Entry/Exit Demand Analysis

ABSTRACT - Past methods of measuring consumer response to the price of a branded good are reviewed and critiqued. A new approach- Entry/Exit Demand Analysis--is described. The method borrows from and improves past methods. Some initial evidence about the technique's test-retest reliability is presented.

Citation:

Peter R. Dickson and Alan G. Sawyer (1984) ,"Entry/Exit Demand Analysis", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 617-622.

Advances in Consumer Research Volume 11, 1984      Pages 617-622

ENTRY/EXIT DEMAND ANALYSIS

Peter R. Dickson, The Ohio State University

Alan G. Sawyer, The Ohio State University

ABSTRACT -

Past methods of measuring consumer response to the price of a branded good are reviewed and critiqued. A new approach- Entry/Exit Demand Analysis--is described. The method borrows from and improves past methods. Some initial evidence about the technique's test-retest reliability is presented.

INTRODUCTION

One of the most troublesome problems that confronts management is estimating the responsiveness of demand to changes in price. In the past, experienced managers may have been able to intuitively assess price elasticity based on a history of strategic price increases and decreases. However, over the last decade, many of the street-wise managers may have lost their touch because consumer demand has become much more volatile. Inflation, recession and cultural trends are three factors that have made the tracking of demand curves and price elasticity an even more elusive exercise.

Inflation has changed the real purchasing power of the dollar and consequently required consumers to frequently change their reactions to price. Consumers have to adjust to both changes in the value of the dollar and changes in income. Inflation's effect on real income has varied drastically across consumers. Consequently, some consumers, because of changes in their real income and/or their awareness of the general depreciation of the dollar, react more systematically and quickly than others and this progressive adjustment to changes in purchasing power has been a major destabilizing factor. A further and infrequently discussed effect of inflation is on relative prices. Economists have convincingly demonstrated that the dispersion in relative price increase is closely correlated with the increase in the general price level (Vining and Elwertowski 1976). In periods of high general inflation, the variance in the increase in prices of individual goods increases dramatically. Some price increases are dramatic, much greater than the general inflation rate, while other individual price increases are somewhat less than the general inflation rate. A wide variation in price changes the relative prices of goods and this changes the demand curve for individual items. Whether the change in relative prices is a short term effect (i.e., goods lagging in price will catch up over time) or is a permanent structural change, the consumers will adjust their budgeting and spending behavior to varying degrees and again at varying speeds.

Recession has reduced general consumer confidence and changed many consumers' attitudes toward what is a necessity and what is a luxury. The short-term economic embarrassment of many households may have only temporarily changed the price elasticity of goods. On the other hand, short-term changes in consumption behavior, such as doing without desserts or switching to cheaper substitutes such as generic brands, may result in permanent structure changes in consumer attitudes and demands. The automobile industry is fearful that middle class America may have learned to live with older automobiles and in the future will find other ways of spending their discretionary income than replacing their nearly new model. Recession has also influenced competitive behavior at all levels of the distribution channel from manufacturer down to retailer. The increase in price deals offered to move slow moving inventory has inevitably increased buyers' price sensitivity.

Finally, there are major structural changes in our economy that are occurring as a result of the introduction of new technology and changing values (see Naisbitt 1982). These changes are affecting some product groups more than others but whatever the extent of such trends, they are inherently destabilizing.

The me sage is clear. We must strive even harder to find efficient and valid ways of monitoring demand and price elasticity. This paper reviews some of the current techniques that attempt to directly measure price elasticity. It does not discuss the use of elegant econometric models fitted to historical data; data which may have little relevance to the realities of today's marketplace. A new method is then presented, called Entry-Exit Demand Analysis. It is a refinement of previous techniques and an advance in terms of its face validity, its operational ease, and its efficiency.

