Do Non-Financial Stakeholders Affect the Pricing of Risky Debt?

Evidence from Unionized Workers*

Huafeng (Jason) Chen University of British Columbia

Marcin Kacperczyk

New York University

Hernán Ortiz-Molina University of British Columbia

Forthcoming at the Review of Finance

Abstract

We study the impact of a powerful non-financial stakeholder – unionized workers – on

the pricing of corporate debt. Firms in more unionized industries have lower bond

yields. This relation is stronger in firms with weaker financial conditions and cannot be

explained by the correlation of unionization with industry characteristics, governance

mechanisms, or financial leverage. Firms in unionized industries implement less risky

investment policies, and are less likely targets of acquisitions. Unionization reduces yields

by more when firms’ takeover barriers are lower. Hence, unions are viewed favorably in

the bond market because, through their influence on corporate affairs, they protect

bondholders’ wealth.

Keywords: debt pricing, non-financial stakeholders, labor unions, shareholder-bondholder conflicts JEL Classifications: G32, G34

* We especially thank Marco Pagano (the editor) and an anonymous referee for several constructive suggestions. We also thank Viral Acharya, Amr Addas (NFA discussant), Ashwini Agrawal, Thomas Bates, Bo Becker, Effi Benmelech, Matthew Billett, Sandra Chamberlain, Alex Edmans, Larry Fauver, Ron Giammarino, Jayant Kale, Sandy Klasa, Ugur Lel, Kai Li, Angie Low, David Matsa, William Maxwell, Vikas Mehrotra, Husayn Shahrur, Krishnamurthy Subramanian, Amir Sufi (AFA discussant), Frank Yu, and seminar participants at the University of Arizona, University of British Columbia, Georgia State University, the American Finance Association Meetings, the UBC Summer Finance Conference, and the Northern Finance Association Meetings. We acknowledge the financial support from the Social Sciences and Humanities Research Council of Canada.

1. Introduction

There has long been an interest in understanding how non-financial stakeholders, especially

labor unions, affect firms’ decisions and ultimately their performance. The literature on labor unions

generally concludes that unions reduce firms’ values (e.g., Ruback and Zimmerman, 1984; Abowd,

1989; Hirsch, 1991; Lee and Mas, 2009). This evidence, however, is solely based on equity values

and thus the above conclusion does not explicitly consider any effect unions may have on debt

values.

In this paper, we focus on unionized workers to address the question of how powerful non-

financial stakeholders, through their influence on corporate matters, affect debt values. Our

conceptual framework is based on the view that non-financial stakeholders can influence firms’

decisions, for example by threatening to withdraw their contributions to the firm, and thus they can

play a prominent role in firms’ internal governance structures. This view is supported by Jensen and

Meckling (1976), who emphasize that firms’ decisions are the outcome of a bargaining process

among all stakeholders. More recently, Acharya et al. (2009) argue that to better understand the

source of agency problems, one has to study a firm as composed of diverse agents with different

horizons, opportunities, and interests, rather than as a monolithic single-employee entity.

Furthermore, the interaction between different types of stakeholders in the political arena may shape

country-level corporate governance systems (e.g., Pagano and Volpin, 2005a; Perotti and von

Thadden, 2006).

This stakeholder view of corporations naturally suggests that labor unions’ influence on

corporate affairs may affect lenders’ assessments of a firm’s default risk and thus the value of its

debt. But how unions’ influence impacts debt values is a priori unclear. On one hand, workers’

interests are often closely aligned with those of bondholders: In solvent states, both workers and

bondholders receive payoffs that are largely insensitive to a firm’s performance; they also receive

lower payoffs in bankruptcy. Thus, both may favor corporate policies that reduce the likelihood of

1

default. Consequently, workers’ influence on corporate matters might benefit bondholders, thus

increasing debt values and lowering bond yields. On the other hand, once the company approaches

bankruptcy, workers become more concerned with a potential loss of their human capital invested in

the firm and their future income; hence, they may oppose an efficient liquidation that benefits

creditors. As a result, workers may align themselves with shareholders in trying to keep the firm alive

by undertaking activities capable of diluting creditors’ wealth. In this case, workers’ influence on

corporate matters may hurt bondholders and increase bond yields.

To resolve this issue, we use the industry labor force unionization rate (the fraction of workers

in a Census Industry Classification (CIC) industry that are unionized) as a proxy for the influence of

workers on corporate affairs and uncover new evidence that unions have a nontrivial impact on

bond yields. Our results suggest that unions are viewed favorably in the bond market because their

influence on corporate affairs serves as a mechanism that protects bondholders’ wealth. Firms in

more unionized industries have statistically lower bond yields, even after controlling for financial

leverage and several firm and debt characteristics that are known to affect yields. The results are also

economically significant: A one-standard-deviation increase in the industry unionization rate

decreases bond yields by about 32 basis points. In addition, we find that creditors view workers’

influence on corporate affairs more favorably when firms are in financial distress. Notably, this

finding holds even when we include CIC industry fixed effects in our regression model; hence,

omitted time-invariant characteristics correlated with unionization are unlikely to explain our results.

Subsequently, we entertain alternative explanations of our findings. First, labor unions are

considered an industry phenomenon and thus unionization may proxy for an industry characteristic,

such as the stage of the industry life cycle, its capital or knowledge intensity. Our empirical model

includes several firm- and industry-level characteristics that could be correlated with such industry

effects, including Tobin’s Q, cash-flow risk, industry concentration, and industry fixed effects. We

additionally include industry-level variables that directly proxy for the stage of the industry life cycle

2

(old- and new-economy status, R&D expenditure, advertising expenditure, growth, and profitability),

capital intensity (capital-labor ratio), and knowledge intensity (the fraction of industry workers with

college education and the average number of working hours per employee in the industry).

Second, the effect of unionization on bond yields may be driven by the correlation between

industry unionization rates and the strength of corporate governance mechanisms in place. Our

results are robust to the inclusion of common proxies for governance, such as the level and

concentration of institutional ownership and the governance index of Gompers et al. (2003).

Third, we examine whether the magnitude of our results is biased due to selection or

measurement error. The estimated magnitude of the unions’ effect might be biased upwards if

unions self-select to less risky industries; conversely, the magnitude might be biased downwards if

unions sort into riskier industries. Also, since we use industry-level unionization rates to proxy for

firm-level rates, another reason for such downward bias could be measurement error. Using the

fraction of female workers in a firm’s CIC industry as an instrument, we show that such bias in our

estimates is statistically insignificant. Another way to evaluate the impact of measurement error is to

conduct our tests at the industry level. We thus convert the bond yields and other firm-level

variables into industry-level averages. The evidence is similar to the firm-level evidence.

Next, we shed light on the mechanism through which unions might affect bond values. One

possibility is that unionized workers might curb shareholders’ incentives to shift risk onto

bondholders. In line with this conjecture, we find that firms in more unionized industries spend a

smaller fraction of their investment budgets on risky R&D than they do on physical assets.

Moreover, this effect is stronger in firms with weaker financial conditions, that is, in firms in which

shareholders have more incentives to take excessive risk.

Another possibility is that unions might deter takeovers, which are associated with changes in

firms’ contracting relationships that shift wealth from workers to shareholders (e.g., Pagano and

Volpin, 2005b). This activity may lower debt risk, since takeovers often entail increases in leverage

3

which reduce the value of existing debt that is not fully protected by covenants, and the managers of

acquisition targets often engage in defensive maneuvers that reduce debt values. Supporting this

view, we find that firms in more unionized industries are less likely to be acquisition targets, and that

unionization reduces bond yields by more when takeover barriers are lower.

Anecdotal evidence also suggests that unions may reduce bond yields by mobilizing the media

and politicians to the rescue of financially troubled firms. Last, in distressed firms unionized workers

may sometimes side with shareholders in an attempt to avoid liquidations that would benefit lenders.

Our evidence cannot rule out this possibility, but suggests that the net effect of unions’ influence on

corporate matters is a reduction in debt risk.

Our results can also be cast in a broader supply and demand framework in which unionization

affects both the equilibrium price and amount of debt. Our arguments suggest that, by helping

protect creditors’ wealth, an increase in unionization causes an increase in the supply of debt to the

firm. Supporting this view, we find that unionization significantly reduces bond yields. But our

further tests show that unionization does not affect firms’ financial leverage. Hence, an increase in

unionization is also accompanied by a decrease in the firms’ demand for debt. This could occur

because more unionized firms choose lower financial leverage to offset higher operating leverage

which results from unions’ opposition to wage cuts and layoffs (Simintzi et al., 2009), or because

unions induce managers to choose lower financial leverage to reduce workers’ exposure to

unemployment risk (Berk et al., 2010).

The rest of the article is structured as follows. Section 2 discusses our conceptual framework

and related literature. Section 3 describes the data and main variables. Section 4 presents our main

results. Section 5 sheds light on the mechanism through which unions might affect bond yields.

Section 6 presents several robustness tests. Section 7 concludes.

4

2. Conceptual Framework and Related Literature

2.1. CONCEPTUAL FRAMEWORK

The role of powerful workers in the pricing of risky debt is a priori uncertain and depends on

creditors’ perception of how workers’ future actions might affect the risk of default on debt

payments. On one hand, workers’ influence on corporate affairs may protect bondholders’ wealth

and thus reduce bond yields. Workers’ contractual claims on firm value resemble the payoff to risky

debt: In solvency, workers receive a largely fixed payoff, equivalent to the present value of all future

wages and benefits and do not gain much from improvements in firms’ performance; this payoff is

lower in bankruptcy, when firms are unable to fulfill their contractual obligations.1 Furthermore,

upon liquidation, workers may lose the value of their firm-specific human capital. Since both

workers and bondholders are largely fixed claimants, their interests are likely to be closely aligned in

supporting actions which decrease the probability that the firm will default on debt payments.2

Specifically, workers may reduce the risk of debt in three main ways. First, both workers and

lenders are adversely affected by shareholders’ incentives to engage in risky investment projects that

shift wealth from fixed claimants to themselves (Fama, 1990). Therefore, influential workers strive

for policies that support less risky investments. Second, workers may oppose takeovers, which are

often associated with changes in firms’ contracting relationships that shift wealth from workers to

shareholders (Shleifer and Summers, 1988; Pagano and Volpin, 2005b). A lower likelihood of a

takeover reduces debt risk, since takeovers often entail increases in leverage which reduce the value

of existing debt that is not fully protected by covenants (Warga and Welch, 1993). Also, managers of

1 In other words, workers have a fixed claim on firm value (the present value of all future wages and benefits) less a put

option with a strike price equal to the expected value of their claim in bankruptcy.

2 In a similar vein, Perotti and von Thadden (2006) study a political economy model in which lenders tend to side with

workers against shareholders in the political arena.

5

target firms often recapitalize the firms, refocus them, and increase payouts to shareholders, all of

which can reduce these firms’ debt values (Klock et. al., 2005).3 Third, in times of distress, organized

labor may mobilize the media and politicians to the rescue of firms in financial trouble: Firms with

stronger labor are more likely to be bailed out.

