University of Southern California Dissertations and Theses

Slashing liquidity through asset purchases: evidence from collective bargaining

(USC Thesis Other)

Slashing liquidity through asset purchases: evidence from collective bargaining

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Slashing Liquidity through Asset Purchases:
Evidence from Collective Bargaining
by
Irene Yi
A Dissertation Submitted to the
FACULTY OF THE GRADUATE SCHOOL,
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree of
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)
August 2017
Copyright 2017, Irene Yi
Abstract
Using a hand-matched data set on 27,284 union contracts, I provide novel evidence on the
strategic use of corporate liquidity in contract negotiations with unions. I focus on the idea
that rms have incentives to hold low levels of liquid assets during union negotiations, because
high liquidity can encourage unions to raise their wage demands. The main nding is that
rms reduce liquidity before contract negotiations primarily through increased asset purchases.
Firms increase asset purchases as a fraction of total assets by one-third before contract negoti-
ations, and nance those purchases by a reduction in cash balances and an increase in leverage.
Firms do not increase investments, R&D, dividends, or repurchases before contract negotia-
tions. Strategic liquidity management is associated with lower wages. The evidence indicates
that rms reduce liquidity to gain strategic advantages in contract negotiations with unions in
ways that simultaneously allow managers to maintain the level of resources under their control.
1
Acknowledgement
I am grateful to my advisor, John G. Matsusaka, as well as Kenneth R. Ahern, Andr as Danis,
Gerard Hoberg, Nan Jia, Arthur Korteweg, Kevin J. Murphy, Vikram Nanda, Paige Ouimet,
Oguzhan Ozbas, Gordon Phillips, Tingting Que, Zacharias Sautner, Andrei Shleifer, and Toni
Whited for helpful discussions and suggestions. I also thank the participants at the Young
Scholars Finance Consortium, the European Finance Association Annual Meeting, the North-
ern Finance Association Conference, the Financial Management Association Annual Meeting,
and the Western Finance Association Annual Meeting, and the seminar participants at USC
Marshall, Louisiana State University, Indiana University, and Loyola Marymount University
for valuable comments. I thank the Katz Research Fellowship and USC for nancial support.
2
Contents
1 Introduction 6
2 Conceptual Framework 11
2.1 Incentives to Reduce Liquidity before Union Negotiations . . . . . . . . . . . . . 11
2.2 Liquidity Reduction Channel: Managerial Control Hypothesis . . . . . . . . . . 12
3 Data 14
3.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.1 Contract-Expiration Years . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.2 Liquidity Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Empirical Specication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4 Main Results 18
4.1 Strategic Liquidity Management . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Liquidity Management Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Asset Purchase as the Main Liquidity Management Channel . . . . . . . . . . . 21
4.4 Asset Sales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5 Robustness and Additional Evidence 25
5.1 Work Stoppages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.1.1 Asset Purchases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3
5.1.2 Asset Sales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.2 Falsication Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3 Contract Expirations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3.1 Alternative Measures of Contract Signicance . . . . . . . . . . . . . . . 28
5.3.2 Length of a Labor Contract . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.4 Are Asset Purchases Indeed Strategic? . . . . . . . . . . . . . . . . . . . . . . . 30
5.5 Can Asset Purchases be Reversed? . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.6 Time-Series Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6 Channels behind the Increase in Asset Purchases 34
6.1 Management Control Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2 Dividends, Investments, and R&D Adjustment Costs . . . . . . . . . . . . . . . 36
7 Collective Bargaining Outcomes 38
8 Conclusion and Discussion 40
References 41
A Description of Data 58
B Additional Tables and Figures 61
4
List of Figures
1 Contract-Expiration Year: Variable Construction . . . . . . . . . . . . . . . . . 16
2 Cumulative Frequency of Asset Purchases . . . . . . . . . . . . . . . . . . . . . 21
3 Asset Purchases and Sales: Time-Series Patterns . . . . . . . . . . . . . . . . . . 33
4 Histogram of AR(1) Coecients . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
B.1 Frequency of Contract-Expiration Years . . . . . . . . . . . . . . . . . . . . . . 63
B.2 Length of a Labor Contract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5
1. Introduction
Companies hold cash and other liquid resources largely to manage routine operational expenses
and exploit unforeseen investment opportunities (Baumol (1952), Graham and Harvey (2001),
Almeida, Campello, Cunha, and Weisbach (2014)). While the benets of nancial
exibility
can be substantial, high levels of corporate liquidity might create strategic vulnerabilities with
respect to the company's stakeholders. In particular, high levels of liquidity can encourage union
employees to raise their wage demands in collective-bargaining negotiations. As such, theory
suggests that rms facing union negotiations will increase leverage to reduce the funds available
to labor unions (Bronars and Deere (1991), Perotti and Spier (1993)). Empirical evidence
indicates that rms in more unionized industries maintain lower liquidity by holding smaller
cash reserves and maintaining higher leverage (Klasa, Maxwell, and Ortiz-Molina (2009), Matsa
(2010)).
1
While empirical ndings show a correlation between union bargaining power and corporate
liquidity, there is surprisingly little evidence on whether rms strategically remove liquidity to
improve their bargaining positions in union negotiations. Most empirical studies use industry-
level unionization rates to proxy for a particular union's ability to in
uence the rm with which
it is negotiating, although there are sucient variations between rms within an industry in
terms of union bargaining power. Moreover, although rms might have greater incentives to
reduce liquidity at times with higher stakes, the literature assumes that the impact of union
bargaining power on corporate liquidity is identical in years with or without contract negotiation
(Klasa et al. (2009), Matsa (2010)). As most studies do not identify precisely when and to what
extent rms have incentives to behave strategically, the causal impact of strategic incentives
on corporate liquidity has not been fully assessed, despite the fact that the connection between
union bargaining power and liquidity is widely accepted.
More importantly, the question of how managers reduce liquidity appears to have been
neglected. Managers have choices as to how to reduce liquidity: paying dividends, repurchasing
1
Additionally, Matsa (2010) and Marciukaityte (2015) show that rms decrease leverage when they expe-
rience a state-level legal shock (i.e., right-to-work law) that weakens union bargaining power.
6
shares, increasing investments, and purchasing assets. The decisions managers make in reducing
liquidity aect the scope of a rm and the nature of its businesses, as well as purely nancial
eects. Agency theory suggests that managers enjoy private benets of control and tend to
maintain the resources under their control. Based on this, I develop the following \managerial
control hypothesis": managers reduce corporate liquidity in order to improve their bargaining
positions in union negotiations, but only in ways that do not reduce the level of resources under
their control. The managerial control hypothesis predicts that managers reduce liquidity by
increasing either investments or asset purchases, but avoid increasing dividends or repurchases.
This is because investments and asset purchases reduce liquidity while maintaining the level
of resources under managers' control, whereas dividends and repurchases reduce the resources
under their control.
The goal of this paper is to provide well-identied estimates of strategic uses of corporate
liquidity in union negotiations, and to assess the managerial control hypothesis. In order
to answer these questions, it is crucial to measure exogenous shifts in rms' incentives to
adjust liquidity. My identication strategy is to exploit the exogenous timing of union contract
expirations, by constructing a hand-matched data set on 27,284 contracts between rms and
unions. Each union contract lasts three to ve years with few exceptions and contract-expiration
dates are essentially exogenous once they are established at the onset of the contract. The
exogenous timing of contract expirations provides an arguably clean setting to identify when
rms have greater incentives to behave strategically. With this setup, I examine whether
corporate nancial policies dier in years when rms negotiate contracts with a large number
of employees (\contract-expiration years") compared to other years (\non-contract-expiration
years").
The main nding is that rms reduce liquidity before union negotiations, mainly through
increased asset purchases. In a sample of 344 rms that negotiated at least one contract
with 500 or more employees between 1996 and 2012, I rst show that rms reduce liquidity
in contract-expiration years. Specically, rms reduce cash as a fraction of total assets by 0.5
percentage points in contract-expiration years. Firms increase leverage by 0.8 percentage points
in contract-expiration years, indicating that rms reduce expected future liquidity. Next, I show
7
that rms increase asset purchases in contract-expiration years. Specically, rms increase asset
purchases as a fraction of total assets by 1 to 1.2 percentage points in contract-expiration years,
which is about one-third relative to the sample mean. The probability of an asset purchase
increases by 2 to 5 percentage points in contract-expiration years. These ndings are robust to
rm and industry-year xed eects, controls, and alternative proxies for contract signicance.
I do not nd any evidence that rms increase investments, R&D expenditures, dividends, or
repurchases in contract-expiration years. Collectively, the evidence supports the hypothesis
that managers reduce liquidity to gain strategic advantages in union negotiations, but only in
ways that do not reduce the level of resources under their control.
The results suggest that rms strategically reduce liquidity before union negotiations. An-
other way to reduce liquidity would be to curtail planned asset sales. If holding more liquid
assets puts a rm into a tough negotiation situation, then the rm would be reluctant to engage
in an asset sale before union negotiations because an asset sale would bump up the rm's cash
balances. Consistent with this idea, I nd that the probability of an asset sale declines by 2 to
4 percentage points in contract-expiration years.
The identication of this paper relies on the assumption that the timing of contract ex-
pirations is exogenous to corporate nancial policies. Though expiration dates are set at the
onset of the contract and therefore provide a clean setting, I perform a number of analyses to
address potential remaining concerns. First, I perform a falsication test and show that the
changes in asset purchases and sales are apparent only in contract-expiration years, but not in
surrounding years when rms do not have a reason to behave strategically. Second, I verify
that contract expirations are exogenous by showing that the length of a labor contract is unre-
lated to macroeconomic or rm-specic conditions. Next, I show that managers further distort
corporate nancial policies during periods with heightened con
ict of interests, measured by
work stoppages (mostly strikes, but also lockouts). In addition, I show that rms with higher
cash balances further increase asset purchases in contract-expiration years. Along the same
lines, the increase in asset purchases during contract-expiration years is pronounced for rms
in manufacturing industries { those with the strongest union power.
8
A key contribution of this paper is to show that asset purchase is the main channel through
which rms remove corporate liquidity before union negotiations. In this regard, I attempt to
provide a better understanding of why rms increase asset purchases but not dividends, R&D,
investments, and repurchases. First, I show that younger managers, having longer careers ahead
of them and presumably greater incentives to expand their businesses, further increase asset
purchases in contract-expiration years. Next, I suggest that adjustment costs could be another
non-mutually exclusive reason behind the increase in asset purchases and non-increase in oth-
ers. For instance, if it is costly for rms to radically increase dividends in one year and curtail
thereafter, rms might choose not to increase dividends in response to temporary bargaining
considerations. Although it is beyond the scope of this study to provide a comprehensive anal-
ysis on adjustment costs, I measure the serial correlation in nancial policies to shed light on
adjustment costs. I nd that dividends, R&D, and investments are much more persistent com-
pared to asset purchases, which is consistent with the literature ndings on dividend and R&D
smoothing (Lintner (1956), Brown and Petersen (2011)). My results suggest that adjustment
costs associated with dividends, R&D, and investments can explain why rms divert corporate
liquidity to asset purchases before union negotiations.
While the main focus of this paper is on rms' strategic behavior, I also provide some sug-
gestive evidence on the outcome of such strategic behavior. Intuitively, one can expect union
contracts to be more favorable to rms that strategically reduced liquidity before union nego-
tiations compared to rms that did not act strategically. I nd that the annual wage increase
rate is 0.39 percentage points lower in contract-expiration years in which rms reduced liquidity
through asset purchases, compared to the 2.74% wage increase rate in a regular contract expi-
ration year. In contrast, if a rm goes against the strategic direction (i.e., increased liquidity
and did not purchase any assets) in a contract expiration years, the annual wage increase rate is
0.45 percentage points higher, compared to a regular contract-expiration year. The evidence is
broadly consistent with the idea that strategic liquidity reduction helps rms at the bargaining
table.
This paper contributes to the literature on the strategic interactions between rms and labor
unions (Matsa (2010), Klasa et al. (2009), Chen, Kacperczyk, and Ortiz-Molina (2011)). To my
9
knowledge, this paper provides the rst direct evidence that the incentives of rms to improve
their bargaining positions have a causal impact on corporate nancial policies. Several recent
papers show that rms in more unionized industries hold less cash (Klasa et al. (2009)) and
maintain higher leverage (Matsa (2010)). Schmalz (2015) shows that an average rm reduces
cash and increases leverage following a new unionization compared to an average rm escaping
from unionization, but the impact varies by rm characteristics. My paper is also related to the
small stream of research that studies strategic behavior by rms and unions before collective-
bargaining expirations (DeAngelo and DeAngelo (1991), Klasa et al. (2009), Huang, Jiang, Lie,
and Que (2015), Matsusaka, Ozbas, and Yi (2017)).
This paper is also related to the literature on the costs and benets associated with holding
liquid assets (Jensen (1986), Kim, Mauer, and Sherman (1998), Opler, Pinkowitz, Stulz, and
Williamson (1999), Duchin, Ozbas, and Sensoy (2010)). Prior literature focuses on the agency
costs coming from holding excess liquidity (e.g., Jensen (1986), Harford (1999)). My results
along with Klasa et al. (2009) show that excess liquidity weakens rms' bargaining position
against labor unions. This paper contributes to the literature on corporate liquidity by providing
novel evidence on how rms reduce liquidity before union negotiations.
Lastly, the paper builds on the voluminous literature on corporate asset purchases, mergers
and acquisitions, and asset sales (Schlingemann, Stulz, and Walkling (2002), Maksimovic and
Phillips (2001), Yang (2008)). This paper oers new insights into the literature by suggesting a
novel and causal determinant of corporate asset purchases. My results also complement several
recent papers that focus on the nancing implications of asset sales (Arnold, Hackbarth, and
Puhan (2015), Edmans and Mann (2013), Borisova and Brown (2013)).
10
2. Conceptual Framework
2.1. Incentives to Reduce Liquidity before Union Negotiations
The rst implication I test is whether rms strategically reduce corporate liquidity before
union negotiations. This subsection provides an illustration of the conceptual framework and a
summary of the assumptions required to predict the main ndings documented in this paper.
The main premise is that high levels of corporate liquidity can encourage unions to raise their
wage demands. Therefore, by reducing corporate liquidity and holding fewer liquid assets, a
rm can make a more credible case that concessions to a union will increase the probability of
bankruptcy. Because bankruptcy is costly to the union, the union is more willing to settle for
lower wages when negotiating with a rm holding low levels of liquid assets.
This view receives support in the theoretical literature, which suggests that rms have
incentives to use debt nancing in order to reduce the expected future liquidity and consequently
preserve shareholder wealth. For example, Bronars and Deere (1991) show that rms have
incentives to issue debt in order to reduce the funds available to a potential union. Perotti and
Spier (1993) also demonstrate that, under certain conditions, rms with higher leverage can
oer lower wages to union employees.
2
The bargaining assumption behind the main implication is that a union takes a constant
portion of the total amount of liquid resources within the rm (Bronars and Deere (1991)). I
assume the portion to be constant, as it is theoretically unclear whether and how the portion
should vary over time or how it will be aected by the amount of corporate liquidity.
3
I assume that the amount of liquidity over the life of the contract, known to the labor union
at the time of its contract negotiation, determines a rm's bargaining position. For example,
2
Also see Dasgupta and Sengupta (1993), Hanka (1998), Sarig (1998) for discussions on the strategic use
of leverage vis- a-vis labor unions. Empirically, Benmelech, Bergman, and Enriquez (2012), Qiu (2016), and
Ouimet and Simintzi (2015) show that rms with low levels of nancial resources can oer lower wages to union
employees.
3
Several empirical studies examine the eects of exogenous shifts in the portion parameter, following the
passage of the right-to-work law (Chava, Danis, and Hsu (2016)) and the passage of union elections (Lee and
Mas (2012), Schmalz (2015)).
11
reported cash holdings, realized earnings, and prots at the time of negotiation contribute to
the increase in liquidity. Alternatively, leverage increases the demand on future cash
ows in the
form of interest payments, thus lowers expected future liquidity over the life of the contract. I
assume the following are not considered measures of liquidity: (i) alternative sources of liquidity,
such as debt capacity or lines of credit, and (ii) uncertain streams of cash
ows over the life
of the contract, such as risky investments. Although it is plausible that such channels could
add to a rm's liquidity and aect its collective-bargaining outcome, the liquidity stream is
subject to managerial discretion as well as macroeconomic and rm-specic risks. Therefore,
the stream is largely unpredictable at the time of a union's contract negotiation.
2.2. Liquidity Reduction Channel: Managerial Control Hypothesis
The key implication I test is how rms reduce corporate liquidity, given that rms plan to
reduce liquidity before union negotiations. Managers can reduce liquidity in a variety of ways,
such as: paying dividends, repurchasing shares, increasing investments, and purchasing assets.
The choices managers make in reducing liquidity have implications for rms' real activities, as
well as purely nancial eects.
Agency theory asserts that managers, whose interests are not perfectly aligned with the
interests of shareholders, tend to overinvest and cause rms to grow beyond the optimal size
because managers enjoy increasing the resources under their control. Jensen (1986) argues that
managers would not voluntarily pay out free cash
ows to shareholders, as payouts reduce the
resources under managers' control, thereby reducing managers' power.
Agency theory can be connected to the liquidity reduction channel in the following way:
managers might be reluctant to reduce liquidity before union negotiations if doing so leads
to a simultaneous decrease in the level of resources under managers' control. In contrast, if
the liquidity reduction can be achieved while still retaining the managers' private benets of
control, then managers would be more inclined to choose such channels. Based on this argument,
I propose the following managerial control hypothesis: managers remove liquidity before union
negotiations, but only in ways that do not reduce the level of resources under their control.
12
The managerial control hypothesis predicts that managers reduce liquidity by increasing either
asset acquisitions or investments, but not by increasing dividends or share repurchases. This
is because asset acquisitions and investments reduce corporate liquidity while maintaining the
level of resources under managers' control, whereas dividends and repurchases decrease both
liquidity and the resources under managers' control.
The managerial control hypothesis is consistent with existing theoretical discussions and
empirical ndings. Jensen (1986) argues that asset acquisition is the major channel through
which managers spend cash instead of paying it out to shareholders. Blanchard, Lopez-de
Silanes, and Shleifer (1994) discuss how managers use cash windfalls, focusing on whether they
choose to pay out the cash or retain it under their control. They predict that managers with
agency problems will use the cash windfall to invest in projects they like rather than distribute it
to shareholders. Specically, managers increase asset purchases and may, or may not, increase
investments depending on investment opportunities, but they do not increase dividends or
repurchases. Blanchard et al. (1994) empirically test these predictions by examining a sample
of 11 rms that won lawsuits, and nd that the sample rms increased asset acquisitions but did
not increase investments, dividends, or repurchases. Harford (1999) nds that rms with excess
cash are more likely to attempt acquisitions and such acquisitions are more value-destroying,
as measured by operational performance and stock returns. Lamont (1997) suggests that rms
tend to overinvest in poorly-performing segments, based on a sample of diversied rms that
were aected by the oil shock in 1986.
It should be noted that the managerial control hypothesis assumes that asset purchases and
investments do not generate predictable streams of liquidity during the life of the contract.
This assumption holds if: (i) the value created by asset purchases or investments is illiquid, (ii)
there is a reasonable time lag between the timing of asset purchases/investments and the cash

