Corporate payout smoothing: A variance decomposition approach

Corporate payout smoothing: A variance decomposition approach

    Corporate Payout Smoothing: A Variance Decomposition Approach Edward Hoang, Indrit Hoxha PII: DOI: Reference: S0927-5398(15)00109-7 ...

213KB Sizes 0 Downloads 114 Views

    Corporate Payout Smoothing: A Variance Decomposition Approach Edward Hoang, Indrit Hoxha PII: DOI: Reference:

S0927-5398(15)00109-7 doi: 10.1016/j.jempfin.2015.10.011 EMPFIN 847

To appear in:

Journal of Empirical Finance

Received date: Revised date: Accepted date:

8 September 2014 26 October 2015 27 October 2015

Please cite this article as: Hoang, Edward, Hoxha, Indrit, Corporate Payout Smoothing: A Variance Decomposition Approach, Journal of Empirical Finance (2015), doi: 10.1016/j.jempfin.2015.10.011

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

CR

IP

T

Corporate Payout Smoothing: A Variance Decomposition Approach

1

CE

PT

ED

MA

NU S

Edward Hoang1 and Indrit Hoxha2

2

AC

University of Colorado at Colorado Springs Department of Economics 1420 Austin Bluffs Pkwy Colorado Springs, CO 80198, Phone: 719-255-3819, E-mail: [email protected]

Pennsylvania State University Harrisburg School of Business Administration 777 W. Harrisburg Pike Middletown PA 17057, Phone: 717-948-6344, E-mail: [email protected]

ACCEPTED MANUSCRIPT

CR

IP

T

Corporate Payout Smoothing: A Variance Decomposition Approach∗

NU S

Abstract

MA

In this paper, we apply a variance decomposition methodology to quantify the smoothness of corporate payouts. We find that firms use debt and investment to smooth a large fraction of shocks to net income to keep payouts less variable. Specifically, our empirical results show that firms keep the growth of payouts relatively small and stable over time. Furthermore, our findings support theoretical work that demonstrates that the dynamics of investment and debt policy should be jointly modeled with payout policy.

ED

JEL Codes: G30, G32, G35



AC

CE

PT

Keywords: Debt, Investment, Lintner Model, Net Income, Payout

We would like to thank Dale DeBoer, Bart Lambrecht, Bent Sorensen, two anonymous referees, and seminar participants at Carnegie Mellon University and Penn State Harrisburg for comments and suggestions. Any remaining errors are our own.

ACCEPTED MANUSCRIPT

Introduction

T

1

IP

This paper empirically examines corporate payout policy in a dynamic model that includes both investment and debt. Much of the theoretical and empirical work on corporate payout policy

CR

has taken investment and debt policies as being independent from payout policy.1 However, the corporate finance literature conjectures that other financing decisions, such as investment and

NU S

debt, may be used to balance the firm’s cash flow if managers want to keep corporate payouts less variable. Although previous research has suggested that corporate payouts are smoothed, we make a further contribution to this literature by using a variance decomposition methodology to measure

MA

the smoothness of payouts along with quantifying the adjustments of debt and investment that are driven by shocks to net income.

ED

Examining the joint dynamics of investment, debt and payout policies is important for understanding firm behavior. Goel and Thakor (2003) argue that managers of firms with compensation

PT

tied to corporate earnings have an incentive to reduce the volatility of their firm’s earnings stream. This suggests that managers may use mechanisms such as investment and debt policies to smooth

CE

shocks to earnings to keep the patterns of their compensation smoothed. Lambrecht and Myers (2012) draw from different agency theories to conceptualize their theoretical model of the dynamics

AC

of firm behavior, which demonstrates that managers reduce the volatility in the distribution of payouts in order to smooth their own compensation. In their model, for a given investment policy, shocks to earnings are primarily absorbed by debt financing in order to keep payouts and, hence, managerial compensation smooth. If it is the case that managers have a preference towards maintaining smooth payouts, then it is of the utmost importance to the corporate finance literature to provide empirical evidence on the matter. The setting of payout policy may reflect managerial risk preferences that in turn

1

Modigliani and Miller (1958) conjectured that in a setting with perfect capital markets, investment and debt financing decisions are made independently. Financial economists have since extended the analysis of firm behavior beyond the assumption of perfect capital markets by incorporating financial frictions in dynamic models examining the interaction between investment and debt financing. For a detailed review, see Strebulav and Whited (2012).

1

ACCEPTED MANUSCRIPT

affects investment and debt financing decisions. For example, managers that prefer to avoid risk

T

may enact less than optimal corporate policies for investment and debt (Gormley and Matsa,

IP

2014). Therefore, measuring the response of payouts may be informative for firm risk management,

CR

investment decisions and debt policy.

Using an intertemporal budget constraint formulated by Lambrecht and Myers (2012), we quan-

NU S

titatively describe the dynamics of firm behavior in response to shocks to net income. In their theoretical models of mature and profitable firms, Lambrecht and Myers (2012, 2014) argue that if the optimal level of investment is determined by investment opportunities and managers are motivated

MA

to keep payouts smoothed then debt policy is solely responsible for absorbing all of the variation in net income. This implies that debt policy not only absorbs all of the shocks to earnings but also smooths any marginal changes in investment to keep the optimal investment policy unresponsive

ED

to net income shocks.

Furthermore, their analysis allows for the possibility that debt policy does not entirely absorb

PT

all of the shocks to net income, thus allowing unabsorbed net income shocks to be potentially

CE

smoothed by investment. For example, if positive shocks to net income reflect the availability of favorable opportunities for economic growth, managers may choose to increase investment and,

AC

therefore, net income shocks will be distributed to investment policy. In this paper, we decompose the variance in net income to estimate the fraction of shocks to net income that are absorbed by debt and investment. The amount left unabsorbed measures the response of payouts after debt and investment have smoothed shocks to net income. The empirical design of our paper uses panel data covering 6,225 public firms from the Compustat database that distributed dividends over the years 1973–2013. Our sample not only includes mature and profitable firms, but also encompasses all of the different types of non-utility and nonfinancial public firms listed in Compustat. Our findings are based on the decomposition of the variance in net income that produces the following relation: βD + βI + βP = 1, where βD and βI measure the fraction of shocks to net income absorbed by investment and debt policy, respectively;

2

ACCEPTED MANUSCRIPT

βP measures the response of payouts to shocks to net income left unabsorbed by debt and invest-

T

ment. After estimating the components of the relation, we find that debt policy and investment

IP

absorbs 56.9% and 40.7% of shocks to net income, respectively. Payouts change by 2.4% in response

CR

to a shock to net income.

The dynamics of debt, investment, and payouts reported in our study fits well with the theo-

NU S

retical findings in Lambrecht and Myers (2012), which shows that managers are motivated to keep payouts stable by using debt policy or investment to smooth net income shocks. In our empirical framework, if shocks to net income are entirely absorbed by debt policy and investment, then pay-

MA

outs should be completely smoothed (i.e., βP = 0). We show that investment and debt together absorb roughly 97.6% of shocks to net income. Since investment and debt financing both smooth a large fraction of shocks to net income, the amount left unabsorbed, 2.4%, is reflected as the change

ED

in payouts. Although we do not find complete smoothing, the relatively small change in payouts suggests that managers attempt to keep payouts less variable.

PT

We provide several additional empirical tests to demonstrate our results are robust to different

CE

exogenous settings. For example, shocks to earnings could be attributed to overall economic conditions and industry conditions that in turn would affect payout policy. Beyond our baseline model,

AC

we estimate the effects of business cycle fluctuations and capital market conditions on firm behavior, and examine the response of debt, investment, and payouts for firms in different industries. We test whether persistent shocks to net income have a larger effect on investment and payout, as predicted in Lambrecht and Myers (2014). Furthermore, we estimate our empirical model by splitting our data into different time periods, stratifying the firms into small and large size categories, grouping firms based on their age, sorting firms into external financing dependent and independent samples, and using different measures of payouts. In these specifications, the estimates of the responses of investment, debt policy, and payouts are consistent with the baseline results. The joint dynamics documented in this paper add to a small but growing literature focusing on the interaction of the three corporate financing decisions. Theoretical work provided by Hennessy

3

ACCEPTED MANUSCRIPT

and Whited (2005, 2007) and DeAngelo, DeAngelo, and Whited (2011) examine the interactions of

T

investment, debt policy, and payouts in models with taxes, and transaction and adjustment costs.