On the other hand, workers’ influence on corporate affairs may hurt bondholders’ wealth and

thus increase bond yields. In pricing debt claims, rational lenders anticipate that their interests could

conflict with those of workers in the event that the firm needs to be liquidated. In liquidation,

creditors recover an amount that is proportional to the firm’s liquidation value. Thus, on the verge

of bankruptcy, creditors would strive for corporate policies that maximize the liquidation value. In

contrast, irrespective of the liquidation value, in bankruptcy workers lose their firm-specific human

capital investments as well as the future income streams they would have received had the firm

remained solvent. Hence, workers may strive for policies that avoid liquidation rather than those

that maximize the liquidation value. It is then possible that in firms that are close to bankruptcy

workers would oppose an efficient liquidation that would benefit creditors.4 Workers would instead

align themselves with shareholders in undertaking risky activities to keep the firm alive.5

In light of the conflicting views, whether workers’ influence on corporate affairs implies higher

or lower bond yields remains an empirical issue. We follow previous work in labor economics (e.g.,

Connolly et al., 1986; Bronars and Deere, 1991) and use labor force unionization rates as a proxy for

the influence of workers on corporate decisions. The choice of unionization is well suited for this

3 Debt values fall in spinoffs (Maxwell and Rao, 2003), share repurchases (Maxwell and Stephens, 2003), takeovers and

acquisitions (Shastry, 1990; Billet et al., 2004), or due to changes in payout policy (Dhillon and Johnson, 1994).

4 However, workers could favor policies that maximize the liquidation value if they are covered by underfunded defined-

benefit pension plans (a form of unsecured debt). See Ippolito (1985) and Benmelech et al. (2010).

5 The idea that workers and managers are natural allies in some circumstances is stressed by Hellwig (2000) and

formalized by Pagano and Volpin (2005b) in the context of takeover threats.

6

context. First, organized labor is better able to coordinate workers’ actions. This improved

coordination leads to a more effective communication between workers and management and allows

unionized workers to use their market power to exert pressure on management (Freeman and

Medoff, 1984). Second, because unionized workers are locked into the firm with their firm-specific

human capital investments, they have strong incentives to monitor management’s actions to ensure

that the firm remains healthy. Moreover, they possess a special monitoring ability that stems from

their unique access to information regarding corporate policies, which in turn comes from their day-

to-day involvement with the firm (Schwab and Thomas, 1998).

In addition, ample evidence illustrates various channels through which labor unions,

representing a large fraction of workers, may affect corporate decisions. First, Gray et al. (1999)

document that about 47% of contemporaneous collective agreements contain some form of

“partnership” that provides unions with a voice in important firms’ decisions. Second, unions

engage in boycotts, picketing, strikes, and public denouncements when firms’ decisions counter their

interests. Third, unions increasingly resort to a wide array of tactics broadly termed as union

corporate campaigns, which range from consumer boycotts, lawsuits, and public relations schemes

to proxy contests. Fourth, in some companies, employees hold a non-negligible fraction of their

companies’ shares, either directly or through their pension plans, and thus they can use their voting

power to advance labor’s agenda (Faleye et al., 2006; Agrawal, 2008). Finally, a handful of large U.S.

firms have boards with union representations, which arise as a result of either employee stock

ownership plans or of collective bargaining agreements.

Our empirical strategy involves the following steps. We first determine the average effect of

labor unionization on bond yields. Since workers’ actions may have more impact on creditors’

wealth when firms are in financial distress, we next study whether the effect of workers on bond

yields differs across financially distressed and healthy firms. Finally, we shed light on the mechanism

through which unions affect bond values.

7

2.2. RELATED LITERATURE

Recent literature examines the market assessment of debt repayment risk through the lens of

agency theory. One strand of this literature focuses on how ownership structure affects the cost of

debt. Bagnani et al. (1994) document a positive relation between managerial ownership and bond

return premium in the 5 to 25 percent ownership range; Anderson et al. (2003) show that founding

family ownership reduces bond yields; Ortiz-Molina (2006) finds that, in pricing debt issues,

bondholders anticipate higher risk-taking incentives from managerial stock options than from

managerial equity ownership; Billett and Liu (2007) find in their sample of dual-class firms that bond

yields increase with managerial voting rights and decrease with cash-flow rights.

Another strand of literature studies how corporate governance affects the cost of debt. These

papers find that bond yields are negatively related to board independence (Bhojraj and Sengupta,

2003), disclosure quality (Sengupta, 1998), antitakeover provisions (Klock et al., 2005), analyst

coverage and forecast accuracy (Mansi et al., 2009a), and more restrictive payout statutes (Mansi et

al., 2009b). Cremers et al. (2007) further consider the interactions of takeover vulnerability with the

existence of large institutional shareholders, and Ellul et al. (2007) study the effect of family

ownership in high and low investor-protection environments. Moreover, Ashbaugh-Skaife et al.

(2006) show that better governance improves bond ratings.

Our study is novel in that it shows that powerful non-financial stakeholders, such as unionized

workers, may influence corporate affairs in ways that have a nontrivial impact on creditors’ wealth.

Our evidence suggests that labor unions increase bond values because, through the various channels

described in Section 2.1, they decrease the probability that firms will default on debt payments. Chen

et al. (2009) show that unions also reduce firms’ operating flexibility and increase the required return

on equity. This suggests that, through this channel, unions may also decrease bond values. Our

results show that the effect of unions on bond values through channels that protect creditors’ wealth

dominates any effect unions may have through the operating flexibility channel.

8

As we note it in the introduction, previous work documents that labor unions reduce equity

values. We extend this work by showing that labor unions increase debt values. Moreover, our

results add to the evidence in previous studies that governance attributes which are not beneficial to

shareholders can nevertheless be beneficial to bondholders (Klock et al., 2005).

More broadly, our paper is related to studies on the importance of labor relations for firms’

decisions and their values (e.g., Fauver and Fuerst, 2006; Faleye et al., 2006; Edmans, 2007;

Atanassov and Kim, 2009). Our work is also related to studies that examine the importance of

various non-financial stakeholders, such as customers and suppliers, for corporate decisions (e.g.,

Kale and Shahrur, 2007). We differ from these literatures in that we examine the importance of

labor unions and their impact on the pricing of risky debt.

Last, our paper is related to the literature on the political economy of corporate governance.

Pagano and Volpin (2005a) show that poor country-level protection of minority investors may result

from a political alliance between workers and controlling shareholders; Pagano and Volpin (2005b)

show that a similar alliance between managers and employees can occur also at the individual

company level (see also Hellwig, 2000). In addition, Perotti and von Thadden (2006) study a model

in which creditors, whose claims are concave in profitability, tend to side with employees against

shareholders. They show that when financial wealth is concentrated among rich voters, a political

majority has more at stake in the form of firm-specific human capital and supports bank dominance

to limit corporate risk taking. Our firm-level evidence that unions’ influence on corporate decisions

reduces the risk of debt supports their key intuition that the incentives of fixed claimants are aligned.

3. Data Sources, Variable Definitions, and Summary Statistics

3.1. DATA SOURCES AND SAMPLE SELECTION

We construct firms’ bond yields using the Lehman Brothers Bond Database (LBBD), which

provides month-end security-specific information on corporate bonds for the period 1973-1998

9

(e.g., market values, yields, credit ratings, duration, and maturity). Although the LBBD does not

contain the entire universe of traded debt, there is no reason to expect any systematic bias in the

sample. Unfortunately, the LBBD ceased to be publicly available at the end of 1998. We obtain

industry unionization data for 1983 onwards from the Union Membership and Coverage Database

maintained by Barry Hirsch and David Macpherson and described in Hirsch and Macpherson

(2003). Most of our additional data come from the CRSP-Compustat Merged Database. Industry

workforce demographics, which we use in our endogeneity tests, are from the CPS Labor Extracts.

To ensure that accounting information and unionization rates are already impounded into bond

yields, we match our bond data for year t with accounting information for the fiscal year ending in

year t-1 and unionization rates for year t-1. Our final data span the period 1984-1998, which is the

period for which both bond and unionization data are available. We further restrict our sample to

firms in the CRSP-Compustat Merged Database. Finally, we exclude financial companies (SIC codes

6000 to 6999) and regulated utilities (SIC codes 4900 to 4999).6 Our final sample contains 996 firms

and a total of 5,557 firm-year observations for which we have complete data on all variables needed.

3.2. LABOR FORCE UNIONIZATION AND BOND YIELD SPREADS

Following previous work in labor economics (e.g., Connolly et al., 1986), we measure labor

force unionization (Union) as the percentage of workers in a Census Industry Classification (CIC)

industry covered by unions in the collective bargaining with the firm. The CIC industries correspond

roughly to the three-digit SIC industries. To the extent that the industry-level unionization is a noisy

proxy for firm-level unionization, this measurement problem should only bias towards zero our

6 We exclude regulated gas and electric utilities because they have considerably less discretion in their choice of

investment and are unlikely to provide shareholders with significant scope for opportunistic behavior that reduces the

value of creditors’ claims. This exclusion is important for our study since utilities are highly unionized but have a low

cost of debt largely due to regulation. Nevertheless, our results are unaffected if we include utilities in our sample.

10

estimates of the impact of unionization on the cost of debt. We measure a firm’s cost of debt in

basis points using the yield spread, Spread, which reflects the market assessment of the firm’s debt

repayment risk. It is defined as the difference between the weighted-average yield to maturity of the

firm’s outstanding debt (the weight on each individual bond yield is the bond’s share in the firm’s

total value of debt) and the yield to maturity of a Treasury bond with a similar duration.

To understand the source and amount of variation in industry unionization and firms’ bond

yield spreads, in Table I we report summary statistics of unionization rates and average bond yield

spreads for the ten most- and the ten least-unionized industries. We observe a significant variation in

unionization rates across different industries. Railroads, steelworks, pulp and paper, and motor

vehicles are among the most unionized industries, with the average unionization rates above 50%

during the period 1983-1997. In contrast, small retail industries are generally not unionized, and their

average unionization rate is about 2%. We also observe a decreasing trend in unionization rates over

time, with aggregate unionization falling from 30% in 1983 to 17% in 1997.

We further inspect the cross-sectional and time-series variation in unionization rates and bond

yields using regression analysis. By regressing Union on CIC industry fixed effects we find that 93%

of the total variation in unionization rates is due to variation across CIC industries, which implies

that the time-series variation for individual industries only accounts for the remaining 7%. Hence,

cross-sectional differences in the industry unionization rates are the primary source of variation in

our key independent variable. A firm’s average yield spread is fairly persistent and it varies more

across firms at a given time than it does for a given firm over time. By regressing Spread on firm

fixed effects we find that about 60% of the variation in spreads is across firms and about 40%

constitutes a within-firm time-series variation. In our analysis, we use the large cross-sectional

variation in unionization rates and bond yields to study the effect of unions on debt risk.

Last, Table I also provides preliminary (univariate) evidence of a negative association between

unionization and bond yields: The median yield spread among the most unionized industries is 215.5

11

basis points while the median yield spread among the least unionized industries is 294.4 basis points.

We thoroughly explore this result using a multivariate regression approach in Section 4.

3.3. CONTROL VARIABLES IN THE YIELD SPREAD REGRESSIONS

Firm characteristics are defined as follows. Size is the natural logarithm of the firm’s book debt

plus the market value of equity. Finlev is the book value of debt over total assets. Tobin’s Q, TobQ, is

the market value of assets (total book debt plus market value of equity plus liquidation value of

preferred stock minus deferred taxes and investment tax credits) over the book value of assets.

Cash-flow risk, Cfrisk, is the standard deviation of quarterly ROA using observations from 15

previous quarters. Intcov1 is an indicator variable equal to one if interest coverage, defined as

operating cash flow divided by interest payments, is less than one, and zero otherwise. Profit is

operating income over assets. Negequity is an indicator variable equal to one if a firm’s book equity is

negative, and zero otherwise. Intang is intangible assets over total assets. Net working capital, NWC,

is current assets minus current liabilities, over total assets. Total payout to shareholders, Totpayout, is

dividends plus repurchases, over total assets. Productivity is sales over assets. HHI is the Herfindahl-

Hirschman Index of sales concentration in a firm’s three-digit SIC industry. Nasdaq is an indicator

variable equal to one if a firm is listed on NASDAQ, and zero otherwise.

