The cash flow/investment relationship: evidence from U.S. manufacturing firms

The cash flow/investment relationship: evidence from U.S. manufacturing firms – includes appendix

Stephen C. Vogt

Substantial empirical evidence documents the strong influence of cash flow on some firms’ investment spending.(1) This well-documented relationship between cash flow and investment spending (after controlling for the cost of capital) is inconsistent with both the Modigliani and Miller (1958) irrelevance theorem and the so-called static trade-off theories of financial behavior.(2)

Two recent explanations focus on imperfect information. The pecking order (PO) hypothesis of Myers and Majluf (1984) identifies the adverse selection problem that arises when firm insiders (owners and managers) have better information than the capital markets about the value of their firm. An important implication of adverse selection is that firms with positive-net-present-value (NPV) investment opportunities will forgo profitable projects to avoid the excessive cost of external financing. This implication has been explored in detail by Fazzari, Hubbard, and Petersen (1988) for capital spending (i.e., fixed plant and equipment) and by Himmelberg and Petersen (1994) for research and development spending. These authors show formally that the excess cost of external finance causes some firms to be liquidity-constrained, so that cash flow becomes an important determinant of investment spending.

The second explanation, the free cash flow (FCF) hypothesis (Jensen (1986)), focuses on the agency issue. Jensen argues that managers can increase their wealth at the expense of shareholders by investing a firm’s free cash flow in unprofitable investment opportunities rather than paying out those funds in the form of dividends, debt-financed share repurchases, and the like. Carpenter (1993), Devereux and Schiantarelli (1990), Oliner and Rudebusch (1992), and Strong and Meyer (1990) study the role that agency problems play in the cash flow/investment relationship. Their findings are contradictory regarding the importance of free cash flow. Oliner and Rudebusch find little evidence that ownership structure affects the cash flow/investment relationship. Strong and Meyer find that stock prices of firms undertaking investment spending with discretionary cash flow experience negative performance.

The aim of this research is to examine whether the importance of cash flow in the firm’s investment decision is because firms waste free cash flow, or because they face excessive costs of external financing created by asymmetric information. First I develop the theoretical implications of both the FCF and PO explanations for the equilibrium level of Tobin’s Q. If the free cash flow theory explains the cash flow/investment relationship, firms with low Q values should rely heavily on cash flow to finance investment. Alternatively, if the PO hypothesis explains the relationship, firms with high Q values will depend more heavily on cash flow. A model of firm investment spending using cross-sectional time-series data with fixed time and firm effects is then used to test the relative importance of the FCF and PO hypotheses for both plant and equipment spending and research and development spending.

The firm’s dividend decision also has implications for both theories. In the FCF theory, dividends are one means of eliminating free cash flow (Lang and Litzenberger (1989)). The model developed here shows that firms with the opportunity to exploit free cash flow will follow low-dividend-payout policies and have a low value of Q, and cash flow will have a strong influence on investment spending. Conversely, if firms are constrained from obtaining external finance because of adverse selection problems (as in the PO theory), those firms with profitable investment opportunities will maintain low-dividend-payout policies in order to conserve on cash flow. In this case, the model is consistent with Fazzari, Hubbard, and Petersen (1988); it predicts that low-payout firms should be associated with high values of Q and a strong cash flow/investment relationship.

Finally, I perform additional tests using asset size as a proxy for both asymmetric information and free cash flow problems. Larger firms with more diverse ownership structures are more likely to suffer agency problems. Consequently, the cash flow/investment relationship should be stronger in large firms with low Q values. Alternatively, asymmetric information-induced liquidity constraints are more likely to be found in smaller firms. Thus, the cash flow/investment relationship should be stronger for small firms with high Q values.

In the case of capital spending, the empirical results tend to support the free cash flow description of the cash flow/investment relationship. Behavior that supports the PO hypothesis, however, is found in small firms paying low dividends. In the case of research and development spending, results are more consistent with the pecking order hypothesis. These results together suggest that the effect that cash flow-financed investment has on firm value depends on asset size, dividend behavior, and the type of investment spending.

I. Background on the Cash Flow/Investment Relationship

Considerable empirical evidence indicates that internally generated funds are the primary way firms finance investment expenditures. In an in-depth study of 25 large firms, Gordon Donaldson (1961, p. 67) concludes that: “Management strongly favored internal generation as a source of new funds even to the exclusion of external funds except for occasional unavoidable ‘bulges’ in the need for new funds.” A more recent survey of 176 corporate managers by Pinegar and Wilbricht (1989) also finds that managers prefer cash flow over external sources to finance new investment; 84.3% of sample respondents indicate a preference for financing investment with cash flow.

Researchers have also discovered the impact of cash flow on investment spending in Q models of investment. Fazzari, Hubbard, and Petersen (1988) find that cash flow has a strong effect on investment spending in firms with low-dividend-payout policies.(3) They argue that this result is consistent with the notion that low-payout firms are cash flow-constrained because of asymmetric information costs associated with external financing. One reason these firms keep dividends to a minimum is to conserve on cash flow from which they can finance profitable investment expenditures. More recently, Fazzari and Petersen (1993) find that this same group of low-payout firms smooths fluctuations in cash flow with working capital to maintain desired investment levels. This result is consistent with the Myers and Majluf (1984) finding that liquid financial assets (slack) can mitigate the underinvestment problem arising from asymmetric information.

Whited (1992) extends the Fazzari, Hubbard, and Petersen (1988) results in a study of firms facing debt financing constraints because of financial distress. She finds evidence of a strong relationship between cash flow and investment spending for firms with a high debt ratio or a high interest coverage ratio, or without rated debt.

Finally, Himmelberg and Petersen (1994) in a study of small research and development firms find that cash flow strongly influences both capital and R & D expenditures. They argue that the asymmetric information effects associated with such firms make external financing prohibitively expensive, forcing them to fund expenditures internally.(4)

An alternative explanation for the strong cash flow/investment relationship is that managers divert free cash flow to unprofitable investment spending. One study assessing the relative importance of such an agency problem is by Oliner and Rudebusch (1992), who analyze several firm attributes that may influence the cash flow/investment relationship. They find that insider share holdings and ownership structure (variables that proxy for agency problems) do little to explain the influence that cash flow has on finn investment spending; firm age, exchange listing, and insider stock trading patterns exhibit a moderately stronger influence. They conclude that weak support exists for the asymmetric information explanation. Data limitations, however, may restrict the generality of these conclusions about ownership structure.

Carpenter (1993) focuses on the relationships among debt financing, debt structure, and investment spending to test the free cash flow theory. He finds that finns that restructure by replacing large amounts of external equity with debt increase their investment spending compared to non-restructured firms. He sees these results as inconsistent with free cash flow behavior, because cash flow committed to debt maintenance should be associated with reductions in subsequent investment spending.