PREVIOUS MEASURES OF CONSUMER RESPONSE TO PRICE

Most of the early published work on measuring consumer's subjective response to price assumed that two price limits exist. The upper limit is the price above which the product was judged to be "too dear" and by implication too expensive to purchase. The lower limit is the price below which the quality of the product was inferred to be suspect and by implication not worth the risk of purchasing. Stoetzel (1969) established these price points for various commodities in studies undertaken in France in the late 1940's. The focus was on a product group such as "a radio set" and not on a particular brand or item that could be examined and whose quality could be directly assessed. Two price acceptance curves were generated from a sample of consumers and the difference between these curves enabled the calculation of the percentage of the sample who considered any particular price acceptable.

Adams (1969) continued this research approach through the early 1950's estimating the optimum price, at least in terms of potential demand, for actual items that are physically presented to respondents, rather than for a generic product class. He observed that the two distributions of lower and upper price boundaries were of similar shape, approximately that of the cumulative normal distribution. Whether the distribution of responses around a central uncertainty over current prices/cost or differences in the actual perceived value of the item was not explored and has not been studied.

Gabor and Granger (1969) compared the potential demand curve derived from upper and lower price limits to the distribution of reported price last paid. The difference between the curves was argued to reflect a lack of adjustment of the consumer's perceived "fair" or "just" price to the prevailing price. This analysis utilized a generic product group level of analysis and implicitly assumed that the two response curves were not for different brands or items within a product group.

While intuitively appealing, the above approach first developed by Stoetzel (and not by Gabor and Granger as frequently implied in the current literature) has two major weaknesses. It is open to question whether today a consumer's lowEr price resistance point comes into play very often, if ever. Consumers can almost always directly assess the quality of goods by inspection or by in-store trial. The great majority of buyers are aware of the existence of manufacturers' warranties, regulated minimum quality performance standards (such as the Underwriters Laboratory Seal for electrical goods) and retailers' responsibility to sell goods of merchandisable quality. A manufacturer's or retailer's name is also often used to infer quality. As a result, consumers are seldom going to use low price as an indicator of unacceptable quality, in the presence of these other cues, and will consequently not be tardy in accepting low prices. They may assume that a $2.00 G. E. Electric Hairdryer (after a manufacturer' s rebate) on sale at a discount store is a discounted line, is being used as an outrageous loss leader, or perhaps even has been priced incorrectly. Consumers will not, however, reject such a price as unacceptable given the manufacturer's or retailer's assurance of quality. An exception is for conspicuous consumption items where the cheapness of the price paid for the good,not to be confused with the cheapness of the good itself) would be apparent to others. It can readily be demonstrated that the lower price limit drops dramatically (often to zero) as assurances are provided as to the integrity and quality of goods.

The second concern which is somewhat related to the first point is the dubious value of undertaking the analysis at the generic product level. When respondents are asked to indicate the least and most they would pay for "a radio", quality and feature preferences are confounded with price acceptability. The results are of almost no value to manufacturers of particular radios because the demand curve is based on opinions about the price acceptability of all radios rather than any single radio. To make the judgments more valid, not only should an actual model be the focus of the study but information about it and the other alternatives available need to be provided. When this is done, the consumer' s responses are more likely to reflect his or her behavior in the marketplace. In short, the shopping circumstances and/or environment need to be simulated.

Such an approach was pursued, if not pioneered, by Pessemier (1960) in the late 1950's. He simulated shopping for toothpaste (7 brand choices), cigarettes (10 brand choices), toilet soap (11 brand choices) and headache remedies (6 brand choices) using 228 student subjects. The choice sets included all the brands available at the student bookstore. Each participant was presented with a photograph of each brand and its price and was asked to make a choice. The student's preferred brand was then increased in price by various intervals and with each price increase a new purchase choice was made. After raising the preferred brand's price in steps, all but the participant's preferred brand were reduced in price by the same amount, in steps, and fresh purchase choices were made. In total, 20 simulated shopping choices were undertaken where the price of the preferred brand systematically rose or the price of all of the initially unchosen brands systematically fell. To increase the realism of the exercise, the students were informed that one in 20 of them would receive the actual merchandise and change from $1.75, as called for by their decision in one of their 20 shopping choices.

