Employment Protection, Investment, and Firm Growth

Abstract

Critics of employment protection contend that it stifles corporate investment and growth.¹ Empirical work, however, does not find a consistent relation between employment protection and capital expenditures (Autor, Kerr, and Kugler 2007; Calcagnini, Giombini, and Saltari 2009; Calcagnini, Ferrando, and Giombini 2014). Moreover, even if greater employment protection lowers capital expenditures, prior work finds that it leads to more innovation (Acharya, Baghai, and Subramanian 2014; Griffith and Macartney 2014). Thus, the overall effect of employment protection on firm growth is unclear. In this study, we examine the relation between employment protection and both U.S. firm-level capital expenditures and sales growth.

From a theoretical perspective, the effect of employment protection on investment is ambiguous. On one hand, greater employment protection could lead to higher investment rates through two channels. First, if firms are less likely to discharge workers when protection increases, a lower threat of dismissal could incentivize workers to invest in firm-specific capital, resulting in greater productivity and higher investment rates (e.g., Nickell and Layard 1999; Belot, Boone, and van Ours 2007). Second, because employment protection makes physical capital a less expensive input relative to human capital, firms may increase investment in assets that are more capital and less labor intensive (e.g., Blanchard 1997; Caballero and Hammour 1998; Koeniger and Leonardi 2007).

On the other hand, greater employment protection makes it costlier for firms to divest or scale back poorly performing projects, resulting in lower investment rates by making projects more irreversible or by constraining firms’ access to external capital. First, the discounted value of a project includes the resale or recovery price from divesting. Consequently, greater investment irreversibility reduces firms’ ex ante incentives to invest by lowering the recovery value of projects and hence the number of projects with positive net present values (e.g., Bernanke 1983; Pindyck 1991; Bertola and Caballero 1994; Abel and Eberly 1996; Abel et al. 1996). Second, greater employment protection makes labor costs more fixed in nature, raising operating leverage and crowding out financial leverage (Simintzi, Vig, and Volpin 2015; Serfling 2016). If firms do not replace debt financing with equity, greater employment protection could limit their ability to finance investments.

Similar to Acharya, Baghai, and Subramanian (2014), we exploit the quasi-natural experiment created by the staggered adoption of Wrongful Discharge Laws (WDLs) by U.S. state courts to test these competing theories. We use these law changes to identify the effect of employment protection on firm-level investment decisions of U.S. public corporations. WDLs matured into three common laws that protect workers against different aspects of unjust dismissal. We focus our analysis on the adoption of one particular WDL: the good faith exception. This law applies in cases when a court determines that an employer discharged a worker out of bad faith, malice, or retaliation. Of the three exceptions, this law represents the largest deviation from at-will employment (e.g., Dertouzos and Karoly 1992; Kugler and Saint-Paul 2004). Importantly, changes in this law have a material effect on firms. For example, employment levels, employment volatility, and firm entry decrease following its adoption (Dertouzos and Karoly 1992; Autor, Kerr, and Kugler 2007). Firms also experience negative cumulative abnormal stock returns when their state adopts the law (Serfling 2016).

For our tests, we utilize a difference-in-differences research design in which the treatment and control groups consist of firms headquartered in states that have and have not adopted the good faith exception, respectively. We use panel regression techniques that control for firm, state, and industry-year fixed effects and find that, following the adoption of this law, capital expenditures as a fraction of lagged book assets decrease by 6.5% relative to the sample mean.² This result holds after controlling for several firm-level characteristics and state-level factors, such as per capita gross domestic product (GDP), per capita GDP growth, and political leaning. We also document that, following the adoption of the good faith exception, investment rates are less sensitive to changes in investment opportunities. Overall, these results suggest that an increase in employment protection lowers corporate investment rates.

Our finding that investment rates decline following the adoption of the good faith exception implies that greater employment protection should constrain firm growth. However, by increasing the enforceability of job contracts, employment protection promotes innovative activities by solving contractual holdup problems between employees and employers (Acharya, Baghai, and Subramanian 2014; Griffith and Macartney 2014). For example, Acharya, Baghai, and Subramanian (2014) show that the adoption of this law leads to firms receiving more patent grants and citations, which should accelerate firm growth. Consequently, these competing effects make the overall effect of employment protection on firm growth unclear. We test the net effect of employment protection on sales growth and find that firms headquartered in states that adopt the good faith exception grow sales at a rate that is 3.1 to 3.8 percentage points slower.

To interpret our results as the causal effect of employment protection on investment rates and growth, our experiment must satisfy the assumption that, in the absence of treatment, the average change in capital expenditures and sales growth would have been the same for both treatment and control firms. We show that our results are robust to accounting for a number of econometric concerns that could threaten this parallel trends assumption.

First, the adoption of the good faith exception and shrinking investment and sales growth rates could be spuriously correlated with underlying economic factors. While all our results hold to controlling for per capita GDP, per capita GDP growth, and political leaning, this set of controls is not exhaustive. Therefore, we also show that our results are robust to controlling for several other economic factors, including unemployment rates, unionization rates, the presence of other labor laws, the number of good faith cases filed in state courts, and the enactment of labor laws in other states in the same federal circuit region.

