THE INVESTMENT OPPORTUNITY SET AND CORPORATE POLICY CHOICES

• 3/27/86 THE INVESTMENT OPPORTUNITY SET AND CORPORATE POLICY CHOICES Clifford W. Smith and Ross L. Watts l 1. INTRODUCTION The determinants of th...
Author: Felix Craig
9 downloads 0 Views 2MB Size


3/27/86

THE INVESTMENT OPPORTUNITY SET AND CORPORATE POLICY CHOICES

Clifford W. Smith and Ross L. Watts l

1. INTRODUCTION The determinants of the finn's capital structure and dividend policy have long been central issues in corporate finance. Recently, attention has also been focused on the determinants of executive compensation policy.I The consideration given these issues is at least partially driven by management demand for policy prescriptions. Existence of optimal financing and compensation policies, at least, is suggested by the stability of policy differences across industries.' Modigliani/Miller (1958, 1961) demonstrated that given the firm's investment policy, in the absence of taxes and contracting costs neither financing policy nor dividend policy affect the firm's market value. Since then, capital structure, dividend policy and compensation policy explanations have relied heavily on taxes andlor contracting costs.'

1 Graduate School of Management, University of Rochester, NY 14627. This work has

been partially supported by the GSM's Managerial Economics Research Center. Previous

drafts of this work were titled, "The Structure of Executive Compensation and the Control

of Management." 2 See Miller/Scholes (1981), Smith/Watts (1982) and volume 7 of the Journal of

Accountin~ and Economics.

3 See Schwartz and Aronson (1967) and Fox (1980) for evidence of the stability of inter­

industry differences in financial and compensation policies respectively.

4 For example, ModiglianilMiller(l963), Black/Scholes (1974), and Miller/Scholes (1981)

suggest taxes explain financial, dividend and compensation policies respectively and Myers

(1977), Jensen/Smith (1985) and SmithlWatts (1982) suggest contracting costs explain

financial, dividend and compensation policies respectively.

!

2



Contracting cost explanations for financial policy (e.g., Jensen/Meckling, 1976; Myers, 1977) and dividend policy (e.g., Rozeff, 1982; Easterbrook, 1984; Smith, 1986) suggest those policies are partially driven by the firm's investment opportunity set. In this paper, we suggest that executive compensation policy is also dependent on the firm's

v

investment opportunity set. If all three policies-financing, dividend and compensation-are driven by the characteristics of firms' investment opportunities, not only should the individual policies be correlated with the investment opportunity set, they should be correlated with each other. However, the correlations among policy choices should not be perfect; the various aspects of the investment opportunity set that determines corporate policy choices are not perfectly colinear across firms. This paper investigates the empirical relations at the industry level among financial, dividend and executive compensation policies and the investment opportunity set. The primary objective is to test various existing incentive-based explanations for corporate policy choices. We find the various policies are correlated with each other and with the investment opportunity set, as expected under the contracting explanations for the policies. We also investigate the extent to which these correlations among the policies and the investment opportunity set can be explained by taxes or regulations. Titman/Wessells (1985) examine a similar set of issues, however they impose a complex structure on the estimated relations among variables. Obviously if the structure they impose is correct, the power of their estimates is increased. However, we are hesitant to impose such a complex structure given the state of knowledge. Without that additional structure, we are unable to separate the component partial effects in the correlations we observe. For example, incentive effects suggest firms with certain characteristics should " have lower debt in their capital structure while tax effects suggest higher debt. If the correlation between the characteristic and capital structure is significantly negative, we assume the incentive effect is significant, while if the correlation is positive, we assume the

3 tax effect is significant. Since we only observe the net effect, we cannot separately identify the significance of less important partial effects. LonglMelitz (1985) also examine the relation between financing policy and certain aspects of investment policy; hence their analysis represents a subset of the range of issues we examine. Our empirical analysis not only includes a broader range of investment policy characteristics, it relates them to dividend and compensation policy as well as financial policy. Section 2 explores contracting, regulation and tax explanations for financing, dividend and compensation policy and generates predictions for the correlations between each policy and the investment opportunity set as well as among the policies themselves. Section 3 describes the data and variables used to represent the investment opportunity set and corporate policies. The evidence on the correlations is presented in Section 4. Section 5 discusses the results' implications for empirical studies and presents our conclusions.

2. RELATION BETWEEN THE INVESTMENT OPPORTUNITY SET AND POLICY CHOICES. Financing policy, dividend policy, execu tive compensation policy, leasing policy, marketing policy and other corporate policies are all chosen by managers and other parties to the firm; they are endogenous. An empirical model of the cross-sectional variation in those policies requires specification of the exogenous variables that drive policy selection. Four potential exogenous variables are:

1) contracting technology-relatively standard contractual forms have been developed (e.g., a sinking fund, a convertible bond, an executive stock option, a cancellable lease, etc.);

4 2) individuals' preferences-the individual's marginal rate of substitution between risk and expected wealth, between consumption today and tomorrow, etc.; 3) firms' investment opportunity sets-the firm's prospective set of investment opportunities and their respective payoffs; and 4) regulation and taxes-rate of return regulation, entry regulation, tax preference items, and the structure of the tax code.

The set of feasible contracts does not appear to vary much across firms; all have access to alternative contractura1 provisions. And, given well functioning labor and capital markets, firms have access to any potential stockholder, bondholder, manager, lessor, or customer. Thus, of the four exogenous variables only the investment opportunity set and regulation/tax variables appear to have potential to explain and predict cross-sectional policy variations, since only they vary across firms.

2.1 Determinants of Financing Policy Investment opportunity set. Myers (1977) describes the firm's future investment opportunities as call options. These options' values depend on whether management can be expected to exercise them. If the firm has risky debt outstanding, it is possible for situations to arise where an option's exercise (taking a positive present value project) reduces the share value because most payoffs go to the debtho1ders. Unless this conflict between the shareholders and debtho1ders is controlled, the probability these real investment options will be exercised is reduced and so is the firm's value. One way to avoid this underinvestment problem and consequent value loss is to issue debt corresponding only to the firm's assets in place. Hence, Myers predicts that (ceteris paribus) the larger the proportion of firm value represented by assets in place, the lower the firm's equity/value ratio.

