Finance Research Letters 58 (2023) 104279

Contents lists available at ScienceDirect

Finance Research Letters

journal homepage: www.elsevier.com/locate/frl

Corporate governance and shareholder-employee risk-shifting:

Evidence from corporate pension plan sponsors
Shingo Goto a, *, Noriyoshi Yanase b
a
The University of Rhode Island, College of Business, 7 Lippitt Road, Kingston, RI 02881, United States of America
b
Keio University, Faculty of Business and Commerce, 2-15-45, Mita, Minato-ku, Tokyo 108-8345, Japan

A R T I C L E I N F O A B S T R A C T

JEL Classification: This study examines how corporate governance affects pension plan sponsors’ conflicting
G22 motives—risk-shifting vs. risk management—and their future pension funding status, R&D in­
G32 vestments, and shareholder values. We focus on managers’ discretionary choices of expected rates
Keywords: of return on plan assets and discount rates for pension liabilities to capture risk-shifting motives
Corporate governance and measure corporate governance quality through foreign equity ownership (strong monitoring)
Corporate pensions
and stable equity ownership (weak monitoring). The evidence shows that well-governed firms use
Ownership structure
managerial risk-shifting to finance R&D and increase shareholder values. Risk-shifting by poorly-
Risk Shifting
R&D governed firms results in larger pension underfunding without increasing shareholder values.
Internal Financing

1. Introduction

Corporate pension policies involve a trade-off between two conflicting motives—risk management and risk shifting—in the
presence of shareholder-employee conflicts. Managers of pension plan sponsors have discretion in managing their pension plans. These
features make defined benefit (DB) corporate pension plans a fertile ground to examine the implications of the two conflicting
managerial motives (e.g., Rauh, 2009).
This study examines how managers’ risk-shifting (vs. risk management) motives precede future pension funding status and risky
corporate investments such as research and development (R&D) projects. Bartram (2017) and Duong et al. (2022) examine the
employee retention and risk-shifting channels through which DB plans may spur or stifle R&D and innovations. We focus on the effects
of corporate governance on managers’ risk-shifting motives in increasing R&D expenditures.
To capture managerial risk-shifting (vs. risk management) motives, we focus on the managers’ discretionary choices of the expected
rate of return on plan assets (“ERR”) and the assumed discount rate for estimating pension liabilities. Risk-seeking managers can
choose higher ERRs by taking riskier pension asset allocation strategies.1 They can also choose higher discount rates to reduce reported
pension liabilities (e.g., Kisser et al., 2017; Armitage et al., 2022; Chu et al., 2023).2 Higher rate assumptions (ERRs and discount rates)

* Corresponding author.
E-mail address: [email protected] (S. Goto).
1
Since Rauh (2009), many studies have used the riskiness of pension asset allocations to measure the degree of risk-shifting. However, Japanese
DB plan sponsors have provided at most eight pension asset allocations since 2014, which is not long enough for our analyses of dynamic (lead-lag)
relationships among variables. Managers typically choose higher ERRs and riskier pension asset allocation strategies jointly.
2
Under the Japanese GAAP, DB plan sponsors do not need to change their discount rates unless the PBO estimates change by 10% or more.

https://doi.org/10.1016/j.frl.2023.104279
Received 4 March 2023; Received in revised form 17 June 2023; Accepted 27 July 2023
Available online 28 July 2023
1544-6123/© 2023 Elsevier Inc. All rights reserved.
S. Goto and N. Yanase Finance Research Letters 58 (2023) 104279