To help further address this concern of omitted correlated variables, we estimate triple-difference regressions that include state-year fixed effects to control for any variables that vary at the state-year level. To the extent that state-level omitted variables are uncorrelated with the interaction term, these results should be unaffected by omitted state-level factors. Specifically, we test whether the recognition of the good faith exception has a larger effect on firms that operate in industries with more volatile cash flows as well as firms that operate in a single industry. Firms in more volatile industries are more likely to need to adjust employment in response to cash flow fluctuations (e.g., Cuñat and Melitz 2012). Similarly, firms that operate in multiple industry segments can use internal labor markets to shift workers from one segment to another and avoid having to discharge workers if one industry segment experiences a temporary negative shock (e.g., Tate and Yang 2015). Thus, the adoption of this law should have a larger effect on the investment and sales growth rates of firms operating in industries with more volatile cash flows and for single-segment firms. Consistent with our prediction, we find that the decline in investment and sales growth following the law’s adoption is more pronounced for firms operating in industries with more volatile cash flows and that the decline in investment is larger for single-segment firms.

Second, it is problematic if lobbying activities influence courts’ decisions to recognize these laws. In our setting, however, this concern is not likely a large problem because the recognition of WDLs is based on judicial rather than legislative decisions and is therefore more likely driven by the merits of the case than political economy considerations (Autor 2003; Acharya, Baghai, and Subramanian 2014). Additionally, we find that changes in investment and sales growth rates appear only after and not before the adoption of the good faith exception, which further alleviates this concern and provides evidence against reverse causality.

The last concern is that treatment and control firms differ along dimensions that affect investment and growth. We address this issue by showing that our results are robust to using a matched sample. We match treatment firms to control firms in bordering states at the time of the treatment based on firm- and state-level characteristics using propensity score matching and find a decrease in investment, but not sales growth, following the adoption of the good faith exception. Overall, our results are robust to addressing a number of econometric issues. However, we cannot completely rule out the possibility that some underlying source of endogeneity clouds our inferences.

Our findings are consistent with greater employment protection reducing investment rates by making investments more irreversible and/or by crowding out firms’ access to capital. We try to disentangle these nonmutually exclusive mechanisms by testing two predictions. First, if employment protection makes it more costly and difficult to scale back poorly performing projects, greater employment protection should reduce the sensitivity of downsizing to negative cash flow shocks. Second, if greater employment protection constrains firms’ access to external capital, the decrease in investment following the adoption of the good faith exception should be greater for firms that depend more on external capital as well as for financially constrained firms. We find that, after the adoption of this law, firms are less likely to divest assets and discharge workers following decreases in industry cash flows. Moreover, when firms do downsize, they are more likely to scale back employment and sell assets in states that have not adopted this law. We also document some evidence that firms dependent on external finance cut investment more following the law’s adoption. Further, we find conflicting evidence of whether financial constraints affect the relation between capital expenditures and the law’s adoption based on how we measure financial constraints. In sum, we conclude that these results provide strong support for the investment irreversibility channel but mixed evidence supporting the financing channel, and therefore, we cannot conclude that the financing channel does or does not drive our findings.

Finally, we examine how greater employment protection affects other types of investment, including research and development (R&D), advertising, and acquisition expenditures. We find that R&D expenditures increase and acquisition expenditures decrease following the adoption of the good faith exception, which is consistent with the findings in Acharya, Baghai, and Subramanian (2014) and Chatt, Gustafson, and Welker (2017). However, total investment defined as the sum of capital, R&D, advertising, and acquisition expenditures decreases following the law’s adoption. We then examine the effect of the law’s adoption on different types of investment conditional on whether the firm engages in R&D or not. We document that, while investment rates decline on average (besides R&D), the reduction in investment rates following the adoption of the law appears concentrated in firms that do not engage in R&D. These results therefore can help reconcile why we find that greater employment protection reduces investment rates on average, while Acharya, Baghai, and Subramanian (2014) find an increase in patenting activity.

Overall, our study contributes to corporate finance research on the determinants of firms’ investment decisions. A predominant stream of research studies how financial frictions affect investment outcomes (e.g., Fazzari, Hubbard, and Petersen 1988; Whited 1992; Kaplan and Zingales 1997; Rauh 2006; Almeida and Campello 2007; Gan 2007; Duchin, Ozbas, and Sensoy 2010; Campello et al. 2011, among others). Our study adds to this work by providing evidence that labor market frictions in the form of employment protection affect real outcomes, such as capital allocation and growth.

We also contribute to work showing that employment laws can have a substantial effect on various corporate policies and valuation (e.g., Matsa 2010; Garmaise 2011; Acharya, Baghai, and Subramanian 2014; Griffith and Macartney 2014; Simintzi, Vig, and Volpin 2015; Serfling 2016; Loderer, Waelchli, and Zeller 2017). By documenting that employment protection laws shape real investment decisions, our study provides further support for the notion that government intervention outside of corporate law can influence the way managers run their firms.