As a matter of interest, the brand demand curves derived from the above point estimates indicated that students were least sensitive to cigarette price changes but for other product classes price changes of one or two cents had a considerable effect on demand and marketshare. Additional price increases or decreases had little additional effect on toilet soap and headache remedy brand loyalty (suggesting the existence of hard-core brand loyalty). On the other hand, additional price increases and decreases continued to substantially influence the demand for specific brands of toothpaste

This simulation approach, as acknowledged by Pessemier, understates the effect of reducing the price on a particular brand because the participant is presented with a situation where several brands are reduced by the same amount at the same time.- It can, however, be argued that it is likely that some other brands will respond to a brand's price decrease, thus increasing the validity of such a general price manipulation. The bias also understates rather than overstates the effect of a price decrease and is hence probably more acceptable to management. The alternative would have been to increase the number of shopping simulations 4 to 5 fold and to have reduced the price of the choices one by one.

A trade-off brand switching approach almost identical to that developed by Pessemier has been used by Market Facts and described by Jones (1975). The percentage of respondents that would buy a particular brand in a particular pricing situation was derived. A transition matrix was calculated which showed how brands would benefit or be burt by price decreases or increases. From this information and the use of appropriate cost accounting the profitability of particular pricing strategies was also computed. The technique was argued to enable the simulation of a great number of market situations and to be very simple, only requiring simple choices rather than subjective evaluations such as likelihood of purchase measure that must be somehow transformed into market shares. The major concern was the possibility that the repetitiveness of the task results in respondent fatigue, game-playing and results of questionable validity.

Mahajan, Green and Goldberg (1982) have taken the Pessemier-Jones methodology and used an elegant and very powerful analysis based on a conjoint analysis to generate elasticities of demand. The only flaw is that their results are based on rather suspect data. Rather than choosing a single brand out of the choice set, the respondent is asked to assign constant-sum probabilities that each alternative will be chosen. However, in an actual choice setting the probabilities will very likely be an array of the form (1.00, 0.00, 0.00,...). Their measure encourages participants to assign non zero probabilities to options that will never be preferred or chosen over more attractive alternatives. and its validity is consequently suspect. [Paul Green (personal communication) feels that subjects are unsure about their choice and that the non-zero probabilities truly reflect intention and do not distort their actual purchase intentions. Our concern remains that an actual discrete choice decision should not be modelled using choice likelihoods.]

The problem of repetitive shopping and artificial response sets was handled by Abrams (1964) by exposing matched representative, mail-panel samples of respondents to only a single treatment condition. Each group was exposed to a different pricing strategy for a test refrigerator brand, imbedded in newspaper advertisements. Demand was demonstrated not to be responsive to $10 and $20 reductions but a sharp break increase in demand occurred when price was reduced by 530. The incremental effect of reducing the price by another $10 to $40 off was not as marked. The study also revealed the competitive brands that would be most vulnerable to the price decreases.

In a unique and telling comparative study, Stout (1969) concluded that estimates of demand based on simulated shopping may well be invalid. The subjects were either presented with photographs and prices of the alternatives before entering a supermarket or played at shopping for the alternative in an especially equipped trailer simulation. These two studies produced conflicting results and both indicated less price sensitivity than the results of an actual in-store experiment. Only the latter produced changes in demand in response to changes in price that were statistically significantly different from zero. Stout suggested that simulation studies fail to create a natural buying environment and also fail to properly capture the consumer's concern over and sensitivity to price when they are actually spending part of their own income. However, Sawyer, Worthing and Sendak (1979) contested Stout's pessimism about the validity of laboratory simulations. None of Stout's lab simulations asked respondents to spend their own money. Sawyer et al. suggested that a crucial factor may be the use of a financial sacrifice by consumers for choosing higher priced products. Such sacrifice can be simulated by giving the subjects a sum of money and allowing them to keep the difference between that sum and the purchase price of their chosen brand as in Pessemier (1960). The validation results of Nevin (1974) and Gabor, Granger and Sowter (1970) lend credence to the optimism of Sawyer et al.