allow the managers to lower future cash contributions to the pension plans and increase risky investments—a blatant risk-shifting from
the shareholders to employees.3
Goto and Yanase (2021) show that higher rate assumptions precede (Granger-cause) larger pension underfunding and investments
(especially R&D expenditures), suggesting that high (or low) rate assumptions reveal managers’ intent to shift risk (or manage risk).
Their findings are primarily driven by firms that likely face high external financing costs, consistent with the arguments that plan
sponsors can use the flexible internal debt financing feature of underfunded DB plans to shift risk (e.g., Bartram, 2017, 2018; Chaudhry
et al., 2017; Kisser et al., 2017; Balachandran et al., 2019; Goto and Yanase, 2021; Duong et al., 2022). However, few studies have
examined how corporate governance affects the managers’ risk-shifting motives to utilize the internal debt financing feature of pension
underfunding. This study seeks to fill the gap by examining how corporate governance affects the economic consequence of managerial
risk-shifting.
Our empirical analyses use a firm-level panel of Japanese DB plan sponsors listed on the First Section of the Tokyo Stock Exchange
(TSE1), mitigating the sample selection concern for studies of US DB plan sponsors.4 Because pension funding regulations are more
lenient in Japan than the US,5 Japanese plan sponsors have larger discretion in pursuing risk-shifting motives than US sponsors do.6
Using Quinn and Rivoli’s (1991) organization theory, Duong et al. (2021) suggest that DB plans resembling a “Japanese system” may
promote innovations. It is thus interesting to study whether and how Japanese firms use DB plans to finance R&D investments.
We use two equity ownership measures for the corporate governance quality: (1) the percentage of shares held by foreign (non-
Japanese) investors (“foreign ownership”) and (2) the percentage of shares held by gray institutions and stable shareholders (“stable
ownership”).7 The choice of these two ownership measures follows the existing evidence that foreign investors have high monitoring
abilities. In contrast, gray institutions and stable shareholders have low monitoring abilities due to business interests (e.g., Chen et al.,
2007). Johnston and McAlevery (1998) show that Japanese firms use stable ownership as a deterrent for investor monitoring and
market disciplines, while those with sizeable foreign (non-Japanese) ownership shares make more value-enhancing investments than
others (e.g., David et al., 2006).
We find evidence for the significant effects of corporate governance on managerial risk-shifting. Higher rate assumptions precede
larger R&D expenditures, especially among firms with sizeable foreign ownership. Among the firms with sizeable stable ownership,
higher discount rates reduce pension liabilities in the current year but increase pension underfunding in the following year. Higher rate
assumptions precede higher shareholder values, but this positive valuation effect disappears for the firms with sizeable stable
ownership.
These results are consistent with the following narrative: Managers monitored by foreign (non-Japanese) investors use higher rate
assumptions to overcome the difficulty of financing R&D projects by lowering plan contributions, which increases shareholder values
by mitigating underinvestment problems. Well-governed firms’ high-rate assumptions are not overly optimistic. They do not increase
pension underfunding, consistent with Phan and Hedge’s (2013) finding for well-governed US plan sponsors. In contrast, higher rate
assumptions (especially discount rates) of poorly-governed firms (with sizeable stable ownership) precede poorer pension funding
status without improving shareholder values.8
This study is subject to various endogeneity concerns. We address the unobserved heterogeneity and simultaneity biases using the
firm- and year-fixed effects. We also control for forward-looking variables to mitigate reverse-causation concerns, including man­
agement forecasts of one-year ahead earnings per current book equity (ROE) and sales growth, as well as market-to-book ratios and
recent stock returns. We use dynamic panel models to address the severe endogeneity concern that rate assumptions and control
variables may depend on lagged dependent variables (Wintoki et al., 2012). We address the additional endogeneity concern due to the
inclusion of lagged dependent variables and firm-fixed effects by implementing the Arellano-Bover (1995)/Brundell-Bond (1998)
system-GMM estimator.

2. Hypotheses

Since Rauh (2009), several studies have investigated factors driving the managerial propensity to shift or manage risk. Among
them, Phan and Hedge (2013) show that well-governed firms with strong investor monitoring tend to achieve higher average plan asset
returns by taking riskier asset allocations (i.e., risk-shifting), resulting in a better pension funding status.

3
Japan has no termination insurance program or entity similar to US PBGC or UK PPF. Consequently, increasing the credit risk of a DB plan
sponsor shifts the risk from the shareholders to the employees.
4
There has been a trend away from DB plans toward defined contribution plans in the US. Still, DB plans are prevalent and economically
important outside the US (Bartram, 2017). According to Goto and Yanase (2021), 96% of TSE1 firms sponsor DB plans.
5
See Sasaki (2015) and Goto and Yanase (2021) for institutional details. No pension plan termination program or entity in Japan (e.g., US PBGC,
UK PPF) enforces mandatory plan contributions. Moreover, Japanese DB plan sponsors facing difficulty in making pension contributions can for­
feit/reduce vested and unvested retirement benefit obligations per regulatory approval. In the US, ERISA strictly prohibits forfeiting/reducing
employees’ vested retirement benefits for any reason.
6
German plan sponsors also have more discretion than US/UK plan sponsors. See Heusel and Mager (2023).
7
Ferreira and Matos (2008) and Aggarwal et al. (2011) define/measure grey institutional ownership as the percentage of shares held by bank
trusts, insurance companies, and other institutions (e.g., pension funds, endowments). Stable shareholders of Japanese firms also include Main Banks
and Keiretsu firms.
8
For example, plan sponsors can use high ERRs to inflate earnings and high discount rates to deflate pension liabilities (e.g., Bergstresser et al.,
2006; Kisser et al., 2017; Qin et al., 2021; Armitage et al., 2022; Chu et al., 2023).