Findings by Strong and Meyer (1990) and Devereux and Schiantarelli (1990) support the free cash flow interpretation. Strong and Meyer (1990) disaggregate the investment and cash flow of firms in the paper industry into sustaining investment (i.e., productive capacity maintaining) and discretionary investment, and total cash flow and residual cash flow (i.e., cash flow after debt service, taxes, sustaining investment, and established dividends). Residual cash flow and discretionary investment are found to be positively and strongly related. Discretionary investment and stock price performance are negatively and strongly correlated. This evidence suggests that residual cash flow is often used to fund unprofitable discretionary investment spending.

Devereux and Schiantarelli (1990) find that the impact of cash flow on investment spending is greater for larger firms. One explanation they provide for this result is that large firms have more diverse ownership structures, and are more influenced by manager/shareholder agency problems.

Finally, in his presidential address to the American Finance Association, Jensen (1993) presents evidence that even though most of the 452 largest COMPUSTAT firms between 1979 and 1990 created more in market value than the compounded value of their dividends, stock repurchases, and capital and R & D expenditures, as many as 25% of the firms did not. He attributes these results to the failure of internal control mechanisms to eliminate value-destroying investment spending associated with free cash flow.

To shed additional light on the cash flow/investment issue, I formally model the relationships among cash flow, dividend behavior, and investment spending to derive testable empirical hypotheses that distinguish between the asymmetric information and free cash flow interpretations.

II. Implications of the Free Cash Flow and Pecking Order Hypotheses

When agency problems associated with free cash flow are the driving influence behind the cash flow/investment relationship, firms will exhibit Tobin’s Q values below unity in equilibrium. This negative association between Q and the effect of cash flow on investment will be stronger for firms that do not shed free cash flow by committing to high dividend payouts. If financing constraints are the source of the cash flow/investment relationship, the equilibrium level of Q will exceed unity; cash flow-constrained firms will exhibit a strong positive relationship between Q and the influence of cash flow on investment.

A. The Relationship Between Free Cash Flow and Tobin’s Q

This section shows the effect that free cash flow has on a firm’s Q value when monitoring management decisions is costly. The model is then expanded to show how dividends (a means of eliminating free cash flow) affect Tobin’s Q.

1. A Simple Q Model with Costly Monitoring

Jensen (1986, p. 323) defines free cash flow as “cash flow in excess of that required to fund all projects that have positive net present value when discounted at the relevant cost of capital.” When free cash flow is present and shareholder monitoring is incomplete, the typical manager-shareholder agency problem arises. Managers have a tendency to overinvest (i.e., invest in negative-NPV projects) in order to capture the pecuniary and non-pecuniary benefits of increased firm size (Jensen and Meckling (1976)), while shareholders would prefer dividends over retention to eliminate free cash flow (Lang and Litzenberger (1989)). The implication of free cash flow for a firm’s equilibrium level of marginal Q can be determined using a simple valuation model. More complicated models provide little additional insight.

I begin by assuming that managers gain utility from expanding the asset base of the firm, whatever the expected profitability of the investment.(5) I further assume that shareholders can extinguish the agency costs associated with free cash flow provided they invest M dollars to monitor manager activity. Expending M dollars on monitoring implies that the value of the firm is

[V.sub.m] = X + [X[prime].sub.m] – M (1)

where [V.sub.m] is the total market value of the firm to existing shareholders, X is the market value of assets in place, and [X[prime].sub.m] is the market value of the firm’s investment opportunity when managerial activities are monitored. The expenditure of M insures that no negative-NPV investments are undertaken, and thus [X[prime].sub.m] [is greater than or equal to] I, where I is the initial cost of the investment.

Conversely, if shareholders do not invest M in monitoring, the market value of the firm is

V = X + X[prime] (2)

where X[prime] is the market value of the firm’s investment opportunity when no monitoring takes place. In this case, X[prime] in Equation (2) may be less than the initial cost of investment. If managers benefit from overinvesting, they will overinvest up to the point where V = [V.sub.m] + [Epsilon], where [Epsilon] is a small positive number. Given the assumptions on the cost of monitoring and the manager’s utility, this investing strategy allows managers to maximize their utility from overinvesting and still prevents shareholder monitoring from being economical. From Equations (1) and (2), managers will overinvest as long as

X[prime] [is greater than] [X[prime].sub.m] – M + [Epsilon] (3)

Dividing Equation (3) by the cost of the investment project (I), and noting that in perfect markets investment expands until the value of marginal Q is unity (i.e., [X[prime].sub.m],/I = 1) yields a relationship for the firm’s marginal Q:

X[prime]/I [is approximately equal to] 1 – M/I [is less than] 1 (4)

Equation (4) shows that agency problems associated with free cash flow and costly monitoring cause the equilibrium value for marginal Q to fall below unity by an amount approximately equal to the monitoring cost per dollar of investment spending. Furthermore, if X[prime]/I [is greater than] (1 – M/I), and free cash flow is available, managers find it in their interest to expand investment. Consequently, cash flow should explain investment spending for these firms and should also be associated with low values of marginal Q.

2. A Q Model with Dividends and Costly External Financing

Paying dividends as a means of committing free cash flow to higher-valued uses will have an effect on a manager’s ability to overinvest free cash flow, and therefore the equilibrium value of marginal Q. The standard NPVGO/constant dividend growth valuation model allowing retentions and external financing to fund growth is:

[Mathematical Expression Omitted]

where V is the value of the firm, [E.sub.1] is expected cash flow next period, k is the cost of capital, r is the return on new investment, b is the cash flow retention rate, and s is the firm’s external financing rate per dollar of cash flow.

Noting that I = [E.sub.1](b + s), the gross market value for a single investment opportunity using this model is:

X[prime] = r(b + s)[E.sub.1]/k = r/k [center dot] I (6)

The corresponding value of marginal Q is,

X[prime]/I = r/k (7)

Profit maximization implies that investment continues until the return per dollar of investment equals the cost of capital (i.e., r = k) or, equivalently, until marginal Q is unity.

One way shareholders can limit overinvestment by managers is to require a high dividend payouts, thereby forcing new investments to be financed externally where the capital market can directly assess (monitor) their value. This is a costly strategy to existing shareholders because new claimholders to the firm (creditors or shareholders) must undertake costly information gathering and screen firms before committing funds.

One function of investment banking consistent with Smith (1986) is its role in processing information about a firm’s prospects–a task for which it must be compensated. This screening function is relevant also during the acquisition of private debt capital (Campbell and Kracaw (1980) and Diamond (1984)). Though screening costs in connection with the acquisition of debt capital are considerably lower than for external equity finance, moral hazard and adverse selection raise the possibility that the firm can be credit-rationed when it holds a profitable investment opportunity.

