THE COSTS AND BENEFITS OF MANAGERIAL INCENTIVES …webuser.bus.umich.edu/westjd/Articles/Costs_and_Benefits_of... · THE COSTS AND BENEFITS OF MANAGERIAL INCENTIVES AND MONITORING

Strategic Management Journal, Vol. 15, 121-142 (1994)

THE COSTS AND BENEFITS OF MANAGERIALINCENTIVES AND MONITORING IN LARGE U.S.CORPORATIONS: WHEN IS MORE NOT BETTER?EDWARD J. ZAJAC and JAMES D. WESTPHALJ. L Kellogg Graduate Sctiool of Management, Northwestern University, Evanston,Illinois, U.S.A.

Recent research and public discourse on executive compensation and corporate governancesuggests a growing consensus that firms can and should increase their control over topmanagers by increasing the use of managerial incentives and monitoring by boards ofdirectors. This study departs from this consensus by offering an alternative perspective thatconsiders not only the benefits, but also the costs of both incentives and monitoring inlarge corporations. The study develops and tests a contingency cost/benefit perspective ongovernance decisions as resource allocation decisions, proposing how and why the observedlevels of managerial incentives and monitoring may vary across organizations and acrosstime. Specifically, the study suggests that: (1) firms that are more risky face greater costswhen using incentive compensation contracts for top managers, thus reducing the expectedlevel of incentive compensation use for such firms; (2) firms facing this problem of lowincentive compensation use can realize greater benefits from higher levels of boardmonitoring, and thus are likely to rely more on board monitoring; and (3) firms with morecomplex corporate strategies face higher costs in using board monitoring, and are thuslikely to rely less on board monitoring as a source of controlling top management behavior.The study also proposes that within this contingency perspective there may be diminishing'behavioral returns' to increases in monitoring and incentives. These hypotheses are testedusing extensive longitudinal data from over 400 of the largest U.S. corporations. Thesupportive findings suggest that maximal levels of incentives and monitoring are notnecessarily optimal, and that a firm's strategy may not only have significant product/marketimplications, but also corporate governance implications.

Research on the control of modern corporationshas increased dramatically in recent years, as therecent strategy and organizational research ontopics such as executive compensation andcorporate governance suggests (cf. Davis, 1991;Eisenhardt, 1989; Finkelstein and Hambrick,1988; Kerr and Kren, 1992). Much of thisresearch uses agency (Jensen and Meckling,1976), managerialist (Berle and Means, 1932;Marris, 1964) and/or political (Wade, O'Reilly,and Chandratat, 1990) perspectives as its basis

Key words: Corporate govemance, CEO compen-sation, boards of directors, agency theory

or point of departure. Interestingly, despitesuch differing disciplinary orientations amongresearchers in this area, there appears to be agrowing consensus that corporations can andshould increase their control over top managersby increasing the use of managerial incentivesand monitoring by boards of directors.

In some research-based and policy-directedanalyses of executive compensation, for example,firms that provide heavy incentive-based compen-sation for top executives are considered exemplarsto be imitated, while those firms relying less onincentive compensation are singled out forcriticism (Jensen and Murphy, 1990b). Similarly,in research-based and policy-directed discussions

CCC 0143-2095/94/100121-22© 1994 by John Wiley & Sons, Ltd.

122 E. J. Zajac and J. D. Westphal

of corporate governance (e.g., Lorsch andMaclver, 1989), firms whose boards are structuredin ways that suggest more vigilant monitoring oftop management behavior are generally viewedas having superior internal control practices thanfirms having lower levels of monitoring. Notsurprisingly, given the apparent benefits ofgreater incentive compensation and board moni-toring, any observed variation across firms in thelevel of incentive compensation and monitoringis typically presumed to be due to politicalactions by influential top managers.

However, this study suggests that the emergingconsensus for greater use of incentive compen-sation and board monitoring may be premature.In particular, while advocates are typicallyexplicit in detailing the potential benefits of suchpractices, there is less willingness to considerexplicitly the potential costs of such practices forfirms. This study seeks to redress this imbalanceby suggesting that (1) there are indeed costsassociated with the heavy use of incentivecompensation and board monitoring, (2) thesecosts (and benefits) vary predictably acrossorganizations and across time as a function ofseveral individual and organizational contingen-cies, (3) there may be diminishing 'behavioralretums' to increases in incentives and monitoring(stemming from reduced internal and externalsalience of such increases), and (4) these contin-gencies and diminishing returns can explain, inpart, the variation in the observed use of incentivecompensation and board monitoring in moderncorporations.

There are very few efforts at examiningempirically the notion that there are costs toincentives or monitoring. One recent (partial)exception is Beatty and Zajac (1994), who drawfrom agency theory to explore the costs of usingincentive compensation, tjsing data from a largesample of firms about to go public, they findevidence consistent with their argument thatmore risky firms face greater difficulties (and thusgreater costs) when using incentive compensationcontracts, given the risk aversion of top managers,and as a result are less likely to emphasizeincentive compensation. They also find that firmsgenerally respond to this problem of inadequateincentive compensation by structuring boards ofdirectors to provide greater levels of monitoring.

The present study also seeks to explainvariation in firms' use of incentive compensation

and monitoring. However, in addition to consider-ing how differences in firms' riskiness may affectthe cost of using incentive compensation for topmanagers, we also propose how additionalorganizational contingencies, such as the differen-tial complexity involved in the pursuit of specificcorporate strategies, may affect the costliness ofusing board monitoring to control top manage-ment behavior.^ Finally, we also propose thatwithin this contingency perspective the benefitsof incentives and monitoring are essentially afunction of their behavioral salience—both tothe manager and external stakeholders—and thatthis salience may be realized at relatively lowlevels. Thus, we propose that there may beconsiderable diminishing 'behavioral returns' toincreases in monitoring and incentives, suchthat the contingent relationships may be morelogarithmic than linear. Empirically, we examinethe proposed contingency relationships usinglongitudinal data over 5 years from a sample ofover 400 of the largest U.S. corporations. Byseeking to explain and predict why incentivecompensation contracts and board-monitoringstructures differ across organizations (rather thanthe typical research approach of treating themas exogenous), the study hopes to contributeto the growing literature on top managementcompensation, ownership, and corporate govern-ance.

COSTS AND BENEFITS OF USINGINCENTIVES TO CONTROL TOPMANAGEMENT

Managerialist theory has long been fundamentallyconcerned with the separation of ownership andcontrol of large U.S. corporations (Berle andMeans, 1932). In this literature, top managersare viewed as essentially running corporationsfor their own benefit, rather than the benefit ofstockholding owners. It is therefore not surprisingthat early formulations of agency theory (e.g.,Jensen and Meckling, 1976) originally definedthe magnitude of the agency problem in terms

' Beatty and Zajac (1994) only emphasized how the cost ofincentives differed across firms (i.e., cross-sectionally), andfocused on IPO firms, which are typically smaller andyounger than the firms studied here. They also did notexamine the possibility of diminishing retums to incentivesand monitoring.

Costs of Incentives and Monitoring 123

of the degree of separation between owner andmanager interests. However, subsequent clarifi-cations suggest that the use of incentive compen-sation to align owner and manager interests ispotentially a double-edged sword. Specifically,linking a manager's compensation too closely tofirm wealth might lead to risk-avoiding behavioron the part of the agent (the point where a linkagebecomes 'too close' remains unclear, however).^

This argument, summarized in Holmstrom (1987)and in other normative principal-agent research,stresses the fact that while contingent compensationmay seem to have desirable incentive/motivationalproperties (relative to noncontingent forms ofcompensation), it also can cause a manager tobear risk that could be more efficiently bome bydiversified stockholders. The manager, unlike theowners, has already invested most of his/hernondiversifiable and nontradeable human capitalin the firm; in this way, the agent can be treatedas risk averse and the principal as risk neutral. Itfollows that agents would be reluctant to bear thisrisk of firm performance or, stated differently,that it is difficult and costly for the principal tohave the agent bear this risk (this argument formanagerial risk aversion does not require ownersto be fully risk neutral, but only that managers bemore risk averse than owners).