5 Another characteristic of the investment opportunity set which affects the capital structure choice is the sensitivity of asset values to use and maintenance decisions. Assets whose values are very sensitive to use and maintenance decisions are more likely to be abused or ill-maintained if the firm gets into financial distress. Thus, firms with such assets should have less debt in their capital structure (see Alchian/Demsetz, 1972 or Smith/Wakeman, 1985). A third investment opportunity set characteristic of potential importance in the capital structure decision is the liquidity of the secondary market for the assets. With more liquid secondary asset markets, the realizable value of the assets in financial distress is greater, and the firm should have more debt in its capital structure (see Smith, 1980). The final investment opportunity set characteristic that might affect the capital structure decision is the extent to which the firm has off-balance sheet claims against the corporation-things like product warranties. The higher the level of these fixed claims against the firm's cash flows, the lower the optimal amount of debt in the firm's capital structure. Re&ulation and taxes. The degree of regulation faced by the firm is potentially important in determining the firm's capital structure. If regulation restricts the firm to particular investments and the regulations are monitored and enforced, the discretion over projects is reduced and more debt should be issued. Hence, it is predicted that regulated firms have lower equity/value ratios.

In some regulatory jurisdictions, regulators

effectively impose minimums on a regulated utility's leverage. If the utility'S leverage is below the minimum, allowed output prices are set to compensate the firm as if it has more debt in its capital structure. (Note: since the return allowed the equityholders generally is not adjusted for the higher leverage, allowed output prices are lower than they would otherwise be.) Conversely, regulation in the banking and insurance industry tends to impose maximums on leverage.

6 Tax code provisions have a potentially important impact on the capital structure choice. DeAngelolMasulis (1980) argue firms that generate substantial non-interest-related tax shields, such as investment tax credits, have a comparative disadvantage in utilizing interest tax shields; thus such firms should have less debt in their capital structures. If a firm whose investment opportunity set contains assets-in-place is more likely to generate investment tax credits than a firm with more growth options, this tax-related prediction differs from the incentive effects. An alternative channel through which taxes might affect policy choices is a positive empirical association between the fraction of intangible assets and the firm's cash flow variance. If such an association exists, progressivity in the tax structure would cause a higher fraction of intangible assets to be associated with higher expected tax liabilities (Smith/Stulz, 1985). This would increase the amount of debt in the firm's capital structure, over the range of tax structure progressivity. Finally, banks are allowed to receive tax-exempt income from municipal bonds while deducting interest paid on CDs. Insurance companies have similar opportunities. Thus, we expect banks and insurance companies to have more fixed claims in their capital structure to take advantage of this form of tax arbitrage.

2.2 Determinants of Dividend Policy Investment opportunity set. Investment and dividend policy are linked through the firm's cash flow identity. Given managers take all positive net present value projects, the larger the investment during the period, the smaller the dividend payout and/or the larger the new equity issued. Given the cost of new equity issues, it is expected that (ceteris paribus) firms with more intangible investment opportunities will have lower dividend payout rates. The preceding prediction can also be generated by considering the administration of debt covenants. Dividend covenants that specify a maximum on payouts also impose a minimum investment requirement (see SmithIWamer, 1979; Kalay, 1982), and so reduce

7 the underinvestment problem by controlling the payout and the agency costs of debt. However, the more binding the dividend constraint, the more likely the managers will be forced to take negative net present value investments. This latter likelihood is a function of the firm's investment opportunities. Firms with many profitable investment opportunities can restrict their dividends more before the lower agency costs are offset by the expected cost of negative net present value investments. Rozeff (1982) and Easterbrook (1984) provide an alternative contracting argument which reinforces the predicted relation between the investment oppportunity set and dividend payout. They argue that firms with new investment opportunities go to the capital market frequently. The new issues market provides effective monitoring and discipline, thus lowering agency costs. Firms with fewer investment opportunities would go to the new issues market less frequently and lose this benefit if they did not pay higher dividends.P

Re&ulation and taxes. Smith (1986) argues the regulatory process provides utility managers with an incentive to pay higher dividends than they otherwise would. By paying higher dividends the utility is constrained (given investment policy) to raise funds in the capital market more frequently. Those new issues provide evidence on the firm's cost of capital that is useful in the regulatory process. Without such evidence, the utility commision would allow a lower rate of return. Hence higher dividends and their implied more intensive use of capital markets monitors and diciplines the members of the regulatory system as well as the firm's managers. Thus, we expect (ceteris paribus) regulated firms have higher dividend payout rates. In the finance literature on dividend policy, we can find no tax-related anlaysis of corporate dividends policy that has cross-sectio:

implications in the dimensions we

examine empirically. 5 In the contracting literature, there is no obvious impact of the sensitivity of asset values

to use and maintenance decisions, the liquidity of secondary asset markets, or the extent of off-balance-sheet claims on the firm's choice of dividend policy.

8

2.3 Determinants of the Level of Compensation Investment opportunity set. We expect the level of chief executive compensation to vary with the firm's investment opportunity set. In particular, we expect (ceteris paribus) the larger the proportion of firm value represented by intangible investment opportunities, the greater the manager's compensation. We expect this relation for three reasons. First,

a

12Iim:i.. we expect the marginal product of an investment decision-maker to be greater than the marginal product of a supervisor and good investment decision-makers to be relatively more scarce than good supervisors. Second, the higher the firm's risk, the higher the risk of the manager's compensation, and given risk-averse managers who cannot diversify away their compensation risk, the higher the manager's equilibrium compensation. We expect that, as an empirical proposition, the larger the proportion of fum value represented by growth options, the greater the the firm's risk. Third, the larger the proportion of firm value represented by intangible assets, the more likely the manager's compensation will be tied to firm value (see below) and the greater the variance of the manager's compensation. Another characteristic of the investment opportunity set of potential importance in setting compensation is the sensitivity of assets' future values to use and maintenance decisions. Managers of firms whose asset values are sensitive exercise more discretion with respect to the use and maintenance of the assets, have a higher value of marginal product, and thus should have higher total cornpensation.v Re~ulation

and taxes. Regulation restricts the manager's investment decision

discretion and thus his marginal product as a decision maker and his compensation. Hence (ceteris paribus) we expect regulated firms have lower manager compensation. However, some rate regulation authorities appear to have directly regulated compensation policy, placing maximums on payments to executives. One effect of this is to increase perk

6 There is no obvious impact on the level of compensation associated with the liquidity of secondary asset markets or the extent of off-balance-sheet claims.