A second cost associated with external financing is the dilution of existing shareholder value that occurs if the stock price falls when new securities are issued. While there are a number of theoretical reasons why share prices may fall upon announcement of new issues, one is that the issue signals poorer firm prospects for the future.

The basic valuation model with retentions and external financing may be adapted to account for the agency problems associated with free cash flow and costly monitoring. The costs associated with external financing are defined as a fixed flotation cost (M*) for the new security issue. The discrete variable,

[d.sub.m] = {1 if s [is greater than] 0

{0 otherwise is defined so that the shareholders expend M* only if they force external financing. As in Section II.A.1., expending M* assures that the manager selects only positive-net-present-value investments, and all benefits of free cash flow to the manager are forgone because negative-NPV investments have been screened out in the process of acquiring external capital.

This model can also accommodate overinvestment costs associated with free cash flow. I define the per dollar agency cost associated with retaining free cash flow as [Mu], where [Mu] may be a function of the level of investment spending. Every dollar of free cash flow retained in the firm is reinvested in a project returning r(1 – [Mu]). In this sense, the agency cost associated with free cash flow acts as a tax on the returns to new investment. The associated gross market value, X[prime], of the investment project is thus,

X[prime] = r(b(1 – (1 – [d.sub.m])[Mu]) + [d.sub.m]s)[E.sub.1]/k – [d.sub.m]M* (8)

From Equation (8), the fact that (b+s)[E.sub.1] = I, and the assumption that sufficient free cash flow is available with which to overinvest, managers will retain earnings to finance overinvestment to the point where

X[prime] = r/k(I – b [Mu] [E.sub.1]) + r/k I – M* + [Epsilon] (9)

This implies an equilibrium value of marginal Q equal to

X[prime]/I = 1 – M* – [Epsilon]/(b + s)[E.sub.1] [is approximately equal to] 1 – M*/I (10)

when r = k. Equation (10) implies a maximum retention rate (b* = (M* – [Epsilon])/([Mu][E.sub.1])) that will maximize the manager’s utility from overinvesting and still deter shareholder monitoring.

Three points of this analysis deserve mention. First, when free cash flow is sufficient so that the maximum retention rate, b*, is achieved, managers must pay out the remaining cash flow rather than reinvest it in order to avoid monitoring by the external capital market. Managers will have exhausted all opportunities to exploit free cash flow and must pay dividends. Furthermore, cash flow will no longer affect investment spending, as additional investment from free cash flow will force external financing.

Second, to the extent that free cash flow is insufficient for the firm to attain the maximum retention rate (i.e., the firm can use all of its free cash flow to finance investment), no dividends will be paid. Any changes in cash flow will then be used to finance additional (negative-NPV) investment.

Finally, when the firm pays no dividends and monitoring does not occur in the external capital market, the associated value of marginal Q is,

X[prime]/I = r/k(1 – b[Mu]/I [[Epsilon].sub.1]) (11)

which is a negative function of [E.sub.1]. This implies that increases in earnings for firms not paying dividends will be associated with declining values of marginal Q.(6)

This model of agency costs of free cash flow and costly monitoring underlies several empirical hypotheses:

1. Since monitoring is assumed costly, and managers can benefit from overinvestment, cash flow will significantly influence investment spending after controlling for the cost of capital.

2. Investment spending of finns not paying dividends will be more influenced by cash flow than investment spending of firms that pay dividends. This follows because no-dividend firms are able to retain all cash flow and still not reach the retention constraint, b*.

3. Firms in which cash flow significantly influences investment spending will be associated with low values of marginal Q. In fact, the equilibrium level of Q for these finns is less than one. This follows directly from Equations (4) and (10).

4. Finally, firms paying no dividends will not only have investment expenditures that are strongly influenced by cash flow, but also will be associated with the lowest levels of marginal Q. This follows directly from Equation (11).

The relationships among cash flow, investment spending, and dividend behavior stated in points (1) and (2) have been discussed in the literature. The results have been interpreted, however, as consistent with the pecking order hypothesis only. My discussion in fact indicates that these relationships are also consistent with the FCF hypothesis. Points (3) and (4) are the reverse of those predicted by the PO hypothesis, however, and are at the root of the test distinguishing between the two theories.

B. The Pecking Order Hypothesis

The pecking order hypothesis focuses on the adverse selection problem that arises because managers (assumed to act in the best interest of existing shareholders) have better information than is known to the capital markets. Internal funds are important because they circumvent the adverse selection problem.

In seeking external funds, managers cannot credibly convey the quality of their firm’s existing assets and available investment opportunities to the market. In the resulting equilibrium, undervalued firms may forgo positive-NPV investments in order not to transfer value from existing owners to new investors. The information cost in Myers and Majluf’s (1984) model is the value of profitable investment opportunities forgone. One implication of this underinvestment problem is that cash flow-constrained firms will have equilibrium values of marginal Q that exceed unity.

To show that asymmetric information implies that marginal Q will exceed unity, I develop a model adapted from Myers and Majluf (1984) and Fazzari, Hubbard, and Petersen (1988). Assume that the economy consists of two types of firms wishing to raise I dollars by issuing new equity: undervalued firms (u) and overvalued firms (o). Firm insiders (managers) know with certainty whether the firm is under- or overvalued. Outside investors know only the proportion (p) of undervalued finns in the economy, but are not able to distinguish between firms.

The value of the firms in a pooling equilibrium is

V = p([X.sub.u] + [X[prime].sub.u]) + (1 – p)([X.sub.o] + [X[prime].sub.o]) (12)

and ([X.sub.u] + [X[prime].sub.u]) [is greater than] V [is greater than] ([X.sub.o] + [X[prime].sub.o]). The variables [X.sub.(.)] and [X[prime].sub.(.)] are defined as the true (full information) value of the firm’s existing assets and investment opportunities, respectively. The subscript identifies the type of firm.

Insiders of overvalued firms would always like to issue and invest because this would transfer value from new investors to themselves. Yet to avoid detection in the market, these insiders mimic the actions of undervalued firms. Rational insiders of undervalued firms will issue and invest if their claim to the total value of the firm exceeds the current value of assets in place.

This occurs when

V/V + I ([X.sub.u] + [X[prime].sub.u]) [is greater than or equal to] [X.sub.u] (13)

Algebraic manipulation of Equation (13) yields

[X[prime].sub.u]/I [is greater than or equal to] [X.sub.u]/V (14)

When total firm value is large relative to the cost of investment (i.e., V is large relative to I), and when Equation (13) holds with equality, the right-hand side of Inequality (14) becomes approximately (V + [Omega]) / V. Here [V.sub.u] = V + [Omega] is the true (full information) value of the undervalued firm, and [Omega] represents the difference between the firm’s true value and its market value.