2
S. Goto and N. Yanase Finance Research Letters 58 (2023) 104279

According to Phan and Hedge (2013), well-governed managers’ risk-shifting behaviors improve the average pension funding status.
This effect of good corporate governance is in stark contrast with the effect of high external financing costs, as Bartram (2017, 2018)
and Goto and Yanase (2021) argue that risk-shifting firms tend to increase pension underfunding to finance investments when they
face high external financing costs.
To reconcile the contrasting effects of corporate governance and external financing costs, we conjecture that well-governed
managers use higher rate assumptions for risk-shifting to increase R&D investments without deteriorating the pension funding sta­
tus. Such risk-shifting may increase shareholder values by mitigating underinvestment problems. However, managers may choose
overly high rate assumptions with weak investor monitoring, resulting in larger pension underfunding without increasing shareholder
values.
Hypothesis 1: Ceteris paribus, well-governed managers use risk-shifting with higher rate assumptions to finance R&D investments.
Among them, higher rate assumptions precede larger R&D expenditures.
Hypothesis 2: Ceteris paribus, poorly-governed managers choose overly high rate assumptions to lower plan contributions, pre­
ceding larger pension underfunding.
Hypothesis 3: Ceteris paribus, higher rate assumptions precede higher shareholder values for well-governed firms but not for
poorly-governed firms.

3. Empirical design

3.1. Data sources, samples, variables, and descriptive statistics

Our sample consists of TSE1 firms with fiscal years ending in March,9 excluding financial firms and firms with negative equity. We
assume a 3-month reporting lag, observing financial statement variables at the end of June in each year. Table 1 (Panel A) summarizes
our variables.
The main explanatory variables are ERR and discount rates (“Disc”). We denote their values in year t by ERR[t] and Disc[t]. The
three dependent variables are the one-year ahead R&D expenditures, pension underfunding, and shareholder values, all divided by
current book equity (BE[t]). We use two equity ownership shares to measure the quality of corporate governance. SO[t] is the “stable
ownership,” i.e., the equity ownership shares of gray institutions and stable shareholders, in year t. FO[t] is the “foreign ownership,” i.
e., the equity ownership share of foreign (non-Japanese) investors. Panel B provides descriptive statistics. Our control variables follow
Goto and Yanase’s (2021).

3.2. Econometric methodology

For each firm (indexed by i), our bhave fiscal years ending in March owing form:

yi [t + 1] = ρ ⋅ yi [t] + α ⋅ RATEi [t] + γ′ ⋅ Xi [t] + ϕi + υ[t + 1] + εi [t + 1], (1)

where yi[t + 1] is the dependent variable in the following year t + 1 and εi[t + 1] is the residual component of yi[t + 1]. RATEi[t] is one
of the rate assumptions—ERR or Disc—chosen by the firm in year t Xi[t] is a vector of firm-level control variables. φi and υ[t + 1]
denote firm- and year-fixed effects.
To test if the coefficients α and γ vary with corporate governance, we include two indicator variables, I(SOi[t]) and I(FOi[t]). I
(SOi[t])=1 when the firm is in the Top-Half subsample of SO (stable ownership) and I(SOi[t])=0 otherwise in year t. I(FOi[t])=1 when
the firm is in the Top-Half subsample of FO (foreign ownership) and I(FOi[t])=0 otherwise. We extend the base model to incorporate
possible interactions with the two indicators:
yi , [t + 1] = ρ ⋅ yi [t + 1] + (α0 + α1 ⋅ I(SOi [t]) + α2 ⋅ I(FOi [t])) ⋅ RATEi [t]
(2)
+(γ0 + γ1 ⋅ I(SOi [t]) + γ2 ⋅ I(FOi [t]))′ ⋅ Xi,t + ϕi + υ[t + 1] + εi [t + 1].