9 consumption-activities which the executive values but are harder for the regulator to monitor. We can find no tax-related analysis of executive compensation policy that has cross­ sectional implications for the level of compensation paid executives.

2.4 Determinants of the Use of Incentive Compensation Investment opportunity set. The typical problem analyzed in the principal-agent literature is that of a risk-neutral principal trying to induce a risk-averse agent to take the action the principal would take.? If the principal can observe the agent's actions, the optimal contract pays the agent a fixed wage and penalizes him if he takes dysfunctional actions. That contract imposes all the risk on the risk-neutral principal. If the principal cannot observe the agent's actions, the optimal contract gives the agent a share in the outcome of his actions. The incentive that contract gives to the agent to expend effort to achieve the principal's objective offsets the increased compensation that must be paid the agent to offset his increased risk. Applying the agency analysis to large firms, shareholders are considered risk­ neutral because they can diversify the firm-specific risk. And, if managers cannot diversify their compensation risk, they are risk-averse in their actions. We argue that the actions of a manager of a firm that has few growth options and whose value arises from assets in place are more readily observable than the actions of a manager of a firm that has many investment opportunities. The supervisory and monitoring activities of the first manager can be observed. However, it is more difficult for shareholders (or their designated monitors, outside members of the board of directors, external accountants, etc.), who do not have the second manager's specific knowledge, to observe the set of alternative investments from which he chooses. Hence, we expect that the larger the proportion of 7 For examples of this literature see Wilson (1968), Berhold (1971), Spence and Zeckhauser (1971), Ross (1973,1974), Mirrlees (1974,1976), Stiglitz (1974,1975) and Holmstrom(1979).

10 firm value represented by investment opportunities, the more likely the firm ties the manager's compensation to the effect of his actions on finn value. Tying the manager's compensation to the firm value effects of his actions doesn't by itself imply the use of formal incentive plans.

The manager's salary could be

informally renegotiated each period based on the previous period's performance. However, the effectiveness of future salary renegotiation depends on expected future employment (e.g., a 64 year old manager facing retirement at 65 would be little motivated by an annual salary renegotiation scheme). It also depends on the probability the person promising the renegotiation will be around when the renegotiation is due. Informal salary renegotiation will be less effective if there is high management turnover and hence less reason to expect future managers to honor unwritten, informal contracts. An alternative reward mechanism is an explicit incentive plan that ex ante ties the manager's compensation to a finn performance measure that reflects the effects of the manager's actions on firm value (e.g., stock price or accounting earningsj.f Hence our prediction is that (ceteris paribus) the larger the proportion of firm value represented by intangible investment opportunities, the more likely the firm has a formal incentive compensation plan. The above contracting arguments not only imply the nature of the performance measure (i.e., a measure of the effect on firm value), they can be extended to explain the structure of the incentive plans. In SmithlWatts (1982) we argue that option-type plans (stock option, stock appreciation, performance and bonus plans) not only provide the manager with incentives to increase firm value, they also control the manager's risk aversion. Risk averse managers will forego a project that increases firm value if it increases firm risk (and manager compensation risk) sufficiently. Similarly, they will undertake a value-reducing project if it decreases firm risk (and manager compensation risk) sufficiently. Since the expected payoff to options increases with the volatility of the price

8 We expect such plans to be more frequent on in industries with higher managerial turnover and less common in family controlled firms where information on informal contracts is likely to be passed from one manager to another.

"

11 of the underlying asset (firm value), option type plans based on firm value measures provide the manager with incentives to undertake projects that increase firm risk. LambertiLarcker (1985) provide evidence that option type plans do provide such incentives, Note that while we predict that the larger the proportion of firm value represented by investment opportunities, the more likely the firm has a formal incentive plan such as a bonus, stock option, stock appreciation or performance plan, we cannot predict whether the firm will have a bonus plan versus a stock option plan or a combination of the two plans or

some other combination. To date, the theory is not developed to yield these detailed predictions. For example, we do not know whether for a particular firm, stock prices or accounting earnings best measure the effects of managers' actions. Another characteristic of the investment opportunity set that should help determine the use of incentive compensation is the sensitivity of assets' future values to use and maintenance decisions. With greater sensitivity to use and maintenance decisions, it is more difficult to directly monitor the implications of the exercise of managerial discretion. It is thus more likely that managers of these firms will be offered formal incentive plans.? Re~ulation

and taxes. Regulation restricts the investment opportunity set and

makes observation of the manager actions easier. Hence, we expect regulated firms are less likely to use formal incentive plans. Finally, taxes are potentially important in determining the use of incentive compensation plans, The Miller/Scholes (1982) analysis of executive compensation indicates that incentive compensation plans frequently have a deferral aspect to them, and they show that if the executive's effective tax rate is higher than that of the corporation, there is an incentive for deferral. As we have indicated above, the ability to utilize tax exempt municipal bond interest suggests that banks and insurance companies should have

9 There is no obvious impact on the use of incentive compensation associated with the

liquidity of secondary asset markets or the extent of off-balance-sheet claims.

12 the lowest tax rate, thus providing strong tax-deferred incentives for use of incentive compensation.

2.5 Relations Among Policies Given that the fraction of intangible assets in the firm's investment opportunity set drives financial, dividend and compensation policies, we can infer relations among those policies from the above predictions. Since the larger the proportion of firm value represented by investment opportunities, the higher the firm's equity/value ratio, the lower its dividend payout rate, the greater its manager's compensation and the more likely it has a formal incentive compensation plan, we expect the following relations to hold:

1) equity/value ratios and dividend payout rates are negatively correlated; 2) equity/value ratios and management compensation levels are positively correlated; 3) equity/value ratios and the existence of formal incentive compensation plans are

positively correlated; 4) dividend payout rates and management compensation levels are negatively correlated; 5) dividend payout rates and the existence of formal incentive compensation plans are negatively correlated; and 6) management compensation levels and the existence of formal incentive compensation plans are positively correlated.

These predictions for relations among policies and the predictions for the relations between the investment opportunity set and policy are summarized in Table 1. However, because of differential impacts on financing, dividend and compensation policies and other characteristics of the firm's investment opportunity set (e.g., sensitivity of the firm's

13 tangible assets to use and maintenance decisions, the liquidity of secondary markets for the firm's tangible assets), and differences in regulation and taxes, we do not expect the correlation among policies to be perfect.

