The Effect of Corporate Social Responsibility on Hard-Freezing of Pension Plan and Firm Performance

Abstract

In recent years, there has been a dramatic increase in the hard-freezing of defined benefit pension plans. Although cost savings associated with a pension freeze are expected to lead to improved future performance, prior studies do not provide conclusive empirical evidence to support enhanced firm performances following pension freezes. In this study, we examine how firms’ commitment to corporate social responsibility (CSR), especially their commitment to employee relations, affects their decision to freeze pension plans, and how firms’ CSR activities affect the association between the pension freeze and their post-freeze performance. First, we find no evidence that firms with a high overall CSR score are less likely to freeze their pension plans, but we find supporting evidence that firms with a high CSR score on the employee relation aspect are less likely to freeze their pension plans. In addition, we find that the pension freeze has a positive impact on the firm’s future performance, as measured by the return on assets, for the firms with high CSR scores.

Keywords

pension freeze corporate social responsibility employee relations pension cost firm performance pension accounting

Introduction

This study examines how firms’ commitment to corporate social responsibility (CSR) affects their decision to hard-freeze defined benefit pension plans (DB plans) and how CSR activities affect firm performance during the post-freeze periods. The number of firms that hard-freeze their DB plans increased dramatically in recent years. According to the Pension Benefit Guaranty Corporation (PBGC, 2013), while only 9.4% of the PBGC-insured single employer DB plans were hard-frozen in the year 2003, that number increased to 30.7% by the end of the year 2012. Among many reasons for DB plan freezing decisions, the literature on pensions reveals that one of the primary motivations is to reduce pension costs (Comprix & Muller, 2011; Munnell et al., 2006).¹ For example, after their decision to freeze their DB plan in 2016, American International Group explained, “AIG’s spending on employee retirement programs is materially higher than most of our peers . . . .”²

Although cost savings associated with pension freezes are expected to lead to better future financial performances, prior studies do not provide conclusive empirical evidence of superior performance for firms that restructure their DB pension plans. While it is possible that the cost savings through a pension freeze may improve a firm’s future performance, it is unclear whether this reduced employee benefit negatively affects employee morale and employee productivity, and the firm’s performance in the end.

To examine the effect of pension freezes on a firm’s future performance, we bring the firm’s level of CSR activities to the forefront for consideration. Pension freezes basically shift the risk associated with future pension payout from the employer to employees, which decreases employees’ anticipated future retirement income (Gelter, 2013; Rauh et al., 2019), thereby reducing employee productivity and affecting firms’ future performance in the end.

The literature on CSR shows that CSR activities build social capital and enhance stakeholder trust, which will have a positive impact on firm valuation (Lins et al., 2017; Putnam, 1993). Benabou and Tirole (2010) argue that CSR activities reduce managers’ behaviors that focus on only short-term-oriented goals. For example, socially responsible firms are less likely to manage earnings (Y. Kim et al., 2012) and less likely to engage in insider trading (Gao et al., 2014).

The literature on CSR based on stakeholder theory also documents that higher CSR activities can lead to better financial performance (Atkinson et al., 1997; Freeman, 1984; Lins et al., 2017; Waddock & Graves, 1997). In particular, Waddock and Graves (1997) document that employee relations policy can affect employee morale and productivity, and eventually affect firms’ competitive advantage. Considering the opposing effects of pension freeze and firms’ socially responsible activities on a firm’s future performance, we expect that the negative impact of a pension freeze on future performance will be mitigated by the trust and cooperation through the firm’s active engagement in socially responsible activities.

As we hypothesize that a firm’s CSR activities could mitigate the adverse effect of pension freeze on firm performance, we also test whether the CSR activity itself can affect a firm’s decision to freeze its pension plans. A priori, the association between a firm’s CSR activities and pension-freezing decisions is not clear. On one hand, it is possible that the firm that already has a high commitment to CSR activities is more likely to freeze its DB pension plans because the high commitment to CSR can mitigate the negative impact of pension freeze on future firm performances. However, it is also possible that firms with a high commitment to employee relations (one component of CSR measure) are less likely to freeze their DB plans because these are the group of firms that pay more attention to their relations with employees. When firms believe the cost savings are not large enough to offset the negative impact on firm performance, firms with a high commitment to CSR activities are less likely to freeze their DB pension plan. Following prior studies that use the Kinder, Lydenberg, and Domini (KLD) database, we construct overall CSR score (CSRCOM) and CSR score concerning employee relations (CSREMP) to measure firms’ CSR activities for the 9-year spanning from 2001 to 2009. Among those firms in the KLD database, we identify 222 unique pension hard-freezing firms in the COMPUSTAT database with the DB pension plans during the sample period.³

Empirically, we do not find any evidence that a firm’s overall CSR activities are associated with a future pension freeze. However, when we focus on a firm’s commitment to employee relations in the CSR index, we find that employee relation scores in the CSR index are negatively associated with future pension freezes. These results imply that firms which show a strong commitment to employee relations are less likely to freeze their pension plans. In an additional analysis, we find no evidence of reverse causality.

We also investigate the impact of a pension freeze on the accounting-based firm performance and how CSR activities affect the association between the pension freeze and firm performance. As a baseline analysis, we first examine whether firms improve financial performance during the post-period of the pension freeze. We find no evidence that the pension freeze helps firms improve their financial performance, on average. However, we find that firms with a higher pension sensitivity to earnings have improved financial performance after the pension freeze, which implies that a pension freeze can facilitate firms to report higher earnings at least for those whose service cost constitutes a significant portion of their total expense.

For our main analysis, we examine how a firm’s CSR activities influence the effect of the pension freeze on firm performance. We find that, on average, pension-freezing firms experience a lower future return on assets than no-freezing firms in the low CSR subsample. However, pension-freezing firms’ future financial performance is much higher than its control group in the high CSR subsample. These results indicate that the negative effect of a pension freeze is mitigated by higher CSR activities, especially for the firms with a higher score in the area of commitment to employee relations. We also find the effect of pension freeze on firm performance lasts at least the next 2 years. Since underlying economic conditions may affect both the pension-freeze decision and the firm performance (i.e., endogeneity concerns), we employ a two-stage least squares regression procedure (2SLS) and the propensity-score-matched control sample approach. We find similar results to those from the ordinary least squares (OLS) regression.

As an additional analysis, we investigate whether labor unionization affects our main findings. We find that the moderating effect of CSR on the relation between the pension freeze and the future firm performance is pronounced in the more unionized industries, but the association is insignificant in the less unionized industries.

This article contributes to the literature on pensions and CSR in several ways. First, the article expands our understanding of the effects of pension freezes on future performance. Prior studies have focused on the market response to a freeze announcement in analyzing the effect of pension freeze on firm value; however, the empirical evidence is inconclusive (McFarland et al., 2009; Milevsky & Song, 2008; Rubin, 2007; Vafeas & Vlittis, 2018). Rubin (2007) attributes these inconclusive findings to lurk variables and confounding effects that are not controlled for.⁴ In addition, McFarland et al. (2009) and Vafeas and Vlittis (2018) document that CSR activities, especially employee relations, could affect a firm’s decision to freeze pension plans. While prior studies have not incorporated the negative effect of pension freeze on employee morale and consequently its effect on firm productivity and firm performance in their research design, we try to address this issue by examining whether CSR activities can influence the effect of pension freeze on firm performance.

Vafeas and Vlittis (2018) also find that a pension freeze increases a firm value when the firm has a higher percentage of independent directors. In the context of shareholders, employee, and management relations, Vafeas and Vlittis (2018) focus on shareholders–management relations and show that the agency cost of free cash flow and fund misuse after freezes can be constrained by independent directors. Our study complements Vafeas and Vlittis (2018) by focusing on employee–management relations measured by CSR activities.

Finally, our study also contributes to the literature on pension freeze by providing an important determinant of pension freeze. Prior studies have focused on financial pressure from firm and pension plan characteristics, such as plan sponsors’ financial distress and funding status of the pension plan as the determinants of pension freeze (Choy et al., 2014; Comprix & Muller, 2011). We examine persisting employee relations in the process of pension-freeze decision because firms have provided the pension plan as a way of organizational commitment to the employee (Muse et al., 2005).

In Section “Background on the Pension Freeze,” we review the background information on pension plan freezes. In Section “Literature Review and Hypotheses Development,” we review prior studies and develop hypotheses.

Section “Method” discusses the research design. Section “Data Description” presents the sample selection and descriptive statistics. We discuss empirical findings in Sections “Empirical Results” and “Additional Analysis.” We conclude with remarks in Section “Conclusion.”

Background on the Pension Freeze

Pension plans in the United States can be largely classified into two basic types: defined benefit (DB) and defined contribution (DC) plans. One of the key differences between the two types of pension plans lies in the bearing of the investment risk. Under a typical DB plan, employees are guaranteed to receive a specific level of benefits upon their retirement. Since the specific amount of benefits is guaranteed, employers bear the risk associated with future cash payouts for the plan. On the contrary, under the DC plan, employees accumulate a specific amount in individual accounts from employers’ contributions, where retirement incomes are determined by the employees’ investment return. Therefore, employers bear no investment risk (Choy et al., 2014).

When employers adopt DB plans, they are obligated to pay pension benefits for employees’ service earned. However, upon freezing DB plans, firms stop accruing benefits attached to future service, which will reduce growth in benefits and employers’ future pension obligations. The cost saving is achieved because only retired and vested participants receive full-expected benefits while active participants will receive only what they had earned. Rauh et al. (2019) demonstrate that annual DB accruals, as a percentage of salary, are larger for older workers than for younger workers. Longer service workers are likely to be compensated more than their marginal product under a DB plan. Therefore, to save on pension costs, firms have incentives to freeze their DB plans. Since the early 2000s, there has been an increase in DB plan freezes among Fortune 1000 firms in their efforts to reduce operating costs, to avoid the volatility associated with investment returns, and ultimately to remain competitive in the global marketplace (Munnell et al., 2006). Firms can save about 2.7%–3.6% of their payroll per year when they replace DB plans with DC plans. The average cost saving is known to be 3.1% of a firm’s current assets over 10 years (Rauh et al., 2019).

