William R. Baber *'a, Surya N. Janakiraman b, Sok-Hyon Kang b ...... J., J. Brickley, and J. Coles, 1993, Stock-based incentive compensation and investment.
JOURNALOF ELSEVIER
Journal of Accounting and Economics 21 (1996) 297-318
Accounting &Economids
Investment opportunities and the structure of executive compensation William R. Baber *'a, Surya N. Janakiraman b, Sok-Hyon Kang b School of Business and Public Management, George Washington University, Washington, DC 20052, USA b Graduate School of Industrial Administration, Carnegie Mellon University, Pittsburgh, PA 15213, USA
a
(Received June 1995; final version received April 1996)
Abstract We extend the contracting paradigm advanced in Smith and Watts (1992) to consider cross-sectional associations between investment opportunities and the sensitivity of C E O compensation to performance measures. We predict stronger associations between compensation and performance for firms with greater investment opportunities. We also predict greater use of market-based, rather than accounting-based, performance indicators as a basis for incentive payments when investment opportunities are substantial components of firm value. Results for specifications of 1992 and 1993 changes in compensation paid to C E O s of 1,249 publicly-traded U.S. firms are consistent with these hypotheses. Key words: measures
Management
compensation;
Investment
opportunities;
Performance
J E L classification: J33; L2; M4
* Corresponding author. The authors appreciate comments and suggestions from Patricia Fairfield, Yuji Ijiri, Shyam Sunder, Richard Sweeney, Sam Tiras, Kumar Visvanathan, Teri Yohn, Jerry Zimmerman (the editor), and workshop participants at Carnegie Mellon University and Georgetown University. The authors are especially indebted to Ken Gaver (the referee) for valuable suggestions. The third author thanks Richard Cyert and Praveen Kumar for their suggestions and support in creating the CEO compensation database. The study is funded by GSIA and the CMU Faculty Development Fund. 0165-4101/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved SSDI 0 1 6 5 - 4 1 0 1 ( 9 6 ) 0 0 4 2 1 - 1
298
W.R. Baber et aL / Journal of Accounting and Economics 21 (1996) 297-318
1. Overview
The proposition that executive compensation and investment opportunity sets (lOS) are related is advanced in Smith and Watts (1992). The underlying premise is that corporate policies, including the structure of executive compensation, are determined by firm characteristics that govern contracting relationships among parties that comprise the firm. One such characteristic is the firm's investment opportunity set. Thus, Smith and Watts, and later Gaver and Gaver (1993) and Skinner (1993), investigate how corporate policies relate with measures that proxy for investment opportunity sets. In this study, we focus specifically on cross-sectional relations between investment opportunities and executive compensation. We argue that, when considered with analyses advanced in other studies (e.g., Lambert and Larcker, 1987; Jensen and Murphy, 1990), the Smith and Watts contracting characterization predicts relations between executive compensation and firm performance that are conditioned on the existence of investment opportunities. Thus, we go beyond prior studies to consider how the sensitivity of compensation-performance relations varies cross-sectionally with investment opportunities. Tests using 1992 and 1993 changes in executive compensation, computed as the sum of salary, bonus, stock options, stock grants, and other incentive compensation paid to Chief Executive Officers (CEOs) of 1,249 U.S. firms, indicate that the sensitivity of executive compensation to firm performance varies directly with investment opportunities. Further analysis reveals that the sensitivity of compensation to security returns, but not to accounting returns, also varies directly with investment opportunities. Both relations are consistent with our predictions based on the Smith and Watts contracting characterization of the firm. Two other features distinguish the study. First, our definition of executive compensation is more comprehensive than that in most prior cross-sectional studies. Data limitations restrict prior investigations either to relatively small samples or to analyses where compensation is computed using only salary plus cash bonuses (i.e., Murphy, 1985; Coughlin and Schmidt, 1985; Antle and Smith, 1986; Lambert and Larcker, 1987; Clinch, 1991). Investment opportunities are likely to be linked with the choice among executive compensation vehicles, and therefore, including all components of compensation specifically, noncash components with values that depend on security prices - is particularly germane to investigations of relations between compensation and investment opportunities. Requirements recently imposed by the U.S. Securities and Exchange Commission expand disclosures of executive compensation to permit reasonably accurate calculations of stock option values, restricted stock, and long-term incentive payments. Thus, we can compute, for a relatively large sample, more comprehensive measures of executive compensation than prior studies. We find that
W.R. Baber et al. / Journal o f Accounting and Economics 21 (1996) 2 9 ~ 3 1 8
299
values of noncash components are, on average, quite substantial, and more significantly, their cross-sectional variation is high relative to cash compensation components. Second, we investigate alternative ways to measure the firm's investment opportunity set, a concept that is relatively new, and thus not yet precisely delineated in the literature. Following G a v e r and Gaver (1993), we employ factor analysis to construct a single measure that we believe captures the investment opportunity set concept. We extend and validate their methodology, however, by demonstrating that factor scores are indeed related ex post to the firm's investment activities. Section 2 presents two hypotheses which become the focus of the data analysis. Section 3 describes the procedures used to construct the investment opportunity set measure and presents the regression specification used to investigate compensation-performance relations. Section 4 describes the data, the sample selection process, and the sample characteristics. The primary results of the study are reported and discussed in Section 5. Concluding remarks are in Section 6.
2. Determinants of executive compensation Predictions that executive compensation varies directly with firm performance follow from the standard agency model (Jensen and Meckling, 1976; Holmstrom, 1979). Since management's actions are unobservable, shareholders offer contracts based on observable performance indicators presumed to be correlated with management's actions. In general, the conditioning of incentive compensation on performance indicators increases as the ability to observe or monitor managers' actions decreases. This characterization is supported by empirical studies that document statistically significant positive associations between executive compensation and performance measures. Such relations are robust with respect to alternative samples and methodologies (e.g., Murphy, 1985; Coughlin and Schmidt, 1985; Lambert and Larcker, 1987; Sloan, 1993). 1 Associations between compensation and investment opportunities also follow from the agency model. In particular, Smith and Watts (1992), G a v e r and Gaver (1993), and Bizjak et al. (1993) argue that the management of investment
1Jensen and Murphy (1990), who find that CEO compensation changes by $3.25 for every $1,000 change in shareholder wealth, question whether such relations are economically substantive. The conclusion is that 'the empirical relation between the pay of top-level executives and firm performance, while positive and statisticallysignificant,is small... [where] ... incentive pay is expected to play an important role' (p. 227). Thus, the extent that performancecontingent compensation actually promotes shareholder interests is debatable.
