GERARD J. TELLIS POINT OF VIEW:INTERPRETINGADVERTISING ANDPRICE ELASTICITIES

GERARD J, TELLIS is asso-ciate professor of marketingat the Schooi of Business Ad-ministratJon, University ofSoulhem California, Los An-geies He has a Ph.D. in busi-ness administration from IheUniversity of Michigan, AnnArbor. He previously earned aB.Sc. and an M B A. andworked as a sales develop-ment manager for Johnson &Johnson His research on ad-vertising response models,pricing models, and pricingstrategies has appeared inmany journals, including theJournal ol Marketing, theJournal ot Marketing Researct).and Marketing Science.

In a recent paper (Tellis,1988a), I noted the "dra-matic" difference in the mean

elasticities estimated for adver-tising and price. Broadbent(1989), in this same issue of theJournal of Advertising Research,criticizes my comparison as defi-cient because it focused only ondemand elasticities and not onthe profitability of potentialchanges in advertising or pricing.I welcome Broadbent's com-ments. They initiate a debate onsome very important aspects ofstrategy in the current environ-ment. The debate will clarifymisunderstandings about elasti-cities and about their implica-tions. However, Broadbent maybe misinterpreting the elasticities.The difference between price andadvertising elasticities may belarger than we previously statedand his recommendations maydepend on too many assump-tions to be practical.

The Meaning ofAdvertising andPrice Elasticities

Broadbent shows that eventhough advertising elasticity issmaller, increases in advertisingmay be more profitable than de-creases in price. For this result,he makes the critical assumptionthat a price cut affects all buyersof the brand. The assumptionmay be invalid and may contra-dict the best principles of goodpricing strategy. He also uses anadvertising change which is 10times larger than the price

change.Basically, a price cut that leads to

an increase in sales needs to affectonly the incremental sales. Incre-mental sales come from twosources: new buyers or more fre-quent purchases by regularbuyers. The essence of goodpricing strategy is to limit theprice cut to only the new buyersor the more frequent purchasesof the older buyers. Attributingthe price cut to all buyers, eventhose who would have bought atthe higher price, is inefficient andprobably not done even by ama-teur managers.

How does one limit price cutsto only incremental sales? Theanswer lies in what I have else-where called differential pricing(Tellis, 1986) and what is oftencalled price discrimination. Theseterms define a set of pricingstrategies by which the pricesetter charges a higher price tothe price-insensitive segment andthe lower price to only the price-sensitive segment. Private pro-viders of service, such as doctors,lawyers, or repair agents, areknown to quote prices on thebasis of what buyers can afford.Retailers of durable goods (auto-mobiles, appliances) frequentlyprovide on-the-floor discounts toprice-sensitive buyers. Manufac-turers and retailers constantlytarget their price promotions(e.g., temporary price discounts[TPDs], rebates, or coupons) to(price-sensitive) segments. Evenwhen buyers would purchasemore if the price was dropped,firms can restrict the discount

40 Journal of ADVERTISING RESEARCH—AUGUST/SEPTEMBER 1989

A D V E R T I S I N G A N D P R I C F , E L A S T I C I T I E S

. . . a price cut that leadsto an increase in salesneeds to affect only the

incremental sales.

only to incremental sales byusing a strategy of volumediscounts.

How exactly can price promo-tions be targeted to price-sensi-tive segments? These promotionscan be targeted to price-sen si tivesegments by use at certain times,places, or media; price-sensitivesegments in turn look for thesepromotions because they havethe time, interest, lower re-sources, or more information. Forexample, end-of-season dis-counts are targeted to buyerswho are willing to wait for thelower price. Random TPDs aretargeted to price-sensitive seg-ments who can afford to watchand wait for them. Coupons arerandomly distributed and tar-geted to consumers who arewilling to spend time and effortto search, clip, and collect them.Indeed, researchers have shownthat coupons or TPDs are pri-marily means by which firms dis-criminate between price sensitiveand insensitive segments (Nara-simhan, 1984; Varian, 1980).Some price-insensitive regularbuyers may be lucky enough toget the brand at the lower price,even though they would havebeen willing to pay the higherprice. The success of goodpricing is to minimize this seg-ment of lucky price-insensitivebuyers while maximizing thenumber of price-sensitive buyers,who might otherwise buy an-other brand.