This suggests the need to (1) recognize explicitlythe potential costs, rather than just the benefits,of using incentives, and (2) start to identify theorganizational and individual contingencies thatcould affect the consideration of the incentivecost/benefit trade-off.^ We begin by focusing onone such organizational contingency, i.e., firmrisk, that may increase the riskiness of incentivecompensation contracts, and therefore may makemanagers particularly reluctant to accept suchcontracts. In other words, differences in firmrisk are associated with differences in the cost

^ This difference in emphasis can be seen in Jensen's (1983:334-335) suggestion that there are two 'almost entirelyseparate' agency literatures: a normative principal-agentliterature that emphasizes the design of compensationcontracts with optimal risk-sharing properties (Holmstrom,1979; Levinthal, 1988), and a positive, empirically based,agency literature on the separation of corporate ownershipand control (Fama and Jensen, 1983; Morck, Shleifer, andVishny, 1989; Weisbach, 1988).' Lambert and Larcker (1987: 85-86) also note that whilecompensation contracts are thought to be functions of'characteristics of the manager, the firm, and the environ-ment,' most compensation studies can be criticized for takingthe form of compensation contracts as simply given.

of using incentive compensation, since it caninfluence managers' willingness to accept compen-sation contracts that include a risk-bearingcomponent (Beatty and Zajac, 1994).

Thus, from an agency perspective, managersin risky firms may be particularly reluctant toaccept substantial long-term incentives (LTIs)in their compensation contracts, and one canhypothesize a negative relationship between firmrisk and the use of long-term incentives. Wepropose further, however, that several behavioralfactors may affect the precise nature of thisrelationship. First, from a goal-setting perspective(cf. Locke et al., 1981), compensation in theform of long-term incentive grants are intendedto motivate CEOs by providing relatively specificperformance goals that can direct their attentionand effort. This suggests that the increasedsalience to a CEO from greater LTI compen-sation, and the resultant changes in his/her'cognitive' attention and motivation, may begreatest when changing from zero or a very smalllevel to a modest level of LTI compensation, ascompared to changing an equal amount from aninitially high level. In other words, there may bediminishing 'behavioral returns' insofar as suchcompensation increases cognitive motivation ata decreasing rate.

Second, the issue of salience suggests thatthere may also be diminishing 'behavioral retums'to increases in incentive compensation that aremore external than intemal in origin, derivingfrom social, as opposed to cognitive, factors.Specifically, from an institutional perspective,LTIs have become accepted and essentiallyexpected components of CEO compensationcontracts. In fact, Westphal and Zajac (1994)and Zajac and Westphal (1995) found evidencethat LTIs (and the agency logic supporting them)have become 'institutionalized' over time aslegitimate and socially expected mechanisms ofincentive alignment (Zucker, 1983). Thus, whileCEOs may generally prefer to avoid heavy long-term incentive compensation (particularly wherefirm risk is high), the increased external salienceof observing some form of CEO incentivecompensation may compel firms to provideminimal or 'token' long-term incentive grants asa symbolic gesture of commitment to shareholderinterests (Westphal and Zajac, 1994). Note thatfrom this perspective (Meyer and Rowan, 1977),the extemal salience (and hence the benefits) of


showing that the firm is conforming to normativecompensation practices can be achieved at lowerlevels of incentive compensation; thus, one cansay that there are diminishing behavioral returnsto increases in incentive compensation.

Taken together, the micro- and macro-behavioral arguments noted above suggest amodification of the negative linear relationshipbetween firm risk and the use of LTIs. Specifi-cally, as risk increases, the use of LTIs maydecrease at a diminishing rate. These argumentssuggest the following hypothesis:

Hypothesis I: Firm risk will be negatively andlogarithmically related to the use of LTIs; i.e.,as firm risk increases, the use of LTIs willdecrease, but at a diminishing rate.'*

Since the notion of financial incentives theoreticallyencompasses all financial rewards that are contin-gent on firm performance, the next two hypothesesextend the notion of incentives to include not onlycontingent compensation, such as executives' LTIs,but also the change in value of the managers'equity holdings, which of course also varies withfirm performance (Jensen and Murphy, 1990a).This represents a more complete calculation of amanager's income that is 'at risk,' i.e., directlyrelated to a firm's equity value.

From the risk-bearing perspective developedearlier, this suggests the need to test twoadditional hypotheses related to Hypothesis 1:namely, that CEOs in more risky firms will have(1) a lower level of equity holdings in their firms,and (2) a smaller proportion of noncash incentives(i.e., incentive compensation from LTIs pluschanges in equity value based on stock ownership)relative to their total wealth change (i.e., non-cash incentives + cash compensation). Moreover,the logarithmic relationship between risk andlong-term incentive compensation proposed inHypothesis 1 should also apply in these relation-ships. Thus, we hypothesize that:

Hypothesis 2: Firm risk will be negatively andlogarithmically related to CEO equity holdings;i.e., as firm risk increases, CEO equity holdingswill decrease, but at a diminishing rate.

Hypothesis 3: Firm risk will be negatively andlogarithmically related to the portion of a CEO'stotal wealth change comprised of noncashincentives; i.e., as firm risk increases, noncashcompensation as a proportion of total compen-sation will decrease at a diminishing rate.

The discussion thus far has treated compensationand wealth change from equity holdings asendogenous managerial choices. However, wealso consider the possibility that equity ownershipis more 'sticky' than compensation, and thustreat it as an independent variable, rather thanas a variable whose level is simultaneouslydetermined with the compensation variable.^Thus, the earlier discussion of risk-bearing impliesanother related hypothesis: namely, that themagnitude of the existing equity positions heldby CEOs may influence their willingness to acceptfurther risk-bearing (e.g., long-term incentivecompensation) in their compensation contracts.This argument is consistent with the previousdiscussion of risk-bearing and incentives, butwith an added emphasis on the specific differencesin risk exposure facing CEOs holding variedamounts of equity in their firms (Beatty andZajac, 1994). In other words, whereas the earlierhypotheses considered how an organizational,firm-specific contingency (i.e., firm risk) canaffect compensation contracts, the followinghypothesis considers an individual, CEO-specificcontingency (i.e., CEO equity holdings). Thehypothesis can be stated as follows:

Hypothesis 4: CEOs having larger equity stakesin their firms are likely to have less long-termincentive compensation in their compensationcontracts, ceteris paribus.*

* As will be shown, our analysis is not limited to LTIs. Also,the precise operationalization of LTIs is discussed in theMeasures section.

' While we recognize that it is possible to consider firm riskas ultimately an endogenous variable, we view risk as likelyto be more exogenous and stable over time than compensationcontracts, which can be easily and quickly modified.Empirically, we also address this issue by lagging ourmeasures of firm risk when predicting the use of incentivecompensation (see later discussion).' The same hypothesis is also suggested by another explanationthat is consistent with our study; namely, that firms whosetop managers have low equity positions face a more severeagency incentive problem, making it more likely that suchfirms will need to use LTIs in their executive compensationcontracts. Such an explanation considers only the incentivecomponent of the agency problem, rather than the incentiveand risk-bearing components addressed in the present study.


COSTS AND BENEFITS OF USINGMONITORING TO CONTROL TOPMANAGEMENT

Monitoring is a central concept in both thenormative and positive agency literatures (Jensen,1983). The former examines abstractly the role(and costs) of monitoring where optimal compen-sation contracts cannot be written, while the latterdiscusses the 'monitoring technology' observed inmodern corporations (e.g., Fama and Jensen's,1983, discussion of boards of directors). In fact,Jensen and Meckling's (1976: 312) originaldiscussion of agency costs begins with an analysisof the problems that arise, and of the value ofmonitoring, when a 100 percent owner/managerreduces his/her equity holdings to 95 percent.From this perspective, agency costs exist whereverthere is no single 100% owner/entrepreneurbearing the full cost of his/her actions, andthese costs increase as top management equityownership decreases. Thus, both literatures gen-erally stress the primacy of incentive contractingas a 'first best' solution to the agency problem,and emphasize that the optimal level of monitor-ing would be based on the magnitude of theincentive gap between principal and agent (Beattyand Zajac, 1994).

