Pension plans’ funded status volatility and corporate credit risk

Abstract

We investigate whether and how U.S. pension plans’ funded status volatility affects firm credit risk. We first show that a firm's funded status volatility is positively related to its bond yield spread. We then find that the adoption of SFAS No.158 (2006) requiring the recognition of pension funding status on the statement of financial position renders the pension plan information more value-relevant, thereby increasing the effect of funded status volatility on bond yield spread. Furthermore, the predictions that funded status volatility affects asset value volatility and incomplete accounting information, which in turn affects corporate credit risk, are empirically supported. Our findings reinforce the need for firms to disclose reliable information about funded status volatility—a major pension plan risk—to external investors.

Keywords:

• Funded status volatility
• Credit risk
• Bond yield spread
• Pension plans
• SFAS No.158 (2006)
• Structural-form credit models

Acknowledgement

We thank the participants at the 2016 EAA Annual Congress and 2016 AAA Annual Meeting for their helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Pension plan underfunding means the level of expected defined benefit obligations exceeds the fair value of assets held by the plan.

2 According to the research report of FTSE Russell (2019), corporate DB plan sponsor surveys demonstrate that effective control of funded status volatility (namely surplus risk) is a top priority for their organizations. Halim et al. (2010) demonstrate that surplus risk exposures (also called balance sheet risk) tend to be materially greater than active management risk exposure (e.g., controlling for tracking errors) for pension funds.

3 The funded status is defined as the fair value of the pension assets less the projected benefit obligation. The definition of funded status follows the Financial Accounting Standards Board (FASB). Underfunded status represents a firm's pension plan liability shortfall.

4 The terms surplus and surplus risk are commonly used in practice and in finance research (Leibowitz 1987, Ezra 1991, Martellini 2006, Halim et al. 2010), which are the same as the terms funded status level and funded status volatility, respectively. To avoid confusion, we use the terms funded status level and funded status volatility in our study.

5 For example, the investment advisory firms that provide ALM and LDI include Willis Towers Watson, Ortec Finance, Milliman, and NISA Investment Advisors. Moreover, NISA Investment Advisors introduces the Pension Surplus Risk Index, a forward-looking estimate of the funded status volatility of US corporate defined benefit pension plans, to monitor the funded status volatility.

6 It is assumed that companies with higher funded status volatility have made rational cost-benefit-based decisions that the costs of managing such volatility (e.g., higher remuneration costs to offset the transfer of pension risk to employees, foregone returns from riskier investments, insurance/hedging premiums, etc.) would exceed the benefits of lower volatility.

7 However, it has to be noted that the firm should recognise the additional minimum pension liability—the difference between minimum pension liability (accumulated benefit obligation (ABO) minus fair value of plan asset) and accrual pension liability, as an increase in accrual pension liability when the amortised underfunded status is relatively smaller.

8 In addition, other pension regulatory changes, such as SFAS No.132R (2003) and SFAS No.132R-1(2008), mainly focus on the disclosure quality improvement of pension plan asset types.

9 Different from Kalogirou et al. (2021) that discuss the credit market reactions to a more transparent disclosure on pension funded status (the adoption of IAS 19 (2005)), this study further explores how the credit market responses to a shift from disclosure to recognition of pension funded status (the implementation of SFAS No. 158 (2006)).

10 The FASB (2006) also requires firms to disclose their estimation of the projected benefit obligation, the fair value of pension fund assets, and the funded status (the difference between the fair value of pension fund assets and projected benefit obligation). Prior research also demonstrates that pension plans’ funded status level can capture the extent of limitations on the use of internal resources and risky investments (Rauh 2006, Jones 2013, Liu and Tonks 2013).

11 Martin and Henderson (1983) discuss the relationship between bond ratings and pension obligations while Maher (1987) explores the effect of net pension asset (liability) on bond rating in the U.S. market. Additionally, McKillop and Pogue (2009) find that pension plan risk (i.e., pension liability level) is negatively related to debt rating in the UK market.

12 McKillop and Pogue (2009) mainly follow Jin et al. (2006) to employ the weighted average of systematic risk exposures of pension assets and liabilities as a pension risk measure, which captures the systematic risk (market risk) of a firm's pension asset and pension liability rather than total risk (both systematic and firm-specific risk).

13 Yu (2005), Francis et al. (2005), Güntay and Hackbarth (2010), and Lu et al. (2010) respectively employ the corporate disclosure ranking assessed by financial analysts, accrual quality, analyst forecasts, and trading-based information asymmetry to explain the variation of bond yield spreads.

14 Other firm- and bond-related determinant variables of bond yield spreads in previous studies include R&D intensity (Shi 2003, Eberhart et al. 2008), amount issued (Yu 2005), bond age (Yu 2005, Warga,1992) and other variables. For the R&D intensity effects on bond yield spreads, Shi (2003) finds that a firm's R&D intensity has a negative impact on bondholder returns while Eberhart et al. (2008) document the opposite effect.