A firm’s debt characteristics are weighted averages of the individual bond characteristics, that is,

in each year, we aggregate individual bonds by weighting their characteristics by their shares in the

total amount of the firm’s outstanding debt. Duration is the weighted-average duration of

outstanding debt. Debtage is the weighted-average difference between the settlement date and the

original bond issue date. Rating is a weighted-average debt rating index that equals 1 for S&P’s credit

ratings D to CC, 2 for CCC- to CCC+, 3 for B- to B+, 4 for BB- to BB+, 5 for BBB- to BBB+, 6

for A- to A+, 7 for AA- to AA+, and 8 for AAA to AAA+. To ensure the robustness of our results

with respect to outliers we winsorize all continuous variables at the 1% level.

12

3.4. SUMMARY STATISTICS

Table II reports the summary statistics. The mean yield spread is about 296 basis points, with a

median of 177.18. The mean and median unionization rates are 23% and 20%, with the 5th and 95th

percentiles equal to 2% and 56%, respectively. The mean debt rating index is about 5, which

corresponds to a rating of BBB according to the Standard & Poor’s classification. The index varies

from 3 (CCC) in the 5th percentile to 7 (AA) in the 95th percentile, showing that our sample covers a

wide spectrum of firms with different degrees of bondholder risk. The distributions of other

variables are similar to those reported in previous studies.

4. Multivariate Analysis

4.1. LABOR UNIONS AND YIELD SPREADS

Since the vast majority of the variation in unionization rates is across industries, throughout the

paper, we identify the effect of labor unions on yield spreads using empirical tests that largely rely on

this contemporaneous industry heterogeneity. Specifically, we regress firm-level yield spreads on

lagged industry unionization rates and several firm- and industry-level controls7:

Spread = β0 + β1Union + β2 Controls + ε. (1)

Our controls include Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout,

Productivity, HHI, and Nasdaq. Notably, including financial leverage in the regression model helps us

control for any effect unions may have on the cost of debt through their possible influences on

capital structure decisions (see Simintzi et al., 2009). We also include three debt characteristics:

Duration, Debtage, and Rating. The duration of debt controls for differences in maturity and coupon

7 One could also convert yield spreads and controls into CIC industry-level averages and estimate the regression models

at the industry level either by OLS or by WLS using the number of firms in the industry as weights. This approach

provides results that are similar to those we obtain in our firm-level analysis.

13

rates of the firm’s debt. The debt age (the length of time the bond has been outstanding) controls

for differences in bond liquidity, since previous work shows that more recently issued bonds are

more liquid than older ones. The rating index controls for differences in default risk. Finally, unless

otherwise noted, in all regressions, we include year and one-digit SIC industry fixed effects.

We estimate our models using pooled (panel) OLS regressions and calculating standard errors

clustered at the CIC industry level. This approach specifically addresses the concern that Union is an

industry-level variable but Spread and most of the controls are firm-level variables. It is thus possible

that the errors, conditional on the independent variables, are correlated within industry groupings.

One reason why this may occur is that industry factors may have a similar effect on the risk of debt

for all firms in the industry. Clustered errors assume that observations are independent across

industries, but not necessarily independent within industries. The clustering method makes minimal

assumptions about the correlation structure of the error term, and thus it is likely to provide the

most conservative standard errors.

Table III reports results from the regression models in which we sequentially add various

control variables. In column (6), our most comprehensive specification, we include firms’ debt

ratings, which further control for various risk characteristics (e.g., debt covenants) that are observed

by rating agencies but not necessarily by econometricians.8 Across all specifications, we find a

negative and highly statistically significant relation between unionization and yield spreads. This

relation is economically significant. Based on the estimates in column (6), a one-standard-deviation

increase in the unionization rate decreases the yield spread by about 32 basis points. For robustness,

in column (7), we report results from estimating a parsimonious specification in which we exclude

control variables that are not statistically significant in our benchmark specification: Intang, Totpayout, 8 Rating agencies may internalize unions’ actions in their rating decisions. However, using an Ordered Probit model we

find no effect of Union on bond ratings controlling for standard determinants of ratings, which supports the view that

bond ratings are imperfect measures of bond risk.

14

Productivity, HHI, Nasdaq, and Debtage. The coefficient of Union and the fit of the model remain

similar. In sum, the results show that the influence of unions on corporate matters is viewed

favorably in the bond market, consistent with the view that such influence serves as a mechanism

that protects bondholders’ wealth.

4.2. CONDITIONING ON FINANCIAL DISTRESS

We now study whether the effect of unionized workers on bond yields differs across financially

distressed and healthy firms. The reason is that, regardless of its nature, workers’ influence on

corporate affairs should have more impact on creditors’ wealth in financially distressed firms. In

particular, in distress, unions may: (1) play a more important role in curbing shareholders’ incentives

to engage in risky activities that increase equity values but reduce debt values, (2) block takeovers of

troubled firms, (3) mobilize political forces to bail out firms, or (4) side with shareholders to avoid

liquidations that would benefit lenders. In line with the first three channels, unionization should

decrease by more yields in distressed firms, but the last force would predict the opposite. Hence,

whether the effect should be stronger or weaker in distressed firms is a priori unclear.

To resolve this issue, we condition the effect of unionization on yield spreads on various

proxies for a firm’s financial distress. We consider five different indicators of financial distress,

broadly denoted Distress, which are based on interest coverage, book equity, Altman’s (1968) z-score,

Ohlson’s (1980) o-score, and the simplified probability of default measure based on Merton’s (1974)

bond-pricing model suggested by Bharath and Shumway (2008). Specifically, firms are classified as

distressed when their interest coverage ratios are less than one (Intcov1), they have negative equity

(Negequity), or their default risk falls into the top tercile of the empirical distribution according to z-

scores (LowZS), o-scores (HighOS), and probability of default (HighPD).9 The advantage of the last

measure is its forward-looking aspect and reliance on a completely specified default model, while the

9 By construction, default risk is higher for lower z-scores and higher o-scores.

15

other ones are largely ex-post measures and some rely on in-sample estimates of predictive

regressions; thus, they are not robust to structural changes.10

Table IV reports the results from estimating regression models with the following specification:

Spread = β0 + β1Union + β2Union×Distress + β3 Distress + β4 Controls + ε. (2)

Controls denotes the full vector of control variables we use in Table III (we omit reporting their

coefficients for brevity). Our coefficient of interest is β2, which measures the effect of Union on

Spread conditional on Distress.11 To facilitate the interpretation of our results, we demean Union and

each of the Distress indicators before forming the interaction terms. The coefficients of the controls

are omitted unless they are interacted with Union.

In addition to year fixed effects, in all empirical models we include CIC industry fixed effects to

address the concern that omitted time-invariant industry characteristics may drive our results.

Unfortunately, this procedure does not allow us to identify the effect of Union on bond yields (β1)

because unionization rates are highly persistent. However, the interactions of Union with Distress

exhibit substantially more time-series variation within each industry, inherited from the distress

variables. Thus, we gain statistical power to identify the coefficients of the interaction terms (β2).

For all the five alternative definitions of financial distress, we find that the coefficient of the

interaction between Union and Distress is negative and statistically significant, that is, unionization

reduces firms’ bond yield spreads by more in financially distressed firms.12 Even though our results

cannot rule out the possibility that labor unions may still favor some corporate policies that reduce

10 Although we use indicator variables to capture financial distress, our results are qualitatively similar if we use

continuous versions of interest coverage, z-scores, o-scores, or probability of default.

11 Our results remain similar if we use the distress indicators to split the sample into distressed and healthy firms.

12 The coefficients of the interaction terms are similar when we use one-digit SIC industry fixed effects as in Table III; in

that case we also find a negative and statistically significant effect of Union.

16

creditors’ wealth when firms are in financial distress, they suggest that the actions of labor unions

which protect creditors’ wealth in distressed firms are more important, and as a result the net effect

of labor unions is a lower risk of debt. Further, the analysis using CIC industry fixed effects gives us

some comfort that our results are unlikely driven by omitted time-invariant industry characteristics

correlated with unionization. If this were the case, then one would have to explain why these

omitted variables have a stronger effect on yield spreads when firms are financially distressed.

4.3. OMITTED INDUSTRY CHARACTERISTICS

Our results may be spuriously driven by the correlation between unionization and an industry

characteristic omitted from our model. In particular, the relation between unionization and yields

may reflect the stage in an industry’s life cycle. In fact, mature industries tend to be more unionized

and less risky; thus, they may have lower bond yields. Moreover, some industries rely more on

unskilled labor that tends to be more unionized; at the same time, these industries have lower costs

of debt due to their ample collateralizable assets and less scope for discretion in investment policies.

In contrast, collective bargaining with unions is less relevant in knowledge-intensive industries, in

which workers are more heterogeneous and often have individually negotiated contracts.

To address these concerns, we expand our main specification with several industry-level

variables that proxy for the stage of the industry’s life cycle, including the industry capital intensity

and the relative importance of human capital. Mature industries differ from other industries in that

they tend to be older and more capital intensive, have lower growth prospects, invest less in research

and development and advertising, are potentially less profitable, and have better access to external

capital. We explicitly control for industry age, defined as the logarithm of the maximum firm age in

that industry (Indlogage), the industry’s old-economy (Oldecon) and new-economy (Newecon) status13,

13 We define new-economy industries as the computer, software, internet, telecommunications, or networking industries.

Old-economy industries are those with SIC codes below 4000 that are not new economy.

17

the industry’s capital-labor ratio (Indkl), median industry R&D expenses to sales ratio (IndR&Dexp),

median industry advertising expenses to sales ratio (Indadvexp), median industry one-year asset

growth (1yr-Indgrwth), median industry ROA (Indprofit), and median industry ratio of retained earnings

to assets (Indretearn). Including Indkl addresses the concern that higher unionization could be

associated with lower yields due to its correlation with an industry’s capital intensity.

Several of the variables above serve as proxies for the nature of the production process in the

industry and also measure the extent to which an industry relies on unionized unskilled labor (e.g.,

Oldecon and Newecon, Indkl, or IndR&Dexp). In addition, we directly control for differences in the

level of education of the industry’s labor force using the fraction of the CIC industry’s workers that

have at least some college education (College). We further control for the average number of hours

worked by the workers in the CIC industry (Hours), since this variable may also proxy for the nature

of the production process or the qualification of the labor force.

The results, presented in Table V, show that the effect of unionization on bond yields is

essentially unchanged when we add the above industry variables. Thus, our findings cannot be

attributed to the correlation between Union and the stage in the industry’s life cycle or the extent to

which an industry’s production process is reliant on unionized unskilled labor.

4.4. SEPARATION OF OWNERSHIP AND CONTROL

Our conceptual framework assumes that managers strictly adhere to share price maximization.

In practice, the extent to which managers will act on behalf of shareholders to maximize the value of

equity at the expense of the value of debt will also depend on firms’ governance structures. Our

empirical tests so far do not account for differences in governance and thus may suffer from an

omitted-variable bias. Specifically, our coefficient estimates could be biased upwards if the extent to

which managers can deviate from share price maximization were systematically higher among firms

18

in more unionized industries. Hence, these firms could have lower bond yields simply because their

risk-averse managers have, for example, more freedom to pursue less risky strategies.