IP

Chang et al. (2014), Gatchev et al. (2010), and Ostergaard et al. (2011) decompose firm cash

CR

flow and measure the sensitivities of cash flow items such as investment, debt, and dividends. In a closely related paper, Leary and Michaely (2011) find empirical evidence of payout smoothing. We

NU S

further contribute to this literature by quantifying the smoothness of payouts, which we measure as the sum of dividends and net stock repurchases, in an empirical model that also estimates the amount of variation in net income that is absorbed by investment and debt policies.

MA

In the next section, we discuss the background of the paper. Section 3 describes the empirical implementation. Section 4 presents the results. Section 5 concludes.

Background: Budget Constraint and Smoothing of Net Income,

ED

2

The Firm’s Budget Constraint and Smoothing of Net Income

CE

2.1

PT

and Related Literature and Hypothesis

To develop the foundation for our empirical analysis, we begin with considering the implications

AC

of the intertemporal budget constraint for an individual firm as described in Lambrecht and Myers (2012) (referred to as LM hereafter):

∆Debtt + N et Incomet = Investmentt + P ayoutt .

(1)

Equation (1) presents several implications for firms in a model without financial constraints.2 If there are fluctuations to net income, the firm can adjust debt or investment to smooth net income. Debt financing is a dominant source of external funds which affects operating income flexibility and the financing of investment activities. Therefore, changes in net debt (∆Debt), which is the repayment of debt, increased borrowing of funds, and changes in cash balances, can help to maintain 2

Equation (1) is extended to include managerial rents in Lambrecht and Myers (2014).

4

ACCEPTED MANUSCRIPT

a balanced intertemporal budget constraint in response to changes to net income.

T

In addition to using debt financing to smooth shocks to net income, the firm can adjust invest-

IP

ment spending to keep the intertemporal budget constraint balanced. In their analysis, LM keep

CR

investment fixed and argue that debt is responsible for smoothing shocks to net income. However, they argue that if shocks to net income present favorable growth opportunities to the firm, man-

NU S

agers may respond by increasing investment and, therefore, both investment and debt policy may smooth shocks to net income. These scenarios imply that payouts do not play a role in balancing the intertemporal budget constraint when there is a change to net income. Ultimately, debt

MA

financing and investment are the key mechanisms the firm can use to absorb fluctuations to net income.

The conjecture that payouts are not used to smooth shocks to net income is based on the idea

ED

that payout policy follows the prediction set forth by the Lintner model (Lintner, 1956). In the Lintner model, payouts are comprised solely of dividends and firm managers infrequently make

PT

changes to payouts from year to year so that dividends paid out to shareholders satisfy a long run

CE

target.3 However, in the LM model payouts will increase in response to shocks to net income, but this response will be larger for persistent shocks than for transitory shocks. The observation that

AC

payouts may increase is not associated with a motivation to balance the budget constraint, but to distribute more income to shareholders when firm earnings are persistent.4 Total payouts used in the LM budget constraint includes cash dividends and stock repurchases net of equity issues. Jagannathan, Stephens and Wiesbach (2000) claim that repurchases are an important source of payouts, and a measure of payouts including both dividends and repurchases performs relatively better in following the Lintner theory of payout policy (Skinner, 2008). Therefore, the use of payouts consisting of dividends and net repurchases may support the implications of the Lintner model for payout policy. Also, a broader measure of payouts beyond dividends may

3

The Lintner Model of payouts is applied to mature firms that distributes dividends. This theory is connected with the motivation to increase managerial compensation. However, we do not further investigate this relationship as it is beyond the scope of this paper. 4

5

ACCEPTED MANUSCRIPT

make the Lintner model more relevant to contemporary firm payout policy since repurchases have

T

supplanted dividends as the primary source of payouts to shareholders for the past 20 years.

IP

To summarize, the LM budget constraint shows that changes to debt and investment policies

CR

can bring balance to the firm’s intertemporal budget constraint in response to a shock to net income. Payouts are not considered to be a mechanism used by the firm to smooth net income

NU S

fluctuations but are influenced to change if shocks to net income are persistent and there is an increase in investment. An increase in investment leads to a permanent increase in net income and, therefore, there will be a new higher target for payouts. Finally, LM derive a joint theory

MA

for debt and payouts while holding investment fixed. For the purpose of our empirical analysis where we explore the smoothness of shocks to net income, we allow both debt and investment to

2.2

ED

simultaneously adjust in response to a shock to net income.

Related Literature and Hypothesis

PT

The determinants of corporate finance policies such as investment, debt and payout have been

CE

extensively studied. To begin with a model without financial frictions, Modigliani and Miller (1958) argue that firm investment depends on the ratio of the value of an additional unit of capital

AC

to its acquisition cost — a ratio known as marginal q. In models with financial constraints, it has been documented that investment is sensitive to cash flow because financial frictions make external finance more expensive than internal finance. In particular, Fazzari, Hubbard and Petersen (1988), Campello, Graham, and Harvey (2010) and Campello et al. (2011) find that financially constrained firms decrease investment in response to a negative shock to cash flow.5 Bolton, Chen and Wang (2011) develop a model which shows that the ratio of marginal q to the marginal value of cash influences investment for financially constrained firms. Generally, in models with financial frictions, the sensitivity of investment to cash flow shocks has

5

In addition to finding firm investment decreases, Campello et al. (2010, 2011) show that financially constrained firms burned more cash, drew more credit from banks, and also engaged in more asset sales in the recent financial crisis.

6

ACCEPTED MANUSCRIPT

implications for debt policy. In DeAngelo, DeAngelo, and Whited (2011) and Denis and McKeon

T

(2012), debt is used to finance investment related capital needs when there is a shortfall in cash.

IP

Specifically, firms temporarily abandon a long run leverage target in favor of increased borrowing to

CR

fund investment. In a model with financial frictions and asymmetric information, Myers and Majluf (1984) find that although firms prefer internal financing to external financing to fund investment,

NU S

external financing primarily takes the form of debt.

Brav, Graham, Harvey, and Michaely (2008) conduct a comprehensive survey of 384 financial executives to ascertain the determinants of payout policy. They find that payouts are preferred

MA

to be smoothed over time but will exhibit positive growth if positive earnings are persistent or investment increases. In addition, in response to an economic shock, executives prefer to use debt to absorb shortfalls in cash before making any changes to their payout policy. The structure of payout

ED

policy has changed over the past few decades where stock repurchases have come to constitute a relatively larger share of payouts. Related studies such as Brown, Liang and Weisbenner (2007) and

PT

Jagannathan, Stephens and Weisbach (2000) examine how the joint distribution of cash dividends

CE

and stock repurchases are set by firms. Surprisingly, work to tie the theories of debt, investment, and payouts into a unified corporate

AC

finance framework is lacking. LM fill this gap by developing a dynamic theory encompassing these three policy instruments. According to LM, a firm’s budget constraint which follows Equation (1) suggests that a dynamic interaction between payout and investment produces debt policy. Likewise, a dynamic theory of debt and investment will yield a dynamic theory of payout. In this paper, we take the first step in empirically examining whether firm behavior is consistent with the implications of the LM budget constraint. Specifically, we attempt to quantify the amount of shocks to net income absorbed through debt and investment, while examining whether payouts are sensitive to changes in net income. Generally, there has been a strong emphasis on the effects of changes in cash flow on debt and investment, but in our analysis we focus on changes to net income. Our hypothesis is that firms absorb a substantial amount of shocks to net income through

7

ACCEPTED MANUSCRIPT

the use of debt and investment and that the sensitivity of payouts to net income is relatively small.