ENTRY/EXIT DEMAND ANALYSIS

The approach we have developed is based on the behavioral assumption that when a buyer considers any item within a choice set s/he has a major critical price point that determines behavior. For a brand currently purchased. the critical price point is the price which if exceeded will lead to rejecting the brand and either the choice of an alternative brand or item from the choice set or the rejection of the entire offering. This is the buyer-brand exit price. If the buyer is not currently Buying a given brand, then the critical price point is that price below which the buyer will abandon his or her current choice and will choose the brand. This is the buyer-brand entry price. Our thesis is that the measurement procedure must recognize the current approach-avoidance relationship between the consumer and a choice option. If the initial response or predisposition is positive then the question must determine the sharp break price point at which the relationship will become negative and, if negative, then the question must determine the sharp break price point at which the relationship will become positive This requires two distinct questions which recognize the distinct current orientation of the buyer.

No strong assumptions are made about the stability of these points. Indeed it is expected that they will depend very much on the alternatives that are available, the perception of their prices, and the perception of their benefits. Casual observation of the behavior of supermarket shoppers suggests that such price points do exist. Shoppers will often exclaim out loud to a friend or even to themselves that the price is very,'too high or (less frequently) that the price is a bargain. This is particularly noticeable with fresh fruits and vegetables whose price fluctuates seasonally quite dramatically. Such reactions are evidence that the price of an item has exceeded the exit break point or the price of an item has dropped below the entry point. The actual measurement procedure we propose is listed step by step in Table 1.

TABLE 1

ENTRY/EXIT DEMAND ANALYSIS

Steps

1. Choose the target groups of buyers whose price sensitivity is of interest.

2. Choose the situation usage/purchase situations that are of interest. The usage setting and the purchase circumstances may influence price sensitivity (e.g., having to purchase an item at a convenience store late at night rather than purchasing it at a supermarket during the week).

3. Choose the set of alternatives that are to be offered to the target group. Ensure that this offering is a valid representation of the offering available at the retail outlets at which the target group shops in the usage/purchase circumstance.

4. Describe the simulated shopping exercise to the subjects--detailing the shopping circumstances and the offering.

5. Present the subjects with pictures and starting prices of the set of alternatives. In addition to price and product information provided alternatives give each item an alphabetical code. Give consideration to using several different randomized orderings of the array of alternatives.

6. As the base question, "Which brand/item would you purchase given the prices as listed?"

7. Then proceed with the following question which is repeated for each item/brand in the choice set. (The example is for item A, but the only change made is to change the alphabetic code throughout ) Require the respondents to refer back to the initial information on the offering before answering each question in this series.

"If you chose brand A in question 1, please answer this question.

If you chose brand A, imagine that brand A, alone, increased in price and all of the other brands did not change their stated price. What is the most you would be prepared to pay for brand A? That is, if it were even one cent more than the price you state you would no longer choose brand A.

If brand A was any higher than $ (please write price), I would no longer buy it.

What brand would you buy if brand A increased to a price above the limit you have just stated (if you would not buy any of the other brands, write none). The brand I would switch to would be

If you did not choose brand A in question 1, please answer this question.

If you did not choose brand A, imagine that brand A, alone, dropped in price and all of the other brands did not change their stated price. At what lower price would you buy brand A, rather than the brand you initially chose.

I would change my choice and buy brand A when its price dropped to $ (please write price)

If you would not buy brand A at any price please write never in the space for the price."

8. Based on the responses, construct demand curves for each alternative that measures the percent of the sample who would choose the alternative at different prices.

9. Consider undertaking test/retest reliability checks over an interval of time and measuring the sensitivity of the results to changes in the target group, purchase circumstances and range of the offering.