The question of what level of monitoringshould be provided is based, of course, on thepremise that monitoring has costs; otherwise,all rational firms would monitor maximally,irrespective of the strength of incentive compen-sation contracts or other factors. There is virtuallyno research, however, that attempts to distinguishhow the costs and benefits of board monitoringmay differ across organizations, and how suchdifferences may predict variation in the level ofmonitoring one would expect to observe. Wepropose that there are at least two organizationalcontingencies that can affect the expected levelof monitoring of top managers by boards ofdirectors: (1) the degree to which the firm usesincentive compensation; and (2) the complexityof a firm's corporate strategy. We discuss eachof these in turn.'

Incentives and monitoring

This organizational contingency follows directlyfrom the normative agency literature's discussionof monitoring, as noted briefly above. Specifi-cally, a higher level of monitoring by boards ofdirectors would be required in the case where amanager does not accept any compensation riskthat is tied to firm performance, relative to thecase where a manager's compensation is tied tothe performance of the firm. In other words, thebenefits of heavy monitoring are particularlyevident in situations where managerial incentivesare only weakly tied to firm performance. Thus,the desired level of monitoring is contingent onthe magnitude of the incentive aspect of theagency problem.

Firms may employ a number of governancedesign features to increase the level of monitoringof top management by the board of directors.Ône governance design issue involves the pro-portion of outside directors on the board. Fama(1980) and Fama and Jensen (1983) view outsidedirectors as professional referees and experts ininternal organizational control. Even if onedisputed this characterization of outside directorsand viewed them instead as often coopted bytop management (Wade et al., 1990), a heavyuse of insider directors still suggests relativelyweak monitoring. Since insiders are 'beholdento CEOs for their jobs' (Fredrickson, Hambrick,and Baumrin, 1988: 262) they may be more willingto accommodate CEO preferences regardingcompensation arrangements. Insider-dominatedboards imply problematic self-monitoring, andparticularly weak monitoring of the CEO, sincethe CEO is likely to be in a position to influencean inside director's career advancement withinthe firm. In general, agency theorists andadvocates of board reform typically assume thatoutside directors will be less conciliatory towardCEOs (Beatty and Zajac, 1994; Schellenger,Wood, and Tashakori, 1989).

' Clearly, these are not the only possible contingencies thatcan affect the cost or benefit of board monitoring inorganizations, nor are they necessarily the most relevant.Given the lack of prior research on tliis subject, however,there is no precedent upon which one can rank suchcontingencies.

* The degree to which boards of directors are effectivemonitors of top management has been the subject ofcontinued debate (cf. Fama, 1980; Fama and Jensen, 1983;Lorsch and Maclver, 1989; Mace, 1971). However, anassumption of fully effective board monitoring is not neededfor the present study, given that our interest is focused noton absolute levels of monitoring, but rather on explainingdifferences in firms' level of monitoring, as predicted bydifferences in the costliness of such monitoring.


A second mechanism for dealing with thismonitoring problem is to choose outside directorswho also hold equity interests in the firm.Outside directors who are also owners wouldseem likely to be more vigilant in their monitoringrole (Morck, Shleifer, and Vishny, 1988). Thelogic of vigilant monitoring based on ownershipinterest can be extended to a third mechanismfor monitoring: the presence of a shareholderwith large equity holdings who, for whateverreason, is not on the board (i.e., a 'blockholder')(Huddart, 1993). Such large-scale owners aremore likely to be keen monitors of managerialbehavior.

A fourth potential mechanism for ensuringadequate monitoring involves choosing whetherthe CEO should also serve as chairman of theboard of directors. The presence of an outsideboard chairman who is not also CEO canrepresent an additional monitor of managerialbehavior. In fact, recent calls for changes orreforms in corporate governance structure haveemphasized the importance of the increasedmonitoring afforded by the separation of CEOand chairman roles (Lorsch and Maclver, 1989).In summary, then, one would expect a negativerelationship between the magnitude of the agencyincentive problem (defined first as the level ofCEO equity ownership in the firm) and the levelof monitoring provided by the firm.

However, the nature of the negative relation-ship may be more complex. Specifically, as wasargued in the prior discussion of the diminishing'behavioral returns' to incentives, there may alsobe diminishing retums to increased levels ofmonitoring. For example, if the desired purposeof monitoring is to make the act of monitoringsalient for the CEO, rather than actually tomonitor CEO behavior, than the desired levelof monitoring is that which 'gets the CEO'sattention.' Firms with an already high level ofmonitoring are likely already to have mademonitoring salient to a CEO, and are thereforelikely to find further increases in monitoring lessbeneficial than firms with low or nonexistentmonitoring activity whose CEOs' cognitive aware-ness of monitoring is much lower.

In addition to this cognitive perspective, theremay be more sociopolitical and institutionalreasons why there are diminishing behavioralreturns to increases in monitoring. First, giventhat CEOs are likely to resist increases in

monitoring that reduce their autonomy anddiscretion, firms already monitoring a substantialamount may be disinclined to increase boardmonitoring further.^ Firms with very low levelsof monitoring (particularly those with lowerlevels of CEO incentive compensation), on theother hand, may feel socially obligated to showsome increase in board monitoring as a symbolicgesture of commitment to shareholder interests(Westphal and Zajac, 1994).

This argument implicitly assumes that monitor-ing structures, as well as incentive compensationarrangements, have symbolic value. Althoughlittle systematic empirical evidence exists regard-ing the 'institutional status' of various monitoringmechanisms, it would appear that shareholdersroutinely and somewhat reflexively advocate theaddition of outside directors to the board andseparation of the CEO and board chair positions(Business Week, 1994), among other changes. Ineffect, therefore, a substitution effect betweenincentives and monitoring exists in the domainof symbolic control (i.e., in terms of managingshareholder perceptions regarding the agencyproblem), as well as the domain of substantivecontrol (i.e., in terms of actually minimizingthe agency problem) (Pfeffer, 1981), and thissymbolic trade-off function is nonlinear.

In summary, then, the expected relationshipbetween the magnitude of the agency incentiveproblem (defined first as the level of CEO equityownership in the firm, cf. Jensen and Meckling,1976) and the level of monitoring can be phrasedin terms of the following hypotheses, eachcorresponding to the specific governance mechan-isms discussed above:

Hypothesis 5: CEO equity holdings in the firmare negatively and logarithmically related tothe level of board monitoring, as representedby

H5a: a larger percentage of outside directorsH5b: a larger percentage of outside director

stock ownershipH5c: a major nonboard member block-

holder

' In fact, Finkelstein and D'Aveni (1994) argue that reducingthe decision-making autonomy of CEOs by separating CEOand Board Chairman positions is 'a double-edged sword.'


H5d: separate CEO and Board Chairmanpositions.10

We can also extend the definition of the magnitudeof the agency incentive problem to include notonly CEO equity holdings but also the incentivecomponent of CEOs' compensation contracts.Consistent with the earlier discussion, managerswho have a lower proportion of their wealthchange (i.e., through long-term incentives andequity stakes) contingent on firm performancecreate a more severe incentive problem for firms,which may be addressed by increasing the level ofmonitoring. Also, as noted above, the relationshipmay be nonlinear. This suggests the followingrelated hypotheses:

Hypothesis 6: The proportion of CEO wealthchange contingent on noncash sources will benegatively and logarithmically related to levels ofmonitoring, as represented by

H6a: a larger percentage of outside directorsH6b: a larger percentage of outside director

stock ownershipH6c: a major nonboard member blockholderH6d: separate CEO and Board Chairman

positions.

Strategic complexity and monitoring

In an agency framework, the costs of monitoring canbe considered in terms of overcoming informationasymmetry. Agency problems typically emergebecause of the two fundamental conditions thatunderlie principal-agent relationships: goalincongruence and information asymmetry (Zajac,1990). Goal congruence is an assumed condition,without which the agency problem reduces to amore easily solvable contracting problem. Thesecond dimension, information asymmetry, is acritical variable in the principal-agent relationship,and has generated a substantial body of researchwithin the normative agency theory literature.

Information asymmetry refers to the fact that inthe typical principal-agent relationship the principalhas less information than the agent about (1) the

characteristics of the agent and (2) the decisionsmade and the actions taken by the agent. Thesetwo aspects of information asymmetry have beenlabeled formally in the agency theory literature asadverse selection and moral hazard, respectively.We focus (as does most of the governanceliterature) on the moral hazard problem, wherethe issue is whether boards of directors are ableto adequately monitor or control the actions anddecisions of self-interested CEOs."