There are two types of DB pension freezes: the hard and soft freezes. When firms hard-freeze their DB plans, the plan stops accruing future benefits to all plan participants, which means any additional future service provided by employees will not increase firms’ reported pension liabilities.⁵ On the contrary, when firms soft-freeze their DB plans, the plan is closed to certain classes of employees, such as new entrants (Choy et al., 2014). Depending on collective bargaining agreements with unions, DB plan freezes can cover all pension funds (total freezing) or some of pension funds in the firms (partial freezing). In this study, we focus on the effect of hard-freeze cases only because reliable information on the soft freeze of DB plans is not available. As most soft freezes of DB plans tend to eventually become hard-freeze plans, the analyses of the hard freeze on DB plans can help us understand the overall effect of DB plan freezes.

We also would like to note that the DB pension plan freeze should be differentiated from the DB plan termination. “DB plan termination” means that the plan-sponsoring firm closes the DB plan by immediately paying out all benefits to plan participants, either as a lump sum or by purchasing annuities from the insurance company for covered employees. According to the Watson Wyatt (2010) report, half of the firms that terminate DB plans were dropped off from the Fortune 1000 list in the following year, implying that termination is often caused by financial distress (Choy et al., 2014). We do not include DB plan termination in our analysis as we investigate the association between managers’ strategic choices and their impact on financial performance.

Literature Review and Hypotheses Development

Two opposing forces will affect firms’ future performances when they hard-freeze their DB pension plans. On one hand, there is a cost-saving aspect associated with the pension freeze. The pension-sponsoring firms tend to freeze their DB plans to improve pension cost structure and to lower financial risk. By freezing pension plans, a firm can actually reduce its pension expenses, as employees’ additional service does not increase its pension liability any longer. For example, Munnell et al. (2006) conduct a case study on pension-freezing and come to the conclusion that a firm’s decision to freeze the pension plan is primarily because of the following four reasons: to cut compensation in the face of increased market competition; to restructure compensation in response to increased health care costs; to avoid an increase in investment risk; and to reduce the proportion of pension benefits in upper managers’ compensation. Comprix and Muller (2011) argue that as DB plans become too costly for employers to maintain, firms’ incentive to reduce the pension obligation increases. Rauh et al. (2019) document that the source of cost savings through the pension freeze comes from reneging on implicit contracts to underpay DB plan participants when they are young and overpay them when they are old. Comprix and Muller (2006) also document that the pension cost amounts to, on average, 16% of a firm’s reported income before extraordinary items, and they argue that managers may have incentives to reduce pension costs to create improved future reported income by freezing their pension plans.

While the cost savings from a pension freeze will have a positive impact on a firm’s future performance, it may have an impact on employee retention and employee productivity. Because employee pension benefits are a form of deferred compensation, a firm’s decision to freeze the DB plan can result in a decrease in employees’ total compensation. Munnell and Soto (2004) estimate that a firm’s contribution to its pension plan can decrease as much as 7% when a DB plan is replaced by a DC plan, indicating that the total employee compensation can decrease at least in the short run (Munnell et al., 2006). This reduced compensation following a pension freeze may affect employee retention and employee productivity, which in turn may affect the firm’s future performance. Muse et al. (2005) document that offering a pension plan has a strong impact on firm performance through increased employee productivity, implying that a pension freeze can negatively affect employee productivity.

Because of these two opposing effects of a pension freeze, it is unclear whether a firm’s pension-freezing decision will have a positive or negative impact on the firm’s future performance, which may be one of the reasons why prior studies do not find conclusive evidence. For example, Rubin (2007) finds that the investors do not respond to the announcement of a pension freeze in a short window, but positively reacts to the news in a long window study (e.g., next 50 days), which he argues as evidence of market inefficiency. McFarland et al. (2009) find negative abnormal returns around pension-freeze announcements as the stock market interprets the event as a sign of financial distress. On the contrary, Vafeas and Vlittis (2018) find that the market positively responds to pension freezes, and the market reaction is stronger as the percentage of outside directors on the board increases. The PBGC (2008) documents that the DB plan freeze has little impact on firms’ financials in the short-term because firms with frozen plans are still required to pay the PBGC premium and have to make minimum required contributions.

A number of prior studies examine the impact of a firm’s CSR activities on firm value. While many prior studies report a positive relationship between CSR and firm value, they provide different reasons for such relations. For example, Waddock and Graves (1997) suggest that positive relationships with stakeholders can reduce the likelihood of difficult situations that may arise in dealing with stakeholders such as employees, customers, and the community. Hillman and Keim (2001) provide empirical evidence that investing in stakeholder management by building better relations with employees, customers, and suppliers can improve shareholder value. Wans (2020) also argues that high CSR firms have established strong ties with various stakeholder groups by considering and meeting their demands. Lins et al. (2017) find that high CSR firms have higher profitability and gross margins, and experience higher sales growth than low-CSR firms during financial crises because of trust and cooperation that high CSR firms have built with stakeholders.

Since firms’ CSR activities are known to affect their future performance (e.g., Lins et al., 2017), we argue that CSR activities can affect the firm’s pension-freezing decisions and the association between firms’ decision to freeze their pension plans and their future performance. On one hand, it is possible that firms which already have a high commitment to CSR activities, especially to employee relations, are more likely to freeze their DB pension plan. This is because they may believe that their high commitment to CSR can mitigate the negative impact of pension freeze, and therefore, the pension freeze decision may not affect firms’ future performance as significantly especially when the cost savings from the pension freeze are considered.

On the other hand, it is also possible that firms with a high commitment to employee relations (one component of CSR measure) are less likely to freeze their DB plans because these are the group of firms that focus more on their relations with their employees. When firms believe cost savings are not large enough to offset the negative impact on firm performance, firms with a high commitment to CSR activities are less likely to freeze their DB pension plans. Hence, we believe the association between a firm’s decision to freeze DB plans and a firm’s commitment to CSR activities, including employee relations, is an empirical question. We first propose a non-directional null hypothesis as follows:

Hypothesis 1: Ceteris paribus, firms that have engaged in a high level of CSR activities and a low level of CSR activities are equally likely to hard-freeze their DB pension plans.

The literature on strategy and management documents that one of the key sources of sustained competitive advantage comes from human resources—for example, employee commitment, culture, and teamwork (Barney & Wright, 1998; Becker & Gerhart, 1996). Particularly, the theory of the service profit chain suggests that profit and growth are the results of satisfied customers, who are retained by providing high-quality goods and services, which in turn is provided by satisfied, loyal, and productive employees (Heskett et al., 2008). Muse et al. (2005) provide empirical evidence that the organizational commitment to employees by providing a pension plan has a significant positive association with performance.

The studies on CSR also find that high CSR firms tend to have a better financial performance, which is often referred to as the virtuous circle—“doing good” socially leads to “doing well” (Choi & Wang, 2009; Flammer, 2015; Servaes & Tamayo, 2013; Waddock & Graves, 1997). These studies document that good relations with stakeholders, particularly with employees, help firms gain a performance advantage. For example, good relations with employees can create stronger loyalty and commitment and reduce employees’ turnover. If a positive impact of pension-freezing through cost savings can be offset by the deteriorated relations with stakeholders, we hypothesize that the net benefits of the pension freeze depend on the firms’ relations with their stakeholders including employees. That is, high CSR activities may mitigate the negative effect of pension-freezing on a firm’s future performance (e.g., reduced employee productivity), thus enabling firms to achieve improved future performance. Based on the discussion above, we state our second hypotheses as follows:

Hypothesis 2: Ceteris paribus, firms that hard-freeze DB pension plans will have a better future firm performance when these firms engage in a high level of CSR activities than a low level of CSR activities.

Method

To investigate whether firms’ levels of CSR activities affect their decisions to freeze the DB plan, we design the binary-coded probit regression model as follows:

\begin{matrix} P (FREEZ E_{i, t + 1} = 1) = α_{0} + α_{1} CS R_{i, t} + α_{2} FUNDIN G_{i, t} \\ + α_{3} UNDERFUN D_{i, t} + α_{4} PLANSIZ E_{i, t} + α_{5} FIRMSIZ E_{i, t} \\ + α_{6} CF O_{i, t} + α_{7} DLOS S_{i, t} + α_{8} LE V_{i, t} + α_{9} SALESG R_{i, t} \\ + α_{10} DCAPE X_{i, t} + α_{11} DRN D_{i, t} + α_{12} DDI V_{i, t} + α_{13} UNIO N_{i, t} \\ + α_{14 - 24} INDUSR Y_{i, t} + α_{25 - 32} YEA R_{i, t} . \end{matrix}

(1)

We employ the probit regression model used in Comprix and Muller (2011) and Choy et al. (2014), who investigate the determinants of firms’ decisions to freeze DB pension plans. The dependent variable (FREEZE_i,t₊₁) is the firm’s choice of future pension-freezing, which takes a value of 1 if a firm, i, freezes the DB plan in the future (t + 1); 0 otherwise. Our variable of interest in this model is CSR_i,t. CSR_i,t is measured in two different ways. First, CSRCOM_i,t is the sum of all indicators of strengths and concerns for seven categories in the KLD database scaled by the number of strengths and concerns.⁶ Second, CSREMP_i,t is the sum of the indicator variables of the strengths and concerns for the employee relations category scaled by the number of strengths and concerns. The coefficients on both measures of CSR_i,t measure the incremental probability of pension-freezing with the variation in CSR_i,t.⁷

We control for the funding status of the pension plan with FUNDING_i,t (pension plan assets divided by the projected benefit obligation) and UNDERFUND_i,t (if the pension plan is underfunded, it takes a value of 1, 0 otherwise). We expect FREEZE_i,t has a negative relation with FUNDING_i,t and a positive relation with UNDERFUND_i,t. Comprix and Muller (2011) document that firms with poor funding status face lower employee resistance to pension-freezing. PLANSIZE_i,t (the natural logarithm of the fair value of plan assets) and FIRMSIZE_i,t (the natural logarithm of total assets) are included to control for pension plan size and firm characteristics. To control for business and financial risks of sponsor firms, we include several proxies. First, we include CFO_i,t (the cash flow from operations before pension contributions divided by total assets) and DLOSS_i,t (takes a value of 1 if the sum of earnings before extraordinary items in years t and t–1 is less than 1, 0 otherwise). We also include LEV_i,t (the sum of long-term debt and debt in current liabilities divided by total assets) and SALESGR_i,t (the sales growth from year t–1 to t divided by sales at t–1). Following Choy et al. (2014), we include DCAPEX_i,t (the change in capital expenditure from year t–1 to t divided by total assets), DRND_i,t (the change in research and development expense from year t–1 to t divided by the total assets), and DDIV_i,t (the change in dividend payout from year t–1 to t divided by the total assets) to control for managerial incentives of taking a risk. We also include a dichotomous variable, UNION_i,t, which takes a value of 1 if a firm belongs to a more unionized industry, 0 otherwise. We classify firms as being in a more unionized industry when the percentage of employees in the unionized workforce is above the median of all industries. INDUSTRY_i,t comprises 11 indicator variables based on the industry classification used in Fama and French (1997).⁸ Finally, YEAR_i,t is a year indicator variable.

