advertisement

PRICE RESPONSE FUNCTION WITH AND WITHOUT CHOICE SET INFORMATION BY TWO-STAGED CONJOINT ANALYSIS IN THE DENIM JEANS MARKET by YOUNGSIK KWAK, B.B.A., M.B.A. A THESIS IN CLOTHING, TEXTILES, AND MERCHANDISING Submitted to the Graduate Faculty of Texas Tech University in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE Approved Accepted D6€h of the Graduate S/bhool August, 1997 ACKNOWLEDGEMENTS Above all, I want to give my deepest appreciation to my loving father and mother, Sunyoung Kwak and Sukja Chang. Their endless love and support in me led to the completion of this study. And I cannot express my thank with any word to my elder bother and his wife, Yoonsik Kwak and Nanhee Choi for their patient guidance and encouragement during my study in Texas Tech University. I am also greatly indebted to my loving two sisters, Hyunsuk Kwak and Hyimjeong Kwak for considerable supports financially and mentally. My sincere thanks are extended to Professor Pil Hwa Yoo, Professor Yong-June Kim, and Professor Sangman Han for their valuable teaching. I appreciate the help of Drs. Eberspacher, Harp, and Dodd on my thesis. u TABLE OF CONTENTS ACKNOWLEDGMENTS ii ABSTRACT vi LIST OF TABLES vii LIST OF HGURES x CHAPTER I. INTRODUCTION 1 Statement of Purpose 4 Research Questions 5 Limitation of the Study 6 Defmition of Terms 6 II. REVIEW OF RELATED LITERATURE 8 The Criteria used in Evaluating Apparel 8 The Criteria Used in Evaluating Products Variables Used in Evaluating Apparel ~8 - -9 Pricing Research in Economics and Marketing 11 Price Response Function 17 Price Elasticity 18 Demand Fimction versus Price Response Fimction 18 Competitive Price Response Functions 19 Calibration of Price Response Functions 20 iii TABLE OF CONTENTS ACKNOWLEDGMENTS ii ABSTRACT vi LIST OF TABLES ~ vii LIST OF HGURES ~ x CHAPTER I. INTRODUCTION - -1 Statement of Purpose - 4 Research Questions 5 Limitation of the Study 6 Definition of Terms 6 II. REVIEW OF RELATED LITERATURE 8 The Criteria used in Evaluating Apparel 8 The Criteria Used in Evaluating Products 8 Variables Used in Evaluating Apparel 9 Pricing Research in Economics and Marketing 11 Price Response Fimction 17 Price Elasticity 18 Demand Fimction versus Price Response Fimction 18 Competitive Price Response Functions - 19 Calibration of Price Response Functions— iii - 20 Application 24 Conjoint Analysis 25 Model 26 Conjoint analysis procedure 27 Application 28 Data gathering 29 Validity and Reliability 30 Multinomial Logit Model and Choice Set InformationModel 32 35 Summary of Literature Review 37 HI. METHOD 39 Selection of Sample 39 Research Instrument 41 Attributes and Levels Investigated 42 Collection of Data 45 Calibration of Price Response Fimction 45 Model Specification With Choice Set Information —45 Price Response Function and Optimal Pricing 46 Revenue-Maximizing Price and Static Optimal Price 48 Model Specification Without Choice Set Information 50 Statistical Analysis of Data —52 Research Question One —52 iv Research Question Two - 52 Research (Question Three — 54 IV. RESULTS OF DATA ANALYSIS 55 Questionnaire Response Rate and Description of the Sample 55 Statistical Treatment of the Data 58 Research Question One 58 Research Question Two — 59 Research Question Three - 71 V. SUMMARY, CONCLUSIONS, AND FUTURE RESEARCH Summary of the Study Interpretation of Results Research (Question One -—73 74 - - 75 —- 75 Research (Question Two 76 Research Question Three 79 Conclusions and Implications —80 Suggestions for Future Study 81 REFERENCES 83 APPENDIX: QUESTIONNAIRE - 90 ABSTRACT The primary purposes of this study were to develop a new methodology for calibration of the continuous price response function and to compare differences in the price response functions with and wdthout choice set information. The new methodology, identified as the two-staged conjoint analysis, incorporated the logit transformation into the conjoint analysis and added the choice set formation step in the conjoint questionnaire. Two conjoint models were specified to compare the differences in the price response functions with and without choice set information. These two models were tested in a study of the denim jeans market. Participants included 103 students at a major state-supported southern university. In a conjoint model, the independent variables consisted of brand, price, color, and style, while the dependent variables were equal to each respondent's purchase intention data. The two-staged conjoint questionnaire consisted of three steps. Step one provided respondents with both a written and a visual description of two randomly selected styles and colors of denim jeans. In step two, respondents were asked to choose the combination of attributes they intended to piu-chase. In step three respondents were asked to rate each combination chosen at step two on a scale of 1-100, with one as least likely to be purchased and 100 as most likely to be purchased. Three research questions were developed to accomplish the purposes of this study. These questions were intended to; (1) examine the differences in the price response function among brands; (2) compare the differences in the price response functions with VI and without choice set information; and (3) compare the optimal price of the price response functions with and without choice set information for each brand. A variety of statistical methods, including ANOVA, t-test, and regression analysis, were employed to analyze the research questions. The results of ANOVA revealed no significant differences in the purchase probabilities among the eight identified brands at $20, $40, $60, and $80. However, results of the t-test and price elasticity showed that the price response functions with and without choice set information had different price response structures. The different shapes of the price response functions with and without choice set information resulted in the differences in optimal price. vu LIST OF TABLES 2.1. Evaluated Cues in Investigating Apparel Quality 2.2. Criteria Types of Apparel Quality Evaluation -10 12 2.3. Eckman et al's (1990) Summary of Intrinsic/Extrinsic Cues Used by Researchers for Evaluating Apparel Quahty 13 2.4 Pricing Research In Economics and Marketing 14 2.5 Yoo's (1989) Pricing Classification 16 2.6 Economically Calibrated Price Response Function Based on Actual Market Data 2.7 Comparison of Methods for Collecting Price Response Fimction Data 22 23 2.8 Examples of Segmentation and Market Structure Using Multinomial Logit Model —33 2.9 Examples of Choice Set Information Using Multinomial Logit Model 36 3.1 Brands and Price Ranges Observed in One Shopping Mall July, 1996 44 3.2 Product Profile of Selected Brands -47 3.3 Dependent and Independent variables in Price Response Fimction 4.1 Demographic Characteristics of Sample 53 —56 4.2 Summary ANOVA for $20 Price Point 60 4.3 Summary ANOVA for $40 Price Point 61 4.4 Summary ANOVA for $60 Price Point 62 4.5 Summary ANOVA for $80 Price Point 4.6 The Results of T-Test at various Price Points 4.7 The Differences in Price Elasticity on the Price Response Function viii 63 69 70 4.8 Estimated Price Response Function and the Optimal Price — 72 5.1 The Coefficient Mean of the Conjoint Model 77 5.2 The Relative Inqjortance Among the Attributes 78 IX LIST OF HGURES 2.1 Historical Development of Pricing Research in Marketing Between the U.S. and Europe 15 2.2 Competitive Price Response Functions 21 2.3 Basic Components of the Pre-purchase Ahemative Evaluation Process 34 3.1 Flow of Calibration of Price Response Function and Optimal Price 40 3.2 Linkage Among the Mean Choice Probability for Price Response Function 49 4.1 Price Response Function With and Without Choice Set Information 65 CHAPTER I INTRODUCTION Today's retail environment is in a state of flux. Caught in a profit squeeze, retail organizations are in a constant battle for market share. To succeed in a highly competitive market dominated by strategic shoppers, marketers in the 1990's must focus on the range of factors or attributes consumers use to make judgements or trade-offs in determining their final purchase choice. In this context, it is important for manufacturers and retailers to understand how price, promotion, and other marketing variables affect product sales. Such information is the basic material for marketing mix decisions. How consumers respond to price arid price changes and how manufacturers and retailers determine a set of optimal prices can be used to identify and maximize market opportunities. Within the marketing mix, price is distinguished from other components by its revenue producing characteristic. A 1996 (Dolan & Simon) study illustrated the increased significance of price. When 186 managers (57 from the United States and 129 from Europe) representing a broad range of industries ranked 13 marketing issues, results showed that managers rated pricing as the most pressing problem. Despite the importance of the pricing decision, a number of researchers point out that price is still mainly established by rule of thumb or by using the cost-plus approach (Bonoma, Crittenden, & Dolan, 1985; Kucher & Hilleke, 1993). Pricing based on marginal costs has the potential for missing a critical element in maximizing profits; the relative perceived value of the product being priced. Therefore, managers are advised to set prices for products according to customers' perceptions of product benefits and costs. It means that the product should be seen through the customers' point of view and priced accordingly. The relationship between alternative prices and the resulting sales quantity is called the price response function (PRE) (Simon, 1989). The price response function of a product or a brand is a tool to understand the effect of price on sales for a product or a brand. Researchers have recommended calibrating the price response function for a product or a brand to find the optimal price for the product or the brand based on conjoint analysis (Dolan & Simon, 1996; Kucher et al., 1993; Simon, 1989, 1992; Yoo, 1991). Conjoint analysis has been recommended as an effective tool for determining the value of a product and deriving price response functions and optimal prices (Dolan & Simon, 1996; Geurts & Whitiark, 1993; Simon, 1989, 1992; Weiner, 1994). This conjoint measurement uses individual customer's preference, which also satisfies the condition of setting the prices from the customers' viewpoint. However, one is still faced with the problem of small data points in the price response function when following the guidelines for calibrating the price response function recommended by Simon (1989) and Yoo (1991). If using four price attribute levels, only four data points on the price response function can be attained. Therefore, the price response function based on traditional conjoint analysis cannot be regarded as a continuous function but rather a discrete function. Yoo and Ohta (1995) use the multinomial logit model to calibrate the continuous price response function. However, under the assumption of Hedonic pricing, they do not use price as an independent attribute to evaluate subject product. Because price has been regarded as one of the most important attributes to evaluate products, Yoo and Ohta's (1995) optimal pricing method can be falsified (Engel, Blackwell, & Miniard, 1995; Jacoby, Olson, & Haddock, 1971; Szybillo & Jacoby, 1974; Wheatiy, Chiu, & Goldman. 1981). The multinomial logit model has also been used to examine various market phenomena such as a price cut effect on a market, market segmentation, choice set change, brand switching, and so on (Bucklin, Gupta, & Han, 1995; Buckiln & Lattin, 1991; Hardie, Johnson, & Fader, 1993; Gupta, 1988). Previous studies have primarily used non-durable products where researchers could obtain the purchase histories of the subject consumers with scanner data. Using the purchase history, researchers were able to estimate parameters suitable to explore the chosen phenomenon. However, obtaining similar consumer information for durable products is difficult due to the relatively long interpurchase time. Several studies investigating individual consumer choice decision-making have concluded that individuals have their own choice set for selecting a certain product category (Han et al., 1995; Shocker, Ben-Akiva, Boccara, & Nedungadi, 1991; Siddarth, Bucklin, & Morrison, 1995). However, the traditional conjoint analysis has only one stage in which respondents are required to express their purchase intentions among the prespecified combinations, although respondents may not consider one of the prespecified combinations as the one they want to buy. That is, the one-stage conjoint analysis can not consider a choice set (Shocker, Ben-Akiva, Boccara, & Nedungadi, 1991). The current research employs a two-staged conjoint analysis. At the first stage, using the profile with researcher specified multiattributes, respondents select the combinations he/she intends to purchase, allowing for identification of each respondent's choice set. Based on the selected choice set, the second stage obtains preference data from each respondent by which the researcher can obtain the unique utility of subject brands. Based on the utility for each respondent, the choice probability is calculated by logit transformation (Green &'Krieger, 1997). Park (1994) and Urban and Hauser (1993) described the method for transforming utility based on conjoint analysis into probability. Using the choice probability, the researcher can calibrate the continuous price response function for each brand. In an effort to overcome the problem of purchase history for durable products, this study purposed using a two-staged conjoint analysis. Instead of purchase history data obtained through scanner data, purchase intention data obtained through conjoint analysis were used to calibrate a price response function. Accordingly, the two-staged conjoint analysis used in this study allowed for calibration of the continuous price response function including choice set information. Statement of Purpose The primary purposes of this study were to develop a new methodology for calibration of the continuous price response function and to compare the price response functions with choice set information to the price response function without choice set information. The purposes were achieved by completing three objectives: (1) to empirically calibrate the price response function of the denim jeans market utilizing a two-staged conjoint analysis with a multinomial logit formation; (2) to compare the differences in the price response functions with and without choice set information; and (3) to compare the optimal price for each brand with and without choice set information. Accordingly, the study endeavored to: 1. Identify the criteria used in evaluating apparel which were used as the attributes in the conjoint analysis; 2. Calibrate the price response function of the denim jeans market by the conjoint analysis and logit transformation; 3. Compare the price response function including choice set information to the price response function without choice set information; 4. Compare the optimal price for each brand with choice set information to the optimal price for each brand without choice set information. Research Ouestions The following research questions were developed to guide the study: 1. Are there differences in the price response functions among brands in the denim jeans market for college students enrolled at a major state supported university? 