The Marketing Mix Decision under Uncertainty Author(s): Harsharanjeet S. Jagpal and Ivan E.

Brick Source: Marketing Science, Vol. 1, No. 1 (Winter, 1982), pp. 79-92 Published by: INFORMS Stable URL: http://www.jstor.org/stable/184075 Accessed: 08/12/2009 10:51
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=informs. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Marketing Science.

http://www.jstor.org

THE MARKETING MIX DECISION UNDER UNCERTAINTY*

HARSHARANJEET S. JAGPALt
AND IVAN E. BRICKt

This paper develops a marketing mix model under uncertainty using the Capital Asset Pricing Model (CAPM) valuation framework. The model is general because it treats price, advertising and personal selling simultaneously, and allows for general patterns of uncertainty. Because the manager often lacks precise quantitative information about the sales response function, the analysis focuses on the qualitative properties of the model. The methodology of comparative statics is used to determine how the marketing mix should be modified when market conditions change. Specifically, the comparative statics are shown to depend on how advertising and personal selling influence price-sensitivity, the interaction between advertising and personal selling and the relationship between sales and the return earned in the capital markets. The comparative statics for risk depend, in addition, on the precise mathematical form with which uncertainty enters the random demand function (i.e., additive, multiplicative or generalized). For example, suppose demand uncertainty increases in the additive case and sales covary positively with the return of the market portfolio. Then, if advertising and personal selling are complementary, advertising decreases if increased advertising and increased personal selling decreases price-sensitivity. However, if advertising and personal selling are substitutes, then advertising decreases if increased advertising decreases price-sensitivity but increased personal selling increases price-sensitivity.

1. Introduction Previous marketing mix models under uncertainty have considered only price and advertising (see Horowitz 1970, Dehez and Jacquemin 1975 and Brick and Jagpal 1981). This is restrictive because the firm can, in general, influence sales by changing price, advertising and the personal selling effort. In particular, advertising and personal selling often have different informational roles for the firm. For example, advertising may provide information on product availability and price whereas personal selling is aimed at providing
*Received January 1981. This paper has been with the authors for 2 revisions. Key words. Uncertainty, marketing mix, CAPM. tGraduate School of Management, Rutgers University, Newark, New Jersey 07102.
MARKETING SCIENCE Vol. 1, No. 1, Winter 1982 Printed in U.S.A.

79
Copyright ? 1982, The Institute of Management Sciences --l
-I-

0732-2399/82/0101/0079$01.25

80

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

detailed information on product attributes to target user groups (e.g., heavy and light users) according to their specific needs. Furthermore, advertising and personal selling do not necessarily have a symmetric effect on pricesensitivity. For example, advertising can inform consumers about prices, hence increasing price-sensitivity. However, personal selling may reduce the consumer's uncertainty about quality and hence lower price-sensitivity (see Ehrlich and Fisher 1982). Finally, except for Brick and Jagpal, previous models are difficult to operationalize because they use the expected utility framework.1 Furthermore, they make the restrictive assumption that uncertainty is exogenous and do not show how price and advertising should be changed when market conditions are altered. This paper develops a marketing mix model under uncertainty which treats price, advertising and personal selling simultaneously. The treatment of uncertainty is general because the model allows the variability of sales to depend nonmonotonically on the marketing mix variables. Furthermore, in contrast to previous models, we use the methodology of comparative statics to determine how the marketing mix should change when market conditions are altered. This is important because the marketing manager often does not know the precise empirical form of the sales response function. Specifically, we show using the Capital Asset Pricing Model (CAPM) valuation framework, that the comparative statics for shifts in demand, input costs and the productivities of marketing policy instruments depend on (a) whether advertising and personal selling affect the risk-adjusted price elasticity symmetrically;2 (b) whether advertising and personal selling are complementary, substitutes or independent and (c) whether the covariance of sales with the return on the market portfolio is positive or negative. The comparative statics for risk depend, in addition, on the precise mathematical form with which uncertainty enters the random demand function (i.e., additive, multiplicative or generalized). Welldefined comparative statics are obtained for all policy variables for shifts in demand, input costs and changes in marketing productivities. For changes in risk, the results are well-defined in the additive uncertainty case. However, in the multiplicative and generalized uncertainty cases, well-defined comparative statics are obtained for price only. 2. Value Maximization and Expected Utility Models Previous pricing models under uncertainty assume that the firm maximizes expected utility (see Horowitz 1970 and Dehez and Jacquemin 1975). This valuation approach, however, is subject to several important limitations.
1A key empirical problem in the measurement of utility functions is intransitivity (see, for example, Alchian 1953 and Allais 1953). Furthermore, many decisions are made by groups rather than individuals. In such cases, there is a theoretical problem because group preferences do not always satisfy the intransitivity assumption (see Sandmo 1971). 2Price elasticity is a random variable under uncertainty. The concept of risk-adjusted elasticity is defined later in the paper.