While our paper is not the first to study the effect of employment protection on capital expenditures, it is the first to show that U.S. employment protection laws lead to U.S. firms reducing investment rates, resulting in subsequent declines in firm growth. In a related paper, Autor, Kerr, and Kugler (2007) use plant-level U.S. Census data and document an increase in the dollar amount of capital expenditures after the adoption of the good faith exception. In our paper, we use the |$q$|-theory model of investment as a benchmark model for our empirical analyses, which shows that research should focus on explaining investment rates rather than the dollar amount of investment. In this theory, a firm chooses a level of investment to maximize firm value subject to a capital accumulation equation. The underlying assumptions are that the firm produces with constant returns to scale in perfectly competitive markets, there are capital adjustment costs, and the adjustment cost function is linear homogenous in investment and capital (e.g., Summers et al. 1981; Hayashi 1982). With these assumptions, the derived optimal investment rule is that the investment rate is a function of only |$q$| and that this transformation is done so that the growth rate of the capital stock is independent of the scale of the firm.³ With this benchmark model, our results can be interpreted in a similar fashion as work that explores whether capital market frictions matter for investment decisions (e.g., Fazzari, Hubbard, and Petersen 1988). Specifically, our result that investment rates decrease after the adoption of the good faith exception after holding |$q$| constant suggests that labor market frictions matter for investment decisions.⁴

Our results are consistent with Calcagnini, Giombini, and Saltari (2009) and Calcagnini, Ferrando, and Giombini (2014), who find a negative relation between the capital expenditures of firms in a number of European countries and an index of country-level employment protection legislation. However, by studying the effect of an individual employment law in only the United States, our empirical setting has two advantages. First, our setting allows us to employ a relatively homogeneous sample in terms of financial and economic development, infrastructure, and legal structure, which reduces concerns that omitted economic factors drive our results. Second, the United States has relatively less stringent employment protection laws than European countries. Consequently, even with an upward trend in the degree of employment protection in the United States (Boxold 2008; Haider and Plancich 2012), less research has been done on the effects of U.S. labor protection laws. Thus, our setting provides insights on how U.S. employment protection laws can shape corporate investment and growth.

Last, our findings provide insights for policy makers regarding the net societal effect of increasing employment protection. Undoubtedly, employment protection laws benefit workers by protecting them from unexpected or unfair dismissal practices despite adequate performance. Our findings suggest that these benefits should be considered in conjunction with the potential costs of slower economic growth resulting from reduced investment rates.

1. Theoretical Motivation and Institutional Background

1.1 Theoretical motivation

Why would greater employment protection affect investment decisions? On one hand, greater employment protection could lead to more investment by either increasing productivity or encouraging investment in relatively cheaper physical capital. Nickell and Layard (1999) and Belot, Boone, and van Ours (2007) suggest that greater employment protection could increase productivity by mitigating holdup problems between firms and employees. In the model of Belot, Boone, and van Ours (2007), workers can invest in noncontractible firm-specific knowledge. Because workers bear the entire cost of exerting effort in learning firm-specific knowledge but only receive a fraction of the gains from such effort, a holdup problem is created and can lead to underinvestment in firm-specific knowledge. Thus, employment protection can encourage workers to accumulate firm-specific human capital by reducing the likelihood of dismissal, which could lead to greater productivity and investment rates.

Greater employment protection can also increase investment rates through a substitution effect in which firms substitute human capital with relatively less expensive physical capital. Employment protection can act as a transfer of benefits from employers to employees that is equivalent to mandated employee benefits. Under the Coase principle, if labor markets are perfect, wages fall to cover the cost of the benefit without productivity or employment consequences (Coase 1960). However, labor markets are not frictionless, and in general, employment protection likely increases firms’ labor expenses.⁵ For example, some investment decisions are made after workers have been located and hired, making it costlier to replace workers when employment protection is greater. As such, workers can generally try to bargain for higher wages. In the long-run, firms will substitute relatively more expensive labor-intensive assets with more capital-intensive assets, which could lead to higher investment rates (e.g., Blanchard 1997; Caballero and Hammour 1998; Koeniger and Leonardi 2007; Cingano et al. 2010).

On the other hand, greater employment protection could lead to less investment. By making it costlier to discharge workers, greater employment protection potentially makes investments more irreversible, and ample evidence shows a negative relation between investment irreversibility and investment activity (e.g., Bernanke 1983; Pindyck 1991; Bertola and Caballero 1994; Abel and Eberly 1996; Abel et al. 1996). Investments are more irreversible if, once undertaken, they cannot be undone or made into a different project without high costs (Bernanke 1983). In an option-based approach to calculate the net present value (NPV) of an investment, the ability to divest a project at a later point in time creates a put option that increases the NPV of the project, and the value of this put option is increasing in the resale or recovery price from divesting. Consequently, greater investment irreversibility lowers the value of the put option, reducing the NPV of the project and ex ante incentives to invest (Abel et al. 1996).

Greater employment protection can also lead to lower investment by constraining firms’ access to external capital. Specifically, greater employment protection makes it more difficult to reduce employment when firms need to do so, such as during economic downturns (e.g., Bentolila and Bertola 1990; Autor, Donohue, and Schwab 2006; Messina and Vallanti 2007). This effect makes a firm’s labor costs more fixed in nature, raising operating leverage and crowding out financial leverage (Simintzi, Vig, and Volpin 2015; Serfling 2016). Market frictions may make it costlier to replace this financing with equity. If higher operating leverage increases external financing costs, firms may decrease investment as a result of constrained access to external capital (Fazzari, Hubbard, and Petersen 1988).

