Asymmetries in Price and Quality Competition: Experimental Test of Underlying Mechanisms

ABSTRACT - Discounting national brands typically attracts consumers from store brands more than discounting store brands attracts consumers from national brands. The current study reviews and tests four mechanisms potentially involved in such asymmetric price competition: (1) differential attribute weights, (2) choice-anchored loss aversion, (3) differential loss aversion to quality and price, and (4) differential distances from indifference. In a choice experiment, price reductions moved subjects from lower to higher quality more than from higher to lower quality, replicating asymmetric price competition. Quality improvements, however, moved subjects from higher to lower quality more than from lower to higher quality. Attribute weights explained much but not all of this variance, whereas differential distances from indifference could not. The results suggest that asymmetric switching arises from differential attribute weights as well as other mechanisms such as loss aversion.

Citation:

Timothy B. Heath, Gangseog Ryu, Subimal Chatterjee, and Michael S. McCarthy (1997) ,"Asymmetries in Price and Quality Competition: Experimental Test of Underlying Mechanisms", in NA - Advances in Consumer Research Volume 24, eds. Merrie Brucks and Deborah J. MacInnis, Provo, UT : Association for Consumer Research, Pages: 366-374.

Advances in Consumer Research Volume 24, 1997      Pages 366-374

ASYMMETRIES IN PRICE AND QUALITY COMPETITION: EXPERIMENTAL TEST OF UNDERLYING MECHANISMS

Timothy B. Heath, University of Pittsburgh

Gangseog Ryu, University of Pittsburgh

Subimal Chatterjee, State University of New York at Stony Brook

Michael S. McCarthy, Miami University

[This study was supported by a research grant from the Joseph M. Katz Graduate School of Business, University of Pittsburgh. The first author wished to extend a very special thanks to Dean Daniel S. Fogel for his strong and continued support of research at the University of Pittsburgh. The authors also wish to thank the following people for their invaluable assistance in data collection: Sandy Milberg, Dave Mothersbaugh, Mike Rich, Jim Patton, Steve Silverman, Krish Krishnan, Lisa Sciulli, Rob Morgan, and Anthony Allred.]

ABSTRACT -

Discounting national brands typically attracts consumers from store brands more than discounting store brands attracts consumers from national brands. The current study reviews and tests four mechanisms potentially involved in such asymmetric price competition: (1) differential attribute weights, (2) choice-anchored loss aversion, (3) differential loss aversion to quality and price, and (4) differential distances from indifference. In a choice experiment, price reductions moved subjects from lower to higher quality more than from higher to lower quality, replicating asymmetric price competition. Quality improvements, however, moved subjects from higher to lower quality more than from lower to higher quality. Attribute weights explained much but not all of this variance, whereas differential distances from indifference could not. The results suggest that asymmetric switching arises from differential attribute weights as well as other mechanisms such as loss aversion.

Brands commonly reduce prices to steal consumers from other brands. Recent scanner research shows that the ability to stea consumers is asymmetric across brands differing in relative quality. Consumers of store brands are more likely to switch to a discounted national brand than consumers of national brands are likely to switch to a discounted store brand (Allenby and Rossi 1991; Bemmaor and Mouchoux 1991; Blattberg and Wisniewski 1989; Kamakura and Russell 1989; Sethuraman 1995; Walters 1991). A number of theoretic mechanisms have been proposed to explain this so called asymmetric price competition (for a review of seven mechanisms, see Heath et al. 1996). The current study reviews four such mechanisms and experimentally tests a number of their implications. Since these mechanisms generally apply to attributes other than competition between store and national brands, we focus, when possible, on the broader distinction between higher-quality and lower-quality brands.

POTENTIAL MECHANISMS AND HYPOTHESES

At least four mechanisms have been implicated in asymmetric switching. The first is differential attribute weights. Consumers of lower quality may be more price sensitive than consumers of higher quality (Bronnenberg and Wathieu 1995; Kamakura and Russell 1989). Those consuming lower quality would then naturally react more strongly to price reductions by competitors than would consumers of higher quality.

H1: Reducing a higher-quality competitor’s price will move consumers to higher quality more than reducing a lower-quality competitor’s price will move consumers to lower quality (asymmetric price competition).