ow generated by asset purchases/investments, or (iii) the cash
ow cannot be predicted by
labor unions. In the empirical analyses that follow, I explore the channels through which rms
reduce liquidity and assess the managerial control hypothesis.
13
3. Data
3.1. Data Sources
The data on labor contract expirations, work stoppages, collective bargaining outcomes, cor-
porate liquidity, asset purchases and sales, and rm nancial information come from several
dierent sources. In this section, I outline the main features of the data sources. The details
are in Appendix A.
The data on labor contract expiration, work stoppages, and collective bargaining outcomes
come from BNA Labor Plus, maintained by the Bureau of National Aairs (BNA). Under the
National Labor Relations Act, rms with union contracts are required to le notices of contract
expiration with the Federal Mediation and Conciliation Service (FMCS). These lings report
employer names, union names, contract expiration dates, and the number of employees involved
in the collective bargaining (henceforth \contract employees"). I focus on rms that have at
least one contract with 500 or more contract employees, but my sample includes all contracts
in which these rms were engaged, even if a contract involves a small number of employees. I
identify 27,284 contract expirations from 377 rms during the period 1995{2014. An example of
a contract expiration is in Appendix Table B.1. I also collect information on 475 work stoppages
and collective bargaining outcomes from the same 377 rms during the period 1997{2014. A
total of 172 rms experienced at least one work stoppage. My sample contains 1,389 wage
settlement observations for 261 rms during the period 1996{2014.
The data on asset purchases and sales come from two separate sources: SDC Platinum
and Compustat. The unit of observation is a transaction in SDC Platinum. Each observation
reports the buyer and the seller, transaction date, transaction value, and the type of transaction
(e.g., merger, asset acquisition, buyback, or exchange oer). The sample consists of both U.S.
and non-U.S. transactions that are categorized as a merger or an asset acquisition. Either the
buyer or the seller can be a private rm. The transaction value is missing for nearly half of
the transactions, especially for transactions that involve private or non-U.S. rms. Instead of
14
relying on transaction value in SDC Platinum, I obtain data on the amount of annual asset
purchases from Statement of Cash Flows. Firm nancial information comes from Compustat.
Variable denitions can be found in Appendix Table B.2.
There were several challenges in the merging the SDC Platinum, BNA, and Compustat
databases. First, the contract-listing database does not have any rm identiers, so rms need
to be identied by their names as they appear on the BNA lings. Hand-matching is necessary
for this process because dierent names and abbreviations are used in the BNA and Compustat
databases.
4
Second, the names in the BNA database are often at a plant or subsidiary level,
and third, some plants change ownership during the sample period. In such cases, I identify
the correct owner at the point in time by referring to online resources. Finally, six-digit CUSIP
numbers in the SDC Platinum database are often dierent from the ones that can be found in
Compustat, in which case I manually matched them using ticker as an additional source of the
matching identier. The nal panel data set has 4,946 rm-year observations from 344 rms
during 1996{2012. My sample covers a signicant fraction of major U.S. companies: About
60% of the sample rms (=206/344) were part of the S&P 500 index at some point.
3.2. Variables
This section provides descriptions of the key variables. Table 1 presents summary statistics for
all major variables examined in this paper.
3.2.1. Contract-Expiration Years
In order to identify years in which rms have greater incentives to engage in strategic behavior,
I construct my key explanatory variable, contract-expiration year, in the following way: for
each rm i, I compute the time-series average
i
and the standard deviation
i
of the number
of contract employees (i.e., employees under contract expirations). If the number of contract
4
The examples of the names that appear in the BNA ling for General Motors Company include obvious
cases such as Metal Fabricating Division GM and General Motors Service Parts PDC 6, and less obvious cases
such as GMPT Livonia and Union Pontiac GMC Buick.
15
Figure 1. Contract-Expiration Year: Variable Construction. The gure shows how the
key explanatory variable, contract-expiration year, is constructed, using an example of a sample
rm. First, I sum the number of contract employees (i.e., employees under contract expirations)
over all contracts expiring in the same year (as stacked in the gure). The height for each year
shows the total number of contract employees in a given year. Next, I compute the rm-specic
time-series mean
i
and standard deviation
i
of the number of contract employees. For the
sample rm,
i
+
i
= 2253 + 3617 = 5; 870. There were more than 5,870 contract employees
in 1996 and 2010. The years before, 1995 and 2009, are contract-expiration years. See Section
3.2.1 for details.
employees in any given year is greater than one standard deviation above the mean (i.e.,
i
+
i
)
in year t+1, then year t is identied as a contract-expiration year for the rm i. Figure 1
graphically shows the variable construction process using a sample rm: Alcoa Inc. Note that
there is a one-year dierence between the actual year of contract expiration and the year that
is identied as a contract-expiration year. This setup is to ensure that regression coecients
do not capture rms' actions that happen after contract expiration, and to identify years when
strategic actions are most likely to take place. Also note that 14% of rm-years are contract-
expiration years (Table 1). Appendix Figure B.1 presents the histogram and the incidence of a
contract-expiration year.
16
3.2.2. Liquidity Measures
The main measure of corporate liquidity in this paper is cash (and also cash equivalents) as
a fraction of rm total assets, which comes from Compustat. While rms have alternative
mechanisms to manage corporate liquidity such as lines of credit, debt capacity, or derivatives,
these options do not oer the same degree of protection as cash holdings.
5
Hence, the conclusion
from the literature is that cash remains the predominant measure of corporate liquidity.
In addition to cash, I also consider leverage as a measure of corporate liquidity. Although
leverage is not commonly considered a measure of liquidity, it serves as a good measure in
the context of this study: leverage increases the demands on rms' future cash
ows, and
thereby credibly reduces expected future liquidity. Not surprisingly, leverage has been discussed
repeatedly in the literature as a strategic bargaining tool against labor unions. While leverage
reduces corporate liquidity after the time of debt issuance, it increases cash
ow at the time of
debt issuance. Therefore, in order for leverage to empirically work as a bargaining tool, it is
crucial to establish that the increase in cash
ow at the time of debt issuance is expensed away.
Section 4.3. presents evidence on whether, and how, the cash
ow is expensed away.
3.3. Empirical Specication
The following regression is the main specication used to examine a rm's corporate nancial
policies:
y
ijt
= +D
ijt
+X
ijt1
+
i
+
jt
+
ijt
(1)
where i indexes a rm, j indexes an industry, and t indexes time. y
ijt
is a corporate nancial
policy variable (e.g., asset purchases as a fraction of total assets, cash as a fraction of total
assets). D
ijt
is a dummy equal to one if year t is a contract-expiration year for rm i in
industry j, and zero otherwise. The vector X
ijt1
captures various controls, lagged by one
year. The rm and industry-year xed eects are
i
and
jt
, respectively, and
ijt
is the error
term. Industry-year xed eects are applicable for industry-year pairs that contain more than
5
See the survey by Almeida et al. (2014) for a detailed discussion.
17
ve observations per industry-year, based on the Fama{French 49 industry classication. This
restriction is necessary because my sample size is relatively small compared to the number of
available industry-year pairs, in which case high-dimensional xed eects can be impractical.
The identication of this paper relies on the assumption that the timing of union contract
expirations is exogenous and therefore, rms do not have any reasons to alter their corpo-
rate nancial policies in contract-expiration years, except for the union contract expiration.
Contract-expiration dates are set at the initiation of the contract, essentially exogenous once
they are established at the onset of the contract. Each contract typically lasts three to ve years
with few exceptions and there is little evidence that contracts are renegotiated before the stated
expiration date. The exogenous timing of labor contract expirations provides an arguably clean
setting to identify when rms have greater incentives to engage in strategic behavior.
I estimate equation (1) with a linear regression except when the dependent variable is the
probability of an asset purchase or the probability of an asset sale, in which case I use a linear
probability model. The linear probability model makes it easier to (i) implement xed eects,
(ii) interpret coecients, and (iii) cluster the standard errors, compared to logit or probit
models. In all regressions, standard errors are clustered at the rm level.
4. Main Results
4.1. Strategic Liquidity Management
The rst implication I test is whether rms remove liquidity before union negotiations. Table
2 reports linear regressions of corporate liquidity measures on contract-expiration years. In
regressions (1) and (2), the dependent variable is cash and cash equivalents as a fraction of rm
total assets. In regressions (3) and (4), the dependent variable is the leverage ratio, dened as
the total amount of debt as a fraction of rm total assets. The unit of observation is a rm-
year and the panel runs from 1996 to 2012. All regressions include rm and industry-year xed
18
eects, so that the key coecients are based on within-rm and within-industry-year variation
in contract-expiration status. In all regressions, standard errors are clustered at the rm level.
Regression (1) indicates that rms decrease cash holdings as a fraction of total assets by
approximately 0.5 percentage points in contract-expiration years. This decrease is about 8%
relative to the sample mean. Regression (2) adds the lagged logarithm of rm total assets and
thereby controls for rm size. The conclusion is similar in terms of coecient magnitude and
statistical signicance: cash as a fraction of total assets declines approximately 0.4 percentage
points in contract-expiration years. Note that the estimates in regressions (1) and (2) are of
similar magnitude to the one reported in Klasa et al. (2009). They obtain an estimate of 0.4
percentage points using a similar identication strategy, although their sample and denition of
contract expiration are dierent from the ones in this paper. Their evidence suggests that rms
strengthen their bargaining position by holding less cash at all times rather than managing
cash levels downward prior to negotiations.
Regression (3) and (4) report the results for leverage.
6
Regression (3) indicates that rms
increase leverage by 0.9 percentage points in contract-expiration years, which is about a 3%
increase relative to the sample mean. Regression (4) indicates that rms increase leverage by
approximately 0.8 percentage points in contract-expiration years, after controlling for rm size.
To summarize, Table 2 shows that rms decrease cash holdings and increase leverage in
contract-expiration years. The evidence is consistent with the widely-accepted hypothesis which
has not been directly tested: rms strategically remove current and expected future liquidity
before union negotiations in order to gain strategic advantages in union negotiations.
4.2. Liquidity Management Channels
The evidence in the previous section indicates that rms strategically remove liquidity in
contract-expiration years. Managers can reduce liquidity in a variety of ways. For example, a
6
Leverage has been discussed repetitively as a strategic variable in labor markets (Matsa (2010), Myers and
Saretto (2010), Marciukaityte (2015)). To the best of my knowledge, no existing studies have explored whether
rms strategically adjust leverage before union negotiations.
19
rm can increase asset purchases and grow the scope of the rm, spend internally by increasing
investments or R&D, return funds to shareholders by paying dividends, or restructure the rm
through share repurchases. The managerial control hypothesis states that managers reduce
liquidity in ways that allow them to maintain the level of resources under their control. The
hypothesis predicts that managers increase either asset purchases or investments, but do not
increase dividends or share repurchases, in contract-expiration years. In this section, I pro-
vide an assessment of the managerial control hypothesis by investigating how managers reduce
liquidity in contract-expiration years.
Table 3 reports the results. Regression (1) indicates that asset purchases as a fraction of
total assets is 0.96 percentage points greater in a contract-expiration year compared to a non-
contract-expiration year. Regression (2) indicates that rms decrease dividends by 0.003 per-
centage points in contract-expiration years, but the coecient is statistically indistinguishable
from zero.
7
In regressions (3){(5), I nd no evidence that rms change the levels of share repur-
chases, capital expenditures, or R&D expenditures in contract-expiration years. Specically,
the point estimates in regressions (3){(5) indicate that rms decrease share repurchases by 0.06
percentage points, decrease capital expenditures by 0.4 percentage points, and increase R&D
expenditures by 0.006 percentage points in contract-expiration years, respectively. All the coef-
cients in regressions (3){(5) are statistically indistinguishable from zero. Regressions (6){(10)
are the same as regressions (1){(5), except that rm size is included as a control variable. The
conclusion remains the same: a signicant increase in asset purchases in contract-expiration
years, but no signicant increase or decrease in dividends, share repurchases, investments, and
R&D expenditures. The evidence supports the managerial control hypothesis: managers reduce
liquidity only in ways that do not reduce the level of resources under their control.
It should be noted that my results indicate that managers do not increase investments in
contract-expiration years. The managerial control hypothesis predicts that managers might
increase investments, because investments reduce liquidity while not reducing the level of re-
sources under managers' control. Blanchard et al. (1994) assert that in agency models, the
7
An increase in dividends can be a clear signal to the labor union that the rm is in good nancial health
and is trying to remove excess liquidity, creating an even tougher negotiation situation. DeAngelo and DeAngelo
(1991) nd that rms reduce dividends before union negotiations.
20
Figure 2. Cumulative Frequency of Asset Purchases. The gure presents the cumulative
frequency of asset purchases as a fraction of total assets.
rst-order objective of managers is asset purchases (termed `diversication' in their paper) and
rms might not increase investments if internal investment opportunities are poor. My evidence
on the increase in asset purchases and non-increase in investments is consistent with their inter-
pretation of agency theory. Section 6 discusses potential non-mutually exclusive reasons behind
the non-increase in investments.
4.3. Asset Purchase as the Main Liquidity Management Channel
The previous sections show that managers reduce liquidity in contract-expiration years, mainly
through increased asset purchases. In this section, I examine asset purchases in further detail
and quantify the extent to which rms increase asset purchases in contract-expiration years.
Before diving into parametric analysis, I rst compare asset purchases in contract-expiration
years and non-contract-expiration years (i.e., years that are not contract-expiration years).
Figure 2 presents the cumulative frequency of asset purchases as a fraction of total assets, in
21
contract-expiration years and non-contract-expiration years. The vertical distance between the
red and blue curves indicates that, on average, asset purchases as a fraction of total assets is
greater in contract-expiration years. For example, in contract-expiration years, roughly 17% of
the rm-year observations involved asset purchases greater than 5% of rm total assets. During
non-contract-expiration years, approximately 14% of rm-years involved asset purchases with
transaction values greater than 5% of rm total assets. This dierence is consistent throughout
most percentiles. The evidence indicates that rms purchase more assets in contract-expiration
years. However, it is possible that large transactions and contract-expiration years are both
correlated with unobserved variables, or concentrated in certain rms, years, or industries. In
the analyses that follow, I examine the pattern in Figure 2 in detail with multivariate regressions.
Regressions (1){(3) of Table 4 report linear regressions of asset purchases as a fraction of
total assets. Regressions (4){(7) are linear probability regressions of the probability of an asset
purchase. In all regressions, the key explanatory variable is a dummy equal to one if a given
year is a contract-expiration year. The unit of observation is a rm-year and the panel runs
from 1996 to 2012. All regressions include rm and industry-year xed eects. Standard errors
are clustered by rm in all regressions.
Regression (1) indicates that asset purchases as a fraction of total assets is 0.96 percentage
points greater in a contract-expiration year compared to a non-contract-expiration year.
8
The
economic magnitude is material: the increase during contract-expiration years (0.96 percentage
points) is about one-third relative to the sample mean (2.9%). Regression (2) adds the lagged
logarithm of rm total assets. After controlling for rm size, the coecient on a contract-
expiration year increases in magnitude and becomes more statistically signicant. Specically,
asset purchases as a fraction of total assets is 1.2 percentage points greater in a contract-
expiration year compared to a non-contract-expiration year. Regression (3) adds several nance
variables as controls that are commonly used in corporate nance research: cash, leverage, ROA,
and capital expenditures, all lagged one year. The results show similar patterns.
Regressions (1){(3) examine whether the level of asset purchases diers in contract-expiration
years. They do not necessarily imply that companies are motivated to make a purchase that
8
Regressions (1) and (2) of Table 4 are identical to regressions (5) and (10) of Table 2, respectively.
22
they would not have made if it were not for the contract expiration. In regressions (4){(7), I
examine whether the probability of an asset purchase diers during contract-expiration years,
using the transaction-level data from SDC Platinum. Specically, I estimate equation (1) with
linear probability regressions.
9
In regression (4), the dependent variable is a dummy equal to
one if a rm made at least one asset purchase in a given year, and zero otherwise. Regression
(4) indicates that a company is 3.5 percentage points more likely to purchase an asset in a
contract-expiration year than in a non-contract-expiration year. It should be noted that the
unconditional probability of an asset purchase in any given year is about 45% (Table 1), so
the increase in probability is more modest compared to the increase asset purchases shown in
regressions (1){(3). The pattern is similar in regression (5) that adds control variables.
In regressions (6) and (7), the dependent variable is equal to one if a rm made at least
one asset purchase but did not make any sales in a given year, and zero otherwise. I use this
specication because there are many rm-years with both purchases and sales, in which case it is
hard to determine if a rm's liquidity is increasing or decreasing as a consequence of purchases
and sales. Regression (6) indicates that a company is 4.7 percentage points more likely to
purchase an asset without making any sales in contract-expiration years. The pattern is similar
in regression (7) that adds nance control variables. To summarize, Table 4 shows that rms
increase asset purchases and the probability of an asset purchase in contract-expiration years.
It is worth putting together the results on cash, leverage, and asset purchases. Taking the
estimates at face value, the sum of the changes in cash balances (0.4{0.5 percentage points) and
leverage (0.8{0.9 percentage points) during contract-expiration years is roughly comparable to
the increase in asset purchases during contract-expiration years (1{1.2 percentage points). This
can be interpreted as rms combining (i) cash, and (ii) nancing from leverage increases, in
order to purchase assets in contract-expiration years. This result also alleviates the potential
concern that the increase in leverage might lead to an increase in cash
ow at the time of debt
issuance, therefore weakening, not strengthening, a rm's bargaining position: the nancing
from a leverage increase is expensed away through asset purchases.
9
As noted in Section 3.3, I estimate linear probability regressions because it is easier to (i) implement xed