1.2 Institutional background on wrongful discharge laws

When firms dismiss workers, they can incur substantial firing costs, which are any costs associated with discharging or firing employees. These costs include, but are not limited to, legal fees and settlements associated with lawsuits arising from violations of employment protection laws. Under the traditional employment “at-will” rule in the United States, employers are free to terminate any employee without warning and for any reason without the risk of legal liability. However, in an attempt to protect employees from unfair dismissal practices, legislation and common laws adopted over the last half-century have created a legal environment that allows employees to sue employers for wrongful termination. This shift in the legal environment has resulted in an increase in dismissal-related lawsuits and the costs associated with discharging workers, with nearly half of surveyed public firms expressing concerns regarding financial losses arising from such lawsuits.⁶

An important part of this shift in the legal environment came as many state courts, beginning largely in the 1970s, recognized exceptions to the terminate at-will rule. These common law exceptions, typically known as wrongful discharge laws, pertain to workers not already covered by explicit contractual agreements or by federal legislation aimed at protecting a particular group of workers, such as union members, racial minorities, women, and the aged (Miles 2000). These common laws evolved into three exceptions called the good faith, implied contract, and public policy exceptions. State courts can choose to adopt none, any, or all three of these exceptions.

The good faith exception requires that employers treat workers in a fair manner. In its broadest sense, this law protects employees from termination for any reason other than for a “just cause.” The implied contract exception protects employees from termination when the employer has implicitly promised the employee not to discharge the worker without good cause. Finally, the public policy exception protects employees from termination for refusing to violate an established public policy or commit an illegal act.⁷

Of the three exceptions, the good faith exception is arguably the most far reaching, as it represents the largest deviation from at-will employment (e.g., Dertouzos and Karoly 1992; Kugler and Saint-Paul 2004). This law should therefore have the largest effect on firm outcomes. The law applies in cases when a court determines that an employer has discharged a worker out of bad faith, malice, or retaliation. The law also serves to prevent employers from denying employees their contractual rights. For example, if an employer fires a salesperson just before a commission is due to deprive the employee of his commission or discharges an employee just before her pension vests, the employee could sue the employer under the good faith exception.

Employees have both a contract and tort cause of action under the good faith exception. This means that employees can recover compensation for punitive damages and emotional distress in addition to contractual losses. Importantly, punitive damages tend to be a large percentage of settlement awards and can substantially increase an employer’s liability. Further, with punitive damages, firms face more uncertain settlement amounts because a jury determines these damage awards without a clear formula.

Prior work suggests that the public policy and implied contract exceptions may not have material effects on firms (e.g., Miles 2000; Autor, Kerr, and Kugler 2007). Courts typically limit recovery under the public policy exception to dismissals in which the employer violated or encouraged the violation of an identifiable statute or constitutional provision. Firms can also largely prevent lawsuits under the implied contract exception by including disclaimers in their employee handbooks and personnel manuals that employment contracts are at-will.

1.3 Wrongful discharge laws in practice

An important assumption for our study is that, following the adoption of WDLs, employment protection and expected firing costs increase due to an increase in expected litigation. Consistent with this assumption, prior work finds that the adoption of the good faith exception results in decreases in employment levels, employment volatility, and firm entry (Dertouzos and Karoly 1992; Autor, Kerr, and Kugler 2007). This assumption, however, raises three questions worth discussing.

First, how large is the financial impact of wrongful termination lawsuits on firms? While it is difficult to estimate the true financial impact of wrongful termination cases because many court decisions are never published and are often settled before trial, several studies and anecdotes suggest that these laws can impose substantial costs.⁸ For example, Dertouzos, Holland, and Ebener (1988) analyze jury trials of wrongful discharge claims in California between 1980 and 1986 and find that plaintiffs won in 68% of cases and that the average award was |${\$}$|0.656 million. Similarly, Jung (1997) estimates that plaintiffs prevailed in 46.5% of wrongful dismissal cases that reached the trial stage in 1996 and won |${\$}$|1.29 million on average. Further, Boxold (2008) documents average and maximum awards of |${\$}$|0.59 million and |${\$}$|5.4 million, respectively, over the 2001 to 2007 period.

While these average settlements are arguably small for large firms, the fear of very large settlements could alter the behaviors of risk-averse managers (Dertouzos, Holland, and Ebener 1988). Moreover, firms can have multiple lawsuits pending at any point in time, which can substantially raise their legal liability. For example, in a case against Lawrence National Security, 130 of 430 laid-off employees sued the firm. The firm claimed the layoffs were economically motivated, but the employees argued the layoffs were a “pretext to get rid of older employees who have higher salaries, larger medical costs, and are closer to collecting their pension.” A jury sided with five employees selected as test cases for the other 125 employees, awarding these five individuals a total of |${\$}$|2.7 million for breach of the good faith exception and breach of contract.⁹ Following this decision, Lawrence National Security ultimately settled with the remaining employees for a total settlement of |${\$}$|37 million.