300
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 2 9 ~ 3 1 8
opportunities is particularly difficult to monitor; and therefore, firms with substantial investment opportunities are more likely to link compensation to indicators of firm performance (Smith and Watts, 1992, p. 275). Thus, both Smith and Watts (1992) and Gaver and Gaver (1993) examine the existence of incentive compensation plans. In contrast, we focus on the sensitivity of compensation to firm performance measures. This approach is motivated by two observations. First, incentive plans typically include a portfolio of compensation vehicles such as a bonus plan, a stock (option) plan, and/or other long-term performance plans. Hence, focusing on one or two elements of incentive plans provides an incomplete picture of the incentive arrangement. Second, the existence of an incentive plan does not guarantee that the plan is actually used, or that the plan is important relative to the overall compensation package. We therefore consider the impact of the total incentive plan package in dollar terms and address compensation-to-performance sensitivity as a measure of the incentive strength of the total compensation package. Other things equal, the greater the sensitivity, the greater the reliance on incentive plans that relate compensation with performance. The following hypothesis applies.
Hypothesis 1: The sensitivity of compensation to performance varies directly with the relative abundance of investment opportunities. Next we consider the use of market-based, as opposed to accounting-based, performance indicators as a basis for setting executive compensation. The argument for accounting-based plans is that, because security prices are affected by factors beyond management's control, accounting information can be more informative with respect to management's actions (Gjesdal, 1981). Moreover, because accounting returns are the lower variance measure, their use as performance indicators promotes efficient risk-sharing among contracting parties (Sloan, 1993). Hence, there can be circumstances where accounting-based plans are preferred to market-based plans. On the other hand, management enjoys an element of discretion in choosing among financial reporting alternatives, and therefore, accounting numbers can be manipulated (Rosen, 1993, p. 199). Moreover, arbitrary accounting rules can distort accounting profits as meaningful indicators of economic returns (Fisher and McGowan, 1983) and, in the extreme, discourage positive net present value investments. For example, research and development spending reduces current-period accounting earnings yet generates substantial, albeit uncertain, returns in the future. Security returns, because they anticipate future cashflows and because they are likely to be invariant to accounting distortions, better reflect economic consequences of management actions. Hence, the case for using market-based performance measures is that all value-relevant actions, regardless of their short-run versus long-run consequences, are properly reflected in security returns.
W.R. Baber et al, / Journal o f Accounting and Economics 21 (1996) 297-318
301
In general, accounting returns are less informative with respect to management's actions when investment opportunities are a substantial portion of firm value (Smith and Watts, 1992; Gaver and Gaver, 1993; Skinner, 1993). Thus, if the management of investment opportunities is important, then we expect greater reliance on market-based compensation, which translates into stronger relations between compensation and stock performance.
Hypothesis 2: The relative sensitivity of compensation to security returns, versus accountin9 returns, varies directly with the relative abundance of investment opportunities. To summarize, Hypothesis 1 predicts a greater use of all incentive plans, market- or accounting-based plans, for firms with greater growth opportunities. In contrast, Hypothesis 2 is specific with respect to the relative importance of these performance measures. That is, Hypothesis 2 posits that the relative abundance of investment opportunities is associated with greater use of market-based, rather than accounting-based, performance measures.
3. Methodology 3.1. Measurin9 investment opportunities G a v e r and Gaver (1993) advance the use of factor analysis to gauge the extent that investment opportunities comprise firm value. Factor analysis condenses pairwise correlations between observable variables that are presumed to derive from a c o m m o n unobservable construct to obtain one or more measures, called factors, that capture variation c o m m o n to the observable variables. Investment opportunities can take alternative forms; thus, an appealing feature of this approach is that a variety of observable variables can be reduced to a single factor. Some variables used by Gaver and Gaver are frequently unavailable for m a n y firms, and therefore, a straightforward application of their approach substantially reduces the sample size. We therefore extend the G a v e r and Gaver approach in two ways. First, we use a set of variables that are more readily available and that are less likely to result in the omission of firms from the sample. Second, we use a test period to construct the factor scores, and then evaluate the efficacy of the factor scores by testing whether they are indeed related ex post to the firm's investment activities. We begin by posing a measure of a firm's investment activity, denoted investment intensity ( I N V I N T ) , as the sum of acquisitions ( C O M P U S T A T item # 129), research and development (item #46), and capital expenditures
302
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 29~318
(item # 30) deflated by depreciation expense (item # 14). 2 We consider c u r r e n t i n v e s t m e n t intensity to be the realization of prior i n v e s t m e n t opportunities. Thus, we seek a c o m m o n factor, d e n o t e d F A C ( I O S ) , that predicts i n v e s t m e n t intensity using i n f o r m a t i o n available at time t. F o u r criteria guide our choice of variables used to c o m p u t e F A C (IOS). First, we restrict o u r search to variables that can be b o t h clearly interpreted a n d c o m p u t e d using readily available data. Second, we prefer a relatively small set of variables to construct F A C (lOS). Third, we avoid variables that require lengthy time series to calculate. The second a n d third criteria are imposed to maximize the sample size a n d m i n i m i z e sample biases, which can result from the omission of o b s e r v a t i o n s from small, new, or poorly performing f i r m s ? Finally, we select variables that yield factor scores that predict future i n v e s t m e n t activity at least as well, a n d ideally better, t h a n i n v e s t m e n t o p p o r t u n i t y proxies used in prior studies. W e begin with a set of sixteen c a n d i d a t e variables that can be plausibly defended as proxies for i n v e s t m e n t opportunities. 4 Some of these are used in prior empirical studies, p r i m a r i l y in G a v e r a n d G a v e r (1993), b u t also in Smith a n d W a t t s (1992) a n d S k i n n e r (1993). W e s u p p l e m e n t these with measures of prior growth rates a n d prior I N V I N T . 5 We then assess the predictive ability of these variables a n d factor scores based on subsets of these variables using a eleven-year e s t i m a t i o n period 1978 1988. M o r e specifically, for each year 1978-88, we use stepwise regression to identify the most p a r s i m o n i o u s set of variables yielding the factor scores that best predict i n v e s t m e n t intensity over
/If depreciation expense approximates the investment required to sustain existing assets, then a crude interpretation of I N V I N T is that measures greater than unity indicate that long-term investment exceeds the investment required to maintain existing assets. As a practical matter, current depreciation expense computed using historical costs likely understates the investment that is required to maintain existing assets, especially for firms that use straight-line, rather than accelerated, methods. Also, research and development spending, but not its related depreciation, is considered in the computation of IN VINT. Therefore, stable firms, or even firms that are contracting rather than growing, can achieve values that exceed unity. 3As examples, computations that require long time series (e.g., variance) cause the elimination of new firms; calculation that require positive earnings (e.g., E/P ratios) potentially bias the sample away from financially distressed firms. 4These include various measures of market-to-book assets, market-to-book equity, Tobin's Q, E/P, R&D/assets, R&D/sales, net property, plant, and equipment/market value of assets, variance of return on total market value, investment/revenue, and the variability of earnings per share. 51n particular, we use growth of revenue, operating income, the book value of assets, capital expenditure, the market value of equity, and the market value of assets. The use of past I N V I N T and growth rates to estimate investment opportunities assumes that current and future investments are related. Empirically, we observe substantial positive autocorrelation in investment intensity, which suggests that, in general, firms with higher current investment also have greater investment opportunities.