T

To explore our hypothesis, we use a novel variance decomposition methodology that allows us to

CR

IP

estimate the responses of debt, investment and payouts to net income shocks.

3

Empirical Implementation: Methodology and Data

NU S

The focus of this paper is to measure the variability of corporate payouts in response to shocks to net income. Specifically, in an empirical framework that incorporates all three corporate financing decisions, we measure the smoothness of payouts by estimating the amount of shocks to net income

MA

that are absorbed by investment and debt policies. In this section, we describe the empirical

3.1

ED

methodology and data used to estimate the smoothness of payouts.

Methodology

PT

Asdrubali, Sorensen and Yosha (1996) and Sorensen and Yosha (1998) develop an empirical strategy to decompose the variance of shocks to Gross Domestic Product that are absorbed by fiscal policy

CE

and private markets. We adapt this same methodology in the current paper to measure the fraction of shocks to net income (N I) that are smoothed through changes in debt and investment for 6,225

AC

COMPUSTAT firms over the period 1973–2013. The amount of shocks to net income that are left unabsorbed in this empirical framework represents the variability in corporate payouts. We begin by considering the firm’s intertemporal budget constraint developed in LM, which we then use to construct the following identity for N I:

NI =

N I + ∆D NI × × P. N I + ∆D P

(2)

In the above identity, N I stands for net income, ∆D is the change in net debt, and P is the payout of the firm. Shocks to net income can be smoothed by corporate financing decisions such as debt policy and investment. Firms may smooth net income shocks through debt policy, which is reflected

8

ACCEPTED MANUSCRIPT

by the difference between N I and N I + ∆D. If shocks to net income are not fully smoothed by

T

debt policy, then further smoothing can be achieved through investment, which is reflected as the

IP

difference between N I + ∆D and P .

CR

To obtain a system of regression equations that will enable us to estimate the amount of shocks to N I absorbed by changes in debt and the level of investment, we construct a complete decomposition

NU S

of the cross-sectional variance of firm net income growth. We log transform and first difference the variables in Equation (2) to express them as growth rates. Finally, we multiply both sides by ∆ log N I and take expectations of the transformed identity to obtain the following decomposition

MA

of the cross sectional variance in N I:

ED

var{∆ log N I} = cov{∆ log N I, ∆ log N I − ∆ log(N I + ∆D)} + cov{∆ log N I, ∆ log(N I + ∆D) − ∆ log P }

PT

+ cov{∆ log N I, ∆ log P }.

(3)

CE

Dividing both sides of Equation (3) by the variance of ∆ log N I obtains the slope coefficients

AC

from three different panel univariate regressions which sum to 1:

1 = βD + βI + βP .

(4)

Our estimates of net income smoothing are based on two slope coefficients, which are βD and βI . βD , the slope coefficient in the regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I, is represented as the percentage of smoothing of a net income shock by debt; βI , the slope coefficient in the regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I, is represented as the percentage of smoothing of a net income shock by investment. The third slope coefficient, βP , is obtained from a regression of ∆ log P on ∆ log N I and represents the amount of shocks to N I that is left unsmoothed; specifically, it measures the sensitivity of firm payouts to shocks to N I. 9

ACCEPTED MANUSCRIPT

According to a theoretical prediction in LM, if investment is held fixed and there is a preference

T

for payout smoothing, debt absorbs all of the shocks to net income. This prediction implies that

IP

Equation (4) yields βD = 1 and βI = βP = 0. Furthermore, Lambrecht and Myers (2014) argue

CR

that once an optimal investment policy is implemented, debt policy is the main shock absorber of net income shocks and, thus, smooths the shocks in investment and payouts. Since debt smooths

NU S

all of the variation in net income, the coefficient βD takes on a value of 1; the other coefficients, βI and βP , are both equal to 0 because there are no remaining shocks to be passed through to investment and payouts. However, Lambrecht and Myers (2012, 2014) allow for the possibility that

MA

both debt policy and investment can absorb net income shocks. The main purpose of our analysis is to examine the adjustments of debt and investment to keep payouts smooth in response to net income shocks.

ED

The coefficients from Equation (4) are estimated according to the following three panel regres-

PT

sions:

(5a)

∆ log(N I + ∆D)it − ∆ log Pit = αIi + γIt + βI ∆ log N Iit + ǫitI ,

(5b)

∆ log Pit = αPi + γPt + βP ∆ log N Iit + ǫitP .

(5c)

AC

CE

i t ∆ log N Iit − ∆ log(N I + ∆D)it = αD + γD + βD ∆ log N Iit + ǫitD ,

In the above equations, i indexes firms, t indexes years, αi are firm fixed effects and γ t are year fixed effects. Since the variables in the panel equations above are expressed in growth rates we can economically interpret a hypothetical 100% increase in the growth rate of N I. According to Equation (5a), if shocks are perfectly smoothed by paying down debt, the growth rate of ∆D is −100% which induces the term N I + ∆D to grow at a rate of 0. A regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I yields the coefficient βD equal to 1 in the case where shocks to net income are completely absorbed by a policy of paying down debt. If N I increases by 100% and there is no smoothing at all at the debt level, then N I + ∆D grows at the same rate as N I and,

10

ACCEPTED MANUSCRIPT

therefore, regressing ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I yields βD equal to zero.

T

The coefficient βD represents the percentage of smoothing of a net income shock carried out by

IP

debt (i.e., through paying down debt). If debt does not fully smooth shocks to net income, then

CR

there is scope for further smoothing of the remaining shocks by changes in investment. Moving beyond Equation (5a), the next level of smoothing occurs in Equation (5b) where the dependent

NU S

variable is investment which is constructed as the difference between ∆ log(N I + ∆D) and ∆ log P ; the coefficient βI measures the incremental amount of shocks to net income that are passed through to investment. In this equation, the term P growing at the same rate as N I +∆D implies that there

MA

is no smoothing of shocks to net income at the investment level; therefore, regressing ∆ log(N I + ∆D)−∆ log P on ∆ log N I yields βI equal to zero. In the event that debt policy does not completely smooth net income shocks, complete smoothing by investment will imply that βI is equal to 1 − βD .

ED

In other words, if the combination of debt and investment completely smooths net income shocks then βD + βI = 1.

PT

However, if βD + βI < 1, the combination of debt and investment incompletely absorbs shocks

CE

to net income and, therefore, payout is sensitive to changes in net income. By construction of the variance decomposition equation, the sensitivity of payouts, βP , is equal to 1 − βD − βI . In

AC

Equation (5c), βP is estimated; this coefficient reflects the growth of payouts or percentage of net income shocks left unabsorbed. To demonstrate the economic implication of βP consider a firm that prefers not to use investment to absorb shocks to net income (i.e., βI = 0) but allows payouts (βP ) to constantly grow at 3%. This payout policy would imply that debt (βD ) absorbs 97% of shocks to net income. Finally, the coefficients βD , βI , and βP are estimated from panel regressions, which include both year and firm fixed effects. These fixed effects are included to demean each observation by its yearly mean and firm level mean to control for macroeconomic effects and firm-level unobservables.

11

ACCEPTED MANUSCRIPT

3.2

Data

T

To estimate the amount of shocks to net income absorbed by debt and investment, and the amount

IP

left unabsorbed (i.e., growth in total payout), we collect annual data on U.S. firms from the Com-

CR

pustat database for the period 1973–2013. Specifically, we obtain data on variables such as net income, short term and long term debt, cash balances, cash dividends, stock repurchases and equity

NU S

issues. Our measure of total payout is cash dividends plus stock repurchases minus equity issues. We exclude utility firms with SIC codes between 4900 and 4999 and financial firms with SIC codes between 6000 and 6999. We also collect data on US gross domestic product (GDP) and deflate

MA

our variables using the GDP price deflator, which are both obtained from the Bureau of Economic Analysis (BEA). To reduce the potential impact of outliers, we winsorize our sample at the 1st and

ED

99th percentiles. Our baseline sample is an unbalanced panel with 6,225 firms where we exclude firms that do not distribute cash dividends. In Panel A of Table 1, the means and standard devia-

PT

tions for net income, debt, investment, and payouts are reported. A correlation matrix is reported

4.1

Results

AC

4

CE

in Panel B which shows the correlation between each of the variables.