If differential attribute weights account for asymmetric price competition, it raises the possibility of asymmetric quality competition in the opposite direction. If consumers of higher quality weight quality more than do consumers of lower quality, then quality competition should be asymmetric favoring lower-quality brands.

H2: Improving a lower-quality competitor’s quality will move consumers to lower quality more than improving a higher-quality competitor’s quality will move consumers to higher quality.

Combined, Hypotheses 1 and 2 predict an interaction between competitor’s quality (lower vs. higher) and attribute improved (price vs. quality). Moreover, if the interaction arises from differential attribute weights, then controlling for attribute weights should (1) yield significant attribute-weight effects, and (2) eliminate the interaction between competitor quality and attribute improved.

H3: Greater weight on quality relative to price will reduce switching among choosers of higher quality, whereas greater weight on price relative to quality will reduce switching among choosers of lower quality (regardless of which attribute competitors improve).

H4: Controlling for attribute weights will eliminate asymmetric competition.

The second mechanism is choice-anchored loss aversion. Loss aversion refers to the fact that losses are more unpleasant than equivalent gains are pleasant (Bernoulli 1738; Kahneman and Tversky 1979). This phenomenon is typically represented with value curves that are steeper for losses than for gains from reference states. Loss aversion holds implications for switching behavior if we assume, as has been shown elsewhere, that consumers anchor on the brand they chooe initially (Heath et al. 1996; Sen and Johnson 1995). A brand’s relative strengths and weaknesses can then be translated into gains and losses, respectively (e.g., a higher-quality brand’s quality and price).

Consider the following microwave ovens.

                  Price   Quality

Brand A     $319    65

Brand B     $269    55

If you prefer and anchor on B, then switching to A constitutes a gain in quality and a loss in price (lost money). Reducing A’s price, therefore, involves reducing a prospective loss for consumers of B, whereas reducing B’s price involves increasing a prospective gain for consumers of A. By loss aversion, reducing losses carries more value than increasing gains. Therefore, A benefits more from reducing its loss on price than B benefits from increasing its gain on price. In this context, therefore, loss aversion is strictly a cross-attribute phenomenon (see Heath et al. 1996 for further explanation).

Choice-anchored loss aversion, like differential attribute weights, predicts asymmetric price competition favoring higher-quality brands (Hypothesis 1) and asymmetric quality competition favoring lower-quality brands (Hypothesis 2). Increasing a higher-quality brand’s quality improves its relative strength, whereas increasing a lower-quality brand’s quality improves its relative weakness. Although both mechanisms yield Hypotheses 1 and 2, they can be disentangled in part statistically. If controlling attribute weights fails to eliminate asymmetric switching (failure to support Hypothesis 4), the data would support some mechanism other than differential attribute weights which, in the current study, we assume to be loss aversion or differential loss aversion discussed next. [The other three mechanisms implicated in asymmetric switching are not likely to be involved in the experiment reported here: No differential asymmetric dominance existed within the stimuli, differential income effects were probably not involved because, among other reasons, the savings were minuscule, and differential changes in perceived value were not likely because price and quality improvements did not leave the competitors' levels close to those of brands initially chosen (for a review of these mechanisms see Heath et al. 1996).]

The third mechanism is differential loss aversion for quality and price (Hardie, Johnson, and Fader 1993; Tversky and Kahneman 1991). Modeling the effects of price and quality changes within household-level scanner data, Hardie et al. estimated a larger loss-aversion coefficient for quality than for price, where coefficients reflect the differential effects of losses and gains within a particular attribute. Distinguishing choice-anchored loss aversion from differential loss aversion is extremely difficult and is not attempted in the current study.