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 29~318
303
the next five years. 6 These investigations suggest FAC(IOS)i,t be constructed from four variables: (1) prior investment intensity (INVINT), computed for years t - 2 through t; (2) year t - 2 through year t geometric growth in the market value of assets (MVAGR); (3) the end of year t ratio of the market value to the b o o k value of assets (MKTBKASS); and (4) the year t ratio of research and development expense to the b o o k value of assets (R&D). Our assessment of FAC (IOS)i,t involves comparisons with a factor score suggested by Gaver and G a v e r (1993), designated FAC(GAVER)i,t. 7 Exhibit 1 indicates how variables that comprise both FAC(IOS) and FAC(GAVER) are computed. To demonstrate the efficacy of FAC(IOS)i,,, we display in Table 1 how FAC(IOS)~,, FAC(GAVER)~,t, and the individual variables that are used to compute these factor scores, correlate with year t + 1 through year t + 5 measures of realized investment intensity, and revenue and asset growth. 8 Notice that FAC(IOS) usually outperforms, or performs at least as well as, FAC(GAVER) or the individual variables that comprise these factors. 9 To illustrate differences between FAC (IOS) and FAC (GA VER), consider the three variables used to compute FAC(GAVER) but excluded from the FAC (lOS) computation. Two of these variables, EP and VAR, exhibit relatively low correlations with future investment. Also, negative EP values lack a straightforward interpretation and the computation of VAR imposes data requirements that causes the omission of relatively new firms. The third measure MKTBKEQ is highly correlated with MKTBKASS, and therefore, including both measures is to a large extent redundant. 1° Thus, FAC (lOS) replaces these three variables with two that are not considered in FAC(GAVER), prior investment intensity I N V I N T and prior growth in the market value of assets MVAGR. Observe that these two variables exhibit higher correlations with 6Details of these procedures are available on request.
7FAC(GAVER) is computed using five variables: (1) the year t ratio of research and development expense to the book value of assets (R&D); (2) the end of year t ratio of the market value to the book value of assets (MKTBKASS); (3) the end of year t ratio of the market value to the book value of equity (MKTBKEQ); (4) the end of year t earnings to price ratio (EP); and (5) the variance of the annual return on the market value of the firm up to year t (VAR). Time series of four to twenty observations are used to compute VAR. The actual measure used in Gaver and Gaver (1993) is constructed using these five variables plus a sixth variable, the n u m b e r of growth-oriented mutual funds holding the firm's shares (Gaver and Gaver, 1993, p. 134). Data availability and comments in Baker (1993, p. 162) discourage our use of this variable. 8Correlations are computed using all 1978 to 1988 firm-year observations on the 1993 C O M P U STAT primary, secondary, tertiary, and full-coverage files that satisfy the data requirements. Fiscal 1993 is the last year on the file, and thus, fiscal 1988 is the last year for which investment opportunity set measures can be compared with investment over the next five years.
9FAC(IOS) also typically outperforms FAC(GAVER) in each of the eleven years 1978-1988, although its relative performance varies. 1°Baker (1993) describes the advantages of using firm growth rather than equity growth.
304
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 297 318
1. Investment intensity (INVINTt) a i=t Y [Capitalexpenditure(30) + R& D expense(46) + Acquisitions(129)] i=t-2 1 i=t ]~ Depreciation(14) i-t-2 i 2. Geometric mean annual growth rate of market value of assets (MVAGR,) a ~ Market value of assets(6- 60 + 199x25), Market value of assets(6 - 60 + 199x25),., where n=max{1, 2, 3}, for which data is available. 3. Market-to-book value of assets (MKTBKASS0 a,b Market value of assets(6 - 60 + 199x25), Book value of total assets(6)~ 4. R&D expenditure to total assets (R&D0 a,b R&D expense(46) t Book value of total assets(6)t 5. Market-to-book value of equity (MKTBKEQt) b Market value of equity(199x25), Book value of equity(60), 6. Earnings-to-price ratio (EPt) b Earnings per share (58)t Price per share(199), 7. Variance of return on market value (VARt) b z/[h(Market value of equity) + Common & Preferred dividends(26x25+ 19)+ Interest expense(15)],/ o ~
[Total assets(6)-Common equity(60) + Market value ofequity(199x25)],_,
)
a. Variables used to calculate FAC(IOS). b. Variables used to calculate FAC(GAV). Exhibit 1. Definitions of variables used to calculate the investment o p p o r t u n i t y factor scores FAC(IOS) and FAC(GAVER); C O M P U S T A T data item n u m b e r s in parentheses.
measures of future investment. Finally, FAC(IOS) can be c o m p u t e d for, on average, 1,740 more observations per year (4,136 versus 2,396) than
FAC (GA VER). 11
11 The Pearson correlation between FAC (lOS) and the measure FAC(GA VER) proposed by the Gaver and Gaver study is 0.862. We find similar correlations for the 1978 1988 estimation period.
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 297-318
305
Table 1 M e a n a n n u a l P e a r s o n c o r r e l a t i o n s between time t i n v e s t m e n t o p p o r t u n i t y set m e a s u r e s [FA C (IOS), FAC(GAV)], a n d t + 1 to t + 5 i n v e s t m e n t a n d g r o w t h a (t = 1978 1988) Investment intensity
(1NVINT) b
lOS measure
FAC (lOS), FAC(GA VER)t INVINTt e MVAGRt f MKTBKASS~ g R&Dth MKTBKEQ~ i EP,j VAR, k
G r o w t h of b o o k value of assets c
0.3192 0.1721 0.2757 0.1014 0.1571 0.2620 0.1495 -- 0.0547 0.0699
0.1889 0.2112 0.0562 0.1850 0.2147 0.0477 0.1966 - 0.0397 0.0454
Revenue growth d 0.1562 0.1410 0.0932 0.1468 0.1384 0.0372 0.1299 - 0.0631 0.0455
a Entries are m e a n s of eleven 1978-1988 a n n u a l c o r r e l a t i o n s between the i n v e s t m e n t or g r o w t h v a r i a b l e a n d the possible I O S measure. FAC(IOS), is a c o m m o n factor e x t r a c t e d from I N V I N T t , M V A G R , M K T B K A S S , R&D,, a n d FAC(GAV)t is a c o m m o n factor e x t r a c t e d from MKTBKASSt, R&D, M K T B K E Q , EPt, V A R , Both factors are c o m p u t e d using o b s e r v a t i o n s a v a i l a b l e up to t. An o b s e r v a t i o n is included only w h e n b o t h FAC(IOS) a n d FAC(GAVER) can be c o m p u t e d . O b s e r v a t i o n s per year r a n g e between 1,925 a n d 2,488. bThe ratio of the sum of year t + 1 to t + 5 capital expenditure, R & D expense, a n d acquisitions to the sum of year t + 1 to t + 5 d e p r e c i a t i o n expense [ C O M P U S T A T d a t a items i=t+5 ~ =i =, t++,5 {(30) + (46) + (129)}/5~=,+ 1 (14)i].
