Corporate Governance over the Business Cycle

Corporate Governance over the Business Cycle ∗
Thomas Philippon†
New York University, CEPR and NBER
July 2003, revised March 2005
Abstract
I provide empirical evidence that badly governed firms respond more to aggregate shocks
than do well governed firms. I build a simple model where managers are prone to
over-invest and where shareholders are more likely to tolerate such a behavior in good
times. The model successfully explains the average profit differences as well as the
cyclical behavior of sales, employment and investment for firms with different governance
qualities. The quantitative results suggest that governance conflicts could explain up
to a third of aggregate volatility.
JEL: E3, L16.
Keywords: Corporate Governance, Business Cycles, Counter-Cyclical Markups.
∗I am grateful to Wouter DenHaan and two anonymous referees for their comments, and to Olivier
Blanchard and Ricardo Caballero for numerous discussions. I have also benefited from the comments of
Manuel Amador, Mark Gertler and Ivan Werning. I thank Daron Acemoglu, David Bowman, John Faust,
Francesco Franco, Augustin Landier, Gordon Phillips, John Reuter, Roberto Rigobon, Bernard Salanie and
Michael Woodford, as well as seminar participants at the NBER summer institute, the ReStud tour 2003,
the Federal Reserve Board, the IMF, MIT, Harvard, NYU, Wharton, Delta and CREST for their comments.
†Contact: New York University, Stern School of Business, Department of Finance, Suite 9-190, 44 West
4th Street, New York, NY 10012. Email: tphilipp@stern.nyu.edu. Phone: 212-998-0490. Fax: 212-995-4256.
1

---

1
Introduction
I propose a simple model to study the implications of corporate governance for the busi-
ness cycle, based on the idea that managers tend to expand their firms beyond the profit-
maximizing size. What matters for aggregate dynamics is whether these deviations from
profit maximization are more likely to happen in booms or in recessions. This, in turn,
depends on how the relative costs and benefits of monitoring firms’ decisions change with
the state of the economy. I take the view that the comparative advantage of managers is
to come up with new ideas to seize profit opportunities, and that scrutinizing managerial
decisions is a time consuming process. Since it is particularly costly to miss a profit op-
portunity when the demand for the firm’s product is high, shareholders are more likely to
leave discretionary authority in the hands of managers in good times than in bad times.
I study the implications of these governance conflicts in a standard dynamic macro
model with imperfect competition in the goods markets. Managerial tendencies to increase
investment, employment and output — together with the proposition that shareholders leave
more discretion to managers in good times — implies that corporate governance conflicts am-
plify aggregate fluctuations. Quantitative simulations, based on the new empirical evidence
presented in section 2, suggest that aggregate volatility would be 30% lower if all firms were
always perfectly governed.
This research is related to the microeconomic literature on governance conflicts between
managers and shareholders. Jensen (1986) emphasizes the idea that managers tend to
expand their firms beyond the profit-maximizing size. On the macroeconomic side, I build
on Blanchard and Kiyotaki (1987) and on Rotemberg and Woodford (1992)1 for the role of
imperfect competition and counter-cyclical markups in explaining aggregate fluctuations.
Chevalier and Scharfstein (1996) show how financial constraints can lead to counter-cyclical
markups in a customer market model.
The macroeconomic literature has traditionally focused on models of costly external
finance (see Bernanke, Gertler, and Gilchrist (1999) for a recent survey). These models
best describe entrepreneurial firms, firms with limited access to bonds markets, and firms
1 The empirical finding that markups of prices over marginal costs are counter-cyclical is relevant for my
paper because a firm operating on its demand curve can expand its output only by lowering its markup. See
Rotemberg and Woodford (1999) for a survey, and Bils and Kahn (2000) for recent evidence.
2