THE MARKETING MIX DECISION UNDER UNCERTAINTY

81

Firstly, as shown by Arrow (1951, p. 3), the group utility function may be undefined when there are several decision-makers in the firm. Secondly, the expected utility model assumes unrealistically that the decision-maker can maximize his own expected utility regardless of the effect of his policies on the value of the firm (see Fama 1980). Thirdly, as Jensen and Meckling (1976) note, firms are a nexus of contracts, not some anthropomorphic entity. Consequently, it is unrealistic to speak of the expected utility of the firm. Finally, as noted earlier, the expected utility model is difficult to operationalize. We assume in this paper that managers choose policy to maximize the value of the firm. This approach has three important advantages over the expected utility framework. Firstly, the value maximization approach is more general because it integrates decision-making by the firm with financial market equilibrium. Secondly, from a normative viewpoint, value maximization is optimal for every stockholder in the firm regardless of his attitude to risk. (See Fama 1978 and Fama and Miller 1972, pp. 150-156). This is because any policy which maximizes the value of the firm automatically maximizes the value of the stockholder's ownership in the firm and hence his wealth. Given that value maximization by the firm implies wealth maximization for all stockholders, any given stockholder can choose the optimal combination of consumption and investment based on his own preferences and attitude to risk. Thirdly, the value maximization approach can be operationalized (see, for example, Brealey and Myers 1980, Chapters 2-5). 3. The Capital Asset Pricing Model (CAPM)

The most widely used value maximization model is the Capital Asset Pricing Model (CAPM) developed by Sharpe (1964), Lintner (1965) and Mossin (1966). The CAPM has been used to determine the firm's optimal capital structure and cost of capital (see Kim 1978, Chen 1978 and Brick and Statman 1981) and to evaluate portfolio performance (see Jensen 1968 and Fama 1972). Previous applications of the CAPM to managerial decisionmaking include papers by Meyer (1976), Greenberg, Marshall and Yawitz (1978) and Brick and Jagpal (1981). A brief outline of the CAPM is presented below (see Sharpe 1980, Chapters 6-8 for a detailed exposition). Modern portfolio theory assumes that investors choose their holdings of securities based on the expected return and the risk of each security during a given time period (i.e., the holding period). Most importantly, the relevant aspects of a security are its contribution to the return and risk of an efficient portfolio, i.e., a portfolio such that it is impossible to obtain a higher expected return for a given risk or a lower risk for a given expected return. Suppose the ith security is a member of the efficient portfolio p. Then the contribution of the ith security to the risk of the efficient portfolio is defined by its "beta" ,Ai, which is equivalent to the regression coefficient of the return of security i, R, (dependent variable) on the return of the efficient portfolio Rp. Furthermore,

82

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

the expected return on security i is linearly related to its risk fip (see Fama 1976, p. 257-270). Given the assumptions of the CAPM,3 it can be shown that the relevant efficient portfolio is the market portfolio, m, defined as the combination of all securities each in proportion to the market value outstanding. Hence E(Ri) is linearly related to its risk /im defined with respect to the market portfolio. Specifically, in equilibrium, the relationship between the rate of return on the ith security (Ri) and the market portfolio (Rm) is given by
- rf] = Pim[E(Rm)(1)