We implement the base and extended dynamic panel regression models using the system-GMM estimator.10 Using the model for
conditional expectations, rejecting the null hypothesis, α∕ =0 implies E(yi[t + 1]|yi[t],Xi[t],RATEi[t]) − E(yi[t + 1]|yi[t],Xi[t]) ∕ = 0, i.e.,
∑
RATEi[t] precedes (Granger-causes) yi[t + 1]. The long-run total multiplier effect of RATEi[t] is: ∞ h=1 (E(yi [t + h]|yi [t], Xi [t],RATEi [t]) −
E(yi [t + h]|yi [t], Xi [t])) = α
̂ × (1 + ρ + ρ2 + ⋯) × RATEi [t]

α
̂
= α=α
× RATEi,t where ̂ α 1 ⋅ I(SOi [t]) + α
̂0 + ̂ ̂ 2 ⋅ I(FOi , [t])
(1 − ρ)

9
About 80% of TSE1 firms have fiscal years ending in March.
10
Following Goto and Yanase (2021), we treat rate assumptions and control variables as predetermined variables. They are uncorrelated with
current and future residuals but can be correlated with past residuals.

3
S. Goto and N. Yanase Finance Research Letters 58 (2023) 104279

Table 1
Variable definitions and descriptive statistics.
Panel A: Variable Definitions

Rate Assumptions
ERR[t] Assumed long-term expected rate of returns on plan assets (ERR).
disk[t] Discount rate assumption for evaluating pension liabilities.
Dependent Variables
rd[t + 1]/BE[t] One-year ahead R&D expenditures (rd) per current book equity (BE).
(PBO-PA)[t + 1]/BE[t] One-year ahead pension underfunding (PBO-PA) per current BE. PBO and PA are the projected benefit obligations
(pension liabilities) and DB plan assets, respectively
ME[t + 1]/BE[t] One-year ahead market equity value (ME) per current BE.
Corporate Governance Variables (Equity ownership shares)
SO[t] Stable Ownership. The percentage of equity shares held by gray institutions and stable shareholders, including
"Keiretsu" cross-shareholders, "Main Banks" and life-insurance companies, and the management.
FO[t] Foreign Ownership. The percentage of equity shares held by foreign (non-Japanese) investors.
Control Variables
Cash[t]/BE[t] Cash and cash equivalents per BE.
Debt[t]/BE[t] Financial leverage – the debt-to-equity ratio.
E(roe[t + 1]) Management forecast of one-year ahead earnings per BE – the forward ROE estimate.
E(sales.growth[t + 1]) Management forecast of one-year ahead sales growth, (sales[t + 1]-sales[t])/Sales[t].
ME[t]/BE[t] The market-to-book ratio.
stock.return[t] Annual trailing total stock returns at the end of June.
1/BE[t] Common denominator control.
Panel B: Descriptive Statistics
Variables Summary Statistics in Pooled Sample in% Pooled Correlations w/
n. obs. mean st.dev. min p05 p50 p95 max Own lag ERR[t] disk[t]
Rate Assumptions observed at t
ERR[t] 13,952 2.35 1.35 0.00 0.50 2.00 5.00 8.00 0.90 1 0.50
Dist[t] 13,952 1.68 1.20 0.00 0.15 1.90 3.10 8.40 0.94 0.50 1
Dependent Variables observed at t + 1
rd[t + 1]/BE[t] (R&D expenditures) 13,952 3.8 5.0 0.0 0.0 2.1 13.7 33.0 0.97 0.23 0.21
(PBO-PA)[t + 1]/BE[t] (Underfunding) 13,952 9.4 13.8 − 12.6 − 2.7 5.4 34.9 117.0 0.93 0.11 0.18
ME[t + 1]/BE[t] (Market equity value) 13,952 124.6 105.6 14.8 40.3 96.2 298.4 1853 0.77 0.06 0.04
Corporate Governance Variables (Equity Ownership Shares) observed at t
SO[t] (% Stable Ownership) 13,952 36.0 15.5 0.6 11.5 34.7 62.7 84.6 0.94 − 0.09 − 0.03
FO[t] (% Foreign Ownership) 13,952 15.5 11.5 0.1 1.4 13.3 37.5 55.5 0.95 0.16 0.49

Notes: Panel A provides the notation and definition of the variables used in this study. We source data from the Astra Manager (Quick Corp.) and
Nikkei NEEDS-Cges. All variables are winsorized at the bottom and top 1 percentiles each year. The variables starting with upper-case letters
represent values at the end of the fiscal year (observed in June). Those starting with lower-case letters are measured over a year. Panel B summarizes
descriptive statistics for rate assumptions, dependent variables, and corporate governance variables using a pooled sample of 13,952 firm years for
which all variables are available. We exclude firms with negative equity value. We also exclude financial firms according to TSE’s industry classi­
fication. The sample period runs from 6/2005 to 6/2021 for t (and from 6/2006 to 6/2022 for t + 1). The last three columns report correlation
coefficients with its own annual lag, ERR[t], and Disc[t].