Blattberg and Wisniewski (1989) proposed the fourth and final mechanism reviewed here: Consumers of higher quality are further from indifference than are consumers of lower quality. We test this mechanism by measuring weights on quality and price and then testing relative attribute weights (RAW) which we define for consumers of lower and higher quality in the following way:

RAW_{LQ Consumers}=Weight_Price - Weight_Quality

RAW_{HQ Consumers}=Weight_Quality - Weight_Price

Differential distances from indifference suggests that relative attribute weights are larger among consumers of higher than lower quality given the following conditions, all of which held in the current study (see Heath et al. 1996): (1) Perceptions of quality are constant across consumers, (2) trade-offs in price and quality are symmetric across brands (e.g., if going from lower to hiher quality involves a 15% increase quality, it also involves a 15% increase in price), and (3) consumers had no sense of what price to pay per unit quality when entering the task (unfamiliar and abstract scales were used to represent quality). [The data support this operationalization of distance from indifference since virtually everyone had positive relative preferences (e.g., choosers of higher quality rated quality as more important than price). If price-quality trade-offs had been skewed in one way or another, choices and the simplified expressions of relative preference would not correspond and more complex operationalizations would be required.]

H5: Choosers of higher-quality brands will exhibit larger relative attribute weights than choosers of lower-quality brands.

EXPERIMENT

The experiment was designed to test (1) if asymmetric price competition could be replicated with laboratory stimuli (Hypothesis 1) and (2) if asymmetric quality competition could be demonstrated (Hypothesis 2). Importance weights on price and quality were measured to assess the effects of differential attribute weights (Hypotheses 3 and 4) and to calculate relative attribute weights with which to test differential distances from indifference (Hypothesis 5).

Method

Subjects. Two-hundred-ninety-nine undergraduate and graduate business students served as subjects on a voluntary basis during class time.

Design. Three variables were manipulated between subjects: competitor (lower quality, higher quality), attribute improved (price, quality), and product class (orange juice, microwave ovens, light bulbs). Subjects were randomly assigned to the conditions defined by the crossing of attribute improved and product class. Subjects self selected themselves into competitor conditions by first choosing either a higher-quality brand (lower-quality competitor) or a lower-quality brand (higher-quality competitor).

Overview of Stimuli. Table 1 summarizes the stimuli. Although brands were labeled A and B, their attribute levels were taken from the marketplace to enhance realism. For orange juice, quality levels for the lower-quality and higher-quality brands were borrowed from Hardie et al.’s (1993) Tropicana Regular and Minute Maid brands, respectively. The authors derived these quality levels from Consumer Reports ratings reflecting freshness and taste. Orange juice prices were taken from the prices of these brands in local grocery stores. For microwave ovens, prices and power levels were taken from mid-sized microwave ovens in local appliance stores. Prices and expected life for light bulbs were taken from regular and long-life 75W bulbs, respectively, in local grocery stores.

For the quality-related attributes, we chose overall quality in orange juice to mirror the differences in overall quality that exist in the marketplace. Although overall quality ratings are not listed on packages of orange juice, perceptions of differential quality are likely to exist and are supported in Hardie et al.’s data. We chose power levels in microwave ovens and expected life in light bulbs as the quality-related attributes to enhance generalizability beyond quality ratings and because each is a product feature consumers face in the marketplace.

Scaling Changes in Price and Quality. To facilitate comparisons across brands and attributes, changes in attributes must be scaled carefully. A competitor’s weakness was improved by splitting the difference between the competitor and the chosen brand on the relevant attribute. Consider microwave ovens:

                   Price       Power

Brand A      $220       650W

Brand B      $160       550W

Weaknesses were improved by reducing A’s price from $220 to $190 or by improving B’s power from 550W to 600W. These values were then used as a basis for proportional improvements in strengths.

For improvements in the higher-quality brand’s strength (quality), proportions were used to mitigate contamination from perceptual relativity and diminishing marginal sensitivity. For example, the improvement in Microwave B’s power, its weakness, from 550W to 600W constituted an increase of 8.3% (50/550). Since we predicted that improving B’s power (weakness) would induce more switching than improving A’s power (strength), it was important to reduce any tendencies to see A’s power increase as smaller than B’s due to A’s higher initial power level. To that end, proportional increases in the higher-quality brand’s quality-related attribute (strength) were greater than or equal to proportional increases in that attribute in the lower-quality brand (weakness). For example, Microwave B improved from 550W to 600W (8.3%), whereas Microwave A improved from 650W to 710W (9.2%).