3. DATA 3.1 Data Sources Estimates of the correlations among investment opportunity set and policy variables are obtained using industry level data. The primary reason for the use of industry level data is the ready availability of data on the level of top management compensation and the use of formal incentive plans by industry in the Conference Board surveys of executive compensation. We use the survey data for 1981 as reported in Fox (1982). The use of industry level data should reduce measurement error in the variables. It should also maintain dispersion among the variable estimates if Fox's classification of industries using SIC codes effectively groups firms by the nature of their investment opportunity set. Further, since this grouping procedure doesn't use any of the variables whose correlations we examine, it doesn't introduce the bias present in many studies that use alternative grouping procedures to reduce measurement error. 10 The industries for which Fox reports compensation data and their definition (in terms of SIC codes) are reported in Table 2. That table also reports the number of firms in each industry that provide 1981 compensation data for Fox's survey. We are unable to obtain finn-level data from Fox, so we estimate investment opportunity set, financial policy and dividend policy variables for each of Fox's industry definitions using Compustat quarterly and annual firm data for the period 1976-1981. In calculating each variable for Fox's industries we use every firm that has available data. Hence, the number of firms

Examples of studies that introduce bias by employing grouping procedures are given in Wheatley(1982), BeaverlLambertJRyan (1985) and Wattsrzimmerman(1986). 10

14 used for a given industry can vary for different variables, although for most variables (research and development being the major exception), the number is the same. Table 2 reports the modal number of firms used in calculating investment, financing and dividend variables for each industry (reported as "Compustat sample number of firms"). We try to minimize the differences between our sample and Fox's. Because the shares of most (47 out of 70) of Fox's construction firms are not traded and the shares of all our construction fmns (15) are traded, we use only compensation data for the 23 traded firms in Fox's sample.U The other major systematic difference that we can detect "a priori" in the two samples' selection criteria is for the insurance industry. Fox's sample includes 76 stock companies and 67 mutual companies, while our sample consists entirely of (14) stock companies. Unfortunately, while Fox does disaggregate the compensation data out by life versus property and casualty companies, he does not report compensation by stock versus mutual companies. Hence, we use compensation data for Fox's overall insurance industry sample. To the extent that the compensation data of stock and mutual companies differ (as -ve expect they do I2 ) , this will introduce error into our analysis. However, since the insurance industry is regulated, this error will not affect our analysis of unregulated industries. The only attribute Fox reports that we can use to compare the two samples "ex post" is median 1981 sales. Table 2 provides the median sales for the two sets of industry samples. For the last three industries in Table 2 (insurance, gas and electrical utilities, and commercial banking), the reported median sales numbers are median premium income, median operating income and median deposits, respectively. Except for the insurance

11 Fox provides compensation data separately for the 23 traded firms, except for the frequency of use of bonus plans. There we use the overall sample frequency (90%) as an estimate of the frequency of bonus plan use for the traded construction firms. This estimate is biased downward, but the extent of that bias is limited by the maximum frequency being 100%. 12 Mayers/Smith (1981) argue that mutuals will predominate in linesof insurance requiring less managerial discretion and thus lower expected compensation and less incentive compensation.

15 industry, Compustat sample sales numbers are obtained from Compustat. The premium income for the 14 insurance firms with financial data available on Compustat are obtained from Moodys' 1982 Bank and Finance Manual. Not knowing Fox's sample firms makes it difficult to test for differences in the two samples' firm sizes in each industry. However, the numbers in Table 2 indicate the samples' firm sizes differ in several industries, in particular the food, consumer chemical, industrial chemical, insurance and commercial banking industries. In the first three of those industries, Fox's sample firms are larger than the Compustat sample firms and in the last two, the Compustat firms are larger. Since Fox's sample is used for compensation policy variables and the Compustat sample for all other variables, these differences only affect estimated relations involving compensation variables. The effect can produce a bias toward the predicted relations between compensation variables and other variables if compensation and other variables are correlated with the size differences. To test this possibility, in Section 4 we report the correlations between size differences and the various policy and investment opportunity set variables. Since some of those correlations are significant, in that section we also estimate the relations between the various variables excluding industries with "substantial" size differences in the two samples (i.e., the food, consumer chemical, industrial chemical, insurance and commercial banking industries).

3.2 Variables i) Investment opportunity set One characteristic of the investment opportunity set our analysis identifies is the ratio of assets in place to firm value. The variables used in this study to proxy for this variable are:

1) Book Value of Assets / Firm Value (A/V). Firm value (V) is measured as the market value of equity (E) plus the book value of debt (B). The book value of assets (A) is

16 used as a surrogate for assets in place. Hence, the higher this ratio, the higher the ratio of assets in place to fIrm value and the lower the ratio of the value of investment opportunities to firm value. The ratio is calculated for each industry i using five years of data:

1981

A/Vi

L

= {

(1)

t=1977

where

Nj,t is

the number of firms j in industry i with data available in year t.

2) Depreciation / Firm Value (Dep/V), The annual depreciation charge (Dep) is used as a surrogate for assets in place. The higher the ratio Dep/V, the higher the ratio of assets in place to finn value and the lower the ratio of the value of investment opportunities to firm value. The ratio is calculated for each industry i using five years of data:

1981

Dep/V j = {

L

(2)

t=1977

where

Nj,t is

the number of firms j in industry i with data available in year t.

3) Research and Development / Firm Value (R&D/V). Expenditures on research and development are investments that do not immediately result in assets in place. They do result in options for future investment. Hence, we expect firms with relatively high research and development expenditures to have relatively low assets in place and relatively high investment opportunities. This expectation is supported by the negative relation between the rate of investment in research and development and the level of borrowing reported by Long/Malitz (1985).

17 There are, however, problems with this measure. Reporting of research and development expenditures is not standardized and some firms do not report research and development expenditures even though they incur them. However, unless there is some systematic association between reporting and policy choice, these measurement problems should reduce the likelihood of observing the predicted associations between investment opportunity sets and policy choices. As with the other ratios, R&D/V for each industry i is calculated using five years of data:

1981

R&D/Vi = {

L

(3)

t=1977

where Ni,t is the number of firms j in industry i with data available in year t.