To address this concern, we focus on two key aspects of a firm’s governance structure: the

presence of large shareholders and shareholder rights. Managers of firms with large investors and

with stronger shareholder rights have less freedom to deviate from share-price maximization. First,

large shareholders have incentives and power to limit managerial discretion (Shleifer and Vishny,

1986), even if they lack substantial control rights (Edmans, 2009). In particular, large institutional

investors are active in monitoring managerial actions (e.g., Gillan and Starks, 2000) and in enforcing

compensation contracts with high pay-performance sensitivities (Hartzell and Starks, 2003). In

addition, shareholder rights are beneficial to shareholders because they constrain managerial

discretion. Gompers et al. (2003) find that firms with strong shareholder rights have higher long-run

stock returns and higher profits relative to firms with strong management rights. Also, Klock et al.

(2005) show that strong management rights are viewed favorably by bondholders.

To capture the differences in various aspects of corporate governance, we expand the set of

control variables in our main specification. Like Hartzell and Starks (2003), we measure institutional

investors’ influence on corporate matters using the concentration of their holdings, defined as the

fraction of the firm’s institutional ownership accounted for by the top five institutional investors

(Top5). We also include the fraction of shares held by institutions (InstitOwn). Our control for

shareholder rights uses the corporate governance index (G-Index) of Gompers et al. (2003), which is

based on twenty-four antitakeover provisions.14 The results in Table VI indicate that the effect of

Union on Spread remains statistically significant and is similar in magnitude to that reported in Table

III, suggesting that the omission of governance variables does not significantly bias our results.

14 The G-Index is available for 1990, 1993, 1995, and 1998. Since antitakeover provisions vary little over time, we fill in

the gaps using the most recent available values. For the period 1983-1989, we use values from 1990.

19

4.5. ENDOGENEITY OF UNIONIZATION

The magnitude of our results may be poorly estimated due to selection bias or measurement

error. The magnitude of the effect of unions on bond yields that we measure would be biased

upwards if unions self-select to industries which are less risky. For example, since fixed costs of

setting up a union are large, unions may be more likely to form in industries that are more likely to

survive in the long run and thus are less risky. Conversely, the magnitude would be biased

downwards if unions sort into riskier industries. For example, the benefits of unionization may be

larger in riskier industries, in which uncertain economic conditions threaten workers’ employment

and working conditions. Another potential reason for a downward bias could be measurement error,

since we use industry-level unionization rates to proxy for firm-level unionization rates.

We use a two-stage least squares (2SLS) regression to examine whether such endogeneity issues

cause a significant bias in the estimation of the effect of Union on bond yields. Our instrument for

Union is the logarithm of the fraction of female workers in the firm’s CIC industry (Female). The

labor economics literature has shown that female workers typically demand less union services than

male workers, and as a result, industries with a larger fraction of female workers tend to be less

unionized (e.g., Hirsch, 1980). Female workers are also less likely to unionize because women tend

to have less attachment to the labor market and to specific internal job ladders than do men. In

addition, the expected benefits (particularly non-wage benefits) from being a union member may be

smaller for female workers and their costs of organizing may be higher. Thus, we expect a negative

relation between Female and Union in our first-stage regression model. At the same time, Female is

unlikely to be correlated with the error term in the second-stage regression model, since there is no

obvious reason why it would directly affect bond yields.

In Table VII, we treat Union as an endogenous variable and estimate the same model as in

column (6) of Table III using a 2SLS approach. Consistent with our prediction, Panel A shows that

the effect of Female on Union in the first-stage regression is negative and statistically significant. The

20

partial R2 indicates that Female explains around 26% of the variation in unionization (net of any

effect it may have through other explanatory variables) and thus it is a powerful instrument. In

unreported tests, we find that Female has no statistically significant effect on Spread when it is added

as an additional predictor in our full OLS specification. This evidence suggests that Female is

uncorrelated with the error term in the second-stage regression and thus it is a valid instrument.

The second-stage results in Panel B continue to show a statistically significant and negative

relation between Union and Spread. The magnitude of the estimated 2SLS coefficient of Union

(-223.64) is larger than that of its OLS counterpart (-186.86), thus suggesting that any potential

endogeneity problem only biases the magnitude of the coefficient downwards. Nonetheless, based

on the Hausman test we cannot reject the hypothesis that the OLS and 2SLS estimates are equal.

4.6. THE EFFECT OF UNIONS ON THE PRICE AND AMOUNT OF DEBT

Our results can be cast in a general supply and demand framework in which unionization

affects both the equilibrium price and amount of debt. In Section 2.1, we argue that, by protecting

creditors’ wealth, unions cause a reduction in required bond yields for any given level of debt, that is,

they cause an increase in the supply of debt to the firm. If this is the only effect of unions, then

more unionized firms should exhibit both a lower cost of debt and higher financial leverage.

However, as Simintzi et al. (2009) show, unions may also shift firms’ demand for debt in ways that

are a priori ambiguous. Hence, the overall effect of unions on bond yields could result from both

supply and demand effects. To better understand these effects, we thus need to examine how unions

affect the financial leverage of the firms in our sample.

For this purpose, we regress book leverage (Finlev) on unionization (Union) and the standard

control variables: firm size (Size), market-to-book ratio (MB), profitability (Profit), asset tangibility

(Tangible), cash-flow risk (Cfrisk), and an indicator variable equal one if the firm is a dividend payer

(Paysdiv) and zero otherwise. We also include year and one-digit SIC fixed effects, and cluster the

21

standard errors at the CIC level. Although we find that unionization is negatively correlated with

book leverage in a univariate regression, the coefficient becomes very close to zero and statistically

insignificant once we include other control variables (t-statistic of 0.07). We also estimate year-by-

year cross-sectional regressions and find no association between book leverage and unionization.

Our results are similar if we use market leverage instead. Hence, unionization has no impact on the

equilibrium leverage of the firms in our sample. This result is in line with those in Lee and Mas

(2009), who find no effect of union election results on book leverage.

We find that unionization reduces bond yields, but does not have a significant impact on

financial leverage. Therefore, the supply and demand framework suggests that the increase in the

supply of debt to more unionized firms is likely to be accompanied by a simultaneous downward

shift in the demand curve. Such an increase in demand would unambiguously generate a reduction in

the equilibrium cost of debt, while possibly keeping the equilibrium amount of leverage unchanged.

The reasons for this increase in the supply of debt to more unionized firms follow directly from our

conceptual framework in Section 2.1 and our evidence provides strong support for this effect.

A decrease in the demand for debt could arise for two reasons. First, unions make wages sticky

and layoffs costly; thus, they increase firms’ operating leverage and risk (Chen et al., 2009).

Consequently, more unionized firms may choose lower financial leverage to offset their higher

operating leverage, effectively reducing their demand for debt (Simintzi et al., 2009). In line with this

view, Mandelker and Rhee (1984) provide evidence of a negative association between financial and

operating leverage. Second, workers are concerned with unemployment risk because they face large

costs from layoffs, and are reluctant to invest in firm-specific human capital when this risk is high.

This suggests that unions may induce firms to choose lower financial leverage to reduce the

likelihood of financial distress or bankruptcy and thus workers’ exposure to unemployment risk (see

the theoretical work of Berk et al., 2010, and evidence in Agrawal and Matsa, 2010).

22

5. Evidence on the Mechanism through which Labor Unions Affect Bond Yields

5.1. THE EFFECT OF UNIONS ON INVESTMENT POLICY

One important way in which labor unions might mitigate shareholder-bondholder conflicts is

by curbing shareholders’ incentives to embark on risky projects that expropriate fixed claimants. In

particular, shareholders can expropriate bondholders’ wealth by investing in negative NPV risky

projects that increase share prices but reduce bond values. We argue that such a risk-shifting

behavior may be an important concern for workers, because their contractual claims resemble the

payoff to risky debt and also because they have a substantial fraction of their wealth invested in

firm-specific human capital. As a result, workers’ preferences for risk taking are likely to be more

closely aligned with those of bondholders than with those of well-diversified shareholders.

If labor unions curb shareholders’ incentives to take risk, then firms in more unionized

industries should undertake less risky investment policies. However, this prediction is difficult to test

because firms’ choices of investment projects and their associated risks are not directly observable.

Therefore, we provide some indirect evidence about the nature of firms’ project choices. R&D

expenditures are typically viewed as riskier investments compared to expenditures on fixed assets.

For example, Kothari et al. (2002) show that R&D expenses increase the variability of future

earnings by three to four times more than do expenditures in fixed assets. Thus, we expect more

unionized firms to spend a lower fraction of their investment budgets on R&D. Moreover, we

expect this effect to be stronger in firms with weaker financial conditions, for which the potential

for risk-shifting behavior is larger.

Table VIII reports the results from estimating Tobit models in which the dependent variable is

the share of R&D spending in total spending on R&D and fixed assets (R&D/(R&D+Capex)). Our

controls include Size, Logage, Finlev, TobQ, Profit, Budget, Salgrowth, HHI, and Busconc. Logage is the

logarithm of the number of years since a firm was first listed in CRSP. Budget captures the size of a

firm’s investment budget, defined as expenditures on fixed assets plus expenditures on R&D divided

23

by assets. Salgrowth is the change in the natural logarithm of firms’ sales. Busconc is the Herfindahl

index measuring the concentration of a firm’s sales across its business segments. Other variables are

defined in Section 3.3. All regressions include one-digit SIC industry and year fixed effects. Standard

errors are clustered at the CIC industry level. Supporting our conjecture, the results in columns (1)-

(3) show that firms in more unionized industries spend a smaller fraction of their budgets on R&D.

In columns (4)-(6), we further include interactions of Union with three variables associated with

distress risk, previously defined in Section 4.2: LowZS, HighOS, and HighPD. All the variables are

demeaned before forming the interactions. The coefficients of all three interaction terms are

negative; those of Union×HighOS and Union×HighPD are also statistically significant. These results

support the view that unions curb firms’ incentives to engage in risky projects that expropriate

bondholders, and that their role is especially important in firms in which shareholder-bondholder

conflicts of interest regarding the risk of investment decisions are more severe. Notably, we find no

evidence that when firms are in or near financial distress labor unions align themselves with

shareholders in implementing riskier investment policies that could reduce debt values.

5.2. THE EFFECT OF UNIONS ON TAKEOVER ACTIVITY

Unions may also deter takeovers, which are often associated with changes in firms’ contracting

relationships that shift wealth from workers to shareholders.15 This action could lower debt risk

because takeovers often entail increases in leverage which reduce the value of existing debt that is

not fully protected by covenants. In addition, managers of target companies often engage in

defensive maneuvers that can reduce debt values. To explore this possibility, we proceed in two

steps. First, we examine whether unions deter takeovers. Second, we examine whether they have a

stronger negative effect on bond yields when firms have lower barriers to takeovers. For this

15 At the end of 2009, British and Irish workers at Cadbury, largely organized as a labor union, announced a campaign to

resist Kraft’s ₤9.8bn ($16.3bn) hostile takeover bid by appealing to shareholders and politicians to block the deal.

24

purpose, from SDC Platinum we obtain a list of all public and private firms operating in non-

financial and non-regulated industries that were acquisition targets during the period 1983-2007.