---

with no outside equity. The traditional assumption in the business cycle literature has been
that financial imperfections matter mostly for these small firms. By contrast, I emphasize
the role of managerial misbehavior, and I argue that conflicts with shareholders matter for
the behavior of large firms over the business cycle. Since the literature has traditionally
estimated the importance of financial constraints from the difference between small and
large firms (see for instance Bernanke, Gertler, and Gilchrist (1996), page 12), my esti-
mates should be added to the existing ones. Recently, Dow, Gorton, and Krishnamurthy
(2003) have explored the asset pricing implications of imperfect corporate governance, and
Castro, Clementi, and MacDonald (2004) have studied the extent to which weak share-
holder protection can limit economic growth. Caselli and Gennaioli (2004) have explored
the consequences of dynastic management for economic growth
Section 2 provides new evidence on the business cycle properties of firms with different
governance qualities. Section 3 and 4 present the model. Section 5 explains the intuition
for the amplification mechanism. Section 6 discusses the calibration method and how it
relates to the existing empirical literature about governance conflicts. Section 7 presents
the impulse responses and the simulations of the model. Section 8 concludes. Derivations
and technical details are in the appendix.
2
Evidence
Figure 1 shows that firms with bad governance have lower profit margins. The governance
data come from the Investor Responsibility Research Center and are based on 24 distinct
corporate-governance provisions. Gompers, Ishii, and Metrick (2003) construct an index
by adding one for every provision that reduces shareholders rights, so that higher values
mean worse governance. The index is constructed for the 1990’s. The profit margin is the
ratio of median income during the period 1989-2001 to median capital expenditures during
the same period, relative to the average of firms in the same industry (one digit SIC code)
and age group (the five age groups are defined below). The figure shows that the profit to
investment ratio of badly governed firms is 6% below average, while the profit to investment
ration of well governed firms is 7% above average. The difference is significant at the 5%
level. These results are in line with Gompers, Ishii, and Metrick (2003) who report that
3

---

badly governed firms have lower profits to sales ratios.
The original governance index ranges from 5 to 14, and I have created three groups with
cutoffs at 8 and 12, corresponding to the 25th and 75th percentiles of the distribution of
the index. I will use these three groups in the calibration exercise. The governance index is
not available for all years and all firms, and it can vary over time. To be on the safe side, I
decided to compare firms with persistently bad governance to firms with persistently good
governance, and I kept only the firms whose index has a standard deviation of less than one
over the sample period. In practice, the index is very persistent over time and the results
are robust to keeping all firms. I sort firms among the three groups according to the earliest
available index.
Figures 2a and 2b show that the capital expenditures and sales of firms with bad gov-
ernance are more cyclical than the investment and sales of firms with good governance.
Sensitivity to aggregate shocks is defined as the regression coefficient, β, of the growth rate
of capital expenditures (or sales) git on the growth rate of aggregate private non-residential
investment (or GDP) ¯
gt:
git = αi + ¡βGov +γIndustry +δAge¢×¯gt+εit
γIndustry is a set of dummies for the one digit SIC code of firm i. δAge is a set of dum-
mies for the age group of firm i, using 5 groups and cutoffs at percentiles (10, 25, 50, 75).
αi is a firm fixed effect. The regression results (using fixed effects and the within re-
gression estimator) are reported in table 1.
Columns 1-5 refer to investment growth,
columns 6-10 to sales growth and columns 11-15 to employment growth. Note that the
specification allows for systematic differences of growth rates across firms (αi), and for sys-
tematic differences of business cycle behavior across industries (γIndustry) and age groups
(δAge). I use the governance index as a linear regressor (βGov = β × Gindex) in regres-
sions 1-3, 6-8 and 11-13. I use the governance index to construct governance dummies
(βGov = β1 × [8 ≤ Gindex ≤ 12] + β2 × [13 ≤ Gindex]) in regressions 4,5,9,10,14 and 15. Fig-
ure 2 is based on regressions 5 and 10. On average, when aggregate investment increases
by 1%, the capital expenditures increase by 1.35% for firms with bad governance, and by
0.65% for firms with good governance: The difference is 0.708 as reported in table 1, column
5. Quite remarkably, the same is true for sales growth and for employment growth: Along
4