4) Variance of Return on Investments (VAR). We have no theoretical reason to

expect the variance of the firm's return on investments to be related to assets in place or investment opportunities. However, if there is an empirical association (for example, if assets in place are lower variance and investment opportunities are higher variance assets) then variance and the fraction of assets-in-place in the firm's investment opportunity set would proxy for each other. In that case, the progressivity of the tax code would produce a tax-based explanation for an observed relation between both assets-in-place and financing policy, and the use of incentive compensation provisions. We estimate the variance of the rate of return to see if there is this empirical association with the above investment opportunity set measures and, to the extent there is, use the estimated variance as an investment opportunity set measure.

The variance is estimated four ways (using

accounting and market returns and quarterly and annual data). For convenience, given the

18 results are similar under the different estimation methods, only the annual accounting data results are reported. The rate of return on investments for firm j for period t (rj,t) is calculated using annual accounting data as follows:

rj,t

= ( OIj,t + INTj,t ) / (Ej,t-l + Bj,t-l )

(4)

where OIj,t is the operating income of firm j in year t, INTj,t is interest expense for firm j in year t, Ej,t-l is the market value of equity of firm j at the end of year t-l and Bj,t-l is the book value of debt of firm j at the end of year t-1. Using quarterly data the period is defined as a quarter rather than a year and using market data the numerator in equation (4) is the dollar return on equity for year t plus interest expense for year 1.

5) Other Variables. We use a dummy variable for regulation. Apart from whether or not the industry is regulated, two other variables are used to represent the investment opportunity set: the earnings/price ratio, and the ratio of capital expenditures to value. The earnings/price ratio reflects the firm's investment opportunities: the greater the investment opportunities, the lower the earnings price ratio. The correlations between the earnings/price ratio and policy variables are very similar to the correlations between the book value of assets/firm value ratio and policy variables, hence the earnings/price correlations are not reported except in the few cases where they differ from the asset/firm value correlations. The capital expenditures/value ratio might represent the flow of the firm's investment opportunities or it might represent investment in assets in place. 'A priori' it could be positively or negatively correlated with investment options. Long/Malitz (1983) find the rate of capital expenditure negatively correlated with the level of borrowing. That

19 suggests the capital expenditure/value ratio will reflect assets in place. We find the same correlation. Indeed, the correlations between the capital expenditures/value ratio and policy variables are, like those for the earningslprice ratio, very similar to the correlations for the asset/value ratio. Hence we report the capital expenditures/value ratio correlations only in the few cases where they differ from the asset/value correlations. While our analysis suggests that other characteristics of the investment opportunity set are also relevant-the liquidity of the secondary market for the firm's assets and the sensitivity of the future value of the firm's assets to use and maintenance decisions-we have no satisfactory proxy for either variable.

ii) Financing policy

The finn's financial policy is represented by the its equity/value ratio (E/V). The equity/value ratio for industry i is calculated using five years of data:

E/V i ={

1981

I.

t~1977

Ni,t

[I.

(5)

Ej,t/(Ej,t+Bj,t)] / Ni,t } 15

j=l

where Ni,t is the number of firms j in industry i with data available in year t.

iii) Dividend policy

The firm's dividend policy is represented by its dividend yield and its dividend payout rate.

To avoid the zero or negative denominator problem, we use the

earnings/dividend ratio to measure the dividend payout rate.

The higher the

earnings/dividend ratio, the lower the dividend payout rate.

1) Dividend Yield (DIP). The dividend yield for industry i (DIP i) is calculated as

follows:

20

D/P j = {

1981

L

N·1, t

[ L Dj,tlPj,t] / Nj,t } / 5

(6)

j;1

t;1977

where Nj,t is the number of firms j in industryi with data available in year t, Dj,t is dividends per share for firm j in year t, and Pj,t is firm its share price at the end of year t.

2) Earnings/Dividend Ratio (X/D). The earnings/dividend ratio for industry i (XIDj) is calculated as follows:

1981

xo, = {L

Nj,t

Nj,t

[L Xj,t/ L Dj,t] } / 5

t;1977

j;l

(7)

j;1

where Nj,t is the number of firms j in industry i with data available in year t and Xj,t is earnings per share for firm j in year t. Dividends and earnings per share are summed across firms before the ratio is calculated to ensure that the denominator is positive and not close to zero.

While the dividend yield is generally correlated with investment opportunity set, financing policy and compensation policy variables, the earnings/dividend ratio is not. Hence, we only report the earnings/dividend ratio correlations in the very few cases in which they are significant.

iv) Compensation policy Compensation Level. We use the chief executive's salaty as a surrogate

for the

.1kve1 of maU@ilRltpt compensation. Since it ignores compensation under incentive plans, this surrogate measures compensation level with error. However, as we shall see, both the use of, and payments under, incentive plans are positively correlated with salary. Hence,

21 ignoring incentive compensation probably reduces the likelihood of observing the predicted relation between the investment opportunity set and compensation level because it reduces the variation in measured level of compensation. AcrosS firms, the chief eIM'Itiug's

8iiR'8Aliiti~A: uiiiltii

willi firm

iii"

~.g.,

see

Fox, 1982 and Murphy, 1984). We adjust for this size effect using a regression estimated by Fox. For each industry Fox regresses the log of the chief executive's 1981 salary on the log of firm sales. The estimated regression parameters are reported in Table 3. Using these parameters, the chief executive's salary is calculated for the same sales figure for each industry. Table 3 gives these size-adjusted salary numbers for five sales figures: the lowest median industry sales number-$190 million (SMin); $530 million (S530); the median sales number for all manufacturing industries-$730 million (S730); $930 million (S930); and the highest median industry sales-$1,900 million (SMax). As can be seen in Table 3, the industry size-adjusted salary numbers are robust to the level of sales chosen as the adjustment point; the lowest correlation coefficient between the alternative size-adjusted salary measures (that between SMin and SMax) is .86. To minimize problems with out-of­ sample forecasts, we use salaries adjusted to the median sales for all manufacturing industries (S730).

Use of Incentive Plan;.., The variables for the use of incentive plans are the percentage of firms in each industry with each type of plan:l:wnus plan (Bonus); restrjcte;!

&tock pkm (RSP); lang-term perfQ[mapge~an (performance shares or units, LTP); s10Ck ~tion

P1&P (SOP); and stock aepreciation right plan. (SAR). Since we cannot predict the

type of incentive plan or combination of plans, we would also like to use the percentage of firms in each industry with at least one incentive plan. However, that number is not available in Fox (1982).

