The Search for the Best FinancialPerformance Measure

Jeffrey M. Bacidore, John A. Boquist, Todd T. Milbourn, and Anjan V. Thakor

Refined economic value added (REVA) provides an analytical framaoork forevaluating operating performance measures in the context of slmreholder valuecreation. Economic value added (EVA) performs quite well in terms of itscorrelation with shareholder value creation, but REVA is a theoretically superiormeasure for assessing whether a firm's operating performance is adequate from thestandpoint of compensating the firm's financiers for the risk to their capital. In thisarticle, comprehensive statistical analysis of both REVA and EVA is used toestimate their correlatio)! with and their ability to predict shareholder value creation.REVA statistically outperforms EVA in this regard. Moreover, the realized retumsfor the 1988-92 period for the top 25 REVA firms were higher than the realizedreturns for the top 25 EVA firms.

I n the 1980s, shareholder activism reachedunprecedented levels and led to increased pres-

sure on firms to maximize shareholder value con-sistently. For example. Time n:\agazine summarizedthis activism as "Angry investors closed out theDecade of Greed with demands that executive com-pensation should be tied to company performance"(Smolowe 1996). The basic idea is that if managersare offered compensation contracts that are tied toshareholder wealth changes, their incentives willbe better aligned with those of shareholders than isthe case for other types of contracts. In designingsuch contracts, however, an important issue iswhich measure of shareholder performance to usein designing the contract.

Tlie obvious metric for judging firm perfor-mance is the stock price itself (see, e.g., Jensen andMurphy 1990 and Miiboum 1996). Stock price, how-ever (or returns based on stock price), may not be anefficient contracting parameter because it is drivenby many factors beyond the control of the firm'sexecutives. Moreover, as one moves down the orga-nizational ranks, the inefficiencies of stock-basedcompensation as a means of aligning managerialinterests with those of stockholders become evenmore evident because managers at lower levels have

Jeffiey M. Bacidore is a doctoral candidate in finance atIndiana University. John A. Boquist is the Edward E.Edwards Professor of Finance at Indiana University.Todd T. Milbourn is an assistant professor of finance atLondon Business School. Anjan V. Thakor is theEdward ]. Prey Professor of Banking and Finance at theUniversity of Michigan.

even less impact on the stock price than the CEO.Tying managerial compensation to stock price mayimpose excessive risk on managers and may detractfrom the ability of such compensation to provideincentive for managers to maximize shareholderwealth.

Any financial performance measure used inmanagerial compensation, on the one hand, mustbe correlated highly with changes in shareholderwealth and, on the other, should not be subject toall of the randomness and "noise" inherent in afirm's stock price. This dichotomy is the fundamen-tal tension a good performance measure mustresolve. A recent example of a performance mea-sure that seeks to resolve this tension is economicvalue added (EVA). This measure, proposed byStern Stewart Management Services, creativelylinks the firm's accounting data to its stock marketperformance (Stewart 1991).

Before examining the correlation betweenshareholder wealth and a performance measure,one must first define the appropriate way to mea-sure changes in shareholder wealth. We contendthat shareholders are concerned with the abnormalreturn they earn in any period—that is, the returnthey eam in excess of what they expected to earn fora firm within a given systematic risk class. Wlienthis return is positive, shareholders have more thancovered their risk-adjusted opportunity cost of pro-viding their capital. Conversely, when this return isnegative, they have been inadequately compen-sated for risk." Given this relationship, a goodfinancial performance measure should correlatehighly with abnormal stock returns.

Financial Anaiysts Journai • May/June 1997 11

In this article, we present an empirical analysisof the ability of EVA to predict abnormal return andof the contemporaneous correlation between EVAand abnormal return. In addition, we define a newperformance measure, a refinement of EVA, andexamine its statistical properties. This new perfor-mance measure, refined economic value added(REVA), complements EVA in that it could be usedin conjunction with EVA, with the choice of mea-sure dictated by the level of the organization atwhich the performance measure is used. We arguethat REVA is a better measure of performance fortop management, although EVA may be useful atlower levels. Our empirical tests are extensive, andwe also focus on nonparametric tests of the predic-tive abilities of EVA and REVA.

Although many companies use EVA for bothresource allocation and compensation purposes,virtually no research reported in the published lit-erature provides information about the statisticalrelationship ofEVAtoshareholdervalue.'^ One goalof this article is to fill that void. Moreover, muchremains to be learned about alternatives and com-plements to EVA, particularly for managers at dif-ferent levels of an organization. Although ourefforts in this direction are by no means exhaustive,we hope that the introduction of REVA is a usefulfirst step.

