unfunded pension liabilities can result in an increase in the probability of default. Thus, firms with huge unfunded pension liabilities are likely to choose a lower salary progression rate to moderate the magnitude of the reported pension liability, because higher rates of salary progression can lead to higher pension obligations. Likewise, firms with large unfunded pension liabilities and leverage tend to assume higher discount rates to downplay the magnitude of the reported liability. They also conclude that firms are likely to select a package of actuarial rates that are advantageous to them. 195

Salah & Smaoui Bergstresser et al. (2006) document that the size of DBPP and managers latitude in presenting them to capital markets make pension accounting a fertile area for earning manipulation. They notice that managers opportunistically select long-term rates of return on DBPP asset. Bergstresser et al. (2006) find that managers choose higher rates of return when they prepare to acquire other firms, when they exercise their stock options, and when the firm is near critical earnings levels. Dechow et al. (2011) examine firms that are suspected to have misstated their financial statements. They focus on the use of the expected return assumption on plan assets for DBPPs. Their results show that misstating firms have greater expected returns on their pension plan assets than other firms. They conclude that the effect of higher expected return assumptions is to reduce reported pension expense. Recently, Comprix et al. (2011) examine whether employers assume downward biased expected rate of return and discount rate to make their plans appear more costly in the year they decide to freeze their DBPP. They find that prior to the adoption of the Sarbanes Oxley Act (SOX), the expected rate of return and the discount rate are downward biased when firms freeze their plans whereas after SOX the bias is lower. The choice of assumptions affects the reported pension plan status and pension cost in a number of ways. The most immediate impact of assumptions is on the benefit obligation and, hence, the reported plan status. There is an indirect impact on the balance sheet as assumptions affect pension cost. Three key economic assumptions- the discount rate, salary growth rate and expected long term rate of return on plan assets- are disclosed. Comparisons of the obligation with the plan assets require consideration of the choice of both the discount rate and salary growth rate. Aggressive assumptions (high discount rate and low salary growth rate) improve the reported status of the plan, whereas conservative assumptions (low discount rate and high salary growth rate) make the plan appear less well funded. Finally, a higher assumed rate of return on plan assets lowers pension cost since the expected return on assets is an offset to other components of that cost. The choice of these actuarial assumptions poses significant problems for financial reporting. First, a change in the discount rate, salary growth rate or rate of return could have a significant impact on the employer‟s financial statements particularly on pension expense and pension liabilities. Second, changes in the actuarial assumptions could affect comparability of not only intra-firm but also inter-firm financial statement analysis. Evidence from the finance and accounting literature indicates that in an efficient market, a firm‟s unfunded pension liabilities are considered like debt and other liabilities of a firm for equity valuation and also for determining the systematic risk (Dhaliwal (1986), Feldestein and Seligman (1982)). The presence of large unfunded pension liability could mean lower credit rating and higher cost of debt for the firms. The discount rate assumption used to compute the present value has the greatest impact on the reported benefit obligation. A higher (lower) discount rate decreases (increases) the benefit obligation. Blankley and Swanson (1995) show that firms do not change discount rates as often as would be warranted by movements in general interest rate level. Furthermore, they note that this lack of conformity is greatest when rates are declining, as firms want to avoid increases in the obligation. The assumed salary growth rate also 196

Salah & Smaoui affects the benefit obligation, but less than the discount rate. Higher (lower) rate increases (decreases) pension obligations. Based on prior research, it appears that the funding status has an impact on actuarial choices (Asthana, 1999), hence we state our first hypothesis: Hypothesis 1: Underfunded (overfunded) firms tend to select liberal (conservative) actuarial assumptions Section 3461 requires more disclosure and more information that has a potential impact on lenders, one major users of accounting report. The magnitude of the unfunded pension liabilities combined with the size of the long term debt of a firm could influence the choice of the actuarial assumptions required under Section 3461. Lending institutions often impose several covenants in the lending agreements entered into between the lenders and the borrowing firms. When these covenants are violated, that would place the borrowing firm in “technical default” or renegotiation at costly terms. Therefore, in order to avoid the costly renegotiation process, firms would understate the degree of their unfunded pension liabilities by choosing a higher discount rate or by decreasing the salary growth rate or both. Hence, we state our second hypothesis: Hypothesis 2: High (low) leveraged firms tend to select liberal (conservative) actuarial assumptions Larger firms are more politically visible than smaller firms and, consequently, are more likely to be candidates for control of their actions by regulators and politicians (Watts and Zimmerman, 1978; Zimmerman, 1983; Hagerman and Zmijewski, 1979). The control might take the form of more regulation or higher taxes, which would result in wealth transfers away from the politically visible firms. The empirical evidence from this line of research seems consistent with the argument that large firms choose conservative actuarial choices to avoid visibility costs. Hence based on the argument of visibility costs, we state our third hypothesis: Hypothesis 3: Big (small) size firms tend to select conservative (liberal) actuarial assumptions