2. Are there differences in the price response functions with and without choice set information for each subject brand in the denim jeans market for college students enrolled at a major state supported southern university? 3. Are there differences in optimal prices with and without choice set information for each subject brand in the denim jeans market for college students enrolled at a major state supported southern university? Limitations of the Studv This study was limited to the case of a profit-maximizing firm that operates in a static situation with no varying time effect in cost and competitor's response. Further, the study was limited to the products and price ranges selected by the researcher. Therefore, conclusions were limited to the sample selected, specific product categories, and price range. Generalizations beyond this study must be made with caution. Definitions of Terms Choice Set: Brands/services seriously considered just prior to actual purchase (Desarbo & Jedidi, 1995). Conjoint Analysis: A technique which attempts to determine the relative importance consumers attach to salient attributes and the utilities they attach to the level of attributes (Malhotra, 1993). Denim Jeans: Pants made of blue or indigo 100% denim worn by both males and females (Kwon, 1993). Heuristic Optimal Price: Optimal price calculated from profit function within the levels in the price attribute of the conjoint analysis. Mathematical Optimal Price: Mathematical solution derived from profit function consisting of price response function and cost function. National Brands: Manufacturers of a product who sell their brand lines to retail stores across the country (Stone & Samples, 1990). Price Response Function: The relationship between alternative prices and the resulting sales quantity (Simon, 1989). Price Response Function With Choice Set Information: Price response function including brands and services seriously considered just prior to actual purchase (choice set) for calibration of the price response function. Price Response Function Without Choice Set Information: Price response function excluding brands and services seriously considered just prior to actual purchase (choice set) for calibration of the price response function. CHAPTER n REVIEW OF RELATED LFTERATURE This study consisted of three tasks. Thefirsttask was to identify criteria used in evaluating apparel which were used as attributes in a conjoint analysis. The second task was to calibrate the price response function. The last task was to identify the optimal price for each subject brand. To implement thefirsttask criteria used in apparel evaluation were reviewed. To accomplish the second and third task the researcher reviewed literature relating to the price response function, the conjoint analysis as a measurement tool for the price response function, the reliability of the conjoint analysis, the multinomial logit model, and choice set. The Criteria Used in Evaluating Apparel The Criteria Used in Evaluating Products Purchase behavior requires that the customer be able to make judgments and comparisons across products and services. Such judgments are associated with some or all of the various items of information in which the product or service are identified, evaluated, and integrated to form a composite judgment. These judgments are referred to as information cues (Jacoby, Olson, & Haddrock, 1971). Jacoby et al. (1971) classified cues as either intrinsic or extrinsic, suggesting that the attribute of quality can be similariy divided. Intrinsic cues derive from the actual physical product and cannot be changed without changing the product itself (e.g., color or texture). Extrinsic cues are product-related but are not part of the physical product itself. 8 Price, brand name, and level of advertising are examples of extrinsic cues that appear to serve as indicators of quality (Jacoby et al., 1971). There is controversy among studies regarding the effect of extrinsic cues as a determinant of quality. In several multi-cue experiments, some researchers reported that extrinsic cues such as price had a small effect on perceptions of quality (Jacoby et al., 1971; Szybillo & Jacoby, 1974), while other researchers reported the positive effect of extrinsic cues (Wheatiy, Chiu, & Goldman, 1981). In the case of intrinsic cues, however, studies revealed consistent results suggesting that intrinsic cues had a positive effect on the process of product evaluation (Jacoby et al., 1971; Szybillo & Jacoby, 1974; Wheatiy, Chiu, & Goldman, 1981). Variables Used in Evaluating Apparel Using the intrinsic versus extrinsic cues classification, numerous researchers have investigated the importance of the impact of specific variables in evaluation of the perception of quality in apparel. Table 2.1 provides a summary of the research investigating the use of intrinsic and extrinsic cues in the evaluation of apparel quality. Based on the results of these research studies, one can define and categorize intrinsic and extrinsic cues which are statistically significant when evaluating the perception of apparel quality. Examples of intrinsic cues include style, color, fabric content, ease of care, and fit. Extrinsic cues include price, brand name, and store type. Table 2.1. Evaluated Cues in Investigating Apparel Quality. Reference Products Gardner (1971) Men's dress shirts. Suit Price, Brand Davis (1985) Skirt Brand Labeling* Dardis, Spivak, and Shih (1985) Men's dress shirts Durability i^pearance Country of Origin, Price, Brand Hatch and Roberts (1985) Socks, Sweaters Fiber Content Price*, Care, Brand, Seal of Content*, Guarantee,* Country of Origin Stemquist and Davis (1986) Sweaters Lee (1987) Suits Color*, Design* Kim(1988) Jeans, Suits, Coat Fiber Content*, Fit*, Color* Norum and Clark (1989) Women's blazers Fiber Content* Design* Country of Origin, Price*, Store Type*, Store Location* WaUand Heslop (1989) Men's, Women's, and Children's Clothing, Men's and Women's Footwear Value, Styling, Construction, Ease of Care, Fit, Durability, Color Brand Name, Country of Origin Heisey (1990) Sweaters Fiber Content*, Care Country of Origin, Store Type* Workman (1990) Jeans' Hangtag Information Easy of care*, Price* Fit*,Style*,Fabric Lee& Lee (1995) Jeans Pants Fiber Content*, Care*, Design* Intrinsic Cues Extrinsic Cues Price*, Vender Image* Country of Origin Note. * refers to significant variables. 10 Price*, Brand Name* Price*, Brand Name* In addition to Jacoby et al.'s (1971) classification, Eckman, Damhorst, & Kadolph (1990) categorized the evaluation of apparel quality into four criteria types (Table 2.2). Further, they summarized the intrinsic and extrinsic cues used by respondents for evaluating apparel quality from 21 separate studies (Table 2.3). The most frequently used variables in these studies were style, price, brand name, fabric and fiber content, fit and color/design. Pricing Research in Economics and Marketing Price is an important topic in economics and marketing. Price is viewed and determined differentiy depending upon the approach. Simon (1989) suggests a taxonomy of pricing using economics and marketing classifications. He categorizes pricing into three approaches: economics, qualitative/descriptive, and quantitative/methodologybased. The quantitative/methodology-based approach deals with measurement issues associated with optimal price and price response. Because the purpose of the study was to calibrate a price response function in an apparel product, the quantitative/methodology -based approach was employed (Table 2.4). Yoo (1989) reviewed the historical development of pricing research in marketing in Europe and the United States (Figure 2.1). According to Yoo's (1989) classification, pricing issues can be divided into three areas; normative, descriptive, and behavioral (Table 2.5). Researchers utilizing the normative approach have dealt with the setting and measurement issue of price. 11 Table 2.2. Criteria Types of Apparel Quality Evaluation. Criteria Type Examples Aesthetic Stylish, Fashionable, Attractive, Looks Good on Me, Pattern Match Performance Launders Well, Holds Its Shape, Color Fast, Doesn't Shrink, Durable Usefulness Fits Well in Wardrobe, Versatile, Usable Extrinsic Price, Designer Label, Brand Name, Store Where Purchased, Hang Tags, Competition Note. From Clothing and Textile Research Journal, 8(2) (p. 14). by M. Eckman, M. L. Damhorst, & S. J. Kapolph (1990). Toward a model of the in-store purchase decision process: Consumer use of criteria for evaluating women's apparel. Adapted. 12 Table 2.3. Eckman et al.'s (1990) Summary of Intrinsic\Extrinsic Cues Used by Researchers for Evaluating Apparel Quality . Extrinsic criteria No. of citations Intrinsic criteria No. of citations Price 9 Style 13 Brand name or label 9 Fabric & Fiber content 7 Country of origin 5 Color/Design 5 Store 4 Fit 5 Coordination with wardrobe 2 Durability 4 Salesperson's evaluation 2 Comfort 3 Department in store 1 Safety 2 Warranty 1 Care 1 other 12 Total 52 35 Note. From Clothing and Textile Research Jpumal, 8(2) (p. 14). by M. Eckman. M. L. Damhorst, & S. J. Kapolph (1990). Toward a model of the in-store purchase decision process: Consumer use of criteria for evaluating women's apparel. Adapted. 13 Table 2.4. Pricing Research in Economics and Marketing. Approach Issues Researchers Economics The Determination of Prices by Supply and Demand, Economic EflBciency of the Price System, Problems of Market Equilibrium. Coumot, Chamberlin, KreUe, Stackelberg. Qualitative/ Descriptive Directional Recommendation for Pricing. Shapiro, Monroe, Oxenfelt Quantitative/ Methodology-based Pricing Oriented from Econometric Analysis, Measurement Issues (the Conjoint Measurement). Rao, Nagie, Simon, Kucher. Note. From Price managemient (p. 8). by H. Simon, 1989, New York: North-Holland. Adapted. 14 Microeconomics Pricing Theory Qualitative Econometric Business-oriented Approach Approach Pricing Research Descriptive Behavioral Normative Normative Behavioral Descriptive Approach Approach Approach Approach Approach Approach Europe U. S. A. Figure 2.1. Historical development of pricing research in marketing between the U.S. and Europe. Note. From 'Tricing research in marketing: The integration of European and American literature." (p. 171). by P. H. Yoo, 1989, Korean Marketing Review. 4(1). Adapted. 15 Table 2.5. Yoo's (1989) Pricing Classification. Approaches Topics Normative Defensive Pricing Strategy, Pricing a Product Line, Price Determination Via Conjoint Analysis, Quantity Discount, Price Bundling. Behavioral Price Awareness, Price Consciousness, Price as an Indicator of Quality, Perceived Price, etc. Descriptive Decision Process of Price, Determinants for Price. Note. From "Pricing research in marketing: The integration of European and American literature." by P. H. Yoo, 1989, Korean Marketing Review, 4(1). Adapted. 16 Descriptive studies have explored the process of determining price, while behavioral studies have examined the psychological issues associated with price. Based on Yoo's (1989) classification scheme (Table 2.5), the current study utilized a normative approach as a price determination via conjoint analysis. Additionally, because the process of price determination often incorporates psychological aspects of pricing such as price as an indicator of quality, the researcher explored various behavioral characteristics of price. Conclusively, based upon Simon (1989) and Yoo's (1989) classifications, this study utilized both the quantitative/methodology-based approach and the normative approach. Price Response Function In the quantitative/methodology-based approach pricing is dependent on external and internal determinants. External determinants are those beyond the firm's control (e.g., consumer characteristics, market stmcture, legal conditions), while internal determinants are those under the firm's control (e.g., advertising, price, product, distribution). These factors determine the number of product units that can be sold at alternative prices (Yoo, 1991). The relationship between alternative prices and the resulting sales quantity is called the price response function (Simon, 1989). The dependent variable of the price response function is sales volume or market share. Market share is an aggregation of individual customer choice or preference. This individual customer choice or preference can be explored by conjoint measurement. To supplement the price response function two concepts were reviewed; price elasticity and the demand function of price. 