c Year t + 1 to t + 5 g e o m e t r i c m e a n a n n u a l b o o k value of assets g r o w t h [ - C O M P U S T A T d a t a item 6]. a Year t + 1 to t + 5 g e o m e t r i c m e a n a n n u a l g r o w t h rate of revenues [ C O M P U S T A T d a t a item 12]. e The ratio of the sum of year t -- 2 to t capital expenditure, R & D expense, a n d a c q u i s i t i o n s to the sum of year t - 2 to t d e p r e c i a t i o n expense. f Y e a r t - 2 to t g e o m e t r i c m e a n g r o w t h of m a r k e t value of assets [ C O M P U S T A T (6 - 60 + 199 × 25)]. g M a r k e t value of assets to b o o k (6 - 60 + 199 × 25)/6].
value of assets for year t [ C O M P U S T A T
d a t a items data
items
h R & D e x p e n d i t u r e to t o t a l assets for year t [ - C O M P U S T A T d a t a items 46/6]; R & D is treated as zero if it is m i s s i n g w h e n sales revenues are n o t missing. M a r k e t - t o - b o o k value of e q u i t y for year t [ C O M P U S T A T J E a r n i n g s - t o - p r i c e r a t i o of year t [ C O M P U S T A T
d a t a items (199 × 25)/60].
d a t a items 58/199].
k V a r i a n c e of t o t a l r e t u r n on the m a r k e t value of the firm using all o b s e r v a t i o n s a v a i l a b l e u p to t. T o t a l return is defined as the sum of (1) price a p p r e c i a t i o n on the c o m m o n equity, (2) c o m m o n a n d preferred dividends, a n d (3) interest p a y m e n t s , all deflated by the b e g i n n i n g total value of the firm, defined as the s u m of the m a r k e t value of e q u i t y and b o o k value of d e b t [ C O M P U S T A T d a t a items {(199 x 25), -- (199 x 25), 1 + ( 2 6 x 2 5 ) , + 1 9 , + 1 5 t } / { 6 - 6 0 + 1 9 9 x 2 5 ) } , 1JF a c t o r s are n o t c o m p u t e d if one or m o r e of the required variables exhibit e x t r e m e values, defined as follows: I N V I N T > 100, M V A G R > 5, M K T B K A S S > 30, R&D > 1, M K T B K E Q > 30, EP < O, VAR > 10.
W.R. Baber et al. / Journal o f Accounting and Economics 21 (1996) 297 318
306
3.2. Specifications of CEO compensation The extant literature suggests various compensation-performance specifications. Our focus on compensation-to-performance sensitivity is similar to that in Jensen and Murphy (1990) and Kaplan (1994), and our interest on security return versus accounting return measures is similar to that in Lambert and Larcker (1987). The complete specification is
A Compi,t = flo + fll RETi,t q- f12 AROEi,t + flaFAC ( I O S ) i , t * RETi,t + fl4FAC(IOS)~,, * A ROE~,t + flsFAC(IOS)i,t + e~,t, where for firm i and t = fiscal year 1992 or 1993,
A Compi,t
= year t change in C E O compensation deflated by year t - 1 base salary, RETi,t = fiscal year t c o m m o n stock return, ROEi,t = year t return on the b o o k value of equity, FAC(IOS)~,t = factor score representing investment opportunities computed using data available at the beginning of fiscal year t, 12 flj = parameters to be estimated (j = 0, 1, 2, 3, 4, 5), ei,, = error term. The intuition behind the model is that C E O compensation increases when shareholder wealth increases, that is, when R E T > 0, or when accounting rate of return exceeds that for the prior year. The specification resembles that in Lambert and Larcker (1987), but it differs from theirs in two ways. First, because they conduct time-series analyses at the firm level, Lambert and Larcker use undeflated changes in compensation as the dependent variable. 13 In contrast, we address cross-sectional data, and therefore, we deflate year t change in executive compensation by the prior year's ( t - 1) base salary to consider potential effects of cross-sectional differences. ~4 Second, Lambert and Larcker consider compensation computed as the sum of salary and cash bonuses, whereas we consider not only changes in salary plus bonus but also changes in
12Using year t - 1, rather than year t, data to compute FAC(IOS) mitigates spurious relations that can result from using contemporaneous data. 13This feature is similar to Kaplan (1994). Other early studies of executive compensation, as examples, Jensen and Murphy (1990) and Clinch (1991), are not scaled for size. 14Tota 1 compensation includes both a fixed component and a performance contingent component, and therefore, using year t - 1 total compensation as the deflator potentially causes dCompl a to vary inversely with period t - 1 performance. If so, then base salary, presumed to be the fixed component of compensation, is the more appropriate deflator. Consistent with this reasoning, results when total t - 1 compensation is the deflator are somewhat weaker, but similar qualitatively to those which are reported later.
307
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 297 318
c o m m o n stock-related c o m p e n s a t i o n c o m p o n e n t s and in total c o m p e n s a t i o n defined as the sum of all forms of compensation. The interactions F A C (lOS)ia * R E T i , t and F A C (IOS)i,t * A ROEi,t, similar to those used in Clinch (1991), indicate h o w investment opportunities affect C E O compensation. F o r example, f13 > 0 (f14 > 0) implies that the sensitivity of c o m p e n s a t i o n to c o m m o n stock performance (changes in accounting rate of return) increases as investment opportunities increase. Thus, estimates of the parameters f13 and f14 are the principal focus of the investigation. Recall that Hypothesis 1 asserts that the sensitivity of c o m p e n s a t i o n to firm performance varies directly with investment opportunities. Thus, Hypothesis 1 is supported when both r3 and r4 are positive or when either r3 or r4 is positive and the other is zero, but not when both f13 and fig are zero. Because m a r k e t and accounting returns are scaled differently, the magnitude of f13 cannot be c o m p a r e d with the magnitude of f14. Hence, the evidence is inconclusive when either r3 or fig is negative. Next consider Hypothesis 2 which posits that the relative sensitivity of c o m p e n s a t i o n to security returns versus accounting returns varies directly with investment opportunities. This hypothesis is supported when f13 > 0 and r4 < 0. Again, because m a r k e t and accounting returns are scaled differently, if both f13 > 0 and r 4 > 0, then statistical comparisons of the relative impact of the alternative performance measures c a n n o t be made. Hence, only r3 > 0 and r4 = 0 are uniquely and definitively consistent with b o t h hypotheses, as F A C (IOS)i,t is included to control for the possibility that investment o p p o r tunities (or factors correlated with investment opportunities) are related to c o m p e n s a t i o n changes. Smith and Watts (1992, p. 274) predict that executive c o m p e n s a t i o n varies positively with levels of investment opportunities. Relations between investment opportunities and c o m p e n s a t i o n changes are not straightforward, however, and thus, we offer no expectation a b o u t the sign of the estimate of fls.
l SThe following table summarizes the implications with respect to each of the hypotheses for all possible outcomes. Observe that the results are consistent with both hypotheses if and only if f13 > 0 and 174= 0. Hypothesis 1
r3 < 0 r3 = 0 f13>O
Hypothesis 2
B, < 0
f14 = 0
/L > 0
/L < 0
/L = 0
/L > 0
X X ?