[E(Ri)

rf]

where rf: the single period risk-free interest rate (e.g., the rate on a one year Treasury bill4); 3im: the beta of security i, which measures the responsiveness of the return of security i to that of the market portfolio. Because the expected rate of return of any asset is derived from its value, we can rewrite equation (1) in terms of the value of the ith security (or equivalently the ith firm), Vi, defined by: ,R) E(7Ti)- amCov(Qi (1 + rf) where: 7i,: random cash flow for the period; Rm: random rate of return of the market portfolio (i.e., the value-weighted return of all securities); rf: single-period risk-free interest rate; am: market price of risk defined by am= (E(Rm) - rf)/Var(Rm) where Var is the variance operator. We assume that policy is chosen to maximize the value of the firm Vi. Alternatively, as equation (2) clearly shows, policy is chosen to maximize the present value of the risk-adjustedcash flow during the current period.
3The CAPM assumes that (a) investors are risk-averse and seek to maximize their end of period wealth; (b) returns from risky assets are normally distributed; (c) investors have homogeneous expectations and (d) capital markets are perfectly competitive. Although these assumptions are somewhat less general than those of the expected utility model, they can be relaxed to varying extents without changing the basic properties of the model (see Rubinstein 1974). An important property of the CAPM, shown by Rubinstein, is that all investors in the firm can maximize their expected utilities provided their utility functions belong to the linear risk tolerance (HARA) class and there are no production externalities. 4The risk-free interest rate is generally measured by the rate on Treasury bills. Assuming that the holding period is one year, the appropriate risk-free rate is the rate on a one year Treasury bill.

(2)

THE MARKETING MIX DECISION UNDER UNCERTAINTY

83

4.

Marketing Mix Model

Let Q(p,X,,X2, u) be the random sales function where Q denotes demand, denotes price, XI is the number of advertising messages, X2 is the number of p salesmen and u7is the stochastic disturbance term. In order to allow for a general pattern of uncertainty, consider the random demand function: Q(p,X, ,X2, ) = a(p,X ,) +(p,X,X)+ ,X ,X 2)u (a > 0, > 0) (3)

where a and /, are deterministic functions. Equation (3) allows the variability of sales to be endogenously determined in a nonmonotonic manner. Furthermore, this specification is general because (a) it does not force the variability of revenue to increase with expected revenue as in Leland (1972), (b) the additive model (/ = 1) treated by Horowitz, and Dehez and Jacquemin is included as a special case and (c) well-known econometric specifications such as the Cobb-Douglas, multiplicative nonhomogeneous (see Vinod 1972) and translog functional forms (see Christensen, Jorgenson and Lau 1973) are special cases. Let W, and W2 respectively be the unit costs of advertising and personal selling assumed to be competitively determined and non-stochastic. Finally, assume that the density functions of u and Rmare not affected by the firm's policy. Assuming that unit costs (c) are constant for feasible demand levels, the random profit of the firm is -7 = (p - c)(a + f/u) - WlXl - W2X2 using equation (3), and Cov(7, Rm)= /(p - c) Cov(u, Rm). Substituting in equation (2),

the value of the firm V is: V(=

(p C)(a + -f(E(u) amCov(7,Rm )) W1X, W2X2

) (1 + rf)

(4)

The firm now chooses the marketing mix vector (p,X,,X2) which maximizes V. Differentiating equation (4), the first-order conditions are: V=(+ and Vx, = ((p - c)Qx W)(l + rf)- = 0 (i = 1,2) (6) (p - c)Up)(l + rf)-l=0 (5)

where Q = a + fi(E(u) - amCov(u,Rm)) is defined as the risk-adjusted sales5

5Under uncertainty, sales is a random variable. Risk-adjusted sales simply measures the certainty equivalent of a risky sales distribution (i.e., the certain sales level which is equally preferred by the market to the uncertain sales distribution).

84

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

and the subscripts denote derivatives. We assume that V is concave so that the second-order conditions are satisfied. Equation (6) is similar to the first-order conditions under certainty and implies that at optimality, the risk-adjusted value marginal product of advertising6 (personal selling) equals the unit cost of advertising (personal selling). Define the risk-adjusted price elasticity of demand7 as E = -(Qp)(p/Q) and the risk-adjusted advertising and personal selling elasticities as , = (Qx,)(XI/Q) and X2= (Qx)(X2/Q) respectively. Then equations (5) and (6) imply that: Exi w xi WX fxjp

(i = 1,2).