TABLE 1

ORIGINAL CHOICE SETS AND IMPROVEMENTS

For improvements in the lower-quality brand’s strength (price), we used the absolute price reduction applied to the higher-quality brand. This means that the proportional reduction in the lower-quality brand’s price was always larger than the proportional reduction in the higher-quality brand’s price. Since Microwave A reduced its price from $220 to $190 (a 13.6% discount), Microwave B reduced its price from $160 to $130 (an 18.7% discount). In raw (nonproportional) units, improvements to strengths were generally greater than or equal to improvements to weaknesses.

The proportion-based changes discussed thus far speak to comparisons of strengths and weaknesses within attributes across brands (e.g., comparing price reductions across higher-quality and lower-quality brands). To compare changes in strengths and weaknesses across attributes within brands, we face the problem of noncomparable dimensions such as price and wattage. Ideally, two criteria should be met to reduce the possibility that improvements in weaknesses are perceived as larger than improvements in strengths. The first is that the proportional improvement in a brand’s strength be greater than or equal to the proportional improvement in its weakness. For example, we might compare the effect of a 15% increase in quality with a 15% reduction in price.

The second criterion is that changes in strengths and weaknesses comprise comparable proportions of between-brand differences on their respective attributes. This is important because consumers commonly evaluate attributes and changes in those attributes with respect to differences in attributes across alternatives (Bronnenberg and Wathieu 1995; Holman 1995). Consider the higher-quality microwave oven (Brand A). To improve its weakness, it reduces its price by $30, or by 50% of the $60 between-brand price difference ($220 vs. $160). To help ensure that the improvement in Brand A’s strength (power) is not perceived as smaller, Brand A should improve its power by at least 50% of the 100 watt between-brand difference in power (650W vs. 550W). These criteria were generally met except for a couple of violations of the second criterion which we address in the experiment’s discussion.

Procedure. Subjects received a booklet that initially presented one of the original choice sets seen in Table 1. They were asked to imagine being in the market for that product and to make a choice by circling their preferred brand. They then turned to the next page where instructions led them to one of two other pages depending on which brand they chose initially. At the top of the new age, the original choice set was repeated and subjects were asked to repeat their initial choice. They then read that after entering their favorite store, they find that the competitor (non-chosen brand stated as Brand A or Brand B) recently reduced its price or improved its quality-related attribute. The choice set was then re-stated with the one attribute improved, and subjects circled their choice from the revised set. The primary dependent measure was whether subjects switched brands.

Subjects then filled out a brief questionnaire. It contained filler questions to distract subjects from the key measure which was the importance of the price and the quality-related attribute in the product class from which they chose. Subjects allocated 100 points across the two attributes according to their relative importance. Attribute weights were measured after choices for two reasons. First, the primary dependent measure was switching, the measure most in need of insulation from carryover effects. Second, whereas carryover from attribute weights to choices would take the form of unwanted consistency biases, carryover from choices to attribute weights might also have a positive component (learning). Since consumers are often uncertain about their attribute weights, making choices first might help them discover their true weights (see Simonson 1990).

RESULTS

We test the data with respect to three issues (see Table 2): (1) asymmetric price and asymmetric quality competition, (2) differential attribute weights, and (3) differential distances from indifference.

Asymmetric Price and Asymmetric Quality Competition. Switching data (no, yes) were subjected to a logit (SAS CATMOD) testing the main and interaction effects of competitor (lower quality, higher quality), attribute improved (price, quality), and product class (orange juice, microwave ovens, light bulbs). Hypothesis 1 predicted that price reductions would yield more switching up than down in quality (standard asymmetry), whereas Hypothesis 2 predicted that improvements in quality would yield more switching down than up in quality. As predicted, the competitor by attribute-improved interaction was statistically significant (c²₍₁₎=13.38, p<.001; see Figure 1). Consistent with reports of asymmetric price competition in the marketplace (Hypothesis 1), price reductions attracted more consumers from lower quality to higher quality (67.1%) than from higher quality to lower quality (39.7%; c²(1)=11.70, p<.001). Moreover, consistent with Hypothesis 2, improvements in quality benefited lower-quality brands more than higher-quality brands. Improvements in quality moved consumers from higher to lower quality (59.9%) more than to from lower to higher quality (43.4%; c²₍₁₎=3.21, p=.073). Collapsing across improvements to price and quality, improving a weakness was more persuasive than improving a strength (62.5% vs. 41.2%).