STRATEGY, VALUE, AND THE CHOICEOF PERFORMANCE MEASUREAn appropriate performance measure gauges howmanagement strategy affects shareholder value asmeasured by the risk-adjusted return on investedcapital. An appropriate performance measure togauge the effectiveness of a given strategy must,therefore, incorporate the required rate of return oninvested capital, accurately measure the amount ofcapital used by the company, and correlate highlywith the risk-adjusted rate of return earned byshareholders. In this section, we formalize howinvested capital and the required rate of return onthat capital should be measured, as well as what ismeant by an appropriate risk-adjusted return toshareholders.

Figure 1. The Components of Value

Net Present Valueof Current and

Future Investment* Opportimities

As5ets in Place

Market Valueof Firm

Econoniic BookValue of Assets

Invested Capital, Strategy and ValueFigure 1 illustrates the components of firm

value. The most transparent component of a firm'svalue is its physical assets in place: plant and equip-ment, real estate, working capital, and so forth.Another component is the net present value of thefirm's current and future investment opportunities.This component's value is less tangible than itsphysical assets, is driven significantly by the firm'sstrategy, and is sizable for many firms. Tlie totalvalue of the firm is the sum of these two compo-nents of value. The question is: How do we deter-mine these values?

The firm's balance sheet contains one measureof the value of the firm's assets in place. Considerthe accounting-based balance sheet in Table 1, forexample. Because of a variety of accounting distor-tions, the total asset value on the typical balancesheet does not accurately represent either the liqui-dation value or the replacement-cost value of theassets in place. Therefore, it is of limited use forasset valuation purposes and must be transformed.

The proponents of EVA, most notably SternStewart, are careful to adjust this accounting bal-ance sheet before arriving at an estimate of thevalue of the firm's assets in place.^ The adjustmentsinclude netting the non-iiiterest-bearing currentliabilities against the current assets, adding back to

Table 1. Accounting-Based Balance Sheet

Assets

Current assetsMet goodwillFixed assets

Total assets

s and Nd Worth

Non-interest-bearing current liabilities CNIBCLs)Interest-hearing current liabilities (IBCLs)Long-term debtEquity (net of write-offs)

Total liabilities and net worth

12 ©Association for Investment Management and Research

equity the gross goodwill (i.e., adding cumulativeamortized goodwill back to total assets), restructur-ing and other write-offs, capitalized value of R&D(and possibly advertising), LLFO reserve, and soforth. The debt balance is increased by the capital-ized value of operating lease paynients. The goal isto produce an adjusted balance sheet that reflectsthe economic values of assets in place more accu-rately than the inherently conservative, historical-cost-based balance sheet guided by generallyaccepted accounting principles.

After these adjustments, the typical firm's"economic book value" balance sheet would lookas it does in Table 2. Although this economic bookvalue balance sheet represents the value of theassets in place more accurately than the balancesheet produced by the firm's accountants, it stilldoes not determine the total value of the firm,which includes the value of future opportunities.

The total value of the firm—the sum of the twocomponents of value in Figure 1—is also equal tothe market value of equity plus the market value ofdebt. Thus, the difference between the marketvalue of the firm and the economic book value ofits assets in place represents the market's assess-ment of the value of the firm's current and futureinvestment opportunities. This difference can beconsidered an assessment of the value of the firm'scompetitive strategy and its deployment of humanresources. If we recast the firm's balance sheet inmarket value terms, we have a balance sheet suchas that presented in Table 3.̂

The market value of assets can be either aboveor below the economic book value of the assets inplace. In particular, if the firm executes a poorstrategy in the opinion of the market or if it doesnot possess the human resources needed to imple-ment a good strategy successfully, the market willlower the value of the firm's assets, perhaps belowthe economic book value of the assets in place.

How Much Capital Is Invested in theFirm?

A good financial performance measure shouldask how well the firm has generated operatingprofits, given the amount of capital invested toproduce those profits. The idea is that the firm'sfinanciers are free to liquidate their investment inthe firm and invest the liberated capital elsewhere.Thus, the financiers must earn at least their oppor-tunity cost of capital on the invested capital. Thiscondition implies that this cost of capital must besubtracted from operating profits to gauge thefirm's financial performance, EVA, for that reason,defines net operating profit after tax (NOPAT) andsubtracts a capital charge for the economic bookvalue of assets in place. The economic book valueof assets in place is the measure of the capital pro-vided to the firm by its financiers. But does thisamount truly represent the capital used to generatethe operating profit?

We beheve the answer to that question is nega-tive. At the beginning of any period, the financiersas a group could sell the entire firm for its marketvalue. They could then invest their proceeds in assetsidentical in risk to the firm and earn an expectedreturn equal to the firm's weigh ted-average cost ofcapital (WACC).̂ By not liquidating their holdingsin the firm, these financiers are forgoing this oppor-tunity to earn the weighted-average cost of capitalon the market value of the firm at the beginning ofthe period.