eects, (ii) interpret coecients, and (iii) cluster the standard errors. The patterns and signicance levels are
essentially the same with a conditional logistic specication.
23
4.4. Asset Sales
If rms have incentives to reduce liquidity before union negotiations, as shown in the previous
sections, then they might be reluctant to engage in an asset sale in contract-expiration years.
This is because the proceeds from asset sales would lead to an increase in liquidity and might
create a tougher negotiation situation. As an additional check on rms' strategic incentives, I
examine whether rms distort their asset sales in contract-expiration years.
Table 5 reports the results. Each column is a linear probability regression. In regressions
(1) and (2), the dependent variable is a dummy equal to one if a rm made at least one asset
sale in a given year, and zero otherwise. As before, the key explanatory variable is a contract-
expiration year. Regression (1) indicates that a company was 3.5 percentage points less likely to
sell an asset in contract-expiration years. Recall from Table 1 that the unconditional probability
of an asset sale is 40%, so the decrease is about 9% relative to the sample mean. The pattern is
similar in regression (2), which adds nancial control variables. In regressions (3) and (4), the
dependent variable is a dummy equal to one if a rm made at least one asset sale but did not
make any purchases in a given year. Regression (3) indicates that the probability that a rm
sells an asset without making any purchases decreases by 2.3 percentage points in contract-
expiration years. The coecient on a contract-expiration year is not statistically dierent from
zero at conventional levels of statistical signicance, but the decrease in economic magnitude is
about 15% compared to the sample mean. The conclusion remains the same in regression (4)
that adds nancial control variables.
The evidence suggests that rms are reluctant to engage in an asset sale during contract-
expiration years. The situation resembles a debt overhang problem in a temporary manner.
Even if rms have good asset sale opportunities, if such opportunities arrive right before union
negotiations and if the proceeds from sales are subject to haggling with labor unions, then rms
might forgo some protable sale opportunities at the margin. The evidence on asset sales tells
a coherent story with the evidence on asset purchases: rms have strategic incentives to reduce
liquidity before union negotiations.
24
5. Robustness and Additional Evidence
5.1. Work Stoppages
Not all contract negotiations turn into a war. It is possible that the management and the union
have established a good rapport and there is no dispute at the bargaining table. Alternatively,
a lead company might negotiate a master agreement with a union, and other companies ne-
gotiating with the same union might simply follow the lead pattern. Firms have incentives to
distort their nancial policies only when there is a divergence of interests between the rm and
the union. As an additional check on the main implications, I examine asset purchases and
sales during contentious negotiations, as measured by the years with a work stoppage (typi-
cally a strike, but also includes lockouts). Although the decision to stop work is endogenous
as opposed to exogenous contract expirations, this test has an advantage of identifying times
with heightened con
ict of interest.
5.1.1. Asset Purchases
Panel A of Table 6 presents the estimates from linear regressions of asset-purchase variables
on an indicator variable as to whether a work stoppage occurs in a contract-expiration year.
The unit of observation is a rm-year and all regressions include rm and industry-year xed
eects. In regressions (1){(3), the dependent variable is asset purchases as a fraction of rm
total assets. Regression (1) indicates that a rm increased asset purchases as a fraction of total
assets by 1.45 percentage points in a contract-expiration year with a work stoppage, compared
to a non-contract-expiration year. This 1.45 percentage points increase is about 50% relative
to the sample mean (2.9 percent). In a contract-expiration year without a work stoppage, a
rm increased asset purchases as a fraction of total assets by 1.1 percentage points, compared
to a non-contract-expiration year. The coecient on the indicator variable for a contract-
expiration year with a work stoppage and the 0.35 percentage points dierence between the
two indicator variables are not statistically dierent from zero; one possible explanation for the
25
lack of statistical signicance is that only 1.2% of total rm-years are both contract-expiration
years and work-stoppage years (Table 1). Regressions (2) and (3) introduce rm size and
nancial variables as additional explanatory variables and show a similar pattern.
Regressions (4){(7) examine the probability of an asset purchase in a contract-expiration
year with a work stoppage. In regressions (4) and (5), the dependent variable is a dummy equal
to one if a rm made an asset purchase in a given year. Regression (4) indicates that a rm was
18.4 percentage points more likely to purchase an asset in a contract-expiration year with a work
stoppage, compared to a non-contract-expiration year. In a contract-expiration year without a
work stoppage, a rm was 2 percentage points more likely to purchase an asset compared to a
non-contract-expiration year. The 16.4 percentage points dierence is statistically signicant
at the 1% level. I nd similar results in regression (5) that adds several nance controls to
regression (4), and in regressions (6) and (7) where the dependent variable is a dummy equal
to one if a rm made at least one purchase but did not make any sales in a given year. The
evidence indicates that rms further increase asset purchases when contract negotiations are
contentious, giving additional support to the hypothesis that managers distort nancial policies
in order to gain strategic advantages in union negotiations.
5.1.2. Asset Sales
Panel B of Table 6 examines the probability of an asset sale during contentious negotiations.
Regression (8) indicates that a rm was 12.7 percentage points less likely to sell an asset in a
contract-expiration year with a work stoppage, compared to a non-contract-expiration year. In
a contract-expiration year without a work stoppage, a rm was 2.1 percentage points less likely
to sell an asset compared to a non-contract-expiration year. The pattern is similar in regression
(9) that adds nancial controls. In regressions (10) and (11), the dependent variable is a
dummy equal one if there is at least one asset sale but no asset purchases in a given rm-year.
Regression (10) indicates that in a contract-expiration year with a work stoppage, a rm was
14.9 percentage points less likely to sell an asset without making any asset purchases, compared
to a non-contract-expiration year. In a contract-expiration year without a work stoppage, a
rm was 0.7 percentage points less likely to sell an asset without making any purchases. The
26
coecient on the indicator variable for a contract-expiration year with a work stoppage and the
14.2 percentage points dierence between the two indicator variables are dierent from zero at
the 1% level. The evidence indicates that rms are especially reluctant to engage in an asset
sale during contentious negotiations, providing additional support to the idea that rms have
incentives to decrease liquidity before union negotiations.
5.2. Falsication Tests
Although the exogenous timing of labor contract expiration provides a clean setting to explore
when rms have greater incentives to engage in strategic behavior, it is still possible that both
contract expirations and asset purchases are correlated with variables that are unobservable to
researchers. For example, a rm might decide to increase its market share within the next few
years and hire thousands of employees, whose contracts will expire within the next three to ve
years. If the rm is actively purchasing assets when the contracts expire, one will observe a
correlation between contract expirations and asset purchases, but the purchases are not driven
by rms' incentives to gain strategic advantages in union negotiations.
In order to address this concern, I examine asset purchases in years surrounding contract-
expiration years. This falsication test has two important advantages over other placebo tests.
First, my falsication test is designed to verify that rms do not signicantly change their asset
purchases when they do not have incentives to reduce liquidity (e.g., the year after a contract
negotiation). Second, if any omitted variables (e.g., long-term growth plan) are spuriously driv-
ing the results, then it is likely that those omitted variables are correlated between consecutive
years. In such cases, one would nd a signicant relation not only in contract-expiration years
but also in surrounding years, and the signicant relation during contract-expiration years be-
comes spurious. Conversely, if I do not nd a signicant relation in surrounding years, then my
ndings are not likely to be driven by any persistent omitted variables.
Table 7 presents the falsication test results. In each column, the key explanatory variable
is the contract-expiration year, but with one key dierence: it lags or leads few years. For
example, if years 2000 and 2003 are contract-expiration years for a given rm, then the key
27
explanatory variable in regression (1) is equal to one in years 1999 and 2002, and zero otherwise.
Using the same logic, the key explanatory variable in regression (3) is equal to one in years
2001 and 2004, and zero otherwise. In all regressions, the dependent variable is asset purchases
as a fraction of total assets. Note that regression (2) of Table 7 is identical to the regression
(2) of Table 4, serving as a baseline for comparison. The purpose of the tests is to compare
the coecients in regressions (1), (3), and (4) with the coecient in regression (2) and conrm
that the coecient is statistically signicant in regression (2), but not in the other regressions.
Regression (1) indicates that rms make 0.3 percentage points more purchases in the years
before contract-expiration years compared to the other years. Regression (3) indicates that rms
make 0.12 percentage points less purchases in the years right after contract-expiration years.
Regression (4) indicates that rms make 0.27 percentage points less purchases two years after
contract-expiration years. The key coecients in regressions (1), (3), and (4) are all statistically
indistinguishable from zero.
10
The conclusion is that the increase in asset purchases is apparent
in contract-expiration years but not in surrounding years, conrming that my main ndings are
not spurious. The evidence also indicates that they are less likely to be driven by any omitted
variables that are persistent. In such cases, it is hard to explain why asset purchases increase
only in contract-expiration years but not in the previous or following years.
5.3. Contract Expirations
5.3.1. Alternative Measures of Contract Signicance
The ndings so far indicate that rms reduce liquidity in contract-expiration years, mainly
through increased asset purchases. In Table 8, I replace my key explanatory variable, the
contract-expiration year, with dierent proxies for contract signicance and verify that my
results are robust to alternative specications. The dependent variable is indicated at the top
of the table. Each number is a coecient from a single regression. To conserve space, I only
10
Results regarding the probability of an asset purchase and the probability of an asset sale exhibit a similar
pattern. They are presented in Figure 3.
28
report the coecient on the key explanatory variable. All regressions control for rm size and
include rm and industry-year xed eects.
The second row of Table 8 shows that in years with more than 3,000 contract employees,
rms decrease cash as a fraction of total assets by 0.69 percentage points, increase leverage
by 1.34 percentage points, and increase asset purchases as a fraction of total assets by 1.29
percentage points. It should be noted that the absolute magnitude of coecients is greater
than the baseline specication. This is because negotiations involving 3,000 or more contract
employees are rarer (11% of rm-years) compared to the incidence of a contract-expiration year
(14% of rm-years). In the next two specications, the contract employee thresholds are 1,000
and 500, respectively. In years with more than 1,000 contract employees, rms decrease cash
holdings by 0.57 percentage points, increase leverage by 0.78 percentage points, and increase
asset purchases as a fraction of total assets by 0.82 percentage points. In years with more
than 500 contract employees, rms decrease cash holdings by 0.41 percentage points, increase
leverage by 0.47 percentage points, and increase asset purchases as a fraction of total assets by
0.63 percentage points. The magnitude of the coecients decreases as the contract-employee
threshold becomes lower. This is expected, as relatively less important contracts are included
as the threshold is lowered. In the last specication, I consider a continuous measure of contract
importance, which is a variant of my key explanatory variable, the contract-expiration year. The
results indicate that as the number of contract employees increases by one standard deviation,
rms decrease cash balances by 0.29 percentage points, increase leverage by 0.37 percentage
points, and increase asset purchases as a fraction of total assets by 0.38 percentage points.
5.3.2. Length of a Labor Contract
The assumption maintained throughout the paper is that the timing of labor contract expira-
tions is exogenous and is not manipulated by rms. One can argue that rms choose dierent
contract lengths depending on macroeconomic conditions. For example, during nancial crisis,
rms might either want to reduce labor market uncertainty by choosing shorter contract length,
or lock in lower wages by choosing longer contract length. In order to address this concern, I
show that there is little variation in contract lengths across time, using a subsample of labor
29
contract agreements in the BNA Settlements database, which has information on both contract
initiation and expiration dates. Appendix Figure B.2 summarizes the information on the length
of a contract by contract-initiation year. The gure demonstrates that there is little dierence
in contract length across dierent initiation years.
In order to verify that contract lengths are unrelated to rm-specic conditions or macroe-
conomic conditions, I compute the correlation between contract length and rm characteristics,
such as age, size, and protability. Unreported analysis conrms that there is no relation be-
tween contract lengths and rm-specic variables. With a smaller subsample of renegotiated
previous contracts, the length of the second contract is exactly the same as the length of the
rst contract in 51% of cases, diering by one year or less in 83% of cases. If rms are choosing
contract lengths according to time-varying conditions, there is little reason for rms to choose
a contract length almost identical to the previous contract. To summarize, there is no evidence
showing that contract lengths are chosen by rms on a whim, and the conclusion is that my
key explanatory variable, the contract-expiration year, remains exogenous.
5.4. Are Asset Purchases Indeed Strategic?
Although the evidence so far supports the idea that rms strategically reduce liquidity through
increased asset purchases during contract-expiration years, no rm is going to admit that the
rm purchased an asset in order to improve its bargaining position in a union negotiation.
Therefore, among many asset purchases conducted for dierent reasons (e.g., to increase market
power, obtain synergy gains, exploit undervaluation), it is empirically impossible to identify
which purchase is motivated by strategic considerations. Nonetheless, I oer some additional
evidence supporting the view that asset purchases during contract-expiration years are strategic.
First, I examine whether the increase in asset purchases during contract-expiration years
is pronounced for rms with higher cash balances. Firms with high liquidity would be more
vulnerable to unions' wage demands, having greater incentives to reduce liquidity before nego-
tiations. By the same token, rms that are already low in liquidity, purposely or not, would
not have a strong incentive to reduce liquidity before negotiations. Table 9 reports coecients
30
from linear regressions of asset purchases on contract-expiration year, lagged value of cash bal-
ances, and their interaction. Regression (1) shows that the coecient on the interaction term
is positive and statistically signicant, indicating that rms with higher cash balances further
increase asset purchases during contract-expiration years. Regression (2) controls for rm size
and shows similar results.
Next, I show that the increase in asset purchases during contract-expiration years is pro-
nounced for rms in manufacturing industries (Table 10). It is widely believed that union
power is strongest in manufacturing industries, and contracts are usually larger for manufac-
turing rms. If the tension is especially strong for manufacturing rms, one can expect that
asset purchase becomes more of a strategic tool for those rms. Table 10 shows that the increase
in asset purchases during contract-expiration years is 1.56% for rms in manufacturing indus-
tries, compared to 1.2% for rms in all industries (regression (1)). The evidence in Tables 9 and
10 indicates that the increase in asset purchases during contract-expiration years is particularly
pronounced for rms that are more vulnerable if they do not engage in strategic behavior,
supporting the idea that asset purchases are indeed motivated by strategic considerations.
5.5. Can Asset Purchases be Reversed?
The evidence so far is based on the assumption that once liquidity is gone, it is gone for good.
If there are few costs involved in undoing completed asset purchases (i.e., reselling purchased
assets), or selling other existing assets, then an asset purchase before union negotiation might
not improve a rm's bargaining position. This is because rms that reduce liquidity through
asset purchases can simply resell the purchased assets, refuel liquidity, and prevent bankruptcy.
Although the costs associated with reselling assets are not directly observable, I explore whether
the reversibility can prevent rms from strategically purchasing assets in the rst place.
First, I examine the prevalence of an asset resale. Specically, I ask whether assets acquired
during contract-expiration years are more likely to be resold later, compared to assets purchased
during non-contract-expiration years, in order to test whether strategic asset purchases are more
likely to be reversed compared to regular asset purchases. For my sample rms, I identify 281
31
assets that were purchased and later resold to another company during 1996{2012. I do not
nd a statistically signicant dierence in the resale probability between the assets purchased
in contract-expiration years and assets purchased in non-contract-expiration years.
Another way to interpret the lack of signicant dierence in the resale probability is that
rms only engage in strategic behavior if they can benet from it. In other words, rms that
can easily resell assets do not use asset purchase as a strategic channel to reduce liquidity.
In a subsample of rms operating in manufacturing industries, I explore whether there is a
heterogeneity in the extent of strategic asset purchases across rms facing dierent degrees
of asset redeployability, using the measures of asset redeployability developed by Kim and
Kung (2014). Regression (1) of Table 11 shows that rms operating in industries with low
asset redeployability increase asset purchases by 2.86 percentage points in contract-expiration
years, compared to non-contract-expiration years. However, rms that operate in industries
with high asset redeployability increase asset purchases by 0.05 percentage points in contract-
expiration years (as shown in the bottom row), indicating that rms whose assets are highly
redeployable are less likely to increase purchases during contract-expiration years. Although the
analysis is based on a subsample of rms operating in manufacturing industries, the analysis
is still meaningful because rms in manufacturing industries are most likely to increase asset
purchases during contract-expiration years, as shown in regression (2) of Table 10.
Overall, the evidence relieves the concern that an asset purchase would not work as a
bargaining tool because rms can resell purchased assets to refuel liquidity: I do not nd any
evidence that asset purchases during contract-expiration years are more likely to be reversed
than those in non-contract-expiration years. Furthermore, rms that strategically purchase
assets are the ones facing higher barriers to reselling purchased assets.
5.6. Time-Series Patterns
Next, I examine the time-series patterns in asset purchases and sales in years surrounding
contract-expiration years. The left graph in Panel A of Figure 3 plots the point estimates from
regressions (1){(4) of Table 7. For example, the rst dot in the graph takes the value of 0.3,
32
Panel A. Asset Purchases
Panel B. Asset Sales
Figure 3. Asset Purchases and Sales: Time-Series Patterns. The gure presents
the time-series patterns in asset purchases and sales in years surrounding contract-expiration
years. Each dot represents the coecient estimate from a linear regression of the dependent
variable (indicated at the top of each graph) on an indicator variable for the year relative to
the contract-expiration year.