22 3.3 Variable Values Table 4 reports the mean investment opportunity set, financing policy and dividend policy variables for each industry. It also reports compensation variables for each industry: the estimated chief executive's salary for each industry at a sales level of $730 million; and the percentage of firms in each industry with bonus, restricted stock, long-term performance, stock option and stock appreciation plans. Finally, Table 4 reports the median bonus and the median long-term perfprrnapce plan payment for the manufacturing

-

i!.ldustries as a group and for the other industrieS jpdjvjdIJ?1].r. 13 Because data is not available to calculate them, Table 4 does not report R&DN and VAR for the insurance and banking industries. Hence, correlations are not reported below for those variables for data sets that include regulated firms. Insurance firms and banks undoubtedly have research and development expenditures (for example they develop new procedures and new products), but those expenditures are not separately reported. The mediaD bopm and 19118 salary) reported in Table

4

are

tifiil tJiii8fMllftft81

pgsjtiuily 89"81&t811

plan paymepts (as a percentage of

'ddl ulaQI lexel suggesting use of

salary level as a surrogate for compensation level is conservative. The correlations between median bonus and 5730 and between median long-term performance plan payment and 5730 are .68 and .22 respectively. In Table 4 the E/V ratio is lower for regulated firms as predicted. For all regulated firms the average E/V is .166 and for unregulated firms it is .442. The t statistic for the difference is 4.4 and is significant at any reasonable probability level. This difference is not due to regulated firms having relatively more assets in place than unregulated firms. The average A/V and Dep/V variables are 1.048 and .018 respectively for regulated firms 13 forma a m n o t ' ble f jndj"iQual mapufact"F1A8 industries--- ote that our measure of incentive compensation is likely to understate the dispersion in policy across industries. For example, the average percentage of manufacturing firms with a bonus plan is 94% while for regulated industries it is only 44%. Moreover, conditional on having a bonus plan, the median ratio of bonus payment to salary for manufacturing CEO's is 50%, but for regulated firms' CEO's it is only 23%. Thus both plan frequency and level of payment differ across these firms.

23 and 1.146 and .043 respectively for unregulated firms. While this evidence is consistent with the prediction that ceteris paribus regulated firms have lower equitylvalue ratios than unregulated firms, it must be treated with some scepticism. By their very nature, insurance firms and banks have high debt and low E/V ratios. Note that this result for banking and insurance could be related to the other two characteristics of the investment opportunity set that we argue in Section 2 are important, but for which we have no satisfactory proxy-the liquidity of the secondary market for the firm's assets and the sensitivity of the future asset value to use and maintenance decisions. Banks and insurance companies invest primarily in loans and securities. Hence, they should represent extreme observations in both dimensions. Moreover, their tax treatment of municipal bond interest also argues for high debt in their capital structures. Our evidence does not allow us to distinguish between the incentive and tax effects. We also predict in Section 2 that regulated firms have higher dividend payout rates than unregulated firms. In Table 4 the DIP ratio is .078 for regulated firms and .035 for unregulated firms. The t statistic for the difference is 3.6 and is significant at reasonable levels.

As for the E/V difference, this difference is not explained by asset in place

differences since the regulated firms in this sample have lower assets in place than unregulated firms. As predicted, regulated firms have lower compensation and use incentive plans less than unregulated firms. S730 is 170 for regulated firms and 239 for unregulated firms. The t statistic for the difference is 3.0 and is significant at standard probability levels. Forty-four percent of regulated firms have a bonus plan compared to 91 percent of unregulated firms. The t statistic for the difference is 3.8. The other common incentive plan, the stock option plan, is employed by 30 percent of regulated firms and 76 percent of unregulated firms. The t statistic for that difference is 7.2. As with the differences in E/V and DIP, these compensation level and incentive plan use differences between regulated and unregulated firms are not explained by asset in place

\

24

'.

differences. It could be argued that the compensation and inceptive plan differences aq Que to regulatory commissiops restriering compensation and inceptive plan usage

djrectl~

However, discriminating between that hypothesis and our hypothesis requires evidence on which commissions do, and which commissions do not restrict compensation and incentive plan usage.

4. EVIDENCE The correlations between the investment opportunity set variables and policy variables and among the investment and policy variables themselves are presented in Tables 5,6, 7 and 8. Table 5 gives the correlations using available data for all industries. Table 6 gives the correlations using available data for all unregulated industries. Tables 7 and 8 provide correlations for all industries and unregulated industries respectively when industries with substantial size differences are excluded. Simply separating the sample into regulated and unregulated firms is admittedly crude since regulation is not homogeneous. For example, the nature of regulation of utilities differs from that of banks and insurance companies. Utilities are subject to rate of return regulation on a firm-by-firm basis. Banks and insurance companies are required to obtain licenses, and they have some industry-wide pricing restrictions, but they are generally not restricted in allowed rates of return. Therefore, if utilities over-invest in physical capital (see Averch/Johnson, 1962) compared to banks and insurance companies because of differences in their regulation, then some of our tests will be less powerful because we do not account for these regulatory differences. There are some significant correlations between the samples' size differences and investment opportunity set and policy variables in the data underlying Tables 5 (all industries) and 6 (unregulated industries). In the all industries sample the difference between the compensation and Cowpustat samples' median firm sizes is significantly

25 positively correlated (at the .10 level) with equity to firm value ratio (E/V), depreciation to firm value ratio (Dep/V), return variance (VAR), research and development to firm value (R&D/V), use of long-term performance plans (LTP), use of stock option plans (SOP) and use of stock appreciation rights (SAR).

When the industries with substantial size

differences are discarded (i.e. the data underlying Table 7), none of the investment opportunity set and policy variables are significantly correlated with the size difference. In the unregulated industries sample (Table 6) the size difference is significantly positively co~~!~t~d

with R&D/V, the use of bonus plans (BO:t\'1JS) and LTP. When the substantial

size difference industries are discarded (Table 8), the size difference is significantly positively correlated with BONUS, VAR and SOP. Hence, discarding the industries with substantial size differences is effective for Table 7, but not for Table 8.