For the firm to create a true "operating" surplusfor its financiers in a given period, its operatingprofit at the end of the period must exceed a capitalcharge that is based on the total market value of thecapital used at the beginning of the period, notsimply the economic book value of its assets inplace. The capital commitment of the firm's finan-ciers is represented by the total market value of thefirm, not simply the economic book value of the

Table 2. Economic Book Value Balance Sheet

Assets Udbilitics ami Net Wortli

Currentasset5(withinventoryatFIFO)—NIBCLs lBCLsGross goodwill Debt (+ capitalized leases)Fixed assets Equity (+ adjustments for deferred taxes,

Economic book value of assets in place

gaodwill amortization, write-offs, LIFOreserves, etc)Total liabilities and net wortli

Table 3. Market-Value-Based Balance SheetAssets

Market vaiue of assets

Total market value of assets

Liabilities and Net Worth

Market value of debt (including leases)Market value of equity

Total market value of debt and equity

Financial Analysts Journai • May/June 1997 13

assets in place. This "investment" in the firm in anygiven period constitutes the capital that the firm hasused to produce profits. This perspecfive mofivatesour development of REVA in the next section.

Operating versus Trading-BasedPerformance

Measures of shareholder wealth creation focuson the firm's stock price performance and seek todetermine how much shareholders increase theirwealth from one period to the next based on thedividends they receive and the appreciation in thefirm's stock price. Essentially, such trading-basedperformance measures assess how well an investorwould have done if he or she had purchased a shareof stock at the beginning of the period and sold itat the end. This type of measure of shareholderwealth creation is called a trading-based measureof performance. In contrast, performance measuressuch as EVA focus on the firm's operating perfor-maaice from the standpoint of its financiers.

An operating measure of current performancefocuses solely on the performance of the firm in agiven period; a trading-based measure of perfor-mance captures revisions in the market's beliefsabout the firms's entire future stream of operatingperformances. If stock markets are efficient and weexamine a sufficiently long time horizon, these twomeasures will converge. We usually assess perfor-mance over shorter horizons, however: a year, aquarter, or a month. Therefore, any operating mea-sure of performance will diverge somewhat from atrading-based measure of performance. Obviously,we do not want to use a trading-based measure ofperformance for conripensafing all of our managers,particularly those at lower levels of the organiza-tion, whose decisions have less impact on stockprice. We do, however, want a measure of perfor-mance to be a barometer of shareholder wealthcreation against which we can judge the efficiencyof any operating performance measures.

Shareholders can earn a return on their invest-ment in two ways; through dividends and throughcapital gains. Over any period of time, t, the share-holder return for firm / can be specified as

D 4- t P P \: , + \r ; , — r : , 1 J

(CAPM) captures the first two factors by specifyingthat the expected return on a stock investment is

j . I'l

where Dj; is the dividends paid during the periodf - 1 to f and Pjf is the price of the shares at the endof period t.

Many factors influence Rj j , most notably, therisk of the investment, the interest rates prevailingin the capital markets, and the expertise of thefirm's managers. The capital asset pricing model

where

Rj = the risk-free bond yield at tim.e t

P. = firm /'s beta, a measure of thefirm's systematic risk

E{Rmi) - R. = the expected equity market riskpremium, usually taken as thelong-run, average realized returnon the market in excess of risk-freebond retums

The CAPM thus helps to determine the abnormalreturn firmy earned in period t. We call this returnalpha and calculate it as

Thus, a , measures the actual shareholder returnin excess of the return that was expected in a period,given the firm's systematic risk. In that sense, thereturn is abnormal. Alpha, we propose, is thus theappropriate measure of shareholder wealth cre-ation in any given period. Consequently, alterna-tive operating performance measures will bejudged by their correlation with this abnormalreturn.

This measure is indeed a high hurdle for per-formance. One would not expect a large number offirms to achieve positive abnormal returns formany periods. In fact, because a firm with a stringof positive alphas is one that continues to produceshareholder returns in excess of the risk-adjustedexpected return, this firm is consistently "beadngmarket expectations." Firms that perform this wellare truly exceptional and quite rare. The question,then, is: What operating measure of performancecorrelates highly with tliis measure of shareholderwealth creation?

REVA AS A HIGH-LEVELPERFORMANCE MEASUREThe operating measure of performance shouldobviously capture how well a given company hasperfornied in terms of operating income. Net oper-ating profit after taxes alone, however, is not anappropriate measure because it neither captureshow much capital is used to generate a given levelof income nor accounts for the required rate ofreturn on invested capital. Tl̂ ius, an appropriateperformance measure is one that includes net oper-ating income after taxes, the amount of capitalinvested, and the required rate of return on capital.

14 ©Association for Investment Management and Research

What Is REVA?EVA is defined as

EVA = NOPAT-fc .̂ (1)

where k^ is the weighted-average cost of capitaland NA is defined as adjusted book value of netcapital at the beginning of the period.