(7)

pQ

Equation (7) states that at equilibrium, the ratio of the risk-adjusted advertising (personal selling) and price elasticities equals the ratio of the total advertising (personal selling) outlay to the risk-adjusted value of the future sales revenue. This result generalizes the Dorfman-Steiner Theorem 1954 to the uncertainty case with several marketing mix instruments. (See Lambin, Naert and Bultez 1975 for a similar development under certainty in an oligopolistic market.) 5. Managerial Implications

From a policy viewpoint, it is important to determine how the marketing mix should be changed when market conditions are altered. In order to conduct the analysis at a general level, we do not assume particular types of functional form for the sales response function. This is important because in many cases the manager can only weakly specify the shape of the sales response function. For example, he may believe that increased advertising lowers (raises) the price sensitivity of the market. Similarly, he may believe that increased personal selling makes advertising more (less) productive. As will be shown, the comparative statics approach is particularly useful in such cases. This is because the manager can determine the directions in which optimal marketing policy changes when market conditions are altered, even if the only informationabout the sales responsefunction is qualitative. a. Demand Shifts8 Consider an exogenous shift in demand because of market growth or
contraction. Let Q '(p, X1, X2, u) = Q (p, X,X2, u) + 6 define the new demand

curve where 8 is the shift parameter. In order to determine the effect on the marketing mix, we use the methodology of comparative statics. Totally differ6The risk-adjustedvalue marginal product of advertising (personal selling) is the value marginal product of advertising (personal selling) measured with respect to the certainty equivalent of sales. See also footnote 5. 7Under uncertainty, sales and hence price elasticity is random. The risk-adjusted price elasticity measures the sensitivity of the risk-adjusted sales to price. See also footnote 5. 8See Tables 1 and 2 for summary of results.

THE MARKETING MIX DECISION UNDER UNCERTAINTY

85

entiating the first-order conditions (5) and (6) and solving for ap/la obtain:
- vps
ap/a8= -

we

VpX,
Vx,x,x2
VX2X,

VpX2
I-1
Vx2x2

Vxs
VX28

(8)

where D is the matrix of the second-order cross-partials of V with respect to p, = X, and X2 respectively. In this case, Vf, = (1 + rf)-l and Vx,a = Vx2^ 0. Hence equation (8) reduces to:
X

/2
/

IDI'(l

+ rf<'

VX2XI

VX2X2|

-D, D(l+ rf)

Since D is negative-definite by the second-order conditions and D, is a second-order principal minor of D, dp/a8 > 0. Hence, price always changes in the same direction as the demand shift. The effects of the demand shift on advertising and personal selling are more interesting. In particular, the results depend crucially on the type of interdependence between advertising and personal selling and the effects of advertising (personal selling) on price-sensitivity. Solving as before, we obtain:

IDl(1 +
and

rf) [ V

Vx2x2

Vpx2Vxl2]

x2/a8= ,

1 IDl(1 + r[)

[Vpx2V'x

Vpx,Vxx]

Because D is negative-definite, sign (aX1/68) and sign (aX2/8) are the opposite of the signs of the terms in the square brackets. In order to interpret these results, it is necessary to examine the economic meaning of the cross-partials Vpx (i = 1,2) and Vxx. Differentiating the risk-adjusted price-elasticity we obtain: a e
_

p(l + rf)

-e

_ p(-+cf)

(i-= 1,2)

86

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

using the first order condition (5). Since Q > 0 and p > c for any realistic pricing policy, sign(Vpx)= -sign(aEp/aXi). That is, the price-advertising interaction effect Vpx, is positive (negative) if price-sensitivity decreases (increases) when advertising is increased. Similarly, the price-personal selling interaction effect Vpx, is positive (negative) if price sensitivity decreases (increases) when the personal selling effort is increased. Now we cannot specify a priori whether increased advertising (personal selling) increases or lowers price sensitivity.9 Furthermore, as discussed earlier, advertising and personal selling can have asymmetric effects on price sensitivity. Hence we cannot assume that sign(aep/aX,) = sign(ap/aX2). Finally, in order to interpret Vxl,X differentiate equation (6) to obtain Vx,x2= (P c)Qx,x. Hence sign (Vxx) = sign( Qx,x) It is therefore necessary to distinguish whether advertising and personal selling are complemen-