Therefore, it is both norma-tively unwise and practically un-true that a price cut to attract in-cremental sales must be appliedto the whole market. Without

this critical assumption, pricecuts could be more profitablethan advertising increases underthe same assumptions thatBroadbent makes. To show thisfact, I rework his main example,without changing any of thecosts and response figures, butonly making the more reasonableassumption that price cuts needapply to only the incrementalsales. (See Table 1.)

In his example, Broadbentchooses to compare a 10 percentincrease in advertising with a 1percent decrease in price, a 10 to1 advantage for advertising. Hisrationale "to get more sales (withthe advertising change) than withthe price change" is not compel-ling because it is an arbitrary cor-rection for the smaller advertisingelasticity. Normally, when onecompares two instruments, theymust be at the same level. More-over, as one increases the changein one variable, because of de-clining returns, the percentagechange in sales is likely to belower. So the assumption of aconstant (linear) response rate fora change of 10 percent as for achange of 1 percent is untenable,while that of a logarithmic re-sponse in the absence of empir-ical evidence is arbitrary. Even

. . . it is both normativelyunwise and practically

untrue that a price cut toattract incremental sales

must be applied to thewhole market.

with a 10 times larger change inadvertising and the assumptionof a linear response rate, pricecuts could be more profitablethan advertising increases, as 1have shown in Table 1.

In conclusion, given a marginof 100 percent on variable costs,a price elasticity of -1.76 per-cent, and an advertising elasticityof .22 percent, both price cuts oradvertising increases would in-crease profits; but a price cutwould be more profitable than anadvertising increase even thoughthe relative price cut was one-tenth as large as the relative in-crease in advertising. In general,because price cuts could be tar-geted to price-sensitive segmentswhile broadcast advertisingcannot be so targeted, and be-cause price elasticities are higherthan advertising elasticities, pricecuts may be a better means to in-

Table 1Rework of Broadbent's Main Example

Condition

Volume in cases

Price

Revenue $,000

$10 per unit var.

Cost $,000Fixed costs $,000

Advertising $,000

Total costs $,000

Contribution $.000

Original

5.000,000

$20

100,000

50,000

20,000

7,000

77,000

23,000

1% Price cut

5,088,000

$20 for old$19.80 for nevi

101,742

50,880

20,000

7,000

77,880

23,862

10% Advertisingincrease

5.110,000

$20

102,200

51,100

20,000

7,700

78,800

23,400

Journal of ADVERTISING RESEARCH—AUGUST/SEPTEMBER 1989 41

A D V E R T I S I N G A N D P R r C E E L A S T I C I T I E S

crease short-term sales and profits.A price cut always makes ashort-term contribution to fixedcosts if the total margins fromnew buyers exceeds the revenueloss due to regular buyers buyingat the lower price.

New Evidence onAdvertising andPrice Eiasticities

Our conclusion that the adver-tising elasticity is .22 comes fromAssmus, Farley, and Lehmann's(1984) review of 122 publishedestimates; the conlusion that theprice elasticity is -1.76 is basedon my review (Tellis, 1988a) of367 estimates of price elasticity.However, both these summaryfigures are arithmetic means ofbiased and unbiased estimates. Ifwe correct for biases in estimates,the advertising elasticity may belower and the price elasticity maybe higher at -2 .5 (TelUs, 1988a).Ir\ addition, new evidence sug-gests that aspects of price elas-ticity may be even higher, whileadvertising elasticity may not beany higher.