---

all margins, badly governed firms expand more in booms, and contract more in recessions
than do well governed firms.
The existing literature has emphasized the role of debt (see Sharpe (1994)), so I ran
all the regressions controlling for initial leverage2. The results are presented in columns
3, 8 and 13. Note that the coefficients on governance are very stable and that leverage is
not significant (one can also see that the within R2 does not improve much). So I am not
capturing the standard effect of leverage on volatility.
I now turn to the model. I will use the quantitative estimates presented in figures 1 and
2 for the calibration of the model.
3
Model
I introduce governance issues into a standard general equilibrium model. I first present
the macro-economic environment in which firms operate. I then describe a simple model
of imperfect corporate governance. The main focus of this paper is not on the details of
the agency costs, but rather on how these costs affect the aggregate economy. To keep the
macroeconomic analysis as transparent as possible, I use a very stylized model of the firm.
3.1
Macro-economic Setup
The macroeconomic setup is very standard. The consumers solve
φ
φ+1
max
E "
φ
0
βt µlog(Ct)−e−zt
L
¶#
t
(1)
K
X
t+1,Lt,Ct,ut
φ + 1
t
subject to the budget constraint
γ (K
(1 + g) K
t+1 − Kt)2
t+1 = (1 − δ (ut)) Kt + WtLt + utRtKt + Πt − Ct −
(2)
2
Kt
Rt is the rental price of capital services, ut is the rate of utilization of the existing stock of
capital Kt, Πt are aggregate profits, g is the trend growth rate of labor productivity and γ
captures adjustment costs for investment. zt is an exogenous aggregate labor supply shock3.
2 I treat leverage exactly like I treat governance: I use the value of the first observation for each firm, in
1989. It is clear that one should not use a time varying measure of leverage, which would be mechanically
correlated with firm investment dynamics.
3 Labor supply shocks provide a convenient way to introduce aggregate shocks that do not directly affect
the technological frontier of the economy. They can be interpreted, for instance, as nominal spending shocks
that interact with nominal wage rigidities. See Chari, Kehoe, and McGrattan (2002)
5

---

The cost of higher utilization is captured by an increase in the depreciation rate δ (ut) as
in King and Rebelo (1999). Capital utilization is introduced only because it is important
in the quantitative analysis. It plays no role for the theory.
The economy produces a final good using differentiated inputs. The final good is pro-
duced competitively and it can be used for consumption and investment. The differentiated
goods are produced by a continuum of mass N of firms indexed from 0 to 1. N will be
determined in equilibrium by a free entry condition. The production function for the final
good is4
σ−1
Y
σ
t = N × µZ 1 y ¶ σσ−1
(3)
it
0
and the final good producers solve
max PtYt
pityit
y
− N × Z 1
it
0
where yit is the production of intermediate good i at time t.
Equation (3) implies that each intermediate producer i faces an iso-elastic demand curve:
Y
y
t
it =
¶−σ
(4)
N × µpit
Pt
The price level, Pt, is such that
for the final good producers. Th R 1
0
ere i³ pit
P
s m ´1−σ = 1. This is also the zero profit condition
t
onopolistic competition in the differentiated goods
sector. The production function for intermediate good i is characterized by constant returns
to variable factors and some fixed costs. The variable factors are the flows of capital and
labor services: kit and lit. Note that kit includes utilization5. The production function for
good i at time t is:
yit = eθt qit k1−αlα
it
it
(5)
θt is an exogenous aggregate technology shock6 and qit is the endogenous firm-specific
4 I choose the measure of firms to be uniform and I omit di when it is obvious that the integration refers
to i.
5 It makes the notations simpler (ut appears only in the aggregate resource constraint below) and, since
capital can be freely traded between firms, it is inconsequential. For the same reason, the assumption that
the utilization rate is chosen by the capital holders (consumers) is also inconsequential.
6 I do not vary Z and θ at the same time. I calibrate using either one or the other. I introduce θ for
the sake of completeness. Since I have verified that the quantitative and qualitative results do not depend
on whether the economy is driven by θ or by Z, I report only the simulations with Z. The one exception
concerns the behavior of the real wage. Without the governance feed-back, labor supply shocks imply a
counter-cyclical real wage, as discussed in section 7, while technology shocks always imply a pro-cyclical real
wage.
6