3. Research Design and Empirical Results 3.1 Sample Selection Our sample is composed of all Canadian firms with available data on defined benefit pension plan in the Compustat annual industrial data base. In order to be included in our sample, firms must have an average ratio of Plan Asset/Total Assets less than 1%, and must have at least 5 Million shares outstanding. This yields a total of 190 Canadian firms. Our sample covers the 2000–2006 time periods. Data on pension assumptions is extracted from the financial statements of each sample firm and gathered from the Compustat Data base.

197

Salah & Smaoui 3.2 Statistical Models To test our hypotheses on the managerial discretion in the choice of the actuarial assumptions used in the computation of the pension data disclosed in the financial statements, we estimate the following time-series cross-sectional equations: IRit   0  1PPFSRit   2 LEVit   3 ROAit   4 LIQUID it   5 AUDITORit 

 6 SIZEit   7 SGRit  8 ERAit  i  it SGRit   0  1 PPFSRit   2 LEVit   3 ROAit   4 LIQUID it   5 AUDITOR it   6 SIZEit   7 IRit   8 ERAit   i   it ERAit   0  1 PPFSRit   2 LEVit   3 ROAit   4 LIQUID it   5 AUDITOR it   6 SIZEit   7 IRit   8 SGRit  i  it

(1) (2) (3)

where i is the firm (i=1,…,N); t is the time indicator that is equal to the number of years (t=1,…,T); IR is the interest rate actuarial assumption; SGR is the salary growth rate assumption; ERA is the expected rate of return on asset actuarial assumption; PPFSR is the pension plan funded status ratio; LEV is the firm‟s leverage; ROA is the firm‟s return on assets; LIQUID is the firm‟s cash flows from operating activities; SIZE is the firm‟s size proxied by the log of sales; AUDITOR is a categorical variable measuring the quality of the auditors;  i are the unobserved individual effects; and it is an error term. Ashtana (1999) shows that as firms become underfunded, they make liberal actuarial choices to avoid visibility costs, and that as firms become overfunded, they tend to make conservative actuarial choices in order to avoid visibility costs. Hence, the pension plan funded status ratio (PPFSR) appears to be a major determinant of the actuarial assumptions. We also include a proxy for debt since as debt increases, the firm may prefer to divert its available cash to pay creditors, avoiding costly violation of debt covenants instead of contributing to its pension plans (Beneish and Press 1995). Alternatively, with increased debt, firms may prefer to tap their cheaper internal sources of funds by contracting contributions instead of relying on costlier external sources. The empirical evidence shows that large firms choose conservative actuarial choices to avoid visibility costs (Watts and Zimmerman, 1978; Zimmerman, 1983; Hagerman and Zmijewski, 1979), so we introduce a proxy for size. Following Gopalakrishnan and Sugrue (1995), we assume that actuarial choices are a function of the firm‟s profitability (ROA) and liquidity (LIQUID). As the profitability and cash availability of the sponsoring firm improves, its pension plan tends to make conservative actuarial choices to maximize tax-deductible contributions. Finally, we control for the quality of the auditors hired by the sponsoring firm using the variable AUDITOR which takes the value of 1 if the auditor is a Big-Eight/Big-Six, and 0 otherwise. Table 1 below shows the definitions and the measurements of the variables used in the analysis. Following Ashtana (1999), we posit that managers select these actuarial rates on a joint basis, so in each model, the other two actuarial choices are included as control variables. Since our data are time-series cross-sectional, auto-correlation and heteroskedasticity are a concern. Using the Breush-Pagan test and the Modified Wald test, we conclude respectively for the presence of serial correlation within panels and heteroskedasticity across panels in the error series. Therefore, OLS estimates will yield biased and inconsistent estimates of the parameters' standard errors. 198