15 See SFAS No.158 in the Introduction section.

16 However, a larger increase in funded status for firms with debt-contracting incentives is not driven by firms taking real economic actions such as amending the plan or freezing the retirement benefits (Jones 2013).

17 In the pre-IAS 19 (2011) period, firms are allowed to either defer recognition of the actuarial gains and losses resulting from defined benefit pension plans (corridor method) or fully recognise the actuarial gains and losses in profit or loss / OCI. The discretion of recognising the actuarial gains and losses is criticized to impede comparability (Glaum et al. 2018). Subsequently, IAS 19 (2011) eliminated the so-called ‘corridor method.’

18 This argument is derived from the statistical principle that the variance of the difference between two random variables equals the sum of their variances minus twice their covariance.

19 The inference, based on Baber et al. (2011), is that due to double-entry bookkeeping under accrual accounting, earnings manipulations can create a bank of cumulative discretionary accruals that carries forward into future periods and constrains the ability to manage future income when prior discretionary accruals eventually reverse back to the income statement.

20 It is possible that ‘smart manipulators’ have multi-period time horizons. The current pension accounting standards under U.S. GAAP allow firms as pension plan sponsors to enjoy the income statement benefits of manipulating actuarial assumptions while avoiding reporting the concomitant risk (Gold 2005, Anantharaman et al. 2021). Since a firm's financial report is not required to report the funded status volatility directly, managers may have less incentive to maintain the stability of the funded status level. That is, managers are more likely to manipulate pension plan assumptions to meet a firm's earnings target (either to inflate earnings or to smooth earnings) and avoid reporting huge funding deficit, with the side effects of greater funded status volatility (as previously discussed). This feature is expected to exist in a multi-period setting, where managers initiating manipulation activities need to consider the potential impacts of such activities on future periods. Thus, their ability to manage toward their objectives (e.g. earnings level) are constrained by the prior manipulation activities.

21 This argument is based on Jin et al. (2006), who indicate pension accounting rules as complicated and opaque. Jin et al. (2006) also indicate that pension plan assets and liabilities are off-balance sheet and are often viewed as segregated from the rest of the firm. In response to the need of outsiders for more information about pension plan assets, SFAS No.132R (2003) was developed and required firms to disclose pension asset allocation strategies. However, FASB (2006) argued that sensitivity information may be misunderstood by investors and may not adequately consider the interdependency of certain assumptions. Therefore, SFAS No.132R (2003) did not require disclosure of sensitivity information about hypothetical changes in discount rates, expected long-term rates of return on assets, or rates of future compensation increase. Moreover, a recent study by Anantharaman and Chuk (2020) indicates that the pension accounting regulation under U.S. GAAP cannot effectively improve pension asset transparency.

22 Christensen and Nikolaev (2017) show that contracts can, and often do, adapt to GAAP changes. Therefore, the argument holds only when firms write fixed GAAP contracts (contracts that are not affected by GAAP changes).

23 This is primarily because bondholders are concerned with a firm's asset value distributions, which are the main credit risk determinants (Merton 1974, Duffie and Lando 2001).

24 The debt book value is defined as the sum of debts in current liability plus long-term debts. It has to be noted that pension-related liabilities are not included in the book value of debt.

25 DISP is calculated as the standard deviation of the analysts’ fiscal year 1 earnings per share forecasts scaled by the absolute value of mean.

26 These sample requirements may create survivorship bias.

27 The ratio of initial investment grade bonds to the initial sample (preliminarily screened 33,205 bond observations) is 64.97%.

28 This suggests that our final sample may be at the risk of sample selection bias. Therefore, we employ a two-stage Heckman model to address this, shown in section 5.4.1.

29 In January 2013, the Federal Reserve Committee (Fed) began to consider the exit mechanism for the QE policy. That is, when the expected inflation rate in the United States is higher than 2.5% and the unemployment rate is lower than 6.5%, the Fed will adopt a gradual reduction in the scale of debt purchases. Then, the Fed decided to start reducing the scale of QE policy from January 2014 by reducing bond purchases. Finally, the Fed withdrew from QE policy in October 2014 and is expected to raise interest rates around mid-2015 (actually in the end of 2015). Based on the above discussions, the scale reduction and exit of QE policy may therefore weaken the incentive of issuing corporate bonds for a firm, which leads to a decreasing trend of corporate bonds from year 2012 to 2014. In 2015, the value of U.S. M&A transactions reached US$2.41 trillion, a sharp increase of 55% over the previous year and a record high since 1985. Since the capital environment with low interest rates can stimulate the large-scale M&A activities, U.S. firms have more incentives to issue more corporate bonds to gather funding liquidity for M&A transactions with a lower financing cost before the Fed's raising interest rate. Therefore, the expectation of raising interest rate and the need for funding for M&A transactions may enhance the incentive of issuing corporate bonds for a firm, which leads to an increase on corporate bonds in 2015.