The present study proposes that corporatestrategy can be an important contingency inpredicting differences in monitoring costs acrossorganizations. We suggest that some corporatestrategies are generally more complex than others,and tliat complex corporate strategies are costlierfor boards of directors to monitor. More specifi-cally, we view the information asymmetry problemthat boards of directors face as an increasingfunction of a firm's strategic complexity. Whilecorporate strategies can be categorized in manydifferent ways, one widely accepted characterizationof a firm's strategy is its level of corporatediversification. In terms of complexity, we suggestthat, ceteris paribus, boards of firms that are inonly one industry are faced with less complexitythan boards whose firms are active in multipleindustries. A board member's ability to understandmultiple industries well enough to adequatelymonitor a firm's top mnagement is likely to belimited, relative to the single business situation.Stated differently, the pool of director candidatespossessing a sufficiently broad and complex 'cogni-tive map' to intelligently monitor top managementis reduced in such firms (cf. Calori, Johnson, andSamin, 1994). and consequently the cost of finding,attracting and retaining sufficiently qualified direc-tors is higher.

However, boards of firms at the highest levelsof diversification, such as boards of conglomeratefirms, may actually face less complexity, giventhat such firms tend to emphasize relativelystraightforward financial oversight, in which differ-ences across industries are deemphasized(Williamson, 1975). This suggests that the relation-ship between degrees of diversification and degreesof strategic complexity (and therefore the cost ofmonitoring) may become negative at the highest

'" As will be discussed, analyses are run separately for eachmeasure of monitoring to assess whether the predictions ofmonitoring hinge on any single measure of monitoring, orare robust across measures.

" Zajac (1990) has examined adverse selection problems inCEO/board relationships, as well, in a discussion of CEOselection.


levels of diversification. In other words, there maybe a curvilinear relationship between the level ofdiversification and the costliness of monitoring,such that the most complex corporate strategy,and thus the most difficult and costly situation formonitoring, exists in those firms pursuing amoderate level of diversification, where top man-agement competes in multiple industries and seeksto capture the elusive 'synergies' across the divisionsactive in these different industries. This suggeststhe following hypothesis:

Hypothesis 7: Firms deviating from intermediatelevels of diversification are likely to have higherlevels of monitoring, as represented by:

H7a: a larger percentage of outside directorsH7b: a larger percentage of outside director-

ownersH7c: a major nonboard member blockholderH7d: separate CEO and Board Chairman

positions.

Resesarch in Security Prices (CRSP) were used tocalculate performance, risk and size measures.

Measures

Firm risk

The first proxy for firm risk is the standarddeviation of return on assets during the prior 10-year period ('accounting risk'). The second measurerepresents the standard deviation of the firm'sweekly stock market retums during the same 10-year period ('stock risk'). Finally, the third measureof firm risk is the commonly used debt-to-equityratio, calculated as total current and long-termdebt divided by total common equity (Altman,1968; Beaver, 1966). Analyses in this study willbe run separately for each of the three measuresof risk, thus showing whether the test of therelation between firm risk and the introduction ofincentives is robust across measures.

METHOD

Sample and data collection

The population for this study includes the largestU.S. industrial and service firms, as listed in the1987 Forbes and Fortune 500 indexes. Firms wereexcluded from the initial sample if proxy statementswere unavailable, or if adequate compensationinformation was not provided in the statements.Firms were also excluded if complete data onboard structure, diversification strategy, or firmperformance were unavailable, yielding a finalsample of 405 firms. T-tests revealed no significantdifferences in size (measured as sales and numberof employees), performance (measured as market-to-book value and return on equity),or risk(measured as stock risk, accounting risk, or debt-to-equity ratio) between the initial and finalsamples.

Data were collected for the years 1987-1991inclusive. Information on board structure, compen-sation and ownership was obtained from bothproxies and Standard & Poor's Register of Corpora-tions, Directors, and Executives. Data on CEOage were obtained from both Standard & Poor'sand the Dun and Bradstreet Reference Book ofCorporate Management. Data from Standard &Poor's COMPUSTAT service and the Center for

Managerial incentives

Three measures of CEO compensation and owner-ship are used in the analyses. First, a continuousmeasure of the proportion of CEO total compen-sation comprised of long-term incentive grants wascreated. As mentioned earlier, LTI compensationincludes grants of multiyear performance incentives,such as performance shares, performance units,and restricted stock (Jarrell, 1993). The measureis calculated as the value of CEO long-termincentives divided by total compensation (i.e.,cash compensation plus the value of long-termincentives). Valuation formulas used for thiscalculation are discussed in detail in the Appendix.

We also used the dollar value of LTIs inour second measure of managerial incentives.Specifically, we calculated the natural log of thepercentage of CEO total compensation derived fromnoncash sources. As discussed earlier, this capturesthe amount of a CEO's total income that is 'atrisk' (i.e., it includes the dollar value of LTIs, aswell as the gain/loss attributable to the shares ofstock held by the CEO). This variable thussummarizes the incentive components of CEOtotal compensation that are directly related toequity value. The measure is constructed as follows:(the value of CEO long-term incentives + thechange in value of stock held by the CEO for theyear)/(cash compensation + the value of CEO


long-term incentives -I- the change in value ofstock held by the CEO for the year).

While our including a CEO's wealth changeattributable to his/her stock holdings as acomponent of a CEO's total incentive compen-sation is consistent with other comprehensivecompensation studies (e.g., Jensen and Murphy,1990a), one could, strictly speaking, considersuch stock-based gains and losses as investmentincome. From an incentive/motivational stand-point, however, we believe that top managers'compensation and shareholding gains, when takentogether, represent a more complete calculationof a CEO's wealth that is 'at risk,' i.e., directlyrelated to a firm's equity value. Since there isno consensus on this issue of how narrowly vs.broadly incentives should be defined (cf. Lambertand Larcker, 1987: 93), it is important tonote that the present study assesses incentivecompensation both ways, i.e. excluding andincluding incentives from equity shares held(Hypotheses 1 and 3, respectively). Finally, wealso include the percentage of a firm's equity heldby the CEO as a third measure of managerialincentives.

Monitoring

Four monitoring variables are used in thisstudy, all of which are straightforward in theirmeasurement. The first monitoring measure isthe outsider ratio, which indicates the portion ofthe board composed of outside directors. Thesecond measure captures outside directors' stockownership, calculated as the percentage of com-mon stock owned by outside directors. The thirdmonitoring measure is the presence (1 = yes,0 = no) of a nondirector blockholder, calculatedas a dichotomous variable and coded as T if anondirector stockholder owns at least 5 percentof the common stock, and '0' otherwise. Thefourth monitoring measure, CEOIchairman split,captures the separation of the CEO and boardchair positions, and is a dichotomous variablecoded T if the CEO and board chair positionswere separate and '0' otherwise.

Corporate strategy

Diversification was calculated using Jacqueminand Berry's (1979) entropy measure. This meas-ure takes into account the number of segments

in which a firm operates, and the relativeimportance of each segment as a portion of totalsales. It is given by

where Pji is the revenue share of segment / andln(l/P,,) is the weight for each segment i, or thelogarithm of the inverse of its sales.

Control variables

Several control variables are also used throughoutthe empirical analyses. Since size may be relatedto our measures of compensation, ownership,and monitoring structures, we control for theeffects of size (i.e., log of total sales) on ourfindings. Also, we control for the age of theCEO. An older CEO might be expected tohave different preferences for contingent vs.noncontingent forms of compensation than wouldyounger executives, and may also be less sensitiveto discipline from the managerial labor market,thus presenting a more severe agency problem(Lewellen, Loderer, and Martin, 1987). We alsocontrol for industry effects to ensure that thecompensation and monitoring structures observedare not simply outcomes of industry practices ortraditions. This was achieved through theinclusion of dummy variables—each at the two-digit SIC code level—in all multivariate analyses(given the large number of variables included ineach analysis, the coefficients for industry dummyvariables will not be reported in Tables 2-(i). Inaddition, we also included prior firm performance,measured as the prior year's return on assets, asa control variable, given that poor prior firmperformance may lead firms to consider higherlevels of incentives and monitoring.