In Table IX, we report the results from industry-level tests in which we use the data for all

acquisitions aggregated at the CIC industry level. For industry-year observations with no acquisition

targets, we set the number and value of acquisitions to zero. In columns (1) and (2), we report the

results from estimating Poisson (count) models relating the number of acquisition bids and the

number of completed acquisitions in the industry to the lagged industry unionization rate and

industry-level control variables. In column (3), we report the results from estimating a Tobit model

that uses the value of completed acquisitions in the industry scaled by the industry’s total assets as a

dependent variable. The control variables follow from Billett and Xue (2007) and include the CIC

industry median of the logarithm of market capitalization (IndLogMcap), the median of the logarithm

of the market-to-book equity ratio (IndLogMBE), the median return on assets (IndROA), the median

asset tangibility ratio (IndTang), the median leverage (IndFinlev), the fraction of dividend paying firms

in the industry (IndDivPay), and the logarithm of the number of firms in the industry (LogNofirms).16

We cluster standard errors at the CIC industry level and include year fixed effects in all

specifications.

We find that the coefficient of Union is negative and statistically significant in all three

specifications we consider: In more unionized industries, we observe smaller numbers of acquisition

bids and completed acquisitions, as well as lower values of acquisition deals. This evidence suggests

that unionization is associated with a less active market for corporate control, and is consistent with

the view that unions may benefit lenders by deterring takeovers.

16 The inclusion of the logarithm of the number of firms in the industry controls for the fact that industries with more

firms tend to have larger numbers and greater values of acquisitions.

25

We also conduct firm-level tests. To this end, we match the merger targets in SDC to the

Merged CRSP-Compustat data, which restricts our sample to publicly traded firms with complete

data. We then estimate Probit and Logit models in which we relate the probability that a firm is

acquired to lagged unionization rates and the control variables in Billett and Xue (2007). We also

include one-digit SIC industry and year fixed effects. We cluster standard errors at the CIC industry

level. The untabulated results from both specifications provide statistically significant evidence that

unionization reduces the probability that a firm is acquired.

Next, we examine whether unions’ opposition to acquisitions reduces debt risk. We argue that if

unions benefit creditors because they deter takeovers which hurt the value of their claims, then their

influence on firms’ decisions should be more beneficial to the creditors of firms which are not

protected by state antitakeover laws. To examine this issue, in Panel A of Table X, we report results

from estimating our benchmark regression model separately for firms with high (above median) and

low (below median) values of the Gompers et al. (2003) governance index (G-Index), which largely

captures antitakeover provisions. The effect of unions on the cost of debt is significantly more

negative for firms with low levels of takeover protection (low G-Index). Furthermore, the coefficient

of Union in the low G-Index subsample is statistically significant at the 1% level, while in the high G-

Index subsample it is only marginally significant, at the 10% level. These results are consistent with

the view that unions reduce creditors’ risk exposure by deterring takeovers, especially when formal

takeover barriers do not exist.

In Panel B, we further examine whether unions’ role in blocking takeovers is more important in

financially distressed firms. If this were the case, then unions should reduce the cost of debt in

financially distressed firms by more when the firm has lower barriers to takeovers. For this purpose,

we estimate our benchmark model separately for firms with G-Index above and below median, and

also include an interaction term between Union and three alternative indicators for whether a firm is

in financial distress: LowSC, HighOS, and HighPD. All our empirical models include CIC industry and

26

year fixed effects. Our evidence is mixed: While the coefficient of the interaction Union x HighOS is

significantly more negative for firms with low G-Index, which suggests that unions’ role in blocking

takeovers is more valuable to lenders in distressed firms which lack antitakeover provisions, the

coefficient of the interaction term does not differ across firms with high and low G-Index if we

measure distress using LowZS or HighPD.

5.3. OTHER POSSIBLE MECHANISMS

Labor unions may also affect creditors’ wealth through other mechanisms, two of which stand

out the most. First, strong unions might have political influence and play an important role in

company bailouts (e.g., during the recent financial crisis, unions at General Motors received

considerable support from President Obama). This suggests that unions may also reduce bond yields

partly because they are very effective in mobilizing the media and politicians to the rescue of firms in

trouble. To evaluate this political-power hypothesis, we start from the premise that, due to a

historical alliance between the labor movement and the Democratic Party, Democratic governments

are more likely to support unionized firms in trouble. If this is indeed the case, then the presence of

strong unions should reduce the cost of debt by more when the Democrats are in power.

To this end, we estimate our main empirical model including an interaction term between Union

and Democrat, an indicator variable for whether unions are likely to enjoy support from Democratic

governments. We consider two versions of Democrat. In the first case, Democrat equals one during the

time periods in which the US President is from the Democratic Party; in the second case, Democrat

equals one during the time periods in which the Governor of the state in which the firm’s primary

operations are located is from the Democratic Party.17 Despite its intuitive appeal, we do not find

17 The advantage of using the first measure is that the President has likely more impact in orchestrating corporate

bailouts; the disadvantage is that it varies only over time. The other measure varies both in the time series and across

states, but governors may have less power to support troubled firms.

27

statistically significant support for the political-power hypothesis: The negative effect of unions on

the cost of debt is not stronger when government officials affiliated with the Democratic Party are

in power. A plausible explanation of this result is that governments often provide financial support

to large firms like GM, but not to smaller firms that are more typical in our sample. This could

reduce the power of our tests.

Second, our arguments largely suggest that unions’ influence on corporate affairs reduces the

probability of default on debt payments, but the interest of workers and creditors may sometimes be

in conflict in financial distress. Specifically, as discussed in Section 2.1, workers who fear losing their

income streams and firm-specific human capital investments may side with shareholders in trying to

keep the firm alive. This bias towards an inefficient continuation instead of an efficient liquidation

could in fact reduce the firm’s liquidation value and reduce the value of creditors’ claims as a result.

It is empirically difficult to examine this plausible mechanism directly. But our evidence that unions

decrease bond yields, and even by more in financial distress, suggests that the net effect of labor

unions’ actions is to reduce the risk of debt. This result implies that workers’ incentives to favor an

inefficient continuation are unlikely to have a large effect on creditors’ wealth.

6. Additional Tests

In this section, we discuss the results of various additional robustness tests for our main

specification in Table III. Details of each test are available upon request.

6.1. SIMULTANEOUS DETERMINATION OF SPREADS AND FINANCIAL LEVERAGE

Both financial leverage and bond yield spreads could be jointly determined by firms’

unionization rates. However, this is unlikely to severely bias the effect of Union on Spread we

document. First, in all regression models we control for financial leverage. Thus, the coefficient of

Union captures the effect of unionization on spreads independently of any effect it may have through

its effect on leverage. If we omit leverage from the regression, the coefficient of Union captures the

28

effect of unionization including any effect it may have through its effect on leverage. However, this

omission has a negligible impact on the coefficient of Union, which suggests that an association

between unionization and leverage does not drive our results.

Second, we replace a firm’s leverage by more exogenous measures: the CIC-industry median

leverage and a firm’s initial leverage (the firm’s leverage recorded the first year the firm’s appears in

the Compustat database). Several studies have shown that industry leverage predicts firm leverage

and Lemmon et al. (2008) show that a firm’s initial leverage has significant power to predict its

future leverage. These leverage variables are less likely to be simultaneously determined with a firm’s

current yield spread. Using these measures of leverage does not affect the coefficient of Union.

Third, we use a firm’s initial leverage as an instrument for its contemporaneous leverage in a

2SLS framework. As discussed above, initial leverage is a strong predictor of future leverage. It is

less obvious why this variable, based on information which is distant in the past, should predict

current yield spreads, other than through its ability to predict a firm’s current leverage. We find that

the coefficient of initial leverage is not statistically significant when we include both current leverage

and initial leverage in the same regression model. This result suggests that initial leverage is correctly

excluded from the second-stage regression model and thus can serve as an instrument in the first

stage. The 2SLS coefficient of Union is almost the same as that from the OLS regression model.

Hence, it is unlikely that the endogeneity of leverage explains our results.

6.2. NONLINEARITIES IN FINANCIAL LEVERAGE AND CREDIT RATINGS

We explore if the results in our main specification are affected by nonlinearities in book

leverage or credit ratings. First, in our regression model, we include a quadratic term of book

leverage. Second, we use a specification in which we split leverage into three indicator variables

based on its empirical distribution (high, medium, and low leverage). Third, we replace the

continuous rating variable by a dummy for each of the following rating categories: D to CCC+, B-

29

to BB+, BBB- to A+, and AA- to AAA+. Fourth, we also compare the results using the continuous

rating variable in our benchmark specification to those we obtain using both this variable and its

squared term. In all the specifications, we find that our main results remain largely unaffected.

6.3. NONLINEARITIES IN UNIONIZATION

The relation between Union and Spread may be non-linear (U-shaped). This can happen for two

reasons. First, an increase in unionization may substantially increase unions’ ability to influence

firms’ decisions in firms that initially have low unionization rates. When unionization is already high,

further increases in unionization rates may not be as effective in increasing the influence of workers.

Second, although some influence of unions on corporate decisions may benefit bondholders,

excessive power may allow unions to pursue their own agendas and destroy firm value. To evaluate

this possibility we add Union2 to our benchmark empirical model; we find no evidence that the effect

of Union on Spread weakens at higher levels of unionization.

6.4. CASH-FLOW RISK

In our main specification, we include a coarse proxy for risk-shifting behavior by shareholders:

cash-flow risk. To the extent unions affect the cost of debt through their impact on this risk-shifting

behavior, one would expect that controlling for cash-flow risk should diminish the effect of Union

on Spread. This is indeed the case: Including Cflowrisk in the regression model reduces the magnitude

of the coefficient of Union from -224.133 to -186.862 (a 15% drop in value). Thus, consistent with

the view that unions affect firms’ risk taking, we find that differences in cash-flow risk across firms

capture some of the effect of unions on bond yield spreads. Nonetheless, the coefficient of Union

remains statistically significant and economically large even after including Cflowrisk. This is likely

because this variable is an imperfect measure of shareholders’ unobservable risk-shifting incentives

and because unions also affect bond yields through other channels such as by deterring takeovers.

30

6.5. SAMPLE SELECTION BIAS

Although we aim to study the relation between unionization and bond yields for the entire

population of firms, our data only includes firms covered by the LBBD. If the criteria for inclusion

in the LBBD favor firms with certain attributes, then our sample may not be a random draw from

the population and our results may not hold for the entire population. We address this concern

using the Heckman (1979) procedure. We first estimate a Probit model of the probability that a firm

is contained in the LBBD data using all firms in the merged CRSP-Compustat database over our

sample period. Our predictors are firm size and indicator variables for whether the firm has rated

debt, is included in the S&P500 index, or it trades on NASDAQ. We then control for selection bias

by including the Inverse Mills Ratio from this Probit in the second-stage regression model projecting

Spread on Union and the control variables. The coefficient of Union remains statistically significant

and similar in magnitude to that in the benchmark OLS specification.

6.6. MULTI-SEGMENT FIRMS

To construct our key test variable, we assign each firm the unionization rate corresponding to

its CIC industry, which we identify using the firm’s primary SIC code. This procedure gives a precise

matching for single-segment firms. However, it may not be accurate for multi-segment firms, whose

operations span several industries with different unionization rates. To address this concern, we

repeat our analysis using the employee-weighted unionization rate averaged across the industries

corresponding to each of the firm’s business segments. The results are similar to our benchmark

results, both in terms of their magnitude and statistical significance.

6.7. INCLUDING TWO-DIGIT SIC INDUSTRY FIXED EFFECTS

In our main specification, we control for unobservable industry characteristics using one-digit

SIC industry fixed effects. We repeat our regressions using two-digit SIC fixed effects instead. The

benefit of this estimation approach is that it mitigates the concern that unobservable differences

31

across industries may bias our inference. However, the cost is that it ignores the large variation in

unionization rates across two-digit SIC industries. The effect of unionization on yields remains

statistically significant and similar in magnitude to that reported in Table III.