Salah & Smaoui To resolve these econometric issues, models (1)-(3) are estimated using the Generalized Least Square (GLS) procedure. When computing the standard errors and the variancecovariance estimates, the disturbances are assumed to be heteroskedastic, autocorrelated within panels, and contemporaneously correlated across panels. Table 1: Variable definitions and measurements Description Variable AUDITOR ERA IR LEV LIQUID PPFSR ROA SGR SIZE

Categorical variables representing the quality of auditors hired by the firm, it takes the value of 1 if the auditor is a Big-Eight/Big-Six, 0 otherwise Expected rate of return on asset actuarial assumption Interest rate actuarial assumption Firm‟s leverage (Total debt divided by total asset). Firm‟s cash from operating activities divided by the number of shares outstanding. Pension plan funded status ratio (Pension Plan asset/Pension obligation) Return on asset ((Net income/total assets) x 100). Salary growth actuarial assumption Firm‟s log of sales.

3.3 Empirical Results Table 2 displays the descriptive statistics for our main variables. We can see that the interest rate assumption (IR) has decreased on average over the study period ranging from 6.91% in 2000 to 5.19% in 2006. This is in line with the general movement of the government bond rates over the period 2000-2006. The salary growth rate (SGR) remained stable between 2000 and 2004, and then decreased on average after 2004. We also notice a steady decrease in the expected return on plan asset (ERA) over the study period, going from 7.8% on average in 2000 to 6.91% in 2006. From Table 2, we notice that the average pension fund status (PPFSR) is negative over the period 2001-2006, ranging from 0.56$ per share in 2000 to -1.22$ per share in 2005. Finally, the average profitability measured by the return on assets (ROA) has increased from 1.56 in 2003 to 3.88 in 2006.

199

Salah & Smaoui Table 2: Descriptive Statistics This table reports the descriptive statistics for our main variables for our sample of 190 Canadian firms.. Variables IR