Finally, we included the prior year's value ofthe dependent variable in all analyses. As aresult, these models effectively analyze changein the relevant dependent variable as a functionof the covariates. Means, standard deviations andcorrelations for all variables are shown in Table 1.

Data analysis

Most hypotheses were tested using pooled cross-sectional time series regression (Sayrs, 1989). Thistechnique is suitable for analyzing continuousvariables in a data set composed of multiple


I8

II

I

IICA

U

a.

I

E2

o oI

00 O m

o d oI

d d d dI I

d d d d dI I

d d d d d d dI I I I I

I I I I I I

I I I I I

s s B

6I I I I

6666o6osI I I I I I

2 rt g 12 8 22dd><S

II

I I I I I


organizations observed at multiple time points.In order to correct for heteroskedasticity andautocorrelation, we used a two-stage generalizedleast-squares (GLS) model (Greene, 1993). Thisprocedure partitions error variance into threecomponents: random error in time, random errorin space (e.g., across organizations), and randomerror not unique to time or space, and usesthis information to derive efficient and unbiasedparameter estimates (Sayrs, 1989). Thus, the GLSestimator takes the form

B =

given

Y = X„,Bk + Un,, with M«, = V, + u« -I- e«,

where v, are random over time, Un are randomover cross-sections, and en, are random over space£md time, and S is the sum of the variances of thethree error components. All independent variableswere lagged by 1 year.

Where the dependent variables were dichot-omous (i.e., CEO/board chairman split and pres-ence of a nondirector blockholder) we used pooledlogit regression analysis. Ordinary least-squares(OLS) analysis is inappropriate where the depen-dent variable is categorical, because OLS assumesa linear additive model with normally distributederror terms, while the true probability model isnonlinear with binomially distributed errors(Aldrich and Nelson, 1984). To correct forpossible unspecified time effects as a source ofheteroskedastic variance, we also included year-specific dummy variables (unreported) for thesetwo models.

RESULTS

Risk-bearing and incentives

Tables 2-4 show the results relating to thehypothesized relationship between organizational(and managerial) risk and the use of managerialincentives. In Table 2, for each of the three riskproxies (debt-to-equity ratio and variation inmarket and accounting performance), GLSregressions are used to test the predicted inverseand logarithmic relation between the level offirm risk and the proportion of CEO totalcompensation comprised of long-term incentive

compensation. For each of the risk measures,the results in Table 2 show that as firm riskincreases, firms' use of contingent compensationin their CEO compensation contracts significantlydecreases (consistent with Hypothesis 1). More-over, this relationship is logarithmic, such thatcontingent compensation decreases at a decliningrate as risk increases. We estimated separateGLS regressions for each measure of risk toshow that the test of the relation between firmrisk and the use of incentives does not hinge onany single measure of firm risk.

Table 3 examines the relation between firmrisk and incentive compensation in greater detailby measuring incentive compensation in termsof the proportion of total managerial wealthchange (i.e., cash compensation + incentivecompensation from LTIs and change in equityvalue) derived from noncash sources (i.e.,changes in the value of LTIs and equityownership). Table 3 reveals that firm risk isnegatively and logarithmically related to theproportion of total managerial wealth changederived from incentives (consistent with Hypoth-esis 3).

Hypothesis 2, which argued that managerialrisk-bearing concerns would result in more riskyfirms having CEOs hold lower levels of equityholdings in their firms, is tested in Table 4. Theresults are consistent with Hypothesis 2, indicatingthat the higher the level of firm risk, the lowerthe level of CEO stock ownership. This resultholds for each of the three risk measures. Finally,the results for Hypothesis 4 are found in Table 2.Hypothesis 4 argued that the level of equityownership held by CEOs is inversely andlogarithmically related to the use of long-termincentive compensation. Again, the results areconsistent with Hypothesis 4 and the risk-bearingexplanation.

Thus, Tables 2, 3, and 4 provide evidence thatfirms with higher risk exhibit significantly lowerlevels of incentives in executive compensationcontracts and CEO ownership. In addition.Table 2 shows that firms whose C!EOs face higherlevels of manager-specific risk (based on thelevel of their current equity holdings in the firm)are likely to have significantly less risk in theirincentive compensation contracts. In summary,then, the consistent pattern of statistically signifi-cant results in Tables 2-4 suggests that thosefirms whose CEOs face substantial risk (due to


CA

is

c

IIBO

_o

"o•o

I8

IIi8

1o

o

a

11O o

.OW

u

8z

206

o

U

I

C/5

+

<

o

the

aoc

dict

ipr

e

Vi

Z

o

Ou

I?Sz

I

cs o >o00 n Qs

V3 S V>

C O--C

l2-2-2-

d d dI I I

00 rt toS "o r-ir-j M Modd

oddI I I

d d dI I I

888

o o oI I I

IS2 2§ S q SviS-2-2-

1.

§SBS§2-22-^ so o >o

Q o o o

II

•2•s

I

T3

IO

8 Md j2V u

rt'JSs so uV g.

q j«o 2V 5."̂ ^• aO c:^ 2


•c

I•g.

c

o

uBoc

IIII.B"aIOU

2

GO

2

aoa.u

.A

as.5o•3

a

1

aen

mD

1̂"̂p̂oa.-Q

" 1

Q

II

MP,

OUK<Z

<

gG

SS

UJ

"O ^H 00

d d d

In 5̂ vo

CA 9 (A

•C o-'C

III«5 Q <

VO 00 ..I•o .-̂ ô o oo d o

I I I

g • * IT) • *

_ d d d

i:::|888'C d d d

.S o o oc I I I

"a • * •

e ^ V ^t: d d d

S

> o o oMil

a

i

Iu

8 |d -.iV «

o ^V I."̂ e

q

fO lV


3

u

sUBOB

cru

sUBO

c/3

2

I

I<iw

o

+

2̂

5:

I

IIIW Q <

o o oI I I

o o o•-! R '-idd d

o o oI I I

8888


either the firm's riskiness or their CEO's levelof equity holdings in the firm) face greater costsfrom imposing more risk-bearing on managersthrough compensation contracts. Furthermore,the observed logarithmic relationships are consist-ent with the hypothesized 'diminishing behavioralreturns' from increases in incentive compen-sation.*^ In conclusion, the results suggest thatorganization-specific and manager-specific factorsaffect managers' willingness to accept contingentcompensation in a way that results in predictablydifferent incentive compensation contracts acrossfirms.

Incentives, strategy, and monitoring

The second set of results address the organiza-tional contingencies that can affect the costs andbenefits of using monitoring. The first contingencyexamines whether firms seek to address thedifferences in incentive compensation contractsdiscussed above by increasing the level ofmonitoring when CEOs do not bear substantialincentive compensation risk. The tests of Hypoth-eses 5a-5d and 6a-6d, all of which address thisrelationship, are shown in Tables 5 and 6,respectively. The results are quite robust insupport of the hypotheses.

For example. Table 5 shows that firms facinga weak alignment of owner and manager interests,i.e., firms whose CEOs hold smaller equitypositions in the firm, are more likely to havegreater monitoring, in the form of (1) a largerpercentage of outside directors on their boards(consistent with H5a), (2) a larger percentage ofdirector stock ownership (consistent with H5b),(3) a major nonboard member blockholder(consistent with H5c), and (4) a separate CEO/Board Chairman position (consistent with H5d).In each case, the greater the magnitude of theagency incentive problem, as defined by lowerlevels of managerial stock holdings, the greaterthe level of monitoring provided. Moreover,these relationships are logarithmic, showing that

'̂ Although we did not formally compare coefficients for thelogged and unlogged variables, the significance levels of thelogged variables were consistently greater than those of theunlogged variables. Moreover, examination of augmentedcomponent-plus-residual plots with cubic splines (Mallows,1986) for relationships between each risk measure and thevarious measures of incentive compensation also consistentlysuggested a logarithmic, rather than linear relationship.

increases in CEO equity holdings in the firmreduce the level of board monitoring at adecreasing rate.