X X 0
? 0 0
X X 0
X X 0
X X ?
X results are not consistent with the hypothesis, 0 results are consistent with the hypothesis, ? results are ambiguous with respect to the hypothesis.
308
VKR. Baber et al. / Journal of Accounting and Economics 21 (1996) 29~318
Finally, observe that the model can be modified to consider alternative measures of market and accounting rates of return, as well as alternative benchmark rates of return used to reward CEOs. We exploit this feature in Section 5.1 to demonstrate that results are robust for alternative market and accounting return metrics and benchmark rates of return.
4. Data and sample selection Executive compensation is obtained from proxy statements which, since 1992, contain information that permits the decomposition of executive compensation into five components: salary, cash bonus, long-term incentive payout, and the value of stock options and restricted stock granted during the year.16 We do not include 'other compensation' - for example, automobile costs. Proxy statements are obtained through a mail request to 4,865 U.S. firms that are both publicly-traded on the NYSE, AMEX, or N A S D A Q exchange and listed on the 1993 C O M P U S T A T primary, secondary, tertiary, or full-coverage files. Table 2 indicates that 4,193 firms respond with proxy statements for either 1992 or 1993. Independent variables are specified as first differences, and therefore we use only 2,006 firms that provide proxy statements for two years. Similar reasoning causes us to omit 172 of these firms where the C E O serves less than two full fiscal years during 1991-1993. As in Gaver and Gaver, we also eliminate 345 firms in regulated industries, in particular, utilities (SIC 49) and banking (SIC 60). 17 Finally, FA C (lOS) scores cannot be computed for 240 of the remaining firms. Thus, 1,249 firms are considered in the data analysis. 18 For each of these firms, we compute FAC (lOS) for each of the years 1992 and 1993.19 Further investigation justifies the use of a single factor, rather than
16Option values are computed using the Black and Scholes (1973) approach adjusted to consider early exercise (see H e m m e r et al. 1994). Data required to execute calculations are typically disclosed, but for cases where details are omitted, vesting is assumed to occur two years after the grant, the exercise period is assumed to be the same as that for the most recent option where an exercise period is provided, and the strike price is assumed to equal the stock price at the end of the year when the options are granted. These procedures yield values which, on average, are about 25% less than those obtained from a straightforward application of the Black Scholes approach used in earlier studies (e.g., Noreen and Wolfson, 1981; Murphy, 1985; Jensen and Murphy, 1990). Results are comparable when option values are computed without the early exercise adjustment. v Results are comparable when the sample is restricted to firms with December 31 fiscal year firms, as in Gaver and Gaver (1993). 18Of the firms in the final sample, 579 lack the data required to compute FAC(GAVER). 19Recall that FAC(IOS)~,~ are computed using data available up to t - 1. Thus, financial data for 1991 and 1992, respectively, are used to compute measures for 1992 and 1993 analyses.
309
W.R. Baber et aL / Journal of Accounting and Economics 21 (1996) 29~318
Table 2 Summary of the sample selection procedures Selection criteria
No. of firms
Total requests
4,865
U.S. firms providing proxy reports Less firms providing only one proxy report (1992 or 1993) Firms providing both 1992 and 1993 proxy reports Firms eliminated: CEO did not serve two full years during 1991-1993 Utilities (SIC 49) and banks (SIC 60) Variables to calculate FAC(IOS) are missing
4,193 (2,187) (2,006)
Final sample
(172) (345) (240) 1,249
m u l t i p l e factors. 2° C o r r e l a t i o n s between F A C ( l O S ) a n d its composite variables are high, typically in the range 0.45 to 0.70. Also, 1992 a n d 1993 correlations are in line with those observed d u r i n g the 1978-1988 e s t i m a t i o n period. T a b l e 3 shows selected characteristics for all sample firms a n d for firms with F A C ( I O S ) greater, a n d firms with F A C ( I O S ) less, t h a n the m e d i a n score. C o m p a r i s o n s in panels A a n d B of the high- versus l o w - I O S groups generally are consistent with p o p u l a r characterizations of growth versus m a t u r e firms. As examples, h i g h - I O S firms are relatively small in terms of sales revenues a n d the b o o k values of assets, yet they exhibit high m a r k e t values of equity. 21 Also, m o r e likely t h a n not, h i g h - I O S firms are listed o n the N A S D A Q . P a n e l C indicates the i n d u s t r y d i s t r i b u t i o n of the samples. P h a r m a c e u t i c a l , c o m p u t e r (both h a r d w a r e a n d software), c o m m u n i c a t i o n s , electronics, a n d ins t r u m e n t s firms tend to be classified as h i g h - I O S firms. Oil a n d gas, construction, textiles, p r i n t i n g / p u b l i s h i n g , chemical, p r i m a r y a n d fabricated metals, a n d t r a n s p o r t a t i o n tend to be classified as l o w - I O S firms. This i n d u s t r y breakd o w n is consistent with c o n v e n t i o n a l perceptions of g r o w t h versus m a t u r e industries, z2 2°In particular, Harman (1976, p. 163) recommends retaining factors equal to the number of principal components with eigenvalues greater than one. Another criterion, used by Gaver and Gaver(pp. 137 139),is that the appropriate number offactors explains the starting communalities of the composite variables, which are computed as the squared multiple correlation of each variable with the remaining composite variables. In both 1992 and 1993, eigenvalues for the first principle component both exceed one and exceed the sum of the starting communalities. Analyses for each year 1978-1988 also suggest the use of a single factor. 21Smaller asset book values for high investment opportunity set firms differ from results in Gaver and Gaver (1993). : 2Industry distributions are comparable to those in Gaver and Gaver (1993),except that Gaver and Gaver classify food products firms and printing and publishing firms as growth firms.
310
PER. Baber et al. / Journal o f Accounting and Economics 21 (1996) 297-318
Table 3 Firm size (mean and [median]), exchange listing, and industry composition for 1,249 sample firms at the end of fiscal year 1993 A. Firm size ($ million)
Low-lOS firmsa
High-lOS firmsa
All firms
Significance test b
Total assets at 1993 fiscal year end
1,850 [204]
815 [ 154]
1,332 [ 171]
3.07** 3.72"*
Market value of equity at 1993 fiscal year end
996 [120]
1,225 [231]
1,110 [165]
0.89 5.37**
1,440 [244]
1,001 [ 191]
1,221 [205]
1.56 3.66* *
624
625
1,249
1993 revenue Total B. Exchanye listiny c
Low-IOS firms
High-lOS firms
NYSE
300 (48.0%)
186 (29.7%)
486 (36.4%)
NASDAQ
231 (37.1%)
383 (61.3%)
614 (52.5%)
AMEX
93 (14.9%)
56 (9.0%)
149 (11.9%)
624 (100.0%)
625 (100.0%)
1,249 (100.0%)
Total
All firms
a High-lOS (low-lOS) firms are those where 1993 FAC(IOS) is greater (smaller) than the sample median. FAC(IOS) for 1992 is used when FAC(IOS) for 1993 cannot be computed. b Significance tests for comparisons of high-lOS firms with low-lOS firms; the first (second) entry is the t-statistic from a parametric test (z-statistic from the Wilcoxon test). ' A Z2 test statistic rejects the hypothesis that the sample firms are equally distributed between exchanges at the 1% level. * Significant at the 5% level, two-tailed.