Recently, Wisniewski andBlattberg (1988) carried out anextensive field study of grocerysales response to changes in storepromotion activity. The observa-tions were weekly scanner datafrom 44 stores, 10 categories,over 200 brands, and over 1 to 6years. They found that theaverage regular price elasticitywas 2.49, identical to the cor-rected mean from my review.However, they found that thesales elasticity of deals is muchhigher, from 4.7 to 9.0. Becausedeals are primarily a short-termchange in the offer price de-signed for price-sensitive seg-ments, this figure more strictlyreflects the elasticity of incre-mental sales that I referred toearlier. Other studies byGuadagni and Little (1983) andGupta (1988) on scanner data also

show high elasticities for pricepromotions (or deals).

At the same time, new studiesthat have examined the sensi-tivity of household purchases ofgrocery products to householdexposure to television advertisingusing the new scanner-TVmeteror "single source" data suggestthat the effects of TV ad exposureare no higher than those ob-tained by the older data (e.g.,Broadbent, 1986; Tetlis, 1988b).The new scanner-TVmeter dataare unobtrusively collected at amuch lower level of aggregationthan the previous sales-adver-tising data which were often ag-gregated at the quarterly level.Such aggregation inflates esti-mates of the effects of advertising(Clarke, 1976; Russell, 1988).Thus we may want to put moreconfidence on the new estimates.

We are left with the conclusionthat newer estimates of priceelasticity may be much higherthan before, while newer esti-mates of TV advertising elasticitymay not be higher than before.For example, if price-promotionelasticity were 8, and advertisingelasticity were .22, they coulddiffer by 40 times. These twofacts together further enhance therelative importance and potentialprofitability of price cuts relativeto advertising increases.

The Profitabiiity ofAdvertising andPrice Changes

In my earlier paper (Tellis,1988a) I refrained from makingany predictions about the profit-ability of various strategies be-cause any such strategies arecontingent on numerous otherfactors about which we do nothave adequate knowledge ingeneral. Particular firms mayhave information about suchfactors in particular situations,and then such analyses may bejustified. But, because I did re-

spond to the above example byBroadbent, we have to makeclear under what assumptionsthey hold. These are discussedunder four headings: cost struc-ture, competitive reaction, indi-rect effects, and long-term effects.

Cost Structure. The structureof costs itself directly affects theprofitability of various strategies.For example, if the per unitmargin is small, than even asmall drop in price could easilyconsume the available marginand any incremental sales wouldnot be profitable; similarly, whenmargins are small, the new salesfrom an advertising increase maynever compensate for the cost ofthat increase. On the other hand,if firms have high absolute pricesor high absolute advertising ex-penditures to begin with, theneven small percentage decreasesin price or increases in adver-tising may be too large relative tothe available margins for profit-able actions.

Competitive Reaction. In mostcases, competitors are likely toreact to new strategies. Suchcompetitors could be larger orsmaller, have more or less re-sources, have larger or smallerresponse elasticities, and couldrespond with larger or smallerchanges in price or advertising.Each of these factors could radi-cally alter the results of our anal-ysis, and therefore mere specula-tion of the possibilities would notbe useful.

indirect Effects. Advertisingmay have indirect effects whichare different and possibly moreimportant than the more notice-able direct effects. First, adver-tising could increase brand loy-alty. Such increases in loyaltymay result not only in more fre-quent purchases of the favoritebrand (reflected in the adver-tising elasticity) but also in de-creases in price elasticity itself.The latter effect is less easily ob-served and could reduce the

42 Journal of ADVERTISING RESEARCH—AUGUST/SEPTEMBER 1989

A D V E R T I S I N G A N D P R I C E E I . A S T I f 1 T ! !-: S

need for price cuts or other morecostly promotions. Second, ad-vertising could increase theimage or value of the brand,which over time could help to el-evate price. Third, manufac-turer's advertising may motivateretailers to better stock or pro-mote the advertised brandswhich could lead to increasedsales. On the other hand, adver-tising may increase brand visi-bility, competitive reaction, and,consequently, consumers' priceelasticity. All these effects of ad-vertising are more subtle and re-quire careful documentation andanalysis.