When extending the definition of the magnitudeof the agency incentive problem to encompassthe total executive wealth change from LTIsand equity ownership, the inverse relationshipbetween the levels of incentives and monitoringis also found for three of the four models. Firmswith managers having a lower proportion of theirtotal managerial wealth change derived fromincentives are more likely to have (1) a largerpercentage of outside directors on their boards(consistent with H6a), (2) a major nonboardmember blockholder (consistent with H6c), and(3) a separate CEO/Board Chairman position(consistent with H6d). Moreover, this relationshipis logarithmic, such that increases in noncashcompensation diminish the level of board moni-toring at a decreasing rate.

Taken together, the findings reported inTables 5 and 6 suggest that firms facing a moresevere managerial incentive problem (due toCEOs having lower incentives from their compen-sation contracts and equity holdings) are generallymore likely to have governance structures thatprovide a higher level of monitoring of managerialbehavior. In addition, the inverse logarithmicrelationship between incentives and monitoringis consistent with the argument that firmsexperience declining behavioral (i.e., motiva-tional and symbolic) returns from increasedmonitoring.^-' Thus, the results suggest that thebenefits of monitoring are predictably differentacross organizations.

The second organization contingency hypothe-sized to affect the costs of monitoring was thecomplexity of a firm's corporate strategy, i.e.,diversification. Again, Tables 5 and 6 indicatestrong statistical support for this hypothesis.Table 5 shows that firms pursuing a more complexstrategy of greater diversification, in whichmonitoring is more costly, relied less on monitor-ing across all four measures (i.e., H7a-H7d)than those firms with a less complex strategy oflittle or no diversification. The hypothesized

" Again, the significance levels of the logged compensationvariables were consistently greater than effects of the unloggedvariables, and examination of augmented component-plus-residual plots with cubic splines for these relationshipsconsistently suggested a logarithmic, rather than linearrelationship.


ers

c^ 1

:ock

<O

S•oa

:rat

t/i

ao•aS

tC

ersi

T3

O

•cO.

ane•a9

orin

g

**

'SoE•H

.gO

"o

lev

Zo<uE5o

cDIV

E+

1

uu<

o

uOH

+

u9•<

•CE

O

+

\c/3s<CO

+

"S

ibO_OT3

"3acS

O Q?5!

•o

2 ̂ H

B >

oooooooo

OOOOOOOO

oo,oo,r4cs^-^^

o o o o o o o oI ^ I I

oooooooo

oooowiorjd

ts .o2 •=

ts -as I J

I8

I

I•a

•Ii

siSi

2o i

1 1 1 1


13u

O

o

u

^

Istc

d.&0B

00

I

BOB

'•5

u u

Ii•a SB 2

O'["S

u

H 13

O

i

u

o

o

II U

oosH W

g>OQ

I

5\ >nd

S

S ?

oooo-ô.— o

oopooppo

8 ^ ^ ^ ^ tSi iZri I A fc.»O O 00 Op f l lO

dddddddd

> o o o

p j p ^ddddwSdrid

o

1 s

I

S3

s

I

V «

V g.

o eo 55

138 E. J. Zajac and / . D. Westphal

curvilinearity (the diversification-squared term)was also found to be statistically significant: firmswith a corporate strategy of conglomerate-likediversification, which is less costly to monitor,also engaged in greater monitoring across allfour measures. '̂* The results in Table 6 (whichuse an alternative measure of incentivecompensation) are also, not surprisingly, consist-ent in their support for H7. In summary, then,the results support the notion that firms makeless use of board monitoring in situations wheresuch monitoring is more costly, i.e., wherecomplex strategies make the reduction of infor-mation asymmetry more difficult.

DISCUSSION

Overall, the empirical results provide strongstatistical support for the study's hypothesesregarding the potential costs of incentive compen-sation and board monitoring. In testing the firstset of hypotheses relating to variation in firms' useof incentive compensation, the results consistentlyshow an inverse relationship between levels offirm risk (measured three different ways) andthe degree to which incentive compensation forCEOs is used (also measured three differentways). This finding highlights how organizationalcontingencies, such as firm risk, can increasethe costs of using incentives in large U.S.corporations, and complements Beatty andZajac's (1994) findings for smaller firms engagedin an initial public offering.

Moreover, the negative relationship betweenCEO equity holdings and the use of LTIs suggeststhat CEO-specific contingencies can also influencethe costliness of using incentive compensation.Taken together, both sets of results support theinterpretation that top managers' willingness toaccept risky compensation (or stated differently,the costs to a firm in forcing incentive compen-sation on top managers) varies across firms andthat the firm-specific and CEO-specific factorsexamined here are important predictors inexplaining this variance. More generally, the

'•' Note that finding this hypothesized curvilinear effect alsoaddresses the potential criticism of reverse causality, i.e.,that lower levels of monitoring lead to higher diversification.Recall also that the present study uses a longitudinal designwith lagged independent variables.

findings support the paper's argument that moreincentive compensation is not always better andthat one can identify and predict the circumstancesunder which this is likely to be true.

Moreover, the observed, inverse logarithmicrelationships between firm risk and levels ofincentive compensation suggest that firms derivediminishing behavioral retums from increasingCEO incentives. These results are consistent firstwith the notion that the salience to a CEO (andthus changes in his or her 'cognitive' attentionand motivation) may be greatest when changingincentive compensation from zero or a very smalllevel to a modest level than when changing anequal amount from an initially large level to aneven larger level. Furthermore, there may alsobe diminishing behavioral returns of a moresocial origin. Specifically, to the extent that long-term incentives and other forms of incentivecompensation have become legitimate and sociallyexpected sources of incentive alignment(Westphal and Zajac, 1994), firms and CEOsmay feel socially obligated to have 'token' CEOincentive compensation, even where firm risk ishigh. In effect, the reputational benefits derivedfrom symbolically conforming to normative com-pensation practices outweigh the costs of havingthe CEO accept a relatively small degree ofcompensation risk (Meyer and Rowan, 1977).Conversely, the symbolic benefits derived fromincreasing incentives beyond some token levelare relatively small in comparison to the costlinessof forcing greater compensation risk upon aCEO.

One implication of the findings discussed thusfar is that the issue of how much risk aCEO should be expected to bear in his/hercompensation agreement is very much an openquestion. While formal models of agency recog-nize the incentive/risk-bearing trade-offs intheory, most compensation studies have, in fact,treated compensation arrangements as exogenous(Lambert and Larcker, 1987). Future compen-sation research might be well served by devotingmore attention to agents' preferences whenexamining incentive contracts (Beatty and Zajac,1994; Lambert, Larcker, and Verrecchia, 1991).In particular, empirical, organizationally basedresearch is needed to establish the firm-specificand CEO-specific contextual factors that canshape CEOs' willingness to accept contingentcompensation, and to then use this knowledge


about the expected variety iti actual competisatioticontracts to address the CEO pay-for-perform-ance question more clearly. Such an expectedvariety of compensation contracts across firmsmight explain, in part, the inability of cross-sectional empirical work (e.g., Jensen and Mur-phy, 1990a) to establish a uniform pay-for-performance relationship for CEOs.

Interestingly, as noted by Beatty and Zajac(1994), organization behavior research on topexecutive compensation has also generally placedgreater emphasis on the importance—from anincentive and control standpoint— of imposingstrong pay-for-performance linkages, rather thanthe possible disadvantages of imposing risk-bearing in managerial compensation contracts.This omission is somewhat surprising, giventhat the organization behavior literature oncompensation has historically recognized thatdifferent forms of compensation, such as pay-for-performance, vary in their attractivenessto individuals, and therefore vary in theirappropriateness as incentive-motivational tools(Lawler, 1971; Mahoney, 1979).̂ ^ Future researchmight also examine other potential costs resultingfrom the heavy use of CEO incentive compen-sation, such as its possibly detrimental effects onmotivation at lower levels of the organization,as might be suggested by a tournament perspectiveon managerial compensation in organizations(Lazear and Rosen, 1981; O'Reilly, Main, andCrystal, 1988).

in terms of the costs and benefits of monitoring,the results are again quite consistent with thehypotheses, showing that the levels of monitoringobserved are inversely related to the levels ofmanagerial incentives used and the complexityof an organization's strategy. These robustresults are obtained using multiple measures ofmonitoring and incentives, and a widely usedmeasure of corporate strategy. More generally,the observed empirical relationships support thepaper's argument that: (1) there are botli benefitsand costs to monitoring top management; (2)these costs/benefits will vary across organizations;(3) the level of monitoring observed will thereforediffer across firms; and (4) firm-specific contin-

" One possible explanation of this may be that organizationalresearch using agency theory has tended to draw fromthe positive, rather than the normative, agency literature(Eisenhardt, 1989).

gencies such as the degree of incentive compen-sation use and the complexity of a firm's strategycan explain the source of this variance.