,

, and

denote statistical signicance at the 10%, 5%, and
1% levels, respectively. 95% condence intervals are plotted for each coecient estimate.
indicating that rms increase asset purchases as a fraction of total assets by 0.3 percentage
points in the years immediately before contract-expiration years. In the middle and right
graphs in Panel A of Figure 3, the dependent variable is a dummy equal to one if there is at
least one asset purchase in a given year, and a dummy equal to one if there is at least one
asset purchase but no asset sales in a given year, respectively. The three graphs in Panel A
tell a coherent story: for all three measures of asset purchases, rms increase asset purchases
in contract-expiration years, and the pattern reverses in the following years.
11
It is not obvious
11
One thing to note is that the signs and the magnitude of the coecients cannot be taken too seriously
since the coecients are usually statistically indistinguishable from zero in non-contract-expiration years. Such
statistical imprecision is exactly what one would expect to see from the previous falsication test in Table 7.
33
whether the reversal is a consequence of exhausting good purchase opportunities or low liquidity
resulting from increased purchases during contract-expiration years.
Panel B of Figure 3 reports the time-series patterns in asset sales in years surrounding
contract-expiration years. The graphs indicate that the probability of an asset sale sharply
decreases in contract-expiration years and increases after contract-expiration years, mirroring
the results in Panel A. One possible explanation for the pattern is that rms hold o asset
sales till union negotiations are over. Although the evidence in Figure 3 is not intended to
be conclusive, the time-series patterns suggest that short-term strategic incentives from labor
markets might in
uence asset purchases and sales not only in contract-expiration years, but
even after the union negotiations are over.
6. Channels behind the Increase in Asset Purchases
The key nding of this paper is that rms remove liquidity before large union negotiations
through increasing asset purchases, but not through increasing dividends, investments, R&D
expenditures, and repurchases. Although the evidence is consistent with the managerial control
hypothesis, it is also possible that other non-mutually exclusive channels are aecting managers'
decision to increase asset purchases, but not dividends, investments, R&D expenditures, and
repurchases. In this section, I provide evidence that further supports the managerial control hy-
pothesis and suggest another channel that might explain why rms only increase asset purchases
before large union negotiations.
6.1. Management Control Hypothesis
If managers indeed remove liquidity through asset purchases before union negotiations because
they enjoy increasing the resources under their control, one can expect the increase in asset
purchases to be more pronounced when managers have greater private benets of control.
Although it is hard to directly measure the degree of private benets of control, studies have
shown that the age of the CEO can be considered a plausible measure: managers with a long
34
career ahead of them have greater incentives to build a grand empire, compared to managers
whose empire have an imminent expiration date.
12
In Table 12, I test whether the increase in asset purchases during a contract-expiration year is
pronounced if the CEO of the rm is relatively young. I run a linear regression of asset purchases
as a fraction of total assets on contract-expiration year, the age of the CEO, and their interaction
term (regression (1)). Note that the variable of interest is the interaction term between the
age of the CEO and the dummy variable for contract-expiration year. The coecient on the
interaction term is negative and statistically signicant, indicating that younger managers are
further increasing asset purchases in contract-expiration years. During non-contract-expiration
years, I do not nd evidence that younger managers purchase more assets, as can be seen from
the non-negative coecient on CEO age.
In regressions (2) and (3), I create age categories instead of using age as a continuous
variable. Regression (2) shows that if the age of the CEO is 52 or below (which is approximately
20%), asset purchases as a fraction of total assets is 3.6 percentage points greater, compared
to a non-contract-expiration year. The 3.6 percentage points dierence is three times larger
than the 1.2 percentage points dierence in the baseline specication (Table 3 regression (6))
and the coecient is statistically signicant at the 1% level. In contrast, if the CEO is older
than 52, asset purchases in contract-expiration years increase to a lesser extent. Specically, if
the age of the CEO is between 53 and 63, or above 63, asset purchases as a fraction of total
assets are greater by 1 and 0.7 percentage points, respectively, compared to a non-contract-
expiration year. Although 1 and 0.7 percentage points are non-negligible considering the 2.9%
unconditional mean, the coecients are no longer statistically signicant, potentially because of
a shrinkage in sample size. Regression (3) allows the amount of asset purchases to vary by age
group in non-contract-expiration years and shows a similar pattern. Note that the coecients
on the last two rows are non-negative, which is in line with the non-negative coecient on
CEO age in regression (1). The interpretation is that the overall amount of acquisition is not
necessarily related to managers' private benets, but when there are funds to be spent for
bargaining considerations, managers use them while enjoying private benets of control.
12
See Yim (2013) for discussions and ndings on how CEO age would matter.
35
Figure 4. Histogram of AR(1) Coecients. Each graph presents the histogram of AR(1)
coecients from equation (2) for my sample rms, for variable names indicated at the top of
each histogram. Vertical gray lines indicate the median value of AR(1) coecients.
6.2. Dividends, Investments, and R&D Adjustment Costs
In addition to the managerial control hypothesis, I suggest that potential adjustment costs
can be another reason behind the increase in asset purchases during contract-expiration year
and the non-increase in dividends, investments, R&D expenditures, and repurchases. If it
is costly for rms to abruptly change a corporate nancial policy (e.g., dividends policy) in
one year and return to the usual levels, then rms might choose not to deviate from their
usual levels in a contract-expiration year. Dividend smoothing is one of the most well-known
empirical regularities in corporate nancial policy (Lintner (1956), Brav, Graham, Harvey, and
Michaely (2005)). Brav et al. (2005) note that 94% of CFOs in dividend-paying rms try to
avoid reducing dividends. Capital expenditures are relatively longer-term expenses compared to
operating expenses. R&D expenditures are also longer-term expenses which might even involve
upfront costs associated with hiring and training employees (Brown and Petersen (2011)). As
36
such, it might be costly for rms to divert the course of dividend policy or investment strategies
in response to temporary bargaining considerations. In contrast, an asset purchase or a stock
repurchase usually does not trigger another asset purchase or another repurchase, potentially
becoming suitable channels for rms to remove liquidity before union negotiations.
Although it is beyond the scope of this paper to examine the adjustment costs associated
with dividends, investments, and R&D expenditure, I examine their autocorrelation and provide
some suggestive evidence: if the costs associated with adjusting dividends, investments, and
R&D expenditures are high, one can expect a high serial correlation in dividends, investments,
and R&D expenditures. Conversely, if the autocorrelation is low, it is less likely that there are
high adjustment costs.
13
I use the following AR(1) framework to measure the autocorrelation:
y
it
= +y
it1
+
it
(2)
wherey
it
is a variable for a potential liquidity reduction channel. i indexes a rm and t indexes
time.
it
is the error term, following a zero-mean white noise process. , the AR(1) coecient,
is the coecient of interest.
Figure 4 presents the histogram of AR(1) coecients from equation (2), for variable names
indicated at the top of each histogram. Note that one AR(1) coecient is computed for each
rm, and the histogram aggregates the information across rms. The autocorrelation is low for
asset purchases (median =0:07). Taking the magnitude at face value, there is little reason
to expect that adjustment costs associated with asset purchases are high. For dividends, the
autocorrelation is much higher: the median of is 0.67, consistent with the empirical evidence
on dividend smoothing. For repurchases, capital expenditures, and R&D expenditures, the me-
dians of are 0.23, 0.48, and 0.67, respectively. The analysis indicates that dividends and R&D
expenditures are the most persistent among the variables examined, and capital expenditures
are less so but still somewhat persistent. The evidence indicates that adjustment costs can be
13
Note that this specication is dierent from what is typically used in the literature on dividend smoothing
(e.g., Lintner (1956)). The literature is based on the assumption that rms have target dividend ratios, while
this study abstracts from such assumption and focuses on measuring the degree of serial correlation.
37
another explanation for why managers increase asset purchases before union negotiations, but
not dividends, repurchases, investments, and R&D expenditures.
7. Collective Bargaining Outcomes
The evidence indicates that rms strategically reduce corporate liquidity through increased
asset purchases during contract-expiration years. An important related question is whether
such strategic behavior yields rms better bargaining outcomes, compared to the counterfactual
where rms did not act strategically. Although it is beyond the scope of this paper to document
the causal impact of strategic behavior on the collective bargaining outcomes, this section
presents some suggestive evidence showing that rms' eorts are related to lower wage increases.
My approach is to divide contract-expiration years into three categories: strategic (4%),
non-strategic (3%), and indeterminate (93%), considering (i) the changes in cash balances and
leverage from the previous year, (ii) the levels of cash balances and leverage in a contract-
expiration year, and (iii) the amount of asset purchases in a contract-expiration year. A
contract-expiration year becomes a candidate for a strategic year if a rm decreased cash bal-
ances, increased leverage, and purchased assets in a contract expiration year compared to the
previous year. Since this criterion inevitably includes many false positives, I also examine the
levels of cash balances and leverage (both the absolute level and the within-rm median level):
a highly levered rm with very low cash balances may experience a decrease in cash balances
and an increase in leverage, while the changes having nothing to do with the rm's strategic
incentives. Alternatively, a rm with high cash balances and no leverage can not expect to
signicantly improve its bargaining position if the decrease in cash balances and the increase
in leverage are both small: those years are reassigned to the indeterminate category. Finally, a
contract-expiration year is considered non-strategic if a rm increased cash balances, decreased
leverage, and did not purchase any assets.
Next, I link the information on strategic/non-strategic years to the hand-cleaned data on
collective bargaining outcomes from the BNA Settlements database. Settlement outcomes con-
38
sist of wage levels (e.g., $8/hour) or wage increase rates (e.g., 9% over three years), bonuses,
lump sum payments, retirement and health care benets, disability insurance, and sick leave. I
focus on wages, the most important component of the contract, because the other components
are hard to quantify. For each contract, I rst compute the average wage increase rate over the
life of the contract. Then, for each rm-year, I compute the average wage increase rate across
contracts renegotiated in the same year, as some rms negotiate multiple contracts with dif-
ferent unions or collective bargaining units. One caveat is that collective bargaining outcomes
are available for only 5% of contract expiration observations, which might lead to a potential
