“Methods for measuring willingness to
pay: The case of high-priced products”
Master’s Thesis Marketing Management
By: Rick Nieters
Completion date: 24 August 2012
University of Groningen
Faculty of Economics and Business
“Methods for measuring willingness to
pay: The case of high-priced products”
Master’s Thesis Marketing Management
By: Rick Nieters
Borkerstraat 8
7891 SG Klazienaveen
Telephone number: 0630062152
Email:
r.b.nieters@student.rug.nl
Student Number: s1640658
Completion date: 24 August 2012
University of Groningen
Faculty of Economics and Business
P.O. Box 800, 9700 AV Groningen, the Netherlands
First supervisor:
Dr. Thorsten Wiesel
ABSTRACT
Knowledge about a consumer’s willingness to pay (WTP) for a product plays a crucial role in many areas of marketing management like pricing decisions and new product development. Not surprisingly, therefore, numerous approaches for measuring WTP with differential conceptual foundations and methodological implications have been proposed in the marketing literature so far. This paper provides a systematic review of the available methods, evaluates their strengths and weaknesses, and thereby provides marketing managers with a basis for selecting a suitable WTP method. In addition, the suitability of two specific methods (open-ended contingent valuation and choice-based conjoint analysis) for measuring WTP for a relatively high-priced, infrequently purchased product category (digital cameras) is empirically tested by examining face validity, internal validity and current market prices.
The results indicate that significant differences exist between the WTP estimates obtained from open-ended contingent valuation and choice-based conjoint analysis. After a comparison with current market prices, choice-based conjoint analysis has been found to perform better than open-ended contingent valuation. Most likely, this result depends on the product category analyzed. Further studies are needed to determine which factors influence the suitability of methods for measuring WTP.
PREFACE
With consumer research as my prime interest during my studies, I wanted to combine a study of consumer behavior with a practical research method. I was particularly interested in conjoint analysis as a method to measure consumers’ willingness to pay. In today’s highly competitive market, conjoint analysis is one of the most important tools to support product development, pricing and positioning decisions in marketing practice. Studying this method and its corresponding analytical techniques has been a great, informative way to finish my studies in the form of this thesis.
TABLE OF COTETS 1. Introduction 6 2. Literature Review 8 2.1 Conceptual definition 8 2.2 Methods to measure WTP 9 2.3 Contingent valuation 11 2.4 Conjoint analysis 12
2.5 Market data analysis 14
2.6 Experimental auctions 15
2.7 Incentive-aligned conjoint analysis 17
2.8 Comparison of methods 17 2.9 Evaluation of methods 18 3. Research Methodology 21 3.1 Data collection 21 3.2 Experimental design 21 3.3 WTP estimation procedure 25 3.4 Validity measures 27 4. Results 29 4.1 Data description 29 4.2 Data quality 30 4.3 Face validity 31 4.4 Open-ended CV 31 4.5 CBC analysis 32
4.6 Comparison with current market prices 34
5. Conclusion & Recommendations 36
References 39
1. ITRODUCTIO
Despite considerable advances in both academic and applied pricing research over the past decades, many companies still make their pricing decisions without a thorough understanding of the likely response of (potential) buyers and competitors to alternative price quotations. As a result of missing adequate knowledge of the consumers’ willingness to pay (WTP) for their products, these companies fail to pursue a pricing strategy that is appropriately customized to their marketing environment and thus risk to ignore valuable sources for increasing profitability of the products offered. Empirical studies show that even minor variations in prices and the corresponding consumer behavior can have a sustainable impact on revenues and profits (Garda & Marn, 1993; Marn, Roegner & Zawada, 2004).
Companies often adopt some business rules and follow a strategy that could be denoted as “intuitive” pricing (Breidert, Hahsler & Reutterer, 2006). Remarkably, such behavior is not limited to retailing or service industries only, where mark-up pricing is still the predominant practice (Levy, Grewal, Kopalle & Hess, 2004; Berman & Evans, 2001; Monroe, 2003). Several recent studies reveal that only 8 to 15% of all companies develop pricing strategies based on buyer response behavior (Monroe & Cox, 2001). In contrast to what seems to be common practice, managers consider the knowledge of consumers’ responses to different prices as a requisite for formulating competitive strategies, conducting value audits, and developing new products (Anderson, Jain & Chintagunta, 1993).
Researchers agree with managers on the importance of valid WTP estimates. Jedidi & Zhang (2002) consider valid estimates of WTP essential for developing an optimal pricing strategy. WTP estimates are also important for forecasting market response to price changes and for modeling demand functions. Similar arguments about the importance of WTP and perceptions of value by customers can be found in many other studies on WTP (Shaffer & Zhang, 1995; Backhaus, Wilken, Voeth & Sichtman, 2005; Jedidi & Jagpal, 2009). Furthermore, various approaches for measuring brand equity (e.g. Farquhar, 1989; Park & Srinivasan, 1994) emphasize consumers’ WTP in terms of the (monetary) added value provided by a brand to a specific product compared with its competitors or an unbranded baseline product.
choose that method which is appropriate for designing an optimal price schedule. There is no simple solution to this problem because consumers’ true WTP is an unobservable construct (Voelckner, 2006). Each method for measuring WTP only represents an attempt to come as close as possible to the truth. Therefore, analyzing and understanding potential sources of differences in WTP estimates that emerge from value elicitation studies represent an important step in research on how managers should estimate consumers’ WTP.