In Equation 1, NOPAT is defined as reportednet operating profits plus any increase in bad debtreserve plus any increase in the LIFO reserve plusamortization of goodwill plus any increase in netcapitalized R&D plus other operating income(including passive investment income) minus cashoperating taxes.

A firm's weighted-average cost of capital inEquation 1 (WACC = k^^) is derived from the follow-ing formula:

'M

•W^^M(2)

'Af

where

Dj^ - market value of the firm's total debt

EjVi = market value of the firm's total equity

kjj - pretax cost of debt

T = the firm's marginal tax rate

kj: = cost of equity

A firm's net asset base in Equation 1 is defined by acompany's total assets minus non-interest-bearingcurrent liabilities, with the adjustments discussedearlier. That is, net assets represent the total eco-nomic book value of the firm's assets in place.

The motivation for the REVA refinement toEVA stems partly from EVA's use of the economicbook value of assets when the capital charge for thefirm is derived from a market-based WACC. Tomake inferences about changes in shareholderwealth, a market-derived cost of capital should beapplied to the market value of the firm's assets.Thus, the RHVA for a given period t is defined as

REVA, = NOPAT,-;-,^,(MV,_j), (3)

where NOPATf is the firm's NOPAT at the end ofperiod t and MWj_i is the total market value of thefirm'sassetsat the end of period t-l (beginning ofperiod 0. MV .̂̂ is given by the market value of thefirm's equity plus the book value of the firm's totaldebt less non-interest-bearing current liabilities, allat the end of period t-l.

EVA and REVA ComparedThe key distinction between EVA and REVA is

that REVA assesses its capital charge for period ton the market value of the firm at the end of period

^ - 1 (or the beginning of period 0 rather than onthe economic book value of the assets in place.

The following example highlights the potentialdifferences in the shareholder wealth implications ofEVA and REVA, Suppose an investor holds one shareof XYZ Company stock at the beginning of the year.The stock is currently trading at $50 a share, and theeconomic book value of the stock is $40 a share.'̂Suppose that the investor and the rest of the marketexpect this stock to eam 10 percent over the next year.The investor expects this 10 percent return on the $50market value of the firm, not its $40 economic bookvalue. That is, the investor expects a payoff of $5 ashare, not $4. The contrast between EVA and REVAcan now be seen in this context. Suppose the firm iscompletely equity financed. With all-equity financ-ing, the WACC for this firm is simply its cost ofequity of 10 percent. Further assume that the firmhas TOO shares of stock outstanding at the beginningof the year (period t = 0). Therefore, the economicbook value of the firm is 100 times $40, or $4,000, andthe market value of the firm is 100 times $50, or$5,000, at / = 0. If during the year the firm generatesa NOPAT of $450 on the firm's invested capital, howhas the firm done in terms of EVA and REVA?

The firm has generated an EVA at time t-loi

EVAT = NOPATi - ($4,000 x 10%)

= S450 - $400

= S50,

According to the EVA perspective, this firm hascreated $50 of value for its shareholders. Now, con-sider the firm's REVA:

REVAi = NOPATj - ($5,000 x 10%)

= $450 - $500

- -S50,

According to REVA, this firm has destroyed $50 inshareholder value.

Should the firm's management be satisfiedwith its performance over this period? It has gener-ated a positive EVA and a negative REVA. The firm,with a cost of capital of 10 percent, has produced an8 percent return on the market value of its assets andani l .25 percent return on the economic book valueof its assets. Because an investor in this firm couldhave taken his or her $50 a share and invested it inanother company of equivalent risk to generate the10 percent required return, the NOPAT of 8 percentof the market value of the firm is, in our view,inadequate compensation.

This example, although simple, illustrates howa firm could be delivering less in terms of operatingeamings than the shareholders require and yetappear to be creating shareholder value based on

Financiai Anaiysts Journai • May/June 1997 15

EVA. Tliis possibility is the first justification forusing REVA.

Flows to Equity versus Total Flows toAll Financiers

Another difficulty with EVA provides a secondjustification for using REVA. Conceptually, EVAshould be the same regardless of whether we useNOPAT and the WACC or (adjusted) net incomeand the equity cost of capital. However, as weshow below, there is no such equivalence when themarket value of a firm differs from its book value.

The flows to equity-based EVA are defined as

EVAf(Equity flows) = Net income, - k^iE),

where net income is simply the NOPAT less thefirm's after-tax interest expense and £ is the eco-nomic book value of equity. We can then write

Net income^ = NOPAT, - ^^(1 - T>D,

where D is the economic book value of debt. There-fore, we should have

EVAf(NOPAT) = EVA,(Equity flows). (4)

Expanding both the left hand side and the righthand side of Equation 4 yields

NOPAT^-k^^,(NA,_•^) = Net income-A-^(£). (5)

Dropping the time subscripts and recognizing thatNet income ^ NOPAT - kijd - T)D, we can writeEquation 5 as

•AD

= Net income-/:/j(1 - 7")£'-A'£(£').