sample selection bias. However, my data set has an advantage in that it provides information
on the actual bargaining outcome, which is not directly observable in other databases that
report wages: plant or rm-level wages inevitably measure (i) the changes in wages caused by
the shifts in the types of employees and (ii) the wages of the employees that did not renegotiate
contract that year (recall that each union contract usually lasts three to ve years).
Table 13 reports the estimates of the average wage increase rate. The unit of observation
is a rm-year. The rst coecient in regression (1) indicates that the wage increase rate is
0.39 percentage points lower in a strategic contract-expiration year, compared to the 2.74%
wage increase rate in a regular contract expiration year. Although the estimate is a correlation
rather than a causal estimate, the evidence nevertheless supports the idea that strategic liquidity
reduction through asset purchases lead to a better collective bargaining outcome. The evidence
is along the lines of empirical ndings; rms with higher leverage or nancial distress oer lower
wages (Hanka (1998), Benmelech et al. (2012)) and are less likely to experience a strike (Myers
and Saretto (2010)). The next coecient in regression (1) indicates that the wage increase
rate is 0.45 percentage points higher in a non-strategic contract-expiration year, compared to
a regular contract-expiration year. The evidence shows that if a rm increases liquidity before
large union negotiations, the rm gives a larger wage increase to union employees. Regression
(2) and (3) control for earnings as there is a strong correlation between wage increase rate and
earnings and regression (3) additionally controls for rm size: the results are similar as before.
The evidence is broadly consistent with the idea that rms strategically reduce liquidity, and
such actions help rms at the bargaining table.
39
8. Conclusion and Discussion
Using a hand-matched dataset on union contracts, I provide novel evidence on the strategic
use of corporate liquidity in contract negotiations with unions. I rst show that rms remove
current and expected future liquidity in contract-expiration years by decreasing cash holdings
and increasing leverage, respectively. Next, I show that managers remove liquidity mainly
through increasing asset purchases, but not through increasing dividends, share repurchases,
investments, or R&D expenditures. I also show that rms decrease the probability of an asset
sale in contract-expiration years. I also provide some suggestive evidence showing that strategic
liquidity management is associated with lower wage increase rates. Overall, the results provide
direct support for the managerial control hypothesis proposed in this paper: managers reduce
liquidity to gain strategic advantages in contract negotiations with unions, while simultaneously
allowing managers to maintain the level of resources under their control.
While previous studies have established connection between industry-level unionization rates
and corporate nancial policies, most studies have not examined rms' actions that are con-
nected to the timing and the gravity of union negotiations. Exploiting the exogenous timing of
union contract expirations, this paper provides a well-identied estimate on the strategic use of
corporate liquidity in union negotiations. This paper also oers new insights into the literature
on asset purchases: by showing that rms increase asset purchases before union negotiations, I
identify a novel and causal determinant of corporate asset purchases and mergers.
This paper also contributes to a growing eld of research that explores how labor and nance
interact. The evidence in this paper suggests that bargaining considerations are perceived
as material to rms, to the extent that rms substantially distort their corporate nancial
policies before union negotiations. The connection between union contract negotiations and
corporate nancial policies illustrates one aspect of how labor aects corporate nancial policies.
Understanding these connections is important because it emphasizes that corporate nancial
policies can not be separated from rms' real activities.
40
References
Almeida, Heitor, Murillo Campello, Igor Cunha, and Michael S Weisbach, 2014, Corporate
liquidity management: A conceptual framework and survey, Annu. Rev. Financ. Econ. 6,
135{162.
Arnold, Marc, Dirk Hackbarth, and Tatjana Xenia Puhan, 2015, Financing asset sales and
business cycles, Boston U. School of Management Research Paper 14{11.
Baumol, William J, 1952, The transactions demand for cash: An inventory theoretic approach,
The Quarterly Journal of Economics 545{556.
Benmelech, Efraim, Nittai K Bergman, and Ricardo J Enriquez, 2012, Negotiating with labor
under nancial distress, Review of Corporate Finance Studies 1, 28{67.
Blanchard, Olivier Jean, Florencio Lopez-de Silanes, and Andrei Shleifer, 1994, What do rms
do with cash windfalls?, Journal of nancial economics 36, 337{360.
Borisova, Ginka, and James R Brown, 2013, R&d sensitivity to asset sale proceeds: New
evidence on nancing constraints and intangible investment, Journal of Banking & Finance
37, 159{173.
Brav, Alon, John R Graham, Campbell R Harvey, and Roni Michaely, 2005, Payout policy in
the 21st century, Journal of nancial economics 77, 483{527.
Bronars, Stephen G, and Donald R Deere, 1991, The threat of unionization, the use of debt,
and the preservation of shareholder wealth, The Quarterly Journal of Economics 231{254.
Brown, James R, and Bruce C Petersen, 2011, Cash holdings and r&d smoothing, Journal of
Corporate Finance 17, 694{709.
Chava, Sudheer, Andras K Danis, and Alex Hsu, 2016, Labor constraints and the impact on
rm investment: Evidence from right-to-work laws, Woring Paper .
Chen, Huafeng Jason, Marcin Kacperczyk, and Hernan Ortiz-Molina, 2011, Labor unions,
operating
exibility, and the cost of equity, Journal of Financial and Quantitative Analysis
46, 25{58.
Dasgupta, Sudipto, and Kunal Sengupta, 1993, Sunk investment, bargaining and choice of
capital structure, International Economic Review 203{220.
DeAngelo, Harry, and Linda DeAngelo, 1991, Union negotiations and corporate policy: A study
of labor concessions in the domestic steel industry during the 1980s, Journal of nancial
Economics 30, 3{43.
Duchin, Ran, Oguzhan Ozbas, and Berk A Sensoy, 2010, Costly external nance, corporate
investment, and the subprime mortgage credit crisis, Journal of Financial Economics 97,
418{435.
41
Edmans, Alex, and William Mann, 2013, Financing through asset sales, Technical report,
National Bureau of Economic Research.
Graham, John R, and Campbell R Harvey, 2001, The theory and practice of corporate nance:
Evidence from the eld, Journal of nancial economics 60, 187{243.
Hanka, Gordon, 1998, Debt and the terms of employment, Journal of Financial Economics 48,
245{282.
Harford, Jarrad, 1999, Corporate cash reserves and acquisitions, The Journal of Finance 54,
1969{1997.
Huang, Qianqian, Feng Jiang, Erik Lie, and Tingting Que, 2015, The eect of labor unions on
ceo compensation, Available at SSRN 1571811 .
Jensen, Michael C, 1986, Agency cost of free cash
ow, corporate nance, and takeovers, Cor-
porate Finance, and Takeovers. American Economic Review 76.
Kim, Chang-Soo, David C Mauer, and Ann E Sherman, 1998, The determinants of corporate
liquidity: Theory and evidence, Journal of nancial and quantitative analysis 33.
Kim, Hyunseob, and Howard Kung, 2014, The asset redeployability channel: How uncertainty
aects corporate investment, Available at SSRN 2022669 .
Klasa, Sandy, William F Maxwell, and Hern an Ortiz-Molina, 2009, The strategic use of corpo-
rate cash holdings in collective bargaining with labor unions, Journal of Financial Economics
92, 421{442.
Lamont, Owen, 1997, Cash
ow and investment: Evidence from internal capital markets, The
Journal of Finance 52, 83{109.
Lee, David S, and Alexandre Mas, 2012, Long-run impacts of unions on rms: New evidence
from nancial markets, 1961{1999, The Quarterly Journal of Economics 127, 333{378.
Lintner, John, 1956, Distribution of incomes of corporations among dividends, retained earn-
ings, and taxes, The American Economic Review 46, 97{113.
Maksimovic, Vojislav, and Gordon Phillips, 2001, The market for corporate assets: Who engages
in mergers and asset sales and are there eciency gains?, The Journal of Finance 56, 2019{
2065.
Marciukaityte, Dalia, 2015, Right-to-work laws and nancial leverage, Financial Management
44, 147{175.
Matsa, David A, 2010, Capital structure as a strategic variable: Evidence from collective bar-
gaining, The Journal of Finance 65, 1197{1232.
Matsusaka, John G, Oguzhan Ozbas, and Irene Yi, 2017, Opportunistic proposals by union
shareholders, Available at SSRN 2666064 .
42
Myers, Brett W, and Alessio Saretto, 2010, Union strikes and the impact of non-nancial
stakeholders on capital structure, Available at SSRN 1553305 .
Opler, Tim, Lee Pinkowitz, Ren e Stulz, and Rohan Williamson, 1999, The determinants and
implications of corporate cash holdings, Journal of nancial economics 52, 3{46.
Ouimet, Paige Parker, and Elena Simintzi, 2015, Wages and rm performance: Evidence from
the 2008 nancial crisis, Available at SSRN 2409496 .
Perotti, Enrico C, and Kathryn E Spier, 1993, Capital structure as a bargaining tool: The role
of leverage in contract renegotiation, The American Economic Review 1131{1141.
Qiu, Yue, 2016, Debt structure as a strategic bargaining tool, Available at SSRN 2762085 .
Sarig, Oded H, 1998, The eect of leverage on bargaining with a corporation, Financial Review
33, 1{16.
Schlingemann, Frederik P, Rene M Stulz, and Ralph A Walkling, 2002, Divestitures and the
liquidity of the market for corporate assets, Journal of nancial Economics 64, 117{144.
Schmalz, Martin C, 2015, Unionization, cash, and leverage, Ross School of Business Paper .
Yang, Liu, 2008, The real determinants of asset sales, The Journal of Finance 63, 2231{2262.
Yim, Soojin, 2013, The acquisitiveness of youth: Ceo age and acquisition behavior, Journal of
nancial economics 108, 250{273.
43
Table 1. Summary Statistics
This table reports summary statistics for the sample rms. The sample period is 1996{2012,
with the exception of work stoppage data in Panel A. For the work stoppage data, the sample
period is 1997{2012. The unit of observation is a rm-year for all variables. Information in
Panel A comes from the BNA. Information in Panel B comes from Compustat. Information in
Panel C comes from Compustat and SDC Platinum. Appendix Table B.2 provides denitions
of variables in Panel B. All variables in Panel B are winsorized at 1% tails. Variables that
cannot take negative values (e.g., cash, leverage) are limited to 0 at the lower bound.
Panel A. BNA
Mean Median S.D. N
Dummy=1 if contract-expiration year 0.14 0 0.35 4,946
Dummy=1 if work stoppage 0.08 0 0.27 4,627
Dummy=1 if contract-expiration year & work stoppage 0.01 0 0.11 4,627
# Contract employees in contract-expiration years 8,673 2,350 25,101 696
# Contract employees in non-contract-expiration years 944 25 4,107 4,250
Annual wage increase rate 2.79 3.00 1.24 777
Panel B. Corporate Financial Policy Variables
Contract- Non-Contract- p-value for
Expiration Years Expiration Years mean dierence
Mean S.D. Mean S.D.
Capital expenditures 0.167 0.096 0.171 0.107 0.33
Cash 0.056 0.067 0.066 0.079 0.001
Dividends 0.017 0.016 0.016 0.017 0.81
Leverage 0.332 0.196 0.314 0.192 0.03
Log(assets) 9.012 1.596 8.879 1.623 0.05
R&D 0.021 0.025 0.022 0.025 0.38
Repurchases 0.019 0.034 0.020 0.036 0.49
ROA 0.129 0.067 0.128 0.065 0.84
Panel C. Asset Purchases and Sales
All Years Contract- Non-Contract-
Expiration Years Expiration Years
Mean S.D. N Mean N Mean N
Asset purchases / total assets 0.029 0.080 4,014 0.038 577 0.028 3,437
Dummy=1 if asset purchase 0.452 0.498 4,946 0.486 696 0.447 4,250
Dummy=1 if asset purchase & 0.201 0.401 4,946 0.227 696 0.197 4,250
no asset sale
Dummy=1 if asset sale 0.402 0.490 4,946 0.397 696 0.403 4,250
Dummy=1 if asset sale & 0.151 0.358 4,946 0.138 696 0.153 4,250
no asset purchase
44
Table 2. Liquidity Management in Contract-Expiration Years
Each column reports estimates from a linear regression. The dependent variable is indicated at
the top of the table. The unit of observation is a rm-year. Log(assets) denotes the logarithm
of the book value of assets, lagged one year. All regressions include rm and industry-year
xed eects. Robust standard errors are clustered at the rm level and are in parentheses
beneath coecient estimates. Coecients and standard errors are scaled by 100 to represent
percentages.