The competitor by attribute-improved by product-class interaction was also significant (c²₍₂₎=7.41, p=.02; see Figure 2). [Given this three-way interaction, experimental convention suggests it is inappropriate to interpret the competitor by attribute-improved interaction as we did initially. However, we do so because product class is an unusual moderator that can be viewed in either of two ways, the firs of which suggests ignoring its effects. First, we can view multiple product classes solely as a vehicle to enhance generalizability to the larger marketplace. Each product class would then be construed as a sampling unit and any effects of product class as error. The competitor by attribute=improved interaction would then be interpreted without regard to any effects of product class (this interaction remains significant if we drop product class and its interactions from the model). Second, we can view multiple product classes as situations across which we expect, or wish to explore, systematic differences. Here, product-class effects are important and we would not interpret the competitor by attribute-improved interaction without considering the moderating role of product class. Since each view has merit, we plot and interpret both the two-way and three-way interactions.] Separate models were then run per product class. In microwave ovens, the competitor by attribute-improved interaction was significant and supported Hypotheses 1 and 2 (c²₍₁₎=17.74, p=.001). Price reductions moved consumers up in quality (75.9%) more than down (47.6%; c²₍₁₎=4.22, p=.04), whereas quality improvements moved consumers down in quality (80.8%) more than up (21.7%; c²₍₁₎=17.07, p<.001).

TABLE 2

SWITCHING PROBABILITIES AND RELATIVE ATTRIBUTE WEIGHTS

In orange juice, the competitor by attribute-improved interaction did not achieve statistical significance (c²₍₁₎=2.02, p=.15). Testing the individual effects of price and quality improvements on theoretic grounds, we find that consistent with Hypothesis 1 and asymmetric price competition, price reductions moved consumers up in quality (48.0%) more than down (23.1%; c²₍₁₎=3.47, p=.06). But in contrast to asymmetric competition, quality improvements moved onsumers up and down in quality comparably (50.0% vs. 54.0%; c²₍₁₎=.06, ns).

The light bulb data supported Hypothesis 1 but not Hypothesis 2. The competitor by attribute-improved interaction was not significant (c²₍₁₎=0.39, ns). Consistent with Hypothesis 1, price reductions moved consumers up in quality (78.9%) more than down (48.4%; c²₍₁₎=4.58, p=.03), and quality improvements tended to do the same (66.7% vs. 46.9%; c²₍₁₎=1.82, p=.18). [The only other significant effects were the main effect of product (X2(2)=5.86, p=.05) and the product by competitor interaction (X2(2)=8.18, p=.02). There was generally less switching within orange juice (44.0%) than within microwave ovens (58.6%) and light bulbs (57.0%). The interaction arose because there was more switching to higher-quality than lower-quality brands within orange juice and light bulbs (48.6% vs. 41.2% in orange juice, and 73.0% vs. 47.6% in light bulbs), whereas the opposite held in microwave ovens (51.9% vs. 66.0%).] Thus, while the overall competitor by attribute-improved interaction supported Hypotheses 1 and 2, Hypothesis 1 was supported in all three product classes, whereas Hypothesis 2 was supported in microwave ovens, reversed in light bulbs, and unsupported by a lack of asymmetry in orange juice.

Differential Attribute Weights. According to Hypotheses 3 and 4, the competitor by attribute-improved interaction arises from differential attribute weights rather than mechanisms such as loss aversion. We tested this by adding attribute weights (and their interactions) to the model. The constant-sum-scaled importance ratings were transformed to a single weight variable by taking quality importance minus price importance. Larger values reflect more weight on quality. The weight variable was mean-centered and standardized within product classes to reduce colinearity and enhance comparability across product classes (Jaccard, Turrisi, and Wan 1990). It was then entered as a continuous variable.

Hypothesis 3 predicted that greater weight on quality would be associated with (1) less switching among choosers of higher quality, and (2) more switching among choosers of lower quality. Hypothesis 3 was generally supported by the significant competitor by weight interaction (c²₍₁₎=20.86, p<.001). Logistic regressions revealed that more weight on quality was associated with less switching among choosers of higher quality (c²₍₁₎=22.62, p<.001) and more switching among choosers of lower quality, although this latter effect did not achieve statistical significance (c²₍₁₎=2.10, p=.15).