SGR

ERA

PPFSR

SIZE

LEV

ROA

STAT

2000

2001

2002

2003

2004

2005

2006

MIN

3.50

3.50

4.00

4.00

3.50

3.25

3.00

MAX

8.10

7.50

7.50

7.00

7.00

6.25

6.10

MEAN

6.94

6.74

6.55

6.18

5.90

5.19

5.19

STD

0.53

0.46

0.41

0.41

0.41

0.38

0.39

MIN

1.25

1.25

2.00

2.00

0.60

0.65

0.59

MAX

7.00

7.00

7.00

7.00

6.10

5.00

5.32

MEAN

4.00

4.00

3.92

3.82

3.69

3.65

3.59

STD

0.83

0.74

0.65

0.70

0.66

0.59

0.59

MIN

4.00

4.00

4.00

4.00

3.50

3.50

3.40

MAX

10.00

10.00

9.50

9.00

9.00

9.00

9.00

MEAN

7.80

7.70

7.47

7.29

7.17

7.02

6.91

STD

0.82

0.89

0.75

0.73

0.75

0.77

0.83

MIN

-1.83

-6.39

-14.75

-15.40

-17.27

-50

-16.95

MAX

9.45

7.42

4.02

3.79

2.64

1.97

2.62

MEAN

0.56

-0.06

-0.66

-0.60

-0.62

-1.22

-0.70

STD

1.42

1.24

1.96

1.94

1.93

4.57

2.33

MIN

2.09

3.30

2.69

0.37

-1.04

-1.23

-0.80

MAX

11.18

11.43

10.22

10.28

10.31

10.35

10.49

MEAN

7.04

7.09

7.06

7.00

7.07

7.12

7.19

STD

1.79

1.72

1.71

1.79

1.90

1.94

1.86

MIN

0.00

0.00

0.00

0.00

0.00

0.00

0.00

MAX

80.61

110.13

128.39

118.99

63.63

86.98

121.30

MEAN

28.00

28.63

27.47

24.95

24.08

24.64

25.11

STD

16.86

18.10

18.21

16.76

15.26

16.25

17.25

MIN

-214.76

-120.85

-25.16

-146.46

-100.74

-71.72

-65.40

30.04

17.36

20.16

27.82

23.72

70.54

46.37

1.40

0.91

2.53

1.56

2.25

3.84

3.88

STD

18.98

12.56

6.18

13.71

12.48

11.71

11.76

MIN

-3.73

-94.15

-10.35

-12.49

-5.63

-30.81

-6.16

MAX

MAX MEAN

LIQUID

21.56

18.71

21.24

32.15

39.79

38.28

63.70

MEAN

2.24

1.01

1.94

2.22

2.46

2.49

2.97

STD

2.95

8.05

2.96

4.04

3.98

4.81

5.77

Table 3 shows the results of the regression (1). We notice that the coefficients of the variable PPSFR are positive and significant in all the regressions. This finding is contrary to our prediction and indicates that the more overfunded the plan is (PPFSR >1) the more likely the manager is to select higher interest rates. Thus, we reject our hypothesis H1. We also note that LEV is significant in the predicted direction. Thus, we confirm our hypothesis H2 that high leveraged firms tend to select higher discount rates in order to understate the degree of their unfunded pension liabilities and avoid the costly renegotiation process. Moreover, as predicted, the coefficient of SIZE is negative and significant at the 1% level. This implies that, consistent with our hypothesis H3, larger firms are more likely to choose conservative actuarial choices to avoid visibility costs (Watts and Zimmerman, 1978; Zimmerman, 1983; Hagerman and Zmijewski, 1979). 200

Salah & Smaoui Table 3 This table shows the results of our model (1) estimated using a GLS method for our sample of 190 firms for the period 2000-2006. The dependent variable is the Interest Rate (IR). The definitions of our explanatory variables appear in Table 1. ***, **, * refer to the 1, 5 and 10% levels of significance respectively. P values are reported between parentheses. Variables

Exp. sign

CONSTANT

PPFSR

-

LEV

+

SIZE

-

SGR

-

ERA

+

ROA

-

LIQUID

-

AUDITOR

-

Wald Chi2 Wald test N

S(1)

Dependant Variable : IR S(2) S(3)

S(4)

2.1248*** (0.000)

2.1309*** (0.000)

1.4189*** (0.000)

1.4617*** (0.000)

0.9641*** (0.000) 0.0043*** (0.000) -0.0489*** (0.000) 0.0873*** (0.000) 0.3989*** (0.000)

0.9991*** (0.000) 0.0042*** (0.000) -0.0452*** (0.000) 0.0822*** (0.000) 0.3940*** (0.000) -0.0037*** (0.001)

0.9175*** (0.000) 0.0042*** (0.000) -0.0667*** (0.000) 0.3062*** (0.000) 0.4092*** (0.000) -0.0036*** (0.005) 0.0018 (0.504)

1223.93 0.000*** 953

1409.52 0.000*** 953

1038.76 0.000*** 941

0.9147*** (0.000) 0.0043*** (0.000) -0.0706*** (0.000) 0.3015*** (0.000) 0.4121*** (0.000) -0.0037*** (0.005) 0.0019 (0.494) -0.0127 (0.760) 1047.32 0.000*** 940

From Table 3, we can see that the coefficients of ROA are all negative and significant at the 1% level. This finding is consistent with Ashtana (1999) and indicates that as the profitability of the sponsoring firm increases, managers are more likely to choose conservative actuarial choices in order to increase tax-deductible contributions. Table 4 presents the results of regression of the determinants of the salary growth rate (SGR). We notice that the coefficients of PPFSR are, as predicted, positive and significant at the 5% level. This result supports our hypothesis H1 and implies that the more overfunded the plan is, the more likely the firm is to make conservative actuarial choices to lower the perceived overfunding. We also note that the coefficients of LEV are negative and significant (at the 10% level) as predicted. Thus, we confirm our hypothesis H2 that the higher the debt, the more likely the firm is to choose lower salary growth rates in order to lower the degree of their unfunded pension liabilities. The coefficients of SIZE have the predicted positive sign, but only significant in regression S(1). Thus, we partially confirm our hypothesis H3. Table 5 displays the results of the regressions of the determinants of the expected rate of return (ERA). We can see that the coefficients of PPFSR are negative and significant as predicted (except in regression S(1)), where PPFSR is not significant). Further, the coefficients of LEV and SIZE are significant in the predicted direction in all the regressions. These results provide support for our hypotheses H1, H2 and H3. In other words, the higher the plan‟s overfunding, the lower the firm‟s leverage, and the higher the firm‟s size, the more likely the firm is to select conservative actuarial assumptions (low IR and ERA, and high SGR). The coefficients of IR and SGR are significant in the predicted directions in all the regressions (except in S(1) and S(2), where SGR is insignificant). These results imply that managers tend to choose either a liberal set of values or a conservative set of 201