To test the impact of a firm’s CSR activities on the link between the pension freeze and the firm performance (Hypothesis 2), we use OLS regressions to estimate an accounting-based performance measurement, ROA_i,t₊₁, with pension freeze and CSR indicators, and other control variables as follows:

\begin{matrix} RO A_{i, t + 1} = α_{0} + α_{1} FREEZ E_{i, t} + α_{2} HiCS R_{i, t} + α_{3} FREEZ E_{i, t} \times HiCS R_{i, t} \\ + α_{4} RO A_{i, t} + α_{5} PLANSIZ E_{i, t} + α_{6} FIRMSIZ E_{i, t} \\ + α_{7} SALESG R_{i, t} + α_{8} LE V_{i, t} + α_{9} M B_{i, t} + α_{10} RO A_{i, t} \times HiCS R_{i, t} \\ + α_{11} PLANSIZ E_{i, t} \times HiCS R_{i, t} + α_{12} FIRMSIZ E_{i, t} \times HiCS R_{i, t} \\ + α_{13} SALESG R_{i, t} \times HiCS R_{i, t} + α_{14} LE V_{i, t} \times HiCS R_{i, t} \\ + α_{15} M B_{i, t} \times HiCS R_{i, t} + α_{16 - 26} INDUSTR Y_{i, t} \\ + α_{27 - 34} YEA R_{i, t} + ε . \end{matrix}

(2)

In Equation 2, we use 1-year ahead of return on assets (ROA_i,t₊₁, the sum of income before extraordinary items and pension expense divided by total assets) as a proxy for firm performance. Since we consider the effect of the pension freeze on ROA_i,t₊₁ through both pension expense and employee productivity, we use future (1-year ahead) response as a dependent variable.

The coefficient for an indicator variable, FREEZE_i,t measures the incremental value of ROA_i,t₊₁ when a firm freezes its pension plans. The effect of CSR on firm performance is measured by binary constructed variable, HiCSR_i,t, which takes a value of 1 if a firm’s CSR score is above the median CSR score, 0 otherwise. To test Hypothesis 2, we focus on the interaction effect between pension freeze and CSR indicators in Equation 2. We expect that the coefficient on FREEZE_i,t×HiCSR_i,t to be positive. As CSR score is measured in two different ways, HiCSR is also defined by CSRCOM and CSREMP. We control for current period ROA (ROA_i,t) because accounting-based performance measures (i.e., earnings) tend to persist into future periods (Frankel & Litov, 2009). We include PLANSIZE_i,t, FIRMSIZE_i,t, SALESGR_i,t, and LEV_i,t as defined before, and BM_i,t (the book value of equity divided by the market value of equity) to control for business and finance risks.

As a firm’s decision to freeze its pension plan can be endogenous, the failure to control for firm characteristics that affect both firm’s decision to freeze the pension plan and future firm performance may bias our inferences. To address this endogeneity concern, we adopt (a) the two-stage least squares regression (2SLS) procedure and (b) the propensity-score-matched control sample approach as alternative methods. In the first-stage probit regression, we include all the control variables used in Equation 1 to model the determinants of a firm’s decision to freeze the pension plan. In the second stage, we regress a firm’s future performance (ROA_i,t₊₁) on the fitted pension-freeze probability from the first stage with control variables. In the propensity-score-matched control sample approach, we find a no-pension-freezing sample that has similar characteristics to those of pension-freezing firms during our sample period. Propensity scores are computed based on firm characteristics of pension plan size (PLANSIZE_i,t), sales growth rate (SALESGR_i,t), firm size (FIRMSIZE_i,t), leverage (LEV_i,t), and book value to market value of equity (BM_i,t).

Data Description

Sample Selection

Our sample selection process begins with measuring the CSR score, using the KLD database, which has been most widely used in academic literature (Chatterji et al., 2009).⁹ KLD has expanded its coverage by including Russell 1000 Index firms starting from 2001 and Russell 2000 Index firms from 2003. The successor of KLD, MSCI, has significantly changed the rating methodology since 2010. To maintain consistency in coverage and measurement, we use the sample period spanning fiscal years 2001–2009.

During the sample period, we obtain KLD’s evaluations regarding firms’ CSR activities in seven categories: community, diversity, employee relations, environment, human rights, product, and corporate governance (MSCI, 2015). Since CSR encompasses all actions that firms take to meet their legal, ethical, and philanthropic responsibility (Jayachandran et al., 2013), we include all available categories from the KLD data set. Each category is binary-coded: a positive and negative performance indicator, which captures the strengths and concerns of CSR activities. Following prior studies (Di Giuli & Kostovetsky, 2014; Servaes & Tamayo, 2013), we construct an overall CSR score using the sum of seven category scores, where each category score is computed by the sum of positive indicators (strengths) minus the sum of negative indicators (concerns) scaled by the total number of indicators (CSRCOM_i,t) for firm i and time t. By scaling the number of indicators, we are able to keep the same weight across the seven categories. We similarly measure the CSR score concerning only employee relations by deducting the sum of negative indicators from the sum of positive indicators (CSREMP_i,t).

We identify firm-year observations that have hard-frozen their DB plans. When firms hard-freeze their pension plans, service costs of pension plans are assigned the value of 0 because the hard freeze eliminates the accrual of additional benefits for all employees (Comprix & Muller, 2011). However, while many firms hard-freeze for the majority of participants (e.g., Brunswick Corp, EMC Corp.), there are several firms that still hold non-zero service costs even after the hard freeze because the salary increase portion continues to be considered for benefit obligation related to prior service (e.g., Bristol Myers Squibb Co. and Coca Cola Co.). Therefore, we include firms in our hard-freeze pension plan sample when firms report a significant decrease in service cost.

To identify firms that decrease their service cost significantly, we, first, identify firms that contain “hard-freezing” related words in their 10-K filings, such as “froze future benefit accruals,”“eliminate future benefits,” and “ceasing future benefit accruals,” during our sample period from 2001 to 2009 in S&P 500 firms (276 unique firms with DB plan). We identified 38 cases of hard-freezing firms. These firms have an average of 52.6% decrease in service cost relative to their prior period service cost. Using this 52.6% as a benchmark, we expand our sample to all COMPUSTAT firms. If any firm’s service cost component of their pension plan decreases more than the benchmark, the decrease is considered a significant decrease, and the firm is included in our hard-freeze pension plan sample.

Our sample starts with the intersection between KLD data and COMPUSTAT database with DB plans, and this process leaves us an initial sample of 22,963 firm-year observations with 4,852 unique firms. When we drop firms that are missing control variables, we have 7,206 firm-year observations (1,254 unique firms) with 1,611 observations of freezing firm-years (222 unique firms). Table 1 summarizes the sample selection process.

Table 1.

Sample Selections.

	Observations	Unique firms
KLD firm-year observations, 2001–2009	22,963	4,852
Less: missing data for constructing control variables	15,753	3,655
Total available observations	7,206	1,254
Total observations of freezing firms	1,611	222

Note. The table describes the sample selection process. KLD = Kinder, Lydenberg, and Domini.

Descriptive Statistics

Panel A of Table 2 reports the distribution of the number of hard-freezing and CSR scores by year. The number of firms that hard-freeze their pension plans has increased over the years. For example, only 14 firms hard-froze their pension plans in 2003 (1.7%), but the number has increased to 46 firms (or 5.1%) in 2007 and continued to stay at a similar level during the later sample period. A sharp increase in hard-freezing may be attributable to the new reporting standard, FAS 158.¹⁰ Munnell et al. (2006) document that the adoption of FAS 158 is one of the accounting standard risks that drive firms to hard-freeze their plans. During the sample period, the mean (median) CSRCOM score is −0.25 (−0.20) and the mean (median) CSREMP score is −0.07 (0.00) indicating that more CSR activities are evaluated as concerns rather than strengths.

Table 2.

Data Descriptions.