4.1 Correlation Among Investment Opportunity Set Variables All the correlations among the investment opportunity set variables have the signs we expect if those variables are surrogates for relative assets in place and investment opportunities. The correlation between the ratio of book to market, NV, and depreciation to market value, Dep/V is positive and significant at the .01 level in Tables 5 and 6 as one would expect if they both measure assets in place. In Table 7 that correlation is significant at the .10 level, In Table 8 it is positive but insignificant. Research and development, R&D/V, is intended to represent investment opportunities and therefore should be negatively correlated with both A/V and Dep/V, Since R&DN is not available for two of the three regulated industries, those correlations are only reported in Tables 6 and 8. Both correlations are negative in both tables, but only the correlation with Dep/V in Table 6 is significant at the .10 leveL However, R&D/V is also significantly negatively correlated with the capital expenditures/value and earnings/price ratios at the .10 level for unregulated industries. As indicated above, the capital expenditures/value ratio has correlations similar to those for A/V and is presumed to

26 measure assets in place. The earnings/price ratio is presumed to vary negatively with relative investment opportunities. While we expected variance of return (VAR) would vary with investment opportunities, in Tables 6 and 8 the only investment variable VAR is significantly correlated with is R&DN. That correlation is positive as expected and significant at the .05 and .01 levels respectively.

4.2 Correlation Between Investment Opportunity Set and Financing Policy We have already seen from Table 4 that regulated firms have lower equity/value ratios as predicted in Table 1. The correlations in Tables 5-8 tend to support the incentive­ based prediction in Table 1 for the relation between the investment opportunity set and financing policy (i.e., the larger the proportion of firm value represented by assets in place, the lower the firm's equity/value ratio). Three of the six relevant correlations reported in Tables 5 and 6 have the expected sign and are significant, two others have the right sign, but are insignificant and one (Dep/V and E/V in Table 5) has the wrong sign and is significant. The correlations have the same signs when the industries with substantial size differences are eliminated (Tables 7 and 8), but the only significant correlation is one consistent with our predictions (R&D/V with E/V in Table 8). Therefore either the tax­ based incentives identified by DeAngelolMasulis (1980) are insignificant, or they are significant but of an algebraically smaller magnitude so that our techniques do not identify them. The correlations with signs contrary to Table 1's predictions are between E/V and Dep/V and are for the samples that include regulated industries (see Tables 5 and 7). Also, the correlations between E/V and the ratio of book to market, A/V, while negative as expected, are much lower in absolute value in the two samples containing regulated industries (Tables 5 and 7). Both sets of correlations are less likely to have the expected negative sign in the regulated samples because the regulated industries have lower E/V

27 ratios (presumably partly due to regulation) and lower assets in place than the unregulated industries in our sample. Unregulated industries are not subject to this effect and therefore stronger correlations are expected for those industries. In Table 6, the correlations between E/V and the assets in place variables (A/V and Dep/V) are negative as expected and

significant at the .01 probability level. In Table 8 those correlations are not significant, but they are much more negative than in Table 7. The correlation between E/V and the investments variable R&D/V is positive as expected and significant at the .01 level in Table 6 and at the .10 level in Table 8.14 The only insignificant correlation between E/V and an investment opportunity set variable in Table 6 is the correlation between E/V and VAR. That correlation is also insignificant in Table 8. These results, in addition to the reported weak association between our measures of variance and the other investment opportunity set variables reduce the plausibility of an effect induced by progressivity of the tax code.

4.3 Correlation Between Investment Opportunity Set and Dividend Policy It is expected that the more the firm's assets in place (and the less its investment opportunities), the higher its dividends. The correlations reported in Tables 5 and 7 are inconsistent with that expectation. Although the correlations between the assets in place variables (A/V and Dep/V) and dividend yield, DIP, are insignificant, they are negative. However, that sample includes regulated industries that we argued have higher dividend payouts, ceteris paribus. Given that our sample of regulated industries has lower AN and Dep/V ratios than our sample of unregulated industries, this regulatory effect reduces the

14 The two investment opportunity set variables (other than regulation) not reported in Tables 5 and 6 are also correlated with E/V as expected. The correlation between the earnings price ratio and E/V is negative as expected and significant at the .01 level for all industries and for unregulated industries. The correlation between the capital expenditureslvalue ratio and E/V is negative as expected and significant at the .10 level for all industries and at the .05 level for unregulated industries.

28 likelihood of observing the predicted positive relation between dividends and assets in place. Unregulated firms are not affected by the regulatory incentive to pay dividends and therefore provide more powerful tests of the expected relations between dividend policy and the investment opportunity set. All the correlations between DIP and investment opportunity set variables reported in Tablse 6 and 8 for unregulated firms have the expected sign and five of the six are significant, In Table 6, the correlations with VAR and A/V are significant at the .10 level and those with R&D/V and Dep/V are significant at the .05 level. 15 In Table 8, the correlations with Dep/V and VAR are significant at the .05 level and the correlation with R&D/V is significant at the .10 probability level. The preceding results support the expected relation between dividend policy and the investment opportunity set. However, the earnings/dividend ratio (XID) is not significantly correlated with any investment set variable for any sample. This result is possibly due to XID measuring dividend policy with error.

4.4 Correlation Between Investment Opportunity Set and Compensation Policy As with the correlations between investment opportunity set variables and other policy variables, the correlations between investment opportunity set variables and ,~ensal.i9~_~level

and incentive plan l!~e are likely to be weaker tests of the expected

relations when the regulated industries are included in the estimation sample. Ceteris paribus, we expect the greater the firm's investment opportunities and the less its assets in place, the greater the compensation level and use of incentive plans. However, in our sample's case, regulation interferes with the expected relations. The regulated industries in our sample have lower assets in place measures than the unregulated industries, while,

15 Capital

expenditures/value and earnings/price ratios are also significantly correlated with

DIP as expected for the unregulated sample. Both correlations are positive and significant at the .10 level.

29 ceteris paribus, we expect regulated firms to pay less compensation and use incentive plans less.