Salah & Smaoui values for all three actuarial rates. Finally, ROA, LIQUID and AUDITOR are significant in all the regressions (except in S(3), where ROA is insignificant), but the signs are not in the predicted direction. 3.3.1. Endogeneity Thus fur, we have estimated the models (1)-(3) assuming that all the explanatory and control variables are strictly exogenous. However, some of our explanatory variables are conceptually endogenous. Indeed, the variable PPFSR obtained by dividing the Pension Plan asset by the Pension obligation depends on the choice of the actuarial rates. Hence, PPFSR is endogenous to the actuarial rates. Moreover, the interest rate assumption cannot be separated from the joint process of selecting other actuarial rates such as the salary growth rates (Asthana, 1999). Consequently, actuarial rates are jointly distributed and a system of simultaneous equations in which the actuarial rates should be treated as endogenous must be employed. Further, since the series of actuarial rates are first order auto-correlated, we include the lag of the dependent variable as a regressor. We estimate the following system of simultaneous time-series cross-sectional equations: IRit   0  1 IRit 1   2 PPFSRit   3 LEVit   4 ROAit   5 LIQUID it   6 AUDITOR it  (4)  7 SIZEit   8 SGRit   9 ERAit  i  it SGRit   0  1SGRit 1   2 PPFSRit   3 LEVit   4 ROAit   5 LIQUID it   6 AUDITOR it   7 SIZEit   8 IRit   9 ERAit  i  it ERAit   0  1 ERAit 1   2 PPFSRit   3 LEVit   4 ROAit   5 LIQUID it   6 AUDITOR it   7 SIZEit   8 IRit   9 SGRit  i  it

(5) (6)

202

Salah & Smaoui Table 4 This table shows the results of our model (2) estimated using a GLS method for our sample of 190 firms for the period 2000-2006. The dependent variable is the Salary Growth Rate (SGR). The definitions of our explanatory variables appear in Table 1. ***, **, * refer to the 1, 5 and 10% levels of significance respectively. P values are reported between parentheses. Variables

Exp. sign

CONSTANT

PPFSR

+

LEV

-

SIZE

+

IR

-

ERA

-

ROA

+

LIQUID

+

AUDITOR

+

Wald Chi2 Wald test N

S(1)

Dependant Variable: SGR S(2) S(3)

S(4)

2.5805*** (0.000)

2.6954*** (0.000)

2.9311*** (0.000)

2.9004*** (0.000)

0.0666** (0.014) -0.0016*** (0.001) 0.0195*** (0.007) 0.1342*** (0.000) 0.0369*** (0.007)

0.1028*** (0.004) -0.0020*** (0.000) 0.0055 (0.536) 0.1297*** (0.000) 0.0376*** (0.006) -0.000 (0.985)

0.2305*** (0.000) -0.0009* (0.083) 0.0071 (0.331) 0.1716*** (0.000) -0.0576*** (0.000) 0.0015** (0.024) -0.0051*** (0.000)

206.78 0.000*** 953

275.24 0.000*** 953

3014.75 0.000*** 941

0.2239*** (0.000) -0.0009* (0.097) 0.0058 (0.455) 0.1712*** (0.000) 0.0514*** (0.000) 0.0012* (0.082) -0.0048*** (0.000) -0.0028 (0.953) 3376.64 0.000*** 940

Table 5 This table shows the results of our model (3) estimated using a GLS method for our sample of 190 firms for the period 2000-2006. The dependent variable is expected rate of return on asset actuarial assumption (ERA). The definitions of our explanatory variables appear in Table 1. ***, **, * refer to the 1, 5 and 10% levels of significance respectively. P values are reported between parentheses. Variables

Exp. sign

CONSTANT

PPFSR

-

LEV

+

SIZE

-

IR

+

SGR

-

ROA

-

LIQUID

-

AUDITOR

-

Wald Chi2 Wald test N

S(1)

Dependant Variable: ERA S(2) S(3)

S(4)

4.5538*** (0.000)

4.5371*** (0.000)

4.6110*** (0.000)

4.5055*** (0.000)

-0.0325 (0.102) 0.0012** (0.025) 0.1056*** (0.000) 0.3209*** (0.000) -0.0001 (0.997)

-0.0687*** (0.008) 0.0014*** (0.010) 0.1048*** (0.000) 0.3281*** (0.000) 0.0011 (0.923) 0.0013*** (0.000)