6.8. ALTERNATIVE CLUSTERING METHODS

Throughout the paper we calculate the statistical significance of our coefficients using standard

errors clustered at the CIC industry level. To evaluate whether this method indeed provides the

most conservative values of standard errors, we estimate our main specification clustering standard

errors by both CIC industry and year. The coefficients retain similar statistical significance. We also

explore clustering by firm, year, and both firm and year, with no detriment to our original findings.

7. Conclusion

We examine how workers’ influence on corporate affairs affects the pricing of corporate public

debt. Using industry labor unionization as a proxy for workers’ influence, we find that: (1) Firms in

more unionized industries have statistically and economically significant lower bond yield spreads;

(2) The negative effect of unions on bond yields is stronger in firms that have weak financial

conditions; (3) Higher unionization is associated with a smaller fraction of investment budgets

allocated to R&D than to physical assets, especially in firms that are close to financial distress; and

(4) Higher unionization is associated with a lower likelihood that a firm is an acquisition target and

unionization reduces bond yield spreads by more when firms’ takeover barriers are lower.

Our evidence supports the view that, by attempting to preserve the value of their contractually

fixed claims and the value of their human capital investments in the firm, unionized workers also

protect bondholders’ and other fixed claimants’ wealth. Hence, our results suggest that the value of

any particular fixed claim on a firm’s cash flow depends on the extent to which other stakeholders

with fixed-payoff contracts are able to monitor the firm’s decision-making process. We also show

that while previous work finds that unions decrease equity values, they nevertheless increase debt

32

values. We provide an economic rationale for why that is the case. Our evidence highlights the need

to study the effects of non-financial stakeholders on debt and equity values separately.

Our work motivates further investigation of the question whether other non-financial

stakeholders who are largely fixed claimants, such as firms’ customers and suppliers, may also

influence corporate policies that affect creditors’ wealth. More broadly, our findings suggest that

more detailed information about the nature of the relation between a firm and its various

stakeholders might be useful for investors in assessing the value and risk of debt-like financial

claims. This is a fruitful area for future research.

References

Abowd, J. M. (1989) The effect of wage bargaining on the stock market value of the firm, American Economic Review 79, 774-800.

Acharya, V. V., Myers, S. and Rajan, R. (2009) The internal governance of firms, unpublished working paper, University of Chicago.

Agrawal, A. K. (2008) Corporate governance objectives of labor union shareholders: Evidence from proxy voting, NYU Stern Working Paper Series No. Fin-08-006.

Agrawal, A. K. and Matsa, D. A. (2010) Labor unemployment risk and corporate financing decisions, unpublished working paper, New York University.

Altman, E. I. (1968) Financial ratios, discriminant analysis and the prediction of corporate bankruptcy, Journal of Finance 23, 589-609.

Anderson, R. C., Mansi, S. A. and Reeb, D. M. (2003) Founding family ownership and the agency costs of debt, Journal of Financial Economics 68, 263-285.

Ashbaugh-Skaife, H., Collins, D. and LaFond, R. (2006) The effects of corporate governance on firm’s credit ratings, Journal of Accounting and Economics 42, 203-243.

Atanassov, J. and Kim, E. H. (2009) Labor and corporate governance: International evidence from restructuring decisions, Journal of Finance 64, 341-374.

Bagnani, E., Milonas, N. T., Saunders, A. and Travlos, N. G. (1994) Managers, owners, and the pricing of risky debt: An empirical analysis, Journal of Finance 49, 453-477.

Benmelech, E., Bergman, N. K. and Enriquez, R. (2010) Negotiating with labor under financial distress, unpublished working paper, Harvard University.

Berk, J. B., Stanton, R. and Zechner, J. (2010) Human capital, bankruptcy and capital structure, Journal of Finance, Forthcoming.

Bharath, S. T. and Shumway, T. (2008) Forecasting default with the Merton distance to default model, Review of Financial Studies 21, 1339-1369.

33

Bhojraj, S. and Sengupta, P. (2003) Effect of corporate governance on bond ratings and yields: The role of institutional investors and outside directors, Journal of Business 76, 455-475.

Billett, M. T. and Liu, Y. (2007) Shareholder-manager alignment and the cost of debt, unpublished working paper, University of Iowa.

Billett, M. T. and Xue, H. (2007) The takeover deterrent effect of open market share repurchases, Journal of Finance 62, 1827-1850.

Billett, M. T., King, D. and Mauer, D. D. (2004) Bondholder wealth effects in mergers and acquisitions: New evidence from the 1980s and 1990s, Journal of Finance 59, 107-135.

Bronars, S. G. and Deere, D. R. (1991) The threat of unionization, the use of debt, and the preservation of shareholder wealth, Quarterly Journal of Economics 106, 231-254.

Chen, H., Kacperczyk, M. and Ortiz-Molina, H. (2009) Labor unions, operating flexibility, and the cost of equity, Journal of Financial and Quantitative Analysis, Forthcoming.

Connolly, R. A., Hirsch, B. T. and Hirschey, M. (1986) Union rent seeking, intangible capital, and market value of the firm, Review of Economics and Statistics 68, 567-577.

Cremers, M. K. J., Nair, V. V. and Wei, C. (2007) Governance mechanisms and bond prices, Review of Financial Studies 20, 1359-1388.

Dhillon, U. S. and Johnson, H. (1994) The effect of dividend changes on stock and bond prices, Journal of Finance 49, 281-289.

Edmans, A. (2007) Does the stock market fully value intangibles? Employee satisfaction and equity prices, unpublished working paper, Wharton.

Edmans, A. (2009) Blockholder trading, market efficiency, and managerial myopia, Journal of Finance 64, 2481-2513.

Ellul, A., Guntay, L. and Lel, U. (2007) External governance and debt agency costs of family firms, unpublished working paper, Indiana University.

Faleye, O., Mehrotra, V. and Morck, R. (2006) When labor has a voice in corporate governance, Journal of Financial and Quantitative Analysis 41, 489-510.

Fama, E. F. (1990) Contract costs and financing decisions, Journal of Business 63, 71-91.

Fauver, L. and Fuerst, M. E. (2006) Does good corporate governance include employee representation? Evidence from German corporate boards, Journal of Financial Economics 82, 673-710.

Freeman, R. B. and Medoff, J. L. (1984) What do unions do?, New York: Basic Books.

Gillan, S. and Starks, L. T. (2000) Corporate governance proposals and shareholder activism: The role of institutional investors, Journal of Financial Economics 57, 275-305.

Gompers, P. A., Ishii, J. L. and Metrick, A. (2003) Corporate governance and equity prices, Quarterly Journal of Economics 118, 107-155.

Gray, G. R., Myers, D. W. and Myers, P. S. (1999) Cooperative provisions in labor agreements: a new paradigm?, Monthly Labor Review, January, 29-45.

Hartzell, J. C. and Starks, L. T. (2003) Institutional investors and executive compensation, Journal of Finance 58, 2351-2374.

34

Heckman, J. (1979) Sample selection bias as a specification error, Econometrica 47, 153-161.

Hellwig, M. (2000) On the economics and politics of corporate finance and corporate control, in Xavier Vives, ed.: Corporate Governance: Theoretical and Empirical Perspectives (Cambridge University Press, Cambridge, UK).

Hirsch, B. T. (1980) The determinants of unionization: An analysis of interarea differences, Industrial and Labor Relations Review 33, 147-161.

Hirsch, B. T. (1991) Union coverage and profitability among U.S. firms, Review of Economics and Statistics 73, 69-77.

Hirsch, B. T. and Macpherson, D. A. (2003) Union membership and coverage database from the Current Population Survey: Note, Industrial and Labor Relations Review 56, 349-54.

Ippolito, R. A. (1985) The economic function of underfunded pension plans, Journal of Law and Economics 28, 611-651.

Jensen, M. C. and Meckling, W. H. (1976) Theory of the firm: Managerial behavior, agency costs and ownership structure, Journal of Financial Economics 3, 305-360.

Kale, J. R. and Shahrur, H. (2007) Corporate capital structure and the characteristics of suppliers and customers, Journal of Financial Economics 83, 321-365.

Klock, M. S., Mansi, S. A. and Maxwell, W. F. (2005) Does corporate governance matter to bondholders?, Journal of Financial and Quantitative Analysis 40, 693-719.

Kothari, S.P., Laguerre, T. E. and Leone, A. J. (2002) Capitalization versus expensing: evidence on the uncertainty of future earnings from capital expenditures versus R&D outlays, Review of Accounting Studies 7, 355-382.

Lee, D. and Mas, A. (2009) Long-run impacts of unions on firms: new evidence from financial markets, 1961-1999, NBER Working Paper 14709.

Lemmon, M. L., Roberts, M. R. and Zender, J. F. (2008) Back to the beginning: Persistence and the cross section of corporate capital structure, Journal of Finance 63, 1575-1608.

Mandelker, G. N. and Rhee, S. G. (1984) The impact of the degrees of operating and financial leverage on systematic risk of common stock, Journal of Financial and Quantitative Analysis 19, 45-57.

Mansi, S. A., Maxwell, W. F. and Miller, D. P. (2009a) Analyst forecast characteristics and the cost of debt, Review of Accounting Studies, Forthcoming.

Mansi, S. A., Maxwell, W. F. and Wald, J. K. (2009b) Creditor protection laws and the cost of debt, Journal of Law and Economics 52, 701-726.

Maxwell, W. F. and Rao, R. P. (2003) Do spin-offs expropriate wealth from bondholders?, Journal of Finance 58, 2087-2108.

Maxwell, W. F. and Stephens, C. P. (2003) The wealth effects of repurchases on bondholders, Journal of Finance 58, 895-919.

Merton, R. C. (1974) On the pricing of corporate debt: The risk structure of interest rates, Journal of Finance 29, 449-470.

Ohlson, J. A. (1980) Financial ratios and the probabilistic prediction of bankruptcy, Journal of Accounting Research 18, 109-131.

35

Ortiz-Molina, H. (2006) Top-management incentives and the pricing of corporate public debt, Journal of Financial and Quantitative Analysis 41, 317-340.

Pagano, M. and Volpin, P. (2005a) The political economy of corporate governance, American Economic Review 95, 1005-1030.

Pagano, M. and Volpin, P. (2005b) Managers, workers, and corporate control, Journal of Finance 60, 843-870.

Perotti, E. C. and von Thadden, E-L. (2006) The political economy of corporate control and labor rents, Journal of Political Economy 114, 145-174.

Ruback, R. S. and Zimmerman, M. B. (1984) Unionization and profitability: Evidence from the capital market, Journal of Political Economy 92, 1134-1157.

Schwab, S. J. and Thomas, R. S. (1998) Realigning corporate governance: Shareholder activism by labor unions, Michigan Law Review 96, 1018-1095.

Sengupta, P. (1998) Corporate disclosure quality and the cost of debt, Accounting Review 73, 459-474.

Shastry, K. (1990) The differential effects of mergers on corporate security values, Research in Finance 8, 175-201.

Shleifer, A. and Summers, L. (1998) Breach of trust in hostile takeovers, in Alan J. Auerbach, ed., Corporate Takeovers: Causes and Consequences. Chicago: University of Chicago Press, 33-56.

Shleifer, A. and Vishny, R. (1986) Large shareholders and corporate control, Journal of Political Economy 94, 461-488.