Despite this fact, the research area today is rather fragmented and there have been relatively few attempts to synthesize the literature on WTP as well as what is known from empirical research regarding the effectiveness and efficiency of the available approaches. The main objective of this paper is to provide some insights into this direction. The strengths and weaknesses of individual approaches will be discussed and evaluated from a managerial point of view. Moreover, due to their practical relevance, special focus will be put on direct surveying techniques (i.e. contingent valuation) and conjoint-based applications. Where many studies have tested these approaches for relatively inexpensive, frequently purchased products (e.g. mineral water, shampoo, shower gel and beer, Natter & Feurstein, 2002; cola and cake, Wertenbroch & Skiera, 2002; cleaning product, Miller, Hofstetter, Krohmer & Zhang, 2011), this study will examine their suitability for a relatively higher-priced, less frequently purchased product category (i.e. digital cameras). The two approaches will be used to measure consumers’ WTP for this product and the resulting WTP estimates will be compared. In summary, the analyses are driven by the following two research questions:
Which method should be used to measure consumers’ WTP for high-priced products? Are there differences between the WTP measured by contingent valuation and conjoint
analysis?
2. LITERATURE REVIEW
In this chapter, I provide a review of methods for measuring WTP and the corresponding estimation procedures. For the sake of clarity, I start with a conceptual definition of WTP, after which I provide references to related theoretical and empirical work, and discuss the most important advantages and drawbacks of each WTP method.
2.1 Conceptual definition
A consumer’s WTP or reservation price1 can be defined as ‘the price at which a consumer is
indifferent between buying and not buying the product’ (Jedidi & Zhang, 2002). This standard
definition is often accompanied by the following equation:
, − − 0, = 0 2.1 In this equation, a consumer with income y considers whether or not to buy a product g priced at p, where U(g,y – p) is the utility from buying the product and U(0,y) the utility from not buying it. Given that the reservation price R(g) for product g is the price at which the consumer is indifferent between buying and not buying the product, the utility of buying the product at R(g) minus the utility of not buying it, is zero.
Jedidi & Zhang (2002) show that, under fairly general assumptions about the consumer’s utility function, the reservation price R(g) always exists, so that for any p ≤ R(g) the consumer is better off purchasing the product. They also show that if the utility function is quasi-linear2, then faced with a choice among G products (g = 1, …, G), a utility-maximizing consumer will need to know only his or her WTPs for the product offerings and the corresponding prices of these products to make the optimal choice decision.
These theoretical claims imply that knowing a consumer’s WTP for the products in a product category is sufficient to predict whether or not a consumer will buy from the respective category and which of these products he or she will choose. Specifically, a consumer will choose the product option that provides the maximum surplus (R(g) – p), given that p ≤ R(g) (Jedidi & Jagpal, 2009). He or she will not buy from the category if the maximum surplus
1
Consistent with the literature, I shall use the term ‘willingness to pay’ interchangeably with ‘reservation price’. 2
across products is negative (i.e. for each product in the category, the consumer’s WTP is less than the price of that product).
Other definitions of WTP have been proposed as well. Hauser & Urban (1986) define WTP as the minimum price at which a consumer will no longer purchase the product. Similarly, Kalish & Nelson (1991) define WTP as the maximum amount of money a consumer is willing to pay for a given quantity of a product. Varian (1992) defines WTP as the price at or below which a consumer will purchase one unit of the good. Ariely, Loewenstein & Prelec (2003) suggest a more flexible definition of WTP. They argue that there is a threshold price up to which a consumer definitely buys the product, another threshold above which the consumer simply walks away, and a range of intermediate prices between these two thresholds in which consumer response is ambiguous. In order to harmonize these alternative definitions, Wang, Venkatesh & Chatterjee (2007) suggest that one should distinguish three reservation prices:
1) floor reservation price, the maximum price at or below which a consumer will definitely buy one unit of the product;
2) indifference reservation price, the maximum price at which a consumer is indifferent between buying and not buying; and,
3) ceiling reservation price, the minimum price at or above which a consumer will definitely not buy the product.
2.2 Methods to measure WTP
Reservation prices can be estimated through numerous approaches and estimation procedures. The primary distinctions among these approaches are whether they determine consumers’ hypothetical WTP or actual WTP and whether they measure WTP directly or indirectly (Miller et al., 2011).
In contrast, methods that elicit actual WTP require that consumers p price of the chosen product
auctions (e.g. Vickrey auctions, Vickrey, 1961) DeGroot & Marschak, 1964) and
Akiva, Bradley, Morikawa, Benjamin, Novak, Oppewal require real economic commitments from consumers, they reveal their true WTP.
In order to structure this paper and organize the various WTP measurement methods, I use the classification framework based on the above
2.1. On the highest level, methods are distinguished whether they determine consumers’ hypothetical WTP or actual WTP
and indirect methods.
Figure 2.1: Classification framework for WTP methods
In the following sections, I address the
their suitability for estimating WTP. In this process, I consider several evaluation criteria. first criterion concerns the validity of estimations
providing an incentive to consumers to reve
related to the conventional criteria of measurement reliability validity. The second criterion
Hypothetical WTP Direct Contingent valuation Indirect Conjoint analysis
In contrast, methods that elicit actual WTP require that consumers pay the stated price or the (Voelckner, 2006). Typical examples of such methods are auctions (e.g. Vickrey auctions, Vickrey, 1961), lotteries (e.g. the BDM procedure, Becker,
and market data analysis (e.g. panel or store scanner data, Akiva, Bradley, Morikawa, Benjamin, Novak, Oppewal & Rao, 1994). Since these methods require real economic commitments from consumers, they provide them with an incentive to
e this paper and organize the various WTP measurement methods, I use the classification framework based on the above-mentioned distinctions as presented in Figure 1. On the highest level, methods are distinguished whether they determine consumers’ or actual WTP. Subsequently, they are further divided into direct methods
Classification framework for WTP methods
, I address the strengths and weaknesses of these approaches to assess their suitability for estimating WTP. In this process, I consider several evaluation criteria.