Therefore, for the equivalence to hold, D^ mustequal D and Ej^ must equal E. That is, the marketvalues of debt and equity should be equal to therespective economic book values. It is uncommonfor the market value of equity to be precisely equalto the economic book value of equity, primarilybecause market value includes an estimate of thevalue of future opportunities. This conceptual dif-ficulty does not exist with REV A, primarily becauseit relies on the market value of capital.

We can thus summarize two key advantages ofREVA relafive to EVA. First, whenever REVA ispositive, incremental shareholder value has beencreated. The operating income flowing to financiersat the end of the period as a percentage of themarket value of their investment at the beginning

of the period exceeds their opportunity cost of cap-ital. This condition does not hold for EVA—thefinanciers could be receiving an operating-income-based return that is less than their opportunity costof capital even when EVA is positive. Second, REVAcan be computed based on total operafing flows todebt and equity or only on the flows to equity. Thiscapability is true for EVA only when the marketvalues of debt and equity coincide with theirrespective economic book values.

Organization Level and Choice ofFinancial Measure

REVA is a more appropriate performance mea-sure than EVA when considering the shareholders'view of the firm. Hence, the senior executives in thefirm (e.g., the CEO and other members of the exec-utive committee) should be evaluated on the basisof the firm's REVA performance. The firm's valuederives both from its physical assets in place andits strategy with respect to future opportunifies.Both of these values are the appropriate domain ofthe firm's senior executives. The market value ofthe firm—which is a component in REVA—includes the values of both the physical assets andthe strategy, whereas the economic value of thefirm—which is a component in EVA—representsonly the values of the physical assets in place. Strat-egy is the primary responsibility of top manage-ment. The firm's economic value is an adequaterepresentation of invested capital from the stand-point of those below top management. Thus, REVAcould be used to compensate senior managementand EVA could be used to compensate divisionalmanagers and those below them.

EMPIRICAL ANALYSISThe empirical analysis used the Stern Stewart Per-formance 1000 database for the years 1982 through1992. We randomly selected 600 of the 1,000 firmsin the database and proceeded to calculate eachfirm's EVA, REVA, total shareholder return, andrisk-adjusted abnormal return for each year. Thesecalculations involved matching the 600 firms tocorporate financial data for those years. Allaccounting and financial market data were takenfrom Standard & Poor's Compustat and the Uni-versity of Chicago's CRSP databases, respectively.

Calculafing a firm's yearly EVA and REVArequires estimates of its NOPAT, WACC, economicbook value of assets, and market value of assets. Inour calculation of REVA, we used the NOPAT dataprovided by the Stem Stewart database. The marketvalue of capital was estimated by summing the mar-ket value of equity, book value of interest-bearing

16 ©Association for Investment Management and Research

liabilities, and book value of preferred stock. Bookvalues were used for debt and preferred stockbecause market values for these variables are notavailable.̂ '-' The WACC was estimated as theweighted average of the cost of equity, debt, andpreferred stock, where the weights are the market-based capital structure weights obtained from thecapital structure components. To estimate the firm'scost of debt, we assigned the bond yield as reportedin Standard & Poor's Industrial Bond Guide for com-mensurate bond rafings. The after-tax cost of debtwas estimated by multiplying the firm's cost of debtby its marginal tax rate, which was estimated bydividingthereported tax expenseby the firm's pretaxincome. The cost of equity was estimated using theCAPM, The cost of preferred stack was estimatedas the average of the cost of equity and debt.

To determine the cost of equity, estimates of therisk-free rate, the expected market risk premium,and the firm's beta are needed. The yield to matu-rity of a one-year discount bond taken from theCRSP Fama-Bliss files was used as a proxy for therisk-free rate. To gauge the expected market riskpremium, historical averages of realized annualmarket risk premiums from 1926 to the )''ear priorto the observation year were calculated. The real-ized market risk premiums were taken from Tbbot-son Associates, 1995 Yearbook of Stocks, Bonds, Bills,and Inflation. Eirm betas were calculated using themethodology outlined by Fama and French (1992).This methodology involves first assigning allNYSE, Amex, and Nasdaq firms available on CRSPto 10 size-based portfolios. Each size portfolio isthen portioned into 10 beta portfolios. The monthlyreturns of each of the 100 equally weighted portfo-lios were used to estimate the portfolios' betasusing the techniques outlined by Dimson (1979).The portfolio beta was assigned to each firm withinits portfolio and used to calculate the firm's cost ofequity.