ENTRY/EXIT ANALYSIS EXAMPLE

Students' price sensitivity to toothpaste and hand-soap was studied using Entry-Exit Analysis. The choice sets consisted of seven brands of toothpaste and eight brands of soap selected from a screening survey that identified the most popular brands. One hundred and nine students were presented with a series of color slides projected onto a screen in the classroom. The first slide was a color photograph of the store where many of the students regularly shopped. The second photograph was of the entrance-exit doors of the store, the third a photograph of the actual shelves displaying toothpaste, the fourth a photograph of the toothpaste choice set without the current listed price, the fifth a photograph of the toothpastes with the actual current prices listed beside each choice. Three similar slides were next presented for the hand-soap choice task. For both products comparisons were made between similar brand sizes (4.6 oz. to 5.2 oz. for toothpaste and 4.5 to 5.0 oz. for hand-soap). The above effort to position the Exit-Entry Analysis within a particular store and time frame (the present) and to make the purchase simulation more realistic appeared to be effective based on the debriefing comments of the students. The major artificial aspect of the exercise was the forced highlighting of the prices. In reality most students would not compare prices as thoroughly as they did in this exercise. This suggests that the results will overstate price sensitivity.

Figures 1 through 4 illustrate the distinctive variations in the 15 demand curves that were generated. Crest has a dominant choice market share (67%) amongst the sample at its stated price of $1.09. Around this price the students are relatively insensitive to price changes. Dropping or raising the price by 5 cents would make little difference to market share. Seven percent of the students would not buy the product even if it were given to them. In contrast, the Colgate priced at $1.39 had a 9% market share. If Crest were priced at $1.39 and ceteris parabus, it would have a 22% market share. However, the curves also differ in that Colgate has a particularly sensitive price point around $1.09-$1.08 (the price of Crest in this exercise). At this 14 price decrease its market share would increase by 18 percentage points. The other alternatives did not exhibit such sensitivity. They were generally much steeper in gradient indicating an insensitivity to price. For example, 60: of the respondents indicated they would never choose Ultra-brite at any price.

The soap demand curves were generally flatter, indicating greater price sensitivity compared to toothpaste. This suggests that students perceive greater product differentiations between toothpastes than between hand-soaps. Figures 3 and 4 again contrast two brands with different demand curves that have strategic implications. Coast is currently priced within an inelastic range on its demand curve. However, Ivory, if it dropped its price by a few cents, could greatly expand its market share. On the other hand, raising its price by a few cents would have little effect on student demand.

Test-Retest Reliability

Previous techniques have not reported the results of test-retest reliability. A test-retest study (five days apart) was undertaken using a sample of 43 students. The two demand curves for Crest are presented in Figure 5. The result is reassuring in that even with such a small sample size the strategic conclusions would be the same based on either study. Table 2 presents the results of a more systematic analysis of the temporal stability of the technique. Although there is variability within subjects, time had neither a main effect nor an interaction effect. Brand obviously was a major determinant of Entry-Exit price. We are currently undertaking a larger and more extensive study of this technique by examining the effect on Entry-Exit price of changing the initial starting price points (i.e., Crest at $1.09, $1.29 or $1.49). It does need to be repeated that generalizability of the findings depends very much on the characteristics of the sample, the choice set, and the purchase context.

FIGURE 1  -  CREST   and    FIGURE 2  -  COLGATE

FIGURE 3  -   COAST   and   FIGURE 4  -  IVORY

FIGURE 5

TEST-RETEST DEMAND CURVES

TABLE 2

TEST-RETEST PRICE DEVIATIONS

SUMMARY

We have described "Exit/Entry Demand Analysis" which we believe is a useful combination of the price limit method of demand estimation. Several previous approaches have been limited to responses to product classes and not individual brand. Other more recent brand purchase choice simulations have measurement problems. The advantages and disadvantages of this method are listed in Table 3.