the validity of estimations. That is, how accurate is the method in providing an incentive to consumers to reveal their true WTP? Note that this issue
criteria of measurement reliability and internal and external criterion relates to the flexibility of the method to include new
WTP measurement Indirect Conjoint analysis Actual WTP Direct Market data
analysis Experimental auctions aligned conjoint
ay the stated price or the . Typical examples of such methods are , lotteries (e.g. the BDM procedure, Becker, .g. panel or store scanner data,
Ben-). Since these methods provide them with an incentive to
e this paper and organize the various WTP measurement methods, I use the mentioned distinctions as presented in Figure 1. On the highest level, methods are distinguished whether they determine consumers’ . Subsequently, they are further divided into direct methods
approaches to assess their suitability for estimating WTP. In this process, I consider several evaluation criteria. The . That is, how accurate is the method in that this issue is closely and internal and external the flexibility of the method to include new
Indirect
price/product combinations. This issue is especially important for estimating WTP for new products that have not yet been made available in the market. The third criterion pertains to the method’s ability to obtain information on consumers’ choice behavior. This information is essential for estimating cross-price effects among new and competing products. The fourth criterion concerns the method’s ability to estimate individual parameters, which can be used to account for heterogeneity in consumer preferences. The fifth and last criterion relates to the data collection procedure, which determines to a great extent the time and the costs of the method. Considering the scope of this research, these are crucial factors in selecting a feasible method for measuring WTP.
2.3 Contingent valuation
With contingent valuation (CV), consumers explicitly state their WTP for a specific product either by indicating whether they would buy a product at a given price (closed-ended or discrete-choice CV) or by directly stating their WTP for a product (open-ended CV). These methods are popular in measuring WTP in agricultural economics and in determining the economic impact of changes in social policies (Jedidi & Jagpal, 2009).
In discrete-choice CV, respondents are asked to decide whether they would buy a product at a certain price (Wertenbroch & Skiera, 2002). A yes response thus indicates that the respondent is willing to pay at least the listed price for the product. When these yes responses are aggregated across respondents, one obtains a demand curve that shows how the proportion of yes responses differs across the experimentally manipulated price levels (Jedidi & Jagpal, 2009). Estimating subsequent WTP levels is straightforward using a binary choice model such as logit or probit (Cameron & James, 1987). In such choice models, the decision of whether or not to buy is modeled through a latent utility function that depends on product characteristics and consumer background variables (Jedidi & Jagpal, 2009).
observe whether an individual’s WTP is higher or lower than the listed price. Hence, it may require large samples or multiple replications per respondent to obtain accurate results.
In open-ended CV, respondents are directly asked to state their WTP (e.g. Mitchell & Carson, 1989, “… the following offers are submitted to you. Please indicate the maximum price you are willing to pay for each offer.”). As a result, individual WTP estimates can be obtained directly from the survey data, which makes it perhaps the easiest method to implement. Another advantage of this method lies in the data generation process: it does not impose high demands on respondents’ cognitive capabilities. As a result, the abandonment rate is relatively low (Backhaus et al., 2005).
However, for a number of reasons, open-ended CV is likely to lead to inaccurate results. Perhaps the most serious problem is the widespread evidence of hypothetical bias of self-stated reservation prices (Frykblom, 2000; Botelho & Pinto, 2002; Voelckner, 2006). This type of bias appears when, placed in a hypothetical situation, particularly in the context of a questionnaire, respondents do not take into consideration all the constraints that would affect their choices in a real situation (e.g. budget available, financial consequences of a poor choice, availability of competitor’s products). Therefore, there is a difference between what the respondents say and what they would accept to pay in a real situation (Le Gall-Ely, 2009).
Other researchers attribute hypothetical bias to the strategic answering behavior of consumers caused by their interest in keeping prices down or in affecting a company’s decision to offer a product (Cameron & James, 1987; Monroe, 2003). A third group argues that the bias may be caused by a measurement error which is ascribed to the assumption that a self-stated reservation price is deterministic (Jedidi, Jagpal & Manchanda, 2003). In particular, this may cause serious problems for new products or products that are infrequently purchased, because consumers have not yet internalized a WTP for these products.
2.4 Conjoint analysis
composed of various attributes (characteristics) and levels. Consumers place a certain utility (value) on each of these characteristics and can determine the overall utility of any product by summing up the value of its parts (levels). By observing how consumers evaluate products in response to changes in the underlying attribute levels, the impact of each attribute on overall product preference, and thus WTP, can be estimated (Orme, 2002).
The literature generally categorizes conjoint analysis into two major families: rating-based and choice-based conjoint (CBC) analysis. In rating-based conjoint studies, researchers present consumers with a number of hypothetical product profiles (concepts) and ask them to rate these profiles on a preference scale (Elrod, Louviere & Davey, 1992; Guyon & Petiot, 2011). The other approach, CBC analysis, requires consumers to make a succession of choices among several sets of hypothetical product profiles and a no-purchase option (see Figure 2.2 for an example). The no-purchase option is critical in this experimental design because it makes the experimental setting more realistic, helps eliminate statistical biases, and improves demand estimates (Parker & Schrift, 2011).