** Significant at the 1% level, two-tailed.
T a b l e 4 i n d i c a t e s fiscal 1993 d i s t r i b u t i o n s for t o t a l e x e c u t i v e c o m p e n s a t i o n a n d f o r c o m p e n s a t i o n c o m p o n e n t s . M e a n b a s e s a l a r y is $ 3 6 0 , 8 0 0 , a l m o s t 4 0 % of mean total compensation. Cash bonuses are awarded by 65.7% of the firms, a n d t h e r a t i o o f m e a n s s u g g e s t t h a t c a s h b o n u s e s ($189,700) a d d a b o u t 5 0 % t o the base salary. T h e d a t a i n d i c a t e t h a t t h e m e a n v a l u e o f o p t i o n s g r a n t e d , $ 3 0 9 , 6 0 0 , is l a r g e r than cash bonuses and about one-third of mean total compensation. Also,
W.R. Baber et a L / Journal of Accounting and Economics 21 (1996) 297-318
311
Table 3 (continued)
C. Industry composition
SIC code
Industry
100 ~ 800 10xx ~ 14xx 13xx 15xx ~ 17xx 20xx ~ 21xx 22xx ~ 23xx
Agricultural Products Mining Oil & Gas Construction Food Products Textile Mills/Apparel & Textile Products Lumber/Wood/Furniture Paper & Allied Printing & Publishing Pharmaceutical/Biological Chemical & Allied excl. 283x Rubber & Misc. Plastics Stone & Clay Iron & Steel Mills Primary Metal Fabricated Metal Industrial Machinery Computers & Office Machines Appliances Communication/Audio/Video Electronic C o m p o n en t s Transportation E q u i p me n t Instruments Misc. Manufacturing Transportation Services C o m m u n i c a t i o n Services Wholesale Trade Retail Trade Finance Services & Brokers Insurance Other Financial Services Services excl. C o m p u t e r & Software C o mp u t er Services & Prepackaged Software Entertainment Health, Medical, Education & Other Services
24xx ~ 26xx 27xx 283xx 28xx ~ 30xx ~ 32xx 331x ~ 333x ~ 34xx 351x ~ 357x 358x ~ 365x ~ 367x ~ 37xx 38xx 39xx 40xx ~ 48xx 50xx ~ 52xx ~ 61xx ~ 63xx 64xx ~ 70xx ~ 737x ~
25xx
29xx 31xx 332x 339x 356x 364x 366x 369x
47xx 51xx 59xx 62xx 67xx 76xx 738x
78xx ~ 79xx 86xx ~ 87xx
Total
L o w - lO S firms
High-lOS All firms a firms
No.
No.
3 8 39 18 19 15
6 2 15 8 19 12
9 10 54* 26* 38 27
0.72 0.80 4.32 2.08 3.04 2.16
11 16 27 0 31 20 6 12 11 31 30 4 18 7 19 18 22 10 18 14 42 58 20 4 7 16 14
11 8 11 48 27 11 2 5 1 8 27 40 15 20 23 18 66 4 13 13 29 47 12 6 12 12 44
22 24 38* 48* 58 31" 8 17 12 39* 57 44* 33 27* 42 36 88* 14 31" 27 71 105 32 10 19 28 58*
1.76 1.92 3.04 3.84 4.64 2.48 0.60 1.36 0.96 3.12 4.56 3.52 2.64 2.16 3.36 2.89 7.05 1.12 2.48 2.16 5.68 8.41 2.56 0.80 1.52 2.24 4.64
10 26
4 26
14 52
1.12 4.16
624
625
1,249
No.
%
100%
a An asterisk indicates that the p r o p o r t i o n of high-lOS firms is significantly different from that of low-lOS firms with less than 5% significance level. Test performed only for industries with greater than 20 firms.
312
W.R. Baber et al. / Journal o f Accounting and Economics 21 (1996) 297-318
Table 4 Profile of 1993 CEO compensation for the 1,249 sample firms Mean [Median] (standard deviation)" (1) Salary ($)
360,800 [300,000] (245,900)
(2) Cash bonus ($)
Firms awarding bonus in 1993
189,700 [62,800] (412,100) 65.7%
(3) Stock option b All observations
309,600 [18,200] (913,000)
Observations with stock options granted
Firms with an option plan Firms awarding option on 1993
587,200 [223,700] (1,191,100) 93.4% 52.7%
(4) Restricted stock ($)c 40,800
[0] (272,100) (5) Long-term incentive plan (LTIP) payouts ($)c
40,000
[0] (404,500) Total pay ($) [sum of (1) through (5)]
942,100 [498,300] (1,333,900)
a Entries rounded to the nearest hundred dollars. bOption values calculated using the Black Scholes model adjusted for early exercise as in Hemmer et al. (1994). c Information about the existence of restricted stock or long term incentive plans is not as complete as the bonus plan or the stock option plan from the proxy reports. In 1993, 9.2% of the sample firms awarded restricted stock and 5.8% awarded LTIP payouts.
I~R. Baber et al. / Journal of Accounting and Economics 21 (1996) 2 9 ~ 3 1 8
313
although almost all firms (93.4%) have stock option plans, only about one-half (52.7%) grant stock options during 1993. Finally, note that the cross-sectional variation in stock options granted is much greater than the variation in other compensation components.
5. Results
Table 5 summarizes results for specifications of 1992 and 1993 salary and cash bonus (column 4) and total CEO compensation (the last column) and individual compensation components (columns 1 through 3).23 Focusing first on the specification of total compensation (column 5), we find statistically significant estimates fll on the security return and on/33 its interaction with FAC (lOS), but not on the accounting return and its interaction (/32 and/34). Moreover, statistically significant relations cannot be attributed to the main effect of investment opportunities (or variables correlated with investment opportunities), as estimates for/35 are not statistically significant. Next, consider column 4 which shows estimates where compensation is restricted to salary plus cash bonuses, a metric commonly employed in prior studies. Notice that fll and/32, estimates for main effects of security returns and accounting returns, are positive and statistically significant. In contrast, the estimate/33 on the interaction of investment opportunities with security returns is positive as predicted, but not significant at the 5% level, and/34 on accounting return is not reliably different from zero. Thus, consistent with prior studies (e.g., Murphy, 1985), we find statistically significant positive relations between compensation and security returns (indicated by ill) both for total compensation and when compensation is defined as salary plus bonus. Relations for interactions differ according to how compensation is specified, however. In particular, restricting compensation to cash components implies that investment opportunities and the sensitivity of compensation to performance are unrelated, or at best weakly related. Results for total compensation suggest, however, that investment opportunities are significantly associated with greater sensitivity of compensation to stock performance. Specifications of compensation components, displayed in columns 1 3, provide additional insights. Estimates of neither f13 nor r4 on the interactions is statistically significant (at the 5% level) when changes in base salary or cash
23Observations are omitted when either independent or dependent variables are in the top or the bottom 0.5% of their respective distributions. To enhance comparability across specifications, we omit observations from all specifications that are considered outliers for any specification. Results are comparable when outliers are included. Results for each year 1992 and 1993 also are comparable.