17 Price Elasticitv Samuelson and Nordhaus (1992) define price elasticity of demand (or price elasticity) as the responsiveness of the quantity demanded of a good to changes in the good's price, certeris paribus. The precise definition of price elasticity is the percentage change in quantity demanded divided by the percentage change in price. percentage change in quantity demanded Price elasticity = ——— percentage change in price Price elasticity, therefore, provides a measurement of the impact of price on sales. From an economic perspective, the demand function may also be employed as a measurement tool, because the demand curve also indicates the quantity as a function of price. Simon (1989) and Yoo (1991), however, recommend the price response function as a tool to measure impact of price on sales when utilizing the quantitative/methodology-based approach. Demand Function versus Price Response Function Both the demand function and price response function are a measurement of the impact of price on sales. In determining the effect of price on sales, the classic economic pricing model assumes that other marketing variables are held constant. This assumption is evident in the usual treatment of the demand function as a relationship only between quantity demanded and price. But this assumption does not consider the effect of other marketing mix components, such as advertising, selling effort, and competitors' response to the firm's marketing action, on price elasticity and similar interactions. In this regard, the price response function has different characteristics from the demand function. The 18 price response function includes several factors which are not considered in the demand function, such as external determinants and non-price marketing activities. Moreover, sales of a product depend not only on its own price but also on the prices of other products. In this regard, in the process of calibrating the price response function researchers can consider competitive situations, develop several competitive price response models, and use various marketing mix (Yoo, 1991). Therefore, employing the price response function allows for the consideration of more variables than does the demand function and, as a result, is a more useful tool for understanding particular market situations. Competitive Price Response Functions A static price response function can be represented by a mathematical equation. Simon (1989) categorized four types of competitive response models and mathematical equations as follows (Sale quantity = dependent variable, price = independent variable): • the linear model ( q = a - bp), • the multiplicative model (q = ap^ a>0, b<0), • the attraction/multinomial logit model (market share/probability of brand i = e' / J^c" where i = ith brand among m brands, e = base of natural logarithm), • the Gutenberg model ( q = a- bp - c 1 sin /2(c2(pi-pj)) a, b, c 1, c2=parameters). Figure 2.2 illustrates the four models while Table 2.6 contains a selection of numerous calibrations for each of the models. As shown in Table 2.6, numerous researchers have calibrated the price response function using the models and actual market data. 19 Calibration of Price Response Functions Calibration of the price response function requires price data and sale/market share data. Simon (1989) classified the method for calibrating the price response function into four categories; expert judgment, customer survey, price experiment, and collection of actual market data. The customer survey method can further be divided into two different types. Researchers can either directiy ask customers how they would react to certain price levels, price changes, or price differentials, or one can ask customers about their preferences and infer the information on price response from these preference data; direct price response surveys and conjoint measurement. Table 2.7 provides an evaluation of these methods on various criteria. Although Simon (1989) identifies the reliability of conjoint analysis as uncertain, Reibstein, Bateson, and Boulding (1987) reported that conjoint analysis was generally reliable with respect to reliability over the stimulus set, the attributes set, and the data collection procedure. From Table 2.7, one can infer that conjoint analysis is a useful method for estimating price response functions. A number of factors must be taken into account in pricing decisions; the objectives of the organization, customers' willingness to pay for the product, the costs of producing and marketing the product, competition, changes in reservation prices, the costs and the competition over time (Lilien, Kotler, & Moothy, 1992). From the static situation approach to pricing, Simon (1989) and Yoo (1991) indicate that prices are determined by the price response, the cost, and the objective function of a firm. 20 Linear low Attraction Multiplicative high low low high high Gutenberg model low high price Figure 2.2. Competitive price response functions. Note. From Price management (p. 22).. by H. Simon, 1989, New York: North-Holland. Adapted. 21 Table 2.6. Econometrically Calibrated Price Response Function Based on Actual Market Data. Methods Research Linear Peles(1971) Houston-Weiss (1974) Prasad-Ring(1976) Multiplicative Wittink(1977) Yon-Mount (1978) Picconi-01son(1978) Attraction/ Multinomial Logit model Urban(1969) Weiss (1969) Bultez(1975) Guadagni& Little (1983)* Gutenberg Simon (1982) Albach(1973) Kucher(1985) * Multinomial logit model: Following publication of this article numerous studies using the multinomial logit model reported price response in the market. The focus of these articles was primarily non-durable goods (Bucklin & Lattin, 1991; Hardie, Johnson, & Fader, 1993; Gupta, 1988). Note. From Price management (p. 36). by H. Simon, 1989, New York: North-Holland. Adapted. 22 Table 2.7. Comparison of Methods for Collecting Price Response Data. Customer Survey Criteria Expert Judgment Price Actual Experiment Market Data Direct Survey Conjoint Analysis Validity Medium Very Low Medium-high Reliability Mediumhigh Uncertain Uncertain Costs Very Low Low-medium Medium Mediumlow High Low Mediumhigh Depends on Availability High & Accessibility Overall Useful for Questionable Evaluation New Situations Very Useful Useful Useful for Established Products Note. From Price management (p. 36). by H. Simon, 1989, New York: North-Holland. Adapted. 23 Application A price response function is initially employed to explore the price response of a market and to allow for calibration. At this point, it can be used to find the profitmaximizing price. In determining the profit-maximizing price, the firm must consider manufacturing costs and sales quantities for altemative prices (Simon, 1989). The sales quantities for altemative prices can be explored using the price response function. Assume that sales are the function of price and costs are the function of sales (Simon, 1989; Yoo, 1991). Revenue is defined as price p times sale q. Accordingly, profit is Profit = Revenue(p*^)"- Cost(c*^) (2.1) =pq(p) - C[q(p)] where p = price, q = quantity sold, c = variable cost. After calibrating a price response function using the price response data, one can derive a mathematical equation for the price response function. Through application of the equation with respect to price, one can find the revenue-maximization price point (Yoo, 1991). Assume the price response function is linear in the simplest case. Accordingly, ^ = a - b/7 (where a, b = coefficients) qp = ap- bp^ Sqp/Sp = a - 2bpr = 0 =>pr = a/2b (2.2) 24 (where pr = revenue-maximization price). Therefore, the revenue-maximization price point, pr, is a/2b. Further, if the marginal cost is known or can be assumed, the optimal static price. P*, can be calculated. The optimal price refers to the profit-maximization price. If the cost function is also linear, i.e., C = fixed cost + k (a-bp) where k = marginal cost. Accordingly, formula 1 can be expressed as follows; Profit = (a-b/7)/7 - fixed cost - k (a-bp) Setting marginal revenue equal to marginal cost, the optimal static price can be obtained. a - 2bp* = -kb P* = 1/2 (a/b + k). (2.3) Conjoint Analvsis Green and Srinivasan (1978, 1990) defined conjoint analysis as any decompositional method that estimates the structure of a consumer's preferences (i.e., estimates preference parameters such as part-worth, relative importance among attributes, ideal points), given his/her overall evaluations of a set of alternatives that are prespecified in terms of levels of different attributes. On the other hand, Malhotra (1993) defined conjoint analysis as a technique which attempts to determine the relative importance consumers attach to salient attributes and the utilities they attach to the level of these attributes. Czepiel (1992) defined conjoint analysis from the manager's perspective. 25 According to his definition, conjoint analysis is a technique tiiat allows managers to determine what elements in an offering are valued by customers. Model Conjoint analysis is one type of compensatory model. It is a multivariate technique used to understand how respondents develop preferences for products or services. Further, the conjoint measurement produces individual levels of information for the subject product. That is, it produces the individual utility for the identified attributes which enables the investigator to find the individual's overall utility for a given product. The basic conjoint analysis model is represented by the following formula (Malhotra, 1993); where U(x) = overall utility of an alternative, aij = the part-worth contribution or utility associated with the jth level (j,j=l,2,..., ki) of the ith attribute (i,i=l,2,...,m), Xij = 1 if the jth level of the ith attribute is present: 0, otherwise, ki = number of levels of attribute i, m = number of attributes. The basic conjoint equation is one variation of the regression model, where the partworth for the conjoint analysis is the same as the coefficient for the regression equation. 26 The importance of an attribute, Ij, is defined in terms of the range of the partworth, ajj, across the levels of that attribute (Malhotra, 1993): Ij = {max(aij) - min(aij)}, for each i. (2.5) Conjoint Analvsis Procedure • Czepiel (1992) summarizes the conjoint analysis procedure as follows: Step 1: Identify relevant attributes of the product/service. Step 2: Specify the levels of each attribute to include in the study. Step 3: Develop conjoint questionnaire. Step 4: Select respondents and administer conjoint questionnaire. Step 5: Estimate utility of each level of each attribute for each respondent. Step 6: (Optional) Develop preference segments by clustering or grouping customers having similar attribute preferences. Step 7: (Optional) Simulate choice share among altemative company and competitor product/service offerings described by the attributes under study. Choice share can be obtained byfirstestimating utility, which will allow for simulating customer's choices of various product offerings and competitive scenarios. For example, assume that there are only two offerings. One is the firm's offering and the other is the competitor's. If respondent 1 's utility for the firm's offerings is 0.2, and utility for the competitive offering is 0.5, the prediction for this example is that respondent 1 will choose the competitor's offering. By calculating the utility for each respondent in the 27 sample, the market share that the constituents in the market would obtain in the simulated market scenario can be estimated (Czepiel, 1992). Applications Since the mid-1970's conjoint analysis has attracted considerable attention as a method to realistically portray consumer's decisions as a trade-off among multiattribute products or services (Wittink & Cattin, 1989). A recent survey reported applications in the areas of new product, concept identification, competitive analysis, pricing, market segmentation, advertising, and distribution (Wittink, Vriens, & Burhenne, 1994). Other applications included an investigation into the relative importance among attributes in a compensation package (Kwon, 1995) and evaluation of the service quality problem (Desarbo, Huff, Rolandelli, & Choi, 1994). Simon (1989) recommended a conjoint analysis as a calibrating tool for the price response function. Kucher et al. (1993) further indicated that conjoint analysis is a more sophisticated and valid pricing method than using the cost-plus approach, which uses the perceived relative value of products. Finally, Wittink et al. (1994) reported that pricing is the single most frequently identified purpose served by conjoint analyses in Europe, whereas in the United States it was third in frequency. Although conjoint analysis has gained widespread acceptance and use in many industry settings, few researchers have applied it to the field of clothing and textiles. Oh and Huh (1995) used conjoint analysis to determine optimal jeans product development in South Korea. However, they failed to satisfy the conditions of mutual exclusiveness between the levels of the various attributes; a condition recommended by Green and 28 Srinivasan (1990). Rather, Oh et al. (1995) used continuous levels in price; which could have created bias in the results of the conjoint measurement. If they were to have followed Green et al.'s (1990) suggestion, the levels of price should have been 30,000 won ($35), 50,000 won ($60), 70,000 won ($85) with the same interval between levels instead of the less than 30,000 won ($35), between 30,000 won ($35) and 50,000 won ($60), and more than 70,000 won ($85) used in Oh & Hub's (1995) study. The levels Oh et al. (1995) used did not differentiate the two turning price points at 30,000 won ($35) and 50,000 won ($60). Virtually no research has been completed in the textiles and clothing field which utilizes calibration of the price response function by conjoint analysis. In this regard, application of this methodology will contribute significantly to the breadth of knowledge in the textiles and clothing discipline. Data Gathering In order to conduct a conjoint analysis, information must be collected from a sample of consumers. There are many different ways of collecting conjoint measurement data, although two dominate; the full profile and Adaptive Conjoint Analysis (ACA). Wittink et al. (1994) reported that 42% of all conjoint measurement studies were carried out by using the ACA method and 24% used the full profile method, while 34% used methods such as paired comparison and trade-off matrices. 