	Number of observations	Number of hard freezes	CSRCOM score		CSREMP score
			M	Mdn	M	Mdn
Panel A: pension-freezing and CSR score by year
2001	416	2	−0.153	−0.033	0.032	0.000
2002	456	8	−0.181	−0.125	0.023	0.531
2003	833	14	−0.168	0.000	−0.050	0.193
2004	851	22	−0.238	−0.200	−0.068	0.000
2005	889	11	−0.259	−0.200	−0.086	0.534
2006	889	36	−0.288	−0.200	−0.089	0.198
2007	934	46	−0.287	−0.200	−0.087	0.000
2008	972	35	−0.275	−0.200	−0.080	0.543
2009	966	48	−0.280	−0.200	−0.033	0.198
Total	7,206	222	−0.247	−0.200	−0.066	0.000
Industry classification	Number of freezing (%)		CSRCOM		CSREMP
			M	Mdn	M	Mdn
Panel B: pension-freezing and CSR score by industry
More unionized industry
1. Food, tobacco, textiles, apparel, toys	26 (4.8)		−0.28	−0.28	−0.10	−0.10
2. Cars, TV’s, furniture, household appliances	8 (3.3)		−0.19	−0.19	−0.06	−0.06
3. Machinery, trucks, planes, off furn, paper	31 (2.4)		−0.23	−0.23	−0.08	−0.08
4. Chemicals and allied products	5 (1.2)		−0.32	−0.32	−0.05	−0.05
5. Telcom telephone and TV transmission	11 (5.1)		−0.29	−0.29	−0.14	−0.14
6. Utilities	3 (0.5)		−0.35	−0.35	−0.04	−0.04
7. Other: construct, trans, hotels, entertainment	18 (3.0)		−0.33	−0.33	−0.08	−0.08
Less unionized industry
8. Energy oil, gas, and coal extraction	5 (1.6)		−0.69	−0.59	−0.10	−0.03
9. Computers, software, and electronic equipment	21 (3.2)		−0.07	−0.06	−0.03	0.00
10. Shops wholesale, retail, and some services	28 (5.8)		−0.39	−0.32	−0.15	−0.20
11. Healthcare, medical equipment, and drugs	10 (3.0)		−0.25	−0.20	−0.04	0.00
12. Money finance	56 (3.8)		−0.10	0.00	−0.03	0.00
	Pension-freezing sample (N = 222)		Non-pension-freezing sample (N = 6,984)
	M	Mdn	SD	M	Mdn	SD
Panel C: summary statistics on key variables
CSRCOM_i,t	−0.274	−0.200	0.465	−0.246	−0.200	0.516
CSREMP_i,t	−0.099	0.000	0.189	−0.064	0.000	0.185
ROA_i,t	0.038	0.028	0.164	0.041	0.039	0.080
FUNDING_i,t	0.775	0.803	0.261	0.768	0.787	0.287
UNDERFUND_i,t	0.838	1.000	0.369	0.873	1.000	0.333
PLANSIZE_i,t	4.332	4.267	2.289	4.988	5.104	2.393
FIRMSIZE_i,t	7.868	7.829	1.660	8.238	8.083	1.674
CFO_i,t	0.084	0.074	0.071	0.093	0.086	0.072
DLOSS_i,t	0.212	0.000	0.409	0.124	0.000	0.329
LEV_i,t	0.222	0.195	0.178	0.248	0.231	0.180
BM_i,t	0.699	0.538	1.197	0.536	0.477	0.719
SALESGR_i,t	0.020	0.011	0.316	0.075	0.061	0.239
DCAPEXP_i,t	−0.003	0.000	0.024	0.000	0.000	0.022
DRND_i,t	0.000	0.000	0.006	0.000	0.000	0.009
DDIV_i,t	0.000	0.000	0.012	0.000	0.000	0.010
UNION_i,t	0.432	0.000	0.497	0.467	0.000	0.499

Note. The table reports descriptive statistics. Panel A reports pension-freezing firms and CRS scores by year. Panel B reports pension-freezing firms and their CSRCOM and CSREMP by industry. Panel C reports summary statistics on key variables used in the study. The list of variable definition is attached in the appendix. CSR = corporate social responsibility.

Panel B of Table 2 reports the distribution of the number of hard-freezing and CSR scores by industry. First, we classify firms into two groups: a member of more unionized industry versus less unionized industry according to the U.S. Bureau of Labor Statistics (2011; 2009 data in Table 3), and they are further classified into industry groups following the classification scheme used in Fama and French (1997). Among the more unionized industries, the non-durable manufacturing industry (food, tobacco, textile, apparel, toys) and the machinery industry have more freezing of pension plans, while the wholesales and finance industries also have more freezing plans among the less unionized industries. The computers, software, and electronic equipment industries have the highest average CSRCOM (−0.07) and CSREMP (−0.03) scores, while the energy industry has the lowest CSRCOM score (−0.69) and the wholesale and retail industry has the lowest CSREMP (−0.15) among industry groups.

Table 3.

The Pearson and Spearman Correlations.

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.	14.	15.	16.	17.
1. FREEZE_i,t	1.0	−.009	−.032	−.031	.004	−.019	−.047	−.038	−.022	.046	−.026	.038	−.039	−.021	−.007	.002	−.012
2. CSRCOM_i,t	−.007	1	.471	.025	−.008	−.045	−.255	−.258	.025	−.059	−.106	.010	.002	−.013	−.005	.009	−.116
3. CSREMP_i,t	−.024	.455	1	.045	.079	−.132	.015	.083	.046	−.076	−.079	.012	−.003	−.009	.010	.013	−.107
4. ROA_i,t+1	−.036	−.020	−.006	1	.016	.045	.081	−.061	.473	−.237	−.103	−.156	.083	.031	−.075	.033	.131
5. FUNDING_i,t	.016	.015	.133	−.015	1	−.536	.451	.207	.010	−.108	.011	−.044	.003	.025	−.014	.012	.027
6. UNDERFUND_i,t	−.019	−.044	−.138	.078	−.577	1	−.158	−.137	.051	.071	.025	.006	−.016	−.013	.005	−.004	.056
7. PLANSIZE_i,t	−.053	−.277	−.020	.109	.381	−.144	1	.671	.064	−.061	.123	−.085	−.068	.004	−.025	−.013	.109
8. FIRMSIZE_i,t	−.042	−.255	.036	−.151	.212	−.117	.683	1	−.153	−.072	.094	.038	−.011	−.003	−.028	−.013	−.194
9. CFO_i,t	−.023	−.015	.001	.592	−.012	.068	.079	−.167	1	−.198	−.089	−.166	.016	.038	−.057	.059	.250
10. DLOSS_i,t	.046	−.052	−.071	−.256	−.139	.071	−.054	−.068	−.188	1	.191	.084	−.131	−.058	.037	−.045	.056
11. LEV_i,t	−.032	−.133	−.082	−.125	−.010	.013	.185	.140	−.077	.138	1	−.169	.009	−.013	−.004	−.038	.119
12. BM_i,t	.038	.009	−.014	−.495	−.081	.012	−.130	.045	−.411	.105	−.097	1	−.071	−.026	−.002	−.015	−.093
13. SALESGR_i,t	−.052	−.006	−.008	.176	.023	−.029	−.073	−.038	.034	−.197	−.030	−.163	1	.139	.004	.032	.014
14. DCAPEXP_i,t	−.024	−.028	−.021	.027	.030	−.004	.017	−.004	.007	−.050	−.008	−.033	.172	1	.047	.020	.011
15. DRND_i,t	.000	−.020	−.003	−.028	−.012	.013	−.016	−.006	−.004	.019	.022	.005	.019	.152	1	.019	−.013
16. DDIV_i,t	−.003	−.017	−.002	.069	.044	−.034	.029	.012	.097	−.066	−.048	−.055	−.021	.134	.127	1	.000
17. UNION_i,t	−.012	−.123	−.129	.240	−.027	.056	.106	−.187	.282	.056	.111	−.169	.028	.021	.001	.033	1

Note. The table reports the Pearson (upper triangle) and Spearman (lower triangle) correlation coefficients among the variables used in the study. The bold values represent the statistical significance at least at 5% level. The list of variable definition is attached in the appendix.

Panel C of Table 2 reports summary statistics of the variables used in our multivariate regression analyses. The average CSRCOM or CSREMP of pension-freezing firms is lower than that of non-pension-freezing firms, while the mean difference between the pension-freezing and the non-freezing firms is insignificant. Firms with pension freezes have, on average, lower FIRMSIZE, PLANSIZE, and CFO, but have higher DLOSS, implying that firms that choose to freeze their pension plans are smaller and have smaller plan sizes, but have higher financial risk. In addition, a lower BM and a higher SALESGR for non-pension-freezing firms imply that growth firms are less likely to freeze their pension plans. Those firms have higher numbers of young participants and the potential other costs associated with a pension freeze may be greater than the pension cost savings from a pension freeze.

Table 3 presents the Pearson and Spearman correlation coefficients between dependent, variable of interests, and control variables. Pension-freezing is negatively correlated with both CSREMP and CSRCOM, but only CSREMP is statistically significant at least at 5% level. These univariate results indicate that firms with higher CSR scores regarding employee relations are less likely to freeze their plans. In addition, pension-freezing is negatively correlated with future ROA, implying that pension-freezing firms tend to have lower profitability 1 year after the freezing. Pension-freezing is also negatively correlated with FIRMSIZE, PLANSIZE, and CFO. However, it is positively correlated with DLOSS. Consistent with prior studies (Cheng et al., 2014; Servaes & Tamayo, 2013), the Pearson correlation coefficients between CSR scores and firm performance are positive.

Empirical Results

Effect of CSR on the Hard-Freezing Decision

We first investigate whether a firm’s commitment to CSR measured by CSR scores affects a firm’s pension plan freezing decision. Columns 1 and 2 of Table 4 report that CSREMP_i,t is negatively associated with FREEZE_i,t₊₁ at less than 1% significance level, where the coefficient on CSRCOM_i,t has the same direction but is insignificant (t-stats = −1.30). These results indicate that firms with high CSR scores, especially with respect to a firm’s commitment to employee relations, are less likely to hard-freeze their pension plans. In addition, this relation between CSR and pension-freezing is less pronounced when the overall CSR score is used. The results do not support the notion that firms that already have a high commitment to employee relations are more likely to freeze their DB pension plans.

Table 4.

Effect of CSR on Hard-Freezing Decision.