FIGURE 1

COMPETITOR BY ATTRIBUTE-IMPROVED INTERACTION

Hypothesis 4 predicted that adding attribute weights to the model would eliminate the competitor by attribute-improved interaction. The competitor by attribute-improved interaction was, in fact, reduced to nonsignificance overall (c²₍₁₎=2.57, p=.11), as was the product-class by competitor by attribute-improved interaction (c²₍₂₎=4.53, p=.10). However, a four-way interaction involving product class, competitor, attribute improved, and attribute weights emerged (c²₍₂₎=5.62, p=.06). Separate models with attribute weights included were then run per product class.

In all three product classes, the competitor by weight interaction supported attribute-weight effects (Hypothesis 3): microwave ovens (c²₍₁₎=6.44, p=.01), orange juice (c²₍₁₎=4.38, p=.036), and light bulbs (c²₍₁₎=10.16, p=.001). In microwave ovens, the significant competitor by attribute-improved interaction remained significant after adding attribute weights and, therefore, was not consistent with Hypothesis 4’s prediction that attribute weights alone drive asymmetric switching (c²₍₁₎=8.14, p=.004). In fact, adding attribute weights moved the competitor by attribute-improved interaction in orange juice from nonsignificance to significance (c²₍₁₎=4.35, p=.037). We explicate this change in a moment. In light bulbs, the interaction remained nonsignificant (c²₍₁₎=0.61, ns), although adding attribute weights failed to eliminate the standard asymmetry (the main effect of competitor; c²₍₁₎=5.92, p=.015). [The four-way interaction among product class, competitor, attribute improved, and attribute weights is largely inexplicable and occurred because within orange juice there was a significant three-way interaction among competitor, attribute improved, and attribute weights (X2(1)=4.21, p=.04). When the competitor improved quality, more weight on quality was associated less switching among choosers of the higher-quality orange juice (X2(1)=5.30, p=.021) and with more switching among choosers of the lower-quality orange juice (X2(1)=2.31, p=.128). However, when price was improved, more weight on quality was not related to switching regardless of whether consumers chose higher or lower quality (X2(1)'s=0.000 and 0.02, respectively, ns). Another way to look at the three-way interaction within orange juice is to test the competitor by attribute-improved interaction at each level of the attribute weights. To do so requires performing a median split on the weight variable which, unfortunately, renders cell sizes too small to obtain reliable tests (e.g., n=2 in two cases.).]

FIGURE 2

COMPETITOR BY ATTRIBUTE-IMPROVED BY PRODUCT-CLASS INTERACTION

The competitor by attribute-improved interaction was significant in microwave ovens and orange juice after taking into account the effects of attribute weights. However, the possibility remains that the nature of the interaction was altered by the weights. Although we cannot easily adjust switching rates for weight effecs using logit’s MLE, we do so using adjusted means from OLS. OLS is generally avoided with dichotomous dependent measures because extreme proportions violate its underlying assumptions. However, when proportions remain within the 25% to 75% range, the relationship between probabilities and log-odds should be relatively linear and OLS should yield results comparable to MLE (e.g., Cleary and Angel 1984). The proportions in our results generally complied with this constraint (see Table 2) and OLS does, in fact, yield findings comparable to those from MLE. [OLS yields a similar pattern of effects, although the p values it generates are generally lower than those produced by MLE. For example, adding attribute weights reduces the overall competitor by attribute-improved interaction to nonsignificance with MLE (X2(1)=2.57, p=.11) but not with OLS F(1.275)=12.95, p<.001).]

For OLS, we coded a switch as 1 and a non-switch as 0 such that means represent proportions of switchers. Figure 2 includes OLS-based switching rates adjusted for weight effects. After factoring in weights, the standard asymmetry remains in tact within all three product categories. Asymmetric quality competition also remains in tact within microwave ovens, becomes slightly more pronounced in orange juice, and is not indicated at all in light bulbs. Although significant weight effects implicate differential attribute weights as a mechanism involved in asymmetric switching, the failure of these effects to eliminate asymmetries implicates other mechanisms as well.