Simintzi, E., Vig, V. and Volpin, P. (2009) Labor and capital: Is debt a bargaining tool?, unpublished working paper, London Business School.

Warga, A. and Welch, I. (1993) Bondholder losses in leveraged buyouts, Review of Financial Studies 6, 959-982.

36

Table I. Average Unionization Rates and Yield Spreads for Selected CIC Industries This table reports the time-series averages (over the 15 years in our sample period) of the industry unionization rate (in %) and the industry debt yield spread (in basis points) for selected Census Industry Classification (CIC) industries. The unionization rate (defined at the industry level) is the fraction of the industry’s workers covered by unions in their collective bargaining with the firm. For each year, the industry debt yield spread is the average yield spreads across all firms in the industry. Based on the time-series average of unionization rates, we identify the ten most and least unionized among the 145 industries corresponding to different CICs contained in our data.

High-Unionization Industries Unionization Spread Railroads 79.3% 175.6 Blast furnaces, steelworks, rolling and finishing mills 55.7% 380.3 Pulp, paper, and paperboard mills 54.3% 166.5 Motor vehicles and motor vehicle equipment 52.0% 190.1 Telephone communications 47.4% 204.5 Engines and turbines 45.8% 226.4 Coal mining 45.7% 627.7 Bus service and urban transit 45.1% 231.1 Primary aluminum industries 42.2% 120.3 Iron and steel foundries 41.8% 472.8 Mean 50.9% 279.6 Median 46.6% 215.5 Low-Unionization Industries Unionization Spread Variety stores 3.1% 195.3 Auto and home supply stores 2.8% 118.8 Eating and drinking places 2.6% 407.0 Farm-product raw materials 2.6% 267.9 Miscellaneous retail stores 2.2% 320.8 Sporting goods, bicycles, and hobby stores 2.1% 386.3 Computer and data processing services 2.0% 176.4 Jewelry stores 1.9% 482.5 Book and stationery stores 1.8% 396.8 Radio, TV, and computer stores 1.8% 262.7 Mean 2.3% 301.5 Median 2.2% 294.4

37

Table II. Summary Statistics The sample spans the period 1984-1998 and contains 5,557 observations corresponding to 996 non-financial and non-utility firms. Spread is the yield spread in basis points, defined as the difference between the weighted-average yield to maturity on the firm’s outstanding debt and the yield to maturity on a Treasury security with similar duration; Union is the fraction of workers in the Census Industry Classification (CIC) that are covered by unions in the collective bargaining with the firm; Size is firm size, defined as the natural logarithm of the firms’ debt outstanding plus the market value of equity; Finlev is financial leverage, defined as the book value of debt divided by total assets; TobQ is the market-to-book ratio, defined as the market value of assets divided by the book value of assets; Cfrisk is the standard deviation of quarterly ROA using the 15 previous quarters; Intcov1 is an indicator variable equal to one if interest coverage, defined as operating cash-flow divided by interest payments, is less than one, and zero otherwise; Profit is profitability, defined as operating income divided by assets; Negequity is an indicator variable equal to one if the firm’s book equity is negative, and zero otherwise; Intang is the ratio of intangible assets to total assets; NWC is net working capital divided by total assets; Totpayout is the total payout to shareholders, including dividends and repurchases, divided by total assets; Productivity is the productivity of a firm’s assets, defined as sales divided by assets; HHI is the Herfindahl index of sales concentration in the firm’s three-digit SIC industry; Nasdaq is an indicator variable equal to one if the firm trades on NASDAQ, and zero otherwise; Duration is the weighted-average duration of outstanding debt; Debtage is the weighted-average difference between the settlement date and the original bond issue date; Rating is an index of the firm’s debt rating, defined as the weighted-average S&P’s credit rating at the date of the yield observation (it takes a value of 1 for ratings D to CC, 2 for CCC- to CCC+, 3 for B- to B+, 4 for BB- to BB+, 5 for BBB- to BBB+, 6 for A- to A+, 7 for AA- to AA+, and 8 for AAA to AAA+).

Variable # obs Mean Std. Dev. Pctile 5 Median Pctile95 Spread 5557 296.44 509.47 49.50 177.18 722.62 Union 5557 0.23 0.17 0.02 0.20 0.56 Size 5557 7.60 1.47 5.22 7.59 10.04 Finlev 5557 0.34 0.20 0.10 0.30 0.72 TobQ 5557 1.10 0.56 0.52 0.94 2.17 Cfrisk 5557 1.32 1.46 0.21 0.83 4.21 Intcov1 5557 0.07 0.25 0.00 0.00 1.00 Profit 5557 0.14 0.07 0.03 0.14 0.25 Negequity 5557 0.05 0.21 0.00 0.00 0.00 Intang 5557 0.07 0.11 0.00 0.01 0.32 NWC 5557 0.13 0.16 -0.07 0.11 0.41 Totpayout 5557 0.03 0.04 0.00 0.02 0.11 Productivity 5557 1.15 0.69 0.29 1.05 2.33 HHI 5557 0.25 0.18 0.07 0.19 0.62 Nasdaq 5557 0.11 0.31 0.00 0.00 1.00 Duration 5557 5.84 2.36 2.04 5.74 9.97 Debtage 5557 4.94 4.68 0.41 3.54 16.18 Rating 5557 4.96 1.50 3.00 5.00 7.00

38

Table III. Unionization and Yield Spreads

The table reports OLS regressions of Spread on Union and the following control variables: Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout, Productivity, HHI, Nasdaq, Duration, Debtage, and Rating. See Table II for detailed definitions of all variables. All regressions include one-digit SIC industry fixed effects (SIC1 FE) and year fixed effects (Year FE). The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

(1) (2) (3) (4) (5) (6) (7) Union -234.795** -152.549** -171.129** -212.905*** -179.545*** -186.862*** -176.190*** (2.08) (2.05) (2.52) (3.06) (2.80) (3.33) (2.88) Size -57.342*** -47.363*** -54.999*** -48.925*** -5.921 -4.297 (10.13) (7.62) (8.97) (8.07) (0.47) (0.32) Finlev 744.147*** 505.076*** 474.032*** 431.893*** 156.213 145.277 (8.69) (6.27) (4.82) (4.32) (1.16) (1.12) TobQ -51.964*** -34.632 -18.416 -28.074* -14.621 -16.454 (2.74) (1.64) (1.08) (1.66) (0.79) (0.85) Cfrisk 64.459*** 45.283*** 43.851*** 41.583*** 35.733*** 36.204*** (5.02) (4.48) (4.61) (4.57) (4.29) (4.31) Intcov1 422.513*** 419.777*** 408.098*** 382.145*** 381.921*** (4.40) (4.38) (4.43) (4.25) (4.26) Profit -337.746** -365.888** -359.403** -166.943 -136.278 (2.18) (2.27) (2.17) (1.08) (0.92) Negequity 168.205 154.030 167.773 158.523 162.688 (1.43) (1.33) (1.46) (1.43) (1.47) Intang -21.914 -35.847 -56.807 -78.522 (0.33) (0.54) (0.75) (1.26) NWC -190.547 -180.674 -186.878 -185.356 (1.54) (1.47) (1.55) (1.51) Totpayout -366.825*** -356.208*** -23.412 (2.76) (2.80) (0.20) Productivity 15.043 11.349 5.514 (0.97) (0.73) (0.36) HHI -66.376* -57.310 -41.895 (1.74) (1.50) (1.09) Nasdaq -36.692 -42.285 -32.060 (1.24) (1.38) (0.98) Duration -24.862*** -20.975*** -20.235*** (3.90) (3.85) (3.72) Debtage -2.510* -1.629 (1.83) (1.06) Rating -95.081*** -96.002*** (4.72) (4.82) Constant 264.362*** 394.646*** 412.963*** 523.145*** 658.229*** 876.794*** 843.269*** (7.63) (6.08) (6.12) (5.40) (5.82) (6.31) (6.61) SIC1 FE Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes # of obs. 5557 5557 5557 5557 5557 5557 5557 R2 0.07 0.25 0.30 0.31 0.32 0.34 0.34

39

Table IV. The Effect of Unionization on Yield Spreads Conditional on Financial Distress The table reports OLS regressions of Spread on Union, Union×Distress, Distress, and the following control variables defined in Table II: Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout, Productivity, HHI, Nasdaq, Duration, Debtage, and Rating. Distress is a generic indicator variable equal to one if the firm is in financial distress, and zero otherwise. In each specification we use an alternative indicator or financial ditress: Intcov1 equals one for firms with interest coverage ratios less than one; Negequity equals one for firms with negative equity; LowZS equals one if a firm’s default risk based on Altman’s (1968) z-score is in the top tercile of the empirical distribution (lower z-scores are associated with higher default risk); HighOS equals one if a firm’s default risk based on Ohlson’s (1980) o-score is in the top tercile of the empirical distribution (higher o-scores are associated with higher default risk); and HighPD equals one if a firm’s probability of default derived from the Merton (1974) bond-pricing model as suggested by Bharath and Shumway (2008) is in the top tercile of the empirical distribution. We demean Union and each of the above financial distress indicators before forming the interaction terms. All regressions include CIC industry fixed effects (CIC FE) and year fixed effects (Year FE). The coefficients of the standard control variables are omitted for brevity unless they are interacted with Union. The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

(1) (2) (3) (4) (5) Union -315.199 -421.644 -420.879 -479.021 -430.312 (1.06) (1.32) (1.27) (1.34) (1.37) Union × Intcov1 -1,551.415*** (4.08) Intcov1 382.717*** (4.72) Union × Negequity -1,628.336*** (2.78) Negequity 156.820 (1.43) Union × LowZS -341.027*** (3.77) LowZS -21.524 (0.46) Union × HighOS -444.751*** (3.65) HighOS -17.147 (0.62) Union × HighPD -395.366*** (4.93) HighPD 8.106 (0.27) Controls Yes Yes Yes Yes Yes CIC FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes # of obs. 5557 5557 5533 5142 5556 R2 0.38 0.38 0.35 0.36 0.35

40

Table V. Regressions of Yield Spreads on Unionization with Additional Industry Controls

The table reports OLS regressions of Spread on Union and an expanded set of control variables. As in our benchmark specification in column (6) of Table III, we include Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout, Productivity, HHI, Nasdaq, Duration, Debtage, and Rating (see Table II for detailed definitions of these variables). We also include industry-level control variables. Indlogage is the logarithm of the age of the oldest firm in a CIC industry; Oldecon is an indicator variable equal to one if a firm operates in an “old-economy” industry, and zero otherwise. Old-economy industries are defined as industries with SIC codes less than 4000 that are not in the computer, software, internet, telecommunications, or networking industries; Newecon is an indicator variable equal to one if a firm operates in a “new-economy” industry, and zero otherwise; New-economy industries are defined to comprise firms in the computer, software, internet, telecommunications, or networking industries; Indkl is the capital-labor ratio in the firm’s CIC industry, defined as fixed assets divided by the number of employees (we divide this number by 1000); Indrdexp is the median ratio of R&D expenses to sales in a CIC industry; Indadvexp is the median ratio of advertising expenses to sales in a CIC industry; 1yr-Indgrwth is the median one-year growth in the logarithm of firm assets in a CIC industry; Indprofit is the median ROA in a CIC industry; Indretearn is the median ratio of retained earnings to assets in a CIC industry; College is the fraction of the CIC industry’s workers that have some college education; and Hours is the average number of hours worked by the workers in the CIC industry. All regressions include one-digit SIC industry fixed effects (SIC1 FE) and year fixed effects (Year FE). We omit the coefficients of the control variables from the benchmark specification for brevity. The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