	Dependent variables
	1. FREEZE_i,t₊₁	2. FREEZE_i,t₊₁
CSRCOM_i,t	−0.177
	(–1.30)
CSREMP_i,t		−1.166***
		(–3.08)
FUNDING_i,t	−0.294	−0.282
	(–0.93)	(–0.87)
UNDERFUND_i,t	0.070	0.016
	(0.28)	(0.06)
PLANSIZE_i,t	0.093**	0.085*
	(1.99)	(1.83)
FIRMSIZE_i,t	−0.222***	−0.205***
	(–3.59)	(–3.38)
CFO_i,t	−0.664	−0.492
	(–0.73)	(–0.53)
DLOSS_i,t	0.325*	0.319*
	(1.76)	(1.72)
LEV_i,t	0.413	0.359
	(1.18)	(1.02)
BM_i,t	0.062	0.065
	(1.01)	(1.06)
SALESGR_i,t	0.026	0.031
	(0.08)	(0.10)
DCAPEXP_i,t	1.954	1.780
	(0.67)	(0.61)
DRND_i,t	1.537	2.251
	(0.23)	(0.34)
DDIV_i,t	−8.725*	−8.475
	(–1.69)	(–1.63)
UNION_i,t	0.750	0.810
	(1.40)	(1.44)
Constant	−1.334**	−1.457**
	(–2.07)	(–2.30)
Industry and year effects	Yes	Yes
Observations	7,206	7,206
Pseduo R²	.066	.068

Note. The table reports the effect of CSR on pension-freezing decision using the probit regression for the fiscal year 2001–2009 with industry and year fixed effects. The t-statistics based on two-tailed tests are reported in parentheses. Standard errors are adjusted for clustering at the firm level. The list of variable definition is attached in the appendix. CSR = corporate social responsibility.

, **, and *** indicate the statistical significance at the 10%, 5%, and 1% levels, respectively.

Among the control variables, FIRMSIZE_i,t is significantly and negatively related with FREEZE_i,t₊₁, indicating that larger firms are less likely to freeze their plans.¹¹ This is consistent with Comprix and Muller (2011), who document that larger firms tend to face more constraints in pension-freezing due to greater resistance from a large number of affected employees. The coefficient on DLOSS_i,t is also positive and significant, implying that firms with an earnings loss are more likely to freeze their plans.¹²

Findings in Table 4 can be driven by a potential reverse causality, that is, the freezing of the pension plan may facilitate the decrease of CSR scores. To address the potential reverse causality concerns, we investigate whether firms’ pension-freezing decisions affect the CSR scores with Granger causality tests. We regress 1-year ahead of CSR scores (CSRCOM_i,t₊₁ or CSREMP_i,t₊₁) on FREEZE_i,t after controlling for current CSR scores and other control variables with a firm-fixed effect regression model. Table 5 reports that the coefficients on FREEZE_i,t are insignificant for both specifications with CSRCOM_i,t₊₁ and CSREMP_i,t₊₁, implying that there is no reversal causality issue.

Table 5.

Effect of CSR on Hard-Freezing Decision: Reverse Causality Test.

	Dependent variables
	CSRCOM_i,t+1	CSREMP_i,t+1
CSRCOM_i,t	0.875***
	(93.72)
CSREMP_i,t		0.822***
		(82.61)
FREEZE_i,t	–0.014	–0.007
	(–0.78)	(–0.92)
FUNDING_i,t	0.044***	0.030***
	(3.37)	(4.07)
UNDERFUND_i,t	−0.008	−0.005
	(–0.71)	(–0.98)
PLANSIZE_i,t	−0.005**	−0.003**
	(–2.07)	(–2.19)
FIRMSIZE_i,t	−0.004	0.003*
	(–0.96)	(1.82)
CFO_i,t	0.152***	0.064***
	(2.92)	(2.71)
DLOSS_i,t	−0.005	−0.007
	(–0.43)	(–1.31)
LEV_i,t	−0.008	−0.020**
	(–0.40)	(–2.20)
BM_i,t	−0.004	−0.001
	(–0.84)	(–0.84)
SALESGR_i,t	−0.024*	0.002
	(–1.86)	(0.34)
DCAPEXP_i,t	−0.166	−0.008
	(–1.30)	(–0.10)
DRND_i,t	−0.072	−0.063
	(–0.26)	(–0.62)
DDIV_i,t	−0.585*	−0.120
	(–1.71)	(–0.81)
UNION_i,t	−0.012	−0.009
	(–0.31)	(–0.49)
Constant	0.025	−0.035**
	(0.74)	(–2.14)
Firm and year effects	Yes	Yes
Observations	6,062	6,062
R ²	.7792	.6798

Note. The table reports the results of Granger reverse causality tests for the effect of CSR on pension-freezing decision with firm- and year-fixed effects. The t-statistics based on two-tailed tests are reported in parentheses. Standard errors are adjusted for clustering at the firm level. The list of variable definition is attached in the appendix. CSR = corporate social responsibility.

, **, and *** indicate the statistical significance at the 10%, 5%, and 1% levels, respectively.

Effect of CSR on the Association Between Pension-Freezing Decision and Firm Performance

Rubin (2007) documents that, in addition to financial risk from the increased volatility of pension assets in a DB plan, mounting pension benefit costs pressure firms to freeze their pension plan. However, firms’ pension-freezing decisions are likely to influence employee behavior leading to a decrease in productivity, and an increase in indirect costs associated with the increased employee turnover (Allen & Clark, 1987). Hence, we investigate whether pension-freezing is beneficial for firms that try to increase their profitability in the hope that the potential costs from the decreased productivity would be offset by the savings in pension cost.

As a baseline analysis we, first, estimate ROA_i,t₊₁ using only firms that already have frozen their pension plans and test whether a firm’s pension-freezing actually increases firm performance during the post-freezing period. We estimate ROA_i,t₊₁ with an indicator variable, POST_i,t, and other control variables, where POST_i,t indicates the post-period of pension-freezing for firms that have frozen their plans. Column 1 of Table 6 reports the insignificant coefficient on POST_i,t, implying that there is, on average, no clear effect of pension freezes on profitability in the post-period of pension-freezing. To further explore the influence of pension-freezing on future firm performance, we include an indicator variable of pension expense sensitivity and its interaction term with POST_i.t. SENSITIVE50_i.t (SENSITIVE75_i,t) takes a value of 1 when a firm’s ratio of the pension service cost to the firm’s revenue is above the median (highest quartile) of the ratio in the preceding year of pension-freezing, 0 otherwise. Therefore, we expect that firms with a high pension cost sensitivity are more likely to show performance improvements in the post-period of pension-freezing. The coefficients on both POST_i,t×SENSITIVE50_i,t and POST_i,t×SENSITIVE75_i,t are positive and significant. However, the magnitude of coefficients on POST_i,t×SENSITIVE75_i,t is much higher than that of POST_i,t×SENSITIVE50_i,t, implying that firms with a higher pension cost sensitivity to earnings are more likely to improve on their accounting profitability (ROA_i,t₊₁) during the post-freezing period.¹³

Table 6.

Effect of Pension-Freezing Decision on Firm Performance (Sensitivity Test).

	Dependent variables: ROA_i,t₊₁
	1.	2.	3.
POST_i,t	−0.005	−0.014	−0.013
	(–0.56)	(–1.45)	(–1.54)
POST_t×SENSITIVE50_i,t		0.019*
		(1.81)
POST_t×SENSITIVE75_i,t			0.036***
			(3.02)
ROA_i,t	−0.136***	−0.140***	−0.144***
	(–3.82)	(–3.93)	(–4.04)
PLANSIZE_i,t	−0.003	−0.003	−0.003
	(–0.54)	(–0.47)	(–0.43)
FIRMSIZE_i,t	0.067***	0.069***	0.070***
	(6.12)	(6.24)	(6.41)
SALESGR_i,t	0.033***	0.031***	0.030***
	(3.29)	(3.17)	(3.04)
LEV_i,t	−0.345***	−0.350***	−0.359***
	(–8.65)	(–8.77)	(–8.99)
BM_i,t	−0.009***	−0.009***	−0.009***
	(–2.98)	(–2.98)	(–3.07)
Constant	−0.413***	−0.427***	−0.440***
	(–4.77)	(–4.91)	(–5.08)
Firm and year effects	Yes	Yes	Yes
Observations	1,060	1,060	1,060
R ²	.2010	.2042	.2098

Note. The table reports the results of firm- and year-fixed effect regressions for the pre- and post-period (spanning 2001–2009) of pension-freezing. POST_i,t is an indicator variable set to 1 if a firm’s fiscal year end is at least one year after freezing year. SENSITIVE50_i,t (SENSITIVE75_i,t) is an indicator variable that takes a value of 1 if a firm’s pension-freezing sensitivity to earnings (the ratio of the service cost to firm’ annual revenue) is above the median (the highest quartile) of the annual pooled sample in the preceding year of pension-freezing. The t-statistics based on two-tailed tests are reported in parentheses. Standard errors are adjusted for clustering at the firm level. The list of variable definition is attached in the appendix.

, **, and *** indicate the statistical significance at the 10%, 5%, and 1% levels, respectively.

Next, we investigate the effect of firms’ commitment to CSR activities on the relation between pension plan freezes and future firm performance using the interaction term between freezing indicator variable (FREEZE) with firms’ CSR level indicator variable (HiCSR). Columns 1, 3, and 5 of Table 7 are based on the CSRCOM measure while Columns 2, 4, and 6 are based on the CSREMP measure.

Table 7.

Effect of CSR on the Relations Between Pension-Freezing Decision and Firm Performance.