30 It is noted that the effect of SIZE (Lnamt) is more pronounced in the pre-SFAS (post-SFAS) No.158 period. It is well known that firm size is negatively associated with proxies for information asymmetry (e.g., bid-ask spread), suggesting that the firm size can be viewed as an adverse proxy of information asymmetry (Yoon et al. 2011). In the post-period of SFAS No.158, debtholders are more informed about a firm's pension funding condition, and their concerns regarding this lack of pension funding information are thus alleviated, which may reduce the effects of firm size on bond yield spread. As a result, the effect of company size is more (less) pronounced in the pre-SFAS (post-SFAS) No.158 period. Moreover, the bond issue amount is an external liquidity proxy in Yu (2005), and higher issue amounts signify greater liquidity of a bond. Adrian et al. (2017) indicate that the financial crisis in 2007–2009 highlighted the need to better understand corporate bond market liquidity, suggesting that debtholders may care more about bond external liquidity after the financial crisis. Since SFAS No.158 was issued in 2006, it is expected that the bond issue amount is more relevant in the post-SFAS No.158 period.

31 Dushi et al. (2010) and Kisser et al. (2013) have documented the importance of life expectancy (at retirement age) assumptions on the valuation of pension liabilities for firms with DB pension plans, and the longevity risk exists due to the increasing life expectancy trends of people reaching retirement age. Considering the context of DB pension plans in our study, we decide to employ life expectancy for people aged 65 in the U.S. to capture the estimated life expectancy at retirement age as the measure of longevity. Data sources for the real-factor proxies are the Federal Reserve Economic Data and the U.S. Centers for Disease Control and Prevention (CDC) website.

32 To measure financial constraints, we employ the Size and Age index (SA, Hadlock and Pierce 2010) as the main proxy and view the Kaplan and Zingales Index (KZ, Kaplan and Zingales 1997) as another proxy. The results of using the SA index and those of using the KZ index lead to similar conclusions.

33 Because the variable of equity allocations is firm-level rather than bond-level, the number of (bond-year) observations in low-PNATE firms may not be exactly three times of that in high-PNATE firms.

34 We compare the equality of FSV5 coefficients from two subsamples using seemingly unrelated estimations.

35 The negative sign and lack of significance for LEV in high-PNATE firms, in contrast to the positive sign for low-PNATE firms that one would expect for both groups, is noted. Goto and Yanase (2021) indicate that firms that choose high ERR tend to exploit flexible internal financing from employees via pension underfunding. Because such firms are more likely to be high-PNATE firms, the finding of Goto and Yanase (2021) suggests that high-PNATE firms are more likely to use internal financing to obtain capital for growth and thus have a lower financial leverage ratio. We compare the leverage ratio between high- and low-PNATE firms. The average leverage ratio for high-PNATE firms is significantly lower than that for low-PNATE firms (the significance level is lower than 1%). Given that high-PNATE firms prefer internal financing over external financing, their leverage ratio may be less sensitive to the bond yield spreads. In contrast, low-PNATE firms tend to use more external financing and have a higher leverage ratio, which may increase debtholders’ concerns regarding the leverage ratio and strengthen the impact of LEV on their bond yield spreads. The above argument can lead to our empirical finding that the impact of LEV is more pronounced in low-PNATE firms.

36 An indirect effect implies a causal hypothesis whereby an independent variable causes a mediating variable, which, in turn, causes a dependent variable (Sobel 1990).

37 It is known that Heckman two-step estimates are highly sensitive to changes in sample and model specification (Clatworthy et al. 2009). We use a different set of independent variables and fixed effects in the first stage, and the results are robust.

38 Specifically, the explanatory variables in the first-stage model include financial leverage (LEV), equity volatility (VOL), firm asset size (SIZE), return on assets (ROA), ROA volatility (ROAV), the natural logarithm of net sales (SALE), firm S&P credit rating (FRAT; a higher FRAT value reflects lower credit quality), firm sales per employee (Sales_EMP), the natural logarithm of the number of firm employees (EMP), pension- and retirement-related expense per unit asset (XPR_TA), unionisation rate (Union_R), industry-level average employee age (EMP_Age), and firm age (Fage). The detailed definitions of the remaining variables are shown in Sections 4.1 and 4.2.

39 We conduct two additional robustness tests. First, we use an interaction model to examine the moderating effects of overfunded status condition and expected return on pension assets/ pension fund equity allocations. Specifically, we add the interaction terms of funded status volatility and the moderating variables into the regression model of Eq. (2) for the pooled sample. Second, we examine the effect of SFAS No. 158 (2006) under different (narrower) time windows, using both interaction variable tests and coefficient difference tests. We examine the three, four, and five years before and after the implementation of SFAS No. 158 (2006), denoted as [-3, +3], [-4, +4], and [-5, +5], respectively. Overall, the results from these tests (unreported, but available on request from the authors) are consistent with our main empirical findings.

Additional information

Funding

This work was supported by Ministry of Science and Technology, Taiwan (Now National Science and Technology Council, Taiwan): [Grant Number MOST 103-2410-H-030 -007-MY3].