Baseline Results

In Table 2, we present our baseline results. We find that both paying down debt and increases in investment absorb 97.6% of positive shocks to net income. Specifically, 56.9% of shocks to net income are absorbed via debt and 40.7% of shocks are absorbed via investment. Our results suggest that both debt and investment play primary roles in smoothing a large fraction of shocks to net income; however, not all of the shocks to net income are smoothed. We find that a 100% increase in net income is associated with a 2.4% increase in payouts, a figure which represents the amount of shocks left unsmoothed. Our results confirm the intuition of the LM budget constraint that debt and investment are the primary mechanisms which smooth fluctuations in net income. Also,

12

ACCEPTED MANUSCRIPT

payouts are sensitive to changes in net income which can be explained by increases in investment

T

or by permanent shocks to net income.

IP

In our empirical methodology, we do not differentiate between temporary and permanent shocks

CR

to net income. However, in a later section, we extend our first difference model to longer difference intervals to capture the responses of debt, investment and payouts to permanent shocks to net

NU S

income. Nevertheless, our observed pattern of debt and investment smoothing is similar to Daniel, Denis and Naveen (2010) who find that debt and investment each cover half of a shock to cash flow. Furthermore, we find that there is a relatively small but economically significant fraction of shocks

MA

to net income left unabsorbed. Specifically, our measure of payouts, which is cash dividends plus repurchases minus equity issues, increases by 2.4% in response to a shock to net income.

Controlling For Business Cycle Fluctuations and Capital Market Condi-

ED

4.2

tions

PT

We have established the smoothing patterns of net income, but this has been done without ex-

CE

amining the potential effects of business cycle fluctuations and capital market conditions. In this section, we present results to show that, conditional on these aggregate measures, the patterns of

4.2.1

AC

smoothing of net income are robust. Business Cycle Fluctuations

To examine the impact of business cycle fluctuations on the smoothing of shocks to net income, we re-estimate the system of panel equations ((5a), (5b), and (5c)) by parameterizing each of the three different slope coefficients as the following:

βZ = βZ,0 + βZ,1 ∆ log GDPt .

(6)

Z is a generic variable denoting whether the slope coefficient is a measure of smoothing via debt (D) or via investment (I), or a measure of the amount of shocks unsmoothed which reflects the 13

ACCEPTED MANUSCRIPT

response of payouts (P ). GDP is US gross domestic product and changes to log GDP represent

T

business cycle fluctuations. Multiplying both sides of Equation (6) by ∆ log N Iit produces the

CR

IP

following expression:

βZ ∆ log N Iit = βZ,0 ∆ log N Iit + βZ,1 (∆ log GDPt × ∆ log N Iit ).

(7)

NU S

According to Equation (7), the slope coefficient on net income growth (βZ ) is a linear combination of the regression coefficients for the main effect term (∆ log N Iit ) and the interaction term (∆ log GDPt × ∆ log N Iit ).6 In the regressions, if the interaction effect captured by the coeffi-

MA

cient βD,1 is significantly positive, this implies that positive shocks to net income will be absorbed through debt when there is positive growth in GDP . Conditional on economic growth, if βI,1 and

ED

βP,1 are statistically negative, there is dis-smoothing by investment and payouts negatively grow in response to positive net income shocks.7

PT

In Panel A of Table 3, we report the cyclical patterns of the channels of net income smoothing. Column 2 shows that positive shocks to net income are absorbed by debt in a procyclical manner:

CE

more debt is paid off during economic expansions and more debt is accumulated during recessions. This finding may be inconsistent with the results documented in Covas and Den Haan (2011)

AC

where it is suggested that debt issuances increase during economic expansions. But their result is conditional on holding shocks to net income constant while we find that firms pay off debt during expansions to absorb positive shocks to net income. McLean and Zhao (2014) support our finding by showing that debt issuances decrease in response to an increase in cash flow during expansions. Interestingly, the pattern of investment is found to be countercyclical with the business cycle. We find that, during expansions, firms substitute the smoothing of shocks to net income away from investment and towards paying down debt. Firms may behave in this manner to pay off debt 6

If log GDP is held constant in Equation (7), then ∆ log GDP is equal to 0 and, hence, βZ is equal to the main effect βZ,0 . 7 We exclude ∆ log GDPt as a non-interacted term in Equation (7) because the inclusion of year and firm fixed effects in the decomposition equations will remove variables which are invariant along both year and firm dimensions.

14

ACCEPTED MANUSCRIPT

obligations and accumulate precautionary savings which they can use to fund investment during

T

recessions. In addition, McLean and Zhao (2013) argue that during economic expansions the cost

IP

of external financing is relatively low and, therefore, firms prefer to use external financing to fund

CR

investment. They empirically show that investment decreases in response to positive cash flow shocks during expansions.

NU S

Conditional on positive shocks to net income, we find that payouts are also countercyclical with economic expansions. Lambrecht and Myers (2012) argue that there is positive comovement between payouts and investment. Therefore, since investment is found to be countercylical with

MA

the business cycle given a positive shock to net income, payouts should also be countercyclical with economic expansions. Additionally, if the cost of external financing is relatively low during expansions, firms may prefer to finance payouts out of external funds rather than internal funds.

ED

The non-interacted main effects reported in Panel A show that both debt and investment absorb roughly 97% of shocks to net income. These results are consistent with the baseline results reported

PT

in Table 2. However, during booms and busts, the channels of smoothing of net income display

CE

different patterns. We find that during expansions there is increased smoothing of net income by debt but dis-smoothing through investment. We also find that positive shocks to net income are

4.2.2

AC

transmitted to payouts less during the business cycle boom. Capital Market Conditions

We have examined the effects of macroeconomic conditions on firm behavior. However, conditions in capital markets might also affect firm behavior because they potentially influence the firm’s ability to raise capital and accessibility to these markets. We apply the same methodology as in the previous section to estimate the effect of capital market conditions on firm behavior. The main variable of interest is the interaction term, where we interact net income with a measure reflecting capital market conditions, rather than GDP, in the regressions. To proxy for capital market conditions, we use the Wilshire 5000 Total Market Index. This data is a market capitalization

15

ACCEPTED MANUSCRIPT

weighted index of the market value of all stocks traded in the US. We collect the annual data from

T

the Federal Reserve Bank of St. Louis FRED database.8

IP

In Panel B of Table 3, we show that the non-interacted main effects are similar to the regressions

CR

where we control for the business cycle. Specifically, debt absorbs 53.83% of shocks to net income; investment absorbs 42.60% of shocks to net income. However, when we control for capital market

NU S

conditions the elasticity of payouts is 3.57%, which is slightly larger than that reported in Panel A. Also, we find the coefficients on the interaction terms follow similar patterns to those reported in Panel A. Conditional on a positive shock to net income, when capital market conditions are

MA

relatively good, as reflected by an increase in the value of Wilshire Index, we report that debt is procyclical. Furthermore, conditional on a positive shock to net income, we find that investment and payouts are countercyclical with upturns in capital markets.

ED

Investment and payout may be countercyclical with capital market conditions because of the potential changes to the cost of external financing. During upturns in capital markets, the cost of

PT

external financing may be relatively low and, therefore, firms may use external sources to finance

Different Sample Periods

AC

4.3

CE

investment spending and payouts rather than own cash flows.

We exploit our long time series data by disaggregating the data into sub-periods. Different events such as the Great Moderation, the Great Recession and other recessionary periods, or strict and lax periods of regulation may have had an effect on the ability of firms to smooth shocks to net income. Although we do not isolate our examination of firm behavior explicitly for specific and significant events, we estimate our variance decomposition equations based on our sample period that is decomposed into time intervals. Starting with the interval between the years 1973–1982, Table 4 shows that firms absorb 53.93%

8

We use the Standard & Poor’s (S&P) 500 index as an alternative proxy for capital market conditions, also. The results using the S&P 500 index in the estimations are similar to the results reported in Panel B of Table 3 which uses the Wilshire 5000 index.