Figure 2.2: Example choice set in a CBC study (digital cameras)
Product 1 Product 2 Product 3 /o-purchase option
16 megapixels 10 megapixels 12 megapixels
I would not buy any of the offered
products
10x zoom 10x zoom 5x zoom
Ultra-compact body Available in bright colors In-camera filter effects
3D photography GSP functionality Touch screen
169 EUR 149 EUR 129 EUR
□
□
□
□
Since the estimated utilities from conjoint analysis are based on simulated choices, instead of real purchases, they are subject to hypothetical bias. Previous research findings, however, show that the responses to CBC questions are generally similar to those from experiments based on real purchase acts (Carlsson & Martinsson, 2001). In the few cases where the differences in the results from the two methodologies are statistically significant, the differences are small (Lusk & Schroeder, 2004). Hence, CBC analysis seems to reduce the hypothetical bias to a minimum compared to other methods.
Another advantage of CBC analysis is that it provides additional information about reservation prices and consumers’ choice behavior. For example, where discrete-choice CV only provides information about whether or not a product is chosen/purchased, CBC provides detailed information about the situation in which a product is not chosen. Specifically, one can distinguish whether a consumer who does not purchase the product chooses another product (brand) alternative or the non-purchase option. This enables a firm to distinguish how much of the demand for a product comes from brand switching, cannibalization and market expansion.
2.5 Market data analysis
Methods based on market data are often used to estimate price-response functions. Generally, these methods analyze panel data (i.e. individual purchase data reported by members of a consumer panel) or store scanner data (i.e. sales records from retail stores). As such, they provide two important advantages. Because the input data come from actual purchases, these methods are incentive compatible and do not suffer from hypothetical bias (Jedidi & Jagpal, 2009). Consumer panel data, for example, provide useful information about consumers’ responses to price changes of an existing brand and those of its competitors. Such information is useful for estimating the effect of a price change on category incidence, brand choice and quantity decisions (Jedidi, Mela & Gupta, 1999).
panel involves high operating costs. Moreover, it is often questionable whether the panel adequately represents the market (Nagle & Holden, 2002).
Furthermore, using historical market data is based on the assumption that past demands can be used to estimate future market behavior. This implies that the product for which future demand is predicted has only been exposed to minor variations in the product profile. A problem then arises if the historical data do not contain the required price variations to cover the desired spectrum of WTPs. This is especially true for new products where price optimization has to be done before any purchase data exist.
For new products, simulated test market methods such as ASSESSOR (Silk & Urban, 1978) and AC Nielsen BASES can be used. These methods provide consumers with the opportunity to buy new products at experimentally manipulated price points. In ASSESSOR, for instance, consumers are first shown advertisements for the new and existing products. Then, they are given an amount of seed money that they can keep or use to buy any of the available products displayed in a simulated store. This experimental design provides data on how the demand for the new product varies across the listed prices.
2.6 Experimental auctions
Auction-based methods are beginning to gain popularity in marketing because they measure real and not self-stated choices. The most-cited among these methods are first-price auctions, Vickrey auctions and BDM lotteries (Wertenbroch & Skiera, 2002; Voelckner, 2006; Jedidi & Jagpal, 2009; Miller et al., 2011).
In first-price auctions, each bidder submits one sealed bid to the seller. This information is not given to the other bidders. Once bids have been made, the person with the highest bid wins the auction and has to pay his or her bid price (McAfee & McMillan, 1987). Given this auction mechanism, each bidder has an incentive to bid less than his or her reservation price in order to get a better deal. In particular, the person with the highest reservation price may not always submit the highest bid. First-price auctions are therefore not incentive compatible.
bid. With this mechanism, Vickrey auctions provide bidders with an incentive to reveal their true valuation, because they must buy the product if they win the auction and the winning bid does not determine the purchase price (Kagel, 1995).
Another incentive compatible auction form is the well-known BDM procedure introduced by Becker et al. (1964). In this procedure, each respondent is directly asked to state the maximum price at which he or she is willing to buy the product. A purchase price is randomly determined from a distribution of prices (e.g. drawn from a lottery). If the respondent’s bid exceeds this value, he or she is required to purchase the product at the purchase price. If the bid is lower than the drawn price, the respondent does not purchase the product. Since the purchase price is randomly drawn, instead of determined by the respondent’s bid, this method is optimal for rational respondents to reveal their actual reservation prices (Wertenbroch & Skiera, 2002).
Nevertheless, experimental auctions still may conceal several background variables that are likely to affect elicited reservation prices. The bidding process, for example, does not naturally mimic the actual decision-making process that a consumer goes through in normal retail settings (Hoffman, Menkhaus, Chakravarti, Field & Whipple, 1993). Contrary to the practically unrestricted supply of products in retail settings, bidders in auctions compete with one another for a limited stock. This is not only unrealistic for many goods, but it also encourages participants to bid more than the true worth of the product to ensure that they are placing the winning bid (Noussair, Robin & Ruffieux, 2004). This kind of gambling behavior may consequently limit the validity of experimental auctions under practical conditions (i.e. WTP estimates elicited in auctions may differ from point of purchase situations).
2.7 Incentive-aligned conjoint analysis
A new stream of literature is emerging that combines the advantages of the stated preference methods (i.e. methods based on survey data) with the incentive compatibility of the BDM procedure. Ding, Grewal & Liechty (2005), for example, extended the open-ended CV and CBC methods using incentive structures that require real financial commitments from respondents. Under CV, respondents were given a menu of 12 Chinese dinner specials (without price information) and were asked to state their WTP for each meal. They were told upfront that a random procedure would be used to select a meal from the menu and that they would receive this meal if their WTP exceeded a randomly drawn price. In the incentive-aligned CBC procedure (ICBC), respondents were presented with 12 choice sets of three Chinese meals each (with price information) and had to choose one meal from each set, or the no-purchase option. Next, a lottery was used to randomly draw one choice set and respondents would receive the meal that they selected from that choice set. With more realistic economic incentives for survey respondents, the out-of-sample predictions of these two methods outperformed those of the standard self-stated WTP and conjoint methods.