---

productivity (to be discussed shortly). The fixed costs for firm i are Φit units of final good.
The (real) profits of firm i are therefore:
p
π
it
it =
yit
P
− Wtlit − Rtkit − Φit
(6)
t
3.2
Corporate Governance
I now describe the governance environment. I assume that managers have a comparative
advantage in running the firms, but that their objectives differ from the ones of the share-
holders. Shareholders are aware of this issue: they can choose to monitor the manager
closely, or they can choose not to interfere with the decisions of the manager.
The productivity of each manager is normalized to one, so that qit = 1 when the man-
ager runs the firm without external interference. Let {km,lm,Φ} be the profit maximizing
solution. The fixed cost Φ is exogenously given by technology, and {km,lm} maximize (6)
subject to (4) and (5) with qit = 1.
However, governance conflicts are such that managers do not always implement the
profit maximizing solution. Specifically, I assume that managers’ favorite implementation
is {(1 + η∗)km,(1 + η∗)lm,(1 + τ∗)Φ}. In words, managers prefer firms that are larger by
η∗ for capital and labor, and by τ ∗ for fixed costs. This is consistent with much of the
corporate finance literature, which I will discuss when I calibrate the model. By making
all the deviations proportional to the profit maximizing solution, I make sure that the
economy has a well defined steady state. I introduce two separate parameters, η∗ and τ ∗,
for both theoretical and empirical reasons. Theoretically, they mean different things: η∗
rises output while τ ∗ does not. Empirically, η∗ will capture the idea that managers prefer
to buy the latest machines even if they are not really needed, or that they over-estimate the
appropriate scale of operation for their firms (for instance because they over-estimate their
own productivity), or that they literally have empire building preferences. On the other
hand, τ ∗ will capture the standard managerial perks (private jets,..) as well as outright
stealing and excessive compensation. Finally, the evidence presented in section 2 suggests
that one parameter is enough to capture the behavior of sales, investment and employment,
so that the same η∗ applies to both labor and capital.
The shareholders of firm i face the following trade-off. On the one hand, they can
7

---

minimize external interference and leave much discretion to the manager: this results in
high productivity, qit = 1, but also in deviations from profit maximization, η∗ and τ ∗. On
the other hand, they can monitor the firm closely, and scrutinize the decisions of the manager
before approving their implementation: this eliminates deviations from profit maximization,
but also lowers the productivity of the firm down to ˜
qi ≤ 1.
I will use ˜
qi as my measure of governance quality for firm i. I assume that it is fixed
over time for a given firm, which is consistent with the evidence presented in section 2, and
that it is distributed across firms according to the c.d.f. F (˜
q) over some interval £q,¯q¤. The
smaller is ˜
qi, the more costly it is to control the manager. In the limit, if ˜
qi is equal to 1,
the governance problem vanishes.
3.3
Equilibrium
At each point in time the shareholders of firm i choose between two technologies: the
“rubber-stamping” technology {1;η∗;τ∗} and the “tight monitoring” technology {˜qi;0;0}.
Shareholders choose the technology that maximizes the value of the firm7, period by period.
A rational expectations equilibrium for this economy is a set of stochastic processes
for the exogenous shocks (either θt or Zt) and for the endogenous prices and quantities.
{lit,kit,pit} solve the intermediate firms’ program described above,
i
{Yt,yit} are deter-
mined by (3), and consumers maximize (1) over {Kt+1,Ct,Lt,ut,}8. All the agents take
{Pt,Wt,Rt} as given, and the following market clearing conditions hold:
Yt = Ct + It + N × Z 1Φit
0
utKt = N × Z 1kit
0
Lt = N × Z 1lit
0
This definition of equilibrium is conditional on the number of firms, N , which is constant
at business cycle frequencies. To pin down N , I impose that a free entry condition holds in
the non-stochastic steady state of the economy (see Rotemberg and Woodford (1999) and
the appendix).
7 The appendix contains a discussion of the role of financial incentives.
8 The assumption that consumers choose ut is immaterial as long as there are no firm specific adjustment
costs.
8