16

ACCEPTED MANUSCRIPT

of shocks to net income through debt and 41.08% through investment, leaving the amount of shocks

T

unabsorbed at 4.99%, which is the elasticity of payouts to net income. In the years 1983–1992,

IP

firms are paying down more debt in response to shocks to net income and investment absorbs a

CR

larger share of shocks during this time period. The sensitivity of payouts falls to 3.42%, indicating that there are less shocks unabsorbed. We observe similar patterns of smoothing during the years

NU S

1983–1992 as for the period 1993–2006. Specifically, shocks to net income are absorbed by debt on the order of 58.90 to 59.05%, and smoothed by investment on the order of 37.67 to 39.89% between these two periods. However, because of the marginal increases in smoothing by debt and

MA

investment, the amount of shocks left unabsorbed falls from 3.42 to 1.06% between 1983–1992 and 1993–2006.

Finally, in the years 2007–2013, we find that firms are using less debt but more investment to

ED

absorb shocks to net income. Also, during this period, the sensitivity of payouts, or the fraction of shocks left unabsorbed, falls to 0.85%. These results are interesting because they reveal that

PT

firms absorbed a larger share of shocks through investment than debt policy. Perhaps due to the

CE

tightening of credit in capital markets, firms preferred to use investment to promote growth in firm income in response to positive shocks to net income, which would leave them less dependent on

4.4

AC

credit in the event of future negative shocks to net income.

Varying Difference Intervals

The empirical framework used in this paper constructs the one year difference of variables in logarithms to yield growth rates. We extend the empirical model beyond the one year differencing to include longer multi-year differences (k years). If firm adjustment to shocks to net income vary depending on whether the shocks are temporary or permanent, specifications with longer differences may yield estimates different from those based on shorter differences. Additionally, if the regressor is measured with error, longer differenced specifications may produce less biased estimates compared to shorter differenced estimates. Thus, as we compare elasticity estimates at

17

ACCEPTED MANUSCRIPT

varying difference lengths, we take into account potential measurement errors in the regressor as

T

well as the “long-run” response of the firm to more persistent shocks to net income.

IP

In Table 5, we show the baseline results along with the results from specifications differenced

CR

up to k = 10. As k increases, the fraction of smoothing at the debt level falls until it reaches 41% at the 10 year interval. In contrast, the percentage of shocks absorbed by investment increases as

NU S

k increases until it reaches a peak of 48%. The fraction of shocks unsmoothed also increases as the difference interval widens. When k = 10, the amount of shocks left unsmoothed, as represented as the coefficient on payouts, is 11%.

MA

Overall, the increase in k is associated with increases in payouts; this is consistent with the prediction set forth by LM budget constraint. As the difference interval increases, the percentage of smoothing captured by debt and investment reflect long run responses. According to Lambrecht

ED

and Myers (2014), if shocks to net income are persistent, investment increases and payouts follows suit by increasing as well. They argue that as investment increases, payouts should also increase

PT

since investment leads to future profitability for the firm. According to Table 5, as k reaches 5

CE

years, investment increases and levels out at 48%. We find that payouts follows the same pattern as investment which increases and eventually levels out.

Different Industries

AC

4.5

The effects of cash flow on corporate financing decisions may differ across different industries. As such, industry type may play a prominent role in the smoothing of shocks to net income by firms. In this section, we re-estimate our variance decomposition equations for firms in each of the 1 digit standard industrial classification (SIC) industries. We compare the patterns in the debt, investment, and payout for firms in the following five sectors: mining and construction, manufacturing, transportation, trade, and services. According to Table 6, debt policy for manufacturing firms absorbs the largest share of shocks to net income compared to other firms. Specifically, 61.12% of net income shocks are absorbed by

18

ACCEPTED MANUSCRIPT

manufacturing firms. Across the other four industry sectors, the amount of shocks to net income

T

smoothed by debt policy is on the order of 49.12 to 60.44%. Table 6 also shows that investment

IP

policy for manufacturing firms smoothes the smallest share of net income shocks. Manufactur-

CR

ing firms absorb 36.77% of shocks while investment policy for firms in the transportation sector smoothes 49.26% of shocks, which represents the largest share of net income shocks smoothed by

NU S

investment across the five different sectors. Consistent with the baselines results reported in Table 2, we find that debt policy is the main absorber of net income shocks for all firms. The amount of shocks left unabsorbed by debt and investment policies reflects sensitivity of firm

MA

payout. Compared to the baseline result for payouts, we find that payouts for firms in the trade sector is highly sensitive to net income shocks. Payouts for trade sector firms grow by 6.78% in response to a 100% increase in net income. For the other remaining sectors, we find relatively less

ED

sensitivity in payouts with the percentage of shocks left unabsorbed ranging from 1.17 to 2.16%. It is interesting to infer why payouts in the trade sector are more sensitive compared to other firms.

PT

Foster et al. (2006) conjecture that retail trade firms attempt to grow by opening new stores rather

CE

than expanding existing stores. Perhaps to finance this expansion through the formation of new stores, retail trade firms issue relatively more payouts to attract investors.

Firm Size

AC

4.6

It is possible that cash flow may be more stable for large firms compared to small firms. Therefore, in response to net income shocks, payouts for larger firms may exhibit relatively less sensitivity to net income shocks. In this section, we explore whether payout smoothing differs among firms conditional on firm size. According to Table 7, we characterize firms as small, medium and large based on their average total assets. We classify firms as small if their total assets are in the bottom 25% of our sample, firms as medium if total assets are in the 2nd or 3rd quartile, and firms as large if total assets are in the top 25%. We find that small firms use debt to absorb a relatively larger fraction of shocks to net income. Specifically, for small firms, debt absorbs 58.25% of shocks to

19

ACCEPTED MANUSCRIPT

unabsorbed is captured by the coefficient on βP which is 4.87%.

T

net income while investment absorbs 36.88% of shocks to net income. The amount of shocks left

IP

The patterns of debt and investment for medium and large firms are similar to that for small

CR

firms. We find that for these firms, debt absorbs a large fraction of changes to net income ranging from 55.02% to 57.61%. For medium firms, investment smoothes 41.98% of shocks to net income,

NU S

while investment for large firms absorbs 40.20%. In addition, we document that payouts become less sensitive to changes in net income where the growth of payouts falls from 3% to 2.19% as we move from medium to large firms.

MA

Overall, we find that the elasticity of payouts is relatively higher for smaller firms than for large firms. In response to a positive shock to net income, small firm behavior may deviate from the LM model by issuing more income to shareholders to attract new investors. Small firms may have more

ED

limited access to credit markets and thus exhibit more volatility in earnings relative to large firms. In response, shareholders might demand more income to compensate them for the risk in investing

PT

in these small firms. Once the firm’s total assets grow, payouts become relatively less sensitive to

CE

shocks to net income. Our result on payout smoothing for small firms is consistent with Leary and Michaely (2011) who find reduced payout smoothing for small firms compared to larger firms.

Young Vs. Mature Firms

AC

4.7

The response of corporate payouts may differ at various stages in the life cycle of a firm. For example, aging firms may issue more payouts than recently established firms. Therefore, in contrast to young firms, mature firms may have a stronger preference to smooth payouts. In this section, we re-estimate the variance decomposition equations by splitting the data to account for firm age. We classify firms as young or mature, depending on their age relative to the date of their Initial Public Offering (IPO). Firms that have been publicly traded for ten years or less are classified as young; otherwise, firms are classified as mature. According to Table 8, young firms absorb 64.38% of shocks to net income through debt policy and investment absorbs 33.45% of shocks. In contrast,

20

ACCEPTED MANUSCRIPT

the patterns of net income smoothing among mature firms is almost identical to the baseline results

T

reported in Table 2. Specifically, debt policy absorbs 55.88% of shocks to net income and investment

IP

policy absorbs 41.72% of shocks.