denotes statistical signicance at the 5% level.
Cash Leverage
(1) (2) (3) (4)
Dummy=1 if contract-expiration year { 0:46

{ 0:43

0:93

0:77

(0.18) (0.19) (0.37) (0.36)
Log(assets) { 0:49 0.82
(0.42) (0.82)
Firm F.E. Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes
Observations 4,410 4,375 4,399 4,365
R
2
0.59 0.59 0.76 0.76
45
Table 3. Liquidity Management Channels
Each column reports estimates from a linear regression. The dependent variable is indicated
at the top of the table. The unit of observation is a rm-year. Robust standard errors are
clustered at the rm level and are in parentheses beneath coecient estimates. Coecients
and standard errors are scaled by 100 to represent percentages.

and

denote statistical
signicance at the 5% and 1% levels, respectively.
Panel A. Baseline Specication
Asset Purchases / Dividends Repurchases Capital R&D
Total Assets Expenditures
(1) (2) (3) (4) (5)
Dummy=1 if 0:957

{ 0:003 { 0:055 { 0:405 0.006
contract-expiration year (0.43) (0.04) (0.12) (0.34) (0.05)
Firm F.E. Yes Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes Yes
Observations 4,014 4,374 4,147 4,012 2,402
R
2
0.20 0.71 0.36 0.53 0.88
Panel B. Specication with Firm Size Control
Asset Purchases / Dividends Repurchases Capital R&D
Total Assets Expenditures
(6) (7) (8) (9) (10)
Dummy=1 if 1:203

0.008 { 0:057 { 0:243 0.046
contract-expiration year (0.41) (0.04) (0.12) (0.33) (0.05)
Log(assets) { 3:98

{ 0:14

0.04 { 3:00

{ 0:52

(0.60) (0.06) (0.19) (0.74) (0.16)
Firm F.E. Yes Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes Yes
Observations 4,014 4,344 4,146 4,012 2,386
R
2
0.24 0.71 0.36 0.54 0.89
46
Table 4. Asset Purchases in Contract-Expiration Years
This table presents coecient estimates from linear regressions of asset purchases as a fraction
of total assets (columns (1){(3)) and linear probability regressions of the probability of an asset
purchase (columns (4){(7)). The dependent variable is indicated at the top of the table. The
unit of observation is a rm-year. All regressions include rm and industry-year xed eects.
Standard errors clustered by rm are in parentheses beneath coecient estimates. Coecients
and standard errors are scaled by 100 to represent percentages.