Under symmetric information, [V.sub.u] = V, [Omega] = 0, firm insiders are indifferent between internal and external finance, and investment continues until the equilibrium level of marginal Q in Equation (14) is unity. Furthermore, firm insiders are value-maximizers, so even if adverse selection pertains (i.e., [Omega] [is greater than] 0), firms with sufficient cash flow will avoid the external capital market and fund investment internally. Therefore, for cash-rich firms, equilibrium Q is also unity, excess cash flow is paid out in dividends, and changes in cash flow will not influence investment, for a surplus already exists.

If firms experience asymmetric information effects ([Omega] [is greater than] 0), and they have limited cash flow available to finance profitable investment opportunities, the equilibrium level of marginal Q becomes

[X[prime].sub.u]/I = 1 + [Omega]/V (15)

Cash-constrained firms will use external finance and expand investment as long as the expected value exceeds the cost of investment plus the premium associated with external finance (i.e., [X[prime].sub.u]/I [is greater than or equal to] 1 + ([Omega]/V)).

When the availability of internal finance is limited, firms with the greatest need to fund profitable investments or those suffering most from information problems will be the most liquidity-constrained and have the largest values of marginal Q.

Four empirical hypotheses result from our analysis of the PO hypothesis:

1. For liquidity-constrained firms, cash flow and changes in the stock of the firm’s liquid assets should have a significant influence on investment spending.

2. Firms with many profitable investment opportunities or large information asymmetries will have investment spending that is most sensitive to changes in cash flow, and should conserve on cash flow by paying low or no dividends.

3. Firms with investment spending that is influenced by cash flow will be associated with high Q values. In fact, the equilibrium level of Q for these firms will be larger than one.

4. Firms indicating a liquidity constraint by not paying dividends will have the most significant cash flow/investment relationship, and will be associated with high Q values in the market.

Again, hypotheses (1) and (2) are not new to the literature. The distinguishing features of the FCF and PO theories are again expressed in hypotheses (3) and (4).

III. The Source of the Cash Flow/Investment Relationship

I have shown that both the FCF and PO hypotheses indicate that cash flow should influence investment spending. Each theory predicts that firms not paying dividends should exhibit the strongest relationship between cash flow and investment spending, while those paying high dividends should show the weakest relationship. The major distinction between the two hypotheses is the predicted interaction between the firm’s Q value and the influence of cash flow in determining investment spending. To test the associations among dividend policy, cash flow, investment spending, and Q, an interaction variable is added to a typical reduced-form Q model of capital spending.(7)

The data used in this study come from 359 manufacturing firms for which the COMPUSTAT tapes have complete datafrom the years 1973 to 1990. Data for 1973 were used only for the construction of lags. Firms were screened for merger and acquisition activity over the period. Firms reporting acquisitions in excess of 10% of their total assets in any year were deleted from the sample. Firms were also screened for defaulted debt to remove the effect that financial distress may have on the firm’s financing decision.(8)

Table 1 presents summary statistics of the data both in the aggregate and when disaggregated by average-dividend-payout behavior. Low-dividend-payout firms have the largest levels of capital spending per dollar of beginning-of-period capital stock, which is consistent with both theories. The mean level of Q is close to unity for firms with low long-run dividend payout policies and is largest for high-dividend-payout firms. This result lends initial support for the FCF view of the cash flow/investment relationship.

A. The Cash Flow/Capital Spending Relationship

A reduced-form model of capital spending behavior similar to that of Fazzari and Petersen (1993) is estimated to establish the empirical significance of cash flow. The equation models the proportion of capital spending to the beginning-of-period stock of fixed plant and equipment (I/K) as a function of: (1) cash flow divided by beginning-of-period gross plant and equipment (CF/K), (2) the change in the firm’s cash divided by beginning-of-period gross plant and equipment (DCASH/K), (3) sales divided by beginning-of-period gross plant and equipment (SALES/K), and (4) beginning-of-period Tobin’s Q (Q). The precise COMPUSTAT data definitions for these variables appear in the Data Appendix.

Since DCASH/K is an endogenous variable, a two-stage least squares model is used first to estimate it as a function of the following instruments; (1) the beginning-of-period stock of cash, [C.sub.t-i]/K, (2) CF/K, (3) SALES/K, and (4) Q. Finally, a fixed-effects model is employed to allow for time-specific and firm-specific intercepts. Direct estimation of the intercepts is eliminated by centering the data around their time-series and cross-sectional means. See Deilman (1983) for a discussion of this technique.

The second stage of the estimation takes the form:

[Mathematical Expression Omitted]

TABULAR DATA OMITTED

where [Mathematical Expression Omitted] refers to the predicted value of DCASH/K resulting from the first-stage estimation, [[Mu].sub.i] and [[Tau].sub.t] are firm- and time-specific fixed effects, and i and t are indexes for firm and time, respectively. Lagged cash flow, [(CF,/K).sub.i,t-i] is included in the regressions to control for possible lagged effects identified by Hall (1992).(9)

The first column in Panel A of Table 2 reports the estimation results of Equation (16). Consistent with the earlier findings, CF/K and DCASH/K have a strong and highly significant impact on capital spending. The negative coefficient on DCASH/K indicates that firms draw down their stock of cash in order to finance additional fixed investment. The parameter estimates on SALES/K and Q also have the correct signs as predicted by the accelerator theory and the Q theory of investment.

Next, in order to determine whether the strength of the cash flow effect varies with dividend policy, firms are separated by their average-dividend-payout behavior. Low-payout (L) firms are defined as firms with an average dividend-to-income ratio of 0.10 or less over the sampleperiod. Medium-payout (M) firms have an average payout rate of less than 0.35 and more than 0.10. Finally, high-payout (H) firms are firms with payout rates greater than 0.35.(10)

Columns 2 through 4 in Panel A of Table 2 report the estimation results of Equation (16) by payout group.(11) The parameter estimates on DCASH/K and CF/K all have the proper sign and are statistically significant. As both the FCF and PO hypotheses would predict, the magnitude of the parameter estimates increases as the long-run payout rate falls. An F-test rejects (at the 0.01 level) the null hypothesis that parameter estimates on cash flow across groups are equal.