29 Validitv and Reliabilitv Malhotra (1982) examined structural reliability and stability of nonmetric conjoint analysis under conditions of severe stmctures perturbation and substantial variation in the number of stimulus profiles. The results indicated that conjoint analysis is a fairly robust procedure for assessing an individual's preferences. Reibstein et al. (1987) categorized data collection methods in the conjoint analysis into three different data collection procedures; full profile, paired comparisons, and tradeoff matrices. They reported that all three data collection methods showed reliable and stable results, although the degree of reliability differed depending upon the type of reliability measured. Further Reibstein et al. (1987) classified the reliability of conjoint analysis as the reliability over stimulus set, over time, over attribute set, and over data collection procedure and reported them reliable. Jain, Malhotra, and Pinson (1980) investigated the reliability over stimulus set between the trade-off versus the full profile method by using a holdout sample procedure. The researchers concluded, in both cases, that reliability was independent of the data collection method. Segal (1982) examined the reliability over time between the trade-off versus the full profile method, computing a correlation coefficient from the part worth across individuals within each attribute. He concluded that there are significant differences between the methods, with the full profile method performing better. A later study by Leigh, Mackey, and Summers (1984) investigated the reliability over time between the paired comparisons versus the full profile method. This study found no significant differences between methods. According to Segal (1982), one can infer an overall preference for the multiple-factor 30 evaluation, or full profile method, among the data collection methods. Thus, conjoint analysis is regarded as a reliable technique to investigate individual preference. Although ACA is the most popular method for gathering data, there is controversy among researchers regarding the validity of ACA . Green, Kriger, and Agarwal (1991) reported several problems such as a difference in scaling between the second and third steps of the ACA process. Further, Green et al. (1991) cited two papers addressing the limitations of ACA methods; longer interview time and cross-validity of ACA (Agarwal & Green, 1991; Johnson, 1991). Huber, Wittink, Fiedler, and Miller (1993), however, demonstrated that ACA yields cross-validity greater than or equal to the full profile conjoint analysis. The full profile procedure utilizes stimuli described in terms of all critical attributes, and the respondent is required to rank order all possible combinations of attribute levels in the questionnaire. Kucher et al. (1993) strongly recommends selecting the full profile method in commercial use cases only if one does not need more than nine hypothetical products. Green et al. (1990) also points out that in the case of the full profile method if one uses more than seven attributes, or more than 20 profiles, respondents may have difficufty rating their preference. After comparing several data gathering methods. Park (1994) suggested three guidelines for selecting the data gathering method for conjoint analysis: • Use the full profile method if one uses less than six attributes; • Use the trade-off method if one uses seven or eight attributes; • Use the ACA method if one uses more than nine attributes. 31 According to Park's (1994) recommendations, data for the current study was gathered using the full profile method. Multinomial Logit Model and Choice Set Formation The multinomial logit model has been used primarily to examine customer choice behavior, although it has also been used to examine various market phenomena such as market share forecasting, store selection behavior, price cut effect on a market, market segmentation, choice set change, and brand switching (An & Chai, 1993; Bucklin, Gupta, & Han, 1995; Buckiln & Lattin, 1991; Gupta, 1988; Hardie, Johnson, & Fader, 1993). The majority of these studies have utilized non-durables based upon scanner data. Table 2.8 provides an overview of the studies exploring choice behavior and competitive market structure by the multinomial logit model. Several additional studies regard the choice set issue as the most important phenomena to be explored in consumer choice behavior (Engel, Blackwell, & Miniard, 1995; Han & Vanhonacher, 1995; Shocker, Ben-Akiva, Boccara, & Nedungadi, 1991; Siddarth, Bucklin, & Morrison, 1995). Engel et al. (1995) suggest that the pre-purchase altemative evaluation process is one necessary process in their consumer decisionmaking model. The pre-purchase alternative evaluation process consists of two basic components; determining evaluative criteria and determining choice alternatives (Figure 2.3). Han and Vanhonacher (1995) and Siddarth et al. (1995) point out that individual consumer choice decision-making has been followed by his/her own consideration set or choice set formation for selecting a certain product category. Choice set is defined by those brands/services seriously considered just prior to actual purchase (Desarbo & 32 Table 2.8. Examples of Segmentation and Market Stmcture Using Multinomial Logit Model. Researchers Content Products Results Grover & Srinivasan Identifying Market Stmcture (^ ^^^) and Market Segmentation Based on Brand Choice Probability with Promotion. Coffee Kamakura & Russell Segmentation Based on the (1989) Price Elasticity. Anonymous Competitive Food Stmcture among Segments, Vulnerability and Clout. Bucklin & Gupta (1992) Brand Choice and Purchase Incidence by Promotion. Detergents Brand Choice and Purchase Incidence by Promotion at Segments Level, and Identification of the Profile Among Segments. Bucklin, Gupta, & Han (1995) Market Segmentation by Market Response Regarding Change of Marketing Mix. Coffee Brand-specific Response at the Segment Level. 33 Identification of the Distribution of Brands Among Segments. Determine Determine Evaluative Choice Criteria Alternatives Assess Performance of Alternatives Apply Decision Rule Figure 2.3. Basic components of the pre-purchase altemative evaluation process. Note. From Consumer Behavior (p.207). by J.F. Engel, R.D. Blackwell, and P.W. Miniard, 1995, Fort Worth, TX: The Dryden Press. Adapted. 34 Jedidi, 1995). Therefore, the researcher should confirm whether the conjoint analysis can overcome the choice set issue. Shocker et al. (1991) indicates that one of the drawbacks of conjoint analysis is the lack of consideration for the choice set issue. Respondents are required to express purchase intention among all prespecified combinations in the conjoint questionnaire. However, occasionally one of the combinations is not within an individual respondent's choice set. That is, the respondent does not consider a specific combination as a product she/he wants to buy. Therefore, Shocker et al. (1991) points out that conjoint analysis can not overcome the issue of choice set. Accordingly, the researcher should develop an altemative; two-staged conjoint analysis. Table 2.9 presents examples where the multinomial logit model is used with the choice set. Model If individual / confronted with a choice from a set of alternatives, Ci, utility can be expressed as follows, where altemative k is one of the alternatives Ci; U.k = Vik + Eik, (2.6) Vik = Uik + pXik (2.7) where Vik = a deterministic component of i's utility, 8ik = a random component of i's utility, Ujk = an intercept for brand k, P = a vector of coefficients for variables X. Both marketing variables and evaluative criteria are included in the Xik vector. The ek are independently distributed random variables with a double exponential (Gumbel type II extreme value) distribution. 35 Table 2.9. Examples of Choice Sets Formation Using Multinomial Logit Model. Researchers Content Products Results Siddarth, Bucklin, & Morrison (1995) Measurement of Changes of Choice Sets by Promotion. Han & Vanhonacker (1995) Using Price-Quality Tier, Detergents Identifying of Reference Identifying the Changes of Dependence and Loss Choice Sets by Promotion. Aversion in Choice Set Formation and Brand Selection Based on PriceQuality Tier. 36 Liquid Promotion can Expand Detergents Choice Sets, Providing Excluded Brands a Means to Gain Entry and Long Term Sales Benefits. P(ek,< e) = exp[-exp(-e)] - « < e <«, (2.8) Assume that individual / chooses the one with the highest utility among the alternatives. Pik = P{Uik>Uij,JGCi} (2.9) Given assumptions (2.6)-(2.9), the conditional probability of choosing brand j can be expressed by the multinomial logit model (k=l, 2,..., m) as follows; m pik = c\p{Vik) / ^ exp(V.m) (2.10) *=i where Pjk = the probability of choosing brand k. This expression is known as the multinomial logit (Ben-Akiva & Lerman, 1993; Guadagni & Little, 1983). Summarv of Literature Review According to the review of literature, the following ideas were verified. First, using the intrinsic versus extrinsic cues classification, numerous researchers have investigated the importance of the impact of specific variables in evaluation of the perception of quality in apparel. Significant intrinsic and extrinsic cues used in the evaluation of apparel quality include style, price, brand name, fabric andfibercontent, fit and color/design. Second, the price response function of a product or a brand is a tool to understand the effect of price on sales for a product or a brand. The multinomial logit model and conjoint analysis can be used to calibrate the price response function. Third, choice set information is a necessary condition in the process of calibration of price response function. Price response function with choice set information reflects consumer 37 purchasing behavior more faithfully than price response function without choice set information. 38 CHAPTER m METHOD Two purposes were set forth to guide this study: one was to develop a new methodology for calibration of the continuous price response function; the other was to compare the differences in the price response functions with and without choice set information. The first purpose was achieved by the new methodology of two-staged conjoint analysis. Figure 3.1 illustrates the procedures to be followed for calibrating a price response function and calculating the optimal price of a brand. The second purpose of the study was achieved by applying the altemative path illustrated in Figure 3.1. The same evaluative criteria and choice alternatives were applied to both price response functions with and without choice set information. The methodology followed in the study is discussed in the following sections: (a) selection of sample, (b) research instrument, (c) collection of data, (d) calibration of price response function, and (e) statistical analysis of the data. Selection of Sample The primary purposes of this study were to develop a new methodology for calibration of the continuous price response function and to compare differences in the price response functions with and without choice set information. Park (1994) and Urban and Hauser (1993) recommend using at least 100 respondents for a relevant conjoint analysis. Accordingly, in this study no criteria were employed to either admit or exclude 39 ( start j i SPECIFYING CONJOINT MODEL Determining evaluative chteha PRFs WITH CHOICE SET INFO. PATH CHOICE SET FORMATION Detemiining purchasing intention PRFs WITHOUT CHOICE SET INFO. PATH CONJOINT ANALYSIS Numbering the combinations Calculating coefficients for attnbutes levels Calculating utility for each brand 1 MULTINOMIAL LOGIT FORMATION AND CALIBRATING PRICE RESPONSE FUNCTION T COMPARISON BETWEEN PRICE RESPONSE FUNCTION Wrm AND WFTHOUT CHOICE SET INFO. i DISCUSSING RESULTS AND IMPLICATIONS 1 Figure 3.1. Flow of calibration of price response function and optimal price. 40 subjects, and the convenience sample consisted of 103 voluntary respondents enrolled at a major state supported southern university. Research Instmment Data were collected using a two part questionnaire developed by the researcher. Section one was designed to collect preference data from each respondent using the full profile method (Park, 1994). However, due to the large number of potential combinations to be evaluated by respondents, a fractional factorial design was used (Appendix). Three steps were required to complete the first section. Step one provided respondents with both a written and a visual description of two different randomly selected styles and colors of denim jeans. In step two respondents were asked to choose the combination of attributes they intended to purchase. Based upon the literature review, these four attributes included: brand, style, price, and color. Respondents were required to assess their purchase intentions for 32 combinations by marking "Yes" if she/he would purchase a given combination and "No" if she/he would not purchase a given combination (Appendix). This allowed for identification of each respondent's choice set. Utilizing methodology recommended by Park (1994) and Urban and Hauser (1993), instructions in step three required respondents to rate each combination marked "Yes" on a scale of 1-100, with one as least likely to be purchased and 100 as most likely to be purchased. This value served as the dependent variable for estimating the parameters in the model. Green (1978) and Park (1994) recommend ordinal, interval, and ratio scales to measure the preference among combinations. 