,

, and

denote statistical
signicance at the 10%, 5%, and 1% levels, respectively.
Asset Purchases / Dummy=1 if Dummy=1 if
Total Assets asset purchase asset purchase
& no asset sale
(1) (2) (3) (4) (5) (6) (7)
Dummy=1 if 0:96

1:20

1:21

3:54

2.18 4:72

3:98

contract-expiration year (0.43) (0.41) (0.45) (1.80) (2.01) (1.70) (1.80)
Log(assets) { 3:98

{ 3:70

{ 1:61 { 0:33 { 4:90

{ 4:82

(0.60) (0.61) (2.28) (2.90) (1.83) (2.20)
Cash 0:24

{ 0:07 0:46

(0.05) (0.19) (0.18)
ROA 0:16

1:11

0:70

(0.03) (0.20) (0.16)
Leverage { 0:03

{ 0:09 { 0:15

(0.02) (0.08) (0.08)
Capital Expenditures 0.01 0.01 0.02
(0.01) (0.03) (0.03)
Firm F.E. Yes Yes Yes Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes Yes Yes Yes
Observations 4,014 4,014 3,355 4,449 3,701 4,449 3,701
R
2
0.20 0.24 0.28 0.39 0.40 0.22 0.25
47
Table 5. Asset Sales in Contract-Expiration Years
This table presents coecient estimates from linear probability regressions of the probability of
an asset sale on an indicator variable for a contract-expiration year. The dependent variables
are indicated at the top of the table. The unit of observation is a rm-year. All regressions
include rm and industry-year xed eects. Standard errors clustered by rm are in paren-
theses beneath coecient estimates. Coecients and standard errors are scaled by 100 to
represent percentages.

,

, and

denote statistical signicance at the 10%, 5%, and 1%
levels, respectively.
Dummy=1 if asset sale Dummy=1 if asset sale
& no asset purchase
(1) (2) (3) (4)
Dummy=1 if contract-expiration year { 3:51

{ 4:05

{ 2:33 { 2:25
(1.90) (2.08) (1.54) (1.74)
Log(assets) 9:00

7:58

5:73

3:11

(2.21) (2.50) (1.39) (1.62)
Cash { 0:67

{ 0:14
(0.18) (0.13)
ROA { 0:25 { 0:66

(0.16) (0.15)
Leverage 0:23

0:18

(0.09) (0.08)
Capital Expenditures { 0:06 { 0:05
(0.06) (0.03)
Firm F.E. Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes
Observations 4,449 3,980 4,449 3,980
R
2
0.39 0.42 0.19 0.22
48
Table 6. Asset Purchases and Sales in Years with a Work Stoppage
This table presents coecient estimates from linear probability regressions of corporate asset
purchases and sales on an indicator variable of whether a work stoppage occurs in a contract-
expiration year. The dependent variable is indicated at the top of the table. Panel A examines
asset purchases and Panel B examines asset sales. The unit of observation is a rm-year. All
regressions include rm and industry-year xed eects. The nance control variables are cash,
ROA, leverage, and capital expenditures, and are suppressed to conserve space. Standard errors
clustered by rm are in parentheses beneath coecient estimates. Coecients and standard
errors are scaled by 100 to represent percentages.

,

, and

denote statistical signicance
at the 10%, 5%, and 1% levels, respectively.
Panel A. Asset Purchases in Contract-Expiration Years with a Work Stoppage
Asset Purchases / Dummy=1 if Dummy=1 if
Total Assets asset purchase asset purchase
& no asset sale
(1) (2) (3) (4) (5) (6) (7)
Dummy=1 if 1.45 1.96 2:64

18:40

18:86

16:21

15:49

contract-expiration year & (1.35) (1.31) (1.30) (5.03) (5.00) (6.03) (6.01)
work stoppage (
1
)
Dummy=1 if 1:10

1:31

1:08

1.97 0.50 3:38

2.82
contract-expiration year & (0.49) (0.47) (0.48) (1.94) (2.01) (1.82) (1.91)
no work stoppage (
2
)
Log(assets) { 4:39

{ 3:70

{ 2:30 { 0:30 { 6:00

{ 4:80

(0.64) (0.61) (2.45) (2.90) (1.94) (2.20)
Finance controls Yes Yes Yes
Firm F.E. Yes Yes Yes Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes Yes Yes Yes
Observations 3,768 3,768 3,355 4,178 3,701 4,178 3,701
R
2
0.20 0.25 0.28 0.39 0.40 0.23 0.25
Dierence between coecients 0.35 0.65 1.56 16:43

18:36

12:83

12:67

on two dummies (
1
-
2
)
49
Panel B. Asset Sales in Contract-Expiration Years with a Work Stoppage
Dummy=1 if asset sale Dummy=1 if asset sale
& no asset purchase
(8) (9) (10) (11)
Dummy=1 if { 12:69

{ 11:60 { 14:88

{ 14:97

contract-expiration year & (7.01) (7.22) (3.52) (3.86)
work stoppage (
1
)
Dummy=1 if { 2:13 { 3:29 { 0:71 { 0:97
contract-expiration year & (2.02) (2.16) (1.72) (1.86)
no work stoppage (
2
)
Log(assets) 9:28

7:56

5:60

3:07

(2.40) (2.52) (1.53) (1.63)
Finance controls Yes Yes
Firm F.E. Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes
Observations 4,178 3,701 4,178 3,701
R
2
0.40 0.42 0.20 0.23
Dierence between coecients { 10:57 { 8:31 { 14:18

{ 14:00

on two dummies (
1
-
2
)
50
Table 7. Falsication Tests
Each column reports estimates from a linear probability regression. The dependent variable is
asset purchases as a fraction of total assets. The unit of observation is a rm-year. Log(assets)
denotes the logarithm of the book value of assets, lagged one year. All regressions include rm
and industry-year xed eects. Robust standard errors are clustered at the rm level and are
in parentheses beneath coecient estimates. Coecients and standard errors are scaled by 100
to represent percentages.

denotes statistical signicance at the 1% level.
Asset Purchases / Total Assets
(1) (2) (3) (4)
Dummy=1 if one year before 0.30
contract-expiration year (0.35)
Dummy=1 if 1:20

contract-expiration year (0.41)
Dummy=1 if one year after { 0:12
contract-expiration year (0.35)
Dummy=1 if two years after { 0:27
contract-expiration year (0.32)
Log(assets) { 3:94

{ 3:98

{ 3:93

{ 3:92

(0.60) (0.60) (0.60) (0.60)
Firm F.E. Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes
Observations 4,014 4,014 4,014 4,014
R
2
0.24 0.24 0.24 0.24
51
Table 8. Robustness: Alternative Measures of Contract Signicance
This table presents coecient estimates from linear regressions of cash, leverage, and asset
purchases on dierent measures of contract signicance. The dependent variable is indicated
at the top of the table. The unit of observation is a rm-year, and all specication controls for
rm size, by including the logarithm of the book value of assets, lagged one year. All regressions
include rm and industry-year xed eects. For brevity, I only report the coecient on the
key explanatory variable. The key explanatory variable varies by each row and is indicated on
the very left column. \Contract Signicance"is dened as the number of contract employees
minus mean contract employees, scaled by the standard deviation of contract employees

i.e.,
#contractemployees{
i

i

. Standard errors clustered by rm are in parentheses beneath
coecient estimates. Coecients and standard errors are scaled by 100 to represent percentages.

,

, and

denote statistical signicance at the 10%, 5%, and 1% levels, respectively.
Key explanatory variable Cash Leverage Asset Purchases / Total Assets
Dummy=1 if { 0:43

0:77

1:20

contract-expiration year (baseline) (0.19) (0.36) (0.41)
Dummy=1 if { 0:69

1:34

1:29

contract employees>3000 (0.22) (0.50) (0.49)
Dummy=1 if { 0:57

0:78

0:82

contract employees>1000 (0.21) (0.38) (0.32)
Dummy=1 if { 0:41

0.47 0:63

contract employees>500 (0.19) (0.37) (0.34)
Contract Signicance { 0:29

0:37

0:38

(0.07) (0.14) (0.14)
52
Table 9. Cash-Rich Firms
Each column reports estimates from a linear regression. The dependent variable is asset pur-
chases as a fraction of total assets. The unit of observation is a rm-year. Lag(cash) denotes the
lagged value of cash as a fraction of total assets. Log(assets) denotes the logarithm of the book
value of assets, lagged one year. All regressions include rm and industry-year xed eects.
Robust standard errors are clustered at the rm level and are in parentheses beneath coecient
estimates. Coecients and standard errors are scaled by 100 to represent percentages.

and

denote statistical signicance at the 10% and 1% levels, respectively.
Asset Purchases / Total Assets
(1) (2)
Dummy=1 if contract-expiration year 0.34 0.51
(0.53) (0.51)
Lag(cash) 0:22

0:21

(0.05) (0.05)
Dummy=1 if contract-expiration year lag(cash) 0:13

0:14

(0.08) (0.08)
Log(assets) { 3:86

(0.60)
Firm F.E. Yes Yes
Industry-year F.E. Yes Yes
Observations 3,993 3,993
R
2
0.21 0.25
53
Table 10. Analysis by Industry
This table presents coecient estimates for dierent industries from regression of asset pur-
chases as a fraction of total assets on an indicator variable for a contract-expiration year. In-
dustries are indicated at the second row of each column. The unit of observation is a rm-year.
Log(assets) denotes the logarithm of the book value of assets, lagged one year. All regressions
include rm and industry-year xed eects. Robust standard errors are clustered at the rm
level and are in parentheses beneath coecient estimates. Coecients and standard errors are
scaled by 100 to represent percentages.

,

, and

denote statistical signicance at the 10%,
5%, and 1% levels, respectively.
Asset Purchases / Total Assets
All Manufacturing Non-Manufacturing Transportation Wholesale
(1) (2) (3) (4) (5)
Dummy=1 if 1:20

1:56

0:92

0.65 1.29
contract-expiration
year
(0.41) (0.60) (0.55) (0.44) (1.32)
Log(assets) { 3:98

{ 4:94

{ 2:81

{ 3:96

{ 2:31

(0.60) (0.92) (0.72) (1.22) (1.35)
Firm F.E. Yes Yes Yes Yes Yes
Industry-year F.E. Yes Yes Yes Yes Yes
Observations 4,014 2,192 1,787 1,029 335
R
2
0.24 0.26 0.22 0.24 0.21
54
Table 11. Asset Redeployability
This table presents coecient estimates from linear regressions of asset purchases on asset rede-
ployability interacted with contract-expiration status, using a subsample of rms that operate
in manufacturing industries. Measures of asset redeploability come from Table 1 Panel B of
Kim and Kung (2014). The dependent variable is asset purchases as a fraction of total assets.
The unit of observation is a rm-year. Log(assets) denotes the logarithm of the book value of
assets, lagged one year. All regressions include rm and industry-year xed eects. Robust
standard errors are clustered at the rm level and are in parentheses beneath coecient esti-
mates. Coecients and standard errors are scaled by 100 to represent percentages.

,

, and

denote statistical signicance at the 10%, 5% and 1% levels, respectively.
Asset Purchases / Total Assets
Dummy=1 if non-contract-expiration year & 1.47
medium asset redeployability (
1
) (1.69)
Dummy=1 if non-contract-expiration year & { 0:11
high asset redeployability (
2
) (2.64)
Dummy=1 if contract-expiration year & 2:86

low asset redeployability (
3
) (1.71)
Dummy=1 if contract-expiration year & 3:17

medium asset redeployability (
4
) (1.83)
Dummy=1 if contract-expiration year & { 0:06
high asset redeployability (
5
) (2.78)
Log(assets) { 4:96

(0.93)
Firm F.E. Yes
Industry-year F.E. Yes
Observations 2,196
R
2
0.26
Dierence between coecients on contract-expiration status 2:86

if low asset redeployability (
3
) (1.71)
Dierence between coecients on contract-expiration status 1:70

if medium asset redeployability (
4
-
1
) (0.77)
Dierence between coecients on contract-expiration status 0.05
if high asset redeployability (
5
-
2
) (1.09)
55
Table 12. Managerial Control Hypothesis
Each column reports estimates from a linear regression. The dependent variable is asset pur-
chases as a fraction of total assets. Log(assets) denotes the logarithm of the book value of
assets, lagged one year. All regressions include rm and industry-year xed eects. Robust
standard errors are clustered at the rm level and are in parentheses beneath coecient esti-
mates. Coecients and standard errors are scaled by 100 to represent percentages.

and

denote statistical signicance at the 5% and 1% levels, respectively.
Asset Purchases / Total Assets
(1) (2) (3)
Dummy=1 if contract-expiration year 12:78

(5.63)
CEO age 0.01
(0.05)
Dummy=1 if contract-expiration year { 0:20

CEO age (0.10)
Dummy=1 if contract-expiration year & 3:59

3:77

CEO age 52 (1.33) (1.32)
Dummy=1 if contract-expiration year & 0.97 0.94
52 < CEO age 63 (0.59) (0.59)
Dummy=1 if contract-expiration year & 0.66 0.51
CEO age > 63 (1.16) (1.30)
Dummy=1 if non-contract-expiration year & 0.31
52 < CEO age 63 (0.39)
Dummy=1 if non-contract-expiration year & 0.45
CEO age > 63 (0.83)
Log(assets) { 4:14

{ 4:16

{ 4:14

(0.68) (0.69) (0.68)
Firm F.E. Yes Yes Yes
Industry-year F.E. Yes Yes Yes
Observations 2,947 2,947 2,947
R
2
0.24 0.24 0.24
56
Table 13. Collective Bargaining Outcome
Each column reports estimates from a linear regression. The sample period is 1996{2012. The
dependent variable is the average annual wage increase rate (%) over the life of the contract.
Earnings denote the lagged value of earnings before interest and taxes, scaled by the book value
of total assets. All regressions include rm and year xed eects. Robust standard errors are
clustered at the rm level and are in parentheses beneath coecient estimates. Coecients
and standard errors are scaled by 100 to represent percentages.