One way to determine what causes the cash flow/investment relationship is to analyze the prediction of TABULAR DATA OMITTED each hypothesis about the effect that cash flow-financed capital spending has on Q. To measure this effect, an interaction variable (CF/K x Q) is created and Equation (17) estimated:

[Mathematical Expression Omitted]

If the parameter estimate on (CF/K x Q) is positive, the results support the PO hypothesis. If the parameter estimate is negative, the results support the FCF hypothesis. The intuition behind this result can be seen by interpreting the parameter estimate on the interaction term as a cross-partial derivative. Since the relationship in Equation (17) is hypothesized to be linear, the parameter [[Beta].sub.5] is [[Beta].sub.5] = [Delta]I/[Delta]CF[Delta]Q = [Delta][[Beta].sub.cf]/[Delta]Q, where [[Beta].sub.cf] is the coefficient on cash flow in Equation (16). If [Delta][[Beta].sub.cf/k]/[Delta]Q is greater than (less than) zero, the PO (FCF) hypothesis is supported.

The first column in Panel A of Table 3 reports the estimation results for Equation (17). While not large, the parameter estimate on [[Beta].sub.5] is -0.0217 and significant. This result indicates that the influence of cash flow on capital spending rises as Q falls, thus supporting the FCF hypothesis.

The estimation results by payout group are presented in columns 2 through 4 in Panel A. The sign of the parameter estimates on the interaction term vary between groups. Capital spending of low-payout firms is negatively and strongly influenced by the interaction term ([[Beta].sub.5,L] = -0.0848), which is consistent with the FCF hypothesis. Medium-payout firms have a positive but insignificant parameter estimate on the interaction term ([[Beta].sub.5,M] = 0.0108), while it is negative for high-payout firms ([[Beta].sub.5,M] = -0.0547) and only marginally significant.

These results suggest that the negative relationship between the magnitude of the CF/I relationship and Q found in the aggregate data is concentrated in firms paying out low or no dividends. This is further evidence consistent with the FCF hypothesis.(12)

The robustness of the two-stage least squares estimates and the quality of the continuous interaction variable (CF/K x Q) are checked by performing a dummy variable interaction regression using ordinary least squares (OLS). This alternative specification is estimated to eliminate possible biases associated with extreme values of Q, which may influence both the sign and the significance of the parameter estimate on the continuous interaction variable. Even though the Q values used in this study are more stable than in previous studies (such as Fazzari, Hubbard, and Petersen (1988)), the maximum Q in the sample is 9.63, while the minimum is 0.28. Such extreme values may affect the regression estimates.

Four dummy variables, QDUM1-QDUM4, corresponding roughly to the four quartiles of the distribution of Q are created. QDUM1 is set to one when [Q.sub.i,t] is below 0.68, and zero otherwise. Similarly, QDUM2-QDUM4 are each set to one when the value of [Q.sub.i,t] falls between 0.68 and 1.0, 1.0 and 1.24, and more than 1.24, respectively. Dummy interaction variables between cash flow and QDUM1-QDUM4 are then created, and after eliminating the endogenous variable DCASH/K the model is estimated using OLS.

Results of the dummy-interaction estimation are reported in Panel B of Table 3. The results are quite similar to those found using the continuous interaction term (CF/K x Q). Aggregate data results presented in the first column show that the firms in the lowest Q-quartile exhibit the strongest association between cash flow and capital spending ([[Beta].sub.CF/K x QDUM1] = 0.1521), and the highest Q-quartile firms have the lowest association ([[Beta].sub.CF/K x QDUM4] = 0.0185). When firms are split by average-dividend-payout behavior, the influence of cash flow on capital spending is seen exclusively in the medium- and low-payout firms.

The largest parameter occurs in the low-Q, low-payout firms ([[Beta].sub.CF/K x QDUM1] = 0.2379) thus adding further support to the FCF hypothesis. However, cash flow also appears to be a factor in influencing high-Q, low-payout firms ([[Beta].sub.CF/K x QDUM4] = 0.1216), which is consistent with the PO hypothesis. Consequently, there is some evidence that both FCF and PO behavior is present in firm financing decisions, although FCF behavior dominates in the aggregate.

A final method of distinguishing between FCF and PO behavior looks at firm size. Larger size is often positively associated with agency problems and free cash flow behavior. Indeed, Devereux and Schiantarelli (1990) argue that the typically more diverse ownership structure of large firms explains their finding that cash flow impacts investment spending more for larger firms than smaller firms. These findings conflict with the idea that small firms suffer from greater asymmetric information problems and are more likely to show a stronger relationship between cash flow and investment.

TABULAR DATA OMITTED

Table 4 presents results from regressions on Equation (17) when firms are split by asset size. Firms are grouped into five asset categories to avoid problems associated with skewness in the distribution of firm size. Group 1 firms are firms with average asset size greater than $1.5 billion. Groups 2, 3, 4, and 5 firms have average asset sizes between $300 million and $1.5 billion, $100 and $300 million, $40 and $100 million, and 0 and $40 million, respectively.

Results show that cash flow strongly influences capital spending for large and small firms, but has less of an impact for medium-sized firms. This result also suggests that both free cash flow and pecking order behavior are potentially important sources of the cash flow/investment relationship and that firm size influences which effect dominates.

Panels A-C of Table 5 report estimation results for firms grouped by asset size and dividend-payout behavior. Panel A reports results for low-dividend (group 1) firms broken TABULAR DATA OMITTED down by asset size. Panels B and C report results for medium- (group 2) and high-payout (group 3) firms, respectively. For brevity, only the parameter estimates for the change in cash (DCASH/K), cash flow (CF/K), and the cash flow/Q interaction term (CF/K x Q) are reported. The low-payout firms, regardless of size, have uniformly larger coefficients on cash flow than do medium- and high-payout firms. The one exception is the smallest firm in the medium-payout group, which has a parameter estimate of 0.2761 versus 0.1819 and 0.0151 for low- and high-payout firms of comparable size.

The most striking evidence regarding the source of the investment/cash flow relationship is shown in the low-payout group (Panel A), where the relationship is strongest. In this group, the parameter estimate on cash flow is largest for firms with average asset sizes of over $1.5 billion and between $300 million and $1.5 billion. A TABULAR DATA OMITTED significant source of the strong relationship between capital spending and cash flow is thus found in large firms paying low dividends over the sample period. This evidence conforms most closely with the free cash flow interpretation, shedding additional light on the Strong and Meyer (1990) finding that agency problems are important.

Cash flow, however, is still an important variable in the capital spending behavior of small, low-payout firms. The parameter estimate of 0.1819, while considerably less than those associated with the larger firms in the low-payout group, is still highly significant. Consequently, the asymmetric information-induced pecking order explanation cannot be dismissed. The most plausible argument is that both free cash flow and asymmetric information are important factors contributing to the influence of cash flow on capital spending.

This interpretation is further supported by considering the parameter estimates on the interaction term (CF/K x Q). With few exceptions, the interaction term is not significant in the medium- and high-payout groups, regardless of asset size. It does enter significantly in four of five asset groups in the low-payout firms, however. Within this low-payout group, the parameter estimates on the interaction term are negative and significant for the three firm groups associated with the largest firm sizes, and positive and significant for the firms constituting the smallest asset group.