Differential Distances from Indifference. Differential preference strengths predict that choosers of higher quality will be more committed to their brands than will choosers of lower quality. We subjected relative attribute weights to an ANOVA using the brand chosen, product class, and attribute improved as predictors (see Table 2). Hypothesis 5 predicted that relative attribute weights would be larger among choosers of higher quality than among choosers of lower quality. In contrast, the main effect of brand chosen was not significant (F_(1,287)=1.96, p=.16).

The product-class by brand-chosen (F_(2,287)=10.34, p<.001), product-class by attribute-improved (F_(2,287)=5.91, p=.003), and product-class by brand-chosen by attribute-improved interactions were all significant (F_(2,287)=2.99, p=.052). In contrast to Hypothesis 5, relative attribute weights were smaller among choosers of higher quality than among choosers of lower quality within microwave ovens and light bulbs, and, therefore, cannot account for the standard asymmetries in those product classes. In orange juice, relative attribute weights were larger for choosers of higher quality (M=27.54) than lower quality (M=15.68; F_(1,287)=7.35, p=.007). While this is consistent with the standard asymmetry in switching in response to price reductions, it is not consistent with the asymmetric quality competition in response to quality improvements (once weight effects were included). Based on both the relative-attribute-weight and switching data, it is unlikely that differential distances from indifference account for the asymmetric switching found in this experiment.

DISCUSSION

We begin with two problems in the orange juice stimuli which may have played a role in the results (see Table 1). First, the proportional improvement in the lower-quality brand’s weakness (quality) was larger than the improvement in its strength (price; 30.4% vs. 14.3%). This poses an alternative explanation for why improving the lower-quality brand’s quality was more effective than improving its price. Second, the improvement to the higher-quality brand’s weakness (price) reduced the between-brand difference on price by 49%, whereas the improvement to its strength (quality) increased the between-brand difference on quality by 80.4%. This may have undermined the improvement in the weakness such that the improvement in the weakness and in the strength yielded comparable rates of switching (48.0% vs. 50.0%). Therefore, in the orange juice stimuli, one scaling problem potentially inflated support for our hypotheses while another potentially reduced it. [The stimuli in our experiment violated the preferred pattern of proportional relationships in two other instances, although neither is likely to have altered the data. First, in microwave ovens, the single violation was small enough as to pose little threat to internal validity. The proportional improvement in the higher-quality brand's price was slightly larger than the proportional improvement in its quality (13.6% vs. 9.2%). In light bulbs, the proportional improvement in the low-quality brand's quality was somewhat larger than that in its price (17.5% v.s 10.0%). However, this pattern cannot account for the findings since, in contrast to our hypotheses, the 10% price reduction induced more switching than the 17.5% quality increase.]

The eperiment generally supports the hypotheses that improving a weakness is more persuasive than improving a strength. Asymmetric price competition favoring higher-quality brands emerged in all product classes, and asymmetric quality competition favoring lower-quality brands emerged in microwave ovens, in orange juice when taking into account attribute weights, but not in light bulbs. These findings suggest that asymmetries in both price and quality competition can be replicated/demonstrated in laboratory settings. It appears that competition can be asymmetric on dimensions other than price, and that asymmetric competition sometimes favors lower-quality rather than higher-quality brands.

Attribute weights had large effects but failed to eliminate asymmetric switching. This pattern implicates differential attribute weights as one, but not the only, mechanism involved in asymmetric switching. Other mechanisms potentially involved include loss aversion and differential loss aversion. The data do not support, however, the hypothesis that choosers of higher quality are further from indifference than are choosers of lower quality.

The experiment holds implications for both theory and practice. For theory, it suggests that standard value functions may apply to dynamic, multi-attribute environments, although the competitor’s relative strengths and weaknesses may have to be translated into prospective gains and losses, respectively. For practice, the experiment suggests that marketers of lower-quality brands have recourse for stealing considerable share from higher-quality competitors by improving perceptions of quality or quality-related attributes.