CR

The growth in payouts for mature firms is slightly larger than that for young firms. Young and mature firms distribute roughly 2.17% and 2.41% of shocks to net income as payouts, respectively.

NU S

The coefficient on payouts for mature firms is very similar to that reported in the baseline results. This is not surprising since the majority of the firms in our baseline sample are mature firms. However, we find that young firms smooth relatively more shocks to net income via debt, and less

MA

shocks are distributed to investment and payout. A possible reason for these findings is that young firms may be trying to pay down more debt in order to finance future investment. Therefore, debt policy for young firms will absorb relatively more shocks to net income compared to investment.

ED

Furthermore, our findings suggest that the response of payouts to net income shocks does not vary much across young and mature firms. Although the coefficient on payouts for mature firms

PT

is slightly larger than that for young firms, these coefficients are within one standard deviation of

CE

each other. Therefore, we conclude that both young and mature firms try to smooth their payouts to roughly the same extent.

External Finance: Independent Firms Vs. Dependent Firms

AC

4.8

Many studies have shown that firms in different industries utilize their cash flows differently. Firms in certain industries are in need of external finance, while other firms can raise the required funds internally to finance projects. The ability to raise the required funds internally or externally may affect the firm’s payout policy. Following Rajan and Zingales (1998), we rank industries based on their external dependency and split the data sample into two categories: firms in externally dependent industries and firms in externally independent industries. The firms in the top 50% of the ranking are considered in the “dependent on external finance” category, and the other half in the bottom of the ranking are considered “independent from external finance.” Our sample in this

21

ACCEPTED MANUSCRIPT

empirical test includes only manufacturing firms.

T

According to Table 9, firms in sectors that do not depend on external financing absorb 61.19%

IP

of shocks to net income through debt policy, and 36.46% through investment policy. Firms in the

CR

sectors in need of external finance distribute 61.25% of a shock to net income to debt policy and 36.85% to investment policy. There is a slight difference in the smoothing of payouts where firms

NU S

independent of external finance have their payouts grow by 2.35% in response to a shock to net income. Further, payouts grow by 1.9% for firms in need of external finance. Firms dependent on external financing might have a relatively lower percentage of net income shocks left unabsorbed

MA

compared to independent firms as they may want to pay down more debt to increase their creditworthiness and to increase investment to become more profitable in the future which will enable

4.9

ED

them to continue to have access to external funds.

Different Measures of Payout

PT

In our estimations, we use a sample of firms that distribute dividends to shareholders. For these

CE

firms, we construct a measure of payout which includes both dividends and net stock repurchases. In Table 10, we examine the fraction of shocks to net income absorbed using different measures

AC

of payouts in our specifications. In Column (1), the data sample is restricted to firms that issue dividends and the measure of payouts consist of only dividends. For firms that distribute only dividends (excluding net repurchases), debt and investment respectively absorb 56.61% and 42.86% of shocks to net income. The response of dividends to shocks to net income left unabsorbed is relatively small; a 100% increase in net income is associated with a 0.53% increase in payouts. In Column (2), the data sample is restricted to firms that do not issue dividends, and the measure of payouts consists of only net repurchases. For firms that only issue or buy back corporate stocks, the responses of debt and investment to shocks to net income are similar to those reported in Column (1); in addition, net repurchases increase by 0.87% in response to a 100% increase in net income, which is slightly larger than the growth of dividends in Column (1).

22

ACCEPTED MANUSCRIPT

In Columns (3), (4), and (5), all firms available in the dataset without our baseline restrictions

T

are used in the sample to re-estimate the baseline model based on different measures for payouts.

IP

In Column (3), we find that when payouts are only comprised of dividends, debt absorbs 57.69%

CR

of shocks to net income and investment absorbs 42.13% of shocks. Also, we find that the fraction left unabsorbed is 0.18%. In Column (4), which considers firm payouts that are only comprised

NU S

of repurchases, firms use debt and investment to respectively absorb 57.47% and 40.9% of shocks to net income. Payouts increase by 1.64% in response to a 100% increase in net income. Finally, in Column (5), where only firms whose payouts consist of both dividends and repurchases are

MA

considered, the patterns of debt, investment, and payouts found are similar to those contained in Column (4).

Conclusion

ED

5

PT

In this paper, we extend the literature on corporate governance by applying a variance decomposition methodology to evaluate the financial adjustments firms make in response to changes in net

CE

income. Based on the theories proposed by Lambrecht and Myers (2012, 2014), we decompose the variance in net income to quantify the amount of shocks to net income absorbed by investment and

AC

debt policy. In our baseline results, firms use debt policy and investment to respectively absorb 56.9% and 40.7% of shocks to net income. Payouts, which is another important decision firms make in this adjustment process, grow by 2.4% in response to a positive shock to net income. Our findings conform with the predictions set forth by Lambrecht and Myers (2012, 2014) in that firms will maintain smooth payouts by using investment and debt policy to absorb shocks to net income. In addition, the findings in this paper support the conclusions of Daniel, Denis, and Naveen (2010) that firms use primarily investment and debt policy to absorb financial shocks while making relatively small adjustments to payouts. Overall, these results suggest that managers have an incentive to keep adjustments to payouts small while allowing investment and debt policy to absorb most of the shocks to net income. Our conceptual and methodological contributions 23

ACCEPTED MANUSCRIPT

open an important avenue for future research to fill in the empirically unexplored relation between

AC

CE

PT

ED

MA

NU S

CR

IP

T

managerial compensation and firm payouts.

24

ACCEPTED MANUSCRIPT

T

References

IP

[1] Asdrubali, P., Sorensen, B.E., and Yosha, O. (1996) “Channels of Interstate Risk Sharing:

CR

United States 1963-1990,” Quarterly Journal of Economics, 111(4), 1081-1110. [2] Bolton, P., Chen, H., and Wang, N. (2011). “A Unified Theory of Tobin’s q, Corporate Invest-

NU S

ment, Financing, and Risk Management,” Journal of Finance, 66(5), 1545-1578. [3] Brav, A., Graham, J.R., Harvey, C.R., and Michaely, R. (2005). “Payout Policy in the 21st

MA

Century,” Journal of Financial Economics, 77(3), 483-527. [4] Brown, J.R., Liang, N., and Weisbenner, S. (2007). “Executive Financial Incentives and Payout Policy: Firm Responses to the 2003 Dividend Tax Cut,” Journal of Finance, 62(4), 1935-1965.

ED

[5] Campello, M., Giambona, E., Graham, J.R., and Harvey, C.R. (2011). “Liquidity Management

PT

and Corporate Investment During a Financial Crisis,” Review of Financial Studies, 24(6), 1944-1979.

CE

[6] Campello, M., Graham, J.R., and Harvey, C.R. (2010). “The Real Effects of Financial Con-

AC

straints: Evidence From a Financial Crisis,” Journal of Financial Economics, 97(3), 470-487. [7] Chang, X., Dasgupta, S., Wong, G., and Yao, J. (2014). “Cash Flow Sensitivities and the Allocation of Internal Cash Flow,” Forthcoming in Review of Financial Studies. [8] Covas, F., and Den Haan, W.J. (2011). “The Cyclical Behavior of Debt and Equity Finance,” American Economic Review, 101(2), 877-899. [9] Daniel, N.D., Denis, D.J., and Naveen, L. (2010). “Sources of Financial Flexibility: Evidence from Cash Flow Shortfalls,” Working Paper. [10] DeAngelo, H., DeAngelo, L., and Whited, T.M. (2011). “Capital Structure Dynamics and Transitory Debt,” Journal of Financial Economics, 99(2), 235-261.

25

ACCEPTED MANUSCRIPT

[11] Denis, D.J., and McKeon, S.B. (2012). “Debt Financing and Financial Flexibility Evidence

IP

T

from Proactive Leverage Increases,” Review of Financial Studies, 25(6), 1897-1929. [12] Fazzari, S.M,. Hubbard, R.G., and Petersen, B.C. (1988). “Financing Constraints and Corpo-

CR

rate Investment,” Brookings Papers on Economic Activity, 1, 141-195.