These results support the findings that both agency problems and asymmetric information may influence firm behavior. The results are also consistent with the hypotheses that large firms are most likely to suffer agency problems, while small firms are likely to suffer information problems.

B. The Cash Flow/R & D Relationship

An alternative, although less tangible, form of investment spending is expenditures on research and development.(13) Two intriguing issues arise when considering this form of investment. First, its intangible and high risk/return nature may make research and development more susceptible to asymmetric information effects than fixed plant and equipment spending. Himmelberg and Petersen’s (1994) study of the R & D spending of small firms in high-tech industries indicates that cash flow is an extremely important source of their financing likely because of liquidity constraints created by asymmetric information.

A plausible alternative explanation, however, is that the intangible nature of R & D spending makes it difficult to monitor, which raises the potential for overinvestment when cash flow is available. Jensen (1993) presents evidence that this may be the case for the largest firms in the U.S., while Chauvin and Hirschey (1993) find that R & D spending creates large and positive increases in market value.

A second issue is the different incentives that R & D and capital spending may generate for managers over time. Research and development represents an expenditure on intangible assets whose impact on the asset size and future cash flows of the firm is (1) extremely uncertain and (2) not likely to be realized in the near future. Fixed plant and equipment spending is likely to produce more certain cash flows in the near future (in part because of accelerated depreciation allowances) as well as increase the tangible asset base of the firm. The effect of plant and equipment spending is to generate free cash flow that can be used in the next period. Consequently, capital spending may be more susceptible to free cash flow problems than research and development spending.

To test the proposition that R & D spending is likely to be more closely related to PO behavior than FCF behavior, I repeat the analysis applied first to capital spending. Because reporting of R & D spending is limited in the early years of the sample, the number of firms with complete data on research and development spending falls from 359 to 229. Since R & D spending is embodied in the definition of cash flow, lagged cash flow is used in place of contemporaneous cash flow to avoid any simultaneity problems.

Results from the regressions on the research and development variable RD/K are reported in Table 6. Interestingly, the results on R & D support the PO hypothesis more than the FCF hypothesis. Cash flow has a very strong influence on R & D spending, and its influence tends to increase in lower-payout groups. The parameter estimate on the continuous interaction term [[(CF/K).sub.i,t-1] x [Q.sub.i,t-1]] is positive in the aggregate sample. The significance of the interaction term reported in Panel A declines appreciably, however, when firms are split into high-, medium-, and low-payout groups, and is marginally negative in the lowest-payout group.

A clearer picture of the influence that cash flow has on R & D is obtained from the results of an OLS regression using dummy variable interactions. As in the case of capital spending, dummy variables QDUM1-QDUM4 are created to define the lowest- to highest-Q quartiles.(14) Lagged cash flow is then multiplied by these dummy variables and included in place of the continuous interaction dummy.

Results of an OLS regression without DCASH/K are reported in Panel B of Table 6. The parameter estimates on the interaction terms for the aggregate data (first column) show that the influence of cash flow on R & D spending rises as the value of Q rises, indicating strong support for the PO hypothesis. This relationship appears to be most extreme in the low-payout group, where the parameter estimates rise from -0.1424 to 0.3404 as Q moves from its lowest to highest quartile. This indicates that among the firms most likely to be liquidity-constrained, the more cash flow is used to finance research and development spending, the higher the value of Q.

IV. Conclusion

The aim of this analysis has been to test the cause of the well-documented relationship between cash flow and investment spending. Two hypotheses about the source of this relationship are considered: the free cash flow (FCF) hypothesis, which assumes managers overinvest free cash flow in unprofitable investment projects, and the pecking order (PO) hypothesis, which suggests that managers underinvest because of an asymmetric information-induced liquidity constraint.

Results from several empirical specifications indicate that the influence of cash flow on capital spending is stronger for firms with lower Q values. This result suggests that cash flow-financed capital spending is marginally inefficient and provides initial evidence in support of the FCF hypothesis. TABULAR DATA OMITTED The negative relationship found in the aggregate data is concentrated in firms paying low dividends over the sample period, in large firms, and most strongly in large firms paying low dividends.

At the same time, the asymmetric information-induced pecking order explanation advanced by Fazzari, Hubbard, and Petersen (1988) and others cannot be dismissed. Small firms that paid low dividends over the sample period relied heavily on cash flow and changes in cash to fund capital spending. The stronger the influence cash flow had on capital spending in this group, the larger the associated value of Tobin’s Q. Consequently, large, low-dividend firms exhibit free cash flow behavior, while small, low-dividend firms exhibit pecking order behavior.

The influence of dividend policy and firm size sheds some light on the finding that financially distressed firms rely primarily on cash flow because they are denied access to the debt markets (Whited (1992)). Although the financial distress argument may explain the strong influence cash flow has on capital spending for low-Q firms found here, it cannot explain why only large firms exhibit this behavior and small firms do not. Thus, the interaction between dividend-payout behavior and firm size appears to distinguish free cash flow and pecking order behavior rather than support the financial distress view.

These results may have important implications for both investors and managers. While the study shows that cash flow-financed capital spending is marginally unproductive for this sample of firms, the potential sources of this inefficiency have also been identified. Cash flow-financed growth by large, low-dividend firms tends to be value-destroying, while cash flow-financed growth is value-creating for small, low-dividend firms. The importance of dividends as a method of mitigating agency costs of free cash flow, moreover, is confirmed. Managers of cash flow-rich companies may consider increasing dividend payouts as a method of increasing the efficiency of their capital spending decisions. A continued high-dividend-payout policy may also signal to shareholders that additional and costly monitoring of capital spending decisions is unnecessary.

The value-decreasing findings associated with cash flow-financed capital spending do not arise in the case of research and development spending. For the subsample of 229 firms with complete data over the sample period, cash flow-financed R & D spending is associated with high-Q firms. Additionally, high-Q firms following a low-dividend-payout policy exhibited the strongest influence of cash flow on R & D.

One implication of these results is that firms with growth opportunities closely tied to research and development are not able to rely on the external capital markets as the primary source of financing, because information costs make external financing excessively costly. Therefore, an appropriate dividend policy for these firms is to pay no dividends and to finance research and development with cash flow.

An explanation for the differing results between capital spending and research and development spending involves the incentives managers have to expand a firm’s tangible asset base. Since capital expenditures typically add to the amount of assets under managerial control and generate more predictable future cash flows, such expenditures create the opportunity to exploit free cash flow in subsequent periods. Research and development spending, however, produces less certain cash flows and increases the firm’s tangible asset base only if successful. Rather, because R & D spending represents investment in intangible and highly uncertain assets, it is more likely to be associated with the asymmetric information problems that create a financing pecking order.