4. TEST results

To test Hypotheses 1–3, we use the dynamic panel regressions to examine if changes in rate assumptions (ERR and Disc) precede
subsequent changes in R&D expenditures (Models 1–4), pension underfunding (Models 5–8), and shareholder values (Models 9–12),
all per lagged book equity (Table 2). In the first row, the three dependent variables are highly persistent, indicating the importance of
controlling their lags.

4.1. R&D expenditures (Hypothesis 1)

A 1% increase in ERR[t] or Disc[t] predicts the firm’s one-year ahead R&D (rd[t + 1]/BE[t]) to increase by 0.080% [t = 2.50] or
0.116% [t = 1.96], respectively (Models 1, 3). The long-run total multiplier effect of a 1% change in ERR or Disc on R&D is 0.40%
[=0.080%/(1–0.802)] or 0.58% [=0.116%/(1–0.800)], suggesting the economic significance of the Granger-causation. Meanwhile,
the significant Granger-causal effect of rate assumptions on R&D expenditures exists only among the well-governed firms with I(FO
[t])=1 (Models 2, 4), supporting Hypothesis 1.

4.2. Pension underfunding (Hypothesis 2)

A 1% increase in Disc[t] precedes a 0.726% [t = 4.19] increase in one-year ahead pension underfunding, (PBO-PA)[t + 1]/BE[t]
(Model 7). The corresponding long-run total multiplier effect of a 1% change in Disc[t] on pension underfunding is economically
significant at 1.43% [=0.726%/(1–0.494)].
The Granger-causal effect of Disc[t] on pension underfunding is much larger, by 0.626 [t = 3.17], for the firms with I(SO[t])=1, i.e.,
those with weak investor monitoring (Model 8). While ERR[t] does not precede pension underfunding significantly in the whole

4
 S. Goto and N. Yanase
Table 2
Dynamic panel regressions.
Dependent Variable: R&D: rd[t + 1] / BE[t] Underfunding: (PA-PBO)[t + 1] / BE[t] Shareholder Values: ME[t + 1] / BE[t]
Rate Assumption: ERR[t] disk[t] ERR[t] disk[t] ERR[t] disk[t]
Model #: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Lagged Dependent .802*** .824*** .800*** .823*** .504*** .520*** .494*** .514*** .362*** .376*** .356*** .378***
Variable [26.15] [34.97] [26.20] [35.00] [17.36] [18.22] [17.20] [18.15] [7.67] [9.46] [7.79] [9.56]
Rate Assumption [t] .080** − 0.002 .116*** .016 .189 .017 .726*** .553** 3.584** 3.380* 2.017 3.770
[2.50] [− 0.06] [2.96] [.31] [1.08] [.09] [4.19] [2.52] [2.07] [1.94] [1.46] [1.53]
Rate Assumption [t] − 0.040 − 0.016 .641*** .626*** − 3.990** − 5.053***
x I(SO[t]) [− 0.90] [− 0.34] [3.26] [3.17] [− 2.46] [− 2.70]
Rate Assumption [t] .118*** .124*** − 0.039 − 0.289 1.788 − 0.882
x I(FO[t] [3.07] [2.62] [− 0.21] [− 1.40] [.67] [− 0.27]
Cash[t]/BE[t] − 0.001 .001 − 0.001 .002 .012 .049** .014 .045** .705*** .530*** .688*** .501***
Cash[t]/BE[t] x I(SO[t]) − 0.007* − 0.008** − 0.057*** − 0.048*** − 0.024 − 0.052
Cash[t]/BE[t] x I(FO[t]) .004 .006* − 0.003 − 0.004 − 0.116 − 0.044
Debt[t]/BE[t] .004*** .002** .004*** .003*** .038*** .034*** .038*** .034*** .221*** .173*** .226*** .169***
Debt[t]/BE[t] x I(SO[t]) .001 .001 − 0.000 − 0.002 .087 .096
Debt[t]/BE[t] x I(FO[t]) .001 .000 − 0.010** − 0.009* .017 .030
5