-0.1918*** (0.000) 0.0027*** (0.000) 0.0850*** (0.000) 0.4205*** (0.000) -0.1267*** (0.000) 0.0011 (0.000) 0.0067*** (0.000)

819.05 0.000*** 953

847.02 0.000*** 953

2011.25 0.000*** 941

-0.1731*** (0.000) 0.0025*** (0.000) 0.0906*** (0.000) 0.4163*** (0.000) -0.1213*** (0.000) 0.0012*** (0.005) 0.0060*** (0.001) 0.0777* (0.064) 1810.64 0.000*** 940

To tackle the endogeneity of the explanatory variables, we use the two-step System GMM estimator proposed by Arellano and Bover (1995) and Blundell and Bond (1998). This estimator combines, within a system, the regression in levels and the regression in first differences and employs a series of instrumental variables to remove the endogeneity of the explanatory variables. The instruments for the regression in levels are the lagged first 203

Salah & Smaoui differences of the endogenous and exogenous variables. For the regression in differences, the instruments are the lagged endogenous and exogenous variables previous or equal to (t-2). In our models (4)-(6), the actuarial assumptions (IR; SGR, and ERA) and PPFSR are treated as endogenous, the lagged dependent variables are considered as predetermined, while the rest of the variables are assumed to be strictly exogenous. The consistency of the system GMM estimator depends on the assumption that the instruments are valid and that the error terms are not serially correlated. To test both hypotheses, we run two specification tests. The first is a test of over-identifying restrictions proposed by Hansen (1980) which tests the overall validity of the instruments. The second test proposed by Arellano and Bond (1991) and Arellano and Bover (1995) tests the null hypothesis that the difference error term shows no second order serial auto-correlation. Our model specification is adequate if we cannot reject the null hypotheses. Table 6 displays the results of the system GMM estimation. We notice that, for all specifications, the test of Hansen (1982) cannot reject the null hypothesis of the overall validity of our instruments at the 1% level. Moreover, the AR2 test cannot reject the null hypothesis of absence of autocorrelation of the second order in the residuals at the 1% level. All our specifications satisfy the diagnostic tests that our instruments are appropriate, and that there is no second order autocorrelation in the differenced residuals. The coefficients of the lagged dependent variable (LDEP), shown in Table 6, are positive and highly significant (p-value less than 1%). This evidence justifies the use of dynamic panel models to analyze the determinants of the actuarial rates and indicates that the choices of the actuarial rates are path-dependent. We note from Table 6 that the coefficients of PPFSR are no longer significant at the 5% level. Thus, after controlling for the endogeneity problem of the explanatory variables and the simultaneity of the actuarial rates, we find no support for our hypothesis H1 that underfunded (overfunded) firms tend to select liberal (conservative) actuarial assumptions. Consistent with our previous results, the coefficients of LEV are significant, at the 5% level, in all the regressions (expect in the regression of SGA) in the predicted directions. This evidence provides further support to our hypothesis H2 that the higher the firm‟s leverage, the more likely the firm is to choose liberal actuarial rates (High IR and ERA, and low SGR) to understate the degree of their unfunded pension liabilities. In other words, as leverage rises, managers may choose to divert the available cash to reimburse debt holders rather than contributing to the firm‟s pension plans (Beneish and Press 1995). ROA, LIQUID, and AUDITOR are insignificant at the 5% level in all the regressions, except in the ERA regression where AUDITOR is significant at the 5% level, but has the unpredicted positive sign. From Table 6, we note that the coefficient of SIZE is significant at the 5% in the IR regression, and at the 10% level in the SGR regression in the predicted directions, but insignificant in the ERA regression. These results provide overall support for our hypothesis H3 that large firms are more likely to select liberal actuarial rates in order to minimize visibility costs and federal taxes and penalties.