Second, can firms be subject to wrongful termination claims if they dismiss workers as part of economically motivated layoffs, such as due to plant closings associated with poor performance? In the context of discrimination, which is similar to wrongful termination, evidence suggests that it is more difficult to prove unjust dismissal when many workers are laid off compared to when a single worker is fired (Donohue and Siegelman 1993; Oyer and Schaefer 2000). However, this point does not remove the potential for wrongful termination lawsuits during layoffs, as highlighted in Andrews et al. v. Lawrence National Security (above) and Robert Coelho v. Posi-Seal International, Inc.¹⁰ In this latter case, Coelho accepted the position of manager of quality control at Posi-Seal after upper-level managers reassured Coelho that he would not be dismissed because of ongoing conflicts between the quality control and manufacturing division (a factor that contributed to the departure of the prior quality control manager). Subsequently, Posi-Seal fired Coelho, claiming it was part of layoffs. Coelho sued Posi-Seal for breach of an implied promise of ongoing employment, claiming he was discharged because of a dispute with a manager of manufacturing.

Posi-Seal argued that termination due to a reduction in work force is, as a matter of law, a just cause. However, the court concluded, “an employer’s contention that some employees were terminated as a result of a legitimate reduction in force does not necessarily establish that all employees were discharged for the same reason.” Further, “an employer may not use a reduction in work force as a pretext to terminate other employees in violation of contractual obligations, public policy grounds, or statutory rights.” In this lawsuit, the court ruled in favor of Coelho because it was clear from the evidence that Posi-Seal used the layoffs as pretext to fire Coelho.

Thus, while layoffs may reduce the risk of wrongful termination lawsuits, they do not eliminate this risk. Further, the risk of wrongful termination lawsuits increases when employers harm employee welfare by concealing information about closings (Rhine 1986). For example, if an employer knows that a plant will close, any false assurances of job security or unreasonable delays in closure notifications could violate the good faith exception.¹¹

The third question is whether firms can completely offset the extent of losses related to employee lawsuits by purchasing Employment Practices Liability Insurance (EPLI). While EPLI can offset losses arising from wrongful termination claims, the impact of EPLI on our findings is likely limited for two reasons. First, firms’ use of EPLI was not widespread during our sample period of 1969 to 2003. The market for EPLI largely emerged in the early 1990s because of greater awareness of the potential for and costs of employee litigation as a result of the Civil Rights Act of 1991 and the Clarence Thomas hearings (Klenk 1999). Moreover, purchasing EPLI was still the exception in the later 1990s (22% of surveyed employers had EPLI in 1997) but has become more widespread in recent years (68% of surveyed employers carried EPLI in 2012).¹²

Second, even if a firm purchased EPLI in the later years in our sample, these early policies typically contained “intentional acts” exclusions that limited coverage for wrongful termination claims. These policies also often had exclusions related to downsizing and retaliation and typically did not cover punitive damages, which can be a substantial amount of the total settlement award. Nevertheless, to the extent that EPLI coverage reduces the risks and costs associated with employee litigation, the presence of EPLI should only reduce the effect of the recognition of the good faith exception on corporate investment and growth.

1.4 Adoption of wrongful discharge laws by state courts

Our analyses rely on identifying which court cases set the precedent that a state has adopted a particular WDL. We base our identification of the recognition of WDLs primarily on the precedent-setting cases provided in Autor, Donohue, and Schwab (2006).¹³ In contrast to Autor, Donohue, and Schwab (2006), we follow Walsh and Schwarz (1996) and Littler (2009) and recognize Utah as adopting the good faith exception since 1989.¹⁴

Table 1 summarizes the dates when each state court ruled on a precedent-setting case for each particular exception, and Figure 1 shows the number of states that have adopted each exception in each year between 1969 and 2003. Besides California adopting the public policy exception in 1959, all states adopted WDLs between the 1970s and 1990s. Of the fourteen states that eventually recognized the good faith exception, the majority of the states recognized the law in the 1980s. A few states also reversed their positions on enforcing the good faith (New Hampshire in 1980 and Oklahoma in 1989) and implied contract exceptions (Arizona in 1984 and Missouri in 1988).

Figure 1

Number of states adopting wrongful discharge laws

This figure shows the number of states that have adopted the good faith, implied contract, and public policy exceptions to the traditional employment at-will rule in each year between 1969 and 2003.

Open in new tabDownload slide

2. Research Design and Sample Selection

2.1 Empirical methodology

All of our models also include firm age fixed effects (AgeFE|$_{j})$|⁠, firm fixed effects |$(\upsilon_{i})$|⁠, state fixed effects (⁠|$\delta_{s}$|⁠), and industry-year fixed effects defined at the 3-digit SIC level (⁠|$\eta_{k}\times \omega_{t}$|⁠). The firm age fixed effects control for potential life cycle effects in which investment and growth slow as the firm matures. The firm fixed effects control for time-invariant omitted firm characteristics and ensure that estimates of |$\alpha_{1}$| reflect average, within-firm changes in investment and sales growth rates over time rather than simple cross-sectional correlations. The state fixed effects control for time-invariant omitted state-level factors. Last, the industry-year fixed effects account for transitory industry- and nationwide factors, such as industry deregulation and macroeconomic conditions that could simultaneously affect corporate investment activity and growth as well as the likelihood that a state adopts the good faith exception. An advantage of our experiment is that different states adopt this law at different times, which allows a firm headquartered in a given state to be in both the treatment and the control group. Thus, the staggered adoption of this law implies that the control group is not restricted to only those firms headquartered in states that never recognize the law.