41 Section two of the questionnaire was designed to elicit demographic information from respondents. Age, gender, marital status, student classification, jean consumption, jean ownership, and ethnic background served as demographic variables. A pretest was conducted to test validity and interpretability of the instmment. Participants consisted of four students at a major state supported southern university. Students were randomly selected and represented both males and females of various ages. Changes in section two of the questionnaire were made as necessitated by the respondents including additional refinements in terms of rewording and reordering of questions. No revisions were required in section one. The revised questionnaire used for data collection appears in Appendix. Attributes and Levels Investigated The research process paralleled the typical procedures for developing a conjoint task (Czepiel, 1993). Selection of product category served as the initial step in the process. Because denim jeans are a frequently purchased product actively marketed by manufacturers and retailers alike, this product category was selected for the current study. After choosing the product category, the investigator selected the attributes for use in the conjoint task. For the model's parsimony only four attributes were selected; intrinsic cues included style and color, while extrinsic cues included price and brand name. These four criteria have frequently been identified as significant cues for evaluating apparel quality (Table 2.1). These four attributes served as independent variables for the conjoint analysis model, while the dependent variable equaled each respondent's purchase intention data numbered 1 to 100. 42 Brand and price levels for the study were determined based upon product availability and brand characteristics at one shopping mall complex with several department stores (Table 3.1). A total of 11 brands were available for consumer purchase. From these 11 brands, eight national brands were selected for inclusion in the study; Levi, Guess, Lee, Rocky Mountain, Pepe, Wrangler, Girbaud, and Dockers. Prices ranged from $23 to $72. Based upon this information, price levels for the study were set at $20, $40, $60, and $80. Two basic styles were randomly selected by the researcher. The "Baggy Fit" featured a straight leg and pleated front, while the "Easy Fit" was designed with a flat front and narrow leg. Examples of both styles were presented to the respondents prior to completion of the questionnaire. Each manufacturer typically uses different terminology for various shades of blue denim. Therefore, the researcher arbitrarily selected two distinctly different shades of blue denim which were available to consumers simultaneously; a deep, unwashed denim referred to as 'Dark Color' and a pre-washed denim referred to as 'Light Color'. Accordingly, the conjoint analysis consisted of eight brands, four prices, two designs, and two colors. Therefore, 128 combinations could be obtained from these levels of attributes. By the fractional factorial design, 32 combinations were identified (Appendix). 43 Table 3.1. Brands and Price Ranges Observed in One Shopping Mall July, 1996. Price Ranges' Maximum Brands Minimum Levi's $30 $42 Guess $42 $68 Lee $23 $30 Gap $29 $58 Calvin Klein $45 $48 Girbaud $49 $72 Wrangler $25 $33 Pepe $49 $49 Arizona $22 $30 Dockers $32 $48 Rocky Mountain $66 $68 * Unit: U.S. dollar 44 Collection of Data Administration of the final questionnaire occurred in August, 1996. The researcher randomly selected two undergraduate courses at a major state supported southern university. Upon receiving instmctor permission students enrolled in two business related courses. Restaurant, Hotel, & Instimtional Management in tiie College of Human Sciences and Information Systems in the College of Business Administration, were asked to voluntarily participate in the study. Of the 104 smdents initially contacted, 103 undergraduate smdents agreed to participate. Of the 103 completed questionnaires, 100% were usable resulting in a 100% response rate. In order to ensure consistency and clarity an assistant was utilized to present verbal instmctions for completion of the instmment. Physical examples of the denim jeans styles and colors selected for the smdy were introduced prior to completion of the questionnaire. Calibration of Price Response Function Model Specification With Choice Set Information By the definition of conjoint analysis and the selected attributes, the conjoint analysis model was specified. The Least Squares method was employed to estimate beta coefficients of the model where the dependent variable equaled each respondent's purchase intention data numbered 1 to 100, and the independent variables were brand name, price, style, and color. Vcik = Ucik + Pcpikpricck + pcsikstylck + pccikcolork where Vck = utility assigned to brand k by consumer i with choice set 45 (3.1) information, Ucik = the intrinsic utility/value of brand k for consumer i with choice set information (brand-specific intercept), pricck = the net available price of brand k for consumer i, stylck = the available style of brand k for consumer i, easy fit is 1; otherwise 0, colork = the available color of brand k for consumer i, dark color is 1; otherwise 0, Pcpik, Pccik, Pcsik = the parameters to be estimated for consumer i with choice set information. The current product profile of each brand investigated is shown in Table 3.2. Using this profile the utility of each brand was calculated and probability was determined based on the utility by the logit transformation. For example, based upon the model and the current profile, Levi's utility function for consumer i was VciLevis — UciLevis "*" Pcpiljevis4U + PcsiLevis I "^ PcciLevis I. The probability of consumer i choosing the Levi brand, according to formula 10, was PiLevis = exp(UciLevis)/{eXp(UciLevis)+exp(UciGuess)+exp(UciLec) + exp(UciGap)+ eXp(UciGirt)aud) + exp(UciWrangler)+exp(UciCK)+eXp(UciArizona)}. Price Response Function and Optimal Pricing Because each brand had unique attributes on offering, it was expected that these differences in attributes would be reflected as differences in the price response. In determining the effect of price on utility, it is assumed that other variables were held 46 Table 3.2. Product Profile of Selected Brands. Brand Price Fit* Color* Levi's $40 Easy Fit Light Color Guess $58 Easy Fit Dark Color Lee $28 Easy Fit Dark Color Pepe $49 Baggy Fit Dark Color Rocky Mountain $68 Baggy Fit Dark Color Girbaud $68 Easy Fit Dark Color Wrangler $30 Easy Fit Light Color Dockers $40 Baggy Fit Light Color Note. * Fit and color are assigned by the researcher arbitrarily. 47 constant. Further, various prices were applied to each brand's utility function thus providing various choice probabilities for each brand corresponding to applied price points. Based upon the continuous linkage among the various mean choice probabilities, the continuous price response function for each brand was obtained. Price response function served as a linkage among the mean choice probabilities corresponding to applied price points. Because the sample size of this study was 103, whenever applying a given price point on the utility function for each brand, the researcher could obtain 103 choice probabilities. Therefore, to calibrate an aggregate price response function, the mean choice probability for each price point was calculated and linked with other mean choice probabilities corresponding to applied price points. These procedures are shown in Figure 3.2. After calibrating the price response function, an approximated function formation, in this case the regression equation, was applied. With this equation and given the cost, the optimal price for each brand was calculated. This study assumed that the cost of each brand was fixed at 30% of the current profile of each brand. This fixed cost method was originally employed in the process of calibration of price response function by Yoo and Ohta (1995). Revenue-Maximizing Price and Static Optimal Price Once the price response function for each brand was determined, the revenuemaximizing price pr was derived. In addition to the revenue-maximizing price, given that marginal cost remained constant at 30% of the current profile of each brand, the optimal static price P* was calculated. Based upon the price response function data, 48 Quantity • Distribution of Choice Probability at a Given Price Point Price Response Function Price Figure 3.2. Linkage among the mean choice probability for price response function. 49 the mathematical price response function was derived by regression analysis. To find the revenue-maximizing price pr for each brand, each mathematical price response function was multiplied by price and derived with respect to price. For example, Levi's price response function was q = 37.91 - 0.2718;? where unit of price is the U.S. dollar, ^p = 37.91 p-0.2718 p ^ (3.2) Formula (3.2) was derived with respect top. 6qp/6p = Zl.91-0.5566 p=Q => pr = 68.159. Optimal static price, according to formula 3, p* = 1/2 (a/b + k)=/7r + k/2 = 68.159 + 6 = 74.159 These results implied that to maximize profit, according to the optimal pricing based on the above equation, the price of Levi jeans should be increased from $40 to $74. Model Specification Without Choice Set Information The conjoint analysis model without choice set information was specified, which was the same formation as the price response function with choice set information. The difference in the conjoint analysis model with and without choice set information stemmed from the use of a different dependent variable. Respondents were required to assess their purchase intention for 32 combinations by marking the box "No" if she/he would not purchase a given combination. This allowed for identification of combinations which were not included in a respondent's choice set. A score of zero was assigned to all purchase intention scores marked "No." The dependent variable of the conjoint analysis model without choice set information 50 equaled each respondent's purchase intention data numbered 0 to 100, while the dependent variable of the conjoint analysis model with choice set information equaled each respondent's purchase intention data numbered 1 to 100. The Least Squares method was employed to estimate beta coefficients for the model with dependent variables equal to each respondent's purchase intention data numbered 0 to 100, and independent variables of brand name, price, style, and color. Vik = Uik + Ppikpricck + PsikStylCk + pcikcolork (3.3) where Vik = utility assigned to brand k by consumer i, Uik = the intrinsic utility/value of brand k for consumer i (brand-specific coefficient), pricck = the net available price of brand k for consumer i, stylck = the available style of brand k for consumer i, easy fit is 1; otherwise 0, colork = the available color of brand k for consumer i, dark color is I; otherwise 0, Ppik, Pcik, Psik = the parameters to be estimated for consumer i. The calibration process from utility to probability for a price response function without choice set information was the same as the calibration of the price response function with choice set information. Using the profile shown in Table 3.2, the utility of each brand was calculated. Probability were determined based on the utility by the logit transformation. Various prices were then applied to each brand's utility function, which allowed for determination of the choice probabilities for respondents corresponding to applied price points. Based upon the continuous linkage among the various mean choice 51 probabilities, a price response function for each brand was obtained. Table 3.3 summarizes the dependent and independent variables used in the price response functions with and without choice set information. Statistical Analvsis of Data Research Question One Research question one sought to compare differences in the price response functions among brands in the denim jeans market for college students enrolled at a state supported southem university. Using ANOVA the differences in mean choice probabilities at four price points for the price response functions among brands were tested. Price levels selected for the study were $20, $40, $60, and $80. Research Ouestion Two Research question two sought to compare differences in the price response functions with and without choice set information for each subject brand in the denim jeans market for college students enrolled in a major state supported southem university. Using the t-test, the differences in mean choice probabilities for the price response functions with and without choice set information for each subject brand were tested. 52 Table 3.3. Dependent and Independent Variables in Price Response Function. Cases Independent Variables Dependent Variable Price Response Function With Choice Set Information Brand, Price, Style, Color Purchase hitention Data (minimum 1, maximum 100) Price Response Function Without Choice Set Information Brand, Price, Style, Color Purchase hitention Data (minimum 0, maximum 100) 53 Research Ouestion Three Research question three was designed to compare differences in optimal price with and without choice set information for each subject brand in the denim jeans market for college students enrolled in a major state supported southem university. After the price response function was derived by regression analysis utilizing the mean choice probability data, optimal price points were calculated for both price response functions with and without choice set information. 