204

Salah & Smaoui Table 6 This table shows the results of our models (4)-(6) estimated with the with the System GMM procedure of Blundell and Bond (1998) for our sample of 190 firms for the period 2000-2006. The dependent variables are the actuarial rates (IR; SGR; and ERA). LDEP is the lag of the dependent variable. The definitions of our explanatory variables appear in Table 1. The Hansen (1982) test tests the validity of our instruments, while AR2 is the Arellano and Bond (1991) test of the absence of second order autocorrelation in the differentiated residuals, ***, **, * refer to the 1, 5 and 10% levels of significance respectively. Two-step system GMM estimator is used. Windmeijer (2005) finite-sample correction to the two-step covariance matrix is employed. Robust standard errors consistent in the presence of heteroskedasticity and autocorrelation within the panel are reported. Variables LDEP PPFSR LEV ROA LIQUID AUDITOR SIZE

IR 0.488*** (0.000) 0.030 (0.976) 0.057** (0.047) -0.006 (0.855) 0.021 (0.714) 1.355 (0.500) -0.475** (0.016)

IR SGA ERA Nbr of Instruments Hansen test AR2 test

0.046 (0.933) 0.453 (0.224) 15 0.130 0.638

SGA 0.287*** (0.004) 0.768* (0.054) 0.020 (0.117) 0.009 (0.485) 0.079 (0.485) -0.105 (0.428) 0.053* (0.097) 0.172*** (0.007)

0.090 (0.548) 31 0.118 0.675

ERA 0.704*** (0.000) 0.465 (0.217) 0.004** (0.013) 0.020* (0.078) 0.001 (0.935) 0.180** (0.027) 0.084 (0.284) 0.711 (0.203) 0.069 (0.659)

25 0.249 0.580

4. Concluding Remarks In this paper, we study the effects of firms‟ financial and pension plan characteristics on their actuarial choices for a sample of 190 Canadian firms over the period 2000-2006. We posit that firms‟ characteristics provide managers with strong incentives to manage the funding of their DBPPs through their actuarial choices. We examine the determinants of three of the most influential actuarial choices: interest rates, salary growth rates, and expected rates of return on plan assets. The research design uses a simultaneous GMM estimation approach, thereby allowing the three actuarial rates to be jointly distributed, and resolving the endogeneity problem of the explanatory variables. The evidence shows that the higher the firm‟s leverage, the more likely the firm is to choose liberal actuarial rates (high IR and ERA, and low SGR) to understate the degree of their unfunded pension liabilities. Indeed, with increased debt, managers may prefer to divert the available cash to pay creditors and bondholders, thereby avoiding costly violation of debt covenants. Furthermore, larger firms are more likely to choose conservative actuarial choices to avoid visibility costs. However, contrary to the received literature, no support is provided for our hypothesis H1. This result suggests that the level of funding status does not appear to be a major determinant of the actuarial choices. 205

Salah & Smaoui The major limitation of this study is that it was carried out over a short time period (from 2000 to 2006). It would be interesting to reexamine the determinants of the actuarial assumptions using more recent data covering a much larger period. All in all, the evidence in this study can help pension plans‟ beneficiaries evaluating and predicting corporate funding behavior, and Canadian regulators to better regulate the DBPPs.

References Arellano, M & Bond, S 1991, „Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations‟, Review of Economic Studies, Vol. 58, pp. 277-297. Arellano, M & Bover, O 1995, „Another Look at the Instrumental Variable Estimation of Error Component Models‟, Journal of Econometrics, Vol. 68, pp. 29-51. Asthana, S 1999, „Determinants of Funding Strategies and Actuarial Choices for DefinedBenefit Pension Plans‟, Contemporary Accounting Research, Vol. 16, No. 1, pp. 3974. Ben Salah, I, Coulombe, D, & Paquette, S 2013, „Market Valuation of Pension Plan Information and Pension Actuarial Assumptions: A Canadian Study‟, Working Paper. Canada: Universite Laval, Quebec City. Beneish, MD & Press, E 1993, „Cost of Technical Violation of Accounting-Based Debt Covenants‟, The Accounting Review, Vol. 68, No. 2, pp. 233-257. Blankley, AL & Swanson, EP 1995, „A Longitudinal Study of SFAS 87 Pension Rate Assumption‟, Accounting Horizon, Vol. 9, No. 4, pp. 1-21. Blundell, R & Bond, S 1998, „Initial Conditions and Moment Restrictions in Dynamic Panel Data Models‟. Journal of Econometrics, Vol. 87, pp.115-143. Bodie, Z, Light, JO, Morck, R, & Taggart, RA 1987, „Funding and Asset Allocation in Corporate Pension Plans: An Empirical Investigation‟, University of Chicago Press, pp.15-47. Bregstresser, D, Desai, M, & Rauh, J 2006, „Earning Manipulation, Pension Assumptions, and Managerial Investment Decisions‟, The Quarterly Journal of Economics, Vol. 121, pp.157-195. Brown, S 2006, „The Impact of Pension Assumption on Firm Value‟, Working Paper. University of Maryland - Department of Accounting & Information Assurance. Comprix, J, & Muller, KA 2011, „Pension Plan Accounting Estimates and the Freezing of Defined Benefit Pension Plans‟, Journal of Accounting and Economics, Vol. 151, pp.115-133. Dechow, PM, Ge, W, Larson, CR, & Sloan, RG 2011, „Predicting Material Accounting Misstatements‟, Contemporary Accounting Research, Vol. 28, No 1, pp. 17-82. Dhaliwal, DS 1986, „Measurement of Financial Leverage in the Presence of Unfunded Pension Obligations‟, The Accounting Review, Vol. LXI, No. 4, pp. 651-661. Ghicas, DC 1990, „Determinants of Actuarial Cost Method Changes for Pension Accounting and Funding‟, The Accounting Review, Vol. 65, No. 2, pp. 384-405. Godwin, JH, Goldberg, SR & Duchac, JE 1996, „An Empirical analysis of Factors Associated with Changes in Pension-Plan Interest Rate Assumptions‟ Journal of Accounting Auditing and Finance, Vol. 11, No. 2, pp. 305-322.