In addition, the observed, inverse logarithmicrelationships between incentive compensationand monitoring suggest that firms experiencediminishing returns to increased levels of monitor-ing. For instance, the findings are consistentwith the interpretation that increasing boardmonitoring by some amount serves to 'get theCEO's attention,' and that such cognitive orattentional benefits are less important wheremonitoring is already high. In addition to thesemotivational considerations, there may be moresociopolitical and institutional factors contributingto diminishing behavioral retums from increasesin monitoring. Specifically, to the extent thatCEOs resist infringements upon their autonomyand discretion, boards may be loath to increasemonitoring further where it is already high.Where monitoring is slight or nonexistent, incontrast, firms may perceive extemal, socialpressures to increase monitoring as a symbolicaffirmation of their commitment to shareholderinterests (Westphal and Zajac, 1994).

For organizational and legal researchers inter-ested in establishing guidelines for structuringboards of directors to ensure maximum monitor-ing capability (cf. Lorsch and Maclver, 1989),an implication of this study's findings is thatsuch well-intended, focused attention may bemisplaced, given that firms' effective use ofmanagerial incentives can substantially lessen theneed for costly monitoring, and given evidencefor declining motivational benefits from increasedmonitoring. Future research could also examinethe possible interaction of monitoring and incen-tives, e.g., the role of monitoring in enhancingthe specific estimators chosen for incentivecontracts (cf. Stiglitz, 1975).

Also, an implication for strategy researchersis that there may be corporate governanceimplications that arise from decisions regardingthe choice of particular corporate strategies.More specifically, the study suggests that thepursuit of more complex corporate strategies isassociated with an increase in the cost ofmonitoring, and thus a reduction in the level ofmonitoring observed. While corporate strategydecisions are traditionally considered to be afunction of competitor and resource situations,the findings of this study imply that governance


costs or govemance capabilities should perhapsalso be included as a relevant consideration inthe decision process leading to the choice ofcorporate strategy.

Taken together, the findings are supportiveof our basic contingency argument: (1) usingincentive compensation contracts for top man-agers is more costly for more risky firms(stemming from the risk aversion of topmanagers); (2) firms facing this potential problemof inadequate incentives experience greater bene-fits from providing higher levels of monitoringby the board of directors; and (3) using monitoringalso has costs, particularly for firms with morecomplex corporate strategies (stemming fromincreased levels of information asymmetry associ-ated with such strategies). More generally,the study suggests the need for governanceresearchers to recognize certain inherent confiicts,trade-offs, and substitution possibilities betweendifferent control mechanisms (Beatty and Zajac,1994).

This need can be seen, for example, inresearch interested in the classic 'problem' of theseparation between ownership and control (Berleand Means, 1932; Morck etal., 1989). Specifically,the present study's results suggest that a narrowfocus on management stock ownership is anincomplete definition of the incentive problem(e.g., Morck et al., 1988), since (1) incentivecompensation contracts for top executives mightbe used to address the problem of the ownership/control separation, (2) increased monitoringcould be used to 'compensate' for the lack ofincentive contracts, and (3) the costliness ofdealing with the problem of risk aversion actuallyimplies the desirability of some separation ofownership and management. Future organiza-tional research could begin to address theseissues jointly, rather than in a piecemeal fashion.

In conclusion, the study suggests that whilethere currently may be some perceived benefitin suggesting that modern corporations sufferfrom insufficient incentive compensation andboard monitoring, there may be some very realcosts in prematurely reaching such a consensus.The present study suggests that future researchconsidering explicitly how strategic, organiza-tional, and individual contingencies can affectthe conflicts, trade-offs, and substitution possi-bilities in the use of incentives and monitoringmay have the greatest potential to develop

a more enduring understanding of importantcorporate governance questions. This study canbe considered a first step in this process byhighlighting that optimal levels of incentives andmonitoring are not likely to be maximal levels.

ACKNOWLEDGEMENTS

The helpful comments of Jerry Davis, EdLazear, Ranjay Gulati, Brian Uzzi, and seminarparticipants at Stanford University are greatlyappreciated in shaping this paper.

REFERENCES

Aldrich, J. H. and F. D. Nelson (1984). LinearProbability, Logit, and Probit Models. Sage, New-bury Park, CA.

Altman, E. I. (1968). 'Financial ratios, discriminantanalysis and prediction of corporate bankruptcy'.Journal of Finance, 23, pp. 589-610.

Beatty, R. P. and E. J. Zajac (1994). 'Top managementincentives, monitoring, and risk sharing: A studyof executive compensation, ownership and hoardstructure in initial public offerings'. AdministrativeScience Quarterly, 39, pp. 313-336.

Beaver, W. H. (1966). 'Financial ratios as predictorsof failure'. Journal of Accounting Research, 4,pp. 71-102.

Berle, A. and G. C. Means (1932). The ModernCorporation and Private Property. Macmillan, NewYork.

Black, F. and M. Scholes (1973). 'The pricing ofoptions and corporate liabilities', /ourna/ of PoliticalEconomy, 81, pp. 637-659.

Business Week (February 21 1994). 'How to handle aCEO', pp. 64-65.

Calori, R., G. Johnson and P. Samin (1994). 'CEOs'cognitive maps and the scope of the organization'.Strategic Mangement Journal, 15(6), pp. 437-457.

Davis, G. F. (1991). 'Agents without principles? Thespread of the poison pill through the intercorporatenetwork'. Administrative Science Quarterly, 36,pp. 583-613.

Eisenhardt, K. M. (1989). 'Agency theory: Anassessment and review'. Academy of ManagementReview, 14, pp. 57-74.

Fama, E. F. (1980). 'Agency problems and the theoryof the firm'. Journal of Political Economy, 88,pp. 288-307.

Fama, E. F. and M. C. Jensen (1983). 'Separation ofownership and control'. Journal of Law andEconomics, 26, pp. 301-325.

Finkelstein, S. and R. A. D'Aveni (1994). 'CEOduality as a double-edged sword: How boards ofdirectors balance entrenchment avoidance and unity


of command'. Academy of Management Journal,37, pp. 1079-1108.

Finkelstein, S. and D. C. Hambrick (1988). 'Chiefexecutive compensation: A synthesis and reconcili-ation'. Strategic Management Journal, 9(6),pp. 543-558.

Fredrickson, J. W., D. C. Hambrick and S. Baumrin(1988). 'A model of CEO dismissal'. Academy ofManagement Review, 13, pp. 255-270.

Greene, W. H. (1993). Econometric Analysis. Macmil-lan, New York.

Holmstrom, B. (1979). 'Moral hazard and observabili-ty'. Bell Journal of Economics, 14, pp. 74-91.

Holmstrom, B. (1987). 'Incentive compensation: Prac-tical design from a theory point of view'. In H. R.Nalbantian (ed.). Incentives, Cooperation, and RiskSharing: Economic and Psychological Perspectiveson Employment Contracts. Rowman & Littlefield,Totoya, NJ, pp. 176-185.

Huddart, S. (1993). 'The effect of a large shareholderon corporate value'. Management Science, 39,pp. 1407-1421.

Jacquemin, A. P. and C. H. Berry (1979). 'Entropymeasure of diversification and corporate growth'.Journal of Industrial Economics, 27, pp. 359-369.

Jarrell, G. A. (1993). 'An overview of the executivecompensation debate', /oarna/ of Applied CorporateFinance, 5, pp. 76-82.

Jensen, M. C. (1983). 'Organization theory andmethodology'. Accounting Review, 58, pp. 319-339.