TABLE 3

ADVANTAGES AND DISADVANTAGES OF ENTRY/EXIT DEMAND ANALYSIS

Our method is based on psychological price perception theory that seeks to identify the break points at which consumers adopt or reject a brand. The method is sufficiently simple to present to respondents that it can be administered by mail. It is also quite flexible and can sample a variety of buyers and buying situations and yields data that are simple to analyze and understand. Like other methods, it seems most appropriate for well-defined product classes that to not contain a large number of brands. Certainly any optimism about this method must await more research into its reliability and an assessment of its validity. It may be that the use of this technique to assess the response to long-term price changes proves to be naive. However, at a minimum, the Entry/Exit demand analysis ought to be a useful measure of the relative, if not absolute, price sensitivity of brands and the likely consumer response to brand specific temporary price promotions. In addition, consumers' reactions to a retaliatory price change by competitors can also be studied by a follow up study in which the competitive brand's starting price is reduced.

REFERENCES

Abrams, Jack (1569), "A New Method for Testing Pricing Decisions," Journal of Marketing, 28 (July), 6-9.

Adam, Daniel (1969), "Consumer Reactions to Price," in Bernard Taylor and Gordon Wills, Pricing Strategy, New York: Brandon Systems Press, Inc., 75-88.

Darden, Bill R. (1968), "An Operational Approach to Product Pricing," Journal of Marketing, 32 (April), 29-33.

Gabor, Andre and Clive Granger (1969), "The Attitude of the Consumer to Prices," in Bernard Taylor and Gordon Wills, Pricing Strategy, New York: Brandon Systems Press, Inc.. 132-146.

Gabor, Andre, Clive W.J. Granger, and Anthony P. Sowter (1970), "Real and Hypothetical Shop Situations in Market Research," Journal of Marketing Research, 7 (August), 355-9.

Guiltinan, Joseph P. (1976), "Risk-Aversive Pricing Policies: Problems and Alternatives," Journal of Marketing, 40 (January), 10-15.

Hawkins, Edward R. (1957), "Methods of Estimating Demand," Journal of Marketing, 21 (April), 428-438.

Jones, D. Frank (197;), "A Survey Technique to Measure Demand under Various Pricing Strategies," Journal of Marketing, 12 (July), 75-77.

Mahajan, Vijay, Paul E. Green and Stephen M. Goldberg (1982), "A Conjoint Model for Measuring Self-and Cross-Price/Demand Relationships," Journal of Marketing Re- search, 14 (August), 334-342.

Naisbitt, John (1982), Megatrends, New York: Warner Books.

Nevin, John R. (1974), "Laboratory Experiments for Estimating Consumer Demand: A Validation Study," Journal of Marketing Research, 11 (August), 261-8.

Oxenfeldt, Alfred R. (1973), "A Decision-Making Structure for Price Decisions," Journal of Marketing, 37 (January), 48-53.

Pessemier, Edgar A. (1959), "A New Way to Determine Buying Decisions," Journal of Marketing, 24 (October), 41-46.

Pessemier, Edgar A. (1960), "An Experimental Method for Estimating Demand," Journal of Business, 33 (October), 373-383.

Sampson, Richard (1969), "Sense and Sensitivity in Pricing," in Bernard Taylor and Gordon Wills, Pricing Strategy, New York: Brandon Systems Press, Inc., 233-243.

Shapiro, Benson P. and Barbara B. Jackson (1978), "Industrial Pricing to Meet Customer Needs," Harvard Business Review, 54 (November-December), 119-127.

Stoetzel, Jean (1960), "Psychological/Sociological Aspects of Price," in Bernard Taylor and Gordon Wills, Pricing Strategy, New York: Brandon Systems Press, Inc., 70-74.

Stout, Roy G. (1969), "Developing Data to Estimate Price-Quantity Relationships," Journal of Marketing, 33 (April), 34-36.

Vining, Daniel R., Jr. and Thomas C. Elwertowski (1976), "The Relationship between Relative Prices and the General Price Level," The American Economic Review, 66 (September), 699-708.

----------------------------------------