NU S

[13] Foster, L., Haltiwanger, J. and Cornell, K.J. (2006). “Market Selection, Reallocation, and Restructuring in the US Retail Trade Sector in the 1990s,” The Review of Economics and Statistics, 88(4), 748-758.

MA

[14] Gatchev, V.A., Pulvino, Todd, and Tarhan, V. (2010). “The Interdependent and Intertemporal Nature of Financial Decisions: An Application to Cash Flow Sensitivities,” Journal of Finance,

ED

65(2), 725-763.

ness, 76(1), 151-192.

PT

[15] Goel, A.M., and Thakor, A.V. (2004). “Why Do Firms Smooth Earnings?” Journal of Busi-

CE

[16] Gormley, T.A., and Matsa, D.A. (2014). “Playing it Safe? Managerial Preferences, Risk, and Agency Conflicts,” Working Paper.

1165.

AC

[17] Hennessy, C.A., and Whited, T.M. (2005). “Debt Dynamics,” Journal of Finance, 60(3), 1129-

[18] Hennessy, C.A., and Whited, T.M. (2007). “How Costly Is External Financing? Evidence from a Structural Estimation,” Journal of Finance, 62(4), 1705-1745. [19] Jagannathan, M., Stephens, S.P., and Weisbach, M.S. (2000). “Financial Flexibility and the Choice Between Dividends and Stock Repurchases,” Journal of Financial Economics, 57(3), 355-384. [20] Lambrecht, B.M., and Myers, S.C. (2012). “A Lintner Model of Payout and Managerial Rents,” Journal of Finance, 67(5), 1761-1810. 26

ACCEPTED MANUSCRIPT

[21] Lambrecht, B.M., and Myers, S.C. (2014). “The Dynamics of Investment, Payout and Debt,”

IP

T

Working Paper. [22] Leary, M.T., and Michaely, R. (2011). “Determinants of Dividend Smoothing: Empirical Evi-

CR

dence,” Review of Financial Studies, 24(10), 1897-1929.

NU S

[23] Lintner, J. (1956). “Distribution of Incomes of Corporations Among Dividends, Retained Earnings, and Taxes,” American Economic Review, 46(2), 3197-3249. [24] McLean, R.D., and Zhao, M. (2014). “The Business Cycle, Investor Sentiment, and Costly

MA

External Finance,” Journal of Finance, 69(3), 1377-1409. [25] Modigliani, F., and Miller, M.H. (1958). “The Cost of Capital, Corporation Finance and the

ED

Theory of Investment,” American Economic Review, 48(3), 261-297. [26] Myers, S.C., and Majluf, N.S. (1984). “Corporate Financing and Investment Decisions When

PT

Firms Have Information That Investors Do Not Have,” Journal of Finance, 13(2), 187-221.

CE

[27] Skinner, D. J. (2008). “The Evolving Relation Between Earnings, Dividends, and Stock Repurchases,” Journal of Financial Economics, 87(3), 582-609.

AC

[28] Ostergaard, C., Sasson, A., and Sorensen, B.E. (2011). “The Marginal Value of Cash, Cash Flow Sensitivities, and Bank-Finance Shocks in Nonlisted Firms,” Working Paper. [29] Rajan, R.G., and Zingales, L. (1988). “Financial Dependence and Growth,” American Economic Review, 88(3), 559-586. [30] Sorensen, B.E., and Yosha, O. (1998) “International Risk Sharing and European Monetary Unification,” Journal of International Economics, 45(2), 211-238. [31] Strebulaev, I.A., and Whited, T.M. (2012). “Dynamic Models and Structural Estimation in Corporate Finance,” Foundations and Trends in Finance, 6(1-2), 1-163.

27

ACCEPTED MANUSCRIPT

1081.41

Debt

786.17

5471.71

Investment

295.53

1053.70

Payout

139.75

996.69 63478

ED

Observations

1

NU S

225.22

MA

Net Income

Panel B: Correlation Matrix Net Income Debt Investment (3) (4) (5)

CR

Panel A: Summary Statistics Mean Standard Deviation (1) (2)

IP

Sample Period: 1973-2013

T

Table 1: Summary Statistics and Correlation Matrix

0.36

1

0.59

0.54

1

0.51

0.31

0.41

63478

In Panel A, the means and the standard deviations of the variables (in millions) are reported. In Panel B, the

AC

CE

covers the period 1973–2013.

PT

correlations of the variables are reported. Sample consists of 6,225 firms taken from the Compustat database and

28

ACCEPTED MANUSCRIPT

CR

Debt (βD )

IP

Sample Period: 1973-2013

T

Table 2: Baseline Results

NU S

Investment (βI )

MA

Payout (βP )

Observations

56.85 (0.71) 40.72 (0.73) 2.43 (0.24) 63478

ED

This table displays the baseline results from estimating the variance decomposition equations. The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973–2013. The numbers displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel

PT

payout (βP ) represents the fraction of shocks unabsorbed. Also, it measures the response of payouts to shocks to net income. The measure for payouts is the sum of cash dividends and stock repurchases net of equity issues. βD is the slope coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI is the slope coefficient

CE

from a regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I. βP is the slope coefficient from a regression of ∆ log P on ∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses. Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied

AC

by 100. All coefficients are significant at the 1% level.

29

ACCEPTED MANUSCRIPT

53.75 (1.13)

123.86 (35.31)

Investment (βI )

43.46 (1.16)

Payout (βP )

2.78 (0.38)

IP

0.12 (0.04)

-109.68 (36.22)

42.60 (1.22)

-0.08† (0.03)

-14.17 (11.79)

3.57 (0.40)

-0.05 (0.01)

MA

63478

63478

ED

Observations

53.83 (1.19)

NU S

Debt (βD )

Panel B: Wilshire Index Main Interaction effect (βZ,0 ) effect (βZ,1 ) (3) (4)

CR

Panel A: GDP Main Interaction effect (βZ,0 ) effect (βZ,1 ) (1) (2)

T

Table 3: Controlling For Business Cycle Fluctuations and Capital Market Conditions

This table displays the results from estimating the variance decomposition equations with the inclusion of interaction effects. The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973–2013. The

PT

numbers displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks unabsorbed. The measure for payouts is the sum of cash dividends and stock repurchases net of equity issues. In the panel regressions in Panel A, βZ is defined

CE

as βZ = βZ,0 + βZ,1 ∆ log GDP , where Z is a generic variable denoting either debt (D), investment (I) or payout (P ). Multiplying βZ by ∆ log N I yields βZ,0 , the coefficient on the main term (∆ log N I), and βZ,1 , the coefficient on the interaction term (∆ log GDP × ∆ log N I). In Panel B, the panel regressions are repeated, except that GDP

AC

is replaced with the value of the Wilshire Index. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses. Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied by 100. All coefficients are significant at the 1% level, except for βP,1 which is statistically insignficant and those labeled with † which indicates they are significant at the 5% level.

30

ACCEPTED MANUSCRIPT

Debt (βD )

53.93 (1.78)

58.90 (1.46)

Investment (βI )

41.08 (1.86)

37.67 (1.51)

Payout (βP )

4.99 (0.62)

Observations

21082

1993-2006 (3)

IP

1983-1992 (2)

CR

1973-1982 (1)

T

Table 4: Different Sample Periods

2007-2013 (4) 54.31 (2.01)

39.89 (1.31)

44.83 (1.99)

3.42 (0.57)

1.06 (0.40)

0.85† (0.44)

15922

18454

8020

MA

NU S

59.05 (1.28)

This table displays the results from estimating the variance decomposition equations separately for different sub-

ED

periods. The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973–2013. The numbers displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks unabsorbed. βD is the slope

PT

coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI is the slope coefficient from a regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I. βP is the slope coefficient from a regression of ∆ log P on ∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses.