	Dependent variables: ROA_i,t₊₁
	OLS regression		Two-stage regression		PSM sample regression
	CSRCOM	CSREMP	CSRCOM	CSREMP	CSRCOM	CSREMP
FREEZE_i,t	−0.020**	−0.018**	−0.155***	−0.156***	−0.021*	−0.014
	(–2.10)	(–2.13)	(–3.44)	(–3.50)	(–1.77)	(–1.19)
HiCSR_i,t	−0.022	−0.024	−0.022**	−0.024***	−0.019	−0.017
	(–1.28)	(–1.43)	(–2.41)	(–2.61)	(–0.38)	(–0.37)
FREEZE_i,t×HiCSR_i,t	0.027**	0.028***	0.026***	0.027***	0.029**	0.022
	(2.56)	(2.80)	(3.93)	(4.08)	(1.98)	(1.64)
ROA_i,t	0.299***	0.317***	0.292***	0.309***	0.090	0.101
	(4.88)	(5.34)	(21.78)	(24.12)	(1.18)	(1.21)
PLANSIZE_i,t	0.004***	0.004***	0.003***	0.004***	0.009***	0.011***
	(3.57)	(4.73)	(4.42)	(5.60)	(3.50)	(3.74)
FIRMSIZE_i,t	−0.005***	−0.006***	−0.005***	−0.006***	−0.009**	−0.011**
	(–3.82)	(–4.70)	(–4.54)	(–5.66)	(–1.98)	(–2.41)
SALESGR_i,t	0.017**	0.017**	0.014***	0.013***	0.016	0.014
	(2.24)	(2.33)	(2.73)	(2.81)	(0.91)	(0.90)
LEV_i,t	−0.043***	−0.039***	−0.045***	−0.041***	−0.061**	−0.052*
	(–4.45)	(–4.02)	(–6.67)	(–6.30)	(–2.01)	(–1.70)
BM_i,t	−0.009	−0.008*	−0.008***	−0.007***	−0.012***	−0.013**
	(–1.64)	(–1.94)	(–4.83)	(–4.92)	(–2.76)	(–2.58)
ROA_i,t×HiCSR_i,t	0.189**	0.166*	0.189***	0.168***	0.400***	0.474***
	(1.98)	(1.72)	(9.48)	(8.16)	(3.39)	(4.14)
PLANSIZE_i,t×HiCSR_i,t	0.000	−0.002	0.000	−0.002*	−0.003	−0.006*
	(0.16)	(–1.54)	(0.18)	(–1.65)	(–0.90)	(–1.91)
FIRMSIZE_i,t×HiCSR_i,t	0.001	0.003**	0.001	0.003**	0.000	0.002
	(0.57)	(2.09)	(0.74)	(2.33)	(0.04)	(0.46)
SALESGR_i,t×HiCSR_i,t	−0.017	−0.019*	−0.018**	−0.019***	−0.034	−0.019
	(–1.63)	(–1.84)	(–2.54)	(–2.62)	(–1.64)	(–0.98)
LEV_i,t×HiCSR_i,t	0.040***	0.034**	0.039***	0.034***	0.049	0.024
	(2.90)	(2.42)	(4.09)	(3.46)	(1.32)	(0.63)
BM_i,t×HiCSR_i,t	−0.002	−0.009	−0.002	−0.008***	0.007	0.011**
	(–0.32)	(–1.14)	(–1.01)	(–3.20)	(1.17)	(2.02)
Constant	0.071***	0.072***	0.080***	0.082***	0.083**	0.086**
	(6.79)	(7.30)	(9.19)	(9.75)	(2.02)	(2.19)
Industry and year effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	7,206	7,206	7,206	7,206	439	439
R² or Wald χ²	0.279	0.278	1,706.2	1,666.1	0.195	0.197

Note. The table reports the effect of CSRCOM and CSREMP on the relations between pension-freezing decision and firm performance using the OLS regressions, the two-stage least squares regressions, and the propensity-score-matched control sample approach for fiscal year 2001–2009 with industry- and year-fixed effects. The high (low) CSR subsample includes firms whose CSRCOM/CSREMP score is above (below) the annual median CSRCOM/CSREMP score. R² is reported for OLS regressions, and chi-square from the Wald test is reported for all other regressions. The t-statistics based on two-tailed tests are reported in parentheses. Standard errors are adjusted for clustering at the firm level. The list of variable definition is attached in the appendix. CSR = corporate social responsibility; OLS = ordinary least squares.

, **, and *** indicate the statistical significance at the 10%, 5%, and 1% levels, respectively.

In the OLS regression analysis, we document that pension-freezing (FREEZE) is negatively related to the future ROA at less than 5% significance level for firms with low CSRCOM scores (coefficient = −0.020, t-stats = −2.10). When CSR is measured by CSREMP scores, FREEZE has also a negative coefficient (coefficient = −0.018, t-stats = −2.13). However, the impact of pension-freezing (FREEZE) on future ROA is positive for firms with high CSR scores. The coefficient on FREEZE×HiCSR is 0.027 (CSRCOM) and 0.028 (CSREMP), and they are significant at least 5% level. These results indicate that pension-freezing has a positive effect on firm performance when the firm has a high CSR score, whereas it has a negative effect on firm performance when the firm’s CSRCOM score is low.

As a firm’s decision to freeze pension plans may not be a random event, we try to address this endogeneity issue through a method suggested by Heckman (1976, 1979) and the propensity-score-matched control sample approach. Columns 3 and 4 in Table 7 report the two-stage regression results that are similar to our findings based on the OLS regression.¹⁴ For example, firms’ future ROA_i,t₊₁ is negatively associated with pension-freezing for firms with the low CSRCOM scores (coefficient = −0.155, t-stats = −3.44). On the contrary, the impact of pension-freezing on future ROA is significantly positive for firms with high CSRCOM scores at less than 1% level (coefficient for FREEZE×HiCSR = 0.026, t-stats = 3.93). When CSR is measured by CSREMP, the results are similar.

In addition, the results from the propensity-score-matched control sample approach are consistent with those from our OLS regressions and from the two-stage regression results (Columns 5 and 6), but the CSREMP results are slightly weaker, though. As the Hausman specification test indicates that a firm’s decision to freeze pension is not free from the endogeneity issue, we only report our results based on the two-stage regressions proposed by Heckman and propensity-score-matched control sample regressions in subsequent tables.¹⁵

In Table 6, we have restricted our sample to the pension-freezing firms only and compare their performance before and after the freezing year. In such a design, the pension-freezing firms act as their own control. To examine how a firm’s CSR activity affects future performance, we conduct a similar test as in Table 6 by including the POST and HiCSR interaction terms with a sample that only includes pension-freezing firms. We find that a positive relationship between a firm’s future ROA and pension-freezing is greater for firms with a high level of CSR activities. The coefficient for the interaction term POST×HiCSRCOM is 0.017 (t = 1.59) and POST×HiCSREMP is 0.025 (t = 2.21). Therefore, these findings based on the sample with only pension-freezing firms corroborate the results reported in Table 7, which is based on the cross-sectional studies that include both pension-freezing and non-freezing firms (untabulated).

Additional Analysis

In this section, we first test whether labor unionization can affect the association between CSR activities and firm performance when firms freeze their pension plans. Prior studies document that firms with unionized labor forces are less able to freeze the plan due to the collective bargaining agreement (Comprix & Muller, 2011; J. J. Kim et al., 2015). To reduce the negative effect of the pension freeze, it is known that managers try to obtain concessions from the unions during the pension-freezing process. Labor unions may accept hard-freezing by requesting substitute compensation programs or resist hard-freezing, which may affect employee productivity and eventually firm performance. For example, Rauh et al. (2019) document that unions help employees to recuperate the losses from the pension freeze through other forms of compensation. Therefore, we conjecture that the effect of CSR on the relation between pension freeze and firm’s performance may be more pronounced for firms in more unionized industry. As the firm level of labor union information is not available, we classify firms into a more unionized industry (less unionized industry) when the percentage of employees in the unionized workforce is above (below) the median of all industries.

In Table 8, we report the impact of labor unions on the relations between pension hard-freezing and ROA_i,t₊₁ along with the level of CSR activities. After controlling for pension-freezing (FREEZE), the labor unions seem to have a positive impact on the firm’s future performance (coefficient = .057 and t-stats = 4.10, Column 1). However, the effect of pension freezes on a firm’s future performance is more negative for firms with more unionized labor than those firms with less unionized labor force. For example, the coefficient for FREEZE×Union is −.033 (t-stats = −3.40) or −.022 (t-stats = −2.61). Interestingly, when a pension-freezing firm has a high CSR score (FREEZE×HiCSR), the labor unions have a positive impact on the firm’s future performance. The coefficient for FREEZE×HiCSR×Union is .045 (t-stats = 3.23) or 0.027 (t-stats = 1.98), which indicates that firms operating in the more unionized industries have a better future performance after pension-freezing only when firms have high CSR scores.¹⁶

Table 8.

Effect of Pension-Freezing Decision on Firm Performance by Unionization.

	Dependent variables: ROA_i,t₊₁
	Two-stage regression
	Full	CSRCOM	CSREMP
FREEZE_i,t	−0.120**	−0.134***	−0.148***
	(–2.54)	(–2.92)	(–3.33)
Union_i,t	0.057***	0.014	0.013
	(4.10)	(1.33)	(1.29)
FREEZE_i,t×Union_i,t		−0.033***	−0.022***
		(–3.40)	(–2.61)
HiCSR_i,t		−0.026***	−0.027***
		(–2.58)	(–2.63)
HiCSR_i,t×Union_i,t		0.003	0.002
		(0.82)	(0.43)
FREEZE_i,t×HiCSR_i,t		0.006	0.015*
		(0.68)	(1.78)
FREEZE_i,t×HiCSR_i,t×Union_i,t		0.045***	0.027**
		(3.23)	(1.98)
ROA_i,t	0.489***	0.295***	0.311***
	(28.98)	(21.83)	(24.24)
PLANSIZE_i,t	0.002***	0.004***	0.004***
	(2.64)	(4.55)	(5.56)
FIRMSIZE_i,t	−0.002*	−0.005***	−0.006***
	(–1.83)	(–4.61)	(–5.65)
SALESGR_i,t	0.003	0.013***	0.013***
	(0.51)	(2.58)	(2.72)
LEV_i,t	−0.003	−0.044***	−0.041***
	(–0.37)	(–6.54)	(–6.26)
BM_i,t	−0.009***	−0.008***	−0.007***
	(–4.89)	(–4.93)	(–4.94)
ROA_i,t×Union_i,t	−0.179***	0.185***	0.165***
	(–8.58)	(9.19)	(7.98)
PLANSIZE_i,t×Union_i,t	0.003***	−0.000	−0.002*
	(2.89)	(–0.16)	(–1.71)
FIRMSIZE_i,t×Union_i,t	−0.005***	0.002	0.004**
	(–3.02)	(1.10)	(2.41)
SALESGR_i,t×Union_i,t	0.007	−0.017**	−0.019***
	(1.04)	(–2.45)	(–2.58)
LEV_i,t×Union_i,t	−0.047***	0.037***	0.032***
	(–4.57)	(3.81)	(3.27)
BM_i,t×Union_i,t	−0.002	−0.002	−0.008***
	(–0.71)	(–0.95)	(–3.16)
Constant	0.045***	0.080***	0.082***
	(4.99)	(9.01)	(9.63)
Industry and year effects	Yes	Yes	Yes
Observations	7,206	439	439
R² or Wald χ²	2,739.0	2,766.3	2,747.1