Finally, the secular behavior of firms may play an important role in determining the strength of free cash flow versus pecking order behavior. This study analyzed a balanced panel of firms for which data were available during an 18-year sample period. Thus, low-dividend-payout firms must be interpreted as having consistently low-dividend-payout behavior. The large, non-financially distressed firms in this study that consistently do not pay out income to shareholders are a priori most likely to be associated with free cash flow behavior. Newer firms for which complete data were not available over the sample period may have greater asymmetric information problems, and thus show pecking order behavior.

General conclusions about the overall behavior of firms must be drawn considering the sample, The research does provide important insight, however, into the behavior of firms that survive the competitive pressures of the marketplace.

1 Evidence dates back to Meyer and Kuh (1957) and Donaldson (1961).

2 See Myers (1984) for a discussion of this issue.

3 Ravid (1988) reviews an extensive body of literature focusing on these and other interactions between investment spending and financing sources.

4 In another test of the pecking order hypothesis, Baskin (1989) analyzes the effect of profits on debt financing for Fortune 500 firms. He finds that profits and debt are negatively related, contrary to the tax-based “static trade-off” theories, which predict a positive relationship. These findings presumably support the pecking order notion that internally generated funds are substituted for external sources whenever possible.

5 The exact form of the manager’s utility function is not an issue here. The most simple characterization is a risk-neutral manager whose utility is linear in the dollar amount of investment. This characterization acknowledges that managerial incentive contracts are written to offset some of the utility gained by controlling a larger amount of assets, but they are assumed to be less than completely effective. More complex utility functions and incentive contracts can also create other adverse incentives. For instance, if managers are risk-averse and incentive contracts are suboptimal, managers may reject positive-NPV investments having sufficient downside risk.

6 The form that dividends take is irrelevant for the theoretical model. The key issue is that cash flow above the retention rate of b* is paid out. In the empirical work, I focus on annual cash dividends as a method of removing free cash flow. Stock repurchases using excess cash flow or commitment to paying out future cash flow through debt-financed stock repurchases are alternative methods. Which method a firm chooses may depend on a number of important issues. For example, firms subject to large earnings swings may find cash repurchases to be preferable because only occasionally will earnings be sufficient to achieve the optimal retention rate, b*. The effect of using my empirical definition of dividends may be to bias the tests away from finding evidence of free cash flow behavior, as some low-dividend-payout firms are in fact paying out cash flows through an alternate source.

7 Hayashi (1982) derives the conditions under which average Q (used in this study) is an appropriate proxy for marginal Q. They include perfect competition in the firm’s product market, scale-expanding investment, and profit-maximizing behavior. The assumption of profit maximization does not hold in the theoretical model when no monitoring occurs. Consequently, marginal Q is less than unity and is likely to be significantly below average Q. A violation of the profit maximization assumption should not, however, seriously change the interpretation of the results. As long as marginal Q is less than unity (i.e., net new investment is unprofitable), average Q will continue to fall and lie above marginal Q when investment is scale-expanding. A potentially more serious problem is that input price and technology shocks may adversely affect the value of the firm’s existing assets. If the firm responds to these shocks by focusing new investment into areas unrelated to the existing asset base but with higher values of marginal Q, the assumption of scale-expanding investment is violated, and the relationship between marginal and average Q disappears. Our method of controlling for this problem in the empirical work is to recognize the potential for measurement error in the Q variable and employ a fixed-effects model to control for time- and firm-specific innovations in the data.

8 For years after 1984, the firm’s bend rating reported in COMPUSTAT data item 228 is used to identify defaulted debt. Bond defaults before 1984 are obtained from Moody’s Bond manuals. An additional screen for finns not having rated debt looked for firms reporting negative cash flow for more than 50% of the sample period; none were detected.

9 Excluding lagged cash flow has no effect on the inferences drawn. Including [(CF/K).sub.i,t-1] in the regression simply reduces the magnitude of the parameter estimate on contemporaneous cash flow and raises the regression [R.sup.2] slightly.

10 The dividend payout cut-offs of 0.10 and 0.35 for low- and medium-payout firms are chosen to maintain a reasonable balance in sample sizes across payout groups. The results reported in the tables are robust to a wide range of cut-off alternatives. An alternative measure of payout behavior, similar to that employed by Fazzari, Hubbard, and Petersen (1988), groups firms by their presence in a particular payout range at least 70% of the time during the period. Results from this grouping criterion are virtually identical to the results reported in the paper.

11 Technically, the estimation was carded out using a dummy/dummy interaction procedure, where dummy variables [d.sub.1], [d.sub.2], and [d.sub.3] correspond to the high-, medium-, and low-payout firms, respectively. This facilitates hypothesis tests of the parameter estimates across firm groups.

12 An alternative explanation for this relationship is that low-Q firms are rationed from the capital market in a manner consistent with the information models of Stiglitz and Wiess (1981) and are forced to finance investment from cash flow. While this explanation includes an asymmetric information argument, it still relies on an agency story to explain the low Q value associated with positive investment spending.

13 I thank an anonymous referee for suggesting the line of inquiry presented in this section.

14 As the distribution of Q changes with the smaller sample, the Q dummies are set to one when Q falls in the ranges: QDUM1 = 1 when Q [is less than or equal to] 0.73; QDUM2 = 1 when 0.73 [is less than] Q [is less than or equal to] 1; QDUM3 = 1 when 1 [is less than] Q [is less than or equal to] 1.37; and QDUM4 = 1 when Q [is greater than] 1.37. Also the mean Q value rises from 1.06 in the capital spending sample to 1.17 in the R & D subsample.

References

Baskin, J., 1989, “An Empirical Investigation of the Pecking Order Hypothesis,” Financial Management (Spring), 26-34.

Campbell, T. and W. Kracaw, 1980, “Information Production, Market Signalling, and the Theory of Intermeditation,” The Journal of Finance (September), 863-882.

Carpenter, R., 1993, “Finance Constraints or Free Cash Flow: The Impact of Asymmetric Information on Investment,” Emory University Working Paper.

Chauvin, K.W. and M. Hirshey, 1993, “Advertising, R & D Expenditures and the Market Value of the Firm,” Financial Management (Winter), 128-140.

Deilman, T.E., 1983, “Pooled Cross-sectional and Time-Series Data: A Survey of Current Statistical Methodology,” American Statistician (May), 111-122.

Devereux, M. and F. Schiantarelli, 1990, “Investment, Financial Factors, and Cash Flow: Evidence from U.K. Panel Data,” in R. Glenn Hubbard, ed., Asymmetric Information, Corporate Finance, and Investment, Chicago, IL, University of Chicago Press, 279-306.