EVA was estimated using NOPAT and the eco-nomic book values of capital from Stern Stewart,and the WACC used to calculate REVA was alsoused to re-estimate EVA. This step was to isolate thetrue difference between REVA and EVA—that is,the different capital bases. Using different WACCscould provide results driven solely by the methodused to estimate WACC and, as a result, wouldcloud the true underlying relationship between thevariables in question. ̂ ^

Empirical Resuits.' Hozv well do EVA and REVA correlate with

shareholder wealth creation? The first set of testsinvolved regressing the CAPM-based abnormal

returns of each security on various combinations ofEVA and REVA measures to determine how welleach performance measure explains abnormalretums—that is, retums over and above theexpected retvim. Both EVA and REVA were scaledby the market value of equity to create two new vari-ables, EVARET and REVARET. This step was doneto make both variables consistent with the abnormalreturn variable, which was measured as a percent-age. The market value of equity was used becauseboth EVA and REVA gauge the value creation forshareholders. First, we regressed abnonnal returnson EVARET and REVARET individually. Ourresults indicate that, on an individual basis, bothEVARET and REVARET are positively related toabnormal retums at the 1 percent significance level(see Table 4). A1 percent increase in H VARET resultsin a 0.27 percent increase in abnormal returns. Sim-ilarly, a 1 percent increase in REVARET leads to a0.58 percent increase in abnormal retums. There-fore, on average, an increase in either EVA or REVAleads to an increase in shareholder wealth.

What about past EVAand REVAaspredictorsoffuttire abnormal returns? Our next set of testsinvolved adding lagged values of EVARET andREVARET into the abnormal retums regressions.These tests examined whether past realizations ofEVA and REVA have a significant effect on abnor-mal retums. In the regression of abnormal returnson EVARETand lagged EVARET, EVARETissignif-icantly positively related to abnormal retums andlagged EVARET is significantly negatively related toabnormal retums (see Table 5). In a mulfiple regres-sion framework, the coefficient on lagged EVARETrepresents the sensitivity of abnormal returns tochanges in lagged EVARET, which are uncorrelatedwith contemporaneous EVARET. Similarly, thecoefficient on EVARET measures the sensifivity ofabnormal returns to changes in EVARET, which areuncorrelated with last period's EVARET. Thus, thecoefficient on EVARET represents how abnormal

Table 4. Abnormal Returns as Explained byEVARET and REVARET(f-statistics in parentheses)

Coefficient

Intercept

Abnormal return

EVARET

11,12047

(16,683)*'

0,26620

(5,942)**

0.0114

REVARET

0,16232

(19,366)**

0.57997

(11.212)*'

0.0393

Note: Sample sizi? - 3,076.

**Significant at the 1 percent level.

Financial Analysts Journai • iViay/June 1997

Table 5. Abnormai Returns on Contemporaneous and Lagged EVARET andREVARET(f-statistics in parentheses)

EVARET REVARET

Coefficient Current Lagged Current Lagged

bitercept

Abnormal return

0,11969

(15,357)**

0,47639 -0,12918

(6,555)** (-1.819)*

0,0205

0.16549

(16.138)'*

0.75111 -0,10277

(9.588)*" (-1,408)

0.0396

Nate; Sample size = 2,574,

"Significant at the 10 percent level,**Significant at the 1 percent level.

returns vary with unexpected changes in EVA, whereexpectations are based on the previous period'sEVA, The results above are consistent if one assumesthat the market uses EVA not only to assess abnor-mal returns this period but also to predict futureperformance. If the firm has a large EVA this period,the market may revise its valuation of the firmupward significantly, refiecting revised expecta-tions about future profitability. If the niarket'sbeliefs are not confirmed in the next period, how-ever, the value of the stock could fall. Thus, even if afirm has a positive EVA this period, the stock maynot eam positive abnormal reti.irns.

The first key implication of this analysis is thatthe market appears to reward unexpected increases inEVA and bases expectations of future profitability onthis period's EVA. The second key implication is thatEVA may bo inappropriate to use as a compensationmeasure for top management because the market isreally rewarding oiily the unexpected porfion of EVA.Basing conipensation on total EVA may result inrewarding managers for subpar performance.

A similar regression was run using REVARETand lagged REVARET. The coefficient on contem-poraneous REVARET is again significant, althoughlagged REVARET is insignificant (see Table 5).This result highlights a key empirical difference

between REVA and EVA. REVA is posifively relatedto abnormal returns, but any revaluafion based onpast period's REVA is impounded into contempo-raneous REVA directly; this result occurs becauseREVA is a function of the market value of equity. Apositive REVA in a given period that leads to revi-sions in the market's expectations regarding futureperformance will "raise the hurdle" by increasingthe market value of equity and, as a consequence,next period's capital charge. Thus, REVA is a moreappropriate compensation measure because it is atruer measure of whether the firm has surpassedthe market's expectation and thereby added share-holder value.