---

4
Monitoring versus Rubber-Stamping
One can think of the governance technology in the following way. Agents inside the firm
(CEO, managers, employees) come up with plans to take advantage of profit opportunities
as they appear. A plan specifies a technology and the amounts of capital and labor that
must be hired to implement it. Supervisors (the board for the CEO, the CEO for the division
managers) can either rubber-stamp the plan proposed by the agent, or they can scrutinize
it. Scrutinizing is time consuming and entails the possibility that the profit opportunity will
be missed, so that the expected productivity under close monitoring drops to ˜
qi. On the
other hand, scrutinizing allows the supervisors to cut wasteful expenses (τ ∗: inefficiencies,
private jets, outright stealing..), and to make sure that the project is implemented on the
right scale (η∗: buying expensive machines, hiring too many employees, refusing to close
down a plant..).
Proposition 1 describes the optimal choice to rubber-stamp or to monitor closely for
firm i at time t.
Proposition 1 The shareholders of firm i rubber-stamp managerial propositions if and
only if
˜
qi < Qt
where
τ ∗Φ
Qt = µκ(η∗)− ¶ 1σ−1
(7)
At
and
At ≡ õe−θt µ Rt ¶1−αµWt¶α!1−σ Yt
1 − α
α
σN
µ (1 + η∗)− 1σ
σ
κ (η∗) ≡ (1 + η∗) ×
− 1 ; µ =
µ − 1
σ − 1
Proof. The decision rule is simple: the shareholders will rubber-stamp if and only if
πm
t (˜
qi) < π∗t (1)
Using some simple algebra, one can see that the maximum profits are
πm
t (˜
qi) = At ˜
qσ−1
i
− Φ
9

---

And the profits from the manager’s favorite choice are
π∗t (1) = Atκ∗ − (1 + τ∗)Φ
So the optimal choice is to rubber-stamp if and only if:
At ˜
qσ−1 < A
i
tκ∗ − τ∗Φ
QED.
Proposition 1 says that governance decisions are characterized by a simple cutoff rule:
strict profit maximization is enforced in all firms with governance quality above Qt, while
managerial decisions are rubber-stamped in all firms below the cutoff. The factor At cap-
tures the influence of the state of the economy on the profits of the firms: higher output
Yt means higher profits, and higher marginal cost e−θt ³ Rt1−α´1−α¡Wtα¢α means lower prof-
its. The influence of the parameter τ ∗ is straightforward. The influence of η∗ is slightly
more subtle: The profit losses are summarized by the function κ (η∗), which is concave and
reaches a maximum for η∗ = 0. Starting from the optimal size (km, lm), a small deviation
by η∗ implies only a second order loss in profits.
The measure of firms that rubber-stamp managerial propositions is F (Qt). The crucial
point is that it is an increasing function of At. This result follows from the assumption
that monitoring costs come from lower productivity: These costs are large when At is large.
On the other hand, the cost of rubber-stamping is less than proportional to At because of
the fixed component τ ∗. As a consequence, shareholders are more inclined to rubber-stamp
managerial propositions in good times.
The profit margins of firms with bad governance are persistently lower than the ones of
better governed firms. This fits figure 1. The model also implies that firms with different
governance qualities have different cyclical properties: Firms with excellent governance
always maximize profits, while firms with bad governance follow the objective function
of their manager when At is large and the objective function of the shareholders when
At is small. As a consequence, their capital spending increases and decreases more than
proportionally with the business cycle. This fits figure 29.
The next step is to investigate the quantitative implications of governance conflicts.
9 Note that the relationship between governance and excess sensitivity is not monotonic. In theory, in
10

---