Diamond, D., 1984, “Financial Intermediation and Delegated Monitoring,” Review of Economic Studies (July). 393-414.

Donaldson, G., 1961, Corporate Debt Capacity: A Study of Corporate Debt Policy and the Determination of Corporate Debt Capacity, Boston, MA, Harvard University Graduate School of Business Administration.

Fazzari, S.M., R.G. Hubbard, and B. Petersen, 1988,” Financing Constraints and Corporate Investment,” Brookings Papers on Economic Activity 1:1988, 141-206.

Fazzari, S.M. and B. Petersen, 1993, “Investment Smoothing with Working Capital: New Evidence on the Impact of Financial Constraints,” The RAND Journal of Economics (Autumn), 328-342.

Hall, B.H., 1992, “Investment and Research Development at the Firm Level: Does the Source of Financing Matter,” NBER Working Paper no. 4096.

Hayashi, F., 1982, “Tobin’s Marginal q and Average q: A Neoclassical Interpretation,” Econometrica (January), 213-225.

Himmelberg, C. and B. Petersen, 1994, “R and D Investment and Internal Finance: A Panel Study of Small Firms in High-Tech Industries,” Review of Economics and Statistics (forthcoming).

Hoshi, T., A. Kashyap, and D. Scharfstein, 1991, “Corporate Structure and Investment: Evidence from Japanese Panel Data,” Quarterly Journal of Economics (February), 33-60.

Jensen, M. and W. Meckling, 1976, “Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure,” Journal of Financial Economics (October), 305-360.

Jensen, M., 1986, “Agency Cost of Free Cash Flow, Corporate Finance, and Takeovers,” American Economic Review (May), 323-329.

Jensen, M., 1993, “The Modern Industrial Revolution, Exit, and the Failure of Internal Control Systems,” Journal of Finance (July), 831-880.

Lang, L. and R. Litzenberger, 1989, “Dividend Announcements: Cash Flow Signalling vs. Free Cash Flow Hypothesis,” Journal of Financial Economics (September), 181-191.

Lindenberg, E. and S. Ross, 1981, “Tobin’s Q Ratio and Industrial Organization,” Journal of Business (January), 1-32.

Meyer, J.R. and E. Kuh, 1957, The Investment Decision: An Empirical Study, Boston, MA, Harvard University Press.

Modigliani, F. and M. Miller, 1958, “The Cost of Capital, Corporation Finance, and the Theory of Investment,” American Economic Review (June), 261-297.

Myers, S., 1984, “The Capital Structure Puzzle,” Journal of Finance (May), 575-592.

Myers, S. and N. Majluf, 1984, “Financing Decisions When Firms Have Investment Information that Investors Do Not,” Journal of Financial Economics (June), 187-220.

Oliner, S. and G. Rudebusch, 1992, “Internal Finance and Investment: Testing the Role of Asymmetric Information and Agency Cost,” Review of Economics and Statistics (November), 643-654.

Pinegar, J.M. and L. Wilbricht, 1989, “What Managers Think of Capital Structure Theory: A Survey,” Financial Management (Winter), 82-91.

Ravid, A., 1988, “On the Interactions of Production and Investment Decisions,” Financial Management (Winter), 82-91.

Smith, C., 1986, “Investment Banking in the Capital Acquisition Process,” Journal of Financial Economics (January/February), 3-29.

Stiglitz, J. and A. Wiess, 1981, “Credit Rationing in Markets with Imperfect Information,” American Economic Review (June), 393-410.

Strong, J.S. and J.R. Meyer, 1990, “Sustaining Investment, Discretionary Investment, and Valuation: A Residual Funds Study of the Paper Industry,” in R. Glenn Hubbard, ed., Asymmetric Information, Corporate Finance, and Investment, Chicago, IL, University of Chicago Press, 127-148.

Whited, T., 1992, “Debt, Liquidity Constraints, and Corporate Investment: Evidence from Panel Data,” Journal of Finance (September), 1425-1460.

Data Appendix

Variables developed from the COMPUSTAT data base include:

1. Cash flow (CF), defined as net income before extraordinary items plus depreciation (COMPUSTAT items 14 and 18).

2. DCASH, defined as the change in the book value of the firm’s cash and equivalents (COMPUSTAT item 1), and [C.sub.i,t-1] is the beginning-of-period stock.

3. SALES and I are taken directly from COMPUSTAT items 12 and 128 respectively.

4. Research and development expenditures (RD) are COMPUSTAT item 46.

5. Beginning-of-period gross plant and equipment (K) is taken from COMPUSTAT item 7 in period T-1.

6. All variables except Q are weighted by K to make comparisons relative to firm size. An alternate weighting scheme using total assets had no substantive effect on the results.

7. COMPUSTAT item 129 was used to identify acquisitions.

8. Finally, Q is defined as the market value of the firm’s assets divided by the replacement cost of assets, MVA/RC, and is calculated in a manner consistent with Lindenberg and Ross (1981).

* The market value of the firm is the market value of common equity (COMPUSTAT item 24 multiplied by item 25) plus the market value of preferred stock calculated as preferred dividends paid (item 19) divided by the average preferred dividend yield on preferred stock of medium-risk companies plus an estimate of the market value of debt.

* The market value of debt is estimated by discounting the proportion of long-term debt maturing in each of the next five years (COMPUSTAT items 44, 91, 92, 93, and 94) by the average yield on equivalent-maturity Baa bonds. Any remaining long-term debt (item 9 minus items 44, 91, 92, 93, 94, and 111) is assumed to mature evenly over the next twenty years and is also discounted at the average Baa rate. Newly issued debt (item 111) is assumed to have a maturity of twenty years. Finally, the market value of short-term debt (item 5) is assumed to equal its book value. The sum of the above market value approximations is the measure of the market value of debt.

* The replacement value of assets is computed as the replacement value of plant and equipment plus the replacement value of inventory.

* An estimate of the replacement value of plant and equipment is obtained by first estimating the average age of plant and equipment. Average age (AGE) is calculated as the ratio of accumulated depreciation (item 7 minus item 8) to current period depreciation (item 14) times a five-year moving average (to smooth the data) of the ratio of gross plant and equipment (item 7) to current depreciation (item 14). The value of net plant and equipment is then adjusted for inflation by multiplying its current value by the ratio of the current year producer price index (PPI) to its value AGE years ago.

* Replacement value of inventory is assumed to equal book value (item 3) unless the firm uses last-in/first-out inventory accounting, in which case inventory is inflated by the ratio of the current PPI to its value in the previous period.

Stephen C. Vogt is Assistant Professor of Finance at DePaul University, Chicago, Illinois.

COPYRIGHT 1994 Financial Management Association

COPYRIGHT 2004 Gale Group