We also regressed abnormal returns on EVARETand REVAl^T simultaneously. Tlie purpose of sucha regression was to determine which performancemeasure does better in explaining a firm's abnomnalreturns when the impact of the other measure isaccounted for. The coefficient on REVARET is againsignificantly positive; the EVARET coefficient is nowsignificantly negative (see Table 6). This result indi-cates that REVA contains information not capturedill EVA that is relevant for predicting abnormalretums and that increases in EVA that are uncorre-lated with changes in REVA result in significantlylower abnormal returns.

Table 6. Abnormal Returns as Explained by Both EVARET and REVARET(f-statistics in parentheses)

Coefficient EVARET REVARET

Intercept

Abnormal retiurn

0,16746

(19-642)**

-0.20407 0.75450

(-3.162)" (9.981)**

0,0424

Note: Sample size = 3,076,

'"Significant at the 1 percent level.

18 ©Association for investment Management and Research

In summary, EVA is significantly related toabnormal returns. As shown in Table 6, however, theinformafion provided by REVA subsumes the infor-mation implied by EVArealizations. Therefore, con-sistent with the theory, REVA better explainschanges in shareholder wealth and also manage-ment performance at the top level of the firm.

Nonlinear Relationships Estimation*- Hozv often does a positive REVA (EVA) predict

a positive excess return? One potential problem withthe regression analysis is that both REVARET andEVARET may be related to abnormal returns in ahighly nonlinear manner. An additional methodol-ogy was used to account for any nonlinearities:nonparametric tests to capture how well REVA andEVA predict the sign of the abnormal retums. Thisapproach is equivalent to calculating the probabil-ity that given that a firm has a positive EVA (orREVA), the firm generates a posifive abnormalreturn.

Our test involved splitting the sample into twogroups: those with positive EVA and those wi tli zeroor negative EVA. The proportion of observations ineach group that had positive abnormal retums wasthen calculated, If EVA is a good predictor of abnor-mal returns, the proportion of firms that experiencedposifive abnormal retums in the positive EVA sub-sample should exceed the proportion of firms thatexperienced positive abnormal retums in the nega-tive EVA subsample. In other words, Probiabnormalreturn > 0,given that EVA > 0) should be greater thanProHabmmnal return > 0given that EVA < 0). We foundthat 68.05 percent of the positive EVA sample exhib-ited positive abnornial returns whereas only 56.01percent of the negative EVA sample exhibited posi-tive abnormal retums. This procedure was repeatedusing REVA. Positive REVA iniplied a positiveabnormal return 77.22 percent of the time; a negativeREVA implied a positive abnormal return only 58.48percent of the time.

These results indicate that the proportion ofpositive REVA that correspond to positive abnor-mal returns is significantly higher than the sameproportion for EVA. Thus, although EVA on itsown predicts abnormal retums fairly well, REVAperforms significantly better. This finding isimportant because senior management should seeka performance measure with the greatest ability topredict correctly directional changes in shareholderwealth. Our results iniply that REVA serves both ofthese roles, as displayed in Table 6 and in the non-parametric tests.

'"• How do REVA and EVA stock portfolios per-form? Our extensive empirical tests docunient thatEVA on its own can predict changes in shareholder

wealth. The simpler measure, REVA, goes even far-ther, however, REVA appears to include all informa-tion relevant to a firm's abnormal retum that iscontained in EVA. Moreover, the statisfical power ofREVA holds up over long time horizons and out-of-saiTiple tests. We formed three portfolios of stocksfrom our sample on the basis of the top 25 EVA firmsover the 1982-87 period, the top 25 REVA firms overthe same period, and a value-weighted market indexfor the same period. We then examined the totalportfolio return over the 1988-92 time period. Theresults of this process, shown in Table 7, documentthat the high-performance REVA firms outper-formed both the high-performance EVA firms andthe market index.

CONCLUSIONThis analytical framew^ork for evaluating operat-ing performance measures in the context of share-holder value creation indicates that the mostappropriate measure of shareholder value is thereturn shareholders eam through price apprecia-tion and dividends in excess of that required tocompensate shareholders for systemafic risk. Weconclude that EVA does quite well in terms of itscorrelation with this measure of shareholder valuecreation.

EVA, however, views the economic book valueof the physical assets in place as the capital finan-ciers commit to the firm. We propose that a moreappropriate measure of the capital used in the firmfor any period of time is the niarket value of thefirn^ at the beginning of the period. This approachled us to a refinement of the EVA measure, REVA,REVA assesses a capital charge for a period equalto the weighted-average cost of capital times themarket value of the firm at the beginning of theperiod. This premise permits computation of REVAusing either flows to equity or flows to all finan-ciers, which is not possible with EVA unless marketand ecoiiomic book values happen to be equal bychance.