Under practical circumstances, however, managers might prefer a hypothetical payment context. They might have this kind of preference because introducing a buying obligation for all respondents can be costly, particularly in terms of administrative costs and the costs of complying with all purchase obligations, or because of severe liquidity constraints that would bias WTP downward in a real context (Wertenbroch & Skiera, 2002; Voelckner, 2006). The need for real products to comply with purchase obligations may also pose problems (e.g. in early stages of product development).
2.8 Comparison of methods
choices derived from the data of the three surveys were compared with the actual choices of the respondents among the holdout products. The results indicate that the predictive validity of both conjoint models clearly outperforms the model fit obtained from the directly-stated WTPs (62% vs. 46% of first choices correctly predicted).
More recently, researchers tested various WTP measurement approaches in terms of their external validity. Voelckner (2006) performed an empirical comparison of the first-price auction, the Vickrey auction, open-ended CV and conjoint analysis. She estimated WTP for prepaid telephone cards, and tested for external validity by requiring a sub-sample of respondents to actually purchase a telephone card at the indicated WTP. The results of her study show that WTP is significantly higher under hypothetical conditions where the respondents do not have to make a purchase. In real conditions, with a purchase at the end, the estimated WTPs are significantly lower. Similarly, Lusk & Schroeder (2004) compared hypothetical and non-hypothetical responses to choice experiment questions involving rib-eye steaks with differing quality attributes. Their results indicate that consumers report substantially higher WTP under hypothetical response formats than under non-hypothetical conditions. These findings are consistent with related studies, for example by Wertenbroch & Skiera (2002) and Sattler & Nitschke (2003).
In addition to comparing hypothetical and non-hypothetical methods, prior studies have compared direct and indirect approaches to measure consumers’ WTP. Several studies confirm the existence of a significant difference between hypothetical direct approaches (contingent valuation) and hypothetical indirect approaches (conjoint analysis). Backhaus et al. (2005), for example, compare the respective hypothetical biases of open-ended CV and rating-based conjoint analysis. They find that rating-based conjoint analysis performs better than open-ended CV, yielding non-significant differences in different purchase situations (real vs. hypothetical contexts). Similarly, studies also confirm that incentive-aligned direct approaches (BDM procedure) and incentive-aligned indirect approaches (ICBC analysis) generate different results (Miller et al., 2011).
2.9 Evaluation of methods
Table 2.1: Evaluation of competing WTP measurement methods Contingent valuation Conjoint analysis Market data analysis Experimental auctions ICBC analysis 1. Validity of estimations -- +/- ++ +/- +
2. Flexibility to include new
combinations +/- ++ - + ++
3. Observed choice behavior -- ++ ++ - ++
4. Individual-level estimations ++ + - +/- + 5. Cost effective/time efficient ++ + +/- -- - + (++) = (strong) advantage - (--) = (strong) disadvantage
+/- = no clear advantage or disadvantage
Table 2.1 shows that market data analysis has a strong advantage over the other methods with respect to the first criterion: validity of estimations. Since the WTP estimates from this method are derived from real purchase data, they are very reliable and show high external validity. Depending on whether the data are readily available and the required size of the data set, this method can be cost effective and time efficient as well. In practice, however, the use of market data may not be feasible for WTP measurement. Specifically for new or hypothetical products that are not yet available on the market, there is no market data available. In addition, market prices frequently do not contain sufficient price variations to estimate demand at different price levels.
Thus, as mentioned before, hypothetical methods (CV and conjoint analysis) may be managers’ preferred measurement approach under practical circumstances. This can be the case when managers face monetary constraints or when the results are required to be quickly available. Conjoint analysis and, to a lesser extent, CV are also very flexible when product features need to be varied or when a larger set of possible prices need to be tested. The only disadvantage of these methods is that real purchase behavior remains unobserved. Hence, WTP estimates obtained from these methods typically show a lower degree of validity as compared to, for example, WTP estimates from market data analysis.
Fortunately, Miller et al. (2011) affirm the usefulness of hypothetical WTP methods. Their analysis shows that even if a hypothetical approach generates hypothetical bias, the approach may still be useful in guiding managers to good pricing decisions. In particular, they show that a hypothetical approach has the potential to forecast an accurate demand curve. Furthermore, both hypothetical approaches they tested (i.e. open-ended CV and CBC analysis) lead to pricing decisions that are not significantly different from the benchmark of real purchase data. Researchers focusing on the hypothetical bias of these approaches may thus have underestimated their value in guiding managerial decision making.
3. RESEARCH METHODOLOGY 3.1 Data collection
An online survey was used to collect data for the study. To recruit participants, consumers were contacted through social networking sites such as Twitter and Facebook, invitation e-mails, and face-to-face conversations. They were encouraged to participate by offering a chance to win a photography workshop of a renowned camera manufacturer in a raffle. The participants were further informed that their chance to win in the raffle was independent of their survey responses.
In total, 242 Dutch consumers participated in the survey. They included people from the student body of the University of Groningen, high school students, and other consumers that were willing to participate in the study. Despite some concerns about their liquidity, student samples are rather common in studies of WTP as such samples are convenient and readily accessible (see Kalish & Nelson, 1991; Botelho & Pinto, 2002; Backhaus et al., 2005; Voelckner, 2006; Miller et al., 2011).