CE

Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied by 100.

AC

All coefficients are significant at the 1% level, except for those labeled with † which indicates they are significant at the 5% level.

31

ACCEPTED MANUSCRIPT

k=3 (3)

k=5 (4)

k = 10 (5)

Debt (βD )

56.85 (0.71)

49.71 (0.69)

44.68 (0.70)

41.18 (0.74)

41.72 (0.93)

Investment (βI )

40.72 (0.73)

44.17 (0.72)

47.18 (0.75)

48.67 (0.80)

47.62 (1.04)

Payout (βP )

2.43 (0.24)

6.12 (0.28)

8.15 (0.33)

10.14 (0.40)

10.65 (0.55)

Observations

63478

59038

53558

44269

27961

IP

k=2 (2)

MA

NU S

CR

k=1 (1)

T

Table 5: Varying Difference Intervals

This table displays the results from estimating the variance decomposition equations separately for each of the

ED

different lengths of the difference interval (k). The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973–2013. The numbers displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks

PT

unabsorbed. βD is the slope coefficient from a regression of ∆k log N I − ∆k log(N I + ∆k D) on ∆k log N I. βI is the slope coefficient from a regression of ∆k log(N I + ∆D) − ∆ log P on ∆k log N I. βP is the slope coefficient from a regression of ∆k log P on ∆k log N I. For a generic variable, Xt , ∆k Xt is represented as ∆k Xt = Xt − Xt−k . Variables

CE

are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses. Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied by 100. All coefficients

AC

are significant at the 1% level.

32

ACCEPTED MANUSCRIPT

Debt (βD )

51.60 (2.31)

61.12 (0.94)

Investment (βI )

46.23 (2.42)

36.77 (0.96)

Payout (βP )

2.16 (0.86)

2.11 (0.31)

Observations

4720

Transportation (3)

Trade (4)

Services (5)

CR

Manufacturing (2)

49.12 (2.18)

60.44 (2.42)

49.26 (2.04)

44.10 (2.25)

37.54 (2.51)

1.17†† (0.67)

6.78 (0.76)

2.02† (0.81)

6205

9084

6375

49.57 (2.01)

MA

NU S

Mining and Construction (1)

IP

T

Table 6: Different Industries

36729

ED

This table displays the results from estimating the variance decomposition equations separately for each 1 digit SIC industry: mining and construction, manufacturing, transportation, trade, and services. The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973–2013. The numbers displayed represent the

PT

percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks unabsorbed. βD is the slope coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI is the slope coefficient from a regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I.

CE

βP is the slope coefficient from a regression of ∆ log P on ∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses. Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied by 100. All coefficients are significant at the 1% level,

AC

except for those labeled with †† and † which indicates they are significant at the 10% and 5% levels, respectively.

33

ACCEPTED MANUSCRIPT

T

Table 7: Firm Size Medium (2)

Debt (βD )

58.25 (2.80)

55.02 (1.14)

57.61 (1.00)

Investment (βI )

36.88 (2.82)

41.98 (1.16)

40.20 (1.03)

Payout (βP )

4.87 (0.81)

3.00 (0.35)

2.19 (0.35)

Observations

6106

28432

28943

MA

NU S

CR

IP

Small (1)

Large (3)

This table displays the results from estimating the variance decomposition equations separately for small, medium,

ED

and large firms. The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973– 2013. The numbers displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks unabsorbed. βD is the slope

PT

coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI is the slope coefficient from a regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I. βP is the slope coefficient from a regression of ∆ log P on ∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses.

CE

Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied by 100.

AC

All coefficients are significant at the 1% level.

34

ACCEPTED MANUSCRIPT

T

Table 8: Young Vs. Mature Firms

Investment (βI )

33.45 (2.44)

CR

64.38 (2.40)

NU S

Debt (βD )

IP

Young (1)

Payout (βP )

55.88 (0.74) 41.72 (0.76)

2.17 (0.70)

2.41 (0.25)

6606

56872

MA

Observations

Mature (2)

This table displays the results from estimating the variance decomposition equations separately for young and mature

ED

firms. Based on a firm’s Inital Public Offering (IPO), firms are characterized as young if they have been publicly traded for 10 years or less; otherwise, firms are characterized as mature. The sample consists of 6,225 firms taken from the Compustat database and covers the period 1973–2013. The numbers displayed represent the percentage of shocks

PT

to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks unabsorbed. βD is the slope coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI is the slope coefficient from a regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I. βP is the slope

CE

coefficient from a regression of ∆ log P on ∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses. Year and firm fixed effects are included in the regressions. Coefficients

AC

and standard errors are multiplied by 100. All coefficients are significant at the 1% level.

35

ACCEPTED MANUSCRIPT

T

Table 9: External Finance: Independent Firms Vs. Dependent Firms

CR

IP

Independent (1)

Dependent (2)

61.19 (1.47)

Investment (βI )

36.46 (1.51)

Payout (βP )

2.35 (0.52)

1.90 (0.38)

Observations

13538

23155

MA

NU S

Debt (βD )

61.25 (1.22) 36.85 (1.25)

This table displays the results from estimating the variance decomposition equations separately for firms in the

ED

sectors that are independent of external financing, and firms in the sectors that are dependent on external financing. To split the sample into these two groups, we follow the methodology in Rajan and Zingales (1998). The sample consists of 3,209 manufacturing firms taken from the Compustat database and covers the period 1973–2013. The

PT

numbers displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout (βP ) represents the fraction of shocks unabsorbed. βD is the slope coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI is the slope coefficient from a regression of

CE

∆ log(N I +∆D)−∆ log P on ∆ log N I. βP is the slope coefficient from a regression of ∆ log P on ∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors appear in parentheses. Year and firm fixed effects are included in the regressions. Coefficients and standard errors are multiplied by 100. All coefficients

AC

are significant at the 1% level.

36

37

42.86 (0.69) 0.53 (0.06) 66272

Investment (βI )

Payout (βP )

Observations

CE

100935

0.87 (0.15)

41.36 (0.51)

57.76 (0.52)

PT NU S

191262

1.64 (0.14)

40.90 (0.39)

57.47 (0.38)

188275

1.55 (0.13)

41.16 (0.39)

57.29 (0.38)

Dividends + Net Repurchases (5)

CR

Full Sample Net Repurchases Only (4)

MA

203179

0.18 (0.05)

ED

42.13 (0.37)

57.69 (0.37)

Dividends Only (3)

are included in the regressions. Coefficients and standard errors are multiplied by 100. All coefficients are significant at the 1% level.

∆ log N I. Variables are deflated using the GDP Price Deflator (2009=100). Robust standard errors in parenthesis. Year and firm fixed effects

is the slope coefficient from a regression of ∆ log(N I + ∆D) − ∆ log P on ∆ log N I. βP is the slope coefficient from a regression of ∆ log P on

(βP ) represents the fraction of shocks unabsorbed. βD is the slope coefficient from a regression of ∆ log N I − ∆ log(N I + ∆D) on ∆ log N I. βI

T

displayed represent the percentage of shocks to firm net income absorbed via the channels debt (βD ) and investment (βI ). The channel payout

issue dividends or not. The final samples consist of firms taken from the Compustat database and cover the period 1973–2013. The numbers

IP

issue or buy back corporate stocks over the sample period. In Columns (3) to (5), the samples include all firms, regardless of whether they

(1), the sample is restricted to firms that issue only dividends; in Column (2), the sample is restricted to firms that do not pay dividends, but

This table displays the results from estimating the variance decomposition equations separately for each different measure of payouts. In Column

56.61 (0.69)

AC

Restricted Sample Dividends Net Repurchases Only Only (1) (2)

Debt (βD )

Composition of Payout:

Table 10: Different Payout Measures

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT Highlights

CE

PT

ED

MA

NU S

CR

IP

T

Firms smooth their payouts Payout smoothing increases with the firm size Payout smoothing has increased over time Persistent income shocks reduce payout smoothing

AC

1. 2. 3. 4.