There is no significant correlation between compensation level (S730) and investment opportunity set measures in Tables 5 or 7 (which include the regulated industries). However in Tables 6 and 8 (unregulated industries), all four correlations have the expected sign and three are significant. The correlation with R&D/V is positive as expected in both tables and significant at the .05 level in Table 6. The correlations with A/V and Dep/V are negative as expected in both tables and significant at the .10 and .05 levels respectively in Table 8. 16

There are six significant correlations between investment opportunity set variables and incentive plan use variables in Tables 5 and 7, and all six have the opposite sign to that expected. The correlations indicate the g~eatY""

rights (SAR) at the .05 and .10 levels respectively. Dep/V is significantly positively correlated with the use of bonus plans (BONUS) and stock option plans (SOP) at the .10 and .05 levels respectively in Table 5, and with LTP and SAR at the .05 and .10 levels respectively in Table 7. We believe that each of these incorrect predictions is due to regulated firms having lower measures of assets in place than unregulated firms. In Tables 6 and 8 (the unregulated sample), there are six significant correlations between investment set and incentive plan use variables and all have the expected sign. SOP is significantly positively correlated with both VAR and R&D/Vat the .05 level in 16 For the unregulated industries, both capital expenditures/value and earnings/price are negatively correlated with compensation level as expected. However, only the capital expenditures/value correlation is significant at the .10 level.

30 Table 6 and at the .01 level in Table 8. The use of restricted stock plans (RSP) is significantly negatively correlated with AN and Dep/V in Table 8. In addition, the capital expenditures/value ratio is significantly negatively correlated as expected with BONUS and at the .10 level. The tax-based explanations for use of incentive plans rests on differences in tax rates between the firm and its managers. Unfortunately, given the levels of management examined and their base salary compensation, there is little cross-sectional variation in tax rates for executives.

The most notable case where the effective tax rates on our

corporations is significantly lower is the case of banks and insurance companies. As noted by Miller/Scholes (1978) and Shelton (1983), banks and insurance companies can hold tax exempt municipal bonds and still fully deduct payments to their liability claim holders. Thus, both banks and insurance companies manage their tax exposure to keep their effective tax rates below the statutory maximum. Hence, the Miller/Scholes (1982) analysis suggests these two industries should have the strongest tax-related incentives to include incentive compensation provisions. Yet, as we have seen, they use such plans less than the other industries in our sample. Again, we cannot distinguish whether this tax­ related effect is insignificant, or significant but with a partial effect of algebraically smaller magnitude than the regulatory effect.

4.5 Correlation Among Policies We argue in Section 2.4 that equity/value ratios and dividend payout rates should be negatively correlated if the investment opportunity set determines both. This prediction is reinforced for Tables 5 and 7 by the associations between regulation and E/V, and dividend payout. The correlations between E/V and DIP are negative in all four tables, but only the Table 5 and 7 correlations are significant at the .10 level. EN is not significantly correlated with the earnings/dividend ratio for either sample.

31 Equity/value ratios and compensation levels are significantly positively correlated, as predicted, for both samples with and without the industries with substantial size differences. The correlations are significant at the .01 level in Tables 5-7 and at the .05 level in Table 8. In Section 2.4 it is predicted that equity/value ratios and the existence of formal incentive plans are positively correlated. The association between regulation and EN and incentive plans strengthens this prediction for Tables 5 and 7. The existence of four of the five plan types in Table 5 (three of the five in Table 7) is significantly positively correlated with E/V. When the regulation effect is avoided by using unregulated firms (Tables 6 and 8), the existence of only one plan (stock options) is correlated with E/V at the .10 level and that is in Table 6. That correlation is positive as expected. The prediction that dividend payout rates are negatively correlated with compensation levels is supported by the significant negative correlations in Tables 5 and 7 between DIP and S730. The correlation is significant at the .01 and .05 levels respectively. In Tables 6 and 8, where the impact of regulation is absent, the correlations are negative, but insignificant. The correlation between the earnings/dividend ratio and compensation level is not significant for either sample. Significant negative correlations (at the .10 level) between DIP and existence of four of the five types of incentive plans in Table 5 and two of the five in Table 7 are consistent with the predicted negative association between dividend payout and the existence of formal incentive compensation plans. However, for unregulated finns (fables 6 and 8), the only significant correlations are of the wrong sign (that with SOP in Table 6 and that with SAR in Table 8). The only significant correlation between the earnings/dividend ratio and the existence of incentive plans is for restricted stock plans for the overall sample and that correlation also has the wrong sign. As expected, e~stence.of

t~J~velofcQrnpensation is

generally positively correlated with the

formal incentive compensation plans. All 20 estimated correlations are

32 positive. In Tables 5 and 7, eight of the ten correlations are significant (the insignificant correlations are for restricted stock plans) and in Tables 6 and 8 four of the ten correlations are significant (those for long-term performance plans and stock options).

4.6 Summary of Evidence i) Financing policy In general the correlations are consistent with our expectations based on the assumption that financing policy is driven by the investment opportunity set. Most of the correlations between E/V and investment set variables are as predicted and many are significant when the industries with substantial size differences are included. When those industries are eliminated the correlations' signs are still generally as predicted, but only the correlation between E/V and R&D/V is significant. The correlations between E/V and compensation variables reinforce the conclusion that financing policy is associated with the investment opportinity set. The correlations with dividend policy variables are less supportive; the only significant correlations are for the samples including regulated firms and could be driven by regulatory incentives to pay dividends.

ii) Dividend policy

The evidence on the relation between dividend policy and the investment opportunity set is weaker than the evidence for financing policy and the investment set. The estimated correlation between DIP and each investment variable has the expected sign for both unregulated samples and all but one are significant.

However, the

earningsldividend ratio is not significantly correlated with any investment set variable. DIP is generally only significantly correlated with other policies' variables for the sample that includes regulated firms, suggesting the correlation is driven by regulatory incentives to pay dividends.

33

iii) Compensation policy Compensation level correlations are generally consistent with our predictions based on the investment opportunity set. Compensation level is correlated with investment variables as expected for the unregulated sample. It is also negatively correlated with financing policy as expected for both samples. However, the expected correlation with dividend policy is only found for DIP in the overall sample and that result could be due to regulatory incentives. The correlations between the use of incentive plans and investment set and other policy variables are less consistent with our expectations.

5.

CONCLUSIONS In general, we find significant associations between measures of the firm's

investment opportunity set and its choice of financing, dividend, and executive compensation policies as expected under contrasting explanations for those policies. Morever, the evidence suggests that the variation in policy choices is not being driven primarily by tax considerations; tax related variables alone have less explanatory power than investment opportunity set variables alone. There are several limitations of this initial analysis and thus there are a number of ~

' ......

-~'''. ",~--_•..

_",._,.

'