We correct estimated standard errors in all regressions for clustering at the state level. The adoption of the good faith exception varies at the state level, so this clustering method accounts for the concern that residuals are serially correlated within a firm and also correlated across firms within the same state (Bertrand, Duflo, and Mullainathan 2004).

Equation (2) is the same as Equation (1), except that we interact the good faith dummy variable (GF|$_{s,t})$| with a variable indicating instances when firms face a larger increase in firing costs (CS|$_{i,s,t-1})$| and also include state-year fixed effects (⁠|$\delta_{s} \times \omega _{t}$|⁠). By identifying a group of firms within treated states that are more likely affected by the adoption of this law and effectively comparing groups of firms within the same state, the triple-difference estimator can alleviate the concern that unobserved factors affect firms headquartered in states that do and do not adopt the good faith exception differently. Further, the inclusion of state-year fixed effects removes all time-varying omitted variables that affect all firms within the same state during a given year by demeaning all variables by state each year. Therefore, to the extent that state-level omitted variables are uncorrelated with the variables we use to identify instances when firms face a larger increase in firing costs, this set of results should be unaffected by omitted state-level factors. We note that the state-year fixed effects do not completely absorb the variables GF|$_{s,t}$|⁠, IC|$_{s,t}$|⁠, and PP|$_{s,t}$| because these variables are defined using calendar month-ends, which do not directly correspond to fiscal year-ends. Thus, the estimated coefficients on GF|$_{s,t},$|IC|$_{s,t}$|⁠, and PP|$_{s,t}$| are not informative in models with state-year fixed effects.

Employment laws typically apply to the state where an employee is working. Compustat, however, provides only a firm’s state of incorporation and headquarters. Consequently, we follow recent studies and match these laws to the state where each firm is headquartered (e.g., Matsa 2010; Agrawal and Matsa 2013; Acharya, Baghai, and Subramanian 2014; Dougal, Parsons, and Titman 2015), which is also typically where major plants and operations are located (Henderson and Ono 2008). Further, Dertouzos, Holland, and Ebener (1988) find that plaintiffs in wrongful termination cases tend to hold executive or managerial positions (53%), who tend to be concentrated at headquarters. Thus, using the headquarters state likely captures a large portion of the increase in employment protection.

A limitation of Compustat is that it provides only the latest headquarters locations. Therefore, we supplement Compustat headquarters data in two ways. First, for firm-years that are after the availability of machine-readable SEC filings (beginning in 1994), we extract the actual state of headquarters from WRDS SEC Analytics Suite. For firm-years prior to the availability of machine-readable SEC filings, we hand-collect the historical headquarters locations from the Moody’s Manuals (later Mergent Manuals) and Dun & Bradstreet’s Million Dollar Directory (later bought by Mergent).¹⁶ In Section 3.3.5, we show that our results are robust to using alternative data sources to obtain a firm’s headquarters locations.

2.2 Compustat sample selection

We use CRSP/Compustat Merged data for firms headquartered in the United States that have nonmissing data for our main variables of interest over the years 1969 to 2003. The sample period starts 5 years before the earliest enactment of the good faith exception by New Hampshire in 1974 and ends 5 years after the last event when Louisiana adopted the good faith exception in 1998. We obtain data on state-level GDP from the U.S. Bureau of Economic Analysis and data on state legislator party affiliations from the Book of the States. For our tests, we exclude utility firms (SIC codes 4900–4999), financial firms (SIC codes 6000–6999), and quasi-public firms (SIC codes greater than 9900). We further require that firms have at least 2 years of data to estimate the firm fixed effects and that 3-digit SIC industries have at least two observations in a given year to estimate the industry-year fixed effects.

These restrictions result in a sample size of 115,432 firm-years for our main analyses. We winsorize continuous variables, except state-level economic variables, at their 1st and 99th percentiles and express dollar values in 2009 dollars. Panel A of Table 2 presents detailed definitions and summary statistics for the variables in our tests. Panel B compares year |$t$|-1 variable means for firms headquartered in states that adopt the good faith exception in year |$t$| to firms headquartered in states that do not adopt this law in year |$t$|⁠. Ideally, treatment and control firms would be similar along each dimension. However, because they are not, we control for each variable in our regressions to account for these differences. In Section 3.3.3, we further address this issue by using a propensity-score-matched sample.