Note. The table reports the effect of pension freeze on firm performance by industry unionization (more unionized vs. less unionized industry) for fiscal year from 2001 to 2009. The results are based on the two-stage least squares regression with the industry- and year-fixed effects. A firm is classified in the more unionized industry sample when the percentage of unionized employees in its industry (Fama-French 12 industry classification) is above the median of all industries, otherwise in the less unionized industry. The Wald chi-square test is reported for the two-stage least squares regressions. The t-statistics based on two-tailed tests are reported in parentheses. Standard errors are adjusted for clustering at the firm level. The list of variable definition is attached in the appendix.

, **, and *** indicate the statistical significance at the 10%, 5%, and 1% levels, respectively.

Next, we examine how long the effect of pension-freezing on firm performance lasts. For this investigation, we replace ROA_t₊₁ with ROA_t₊₂ in Equation 2, where year t+2 indicates a time period of 2 years after hard-freezing and report the results in Table 9. Firms’ performance 2 years after hard freezes shows a similar pattern reported in Table 7. Specifically, the hard freeze has a negative impact on the firms’ performance for firms with low CSRCOM scores in the PSM sample regression setting (coefficient = −0.054 and t-stats = −3.10). However, for firms with high CSRCOM scores, the hard freeze has a positive impact on firms’ future performance. For example, the coefficient on FREEZE×HiCSR is positive and significant (coefficient = 0.059 and t-stats = 2.90), which is similar to what we find with ROA_t₊₁. In addition, we find similar results when CSREMP is used. Finally, we also find consistent results with the two-stage regression approach. However, when we extend a firm’s performance period to year t+3, we do not find the effect of CSR on the relation between hard-freezing and firm performance (untabulated), which may suggest one of two possibilities. First, under the implicit contract assumption, the savings from the pension freeze may be offset by higher wages or increased total employee benefits in the longer period. Another plausible explanation might be a typical mean-reversion of firm performance documented in many accounting studies (e.g., Sloan, 1996).

Table 9.

Effect of CSR on the Relations Between Pension-Freezing Decision and Firm Performance, After 2 Years.

	Dependent variables: ROA_i,t₊₂
	Two-stage regression		PSM sample regression
	CSRCOM	CSREMP	CSRCOM	CSREMP
FREEZE_i,t	−0.057	−0.048	−0.054***	−0.050***
	(–1.05)	(–0.87)	(–3.10)	(–3.31)
HiCSR_i,t	−0.027***	−0.016	0.024	−0.086**
	(–2.74)	(–1.63)	(0.47)	(–1.97)
FREEZE_i,t×HiCSR_i,t	0.042***	0.038***	0.059***	0.059***
	(4.23)	(3.73)	(2.90)	(3.08)
ROA_i,t	0.184***	0.218***	0.075	0.078
	(12.31)	(15.20)	(1.38)	(1.34)
PLANSIZE_i,t	0.004***	0.005***	0.006	0.006
	(5.64)	(6.38)	(1.30)	(1.46)
FIRMSIZE_i,t	−0.006***	−0.006***	−0.004	−0.008*
	(–5.12)	(–5.37)	(–0.76)	(–1.74)
SALESGR_i,t	0.015***	0.013***	0.004	0.001
	(2.82)	(2.62)	(0.24)	(0.03)
LEV_i,t	−0.038***	−0.035***	−0.164	−0.179
	(–5.45)	(–5.15)	(–1.31)	(–1.55)
BM_i,t	−0.012***	−0.011***	0.005	0.005
	(–6.72)	(–7.52)	(0.56)	(0.57)
HiCSR_i,t×ROA_i,t	0.156***	0.087***	0.152	0.254**;
	(7.11)	(3.83)	(1.37)	(2.09)
HiCSR_i,t×PLANSIZE_i,t	0.000	−0.001	−0.005	−0.006
	(0.25)	(–0.58)	(–0.79)	(–1.12)
HiCSR_i,t×FIRMSIZE_i,t	0.002	0.002	−0.007	0.005
	(1.13)	(1.25)	(–1.18)	(0.87)
HiCSR_i,t×SALESGR_i,t	−0.028***	−0.027***	−0.014	0.002
	(–3.86)	(–3.60)	(–0.62)	(0.09)
HiCSR_i,t×LEV_i,t	0.035***	0.032***	0.164	0.207*
	(3.50)	(3.11)	(1.27)	(1.74)
HiCSR_i,t×BM_i,t	−0.001	−0.005*	−0.011	−0.009
	(–0.21)	(–1.87)	(–1.29)	(–1.13)
Constant	0.076***	0.071***	0.082*	0.121***
	(8.14)	(7.84)	(1.96)	(3.86)
Industry and year effects	Yes	Yes	Yes	Yes
Observations	7,027	7,027	428	428
R² or Wald χ²	1,706.2	1,666.11	0.1276	0.1269

Note. The table reports the effect of CSR on the relations between pension-freezing decision and firm performance using the two-stage least squares regressions and the propensity-score-matched control sample approach for the fiscal year 2001–2009 with industry- and year-fixed effects. The ROA is measured 2 years after the pension-freezing year. The high (low) CSR subsample includes firms whose CSR score is above (below) the annual median CSR score. R² is reported for OLS regressions and the Wald chi-square test is reported for all other regressions. The t-statistics based on two-tailed tests are reported in parentheses. Standard errors are adjusted for clustering at the firm level. The list of variable definition is attached in the appendix. CSR = corporate social responsibility; ROA = return on assets; OLS = ordinary least squares.

, **, and *** indicate the statistical significance at the 10%, 5%, and 1% levels, respectively.

Finally, we test whether firms’ pension-freezing news along with CSR levels provides new information to investors. We conduct the event study surrounding the 10-K filing dates with three event windows: 3 days (–1, +1), 5 days (–2, +2), 15 days (–7, +7). Here we assume that pension-freezing news is released through 10-K filing. In the event study, we examine the stock market’s reaction to the pension-freezing news by comparing cumulative abnormal returns (CAR) between the freezing firms and control groups and its interaction with firms’ CSR level (high CSR vs. low CSR firms).¹⁷

We find that the CAR between the freezing and the non-freezing control samples is very small in magnitude (i.e., less than 0.5%) and they are not statistically different from zero at 10% level in 3- and 5-day windows (not tabulated). We find CAR (−7, +7) for the freezing firms is 1.74% (t-stats = 1.66) and 0.96% (t-stats = 1.23) for the control sample, but they are not statically different from each other. We also explore whether a firm’s CSR activity affects the investor’s reaction to pension-freezing news by comparing CARs between high and low CSR subsamples among pension-freezing firms. In both 3-day and 5-day windows, the magnitude of CAR is still very low in both subsamples—all less than 1% and statistically insignificant. Moreover, two subsamples (high vs. low CSR freezing firms) do not show much difference in CARs. CAR (−7, +7) for the high CSR sample is 2.25% (t-stats = 1.70), whereas CAR (−7, +7) for the low CSR sample is 1.33% (t-stats = 0.85), and they are not statistically different from each other.

Our empirical analysis based on the market returns tests does not find evidence that the pension-freezing news provides new information to the investor. We believe the lack of empirical support is attributable to a couple of factors. First, the pension-freezing announcement may not provide any news to the market at least at the time of the 10-K filing date because the market may have found out the news through other channels. Another possibility is that the stock market may not really understand the implication of pension freezes on the firm’s future performance. For example, Picconi (2006) provides evidence that the market does not fully reflect the effects of changes in pension assumptions on the firm’s future earnings that are available in the 10-K reports. Thus, because of the limitations of a short window event study in our setting, our research design relies on accounting-based performance, which provides clear support for our hypotheses.

Conclusion

While the number of firms that freeze their DB pension plans has been increasing in recent years, prior studies have not provided conclusive evidence of the impact of the pension freeze on firm performance. We believe two different factors contribute to the inconclusive relationship between pension freezes and firm performance. First, the pension-freezing firms are expected to improve their financial performance because of cost savings from decreased pension obligations through pension freezes. Second, the decreased overall compensation for employees following the pension freeze may have a negative impact on employee retention and employee productivity, which will affect the firm’s future performance.

We bring a firm’s level of CSR activities into consideration in examining the relations between a firm’s pension-freezing decision and its future performance. We document that firms with a high level of CSR activities, especially employee relations, are less likely to freeze their pension plans. In addition, we find that the firm’s performance after pension freeze measured by the return on assets will be higher for firms with high CSR scores than those with low CSR scores. These findings support the premise that a firm’s engagement in CSR activities (especially the CSR activities associated with employee relations) indeed mitigates the negative impact of pension freezes on employee productivity, and eventually leads to firms’ better future performance.

Footnotes

Appendix

Appendix.

Variable Definitions.