E(roe[t + 1]) − 0.014 − 0.006 − 0.014 − 0.007 − 0.055 − 0.074 − 0.052 − 0.075 .539 .585 .465 .513
E(roe[t + 1]) x I(SO[t]) − 0.000 − 0.002 .042 .053 .837 .809
E(roe[t + 1]) x I(FO[t]) − 0.028** − 0.026* − 0.031 − 0.031 − 0.026 .102
E(sales.growth[t + 1]) .013*** .013*** .013*** .013*** .026** .025 .026** .029 .318** .135 .356** .121
E(sales.g.[t + 1]) x I(SO[t]) − 0.002 − 0.002 .016 .011 .006 .024
E(sales.g.[t + 1]) x I(FO[t]) − 0.001 − 0.002 − 0.034** − 0.034** .187 .181
ME[t]/BE[t] .000 − 0.001 .000 − 0.002 .001 .002 .000 .001
ME[t]/BE[t] x I(SO[t]) .002* .002* − 0.001 − 0.001
ME[t]/BE[t] x I(FO[t]) .001* .002* .004 .005
stock.return[t] .002*** .001 .002*** .001 .002 − 0.002 .001 − 0.003 .858*** .786*** .851*** .786***
stock.ret.[t] x I(SO[t]) .000 .000 .008* .009* .134** .127**
stock.ret.[t] x I(FO[t]) .001 .001 .001 .002 .161*** .159***
1/BE[t], 1/BE[t] x I(SO[t]), 1/BE[t] x I(FO[t]) Y,N,N Y,Y,Y Y,N,N Y,Y,Y Y,N,N Y,Y,Y Y,N,N Y,Y,Y Y,N,N Y,Y,Y Y,N,N Y,Y,Y
Firm-Fixed Effects & Year Fixed Effects Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y Y, Y

Finance Research Letters 58 (2023) 104279

Notes: This table reports system GMM estimation results for dynamic panel regressions of R&D expenditures (Models 1–4), pension underfunding (Models 5–8), and shareholder (market equity) values
(Models 9–12) (all per BE[t]). Numbers in brackets are Arellano-Bond (1991) robust statistics.
***
p<0.01,.
**
p<0.05,.
*
p<0.10.
S. Goto and N. Yanase Finance Research Letters 58 (2023) 104279

sample (Model 5), it precedes pension underfunding significantly among the firms with I(SO[t])=1 (Model 6). Consistent with Hy­
pothesis 2, our results suggest that higher rate assumptions of poorly-governed firms precede larger pension underfunding. Discount
rates are more informative about future pension underfunding than ERR, corroborating Chu et al. (2023) finding.

4.3. Shareholder values (Hypothesis 3)

A 1% increase in ERR[t] precedes a 3.584% [t = 2.07] increase in ME[t + 1]/BE[t] (Model 5). This positive valuation effect,
however, is significantly weaker, by − 3.990 [t=− 2.46], among the firms with I(SO[t])=1 (Model 10), for which the coefficient on ERR
is 3.380–3.990=− 0.610. By contrast, the coefficient on ERR is much larger at 3.380+1.788=5.168 among the firms with I(FO[t])=1.
disk[t] exhibits an insignificant valuation effect (Model 11), but the effect is significantly negative for the firms with I(SO[t])=1 (Model
12). Risk-shifting with higher ERRs precedes an increase in shareholder values, but this positive valuation effect does not appear to
exist for poorly-governed firms with large stable ownership.

4.4. Subsample analyses—R&D sort and underfunding-sort

Bartram (2017) stresses that R&D-intensive firms have higher external financing costs due to higher information asymmetries and
agency costs, lower collateral, and more hard-to-value intangible assets. Balachandran et al. (2019) argue that large pension under­
funding is associated with high external financing costs because it increases the firm’s effective leverage, credit risk, asymmetric
information, and agency costs. Following their arguments, we sort firms by the R&D and pension underfunding levels and report the
results in Table 3.

Table 3
Dynamic panel regressions—Subsample analyses.
Rate Assumption: ERR[t] disk[t]
Subsamples: R&D Sort Underfunding Sort R&D Sort Underfunding
Low High Low High Low High Low High
Model #: (1) (2) (3) (4) (5) (6) (7) (8)

Panel A: Dependent Variable ¼ R&D Expenditures: rd[t þ 1] / BE[t]