Consider first the case where advertising and personal selling are complementary. Then, clearcut results are obtained if advertising and personal selling influence price-sensitivity in the same direction. In particular, both advertising and personal selling increase (decrease) if increased advertising/personal selling increase (decrease) sensitivity. If, however, advertising and personal selling are substitutes unambiguous results are obtained if changes in advertising and personal selling have opposite effects on price sensitivity. In this case, advertising increases and personal selling decreases if increased advertising decreases price sensitivity but increased personal selling increases price sensitivity. These results are reversed if increased advertising increases price sensitivity but increased personal selling reduces price sensitivity. Finally, suppose advertising and personal selling are independent. Then advertising increases (decreases) if increased advertising reduces (increases) price sensitivity. Similarly, personal selling increases (decreases) if increased personal selling reduces (increases) price-sensitivity. Note that, in general, a shift in demand should change all marketing mix variables simultaneously. This is because both advertising and personal selling influence the response of sales to price. Assume that advertising and personal
9Previous empirical research has only examined the relationship between price sensitivity and advertising under certainty. (See Comanor and Wilson 1974 and Eskin and Baron 1977). Unfortunately, the evidence is not consistent. Specifically, Comanor and Wilson (Chapter 5) use observational data on different industries and find that price elasticity falls when advertising increases. Eskin and Baron, in contrast, use test market data and find a significant negative interaction effect between price and advertising in three of four test markets analyzed. Hence Eskin and Baron's results are consistent with price sensitivity increasing when advertising increases. We believe that Eskin and Baron's results are more persuasive because their study was conducted under controlled test market conditions. (For a detailed critique of the ComanorWilson results, see Peterman and Carney 1978).

( tary( Qx,2 > 0), substitutes Qxx2 < 0) or independent Qx,2 = 0). (

THE MARKETING MIX DECISION UNDER UNCERTAINTY

87

selling are complementary and an increase in advertising (personal selling) reduces price-sensitivity. Then, given an increase in demand, the manager will increase price to bolster the unit profit margin. However, in order to reap the maximum benefit of the demand increase, he should increase his advertising and personal selling efforts to decrease the extent to which the increase in price chokes off demand. b. Changes in Risk From equation (4) the risk of any policy depends on Cov(u7, Rm).Proceeding as before, let Cov'(, Rm))= Cov(u7, Rm)(1+ 6) where 8 is the shift parameter. In this case, the comparative statics depend on the precise form with which uncertainty enters the random demand function. Differentiating the new first-order conditions we obtain:
Vp=amCov(u7, m)(/ + (pc)a//ap)(l + rf)-1 (10)

and
V = -am((p c)(a/laxi)Cov(a7,Rm))(l + rf)-.

(11)

Let M = (p - c)Q denote the random gross profit margin. Then aa/lap = (/3 + (p - c)d,3/ap)ca and aao/ax, = (p - c)(al3/aX)a. We assume in the analysis that Cov(ui,Rm) > 0 because the results are exactly reversed if
Cov(J, Rm) < O.

Consider first the additive uncertainty specification ( , - 1 in equation (3)). In this case, the comparative statics are well-defined for all policy variables. Specifically, Vp, < 0 and Vx, = Vx2 = 0 in equations (10) and (11). Hence, if uncertainty is exogenous, the comparative statics for an increase in risk are analogous to those for an exogenous decrease in demand. However, for the multiplicative uncertainty specification (a - 0, a//ap < 0, af3/aXi > 0 in equation (3)), Vp, = 0 using equation (5) and Vx, < 0 for i = 1,2. In this case, well-defined results are obtained for price provided (a) advertising and personal selling are not substitutes (Qx,x2 > 0) and (b) advertising and personal selling influence price-sensitivity symmetrically. Specifically, price increases (decreases) if price sensitivity increases (decreases) when advertising/personal selling is increased. For the generalized uncertainty specification, as in the multiplicative case, the comparative statics for price are well-defined only when advertising and personal selling are not substitutes. We assume in the analysis that advertising and personal selling influence the variance of gross profit margin in the same direction. Consider the case where the variance of the gross profit margin increases with price. Then price decreases if price sensitivity and the variance of the gross profit margin increase (decrease) with advertising/personal selling. Suppose the variance of the gross profit margin decreases with price. Then, price increases if price sensitivity and the variance of the gross profit

88

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

margin with respect to advertising/personal selling change in opposite directions when price is increased. These results show that it is not necessarily a good defensive policy for a manager to increase price when uncertainty increases. This is because an increase in price may actually put the firm in a riskier position than before. Finally, the comparative statics for advertising and personal selling are ambiguous for both the multiplicative and generalized uncertainty specifications. Changes in Productivity Suppose the effectiveness of advertising changes without any change in the effectiveness of personal selling or price. Furthermore, assume that the variability of demand conditional on a given marketing mix is unchanged. Let Q' = Q + 8y(Xi) where ay(X1)/8aX > 0. Then Vp = O, Vx, = ay/aX, > 0 and Vx^ = 0. Consider first the comparative statics for advertising. Then
aXi/a = ID\-I(VpP Vx2x2Vp2 )(+

c.

a/aXI)(

rf)-1.

Now (Vpp Vx2x-2 x2) is a principal minor of second-order for D and hence positive. Hence aX1/a8 > 0. That is, advertising increases (decreases) when its productivity increases (decreases). The comparative statics for price and personal selling, however, depend on the interaction between advertising and personal selling. Suppose advertising and personal selling are complementary. Then the results are well defined if advertising and personal selling influence pricesensitivity symmetrically. Specifically, personal selling is increased. However, price increases (decreases) if price sensitivity decreases (increases) when advertising/personal selling is increased. Suppose advertising and personal selling are independent. Then the effect on price does not depend on the interaction between price and personal selling. In particular, price increases (decreases) if price sensitivity decreases (increases) when advertising is increased. The effect on personal selling is more interesting. Specifically, personal selling increases if increases in advertising/personal selling have symmetric effects on price sensitivity. However, the selling effort is reduced if increased advertising or personal selling influence price sensitivity in opposite directions. Suppose advertising and personal selling are substitutes. Then the results are well-defined provided these marketing instruments influence price-sensitivity in opposite directions. Specifically, price increases (decreases) if price sensitivity decreases (increases) when advertising is increased and price sensitivity increases (decreases) when personal selling is increased. However, personal selling is reduced if increased advertising and increased personal selling influence price sensitivity in opposite directions. Finally, the effect of shifts in the effectiveness of price and personal selling on the marketing mix can be analyzed similarly.
-

THE MARKETING MIX DECISION UNDER UNCERTAINTY

89

d.

Changes in Costs Suppose the cost of advertising changes. Let Wf = W1 + 8 where 6 is the shift parameter. Then V,, = 0, Vx^ = -(1 + rf)- and Vx^ = 0. It is readily verified that the comparative statics for an increase (decrease) in the cost of advertising are analogous to those for a decrease (increase) in the productivity
TABLE 1 ComparativeStatics Propertiesof the Model Increase in Productivity Increase in Cost Exogenous Increase in Demand of Advertising of Advertising Case 1: Advertising Personal Selling Complementary

ap/as:
ax,/as:

+ if a0/axi < o (i = 1,2) - if a?/a, >0 (i = 1,2) + if ap/axi < (i = 1,2) -if aE/axi >0 (i = 1,2) +if a/axi < 0 (i = 1,2) -if ap/axi >0 (i = 1,2)
+ +

Case 2:

Advertising Personal Selling Substitutes

Opposite of a Decrease in
+

ap/a:.

< 0, ay/ax2 > 0 - if aE/ax, > 0, a/ax2 < 0

+if ae/ax,
+

Productivity of Advertising

ax,/a8:

+if aep/ax, < 0, ap/ax2 > 0

- if ap/ax, > 0,

aCe/ax2< 0 ax2/a3:

-if aE/ax, < 0

ay/ax2 > 0 + if ap/ax, > 0, a?/ax2 < 0

- if sign(aEp/axl)= - sign(ae/dx2)

Case 3:

Advertising Personal Selling Independent

ap/ad: ax,/as:
aX2/a8:

+ - Sign(a3e/axl) - Sign(aEp/3x2)

- Sign(a3p/ax1)
+

+ if Sign(ap/axl) = Sign(a3p/3x2) - if Sign(a3p/axl) = - Sign(aEp/ax2)

90

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

TABLE 2 Increased Risk on Pricea (ap/a/) Effect of Additive Uncertainty Multiplicative Uncertainty Generalized Uncertainty Case 1: ad/dap > 0 Opposite of exogenous decrease in demandb + if ap/ax,
-ifaE/axi

> 0 (i = 1,2) < 0 (i= 1,2) > 0

- if ap/axi

> 0 and aa,/axi (i= 1,2)

- if a/axi
Case 2: aoa/ap < 0 + if aEp/axi

< 0 and 3oa/axi < 0

(i= 1,2) > 0 and aa/3Oxi < 0

(i= 1,2) + if aEp/ax, < 0 and aaMo/ax > 0 (i= 1,2)

aAssumes Cov(u, Rm) > 0. bIn this case, dx,/a8 has determinate sign. See Table 1.

of advertising (see previous section). Specifically, advertising is increased (decreased) when the cost of advertising falls (increases). Furthermore, as in the case of shifts in productivity, the effects on price and personal selling depend on the advertising-personal selling interaction. Finally, the effect of a
change in the cost of personal selling on the marketing mix can be similarly determined. 6. Discussion and Summary

This paper develops a marketing mix model under uncertainty using the Capital Asset Pricing Model (CAPM) valuation framework. The model is general because it treats price, advertising and personal selling simultaneously and allows uncertainty to be endogeneously determined. Specifically, we examine the comparative statics of the model to determine how the marketing mix should be modified when market conditions change. Finally, it should be noted that the CAPM value maximization model can be operationalized if time-series data are available on sales, price, advertising and personal selling. In this case, however, the comparative static methodology becomes redundant because specific quantitative information is available regarding the shape of the sales response function. Hence it is possible to quantitatively measure how the marketing mix should be modified for any given change in market conditions. The following steps are necessary in order to operationalize the model (see Larreche and Srinivasan 1982 for details): (a) choose a particular functional form for the demand function in equation (3); (b) estimate the vector of residuals u in the demand function (see Jagpal, Sudit and Vinod 1979, 1982 for empirical applications); (c) estimate the vector of

random rates of return on the market portfolio (R,) for the sample period.

THE MARKETING MIX DECISION UNDER UNCERTAINTY

91

This can be done using the CRSP tapes'? or the Standard and Poor's Index, recalling that Rm is simply the weighted average of all stock prices in the Rm) using the estimated vectors u7and Rm. economy, and (d) estimate Cov(u7, the marketing mix can be chosen to maximize the value of the firm Finally, (see equation (4)) because the unit costs of personal selling and advertising are generally known.
'1The CRSP tapes are available from the Center for Research in Security Prices (sponsored by Merrill Lynch, Pierce, Fenner and Smith, Inc.), Graduate School of Business, University of Chicago.

References
Alchian, A. (1953), "The Meaning of Utility Measurements," American Economic Review, (March), 26-50. Allais, M. (1953), "Le Comportement de L'Homme Rationnel devant Le Risque: Critiques des Postulats et Axiomes de L'Ecole Americaine," Econometrica,(October), 503-546. Brealey, R., and Myers, S. (1980), Principles of CorporateFinance, McGraw Hill, New York. Brick, I., and Jagpal, H. S. (1981), "Monopoly Price-Advertising Decision-Making Under Uncertainty," Journal of Industrial Economics, (March), 279-285. and Statman, M. (1981), "A Note on Beta and the Probability of Default," Journal of Financial Research, (Winter). Chen, A. H. (1978), "Recent Developments in the Cost of Debt Capital," Journal of Finance, (June), 863-877. Christensen, L. R., Jorgensen, D. W. and Lau, L. J. (1973), "Transcendental Logarithmic Production Frontiers," Review of Economics and Statistics, (February), 28-45. Comanor, W. S., and Wilson, T. S. (1974), Advertising and Market Power, Harvard University Press, Cambridge, Mass. Dehez, P. and Jacquemin, A. (1975), "A Note on Advertising Policy Under Uncertainty and Dynamic Conditions," The Journal of Industrial Economics, (September), 73-78. Dorfman, R., and Steiner, P. 0. (1954), "Optimal Advertising and Optimal Quality," American Economic Review, 44, 826-836. Ehrlich, I., and Fisher, L. (1982), "The Derived Demand for Advertising: A Theoretical and Empirical Investigation," American Economic Review, (June). Eskin, G. J., and Baron, P. H. (1977), "Effects of Price and Advertising in Test-Market Experiments," Journal of Marketing Research, (November), 499-508. Fama, E. (1972), "Components of Investment Performance," Journal of Finance (June), 551-568. (1976), Foundations of Finance, Basic Books, New York. - (1978), "The Effects of a Firm's Investment and Financing Decisions on the Welfare of its Security Holders," American Economic Review, (June), 272-284. (1980), "Agency Problems and the Theory of the Firm," Journal of Political Economy, (April), 288-307. and Miller, M. (1972), The Theory of Finance, Holt, Rinehart and Winston, New York. Greenberg, E., Marshall, W. J. and Yawitz, J. B. (1978), "The Technology of Risk and Returns," American Economic Review, (June), 241-251. Horowitz, I. (1970), "A Note on Advertising and Uncertainty," Journal of Industrial Economics, 151-160. Jagpal, H. S., Sudit, E. F., and Vinod, H. (1979), "A Model of Sales Response to Advertising Interactions," Journal of AdvertisingResearch, (June), 41-48. --(1982), "Measuring Dynamic Marketing Mix Interactions Using Translog Functions," Journal of Business, forthcoming. Jensen, M. C. (1968), "The Performance of Mutual Funds in the Period 1945-1964," Journal of Finance, (May), 389-416.

92

HARSHARANJEET S. JAGPAL AND IVAN E. BRICK

Jensen, M. C. and Meckling, W. (1976), "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure,"Journal of Financial Economics, 305-360. Kim, E. H. (1978), "A Mean-Variance Theory of Optimal Financial Structure and Corporate Debt Capacity," Journal of Finance, (March), 45-64. Lambin, J. J., Naert, P. A. and Bultez, A. (1975), "Optimal Marketing Behavior in Oligopoly," European Economic Review, 105-128. Larreche, J. C. and Srinivasan, V. (1982), "STRATPORT: A Model for the Evaluation and Formulation of Business Portfolio Strategies," Management Science, forthcoming. Leland, H. E. (1972), "Theory of the Firm Facing Uncertain Demand," American Economic Review, (June), 278-291. Lintner, J. (1965), "The Valuation of Risky Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets," Review of Economics and Statistics, Vol. 47 (February), 13-37. Meyer, R. A. (1976), "Risk-Efficient Monopoly Pricing for the Multi-Product Firm," Quarterly Journal of Economics, 461-474. Mossin, J. (1966), "Equilibrium in a Capital Asset Market," Econometrica,769-783. Peterman, J. L. and Carney, M. (1978), "A Comment on Television Network Price Discrimination," Journal of Business, (April), 343-352. Rubinstein, M. (1974), "An Aggregate Theorem for Securities Markets," Journal of Financial Economics, 225-244. Sandmo, A. (1971), "On The Theory of The Competitive Firm Under Uncertainty," American Economic Review, (March), 65-73. Sharpe, W. F. (1964), "Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk," Journal of Finance, (September), 424-442. (1980), Investments,Prentice-Hall, Englewood Cliffs, N.J. Vinod, H. D. (1972), "Nonhomogeneous Production Functions in Telecommunications," Bell Journal of Economics and Management Science, (Autumn), 531-545.