The study suffers from various limitations. First, if subjects viewed 650 watts as sufficient power for microwave ovens, improving the higher-quality microwave’s wattage from 650 to 710 may have seemed unimportant despite our use of proportions to mitigate confounding from diminishing marginal sensitivity. This problem can be overcome in future research by using different attribute levels or by replacing specific attributes with overall quality ratings. Second, attribute weights could be measured better by using, for example, revealed preference techniques rather than the current study’s self-explicated scales. Third, richer stimuli might be used in the future, although replicating asymmetric price competition with our relatively austere stimuli suggests that people trade off price and quality in fairly standard fashion regardless of the stimuli’s richness. Fourth, the study could be enriched by data on process and traits (Simonson 1989). Finally, future research may want to address the anomalous finding in the light bulb data for which we have no immediate explanation: In contrast to microwaves and orange juice, quality improvements benefited the higher-quality light bulb more than the lower-quality light bulb.

REFERENCES

Allenby, Greg M. and Peter E. Rossi (1991), "Quality Perceptions and Asymmetric Switching between Brands," Marketing Science, 10 (Summer), 185-204.

Bemmaor, Albert C. and Dominique Mouchoux (1991), "Measuring the Short-Term Effect of In-Store Promotion and Retail Advertising on Brand Sales: A Factorial Experiment," Journal of Marketing Research, 28 (May), 202-214.

Bernoulli, Daniel (1738), "Specimen Theoriae Novae de Mensura Sortis," Commentarii Academiae Scientiarum Imperialis Petropolitanae, Tomas 5, 175-192. [translated in Econometrica, 1954, 22 (January), 23-36.]

Blattberg, Robert C. and Kenneth Wisniewski (1989), "Price-Induced Patterns of Competition," Marketing Science, 8 (Fall), 291-309.

Bronnenberg, Bart J. and Luc Wathieu (1995) "Asymmetric Promotion Effects and Brand Positioning," Working Paper, University of Texas, Austin, Texas.

Cleary, Paul D. and Ronald Angel (1984), "The Analysis of Relationships Involving Dichotomous Dependent Variables," Journal of Health and Social Behavior, 25 (September), 334-348.

Hardie, Bruce G. S., Eric J. Johnson, and Peter S. Fader (1993), "Modeling Loss Aversion and Reference Dependence Effects on Brand Choice," Marketing Science, 12 (Fall), 378-394.

Heath, Timothy B., Gangseog Ryu, Subimal Chatterjee, and Michael S. McCarthy (1996), "Asymmetric Competition and the Elimination of Competitive Disadvantages: Theories and Experimental Tests," Working Paper, University of Pittsburgh, Pittsburgh PA, 15260.

Holman, Frances Galliano (1995), "Relative Value Theory: An Analytic and Predictive Model of Contextual Preference Formation and Choice Behavior," Marketing Science Conference, Sydney, Australia, July 2-5.

Jaccard, James, Robert Turrisi, and Choi K. Wan (1990), Interaction Effects in Multiple Regression, Newbury Park: Sage.

Kahneman, Daniel and Amos Tversky, (1979), "Prospect Theory: An Analysis of Decision Under Risk," Econometrica, 47 (March), 263-291.

Kamakura, Wagner A. and Gary J. Russell (1989), "A Probabilistic Choice Model for Market Segmentation and Elasticity Structures," Journal of Marketing Research, 26 (November), 379-390.

Sen, Sankar and Eric J. Johnson (1995), "Now that I have It Do I Want It? Ownership Effects in Consumer Choice," Working Paper, Temple University, Philadelphia, PA, 19103.

Sethuraman, Raj (1995), "A Meta-Analysis of National Brand and Store Brand Cross-Promotional Price Elasticities," Marketing Letters, 6 (October), 275-286.

Simonson, Itamar (1989), "Choice Based on Reasons: The Case of Attraction and Compromise Effects," Journal of Consumer Research, 16 (September), 158-174.

Simonson, Itamar (1990), "The Effect of Buying Decisions on Consumers’ Assessment of Their Tastes," Marketing Letters, 2 (January), 5-14.

Tversky, Amos and Daniel Kahneman (1991), "Loss Aversion in Riskless Choice: A Reference Dependent Model," Quarterly Journal of Economics, 106 (November), 1040-1061.

Walters, Rockney G. (1991), "Assessing the Impact of Retail Price Promotions on Product Substitution, Complementary Purchase, and Interstore Sales Displacement," Journal of Marketing, 55 (April), 17-28.

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