206

Salah & Smaoui Gopalakrishnan, V & Sugrue, FT 1995, „The Determinant of Actuarial Assumptions under Pension Accounting Disclosures‟, Journal of Financial and Strategic Decisions, Vol. 8, No. 1, pp. 35-41. Hagerman, RL & Zmijewski, ME 1979, „Some Economic Determinants of Accounting Policy Choice‟, Journal of Accounting and Economics, Vol. 1, pp. 142-161. Hann, R Lu, Y & Subramanyam, KR 2007, „Does Discretion Improve or Impair Value Relevance? Evidence from Pricing of the Pension Obligation‟, The Accounting Review, Vol. 82, No. 1, pp. 107-137. Hansen, LP 1982, „Large Sample Properties of Generalized Method of Moments Estimators‟, Econometrica, Vol. 50, pp.1029-1054. Hausman, JA 1978, „Specification Tests in Econometrics‟, Economica, Vol. 46, No. 6, pp. 1251-1271. Healy, PM, 1985, „The Effect of Bonus Schemes on Accounting Decision‟, Journal of Accounting and Economics, Vol. 7, No. 1, pp. 85-107. Healy, PM & Palepu, KG 1990, „Effectiveness of Accounting-Based Dividend Covenants‟, Journal of Accounting and Economics, Vol. 12, No. 1, pp. 97-123. Mittelstaedt, H 1989, „An Empirical Analysis of the Factors Underlying the Decision to Remove Excess Assets from Overfunded Pension Plans‟, Journal of Accounting and Economics, Vol. 11, No. 4, pp. 399-419. Press, EG & Weintrop, JB 1990, „Accounting Based Constraints in Public and Private Debt Agreements, Their Association with Leverage and Impact on Accounting Choice‟, Journal of Accounting and Economics, Vol. 12, No. 1, pp. 65-95. Tepper, I 1981, „Taxation and Corporate Pension Policy‟, The Journal of Finance, Vol. 36, No. 1, pp. 1-13. Thomas, JK 1988, „Corporate Taxes and Defined Benefit Pension Plans‟, Journal of Accounting and Economics, Vol. 10, pp. 199-237. Thomas, JK 1989, „Why Do Firms Terminate Their Overfunded Pension Plans?‟ Journal of Accounting and Economics, Vol. 11, No. 4, pp. 361- 399. Thomas, JK & Tung, S 1992, „Cost Manipulation Incentives Under Cost Reimbursement: Pension Cost for Defense Contracts‟, The Accounting Review, Vol. 67, pp. 691-711. Watts, RL & Zimmerman, JL 1986, „Positive Accounting Theory: a Ten Year Perspective‟, The Accounting Review, Vol. 65, No.1, pp. 131-156. Weidman, CI & Weir, HE 2004, „The Market Value Implications of Post-Retirement Benefit Plans and Plan Surpluses- Canadian Evidence‟, Canadian Journal of Administrative Sciences, Vol. 21, No.3, pp. 229-241. Windmeijer, F 2005, „A Finite Sample Correction For the Variance of Linear Efficient TwoStep GMM Estimators‟, Journal of Econometrics, Vol. 126, pp.25-51.

207