2.3 Census sample selection

We also utilize several databases from the U.S. Census Bureau. Using these data allows us to examine the robustness of our main findings based on firm-level data to using plant-level data. Further, using plant-level data allows us to test the underlying mechanism through which employment protection affects firms’ investment decisions. Specifically, we employ the Longitudinal Business Database (LBD), the Annual Survey of Manufacturers (ASM), and the Census of Manufacturers (CMF). The LBD is compiled from the business register and provides annual coverage on employment for the universe of all business establishments with at least one paid employee in the United States. The ASM provides comprehensive coverage for the universe of manufacturing plants (SIC 2000–3999) in the United States with more than 250 employees, but it maintains a randomly selected rotating panel for the smaller plants. In Census years, the ASM is replaced by the CMF, which surveys the universe of all manufacturing plants in the United States regardless of their size.

3. Empirical Results

3.1 Employment protection and corporate investment rates

We first investigate whether an increase in employment protection arising from the adoption of the good faith exception affects corporate investment, measured as capital expenditures scaled by beginning of year book assets. Panel A of Table 3 reports the results from this analysis. Column 1 includes the indicator variables for whether the state where a firm is headquartered recognizes the good faith, implied contract, and public policy exceptions as well as firm age, firm, state, and industry-year fixed effects. The results show a negative and statistically significant relation between investment rates and only the recognition of the good faith exception. In terms of economic significance, the coefficient estimate implies that firms reduce capital expenditures by 0.54 percentage points following the adoption of the law. Given that the sample mean of the ratio of capital expenditures to book assets is 8.33%, this finding represents a relative reduction in investment rates of 6.5% (⁠|$=0.54/8.33$|⁠).

Column 2 further controls for firm size, investment opportunities, and cash flow. Column 3 also controls for cash holdings and financial leverage. The results remain similar to those in Column 1; investment rates decrease by 0.52 to 0.55 percentage points following the adoption of the good faith exception. Last, Column 4 is our primary model specification and further controls for per capita GDP growth, per capita GDP, and political balance. The inclusion of these state-level economic variables strengthens both the statistical and economic significance of the effect of the recognition of this law on capital expenditures. The coefficient estimate of |$-$|0.66 on the good faith indicator variable implies that investment rates decline by 7.9% (⁠|$=0.66/8.33$|⁠) relative to the sample mean following the adoption of this law.

Next, we test how the adoption of the good faith exception affects the responsiveness of capital expenditures to changes in investment opportunities and present the results in panel B of Table 3. If greater employment protection makes it so that firms do not invest in all profitable projects due to the potential cost of firing workers if the project turns out poorly, we would expect firms to respond less to changes in investment opportunities. To test this prediction, we use our main investment regressions and interact the good faith indicator variable with two measures of investment opportunities—the firm’s beginning of year Tobin’s q and prior year’s sales growth rate. The results show that an increase in investment opportunities is associated with an increase in capital expenditures. Following the adoption of the good faith exception, however, this sensitivity of investment rates to investment opportunities significantly shrinks. Further, these findings are robust to controlling for state-year fixed effects.

3.2 Employment protection and sales growth

The evidence presented so far implies that increasing employment protection discourages investment activity. This lower rate of investment could limit the firm’s ability to grow. However, prior work finds that restrictions to employment at-will can limit a firm’s ability to take advantage of employees that contributed significant effort to successful innovation and in turn encourage employees to engage in more innovative activities. Importantly, an increase in innovation due to an increase in employment protection could suggest that greater employment protection accelerates firm growth. Therefore, the opposing effects of employment protection on capital expenditures and innovation make its overall effect on growth ambiguous. In this section, we first confirm for our sample of firms the finding that increased employment protection leads to greater innovation (Acharya, Baghai, and Subramanian 2014) and then test the effect of employment protection on sales growth.

We confirm the innovation results by regressing the natural logarithm of one plus the number of patents a firm files (number of citations a firm’s patents receive in subsequent years) on the good faith indicator variable and the same set of control variables used in our main investment tests. The results in the Online Appendix Table A1 show that, following the adoption of the good faith exception, the number of patents a firm files increases by approximately 8.7% and the patents receive about 10.3% more citations.

Next, Table 4 presents the results of the analysis examining the relation between a firm’s 1-year sales growth rate and the recognition of the good faith exception. We use the same set of control variables as in our investment regressions. Similar to Table 3, we sequentially add firm- and state-level control variables to the model specifications. Across all of the models, we find that firms’ sales growth is 3.07 to 3.80 percentage points slower after the enactment of this law.

While these magnitudes are large relative to the sample mean sales growth rate of 20.5% (relative decreases of 15.0%–18.5%), these declines represent more reasonable decreases of 4.8% to 6.0% of its standard deviation. These magnitudes are comparable to the decrease in investment rates of 5.5% to 6.8% of its standard deviation. We can also get a sense of how much sales growth is affected by the reduction in capital expenditures found in Table 3 by regressing the sales growth rate on capital expenditures scaled by beginning of year book assets and firm, state, and 3-digit SIC-year fixed effects. The coefficient estimate on capital expenditures is 1.63 (untabulated). Given that we find capital expenditures decline by 0.66 percentage points following the law’s adoption in Column 4 in panel A of Table 3, these estimates suggest that sales growth should decline by about 1.08 (⁠|$=0.66\times 1.63$|⁠) percentage points following the law’s adoption. Thus, the decrease in capital expenditures explains about 28% (⁠|$=1.08/3.8$|⁠) of the decrease in sales growth.¹⁷|$^{,}$|¹⁸