We conducted a comprehensive statisticalanalysis of both EVA and REVA to estimate theircorrelation with and ability to predict shareholdervalue creafion. REVA statistically outperforms EVAin this regard. Moreover, the realized returns for thetop 25 REVA firms were higher than the realized

Table 7. Portfolio Analysis of EVA and REVA(percentages)

I 'ortfolio Total 1988-92 Porttolio Retum

Top 25 EVA firms (1982-87)

Top 25 REVA firms (1982-87)

Value-weighted market index

15.804%

17.013

12,861

Financiai Anaiysts Journai • May/June 1997 18

returns for the top 25 EVA firms for the 1988-92period. We concluded that REVA could be used tocompensate senior executives and EVA could beused to compensate those at lower levels in an

organization. Euture research in this area shouldaddress the important issues of optimal compensa-tion design using both EVA and REVA for firms ofhierarchical organizational ^̂

NOTES

Noise trading, portfolio insurance, and other factors unre-lated to the firm's performance may induce randomness instock prices. See Milbourn (1996) for a theoretical and em-pirical examination of some of these issues.The theoretical statistical expectation of the abnormal return2.

3. Articles in the popular press have mentioned the correlationofEVAtomeasuresofshareholdervalue,but the underlyingresearch has not been reported.

4. Stem Stewartconsidersmorethan250accountingadjustmentsin moving to EVA, In practice, however, most firms find thatno more than 15 adjustments are of material significance,

5. For nonpubhcly traded firms or divisions, a typical ratio ofmarket value to book value for comparable firms could beused to convert existing book values into market values.

6. That is, the firm's weighted-average cost of capital is definedas the financiers' opportunity cost of capital,

7. This disparity between economic book value and marketvalue is representative of the sample of firms used in oursubsequent empirical analysis. Specifically, we document anaverage ratio of adjusted economic book value to marketvalue to be 78.19 percent,

8. Finance companies, for example, may find it appropriate touse this flows-to-equity approach,

9. As recommended in the calculation of EVA and as we defineREVA in Equa tion 3, we did not include a firm's non-interest-bearing current liabilities.

10. Following most finance researchers, we assumed that the book

valuesofdebtand preferred stockapproximate market values,11, See Sharpe (1964).12, See Ibhotson (1995).13, As in Fama and French (1992), we estimated beta as the sum

of the slope coefficients in a regression of portfolio returnson the contemporaneous market return and one-periodlagged market return,

14, We chose to re-estimate the WACC for each firm each yearbecause the WACCs reported in the Stern Stewart Perfor-mance 1000 are based on rolling three-year averages.

15, Because beginning-period market value of equity appearsin the denominator of both the independent and dependentvariables, the independent and dependent variables may bespuriously correlated. As a result, one might find a signifi-cant relationship between the independent variables andabnurmal returns even if the underlying variables are un-correlated. Because the market value of equity undergoes anonlinear transformation, however, the spurious correla-tion should be reduced. To control for any residual spuriouscorrelation, we ran instrumental variable regressions inwhich the lagged values of EVARET and REVARET wereused as instruments. The conclusions drawn from the in-strumental variable regressions were qualitatively the sameas those drawn from the ordinary least squares regressions,

16, Both ofthese differences aresignificant at the 1 percent level,17, This difference is significant at the 1 percent level,18, These nonparametric results fortify our intuition that a firm

could have a positiveEVA and stillbedestroying shareholdervalue,

19, We would like to thank W. Van Harlow III for his helpfulcomments on an earlier draft of this article.

REFERENCESDimson, Elroy. 1979, "Risk Measurement When Shares AreSubject to Infrequent Trading." Journal of Fiimucint Economics,vol. 7, no. 2 (June):l97-226.

Fama, Eugene F., and Kenneth R, French, 1992, "The Cross-Section of Expected Returns," journal of finance, vol, 47, no, 2(June):427-65,

Ibbotson Associates, 1995. 1995 Yearbook of Stocks, Bonds, Bills,and Inflation. Chicago: Ibbotson Associates,

Jensen, Michael, and Kevin Murphy, 1990, "Performance Payand Top-Management Incentives," journal of Political Economy,vol, 98, no, 2 (April) :225-62.

Miiboum, Todd T, 1996, "The Executive Compensation Puzzle:Theory and Evidence," IFA Working Paper No, 235, LondonBusiness School.

Sharpe, William F, 1964, "Capital Asset Prices: A Theory ofMarket Equilibrium under Conditions of Risk," journal ofFinance, vol. 19, no. 3 (Ju]y):425-42,

Smolowe, Jill. 1996. "Reap as Ye Shall Sow," Time, vol, 147, no.6 (February 5):45.

Stewart, G, Bennett UI, 1991. The Quest for Value. New York:Harper Business

20 ©Association for Investment Management and Research