Rate Assumption [t] − 0.009 − 0.018 .088** − 0.055 − 0.004 .023 .055 − 0.023
[− 0.57] [− 0.34] [2.02] [− 1.25] [− 0.21] [.31] [1.28] [− 0.32]
Rate Assumption [t] x I(SO[t]) .023 − 0.013 − 0.074 − 0.018 .021 .055 − 0.055 .007
*
[1.37] [− 0.24] [− 1.74] [− 0.32] [1.16] [.95] [− 1.29] [.12]
Rate Assumption [t] x I(FO[t]) .015 .149*** .014 .168*** .013 .122* .023 .152**
[.96] [2.67] [.38] [3.21] [.60] [1.84] [.66] [2.17]
Lagged dependent var., Control variables, Firm- & Year-
Y Y Y Y Y Y Y Y
Fixed Effects
Panel B: Dependent Variable ¼ Pension Underfunding: (PBO-PA)[t þ 1] / BE[t]
Rate Assumption [t] − 0.290 − 0.043 .104 − 0.076 − 0.125 .494** .460*** .854***
[− 1.36] [− 0.18] [1.09] [− 0.32] [− 0.58] [1.97] [3.29] [2.92]
Rate Assumption [t] x I(SO[t]) .716*** .697*** − 0.115 .976*** .893*** .612*** − 0.050 .781***
[2.86] [2.83] [− 1.30] [3.45] [3.29] [2.85] [− 0.50] [2.89]
Rate Assumption [t] x I(FO[t]) .049 − 0.118 − 0.045 .069 .005 − 0.443* − 0.103 − 0.524*
[.24] [− 0.47] [− 0.63] [.27] [.02] [− 1.72] [− 1.06] [− 1.82]
Lagged dependent var., Control variables, Firm- & Year- Y Y Y Y Y Y Y Y
Fixed Effects
Panel C: Dependent Variable ¼ Shareholder Value: ME[t þ 1] / BE[t]
Rate Assumption [t] − 3.147 3.428* − 1.764 4.117* 0.538 4.118 − 2.705 4.959
[− 1.12] [1.90] [− 0.67] [1.93] [0.14] [1.60] [− 0.92] [1.38]
Rate Assumption [t] x I(SO[t]) − 1.424 − 4.668** − 2.059 − 5.249** − 6.183 − 6.484*** − 4.261 − 4.956**
* *
[− 0.53] [− 2.56] [− 1.11] [− 2.51] [− 1.84] [− 3.09] [− 1.80] [− 2.00]
Rate Assumption [t] x I(FO[t]) − 0.014 1.604 6.738* 2.765 − 3.778 − 1.149 6.103 0.648
[− 0.00] [0.75] [1.69] [0.88] [− 0.61] [− 0.44] [1.57] [0.16]
Lagged dependent var., , Control variables, Firm- & Y Y Y Y Y Y Y Y
Year-Fixed Effects

Notes: This table reports system GMM estimation results for dynamic panel regressions of R&D expenditures (Panel A), pension underfunding (Panel
B), and shareholder (market equity) values (Panel C). Both panels report only the coefficients on the rate assumption (ERR or Disc) and its interactions
with I(SO[t]) and I(FO[t]). In each panel, Models 1–4 use ERR. Models 5–8 use Disc for the rate assumption. All dynamic panel regressions include
lagged dependent variables, all control variables, and fixed effects reported in Table 2, though this table does not show their coefficients. Models 1–2
and 5–6 use subsamples sorted by R&D expenditures, R&D[t]/BE[t] (below and above median in each year). Models 3–4 and 7–8 use subsamples
sorted by the degree of pension underfunding, (PBO-PA)[t]/BE[t] (below and above median in each year). Numbers in square brackets are Arellano-
Bond (1991) robust statistics.
***
p<0.01,.
**
p<0.05,.
*
p<0.10.

6
S. Goto and N. Yanase Finance Research Letters 58 (2023) 104279

Higher rate assumptions of well-governed firms (I(FO[t])=1) precede larger R&D (Panel A), especially among those with large R&D
(Models 2, 6) or large underfunding (Models 4, 8) than others. Higher rate assumptions also precede larger pension underfunding
among poorly-governed firms (I(SO[t])=1) (Panel B). This effect is present only among firms with large underfunding (Models 4, 8).
Higher ERRs precede higher shareholder values for firms with high external financing costs. However, this positive valuation effect
disappears or even gets reversed for poorly-governed firms (I(SO[t])=1) (Panel C, Models 2,4).
In sum, external financing costs and investor monitoring affect firms’ risk-shifting behaviors differently. Both effects are essential to
understand the implications of managerial risk-shifting.

5. Conclusion

This paper shows how the quality of corporate governance affects managerial risk-shifting behaviors and their consequences on
future pension funding status, R&D investments, and shareholder values. This study contributes to the literature by demonstrating the
significant influence of investor monitoring on managerial risk-shifting. Prior studies show that risk-shifting behaviors are associated
with larger pension underfunding and R&D expenditures, especially among firms with high external financing costs. However, this
paper demonstrates that well-governed and poorly-governed managers engage in risk-shifting behaviors differently. Well-governed
managers use risk-shifting to increase R&D expenditures, whereas poorly-governed managers’ risk-shifting results in increased
pension underfunding. While managerial risk-shifting improves shareholder values by addressing underinvestment problems without
aggravating pension underfunding, risk-shifting by poorly monitored firms results in larger pension underfunding without increasing
shareholder values.