3.3 Econometric concerns

Two possible sources of endogeneity could affect the interpretation of our results. In particular, lobbying activities could influence the adoption of the good faith exception, and/or an omitted economic characteristic could be correlated with both the enactment of this law and changes in investment and sales growth rates. As a first step in determining the extent to which the recognition of this law is exogenous, we examine the institutional details around its adoption. Because WDLs are common laws, the enactment of the good faith exception is based on judicial rather than legislative decisions, which are more likely driven by the merits of the case than political economy considerations (Autor 2003; Acharya, Baghai, and Subramanian 2014). Therefore, lobbying activities should not be a major problem for our analyses. In addition, Serfling (2016) finds that firms experience cumulative abnormal stock returns of |$-$|1.05% to |$-$|1.22% when their state adopts this law and that there is no price run-up before the law is adopted, suggesting that the law’s adoption is not only costly but also partially unanticipated.

Further, Walsh and Schwarz (1996) analyze published court decisions to investigate cited reasons for judges adopting the good faith exception and find that these reasons include (1) assuring consistency with established principles of contract law, (2) enhancing fairness in employment relationships, and (3) following other states that have already adopted WDLs.¹⁹ These institutional features imply that lobbying activities and underlying economic factors related to investment decisions and growth were not major determinants of courts’ decisions to adopt the good faith exception and therefore provide initial evidence that the recognition of the law likely represents an exogenous shock. Nevertheless, in the following sections, we conduct various empirical analyses to alleviate residual endogeneity concerns.

3.3.1 Effect of potential omitted variables

While survey evidence suggests that judges’ rationales for adopting the good faith exception are unrelated to factors that affect firms’ investment decisions and sales growth, it is possible that judges are directly or indirectly motivated by economic factors that they do not cite. Thus, we next explore whether several state-level variables that have been hypothesized to affect a court’s decision to adopt WDLs are correlated with the adoption of the good faith exception. For example, courts may be more likely to adopt WDLs when the unemployment rate in the state is higher, because a larger fraction of workers could benefit from employment protection (Dertouzos and Karoly 1992). Also, a state’s decision to adopt WDLs may be influenced by whether states in the same federal circuit court region have already adopted these laws (Bird and Smythe 2008).

To test whether any of these factors predict the adoption of the good faith exception, we follow Acharya, Baghai, and Subramanian (2014) and estimate a Cox proportional hazard model with year fixed effects, where a failure event represents the adoption of this law. Table A5 in the Online Appendix presents these results. All predictor variables are measured as of year |$t$|-1 relative to the law’s adoption. The sample spans the years 1969 to 2003, and states are excluded from the sample once they adopt the good faith exception.

The set of economic predictors include growth in and the level of per capita GDP, political balance, whether the state has adopted the implied contract and public policy exceptions, the level of and change in the unemployment rate, whether the state has passed right-to-work laws, the level of and change in the union membership rate, the fraction of states in the same federal circuit court region that have already adopted the good faith, implied contract, and public policy exceptions, and the number of state court cases filed related to the good faith exception. In Table A6, we then show that our main finding that investment and sales growth rates decrease after the adoption of the good faith exception is robust to controlling for all economic variables that significantly predict its adoption (Columns 1 and 3). Columns 2 and 4 further show that the results are robust to including the full set of economic variables regardless of whether a variable is a significant predictor of the adoption of the law. Overall, these results imply that our findings are robust to controlling for a number of potential omitted variables related to local economic conditions.

To further help address the concern of omitted correlated variables, we next estimate triple-difference regression models by testing whether the recognition of the good faith exception has a larger effect on firms that operate in industries with more volatile cash flows as well as firms that operate in a single industry. Because firms that operate in industries with more volatile cash flows are more likely to need to adjust employment in response to cash flow fluctuations (e.g., Cuñat and Melitz 2012), firms in these industries should expect a larger increase in firing costs. Similarly, firms that operate in multiple industry segments might be able to use their internal labor markets to shift their workers from one segment to another and potentially avoid having to discharge workers if one industry segment experiences a temporary negative shock (e.g., Tate and Yang 2015). Thus, compared to multisegment firms, single-segment firms should expect a larger increase in firing costs. Therefore, the negative relation between the adoption of the law and investment and sales growth rates should be stronger for firms operating in industries with more volatile cash flows as well as for firms operating in a single industry segment.

Table 5 presents the results from tests examining the effect of industry cash flow volatility and industry focus on the relation between the adoption of the good faith exception and investment rates and sales growth. We measure industry cash flow volatility as the average cash flow volatility across all firms in the same 3-digit SIC industry and year. A firm’s cash flow volatility is the standard deviation of the ratio of income before extraordinary items plus depreciation and amortization to book assets over years |$t$|-10 to |$t$|-1. We obtain data on the number of segments a firm operates in from the Compustat segments files. We focus on the number of unique 3-digit SIC industry segments the firm operates in and therefore exclude geographic segments from the calculation. Segment data are available beginning in 1977, but because we use the number of segments the firm operates in during year |$t$|-1 for our analysis, the sample for this test spans the years 1978 to 2003. For this sample, 68.8% of firms operate in a single industry segment.