54 CHAPTER IV RESULTS OF DATA ANALYSIS The purposes of this research were to develop a new methodology for calibration of the continuous price response function and to compare differences in the price response functions with and without choice set information. Three research objectives were developed to achieve these purposes: (1) to empirically calibrate the price response function of the denim jeans market utilizing a two-staged conjoint analysis with a logit transformation; (2) to compare the differences in the price response functions with and without choice set information; and (3) to compare the optimal price for each subject brand with and without choice set information. The results of the study are reported in the following sections: (a) questionnaire response rate and description of the sample, (b) statistical treatment of the data, and (c) analysis of research questions. Ouestionnaire Response Rate and Description of the Sample Data were obtained by means of a hand distributed questionnaire. A total of 103 students enrolled at a major state supported southern university were asked to participate. All completed questionnaires were usable yielding a total response rate of 100%. Table 4.1 provides an overview of the demographic characteristics of the sample. The sample consisted of both male (72.8%) and female (27.2%) participants, ranging in age from 21 to 40 years with a mean age of 23.1 years (standard deviation (SD) of 3.83). In addition the majority of the sample were single (82.5%). Over half of the respondents were seniors (51.4%), 37.9% juniors and 10.7% sophomores. The majority of 55 Table 4.1. Demographic Characteristics of Sample. Characteristics Frequency Percent Gender Male Female 75 28 72.8 27.2 Age 21-25 26-30 31-35 36-40 81 17 4 1 78.7 16.5 3.9 0.9 Marital Status Single Married 85 18 82.5 17.5 Student Classification Senior Junior Sophomore 53 39 11 51.4 37.9 10.7 Ethnic Group White/Non-Hispanic Hispanic Asian/Pacific Islander Black/Non-Hispanic American Indian/Native American 81 11 5 4 2 78.6 10.7 4.9 3.9 1.9 Jean Consumption 7 times per week 6 times per week 5 times per week 4 times per week 3 times per week 2 times per week 1 times per week 22 13 25 20 11 7 5 21.4 12.6 24.3 19.4 10.7 6.8 4.9 56 Table 4.1. Continued. Characteristics Frequency Percent Denim Jean Ownership (pairs) Jeans Pants 16 or above 11-15 6-10 Below 5 1 6 50 46 0.9 5.8 48.6 44.7 Denim Jean Ownership (pairs) Jeans Shorts 11-15 6-10 Below 5 3 9 91 2.9 8.8 88.3 57 participants were White/Non-Hispanic (78.6%), 10.7% Hispanic, 4.9% Asian/Pacific Islander, 3.9% Black/Non-Hispanic, and 1.9% American Indian/Native American. Jean consumption per week ranged from 0 to 7 wearings, with a mean frequency of 4.7 (SD = 1.74). Neariy a quarter of the sample (24.3%) wore denim jeans 5 times per week, 21.4% 7 times per week, and 19.4% 4 times per week. The mean number of denim jeans pants owned by respondents 6.5 (SD = 4.2), while the mean number of denim jeans shorts was 3.1 (SD =2.7). Statistical Treatment of the Data A variety of statistical methods were used for the study, including ANOVA to test Research Question One, t-test and price elasticity to examine Research Question Two and comparison of optimal price to price response function to explore Research Question Three. All data analysis was conducted using the Statistical Analysis System (SAS) package. A probability level of 0.05 or less was considered statistically significant. Research Ouestion One Research Question One sought to compare differences in the price response functions among brands in the denim jeans market for college students enrolled at a major state supported southern university. Data addressing this question were analyzed using a two step process: (1) calculation of respondent's choice probabilities of four price points on the price response function for eight brands; and (2) application of ANOVA, testing the null hypothesis (Ho: pi = p2 = p3 =p4 = p5 =p6 = p7 = p8) against the altemative hypothesis (HI: pi ^A [i2 ^ [13 ^ [i4 :^ \i5 ^ [i6 ^ \il ^ \iS), io determine 58 whether differences existed in mean choice probabilities, at four points, for the price response function among brands. Results of the ANOVA for the four price points both with and without choice set information are presented in summary tables 4.2, 4.3, 4.4, and 4.5. No significant differences in mean choice probabilities were found at the 0.05 level of significance for any of the four price points either with or without choice set information. Research Ouestion Two Research Question Two sought to compare differences in the price response functions with and without choice set information for each subject brand in the denim jeans market. Data addressing this question were analyzed using the following steps: (1) calculation of the utility function of each brand through conjoint analysis: (2) application of various prices to each brand's utility function to determine respondent's choice probabilities, under the assumption that the prices of other brands be fixed; (3) determination of a price response function for each brand from the mean choice probabilities; and (4) application of the t-test to determine differences in mean choice probabilities for the price response functions with and without choice set information for each brand. 59 Table 4.2. Summary ANOVA for $20 Price Point. Source DF Sum of Squares Mean Square F Value Pr>F 0.04314365 0.00616338 0.03 0.9999 0.18237039 0.24 0.9755 With Choice Set InformationModel Error 816 148.81423729 Total 823 148.85738094 Without Choice Set InformationModel 0.21230474 0.03032925 0.12677096 Error 816 103.44510050 Total 823 103.65740524 60 K Table 4.3. Summary ANOVA for $40 Price Point. Source DF Sum of Squares Mean Square F Value Pr>F 0.30 0.9545 With Choice Set InformationModel 0.0096019 0.0013717 0.00459492 Error 816 3.74945085 Total 823 3.75905285 Without Choice Set InformationModel 0.01739671 0.00248524 0.00275666 Error 816 2.24943329 Total 823 2.26683000 61 1.90 0.5045 Table 4.4. Summary ANOVA for $60 Price Point. Source DF Sum of Squares Mean Square F Value Pr>F 0.00372790 0.00053256 0.12 0.9973 0.00461343 0.40 0.9014 With Choice Set InformationModel Error 816 3.76455627 Total 823 3.76828417 Without Choice Set InformatipnModel 0.00714201 0.00102029 0.00253905 Error 816 2.07186491 Total 823 2.07900692 62 Table 4.5. Summary ANOVA for $80 Price Point. Source DF Sum of Squares Mean Square F Value Pr>F 0.00383624 0.00054803 0.01 0.9999 0.06452644 0.02 0.9999 With Choice Set Information-r Model Error 816 52.65357306 Total 823 52.65740930 Without Choice Set InformationModel 0.00728097 0.00104014 0.06805649 Error 816 55.53409708 Total 823 55.54137804 63 Figure 4.1 illustrates the price response functions with and without choice set information for each of the eight brands. In each case as prices increased a decrease in choice probability was revealed. For seven of the brands, this phenomenon was evidenced between the price range of $20-$60. The range for the last brand was slightly greater from $20 to $67.50. There after as price increased mean choice probability increased slightly. Utilizing the t-test, differences between the price response functions with and without choice set information were examined for each brand (Table 4.6). Significant differences in mean choice probabilities were found at the 0.05 level of significance at several price points on the price response functions with and without choice set information. To further examine the differences in the price response functions with and without choice set information, a comparison was made of price elasticities. Although the results of the t-tests indicated differences in mean choice probability at a given price point between the price response functions with and without choice set information, the ttests did not examine the differences between two price points on the price response functions. Table 4.7 illustrates the differences in price elasticity on the price response function. All brands appear to have lower price elasticities with choice set information than without choice set information at the $20 price point, while higher price elasticity is evidenced with choice set information from $25 to $45 than without choice set information. Based upon the results and these of the t-tests, it can be considered that the price response function 64 Levi's Prize Response Functbn 0.8 •H 0.7 XI 0.6 iH •H (0 XI 0.5 0 u 0.4 Ch Q) U 03 •H 0 0.2 CJ 0.1 20 25 30 35 40 45 50 55 prce ($) -•—wih choi:e setiifc>inati)n 60 65 70 75 80 wihoutchoie setiifcmatbn (a) Levi's Price Response Function Guess" Proe Response Function J 20 I 25 I \ 30 1 L ^Hi-4^-i-A-4-4-i^^^ $ I $ r r t 35 40 45 50 55 p i t e ($) 60 65 70 75 80 wih c h o h e setiifcimatin —•—wAoutchoie setiifcmatijn (b) Guess' Price Response Function Figure 4.1. Price response functions with and without choice set information. 65 Lee's P i t e Rei^x^nse Functbn ttJAtiiiiiHI 0 20 25 30 35 40 45 50 55 60 65 70 75 80 prize ($) wJSi choi:e s e t n f o m a t b n —•—wihoutchoire s e t n f o m a t b n (c) Lee's Price Response Function Pepe's Prize Response Function A A m : t-M-U-*-M-M 20 25 30 35 40 45 50 55 60 65 70 75 80 prize ($) wih choize sethfcinatiDn —•—wihoutchoize s e t n f o m a t b n (d) Pepe's Price Response Function Figure 4.1. Continued. 66 kA^^i«4»« Gi±)aud's Prize Response Functbn •H n T U ./ •H 0.6 «0 0.5 g|[ - Xi 0 0.4 04 u 0) 03 V - u 02 xi 0.1 u 0 \ 1 •H O N^ 1 20 25 1 1 30 1 ' 1 7 T ¥ 35 40 W W • 45 50 • • P P T 55 60 p 1 1 65 70 1 1 75 1 1 80 prize ($) with c h o i e s e t n f o m a t b n —»—wfiioutchoize s e t n f o m a t b n (e) Girbaud's Price Response Function W langifer's Prize Re^xDnse Functbn 0.7 0.6 - XI (0 0.5 XJ - iH •H o 0.4 u - 03 - u 02 - x: 0.1 u 0 - Ch 0) •H 0 20 25 30 35 40 45 50 55 prize ($) 60 65 70 75 80 wlh choize setiifomatbn —•—wihoutchoize setiifomatbn (f) Wrangler's Price Response Function Figure 4.1. Continued. 67 R o c l ^ M oimtaii's Prize Response Functijn iiiiiili^-l-4^-l-i 20 25 30 35 40 45 50 55 60 65 70 75 prize ($) 80 ••—wih choize setiifomatbn —•—wiftioutchoize setiifomatbn (g) Rocky Mountain's Price Response Function Docker's Price Response Functbn l^m i H I M M M t 20 25 30 35 40 45 50 55 60 65 70 75 prize ($) -•—wih choize setiifomatiDn 80 wihoutchoize setiifomatbn (h) Docker's Price Response Function Figure 4.1. Continued. 68 Table 4.6. The Results of T-Test at Various Price Points. Price $20 $25 $30 Levi's * * * Guess * j_^g * Pepe * j ^ j ^ * wv rjinp^lpr ^ Dockers * $40 * * ^ $45 $50 $55 $60 ^ ^i Hf * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * Girbaud $35 ^ • • • • * * * * * Note. * denotes that there is a significant difference in mean choice probability at a given price with 95% confidence. 69 Table 4.7. The Differences in Price Elasticity on the Price Response Function. Price Brand Choice Set Information $20 $25 $30 $35 $40 $45 Levi's With Without -0.040 -0.053 -0.063 -0.055 0.000 0.000 -0.017 -0.015 -0.010 -0.010 -0.007 -0.007 Guess With Without -0.040 -0.052 -0.064 -0.051 0.000 0.000 -0.016 -0.013 -0.009 -0.008 -0.006 -0.006 Lee With Without -0.025 -0.049 -0.044 -0.066 0.000 0.000 -0.043 -0.027 -0.025 -0.017 -0.017 -0.012 Pepe With Without -0.039 -0.051 -0.063 -0.048 0.000 0.000 -0.018 -0.012 -0.011 -0.008 -0.007 -0.006 Girbaud With Without -0.039 -0.051 -0.062 -0.048 0.000 0.000 -0.018 -0.012 -0.010 -0.008 -0.007 -0.006 Wrangler With Without -0.040 -0.052 -0.064 -0.050 0.000 0.000 -0.016 -0.012 -0.009 -0.008 -0.006 -0.006 Rocky Mountain With Without -0.039 -0.053 -0.063 -0.057 0.000 0.000 -0.019 -0.017 -0.011 -0.011 -0.008 -0.008 Dockers With Without -0.040 -0.052 -0.063 -0.051 0.000 0.000 -0.017 -0.013 -0.010 -0.009 -0.007 -0.006 70 with choice set information has a different stmcture than the price response function without choice set information. Research Ouestion Three Research Question Three was designed to compare differences in the optimal price of the price response functions with and without choice set information for each subject brand in the denim jeans market. The optimal price points were determined using the following steps: (1) estimation of the price response function for each brand through regression analysis; (2) determination of cost for each brand; and (3) calculation of the profit-maximization price point within the levels in the price attribute. Utilizing regression analysis parameter estimates for the price response functions were attained. With an R~ spread from 0.922 to 0.967, all parameters were significant at the 0.001 level. Under the assumption that the cost of each brand is fixed at 30% of its current profile and based upon the estimated price response function, the optimal price for each brand was attained (Table 4.8). 71 Table 4.8. Estimated Price Response Function and the Optimal Price. Brands Cases Estimated Price Response Fimction Costs Optimal Price Levi's WithCSI* Without CSI 1.753-0.075p + 0.0O08p^ (R^ = 0.933) 2.259-0.095p + 0.001p^ pi^ = 0.966) $12 $12 $25.0 $24.5 Guess WithCSI Without CSI 1.745-0.074p + 0.0007p^ (R^ = 0.931) 2.202-0.094p + 0.001p^ (R^ = 0.962) $17.4 $17.4 $26.0 $25 5 Lee WithCSI Without CSI 1.819-0.074p + 0.0008p^ (R^ = 0.938) 2.315 - 0.094p + 0.0009p^ (R^ = 0.965) $8 4 $8.4 $26.5 $25.5 Pepe With CSI 1.773 - 0.075p + O.OOOSp^ (R^ = 0.937) Without CSI 2.143 - 0.092p + 0.0009p^ (R^ = 0.961) $14.7 $14.7 $25.0 $24.5 R.M. WithCSI 1.764-0.074p + 0.0008p^ (R^ = 0.936) Without CSI 2.157 - 0.093p + 0.0009p^ (R^ = 0.962) $20.4 $20.4 $26.6 $26.5 Girbaud With CSI 1.655 - 0.069p + 0.0007p^ (R^ = 0.922) Without CSI 2.067 - 0.087p + 0.0009p^ (R^ = 0.952) $20.4 $20.4 $26 5 $26.6 Wrangler With CSI 1.778 - 0.075p + O.OOOSp^ (R^ = 0.936) Without CSI 2.275 - 0.096p + O.OOlOp^ (R^ = 0.967) $9 $9 $25.3 $24.9 Dockers $12 $12 $25.0 $22.5 WithCSI 1.756 - 0.075p+ 0.0008p^ Without CSI 2.188- 0.094p + 0.0009p^ (R^ = 0.934) (R^ = 0.964) Note. *CSI(Choice Set Information). All coefficients are significant at 99% level. 