314
W.R. Baber et al. / Journal of Accounting and Economics 21 (1996) 297-318
Table 5 Parameter estimates for regressions of base salary, cash bonus, all other incentive components, and total compensation on market and accounting performance measures, fiscal years 1992 and 1993 (standard errors in parentheses and t-statistics in italics) A (Base salary)
d (Cash bonus)
(1)
(2)
0.0337 (0.0029)
0.0090 (0.0092)
A (All other incentive pay)
(3)
d (Salary and bonus)
(4) [ = (1) + (2)]
(5) [ = (1) + (2) + (3)]
- 0.0295 (0.0481)
0.0427 (0.0091)
0.0132 (0.0493)
- 0.58
4.31
Dependent variable (A Compi) 8o Intercept
8l Stock return
11,33
0.97
0.0246 (0,0059)
0.2366 (0.0184)
0.4930 (0.0965)
0.2613 (0.1987)
12.81
5.10
13.15
4.13
8~ A ROE
- 0,0108 (0,0131 ) - 0.82
83 FAC(IOS)* Stock return
8,* FAC(IOS)*AROE
0.0030 (0.0070) 0.43
0.0181 (0.0116) 1.55
85 FAC(IOS)
0.0251 (0.0034) 7.18
Adjusted R 2
0.030
0.3037 (0.0408) 7.44
0.0310 (0.0217)
A(Total pay)
0.0377 (0.2134) 0.17
0.3089 (0.1138)
0.2929 (0.0439) 6.67
0.0340 (0.0243)
0.26
0.7544 (0.0990) 7.61
0.3306 (0.2190) 1.51
0.3430 (0.116)
1.42
2. 71
1.45
2.93
- 0.0545 (0.0361)
- 0.0205 (0.1888)
- 0.0364 (0.0388)
- 0.0570 (0.1938)
- 1.51
- 0.10
- 0.93
- 0.29
-- 0.0031 (0.0108) - 0.29
0.137
0.0200 (0.0566)
0.02196 (0.0116)
0.0419 (0.0580)
0.35
1.88
0.72
0.015
0.131
0.037
The regression model is: d Compl,t = flo + fll RETia + r2 dROEi,t + r3 F A C (IOS)i,t * RETi., + 84 FAC (lOS)ia * A ROEi.t + 85 F A C (lOS)i, + el,t, where RETt is fiscal year t stock return; A ROE is change in return on book equity (income before extraordinary itemst/total book equity0; FAC (lOS)t is factor score representing investment opportunity set (see Exhibit 1). Comp is defined as either Base salary, Cash bonus, All other incentive pay (value of awarded options, restricted stocks, and LTIP payouts in year t), Salary and bonus (cash salary and bonus), or Total pay (total compensation including salary, bonus, and all cash and noncash incentives including option grants, but excluding 'other' compensation). All dependent variables are deflated by prior year base salary, and all values are in 1993 dollars. Option values are adjusted for early exercise using the method suggested by Hemmer et al. (1994). Observations are included only when all five regressions can be estimated. Independent or dependent variables in the top or the bottom 0.5% of their respective distributions are designated as outliers and are omitted from the regression. The total sample size is 1,951.
IJLR. Baber et aL / Journal o f Accounting and Economics 21 (1996) 297-318
315
T a b l e 5 (continued) Pearson correlations
RET A ROE F A C (IOS) * R E T F A C (IOS) * A ROE FAC(IOS) ROE
0.307* -- 0.068* - 0.037 - 0.137" 0.256*
AROE
-- 0.066* 0.071 - 0.044 0.450*
F A C (lOS) *RET
0.133" 0.132" 0.077*
F A C (10S *AROE
FAC(IOS)
-- 0.018 0.059*
- 0.037*
* C o r r e l a t i o n s significant at ~ < 0.05.
bonus are the dependent variable (columns 1 and 2). If compensation is specified as changes in the noncash incentive compensation component (column 3), however, then the estimate f13 on the security return interaction, but not the estimate f14 on the accounting return interaction, is positive and statistically significant. Finally, the estimate fls for the main effect of investment opportunities is positive and statistically significant for salary (column 1), but not for other compensation components. These results suggest that salary increases, but not changes in incentive compensation, are greater for managers of firms with high levels of investment opportunities. As a whole, these results not only support both Hypotheses 1 and 2, but also they are consistent with prior evidence that high-growth firms are more likely to use incentive compensation than low-growth firms (Clinch, 1991; Smith and Watts, 1992; Gaver and Gaver, 1993). The results are also consistent with some, but not all, of the results in Bizjak et al. (1993). 24 The results also permit an observation about the relative importance of accounting returns as a determinant of executive compensation. In particular, associations for security returns (indicated by fl~) are statistically significant for all compensation components, but relations between compensation and accounting returns (indicated by f12) are significant only for cash bonuses (column 2), but not for other compensation components. Such results do not necessarily imply that accounting returns are unimportant for setting executive compensation. Recall from Table 4 that the cross-sectional variation in option-related compensation is high relative to that in salary and cash bonuses, which suggests that the variation in total compensation that can be attributed to accounting returns is dwarfed by the variation in the component that can be attributed to security returns. On the other hand, at least some of the variance in
24Bizjak et al. (1993) r e p o r t statistically significant n e g a t i v e r e l a t i o n s between the sensitivity of c o m p e n s a t i o n s to t o t a l C E O w e a l t h a n d r a t i o s of m a r k e t - t o - b o o k a n d of research a n d d e v e l o p m e n t expense-to-assets (p. 368). The a u t h o r s recognize these results to be i n c o n s i s t e n t with expectations.
316
W.R. Baber et al. /Journal of Accounting and Economics 21 (1996) 297-318
option-related compensation can be due to error in estimating stock option values. Thus, despite the apparent superiority of security returns as a determinant of total compensation, the results suggest a role for accounting returns as a basis for executive contracting.