Variable name	Definition
FREEZE_i,t	An indicator variable that takes a value of 1 if a firm i freezes defined benefit plan, 0 otherwise at year t.
ROA_i,t	Income before extraordinary items and pension expense divided by total assets at the end of the year t for a firm i.
CSRCOM_i,t	The sum of indicator variables of the strengths and concerns for seven categories in KLD database scaled by the number of strengths and concerns, where seven categories include Employee, Community, Diversity, Environment, Humanity, Product, and Governance at year t for a firm i.
CSREMP_i,t	The sum of indicator variables of the strengths and concerns for employee category in KLD database scaled by the number of strengths and concerns at year t for a firm i.
POST_i,t	An indicator variable that takes a value of 1 if a firm i’s fiscal year end is at least one year after freezing year t, 0 otherwise.
SENSITIVE50_i,t (SENSITIVE75_i,t)	An indicator variable that takes a value of 1 if a firm i’s pension-freezing sensitivity to earnings (the ratio of the service cost to firm’s annual revenue) is above the median (the highest quartile) of the annual pooled sample in the preceding year of pension-freezing at year t.
FUNDING_i,t	Fair value of plan assets divided by projected benefit obligation at year t for a firm i.
UNDERFUND_i,t	An indicator variable that takes a value of 1 if a firm i’s FUNDING is less than 1 (i.e. the pension is underfunded) at year t, 0 otherwise.
PLANSIZE_i,t	Natural logarithm of (1 + FV of plan assets) at the end of the year t for a firm i.
FIRMSIZE_i,t	Natural logarithm of (1 + total assets) of the plan sponsor at the end of the year t for a firm i.
CFO_i,t	Cash flow from operations before pension contributions divided by total assets at the end of the year t for a firm i.
DLOSS_i,t	An indicator variable that takes a value of 1 if earnings before extraordinary items in years t and t–1 sum to less than 0 for a firm i, 0 otherwise.
LEV_i,t	The sum of long-term debt and debt in current liabilities scaled by the total assets at the end of the year t for a firm i.
BM_i,t	The book value of equity divided by market value of equity at the end of the year t for a firm i.
SALESGR_i,t	Sales growth rate measured by the change in sales relative to the prior year sales scaled by the lagged sales at year t for a firm i.
DCAPEXP_i,t	Annual changes in capital expenditures from year t−1 to t divided by total assets at the end of the year t for a firm i.
DRND_i,t	Annual changes in research and development expenses from year t−1 to t divided by total assets at the end of the year t for a firm i.
DDIV_i,t	Annual changes in dividend paid from year t−1 to t divided by total assets at the end of the year t for a firm i.
UNION_i,t	An indicator variable that takes a value of 1 if a firm i operates in a more unionized industry, 0 otherwise at year t. More unionized industries include following six industries based on Fama-French 12 industry classification: (1) food, tobacco, textiles, apparel, toys; (2) cars, TV’s, furniture, household appliances; (3) machinery, trucks, planes, off furn, paper; (4) chemicals and allied products; (5) telcom, telephone, and TV transmission; (6) utilities

Note. The appendix defines the variables used in the study. KLD = Kinder, Lydenberg, and Domini.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Seokyoun Hwang

Notes

References

Allen

S. G.

Clark

R. L.

(1987). Pensions and firm performance. In Kleiner

M. M.

Block

R. N.

Roomsted

Salsburg

S.W.

(Eds.), Industrial relations research association series. Human resources and the performance of the firm (pp. 195-242). Industrial Relations Research Association.

Atkinson

A. A.

Waterhouse

J. H.

Wells

R. B.

(1997). A stakeholder approach to strategic performance measurement. MIT Sloan Management Review, 38(3), 25-37.

Barney

J. B.

Wright

P. M.

(1998). On becoming a strategic partner: The role of human resources in gaining competitive advantage. Human Resource Management, 37(1), 31-46.

Becker

Gerhart

(1996). The impact of human resource management on organizational performance: Progress and prospects. Academy of Management Journal, 39(4), 779-801.

Benabou

Tirole

(2010). Individual and corporate social responsibility. Economica, 77, 1-19.

The Bureau of Labor Statistics. (2011). Union members (News Release USDL-11-0063). U.S. Department of Labor.

Chatterji

A. K.

Levine

D. I.

Toffel

M. W.

(2009). How well do social ratings actually measure corporate social responsibility? Journal of Economics & Management Strategy, 18(1), 125-169.

Cheng

Ioannou

Serafeim

(2014). Corporate social responsibility and access to finance. Strategic Management Journal, 35(1), 1-23.

Choi

Wang

(2009). Stakeholder relations and the persistence of corporate financial performance. Strategic Management Journal, 30(8), 895-907.

10.

Choy

Lin

Officer

M. S.

(2014). Does freezing a defined benefit pension plan affect firm risk? Journal of Accounting and Economics, 57(1), 1-21.

11.

Comprix

Muller

K. A.

III . (2006). Asymmetric treatment of reported pension expense and income amounts in CEO cash compensation calculations. Journal of Accounting and Economics, 42(3), 385-416.

12.

Comprix

Muller

K. A.

III . (2011). Pension plan accounting estimates and the freezing of defined benefit pension plans. Journal of Accounting and Economics, 51(1), 115-133.

13.

Di Giuli

Kostovetsky

. (2014). Are red or blue companies more likely to go green? Politics and corporate social responsibility. Journal of Financial Economics, 111(1), 158-180.

14.

Fama

E. F.

French

K. R.

(1997). Industry costs of equity. Journal of Financial Economics, 43(2), 153-193.

15.

Flammer

(2015). Does corporate social responsibility lead to superior financial performance? A regression discontinuity approach. Management Science, 61(11), 2549-2568.

16.

Frankel

Litov

(2009). Earnings persistence. Journal of Accounting and Economics, 47(1–2), 182-190.

17.

Gao

Lisic

Zhang

(2014). Commitment to social good and insider trading. Journal of Accounting and Economics, 57, 149-175.

18.

Heckman

J. J.

(1976). The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Annals of Economic Social Measurement, 5(4), 475-492.

19.

Heckman

J. J.

(1979). Sample selection bias as a specification error. Econometrica, 47, l53-161.

20.

Heskett

J. L.

Jones

T. O.

Loveman

G. W.

Sasser

W. E.

Schlesinger

L. A.

(2008). Putting the service-profit chain to work. Harvard Business Review, 86, 118-129.

21.

Hillman

A. J.

Keim

G. D.

(2001). Shareholder value, stakeholder management, and social issues: What’s the bottom line? Strategic Management Journal, 22, 125-139.

22.

Jayachandran

Kalaignanam

Eilert

(2013). Product and environmental social performance: Varying effect on firm performance. Strategic Management Journal, 34(10), 1255-1264.

23.

Kim

J. J.

Liu

Yun

J. K.

(2015). Regulatory scrutiny, labor unions, and pension freezes. Journal of Business and Accounting, 8(1), 3-14.

24.

Kim

Park

M. S.

Wier

(2012). Is earnings quality associated with corporate social responsibility? The Accounting Review, 87, 761-796.

25.

Lins

K. V.

Servaes

Tamayo

(2017). Social capital, trust, and firm performance: The value of corporate social responsibility during the financial crisis. The Journal of Finance, 72(4), 1785-1824.

26.

McFarland

Pang

Warshawsky

(2009). Does freezing a defined-benefit pension plan increase company value? Empirical evidence. Financial Analysts Journal, 65(4), 47-59.

27.

Milevsky

M. A.

Song

(2008). Do markets like frozen DB pensions? An event study. York University.

28.

Munnell

A. H.

Golub-Sass

Soto

Vitagliano

(2006). Why are healthy employers freezing their pensions?Center for Retirement Research at Boston College. http://crr.bc.edu/wp-content/uploads/2006/03/ib_44.pdf

29.

Munnell

A. H.

Soto

(2004). The outlook for pension contributions and profits in the US. Journal of Pension Economics & Finance, 3(1), 77-97.

30.

Muse

L. A.

Rutherford

M. W.

Oswald

S. L.

Raymond

J. E.

(2005). Commitment to employees: Does it help or hinder small business performance? Small Business Economics, 24(2), 97-111.

31.

Pension Benefit Guaranty Corporation. (2008). Hard-frozen defined benefit plans: Findings for 2003-2004 and preliminary findings for 2005. https://www.pbgc.gov/documents/Frozen_Plans_0808.pdf

32.

Pension Benefit Guaranty Corporation. (2013). Single-employer DB plan freezes. https://www.pbgc.gov/Documents/Plan-Freeze-2013-EA-meeting.pdf

33.

Picconi

(2006). The perils of pensions: Does pension accounting lead investors and analysts astray? The Accounting Review, 81(4), 925-955.

34.

Putnam

R. D.

(1993). The Prosperous Community: Social capital and public life. In Ostrom

Ahn

(Eds.), Foundations of social capital (pp. 529-536). Edward Elgar.

35.

Rauh

J. D.

Stefanescu

Zeldes

S. P.

(2019). Cost saving and the freezing of corporate pension plans (Columbia Business School Research Paper No. 16–14). http://doi.org/10.2139/ssrn.2706448

36.

Rubin

(2007). The impact of pension freezes on firm value [Wharton research scholars]. University of Pennsylvania.

37.

Servaes

Tamayo

(2013). The impact of corporate social responsibility on firm value: The role of customer awareness. Management Science, 59(5), 1045-1061.

38.

Sloan

R. G.

(1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? Accounting Review, 71, 289-315.

39.

Vafeas

Vlittis

(2018). Independent directors and defined benefit pension plan freezes. Journal of Corporate Finance, 50, 505-518.

40.

Waddock

S. A.

Graves

S. B.

(1997). The corporate social performance–financial performance link. Strategic Management Journal, 18(4), 303-319.

41.

Wans

(2020). Corporate social responsibility and market-based consequences of adverse corporate events: Evidence from restatement announcements. Journal of Accounting, Auditing & Finance, 35(2), 231-262.

42.

Watson

Wyatt

. (2010, September). Pension freezes continue among fortune 1000 companies in 2010. Insider, 20(9), 1-12.