72 CHAPTER V SUMMARY, CONCLUSIONS AND FUTURE RESEARCH The price response function of a product or a brand is a tool to help understand the effect of price on sales for that product or brand. Researchers have recommended calibrating the price response function to find the optimal price for the product or brand based on conjoint analysis (Dolan & Simon, 1996; Kucher et al., 1993; Simon, 1989, 1992; Yoo, 1991). However, utilizing the traditional conjoint analysis limits the results due to the problem of small data points in the price response function and therefore must be regarded as discrete rather than continuous. Several studies investigating individual consumer choice decision-making have concluded that individuals form their own choice set just prior to actual purchase (Han et al., 1995; Shocker, Ben-Akiva, Boccara, & Nedungadi, 1991; Siddarth, Bucklin, & Morrison, 1995). However, the traditional conjoint analysis can not consider a choice set (Shocker, Ben-Akiva, Boccara, & Nedungadi, 1991). In this regard, the price response function with choice set information reflects consumer purchasing behavior more faithfully than does the price response function without choice set information. However, virtually no research has been reported in the pricing field which utilizes the calibration of the continuous price response function with choice set information by conjoint analysis. The primary purposes of this study were to develop a new methodology for the calibration of a continuous price response function and to compare differences in the price response functions with and without choice set information. This chapter contains 73 the summary of the study, interpretation of results, conclusions, implications, and recommendations for future research. Summarv of the Studv A new methodology for calibration of a continuous price response function including choice set information was developed to meet the purposes of the study. The new methodology, identified as the two-staged conjoint analysis, incorporated the logit transformation into the conjoint analysis and added the choice set formation step in the conjoint questionnaire. Two conjoint models were specified to compare the difference in the price response functions with and without choice set information. These two models were tested in a study of the denim jeans market. Participants included 103 students at a major state supported southern university. In a conjoint model, the independent variables consisted of brand, price, color, and style, while the dependent variables were equal to each respondent's purchase intention data. The two-staged conjoint questionnaire consisted of three steps. Step one provided respondents with both a written and a visual description of two randomly selected styles and colors of denim jeans. In step two, respondents were asked to choose the combination of attributes they intended to purchase. In step three respondents were asked to rate each combination chosen at step two on a scale of 1-100, with one as least likely to be purchased and 100 as most likely to be purchased. Three research questions were developed to accomplish the purposes of this study. These questions were intended to; (1) examine the differences in the price 74 response function among brands; (2) compare the differences in the price response functions with and without choice set information; and (3) compare the optimal price of the price response functions with and without choice set information for each brand. A variety of statistical methods, including ANOVA, t-test, and regression analysis, were employed to analyze the research questions. Interpretation of Results Research Ouestion One The first research question investigated whether differences existed in the price response functions among brands in the denim jeans market for college students enrolled at a major state supported southem university. The results of ANOVA revealed no significant differences in the purchase probabilities among the eight brands at $20, $40, $60, and $80. The results can be explained by one of characteristics of model specification; a strong effect on utility by the shared attributes among brands. The model used in this study was based on the model specification of Guadagni and Little (1983), Han (1993), and Bucklin, Gupta and Han (1995). According to their models, the deterministic component of a respondent's utility for brandy is expressed as a linear function of the attributes of brand/ These attributes can then be divided into two categories; attributes unique to brandy and attributes common in all brands (Guadagni & Little, 1983). In this study, the brand specific constant was used as the attribute unique to brand j , while price, style, and color were used as the attributes common in brand;. Accordingly, each respondent shared the same price, color and style coefficient for all 75 subject brands. Table 5.2 summarizes the difference in the relative importance among the attributes used in conjoint analysis. With choice set information the relative importance of price, style, and color is 74.8%, while brand is only 25.2%. Without choice set information, the relative importance of attributes common to the brandy drops slightly (67.1%), while the relative importance of the brand remains fairly low at 32.9%. In conclusion, in the model used, the relative importance of attributes common to the brands had a strong effects on utility, resulting in no differences in the choice probabilities among brands. Research Ouestion Two Research Question Two investigated whether there were differences in the price response functions with and without choice set information in the denim jeans market for college students enrolled at a major state supported southem university. Results of the ttest and price elasticity showed that the price response functions with and without choice set information had different price response stmctures (Tables 4.6 & 4.7). The different shapes of the price response functions with and without choice set information resulted from the different coefficients of the attributes between the two conjoint models (Table 5.1). The different coefficients of the attributes resulted from the difference in the variance of the levels in the attributes. Each respondent's choice set information affected the variance of the levels in the attributes of the conjoint models. The price response function for each brand included a price range where mean choice probability increased with each incremental increase in price. This phenomena 76 Table 5.1. The Coefficient Mean of the Conjoint Model. Attributes Mean Beta Coefficient With Choice Set Information Mean Beta Coefficient Without Choice Set Information Attributes common to all brands Color -0.00489 Style -0.00697 0.07880 Price -0.30509 -0.27871 Wrangler -0.02760 -0.13417 Rocky Mountain -0.04440 -0.08024 Girbaud 0.05814 0.01741 0.06231 -0.0353 0.01008 -0.08602 -0.03111 0.03169 0.04972 0.08229 0* 0* 1.08358 Attribute unique to brand Pepe Lee Guess Levi's Dockers Note. * Dockers is a reference brand to estimate parameters for brand. 77 Table 5.2. The Relative Importance Among the Attributes. Attributes Relative Importance With Choice Set Information Relative Importance Without Choice Set Information Price 72.4% 41.9% Style 1.5% 12.3% Color 0.9% 12.8% Brand 25.2% 32.9% Total 100% 100% Attributes common to all brands Attribute unique to brand 78 can be explained by the signal theory of price, which suggests that if price is a reliable signal for quality of apparel products in the U.S. market, the increase of the choice probability can be interpreted in terms of the price-quality relationships (Gardner, 1971; Gerstner, 1985; Lee, 1996). That is, if consumers regard price as the indicator of quality, when price increases, the choice probability of a product can increase for a productspecific price range. Research Question Three Research Question Three sought to compare differences in the optimal price of the price response functions with and without choice set information. In this study, the optimal price with choice set information was different from the optimal price without choice set information (Table 4.8). The different shapes of the price response functions with and without choice set information resulted in the differences in optimal price. In this study, the optimal price was sought within the levels in the price attribute of the conjoint model. The profit-maximization price points as the mathematical solution of the profit function could be found outside of the levels in price attribute. However, the work conducted by Yoo and Park (1996) concluded that "the Heuristic Optimal Price (HOP)" within the levels in the price attribute is more valid for determination of the optimal price, because the price coefficient of the conjoint model is estimated by the variance of levels in the price attribute. Therefore, the optimal price reported in this study is the Heuristic Optimal Price within the levels of the price attribute, not the mathematical optimal price for subject brand (Yoo & Park, 1996). 79 Conclusions and Implications The purposes of the study were to suggest a new methodology for calibration of the continuous price response function and compare differences in the price response functions with and without choice set information. The research objectives were intended to: (1) examine the differences in the price response function among brands; (2) compare the differences in the price response functions with and without choice set information; and (3) compare the optimal price of the price response functions with and without choice set information for each brand. The results of ANOVA revealed no significant differences in the purchase probabilities among the eight identified brands at $20, $40, $60, and $80. However, results of the t-test and price elasticity showed that the price response functions with and without choice set information had different price response stmctures (Tables 4.6 & 4.7). The different shapes of the price response functions with and without choice set information resulted in the differences in optimal price (Table 4.8). The results of this study have implications for market researchers, brand managers and retailers as they seek to improve consumer satisfaction and increase practitioner profits. Retailers of the 1990's are continually challenged to differentiate product offerings in an intensely competitive market place and understanding consumer choice decision making is fundamental to successful product positioning. The results of this research provide a more effective tool for evaluating consumer purchase practices thereby providing the basis for sound marketing mix decisions. Tested on one apparel product category, calibration of the product response function with choice set information clearly reflected consumer purchase behavior more faithfully than did price response 80 function without choice set information. Further, results suggested that optimal price varies among consumer groups. This would suggest that retailers should consider tailoring price to specific target markets through application of optimal price calculations. Ultimately this would allow retailers greater opportunity to maximize profits and to more effectively price goods to meet consumer expectations. Suggestions for Future Studv The purposes of the study were to develop a new methodology for calibration of the continuous price response function and compare differences in the price response functions with and without choice set information. Although the primary purposes were successfully achieved, the following are suggestions for further research related to the results: 1. Develop a new model for calibration of the price response function to include only the deterministic component with attributes unique to subject brand in the utility function. 2. Explore the price-quality relationship to interpret thefluctuationof choice probability. Future research should explore the price response function, explaining the price-quality relationship. 3. Replicate the study using evaluative cues other than color and style. For parsimony this study used only four attributes in the utility function. However, in the case of the price response function with choice set information, color and style did not play a major role in affecting the utility for each brand. Therefore, other evaluative cues may elicit significantly different results. 81 4. Conduct future research which consider competitive situations. This study assumed that when the price of a brand changes in utility function, other brands do not take any action in price and other marketing variables. Game theory can provide a framework for analyzing the competitive market situation. 5. Replicate the current study utilizing product categories where validity tests are available. This study did not include a procedure to test the validity of the new methodology with market share data. However, the real market data available in the U.S. apparel market was based on the company level rather than brand level. 82 REFERENCES Agarwal, M.K. & Green, P.E. (1991). Adaptive conjoint analysis versus self-explicated models: Some empirical results. International Journal of Research in Marketing, 8 (June), 141-146. An, K.H. & Chai, S.I. (1993), An Empirical study on store selection behavior exploration by multinomial logit model. Korean Management Review, 22(2). 101-120. Bearden, W. O., Netemeyer, R. G., & Mobley, M. F. (1993). Handbook of Marketing Scales: Multi-Item Measures for Marketing and Consumer Behavior Research. London: Sage. Ben-Akiva, M. & Lerman, S. R. (1993). Discrete Choice Analvsis: Theory and Application to Travel Demand. London: The MIT Press. Blattberg, R., Buessing, T., Peacock, P., & Sen, S. (1978). Identifying the deal prone segment. Journal of Marketing Research, 15(Aug.), 369-377. Boulinger, M. S. (1993). Merchandise Buving (4th ed.), Needham, MA: Allyn and Bacon. Bucklin, R.E. & Gupta, S. (1992), Brand Choice, Purchase Incidence, and Segmentation: An integrated Modeling Approach, Journal of Marketing Research, 29 (May), 210-215. Bucklin, R. E., Gupta, S., & Han, S. (1995) A brand's eye view of response segmentation in consumer choice behavior. Journal of Marketing Research, 32(Feb.), 66-74. Bucklin, R. E. & Lattin, J.M. (1991). A two-state model of purchase incidence and brand choice. Marketing Science, IQ(Winter), 24-40. Chowdhary, U. (1989). Fashion information seeking by younger and older consumers. Clothing and Textile and Research Journal, 8( 1), 49-55. Clarke, K., & Belk, R. (1979). The effects of product involvement and task definition on anticipated consumer effort. In W. L. Wilkie (Ed.), Advances in Consumer Research, 6. 313-318, Ann Arbor: Association for Consumer Research. Cohen, J. B., & Goldberg, M. E. (1970). The dissonance model in post-decision product evaluation. Journal of Marketing Research, 7, 315-321. Czepiel, J.A. (1992). Competitive Marketing Strategy. Englewood Cliffs, NJ: PrenticeHall. 83 Davis, L. L. (1987). Consumer use of label information in ratings of clothing quality and clothing fashionability. Clothing and Textile Research Journal. 6( 1), 8-14. Dardis, B., Spivak, S., & Shih, C. M. (1985). Price and quality differences for imported and domestic men's dress shirts. Home Economics Research Journal. 13. 391399. Desarbo, W. S., & Jedidi, K. (1995). The spatial representation of heterogeneous consideration sets. Marketing Science. 14(3), 326-342. Dobson, G., & Kalish, S. (1993). Heuristics for pricing and positioning a product-line using conjoint and cost data. Management Science. 