,

, and

denote statistical
signicance at the 10%, 5% and 1% levels, respectively.
Average Wage Increase Rate (%)
(1) (2) (3)
Dummy=1 if strategic 0:39

0:37

0:44

in contract-expiration year (0.18) (0.19) (0.19)
Dummy=1 if non-strategic 0:45

0:50

0:48

in contract-expiration year (0.24) (0.24) (0.23)
Dummy=1 if 0.03 0.03 0.06
non-contract-expiration year (0.09) (0.09) (0.08)
Earnings 2:70

2:98

(1.56) (1.61)
Constant 2:74

2:47

1:25
(0.18) (0.21) (1.75)
Firm size control Yes
Firm F.E. Yes Yes Yes
Year F.E. Yes Yes Yes
Observations 718 714 714
R
2
0.13 0.14 0.16
57
Appendix A. Description of Data
A.1. Contract Expirations
Information on labor contract expirations comes from the BNA Labor Plus database, main-
tained by the Bureau of National Aairs. Under the National Labor Relations Act, rms with
union contracts are required to le notices of contract expirations with the Federal Mediation
and Conciliation Service. These lings include information on employer and union names,
contract expiration and notice dates, and the number of employees involved in the collective
bargaining. The database does not have any rm identiers, so rms had to be identied by
their names as they appear on the BNA lings. I manually matched these employer names
with the company names in Compustat. The names in the BNA database are often at a plant
or a subsidiary level, in which cases I identied and matched with the ultimate parent. When
a division or a plant changed its ownership during the sample period, I identied the owner
at the point of contract expiration. To make the project manageable, I limited my sample to
rms that have at least one contract with 500 or more contract employees. This lter is needed
because there are more than 210,000 unique names in the contract expirations database. Once
a rm passed this lter, I included all contracts it engaged in, using company-specic keywords
and then manually correcting wrong matches. For example, the keywords used for TJX Com-
panies Inc. are TJ MAXX, T J MAXX, TJX, T.J. MAXX, MARSHALLS, MARMAXX, where
the latter two are subsidiaries of the company. The nal dataset contains 27,284 observations
from 377 rms during the period 1995{2014.
A.2. Work Stoppages
The BNA Work Stoppage database reports employer name, work stoppage start and end dates,
union, and number of employees under a work stoppage. Work stoppages include strikes and
lockouts. As with the BNA Labor Plus database, only employer names, not rm identiers, were
available, so companies had to be matched to the other databases manually. I only include rms
that had at least one contract expiration involving more than 500 employees; among those rms,
I include all work stoppages involving any number of workers. Stoppage dates were assigned
to the year in which the stoppage occurred. I identify 475 work stoppage events from the 377
58
rms that had at least one contract expiration involving 500 or more contract employees during
the period 1997{2014. 172 rms had at least one work stoppage.
A.3. Collective Bargaining Outcomes
The BNA Settlement database reports employer, union, settlement eective date, contract
expiration date, contract term, wage increase, original wage, and a description of other contrac-
tual terms. All the entries are in text format (e.g., \3.66% 1st yr, 2nd yr, 3rd yr, 4th yr, 5th
yr" and \$30 (was $22) per hr for tutors over term"). I extracted information from the text.
Contracts are usually multi-dimensional: they report information on wage levels (e.g., $8/hour)
or wage increase rates (e.g., 9% over three years), bonuses, lump sum payments, retirement and
health care benets, disability insurance, and sick leave. I focus on wages, the most important
component of the contract, because the other components are hard to quantify. For each con-
tract, I compute the average wage increase rate over the life of the contract. I identify 1,389
settlement observations from the 377 rms that had at least one contract expiration involving
500 or more contract employees during the period 1996{2014. 261 rms (and 777 rm-years)
had at least one settlement observation.
A.4. Asset Purchases and Sales Events
Information on asset purchases and sales comes from SDC Platinum. The unit of observation is
a transaction. Each observation reports detailed information including the buyer and the seller,
transaction date, transaction value, method of payment, and the type of each transaction (e.g.,
merger, asset acquisition, or buyback). I remove transactions if the acquiror and the target
share the same six-digit CUSIP or if the announcement dates are missing. The sample consists
of both U.S. and non-U.S. transactions that are categorized as a merger or an asset acquisition.
Either the buyer or the seller can be a private rm. I use the announcement date as the
transaction date. The transaction value is missing for nearly half of the transactions, especially
for transactions that involve private or non-U.S. rms, in which cases the deal of the transactions
is often not disclosed. Asset purchases and sales by subsidiaries are included in the sample.
59
A.5. Amount of Asset Purchases
I obtain data on the annual amount of asset purchases from the Statement of Cash Flows
available in Compustat (variable name aqc). The denition is given as \This item represents
cash out
ow of funds used for and/or the costs relating to acquisition of a company in the current
year or eects of an acquisition in a prior year carried over to the current year." I normalize
the amount of annual asset purchases by the lagged value of total assets. The normalized value
is limited to 0 at the lower bound and winsorized at the upper 1% tail.
A.6. Other Information
Firm nancial information is taken from Compustat. Compustat variable denitions are in
Appendix Table B.2. Information on CEO age comes from Execucomp. Information on asset
redeployability comes from Table 1 Panel B of Kim and Kung (2014).
A.7. Combining the Databases
After manually matching the rms in the BNA databases with Compustat using company
names, I merge the data on asset purchases from SDC Platinum with the matched data using the
six-digit CUSIP of the rm (or their subsidiaries) as the primary identier. six-digit CUSIPs are
often incorrect in the SDC Platinum database, and some rms used multiple six-digit CUSIPs
during the sample period. In such cases, I also: (i) use ticker as a secondary rm identier
and manually verify that each match with the ticker is correct, (ii) manually search the SDC
Platinum database with company-specic keywords (e.g., 3M, MMM, and Minnesota Mining
and Manufacturing for 3M Company) to identify transactions made by my sample rms. The
nal panel data set has 4,946 rm-year observations from 344 rms during 1996{2012.
A.8. Sample Universe
My sample covers a signicant fraction of major U.S. companies: About 60% of the sample
rms (=206/344) were part of the S&P 500 index at some point in time. In my sample,
58% of rms operate in Manufacturing industries (SIC 2000-3999), 24% in Transportation and
Public Utilities industries (SIC 4000-4999), 10% in Wholesale and Retail Trade industries (SIC
5000-5999), and 6% in Service industries (SIC 7000-8999), among others.
60
Appendix B. Additional Tables and Figures
Table B.1. Example of a Labor Contract Expiration Filing
This table shows information reported in a typical labor contract expiration.
Information Example of an Entry in the BNA Database
Employer Name RIO SUITE HOTEL-CASINO
Contract Notice Date 2013-03-01
Contract Expiration Date 2013-07-31
Number of Workers 3,644
Number of Contract Employees 1,640
Employer Location Las Vegas, NV
Union UNITE HERE
Employer Representative Name Gianini, Debbie
Employer Representative Address 3700 W. Flamingo Road, Las Vegas, NV
89103-4043
Union Representative Name Arguello-Kline, Geoconda
Union Representative Address 1630 S. Commerce Street, Las Vegas, NV
89102
Product Description Accommodation and Food Services
61
Table B.2. Compustat Variable Denitions
This table provides the denitions of Compustat variables used in this paper. Texts in italic
denote Compustat variable names.
Variable Denition
Asset Purchases / Total Assets aqc/at(t {1)
Capital Expenditures capx/ppent(t {1)
Cash che/at
Dividends (dvc + dvp)/at
Earnings ebit/at
Leverage (dltt + dlc)/at
Log(assets) log(at)
R&D xrd/at
Repurchases [prstkc { (pstkrv(t) { pstkrv(t {1))]/at
ROA oibdp/at
62
Panel A. Frequency of the Number of Contract-Expiration Years
Panel B. Incidence of Contract-Expiration Years for Firms in the Sample
Figure B.1. Frequency of Contract-Expiration Years. Panel A presents the histogram of
the number of contract-expiration years for my sample rms. Note that the number of contract-
expiration years is a count variable. Panel B presents the incidence of contract-expiration years
for 52 rms in my sample. Each horizontal line indicates whether a year was a contract-
expiration year or not for a given rm. For example, the very top horizontal line indicates that
years 1996 and 1998 were contract-expiration years for the rm. A discontinued line indicates
that the rm was not in my sample before/after the time of discontinuation, due to reasons
including pre-IPO, mergers, and bankruptcy.
63
Figure B.2. Length of a Labor Contract. The gure presents the boxplot for the length of
a labor contract, tabulated by the contract-initiation year. The information on contract length
come from the BNA Settlements database. The gure is based on 1,300 labor contracts that
were initiated between 1997 and 2012. The length of a labor contract is measured in years.
The median value for the length of a labor contract is marked in red for each year. The start
and the end of a box are the lower and upper quartile, respectively. The start and the end of a
whisker are the lower quartile1.5 (upper quartile lower quartile), and upper quartile +1.5
(upper quartile lower quartile), respectively. Outside values are in gray dots.
64

Abstract (if available)

Abstract Using a hand-matched data set on 27,284 union contracts, I provide novel evidence on the strategic use of corporate liquidity in contract negotiations with unions. I focus on the idea that firms have incentives to hold low levels of liquid assets during union negotiations, because high liquidity can encourage unions to raise their wage demands. The main finding is that firms reduce liquidity before contract negotiations primarily through increased asset purchases. Firms increase asset purchases as a fraction of total assets by one-third before contract negotiations, and finance those purchases by a reduction in cash balances and an increase in leverage. Firms do not increase investments, R&D, dividends, or repurchases before contract negotiations. Strategic liquidity management is associated with lower wages. The evidence indicates that firms reduce liquidity to gain strategic advantages in contract negotiations with unions in ways that simultaneously allow managers to maintain the level of resources under their control.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on the firm and stakeholders relationships: evidence from mergers & acquisitions and labor negotiations

PDF

Real asset liquidity and asset impairments

PDF

Essays in financial intermediation

PDF

Shareholder litigation as a disciplining device: evidence from firms' financial policies

PDF

Costs and benefits of "friendly" boards during mergers and acquisitions

PDF

Internal capital markets and competitive threats

PDF

Essays on information and financial economics

PDF

Essays in corporate finance

PDF

Incentives and relative-wealth concerns: theory and evidence

PDF

Evolution of returns to scale and investor flows during the life cycle of active asset management

PDF

Voluntary disclosure responses to mandated disclosure: evidence from Australian corporate tax transparency

PDF

Essays in behavioral and entrepreneurial finance

PDF

How do acquirers govern the deal-making process? Three essays on U.S. mergers and acquisitions 1994 – 2017

PDF

Essays in corporate finance

PDF

The structure of strategic communication: theory, measurement, and effects

PDF

Essays in empirical asset pricing

PDF

“What you see is all there is”: The effects of media co‐coverage on investors’ peer selection

PDF

The term structure of CAPM alphas and betas

PDF

Share repurchases: how important is market timing?

PDF

Executive compensation: the trend toward one size fits all

Asset Metadata

Creator Yi, Irene (author)

Core Title Slashing liquidity through asset purchases: evidence from collective bargaining

School Marshall School of Business

Degree Doctor of Philosophy

Degree Program Business Administration

Publication Date 07/20/2017

Defense Date 08/08/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag asset purchases,labor union,mergers,OAI-PMH Harvest,strategic,unions

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Matsusaka, John (committee chair), Ahern, Kenneth (committee member), Jia, Nan (committee member), Korteweg, Arthur (committee member), Ozbas, Oguzhan (committee member)

Creator Email irene.e.yi@gmail.com,iyi@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-406401

Unique identifier UC11265771

Identifier etd-YiIrene-5567.pdf (filename),usctheses-c40-406401 (legacy record id)

Legacy Identifier etd-YiIrene-5567.pdf

Dmrecord 406401

Document Type Dissertation

Rights Yi, Irene

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA