|
Article Excerpt
We investigate differences in the dynamics of marketing decisions across geographic markets empirically. We begin with a linear-quadratic game involving forward-looking firms competing on prices and advertising. Based on the corresponding Markov perfect equilibrium, we propose estimable econometric equations for demand and marketing policy. Our model allows us to measure empirically the strategic response of competitors along with economic measures such as firm profitability. We use a rich dataset that combines sales, marketing mix, factor cost, and advertising cost data for eighteen geographic markets in the frozen entree category. We find that larger markets tend to be less price-sensitive and more profitable than smaller markets. We also find evidence of positive carryover of own advertising on own demand. In terms of consumer substitution patterns, we find that the role of advertising (in our data) seems to be more category-building (complementary) than share-stealing (competitive). The complementary role is stronger in larger markets. On the supply side, we find that firms make smaller adjustments to own advertising as goodwill goes up. Consistent with crossadvertising effects on demand, firms make smaller (larger) adjustments to advertising in response to competitive goodwill in the less competitive larger (in the more competitive smaller) markets. Finally, we find that consumer welfare decreases (increases) in larger (smaller) markets when firms move to a zero-advertising regime.
Key words: competition; advertising; multiple geographic markets; structural models; Markov perfect equilibrium; dynamics; packaged goods
History: This paper was received September 4, 2001, and was with the authors 10 months for 4 revisions; processed by James Hess.
1. Introduction
Consumer goods manufacturers typically compete in various geographic markets. Recently, research has begun to focus on geographic differences in responses to marketing-mix variables (Boatwright et al. 2004) and the potential dependencies of marketing efforts across neighboring regional markets (Bronnenberg and Mahajan 2001). In consumer goods markets, marketing policies typically consist of three types of strategic instruments: prices, promotions (in this paper, we use "promotion" to refer to nonprice promotion only), and advertising. These instruments usually have strong carryover effects in demand (Leeflang and Wittink 1992, Clarke 1976), requiring firms to be forward looking. Thus, competitive marketing decisions are not only market specific, they are also inherently dynamic. The extant empirical research in marketing has not formally disentangled the supply and demand responses to dynamic marketing effort. Little attention has been paid to the measurement of market-specific differences in optimal marketing decisions, especially in a dynamic environment.
We wish to empirically measure the long-run profitability of marketing effort, pricing, and advertising for a consumer packaged goods (CPG) category and to describe the differences in such efforts across the largest U.S. markets. A novel feature of our empirical analysis is the data we use to estimate the model, collected from the frozen entree industry. Our data comprise three years of weekly sales, prices, promotions, and advertising gross ratings points (GRPs) for 18 major U.S. city-markets. We supplement these data with market-specific information on both production costs (wages and factor prices) and advertising costs (cost per GRP), which provide exogenous sources of variation in the marketing variables.
To guide us in the formulation of the econometric specification, we begin with an economic model of profit-maximizing firms. The model accounts for a decision-making process that spans several strategic marketing instruments and a long-run planning horizon (Vilcassim et al. 1999, Slade 1995). In the model, we account for firms' strategic responses to intertemporal changes in their own and their competitors' marketing effort. We base our econometric specification on the corresponding Markov perfect equilibrium (MPE) in prices and advertising. After fitting the model to marketing data, we then carry out policy experiments to quantify the value of advertising to firms and consumers.
Our results show that the pattern of demand, margins, and profits varies significantly across markets. We find that larger markets tend to be less price sensitive and more profitable than smaller markets in this industry. With respect to advertising, we find that own current and past advertising has a positive effect on own demand. The role of advertising (in our data) seems to be more category building (complementary) than share stealing (competitive). The complementary role of advertising is much stronger in the larger markets relative to the smaller markets, possibly because media availability in smaller markets is much lower, leading to firms behaving more competitively with regard to advertising.
On the supply side, we find that firms make adjustments to own advertising in response to current goodwill--as goodwill goes up, advertising adjustments are smaller. We find that the adjustment to advertising as a function of competitive goodwill is consistent with the cross-advertising effects on demand--we find larger adjustments in the more competitive smaller markets. In particular, the evidence suggests that when competitive advertising is complementary, firms lower their own advertising in response to competitors' goodwill. Because smaller markets tend to have less complementarity, advertising tends to be more competitive. Firms also condition their adjustments of the cost of GRPs in each market, with the three firms adjustments' depending on the cost of different dayparts.
We use the model to compute the change in long-run profits in response to current investment in advertising goodwill. The direct effect of advertising (arising through the carryover in demand) is positive for large markets for all the products in our data. This is not true in the small markets. Interestingly, we find the direct effect of advertising to be an order of magnitude larger than the strategic effect (arising through competitive reactions). Thus, the total effect of advertising tends to be driven more by the direct influence of carryover effects than the strategic influence of competitive response.
We also look at the welfare implications of setting advertising to zero each period. Consumers benefit from advertising in smaller markets where advertising is fairly competitive. Intuitively, competition on advertising erodes market power and thus lowers equilibrium prices. In contrast, in larger markets where advertising is more complementary, we find that consumers are harmed by advertising. While advertising tends to expand the product category, in equilibrium firms tend to free-ride off one another's advertising investments as competitive advertising increases own market power and thus prices.
The rest of the paper is organized as follows. We describe our model and the econometric specification in [section] 2. Section 3 describes the data. We discuss the results in [section] 4 and conclude in [section] 5.
2. Model and Econometric Specification
We present a dynamic oligopoly model to describe the frozen entree industry. We base the econometric specification on the corresponding equilibrium conditions of the model. The link to a model helps us control for the endogeneity of strategic variables during estimation. It also allows us to measure empirically the long-run profitability of advertising while controlling for competitive response.
2.1. Aggregate Demand
We model consumer demand for the J brands using the convenient linear demand system, a model that has been used frequently in the marketing literature (see, for instance, Roy et al. 1994, Vilcassim et al. 1999, Kadiyali et al. 2002). Specifically, we model total consumer demand for brand j in a market at time t as
(1) [Q.sub.jt] = [[alpha].sub.j] + [J.summation over (k=1)][[beta].sub.jk][p.sub.kt] + [[phi].sub.j][F.sub.jt] + [J.summation over (k=1)][[delta].sub.jk][G.sub.kt] + [[gamma].sub.j][y.sub.t] + [[zeta].sub.jt],
where [p.sub.k] is the price of brand k, [F.sub.j] is the current promotion level for brand j, [G.sub.k] is the current stock of accumulated advertising goodwill for brand k, y is the total income in the market, and [[zeta].sub.jt] is a random demand shock capturing aggregate demand-shifting variables that are unobserved by the researcher. As we assume demand and competition are independent across markets, we omit the market subscript to simplify notation. An important feature of this demand system is that it allows for different own and cross-effects of advertising on each of the demand curves because we do not impose any restrictions on the sign of the advertising effects. The existing literature has documented that, in a multiproduct market, a competitor's advertising could be category building or share stealing (e.g., Roberts and Samuelson 1988, Vilcassim et al. 1999, Gasmi et al. 1992). We also include own promotion as one of the demand shifters (e.g., Boatwright et al. 2004, Montgomery 1997). (1) The set [{[[alpha].sub.j], [[beta].sub.jk], [[phi].sub.j], [[delta].sub.jk], [[gamma].sub.j]}.sup.J.sub.j, k=1] consists of parameters to be estimated. (2)
We assume that aggregate consumer response to advertising variables exhibits carryover effects (e.g., Clarke 1976, Lodish et al. 1995). To capture these long-run effects, we model current accumulated goodwill as [G.sub.jt] = [ad.sub.jt] + [[summation].sup.[infinity].sub.[tau] = 1][[eta].sup.[tau]][ad.sub.jt-[tau]], which consists of current advertising effort measured in GRPs, the historic goodwill stock, and [eta] [member of] (0, 1), a depreciation factor (assumed to be the same for all firms) for past advertising. Hence, we can think of goodwill as
(2) [G.sub.jt] = [ad.sub.jt] + [eta][G.sub.jt-1].
2.2. Prices and Advertising
We now discuss the supply side of the model, consisting of competing manufacturers setting both prices and advertising each week. We do not model the downstream retailers' decisions, which could be interpreted as assuming a competitive retail environment or constant retail mark-ups (e.g., Vilcassim et al. 1999). The firms' decisions consist of setting prices, [p.sub.jt], and adjusting their advertising levels, [DELTA][ad.sub.jt], on a market-by-market basis. Our decision to model advertising decisions as an adjustment, versus a level effect, was based on our understanding of media planning in this industry. We expect high-frequency (weekly) changes in advertising to reflect an adjustment to a baseline rate determined at a much lower frequency (quarterly). Our understanding is that firms determine a baseline ad rate well in advance. However, over time, they monitor their own and their competitors' advertising behavior and brand performance. As a result, they periodically adjust their advertising decisions (up or down) accordingly. Similar adjustment models have been used for pricing (Slade 1995), demand (Karp and Perloff 1989, Roy et al. 1994), and market shares (Sorger 1989).
To determine the appropriate solution concept for the model, we spoke with several industry experts about media planning in CPG industries. This discussion indicated that firms adjust marketing instruments on a periodic basis in this category in response to changes in the market. Specifically, even though manufacturers choose a network advertising schedule over an accounting period (typically a quarter), they use spot markets to make adjustments to their chosen advertising levels on a weekly basis across markets (as discussed later, expenditure on spot TV advertising represents a high proportion of all advertising expenditure in this industry). These adjustments are based on tracking of recent own and competitive advertising efforts, where the latter are monitored to account for competitive dilution of own ad efforts. In the model below, we make the additional assumption that firms know the goodwill formulation process (2).
Given this industry behavior, we use...
|