A total of 34 surveys were not fully completed and thus excluded from the analyses. Three more surveys were excluded because they contained seemingly unrealistic results (e.g. always choosing the first answer option). Ultimately, this resulted in a final sample of 205 respondents.
3.2 Experimental design
For the conjoint study, I chose a CBC design because choice tasks are more immediate and concrete than rating or ranking tasks (Huber, 1997). Moreover, CBC designs are frequently applied in pricing studies (Wittink, Vriens & Burhenne, 1994).
3.2.1 Design of the attributes and levels
Defining proper conjoint attributes and levels is arguably the most fundamental and critical aspect of designing a good CBC study. Therefore, I searched for relevant product attributes and attribute levels among potential digital camera buyers, reviewed current offers and interviewed potential buyers to determine price levels. I also conducted a focus group session with three product experts from a renowned camera manufacturer. These efforts resulted in the following list of product attributes and attribute levels (Table 3.1).
Table 3.1: Product attributes and attribute levels
Attribute Level of attribute
Resolution 10 megapixels
12 megapixels 14 megapixels 16 megapixels
Zoom factor 5x zoom
10x zoom 18x zoom 30x zoom
Product features & aesthetics Low-light CMOS sensor
In-camera filter effects Ultra-slim, compact body Available in bright colors
Gadgets Full-HD movie recording
A more extensive description of the product attributes and their levels can be found in Appendix I. During every step of the online survey, respondents could view this explanation by clicking on a question mark next to the respective attribute or level.
A potential problem in conjoint studies is the “number of attribute levels effect” (Currim, Weinberg & Wittink, 1981; Steenkamp & Wittink, 1994). This effect recognizes the fact that the importance of an attribute is influenced by the number of levels: the addition of intermediate levels (holding the extreme levels constant) tends to increase the importance of an attribute in relation to other attributes. As this study approximately balances the number of levels over the attributes (4x4x4x4x3), the number of levels effect is not expected to influence the attribute importances.
3.2.2 Design of the CBC analysis task
The CBC design contained 8 choice sets, from which I estimated the parameters, and 2 holdout sets to assess the predictive validity of the design. Each choice set contained three product alternatives, which were described by the five attributes obtained from pretesting, and a no-purchase option. To implement this design, I generated an efficient randomized design based on complete enumeration strategy that aimed to produce a nearly orthogonal design, using Sawtooth Software CBC/Web 6.4.2. As a result, the product alternatives in each choice set were as different as possible to ensure minimal attribute-level overlap.
Despite the robust statistical qualities of orthogonal designs, some researchers have been bothered that product concepts with the best features sometimes are shown at the lowest prices and products with the worst features at the highest prices. These combinations seem illogical and often lead to obvious choices in the questionnaire. Such choices are less informative and lead to a less realistic experience for the respondent. In order to solve this problem, I made use of conditional pricing. With conditional pricing, incremental amounts are added for premium features, so enhanced products are generally shown at higher prices. Price is still treated as a separate attribute with a few levels, but these levels of price are described with different absolute euro amounts, depending on product characteristics (Orme, 2007).
current offers and focus group session, this price was set at €79. In the same way, I chose price premiums associated with the two product attributes and the price level (see table 3.2). These price premiums were not explicitly shown to respondents next to each attribute level, but were used just to determine the overall average price. Only a single total price was shown within the product concept.
Table 3.2: Conditional price premiums
Attribute level Price premium
10 megapixels + €0 12 megapixels + €20 14 megapixels + €40 16 megapixels + €60 5x zoom + €0 10x zoom + €50 18x zoom + €100 30x zoom + €150 Low price (- 20%)
Medium price (Average price)
High price (+ 20%)
Subsequently, I constructed a look-up table to determine the prices that should be displayed in the survey for each possible product combination at each of the three price levels (low, medium, high price). The first five rows of this table are presented in table 3.3:
Table 3.3: Look-up table for displayed prices
Resolution Zoom factor Price Price to display
For example, row four of the table specifies what price should be displayed when a product with 10 megapixels and 10x zoom appears at the low price. In that case, the price to be shown is €99. This is determined by taking the base price (€79) plus the price increments associated with the two conditional attributes, and then reduced by 20%. Consistent with digital camera prices in the real market, the outcome of this calculation was rounded to the nearest “9”, resulting in the displayed price of €99.
The benefits of conditional pricing are that more reasonable prices are shown to respondents and that the CBC design stays orthogonal and unencumbered by prohibitions. Moreover, the effect of price on consumers’ choices can be estimated using the main effects of (in this case) the three levels of price. However, there are also some challenges when working with conditional pricing. The most important one is that the main effect utilities of the other attributes involved in the conditional relationship no longer can be interpret independent of price. For example, I cannot interpret the levels of resolution as the preference for each of its levels holding everything else constant. The utility of each level of resolution is confounded with the incremental price attached to that level. The levels must therefore be interpreted as the preference for levels of resolution given the average prices shown for those levels. So, it is very possible to achieve a higher average utility for 10 megapixels + €0 than for 14 megapixels + €40, if respondents on average did not feel that it was worth the extra €40 to receive the higher resolution.
3.3 WTP estimation procedure
For open-ended CV, I can obtain respondents’ WTP directly from the survey data. In contrast, determining WTP within CBC analysis requires a special estimation procedure because the CBC method only provides information on preferences and utilities.
Choice experiments can be analyzed with the multinomial logit (MNL) model (Green & Krieger, 1988; Chakraborty, Ball, Gaeth & Jun, 2002), that provides the probability of choosing product i from choice set a given its attribute vector Xia, with number of elements
equal to the number of attribute-level dummies, as:
1234 = 1 =7 exp 5exp 534 6
346
The MNL model, however, is estimated at the aggregate level by maximizing the likelihood function. Hence, individual level estimates cannot be obtained. This is an important disadvantage of this procedure, because models estimated at the aggregate level ignore heterogeneity in consumer preferences.