Data availability

Data will be made available on request.

Financal support

This work was partially supported by the JSPS KAKENHI [grant number 21H04394, 22K01579]; JSDA Capital Markets Forum. All
remaining errors are the authors’ own.

References

Aggarwal, R., Erel, I., Ferreira, M., Matos, P., 2011. Does governance travel around the world? evidence from institutional investors. J. Financ. Econ., 100 (1),
154–181.
Arellano, M., Bover, O., 1995. Another look at instrumental variables estimation of error-component models. J Econom 68, 29–51.
Armitage, S., Gallagher, R., Xu, J., 2022. The elusive relation between pension discount rates and deficits. J Bus Finance Account 1–27.
Balachandran, B., Duong, H.N., Vu, V., 2019. Pension deficit and the design of private debt contracts. J. Financial Quant. Anal. 54, 1821–1854.
Bartram, S.M., 2017. Corporate postretirement benefit plans and real investment. Manage Sci 63 (2), 355–383.
Bartram, S.M., 2018. In good times and in bad: defined-benefit pensions and corporate financial policy. J. Corp. Finance 48, 331–351.
Bergstresser, D., Desai, M., Rauh, J., 2006. Earnings manipulation, pension assumptions, and managerial investment decisions. Q. J. Econom. 121 (1), 157–198.
Brundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel-data models. J Econom 87, 115–143.
Chaudhry, N., Au-Yong, H.W., Veld, C., 2017. How does the funding status of defined benefit pension plans affect investment decisions of firms in the United States?
J Bus Finance Account 44, 196–235.
Chen, X., Harford, J., Li, K., 2007. Monitoring: which Institutions Matter? J Financ Econ 86 (2), 279–305.
Chu, L.M.A.Goldstein, X. Li, and T. Yu (2023) "Do firms set pension discount rates strategically," SSRN Working Paper (3627965).
David, P., Yoshikawa, T., Chari, M.D., Rasheed, A.A., 2006. Strategic investments in japanese corporations: do foreign portfolio owners foster underinvestment or
appropriate investment? Strateg. Manag. J. 27 (6), 591–600.
Duong, H.N., Qiu, B., B, Rhee, S.G., 2022. Golden handcuffs and corporate innovation: evidence from defined benefit pension plans. Rev. Corp. Finance Stud. 11,
128–168.
Ferreira, M.A., Matos, P., 2008. The colors of investors’ money: the role of institutional investors around the world. J Financ Econ 88 (3), 499–533.
Goto, S., Yanase, N., 2021. Pension return assumptions and shareholder-employee risk-shifting. J. Corp. Finance 70.
Heusel, N., Mager, F., 2023. Pension funding and the cross section of stock returns - the case of Germany. J. Bank. Financ. forthcoming.
Johnston, S., McAlevey, L., 1998. Stable shareholdings and Japan’s bubble economy: an historical overview. Strateg. Manag. J. 19 (11), 1101–1107.
Kisser, M., Kiff, J., Soto, M., 2017. Do managers of U.S. defined benefit pension plan sponsors use regulatory freedom strategically? J Account Res 55 (5), 1213–1255.
Phan, H.V., Hedge, S.P., 2013. Corporate governance and risk taking in pension plans: evidence from defined benefit asset allocations. J. Financial Quant. Anal. 48 (3),
919–946.
Qin, Y., Cai, J., Rhee, S.G., 2021. Do Japanese firms systematically inflate expected rate of returns from defined benefit pension plans? Pacific-Basin Finance J. 68,
101321.
Quinn, D., Rivoli, P., 1991. The effects of American-and Japanese-style employment and compensation practices on innovation. Organ. Sci. 2, 323–341.
Rauh, J.D., 2009. Risk shifting versus risk management: investment policy in corporate pension plans. Rev. Financ. Stud. 22 (7), 2687–2733.
Sasaki, T., 2015. The effects of liquidity shocks on corporate investments and cash holdings: evidence from actuarial pension gains/losses. Financ Manage 44 (3),
685–707.
Wintoki, M.B., Linck, J.S., Netter, J.M., 2012. Endogeneity and dynamics of internal corporate governance. J Financ Econ 105 (3), 581–606.