Jensen, M. C. and W. H. Meckling (1976). 'Theoryof the firm: Managerial behavior, agency costs,and ownership structure'. Journal of FinancialEconomics, 3, pp. 305-350.

Jensen, M. C. and K. J. Murphy (1990a). 'Performancepay and top-management incentives'. Journal ofPolitical Economy, 98, pp. 225-264.

Jensen, M. C. and K. J. Murphy (1990b). 'CEOincentives—it's not how much you pay, but how'.Harvard Business Review, 90, pp. 138-153.

Kerr, J. and L. Kren (1992). 'Effect of relativedecision monitoring on chief executive compen-sation'. Academy of Management Journal, 35,pp. 370-397.

Lambert, R. A. and D. F. Larcker (1987). 'Ananalysis of the use of accounting and marketmeasures of performance in executive compensationcontracts'. Journal of Accounting Research, 25(Suppl.), pp. 85-125.

Lambert, R. A., D. F. Larcker and R. Verrecchia(1991). 'Portfolio considerations in valuing executivecompensation'. Journal of Accounting Research,29, pp. 129-149.

Lawler, E. E. (1971). Pay and Organizational Effective-ness: A Psychological View. McGraw-Hill, NewYork.

Lazear, E. P. and S. Rosen (1981). 'Rank-ordertournaments as optimum labor contracts'. Journalof Political Economy, 89, pp. 841-864.

Levinthal, D. (1988). 'A survey of agency models oforganizations'. Journal of Economic Behavior andOrganization, 9, pp. 153-185.

Lewellen, W., C. Loderer and K. Martin (1987).

'Executive compensation and executive incentiveproblems: An empirical analysis'. Journal ofAccounting and Economics, 9, pp. 287-310.

Locke, E. A., K. N. Shaw, L. M. Saari and G. P.Latham (1981). 'Goal setting and task performance:1969-1980', Psychological Bulletin, 90, pp. 125-152.

Lorsch, J. W. and E. Maclver (1989). Pawns orPotentates: The Reality of America's CorporateBoards. Harvard Business School Press, Boston,MA.

Mace, M. L. (1971). Directors: Myth and Reality.Harvard Business School Press, Boston, MA.

Mahoney, T. A. (1979). Compensation and RewardPractices. Irwin, Homewood, IL.

Mallows, C. L. (1986). 'Augmented partial residuals'.Journal of Technometrics, 28, pp. 313-319.

Marris, R. L. (1964). The Economic Theory ofManagerial Capitalism. Macmillan, London.

Meyer, J. W. and B. Rowan (1977). 'Institutionalorganizations: Formal structure as myth and cer-emony', American Journal of Sociology, 83,pp. 340-363.

Morck, R., A. Shleifer and R. W. Vishny (1988).'Management ownership and market evaluation'.Journal of Financial Economics, 20, pp. 293-315.

Morck, R., A. Shleifer and R. W. Vishny (1989).'Alternative mechanisms for corporate control',American Economic Review, 79, pp. 842-852.

Noreen, E. and M. Wolfson (1981). 'Equilibriumwarrant pricing models and accounting for executivestock options'. Journal of Accounting Research, 19,pp. 384-398.

O'Reilly, C. A., B. G. Main and G. S. Crystal (1988).'CEO compensation as tournament and socialcomparison'. Administrative Science Quarterly, 33,pp. 257-274.

Pfeffer, J. (1981). 'Management as symbolic action:The creation and maintenance of organizationalparadigms'. In L. L. Cummings and B. M. Staw(eds.). Research in Organizational Behavior, Vol. 3.JAI Press, Greenwich, CT, pp. 1-52.

Sayrs, L. W. (1989). Pooled Time Series Analysis.Sage, London.

Schellenger, M. H., D. D. Wood and A. Tashakori(1989). 'Board of director composition, shareholderwealth, and dividend policy'. Journal of Manage-ment, 15, pp. 457-467.

Stiglitz, J. E. (1975). 'Incentives, risk and information:Notes towards a theory of hierarchy'. Bell Journalof Economics, 6, pp. 552-579.

Wade, J. B., C. A. O'Reilly and I. Chandratat (1990).'Golden parachutes: CEOs and the exercise ofsocial influence'. Administrative Science Quarterly,35, pp. 587-603.

Weisbach, M. S. (1988). 'Outside directors and CEOturnover'. Journal of Financial Economics, 20,pp. 431-460.

Westphal, J. and E. J. Zajac (1994). 'Substance andsymbolism in CEOs' long-term incentive plans'.Administrative Science Quarterly, 39, pp. 367-390.

Williamson, O. E. (1975). Markets and Hierarchies:Analysis and Antitrust ImpUcations, a Study in theEconomics of Internal Organization. Free Press,New York.


Zajac, E. J. (1990). 'CEO selection, succession,compensation and firm perfonnance: A theoreticaiintegration and empirical analysis'. Strategic Man-agement Joumat, 11 (3), pp. 217-230.

Zajac, E. J. and J. Westphal (1995). 'The symbolicmanagement of CEO compensation: Alternative

APPENDIX: FORMULAS USED INCALCULATING COMPENSATIONCONTINGENCY

Stock options were valued using the following,simplified version of the Black-Scholes (1973)model (Kerr and Kren, 1992; Noreen andWolfson, 1981):

option value = price x shares x [exp(-dOA^(Z)- exp(-rt)N(Z - s t)]

where

pricesharesd

t

N

s

Z

the exercise price of an option;the number of shares granted;the average dividend yield over theprevious 5 years;the time to the expiration of theoption, either 5 or 10 years;the standard normal probability dis-tribution function;the risk-free interest rate, based on5- and 10-year average yields onU.S. government securities;the stock return variance for theprevious 5 years; and[r - d + sm) X (tis t).

When the expiration date was not specified, weassumed 10 years. The Black-Scholes model hasbeen criticized for employing several unrealisticand/or inaccurate assumptions. For instance,although options are nontransferable, the modelassumes that options are marketable and sold onthe date of grant. In addition, it is implicitlyassumed that executives value options in thesame way as shareholders, even though eachparty holds different risk preferences (Lambert etal., 1991). Nevertheless, despite these theoreticalshortcomings, the model has performed well atpredicting the prices of marketable warrants,which resemble stock options (Noreen andWolfson, 1981). Moreover, a widely acceptedalternative is not currently available.

logics and their determinants'. Administrative Sci-ence Quarterly, forthcoming.

Zucker, L. G. (1983). 'Organizations as institutions'.In S. B. Bacharach (ed.), Research in the Sociologyof Organizations. JAI Press, Greenwich, CT,pp. 1^2.

Performance units/cash were valued usingthe following formula, used by several largeconsulting firms in conducting surveys of execu-tive compensation (cf. Towers Perrin 1991 Com-pensation Data Bank):

value = price x shares x target x [l/((r + p +

where

price =

shares =target =

/z

the price at which shares/units weregranted;the number of shares/units granted;the target payout, expressed as por-tion of shares granted;the risk-free interest rate, based on5- and 10-year average yields onU.S. government securities;long term average equity premium(6%);forfeiture risk (3%); andlength of performance period.

Where adequate information was not providedin the proxy statement, we employed a progressiveseries of assumptions. For instance, if the dateof grant was specified but the grant price wasnot, we used the market price at date of grant;if neither the grant price nor the grant date wasprovided, we used the average annual marketprice. Moreover, where the target payout wasnot specified, we assumed 100 percent. Analogousassumptions were employed in valuing otherincentive vehicles. For instance, where the sharevalue was not specified, book units were valuedaccording to the book value per share on thedate of grant. The forfeiture risk is based uponempirical analysis or forfeiture among Fortune500 CEOs conducted by a large compensationconsulting firm.

In valuing performance shares and restrictedstock, we used a simplified adjustment factor: 1/(f). This approach implicitly sets the discountrate equal to expected stock returns (cf. TowersPerrin 1991 Compensation Data Bank).

Documents

THE COSTS AND BENEFITS OF MANAGERIAL INCENTIVES …webuser.bus.umich.edu/westjd/Articles/Costs_and_Benefits_of... · THE COSTS AND BENEFITS OF MANAGERIAL INCENTIVES AND MONITORING