To overcome the limitations of aggregate analyses, DeSarbo, Ramaswamy & Cohen (1995) propose the use of a latent class version of CBC. The authors generalize the Kamakura & Russell (1989) scanner data response methodology to a latent class CBC model considering within subject replications over choice sets. In this model, the respondent’s (segment specific) choice probability of product i in choice set a for segment s, Ps,i,a, is given by:
19,3,4 = exp exp9,3
9,:; + 73∈>?exp9,3
=exp6 exp 7 7@∈K A∈IJ69,@,A∙ C3,@,A− 69,DE3FG∙ H3
9,:; + 73∈>?exp 7 7@∈K A∈IJ69,@,A∙ C3,@,A− 69,DE3FG∙ H3
, ∀M ∈ N, O ∈ P, Q ∈ R, 3.2 where
• Us,i = utility of product i for segment s
• Us,/P = utility of no-purchase option for segment s
• βs,j,m = parameter of the level m of attribute j for segment s
• xi,j,m = binary variable indicating whether product i features level m of attribute j
• βs,price = price parameter for segment s
• pi = price of product i
• βs,/P = parameter for the no-purchase option for segment s
• a = index set of choice sets • s = index set of market segments • i = index set of products
• Ia= index set of products in choice set a (not including the no-purchase option)
Although latent class models are estimated at the segment level, they can also be used to derive individual parameters and WTP estimates (Andrews, Ainslie & Currim, 2002; Natter & Feurstein, 2002). In a latent class model that maximizes the likelihood function, I can use information about the posterior segment membership probability to derive individualized parameters based on the estimated segment-specific parameters (Wedel & Kamakura, 2000). By deriving individual parameters for the product attributes (βh,j,m), price (βh,price), and the
no-purchase option (βh,/P), I can then gain information about a consumer’s WTP, or the price at
which consumer h is indifferent between purchasing and not purchasing product i (Moorthy, Ratchford & Talukdar, 1997; Gensler, Hinz, Skiera & Theysohn, 2012):
S S 6T,@,A∙ 53,@,A− 6T,DE3FG ∙ UV1T,3 = 6T,:; ∀ℎ ∈ X, O ∈ P.
A∈IJ
@∈K
3.3
Rearranging equation (3.3) gives:
UV1T,3 =6 1
T,DE3FG ∙ YS S 6A∈I T,@,A∙ C3,@,A− 6T,:;
J
@∈K
Z 3.4
3.4 Validity measures
In order to assess the validity of the two WTP measurement methods, I will determine the face validity of the WTP estimates by correlating the measured WTP for each method with respondents’ interest in buying a digital camera at the time of the survey (similar to Voelckner, 2006).
hit rate across the sample is the percent of correctly predicted holdout responses using the model (Orme, 2006).
4. RESULTS 4.1 Data description
The final data set contains 205 Dutch consumers, including 117 males and 88 females. The majority of these consumers is younger than 25 years, highly educated (student sample) and owns a digital compact camera. Furthermore, most consumers buy their digital camera at large retail chains specialized in consumer electronics (e.g. Saturn or Media Markt), and base their decision on the recommendations made by salesmen. Another large proportion of consumers searches on the internet for product information, professional reviews, and reviews by other consumers.
Overall, the sample shows a low-to-moderate interest in photography (neutral = 3, mean = 2,86) and most consumers are able to name at least 3 camera brands. More detailed descriptive statistics are listed in Table 4.1 below.
Table 4.1: Descriptive statistics
Variable Category Proportion
Gender Male Female 0,571 0,429 Age < 25 years 25 – 34 years 35 – 44 years 45 – 59 years > 60 years 0,595 0,278 0,059 0,059 0,010 Education Low Medium High 0,200 0,395 0,405
Camera ownership /one
Purchase place /one Online
Photo specialist
Large chain electronics Second-hand
Gift from family/friends
0,239 0,146 0,176 0,366 0,020 0,053
Purchase driver /one
Seen on TV
Online product information Consumer reviews
Professional reviews Advice from family/friends Advice from salesmen
0,239 0,015 0,117 0,151 0,146 0,102 0,230 Interest in photography To a very small extent
To a small extent /eutral
To a great extent To a very great extent
0,102 0,278 0,371 0,151 0,098 Knowledge of brands < 3 brands
3 – 5 brands > 5 brands 0,263 0,483 0,254 4.2 Data quality
The quality of the data set also relies on consumers’ response behavior. Response behavior is affected by the design of the study, especially by the price levels provided. If the prices do not sufficiently overlap with consumers’ WTP, extreme response behavior occurs: consumers always or never choose the no-purchase option. Gensler et al. (2012) show that both types of extreme response behavior result in invalid WTP estimates. Accordingly, a pretest and focus-group session were carried out to determine appropriate price levels, and the data set was scanned for extreme response behavior. In the final data set of 205 respondents, 23,41% of the respondents never selected the no-purchase option, and 10,24% always did. This share of respondents who exhibited extreme response behavior is comparable to percentages from other studies (Gilbride, Guiltinan & Urbany, 2008; Natter, Mild, Wagner & Taudes, 2008).
4.3 Face validity
I determined the face validity of the WTP estimates by correlating the measured WTP for each method with respondents’ interest in buying a digital camera. As expected, I found a highly significant (p < 0,001) relationship between the respondents’ WTPs and their interest in buying a digital camera (see Table 4.2 below). All correlation coefficients possess the expected sign. The results therefore indicate face validity of the measured WTPs.