39(2). 160-175. Dolan, R.J., & Simon, H. (1996), Power Pricing. New York, NY: Free Press. Eckman, M., Damhorst, M.L., & Kapolph, S. J. (1990). Toward a model of the in-sore purchase decision process: Consumer use of criteria for evaluating women's apparel. Clothing and Textile Research Journal. 8(2). 13-22. Engel, J. F., Blackwell, R. D., & Miniard, P. W. (1995). Consumer Behavior (8th ed.). New York, NY: Dryden Press. Gardner, M. D. (1971). Is there a generalized price-quality relationship? Journal of Marketing Research, 8, 241-243. Gay, L. R. (1987). Educational Research: Competencies for Analvsis and Application (4'^ ed.). New York, NY: MacMillian Publishing Company. Gerstner, E (1985). Do higher prices signal higher quality? Journal of Marketing Research, 22, 209-215. Geurts, M. D., & Whitiark, D. (1993). Forecasting market share. Journal of Business Forecasting, 11(4). 17-22. Green, P. E., & Krieger, A. M. (1991). Segmenting markets with conjoint analysis. Journal of Marketing, 55(Oct.). 20-31. Green, P. E., & Krieger, A. M. (1992). Modeling competitive pricing and market share: Anatomy of a decision support system. European Journal of Operational Research. 60(1). 31-44. Green, P. E., & Krieger, A. M, (1997). Using conjoint analysis to view competitive interaction through the customer's eyes. In G.S. Day and D.J. Reibstein (Eds.), Wharton on Dynamic Competitive Strategy. New York, NY: John Wiley & Sons. 84 Green, P. E., Krieger, A. M., & Agarwal, M. K. (1991). Adaptive Conjoint Analysis: Some caveats and suggestions. Journal of Marketing Research. 28(Mav). 215222. Green, P. E., & Srinivasan, V. (1990). Conjoint analysis in marketing: New development with implications for research and practice. Journal of Marketing. 54. 3-19. Grover, R. & Srinivasan, V. (1987), A Simultaneous Approach to Market Segmentation and Market Stmcturing, Journal of Marketing Research. 24(Mav). 139-153. Guadagni, P. M., & Littie, J.D.C. (1983), A logit model of brand choice calibrated on scanner data. Marketing Science. 17(August). 319-350. Gupta, S. (1988). Impact of sales promotions on when, what, and how much to buy, Journal of Marketing Research. 25(Nov.). 342-355. Han, S. & Vanhonacker, W.R. (1995). Reference dependence and loss aversion in choice set formation and brand selection. Working paper. Hong Kong University of Science and Technology. Hatch, K., & Roberts, J. A. (1985). Use of intrinsic and extrinsic cues to assess textile product quality. Journal of Consumer Studies and Home Economics, 9, 341-357. Hardie, B.G., Johnson, E.J., & Fader, P.S. (1993). Modeling loss aversion and reference dependence effects on brand choice. Marketing Science, 12(Fall), 378-394. Heisey, L. F. (1990). Perceived quality and predicted price: Use of the minimum information environment in evaluating apparel. Clothing and Textiles Research Journal, 8, 22-28. Howard, J. A., & Sheth, J. N. (1969). The Theory of Buyer Behavior. New York, NY: Radius Press. Huber, J., Winntik, D.R., Fiedler, J.A., & Miler, R. (1993). The effectiveness of altemative preference elicitation procedures in predicting choice. Journal of Marketing Research, 30 (Febmary), 105-114. Hwang, S., & Jung, C. (1993). An exploratory study on apparel distribution system and its countermeasure in the view of market liberalization. Journal of the Korean Society of Clothing and Textile, 17(2). 265-279. Jacoby, J., Olson, C. J., & Haddock, A. R. (1971). Price, brand name, and product composition characteristics as determinants of perceived quality. Journal of Applied Psychology. 55. 570-579. 85 Jain, A. K., Malhotra, N. K., & Pinson, C. (1980). Stability and reliability of part-worth utility in conjoint analysis: A longitudinal investigation. European Institute of Business Administration. Working Paper 80/05. Johnson, R.M. (1991), Comment on Adaptive Conjoint Analysis: Some caveats and suggestions. Journal of Marketing Research, 28(Mav). 223-225. Kamakura, W. A. & Russell, G.J. (1989), A probabilistic choice model for market segmentation and elasticity stmcture. Journal of Marketing Research, 26 (November), 379-390. Kim, M.Y. (1988), An study on clothing evaluation criteria with apparel product category. Journal of the Korean Consumer Economic. 26(3). 1-12. Kucher, E., & Hilleke, K. (1993). Value pricing through conjoint analysis: A practical approach. European Management Journal. 11(3). 283-290. Kwon, E. Y. (1993). The study on brand image in jeans casual wear. Unpublished master's thesis. Sung Kyun Kwan University, Seoul, Korea Kwon, H. (1995). The study on the employee segmentation and the product policy for the internal customers: Focused on the compensation package for the advertising agency employees. Unpublished doctoral dissertation. Sung Kyun Kwan University, Seoul, Korea Lee, D.H. (1996). The relationship between price and objective apparel quality: A comparison between the United States market and Korean market. Unpublished master's thesis. Texas Tech University, Lubbock, TX. Lee, H. J. (1987). An study on the market segmentation by the life-style of consumers: With women in labor market in Seoul. Unpublished master's thesis, Seoul National University, Seoul, Korea Lee, H. J. (1993). An experimental study on the fashion merchandising system: With special reference to the life-style of consumers and the marketing strategy for the fashion industry. Unpublished doctoral dissertation, Sook Myung Women's University, Seoul, Korea. Leigh, T. W., Mackey, D. B., & Summers, J. O. (1984). Reliability and validity of conjoint analysis and self explicated weights: A comparison. Journal of Marketing Research, 21. 456-462. Lennon, J. S., & Fairhurst, E. A. (1994). Categorization of the quality concept. Home Economics Research Journal. 22. 267-285. 86 Lilien, G. L., Kotler, P., & Moothy, K. S. (1992), Marketing Models. Englewood Cliffs, NJ: Prentice-Hall. Ludbner-Pupert, J., & Winakor, G. (1985). Male and female style preference and perceived fashion risk. Home Economics Research Journal. 13(3). 256-266. Malhotra, N. K. (1982). Stmctural reliability and stability of nonmetric conjoint analysis. Journal of Marketing Research. 19. 199-207. Malhotra, N. K. (1993). Marketing Research: An Applied Orientation. Englewood Cliffs, NJ: Prentice-Hall. Mansfield, E. (1988). Microeconomics: Theory and Applications (6th ed.). New York: W.W. NORTON. Morganosky, M. A. (1990). Store and brand type influences on the perception of apparel quality: A congmity theory approach. Clothing and Textiles Research Journal. 9 (1), 45-49. Nagle, T. (1984). Economic foundations for pricing. Journal of Business. 57(1). 2-25. Nomm, S. P., & Clark, A. L. (1989). A comparison of quality and retail price of domestically produced and imported blazers. Clothing and Textile Research Journal, 7, 1-9. Park, C. S. (1994). Conjoint analysis. In Yoo, P. H. (Ed.). Modern Marketing Science, (pp. 121-185). Seoul: Bummunsa. Park, Y.S. (1996). The study on the service pricing: Focused on the non-linear pricing for the shipping industry. Unpublished doctoral dissertation. Sung Kyun Kwan University, Seoul, Korea Reibstein, D., Bateson, J. E. G., & Boulding, W. (1987). Conjoint analysis reliability: Empirical findings (Report No. 87-101). Cambridge, MA: Marketing Science Institute. Samuelson, P. A., & Nordhaus, W. D. (1992). Economics (14th ed.). New York, NY: McGraw-Hill. Segal, M. N. (1982). Reliability of conjoint analysis: Contrasting data collection procedure. Journal of Marketing Research, 19. 139-143. Shocker, A.D., Ben-Akiva, M., Boccara, B., & Nedungadi, P. (1991). Consideration set influences on consumer decision-making and choice issues, models, suggestions. Marketing Letters. 2(3). 181-197. 87 Siddarth, S., Bucklin, R.E., & Morrison, D.G. (1995). Making the cut; Modeling and analyzing choice set restriction in scanner data. Journal of Marketing Research, 32(Aug.), 225-266. Simon, H. (1989). Price Management. New York, NY: North-Holland. Stemquist, B., & Davis, B. (1985). Store status and country of origin as information cues: consumer's perception of sweater price and quality. Home Economics Research Journal, 15. 124-131. Stone, E. & Samples, J.A. (1990). Fashion merchandising: An introduction (5*'' ed.). New York, NY: McGraw-Hill. Szybillo, J. G,. & Jacoby, J. (1974). Intrinsic versus extrinsic cues as determinants of perceived product quality. Journal of Applied Psychology, 59. 74-78. Tigert, D. J., Ring, L. R., & King, C.W. (1976). Fashion involvement and buying behavior: A methodological study. In B. B. Anderson (Ed), Advances in Consumer Research Vol. 3 , (pp. 46-52), Provo, UT: The Association for Consumer Research. Traylor, M. B., & Joseph, B. W. (1984). Measuring consumer involvement in products. Psychology & Marketing, 1(2), 65-76. Urban, G. L. & Hauser, J. R. (1993). Design and Marketing of New Product (2nd ed.). Englewood Cliffs, NJ: Prentice-Hall. Wall, M., & Heslop, A. L. (1989). Consumer attitudes towards the quality of domestic and imported apparel and footwear. Journal of Consumer Studies and Home Economics, 13, 337-358. Weiner, J. (1994). Forecasting demand: Consumer electronics marketer uses a conjoint approach to configure its new product and set the right price. Marketing Research: A Magazine of Management & Applications, 6(3), 6-11. Wheatley, J. J., Chiu, S.Y. J., & Goldman, A. (1981). Physical quality, price, and perceptions of product quality: Implications for retailers. Journal of Retailing. 5L 100-115. Wittink, D. R., & Cattin, P. (1989). Commercial use of conjoint analysis: An update. Journal of Marketing. 53. 91 -96. Wittink, D. R., Vriens, M., & Burhenne, W. (1994). Commercial use of conjoint analysis in Europe: Results and critical reflections. International Journal of Research in Marketing. 11.41-52. 88 Workman, J. E. (1990). Effects of fiber content labeling on perception of apparel characteristics. Clothing and Textiles Research Journal, 8 (3), 19-24. Yoo, D. & Ohta, H.(1995), Optimal pricing and product-planning for new multiattribute products based on conjoint analysis. International Journal of Production Economics. 38, 245-253. Yoo, P. H. (1989). Pricing research in marketing: The integration of European and American literature. Korean Marketing Review, 4(1). 168-200. Yoo, P. H. (1991). Pricing Policy. Seoul: Pakyoungsa. Yoo, P. H., Dolan, R. J. & Rangan, V. K. (1987), Dynamic pricing strategy for new consumer durables, Zeitschrift fur Betriebswirtschaft, 57(Oktober), 1024-1043. Yoo, P.H. & Park, Y.S. (1996, November). The studv on the service pricing: Focused on the non-linear pricing for the shipping industry. Paper presented at the Fall Conference of Korean Marketing Society, Chunju, Korea. 89 APPENDDC QUESTIONNAIRE 90 Questionnaire This questionnaire is designed to examine your purchase intention of blue jeans. Please answer each items as accurately as possible. Your identity will be kept confidential. The questionnaire and the demographic information sheet will be coded and kept in a secure place. All responses on the questionnaires will be reported as group data. Thank you for your assistance in the study. 91 Section 1. Step 1: You have been shown two pair of blue jeans which are further described below. Use these descriptions to complete the following questionnaire. Style Baggy Fit Easy Fit features straight leg and pleated front narrow leg and flat front color Dark Color Light Color features ' Unwashed Pre-washed Easy Fit + Light Color Baggy Fit + Dark Color 92 Step 2: Based upon the combination of attributes identified in each row, please indicate your purchase intentions by marking the box Yes or No. Step 3: For each of the items marked yes, please assign the number (1-100) which most accurately reflects your purchase intention. 1 =least acceptable No. Brand 100=most acceptable Price Style Color Yes or No Number 1 - 100 1 Levi's $20 Easy Fit light blue 2 Levi's $40 Baggy Fit dark blue 3 Levi's $60 Easy Fit light blue 4 Levi's $80 Baggy Fit dark blue 5 Guess $20 Baggy Fit dark blue 6 Guess $40 Easy Fit light blue 7 Guess $60 Baggy Fit dark blue 8 Guess $80 Easy Fit light blue 9 Lee $20 Easy Fit dark blue 10 Lee $40 Baggy Fit light blue 11 Lee $60 Easy Fit dark blue 12 Lee $80 Baggy Fit light blue 13 Pepe $20 Baggy Fit light blue 14 Pepe $40 Easy Fit dark blue 15 Pepe $60 Baggy Fit light blue 16 Pepe $80 Easy Fit dark blue 17 Girbaud $20 Easy Fit light blue 18 Girbaud $40 Baggy Fit dark blue 19 Girbaud $60 Easy Fit light blue 20 Girbaud $80 Baggy Fit dark blue 93 No. Brand Price Style Color Yes or No Number 1-100 21 Wrangler $20 Baggy Fit dark blue 22 Wrangler $40 Easy Fit light blue 23 Wrangler $60 Baggy Fit dark blue 24 Wrangler $80 Easy Fit light blue 25 Rocky Mountain $20 Easy Fit dark blue 26 Rocky Mountain $40 Baggy Fit light blue 27 Rocky Mountain $60 Easy Fit dark blue 28 Rocky Mountain $80 Baggy Fit light blue 29 Dockers .$20 Baggy Fit light blue 30 Dockers $40 Easy Fit dark blue 31 Dockers $60 Baggy Fit light blue 32 Dockers $80 Easy Fit dark blue 94 Section 2. Please give us some information about yourself. 1. Age: 2. Gender: 1. Female 2. Male 3. Marital status: 1. Married 2. Single 4. Classification: 1. Freshmen 2. Sophomore 3. Junior 4. Senior 5. How often do you wear jeans pants/jeans shorts each week? ( days/week) 6. How many of following do you own? Jeans pants ( pairs) Jeans shorts ( pairs) 7. Race 1. White/Non Hispanic 2. Black/ Non Hispanic 3. Hispanic 4. Asian/Pacific Islander 5. American Indian/Native American 6. Other Thank you. 95 PERMISSION TO COPY In presenting this thesis in partial fulfillment of the requirements for a master's degree at Texas Tech University or Texas Tech University Health Sciences Center, I agree that the Library and my major department shall make it freely available for research purposes. Permission to copy this thesis for scholarly purposes may be granted by the Director of the Library or my major professor. It is understood that any copying or publication of this thesis for financial gain shall not be allowed without my further written permission and that any user may be liable for copyright infringement. Agree (Permission is granted.) Student's Signature Xy / Dafe Disagree (Permission is not granted.) Student's Signature Date