Long-term Effects. The priceand advertising elasticities dis-cussed above are indices of im-mediate short-term sales re-sporise to these variables. Bothadvertising and price promotionsmay have delayed or long-termeffects which are not reflected intheir short-term elasticities. Ad-vertising's effect on purchasesmay be delayed due to delays inbuyer conviction about the mes-sage, the buyer's purchase occa-sion, or message diffusion amongother buyers. On the other hand,repeated price cuts could lowerbuyers' perceived quality of thebrand; or repeated price cutscould increase the number ofprice-sensitive buyers who planto purchase only when the brandis on discount. Any of thesefactors would support heavieradvertising and caution againstindiscriminate price cuts. Becauseprior research often sufferedfrom biases of data aggregation,we do not have conclusive evi-dence about any of these effects.

The ChaiiengeThe key question facing adver-

tisers and reseachers is whether

short-term advertising and priceelasticities truly reflect the totaleffects of these two marketingvariables and whether their dif-ference is really as large as thenew evidence suggests. In partic-ular, is TV advertising primarilya means of inducing current pur-chases or does it play an entirelydifferent role from the other mar-keting variables? While past re-search has provided some an-swers about the short-term elas-ticity of price and advertising, theabove discussion suggests thatmany questions remain unan-swered. For this reason, arbitrarymanipulations of numbers in ar-bitrary problems are unlikely tobe enlightening. What we need isa solid theory for the role of ad-vertising, especially TV adver-tising, in contemporary marketsand good evidence to supportthis theory. Studies based on im-proved data now available fromscanner-TVmeter panels or fromformal field experiments wouldbe most helpful. •

References

Assmus, Gert; John U. Farley;and Donald R. Lehmann. "HowAdvertising Affects Sales: Meta-Analysis of Econometric Re-sults." Journal of Marketing Re-search 21, 1 (1984): 65-74.

Broadbent, Simon. "Quantifyingthe Sales Fffects of Advertisingon Individuals." In TranscriptProceedings of the First Annual ARFFall Conference. New York: Ad-vertising Research Foundation,1986.

. "What Is a Small Adver-tising Elasticity?" journal of Ad-vertising Research 29, 4 (1989):37-39.

Clarke, Darral G. "Sales-Adver-tising Cross-Elasticities and Ad-vertising Competition." journal ofMarketing Research 10, 3 (1973):250-61.

Guadagni, Peter M., and JohnD. C. Little. "A Logit Model ofBrand Choice Calibrated onScanner Data." Marketing Science2, 3 (1983): 203-38.

Gupta, Sunil. "Impact of SalesPromotion on When, What andHow Much to Buy." journal ofMarketing Reseach 25, 4 (1988):342-55.

Narasimhan, Chakravarthi. "APrice Discrimination Theory ofCoupons." Marketing Science 3, 2(1984): 128-47.

Russell, Gary J. "RecoveringMeasures of Advertising Carry-over from Aggregate Data: TheRole of the Firm's Decision Be-havior." Marketing Science 7, 3(1988): 252-70.

Tellis, Gerard J. "Beyond theMany Faces of Price: An Integra-tion of Pricing Strategies." journalof Marketing 50, 4 (1986): 146-60.

. "The Price Elasticity ofSelective Demand: A Meta-Anal-ysis of Econometric Models ofSales." journal of Marketing Ke-

/ 25, 4 (1988a): 331-41.

. "Advertising Exposure,Loyalty and Brand Purchase: ATwo Stage Model of Choice."Journal of Marketing Research 15, 2(1988b): 134-44.

Varian, Hai. "A Model of Sales."The American Economic Review 70,4 (1980): 651-59.

Journal of ADVERTISING RESEARCH—AUGUST/SEPTEMBER 1989 43