5.1. Alternative specifications Sensitivity analysis establishes that the results in Table 5 are robust to alternative specifications and performance measures. First, we consider alternative measures of market and accounting performance. In particular, we specify market returns as (i) unexpected returns, computed as raw return less expected (CAPM-based) risk-adjusted return, (ii) market model residuals using both equally-weighted and value-weighted CRSP indices, and (iii) market-adjusted returns, computed as raw return less both equally-weighted and value-weighted CRSP indices. We specify accounting returns as (i) return on assets (ROA), (ii) ROE less expected ROE, and ROA less expected ROA, where expectations are computed from an AR(1) time-series model. Results for all specifications are consistent with those in Table 5. Second, following Lambert and Larcker (1987), security returns are specified as levels, whereas accounting returns are specified as changes. To evaluate whether the 'inconsistent' treatment of the competing return measures affects interpretations of the results, we re-estimate the regression specifying both market and accounting returns as levels. Similarly, we estimate specifications where the dependent variable is levels of (not changes in) compensation, and where all return measures are specified as levels. The results from these analyses are consistent with those in Table 5. Third, CEOs who are heavily invested in the firm already bear considerable firm-specific risk, and thus, conditioning compensation on security returns can be redundant in such circumstances (Jensen and Murphy, 1990; Cyert et al., 1995). This possibility suggests including the CEO's stake in the firm in the specifications reported in Table 5. Despite multicollinearity caused by the inclusion of these variables, we find that estimates for interactions between FAC (lOS) and the performance measures are comparable to those in Table 5. Fourth, some studies (Antle and Smith, 1986; Jensen and Murphy, 1990) specify compensation to include subsequent changes in the value of stock options or stock held by the CEO, a compensation component not considered in this study. Information available from 1992/93 proxy reports precludes meaningful estimates of changes in the values of stock options granted prior to 1992, but disclosures do permit estimates of changes in the values of CEO stock holdings. Thus, we consider specifications of CEO compensation to include gains or losses computed as the product of beginning-of-the-year value of CEO common stock-holdings and the common stock return during the year. Tstatistics for both estimates of/~x and/~3 are three to five times larger than those
W.R. Baber et al. / Journal o f Accounting and Economics 21 (1996) 2 9 ~ 3 1 8
317
reported, while estimates for accounting returns are not statistically significant. Such results again are consistent with Hypotheses 1 and 2. In particular, security returns realized during the year are used to compute compensation in these analyses, and therefore, we expect statistically significant estimates of/~1 for RETI,t by construction. More substantial relations between compensation and the interaction FAC(IOS)*RETia are not obvious, however. In this regard, these results corroborate the primary results. Finally, our interpretation of the evidence suggests that the relative use of performance-contingent versus fixed-wage remuneration varies directly with investment opportunities. We investigate this proposition by regressing the ratio of incentive compensation (specified as cash bonuses plus all option and stock-based compensation) deflated by current-period base salary on FAC(IOS). As expected, the estimate on FAC(IOS) is positive and highly significant, a result which corroborates Hypothesis 1, and more generally, the Smith and Watts (1992) and Gaver and Gaver (1993) characterization of executive contracts.
6. Concluding remarks The evidence generally supports both Hypotheses 1 and 2 which are the principal focus of the study. Such evidence is consistent with prior studies that report positive cross-sectional associations between investment opportunities and the use of incentive compensation plans. We extend these investigations to document two additional relations that support the Smith and Watts contracting characterization of the firm. First, we find positive cross-sectional relations between investment opportunities and the sensitivity of CEO compensation to firm performance. This contribution is important as our data indicate that the existence of an incentive compensation plan does not necessarily indicate that incentive compensation is paid (i.e., stock option plans). Second, we find that investment opportunities are associated with greater sensitivity to marketbased, as opposed to accounting-based, performance indicators. Related to this, we find that the results are driven primarily by relations for compensation other than cash salary and bonuses, which are the focus of most prior empirical investigations. The implication here is that noncash vehicles, primarily stock options, are an efficient response to contracting frictions that otherwise encourage executive actions inimical to shareholder interests.
References Antle, R. and A. Smith, 1986, An empirical investigation of the relative performance evaluation of corporate executives, Journal of Accounting Research 24, 1-39.
318
W.R. Baber et al. / Journal o f Accounting and Economics 21 (1996) 2 9 ~ 3 1 8
Baker, G., 1993, Growth, corporate policies, and the investment opportunity set, Journal of Accounting and Economics 16, 161 165. Bizjak, J., J. Brickley, and J. Coles, 1993, Stock-based incentive compensation and investment behavior, Journal of Accounting and Economics 16, 349 372. Black, F. and M. Scholes, 1973, The pricing of options and corporate liabilities, Journal of Political Economy 81, 637 654. Clinch, G., 1991, Employee compensation and firms' research and development activity, Journal of Accounting Research 29, 59-78. Coughlin, A. and R. Schmidt, 1985, Executive compensation, management turnover, and firm performance, Journal of Accounting and Economics 7, 42 66. Cyert, R., S. Kang, J. Kim, and P. Kumar, 1995, Agency and performance pay: Theoretical and empirical analysis, Working paper (Graduate School of Industrial Administration, Carnegie Mellon University, Pittsburgh, PA). Fisher, F. and J. McGowan, 1983, On the misuse of accounting rates of return to infer monopoly profits, American Economic Review 73, 82 97. Gaver, J. and K. Gaver, 1993, Additional evidence on the association between the investment opportunity set and corporate financing, dividend, and compensation policies, Journal of Accounting and Economics 16, 125-160. Gjesdal, F., 1981, Accounting for stewardship, Journal of Accounting Research 19, 208-231. Harman, H., 1976, Modern factor analysis, 3rd ed. (University of Chicago Press, Chicago, IL). Hemmer, T., S. Matsunaga, and T. Shevlin, 1994, Estimating the 'fair value' of employee stock options with expected early exercise, Accounting Horizons 8-4, 23-42. Holmstrom, B., 1979, Moral hazard and observability, Bell Journal of Economics 10, 74-91. Jensen, M. and W. Meckling, 1976, Theory of the firm: Managerial behavior, agency costs and ownership structure, Journal of Financial Economics 3, 305 360. Jensen, M. and K. Murphy, 1990, Performance pay and top-management incentives, Journal of Political Economy 98, 225-264. Kaplan, S., 1994, Top executive rewards and firm performance: A comparison of Japan and the United States, Journal of Political Economy 102, 510-546. Lambert, R. and D. Larcker, 1987, An analysis of the use of accounting and market measures of performance in executive compensation contracts, Journal of Accounting Research 25, 85-123. Murphy, K., 1985, Corporate performance and managerial remuneration, Journal of Accounting and Economics 7, 11-42. Noreen, E. and M. Wolfson, 1981, Equilibrium warrant pricing models and accounting for executive stock options, Journal of Accounting Research 19, 384-398. Rosen, S., 1992, Contracts and the market for executives, in: L. Wernin and H. Wijkander, eds., Contract economics (Basil Blackwell, Oxford) 181 211. Skinner, D., 1993, The investment opportunity set and accounting procedure choice: Preliminary evidence, Journal of Accounting and Economics 16, 407 445. Sloan, R., 1993, Accounting earnings and top executive compensation, Journal of Accounting and Economics 16, 55 100. Smith, C. and R. Watts, 1992, The investment opportunity set and corporate financing, dividend, and compensation policies, Journal of Financial Economics 32, 263 292.