(1) (2) (3) (4) (5) Variables in Main Specification (Control Variables in Benchmark Specification Omitted) Union -193.081*** -165.941*** -187.129*** -185.167*** -161.690** (3.38) (2.93) (3.26) (2.97) (2.55) Additional Industry Control Variables Indlogage 15.068 1.994 (0.58) (0.07) Oldecon 54.133 57.869 (1.33) (1.46) Newecon 23.338 26.510 (0.69) (0.72) Indkl -192.329 -224.925 (1.16) (1.40) Indrdexp 28.658* 18.720 (1.79) (1.35) Indadvexp 116.302 151.684 (1.19) (1.37) 1yr-Indgrwth -150.050 -201.786 (0.65) (0.86) Indprofit -166.736 -116.784 (0.95) (0.68) Indretearn -103.066 -165.286 (0.82) (1.41) College 0.080 0.249 (0.09) (0.26) Hours -2.542 -1.042 (0.79) (0.33) Controls Yes Yes Yes Yes Yes SIC1 FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes # of obs. 5557 5557 5549 5557 5549 R2 0.34 0.34 0.35 0.34 0.35

41

Table VI. Regressions of Yield Spreads on Unionization Controlling for Corporate Governance

The table reports OLS regressions of Spread on Union and an expanded set of control variables. As in our benchmark specification in column (6) of Table III, we include Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout, Productivity, HHI, Nasdaq, Duration, Debtage, and Rating (see Table II for detailed definitions of these variables). We also include corporate governance control variables. Top5 is the fraction of the institutional ownership accounted for by the 5 largest institutional investors; InstitOwn is the fractional ownership of institutions; and G-Index is the firm’s corporate governance index based on antitakeover provisions constructed by Gompers, Ishii, and Metrick (2003). All regressions include one-digit SIC industry fixed effects (SIC1 FE) and year fixed effects (Year FE). We omit the coefficients of the control variables from the benchmark specification for brevity. The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

(1) (2) (3) Variables in Main Specification (Control Variables in Benchmark Specification Omitted) Union -211.169*** -184.612*** -214.013*** (3.80) (3.15) (3.62) Additional Corporate Governance Control Variables Top5 2.272*** 2.243*** (3.69) (3.57) InstitOwn -1.338*** -1.356*** (3.18) (3.14) G-Index -3.168 0.169 (1.21) (0.06) Controls Yes Yes Yes SIC1 FE Yes Yes Yes Year FE Yes Yes Yes # of obs. 5539 5518 5500 R2 0.35 0.34 0.35

42

Table VII. Regressions of Yield Spreads on Unionization: Instrumental-Variables Estimation The table reports two-stage least squares (2SLS) regressions of Spread on Union, which is treated as an endogenous variable, and the following control variables from our benchmark specification in column (6) of Table III: Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout, Productivity, HHI, Nasdaq, Duration, Debtage, and Rating. See Table II for detailed definitions of all variables. The instrumental variable is Female, defined as the natural logarithm of the fraction of workers in the Census Industry Classification industry that are female. Both the first-stage and second-stage regressions include one-digit SIC industry fixed effects (SIC1 FE) and year fixed effects (Year FE). The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. Panel A reports condensed information for the first-stage regression of Union on Female and all exogenous control variables. All variables other than Union are omitted for brevity. In addition to the coefficient of Female, we report the partial R2 which indicates the fraction of the variation in Union explained by Female, net of its effect through the exogenous variables. Panel B reports the coefficient of Union from the second-stage regressions of Spread on Union, which we treat as the endogenous variable, and all the exogenous control variables. All variables other than Union are omitted for brevity. We also report the statistics for the Hausman test which examines whether the OLS and 2SLS coefficients of Union are statistically different from each other.

Panel A: Summary of First-Stage Regression Coefficient of Female -0.158*** (5.34) Partial R2 0.261 Panel B: Summary of Second-Stage Regression Coefficient of Union -223.635** (2.29) Statistic for Hausman Test 0.490 p-value 0.628

43

Table VIII. Unionization and Investment Policy The table reports Tobit regressions of the share of R&D expenses in total spending on R&D and physical equipment (i.e., R&D/(R&D+Capex)), on Union and the following control variables: Size, Logage, Finlev, TobQ, Profit, Budget, Salgrowth, HHI, and Busconc. Budget is the firm’s total investment budget, defined as capital expenditures plus R&D expenditures divided by assets. See Table II for detailed definitions of the other control variables. We also include interactions of Union with alternative indicator variables for whether a firm is financially distressed which are defined in Table IV: LowZS, HighOS, and HighPD. We demean Union and each of the above financial distress indicators before forming the interaction terms. All regressions include one-digit SIC industry fixed effects (SIC1 FE) and year fixed effects (Year FE). The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

(1) (2) (3) (4) (5) (6) Union -0.422** -0.412*** -0.406*** -0.400*** -0.424*** -0.409*** (2.56) (3.52) (3.59) (3.55) (3.68) (3.68) Size 0.045*** 0.042*** 0.043*** 0.042*** 0.042*** (4.31) (4.03) (3.96) (3.45) (4.05) Logage 0.046*** 0.033** 0.032** 0.034** 0.032** (3.34) (2.37) (2.27) (2.39) (2.32) Finlev -0.276*** -0.280*** -0.273*** -0.296*** -0.293*** (4.53) (4.80) (4.69) (4.45) (4.67) TobQ 0.034 0.042 0.043 0.048 0.046 (0.99) (1.29) (1.29) (1.47) (1.42) Profit -0.393* -0.394* -0.434** -0.361 -0.401* (1.90) (1.79) (2.08) (1.58) (1.89) Budget 0.802*** 0.859*** 0.857*** 0.933*** 0.849*** (2.69) (3.03) (2.98) (3.01) (3.00) Salgrowth -0.057 -0.060 -0.061* -0.057 (1.59) (1.62) (1.66) (1.58) HHI -0.057 -0.059 -0.032 -0.059 (0.54) (0.55) (0.29) (0.56) Busconc -0.144*** -0.141*** -0.142*** -0.145*** (3.49) (3.43) (3.41) (3.58) Union × LowZC -0.086 (0.97) LowZS -0.009 (0.35) Union × HighOS -0.168* (1.66) HighOS 0.026 (1.34) Union × HighPD -0.187** (2.27) HighPD 0.009 (0.59) Constant -0.065 -0.398*** -0.216 -0.311** -0.329*** -0.309** (0.52) (3.14) (1.63) (2.53) (2.68) (2.49) SIC1 FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes # of obs. 5409 5409 5409 5389 5057 5409

44

Table IX. Unionization and Acquisition Activity in the Industry

The table reports industry-level analyses of how unionization affects acquisition activity. All variables are defined at the Census Industry Classification (CIC) level. The dependent variables capturing the acquisition activity in the industry are calculated using all public and private firms operating in non-financial and non-regulated industries that were acquisition targets during the period 1983-2007. For industry-years with no acquisition targets, the number and value of acquisitions is set equal to zero. Column (1) reports the results of Poisson (count) models relating the number of acquisition bids in the industry to lagged Union and control variables. Column (2) reports the results of Poisson (count) models relating the number of completed acquisitions in the industry to lagged Union and control variables. Column (3) reports the results of Tobit models relating the value of completed acquisitions in the industry scaled by the industry’s assets to lagged Union and control variables. Union is defined in Table II. IndLogMCap is the industry median of the logarithm of market capitalization; IndLogMBE is the industry median of the logarithm of the market-to-book equity ratio; IndROA is the industry median ROA; IndTang is the industry median ration of fixed assets to total assets; IndDivPay is the fraction of dividend paying firms in the industry; IndFinlev is the industry median book leverage; and LogNoFirms is the logarithm of the number of firms in the industry. All regressions include year fixed effects (Year FE). The absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. (1) (2) (3) # Acquisition Bids # Completed Acquisitions Value Acquisitions/Assets Union -0.964* -1.345** -0.045*** (1.95) (2.45) (3.01) IndLogMCap 0.034 0.035 -0.006*** (0.70) (0.66) (3.04) IndLogMBE 0.308*** 0.326*** -0.001 (3.01) (3.15) (0.33) IndROA 0.460 0.522 0.079*** (0.42) (0.46) (2.60) IndTang -0.752*** -0.957*** -0.013 (3.04) (3.46) (1.04) IndDivPay -0.474** -0.438* 0.003 (2.14) (1.94) (0.24) IndFinlev 0.792** 0.776** 0.004 (2.32) (2.23) (0.22) LogNofirms 0.928*** 0.941*** 0.014*** (11.47) (11.01) (7.97) Constant -1.214** -1.646*** -0.035*** (2.50) (3.20) (2.77) Estimation Poisson Poisson Tobit Year FE Yes Yes Yes # of obs. 4427 4427 4427

45

46

Table X. The Effect of Unionization on Yield Spreads and Antitakeover Defenses

Panel A of the table reports OLS regressions of Spread on Union and the following control variables defined in Table II: Size, Finlev, TobQ, Cfrisk, Intcov1, Profit, Negequity, Intang, NWC, Totpayout, Productivity, HHI, Nasdaq, Duration, Debtage, and Rating. All regressions include one-digit SIC industry fixed effects (SIC1 FE) and year fixed effects (Year FE). Column (1) reports the results for firms with G-Index above the sample median and Column (2) reports the results for firms with G-Index below the sample median. Higher values of the G-Index are associated with more antitakeover defenses. The coefficients of the standard control variables are omitted for brevity. Panel B of the table reports OLS regressions of Spread on Union, Union×Distress, Distress, and the control variables. We use three alternative versions of the financial distress dummy variable which are defined in Table 4: LowZS, HighOS, and HighPD. We demean Union and each of the above financial distress indicators before forming the interaction terms. All regressions include CIC industry fixed effects (CIC FE) and year fixed effects (Year FE). Columns (1), (3), and (5) report the results for firms with G-Index above the sample median and Columns (2), (4), and (6) report the results for firms with G-Index below the sample median. The coefficients of the standard control variables are omitted for brevity unless they are interacted with Union. In both panels, the absolute values of the t-statistics, reported in parentheses, are based on standard errors robust to heteroskedasticity and clustered at the CIC industry level. *, **, and *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

Panel A: Benchmark Regressions for High and Low G-Index Subsamples

(1) (2) High G-Index Low G-Index Union -102.453* -252.935*** (1.73) (2.79) Controls Yes Yes SIC1 FE Yes Yes Year FE Yes Yes # of obs. 2387 3170 R2 0.38 0.35

Panel B: The Effect of Unions Conditional on Financial Distress in High and Low G-Index Subsamples

(1) (2) (3) (4) (5) (6) High G-Index Low G-Index High G-Index Low G-Index High G-Index Low G-IndexUnion -235.399 -482.044 -293.828 -575.612 -262.228 -475.441 (0.95) (1.07) (1.25) (1.09) (1.18) (1.06) Union × LowZS -313.631* -301.204** (1.79) (2.55) LowZS 7.023 -26.394 (0.21) (0.45) Union × HighOS -276.526* -509.127*** (1.67) (3.34) HighOS -15.340 -5.548 (0.53) (0.16) Union × HighPD -348.385*** -347.292*** (2.80) (3.18) HighPD 28.803 -3.800 (1.44) (0.10) Controls Yes Yes Yes Yes Yes Yes CIC FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes # of obs. 2378 3155 2273 2869 2387 3169 R2 0.43 0.36 0.42 0.37 0.43 0.36