Table 4.2: Correlation table (Pearson correlation)
1. 2. 3.
1. Purchase intention 1 0,710** 0,874**
2. WTP CV 0,710** 1 0,692**
3. WTP CBC analysis 0,874** 0,692** 1
**. Correlation is significant at the 0.01 level (2-tailed)
4.4 Open-ended CV
In the CV task, respondents were directly asked to state their WTP for 4 different digital cameras with the following characteristics:
1) 16 megapixels, 5x zoom, ultra-compact body, touch screen;
2) 16 megapixels, 10x zoom, available in bright colors, 3D photography; 3) 16 megapixels, 18x zoom, in-camera filter effects, GPS functionality;
These characteristics are applicable to the most popular models of camera manufacturers in the market and are comparable to the attributes used in the CBC study. This facilitates a straightforward comparison of WTP estimates between methods, and with real market prices. Respondents’ WTP estimates were obtained directly from the survey data. The estimated mean WTPs ranged from €147,21 for the first camera to €271,84 for the last camera. Table 4.3 summarizes the results of the analysis.
Table 4.3: Results open-ended CV
N Mean WTP camera 1 205 147,21 WTP camera 2 205 195,82 WTP camera 3 205 258,93 WTP camera 4 205 271,84 4.5 CBC analysis
The CBC analysis required a special estimation procedure. First, I estimated a latent class model in LatentGold Choice 4.5, which is a specialized program designed strictly for estimating CBC models. The optimal number of classes was determined by looking at the information criteria of the different solutions that were created (see Table 4.4). The maximum number of classes considered was five, because a higher number would lead to relatively similar classes or too small classes to obtain valid results. Since the Bayesian information criteria (BIC) statistic has been found to be most effective in selecting the appropriate number of classes (Andrews et al., 2002; Konus, Verhoef & Neslin, 2008), this statistic was used to select the best model (the lower the number, the better the model).
Table 4.4: Information criteria
1 Class 2 Class 3 Class 4 Class 5 Class
As shown by the rectangle in Table 4.4, the BIC statistic suggests a three-class solution. The degrees of freedom (df) represent the number of independent pieces of information that go into the estimate of the parameters. Too few df makes the model very dependent on the observations, so the more df, the more robust the findings are. The df of 161 is considered to be sufficient (Hair et al., 2006), so the results of the three-class solution can be used.
Subsequently, the three-class solution was used to derive individual WTP estimates on the basis of the posterior probabilities of class membership. I used each respondents’ class membership probability to derive individual estimates, which then enabled me to calculate each respondent’s WTP. The estimated mean WTPs for the studied cameras are presented in Table 4.5.
Table 4.5: Results CBC analysis
N Mean
WTP camera 1 205 110,74
WTP camera 2 205 186,19
WTP camera 3 205 224,11
WTP camera 4 205 389,45
I evaluated the predictive validity of the CBC analysis by calculating the holdout hit rate. Using the part-worth utilities from the first eight choice tasks (that were used in the estimation of the part-worth utilities), I predicted the responses to the two holdout tasks and compared them with respondents’ actual choices. Table 4.6 shows the hit rates for each holdout task. The holdout hit rate is lowest for the first holdout task, which is consistent with the notion of Johnson & Orme (2010) that the first tasks contain more randomness. Nevertheless, the values are high, and an average hit rate of 55,85% is good and much higher than the 25% chance criterion that would be obtained without a model.
Table 4.6: Holdout hit rates
Holdout task 1 Holdout task 2 Total
Correct 108 121 229
Incorrect 97 84 181
4.6 Comparison with current market prices
I previously commented on the problem of significant differences between WTP estimates derived from real purchase situations and WTP estimates from CV and conjoint analysis. It has been proven through numerous empirical studies that the values elicited in a hypothetical context are higher than those elicited in a real context. This leads to an overestimation of the real WTP values. In order to demonstrate that, at least in this study, CBC analysis measures WTP more ‘realistically’, I examined current market prices of digital cameras comparable to those that were studied. This enabled me to determine a realistic price range for the analyzed cameras (mean plus/minus standard deviation). The prices can be interpreted as the lower boundary of WTP since the actual WTP may lie above them. The means and price ranges are displayed in Table 4.7, as well as the mean WTPs for open-ended CV and CBC analysis.
Table 4.7: Price ranges of current market prices and WTP means Real prices - mean
(Range) WTP CV - mean WTP CBC - mean Camera 1 116,70 (101,92 – 131,48) 147,21 110,74 Camera 2 151,89 (129,48 – 174,30) 195,82 186,19 Camera 3 219,42 (192,61 – 246,23) 258,93 224,11 Camera 4 324,82 (295,59 – 354,05) 271,84 389,45
The relatively low WTP for camera 4 under open-ended CV may be caused by the low level of knowledge of some consumers about the true worth of the camera. According to Brown, Champ, Bishop & McCollum (1996), it is cognitively easier for a respondent to decide whether a specific price for a product is acceptable than to directly assign a price. Furthermore, because of their little interest in buying a camera at the time of the survey3, respondents may have been even more unwilling to buy a high-class camera for a high price and therefore state a relatively low WTP.
3
5. COCLUSIO & RECOMMEDATIOS
In this paper, I provided a thorough review of available methods for measuring consumers’ WTP, evaluated their strengths and weaknesses, and thereby provided marketing managers with a basis for selecting a suitable WTP method. As discussed, this selection depends on the managerial task underlying the estimation of WTP and is influenced by conceptual considerations (e.g. if individual estimates are required or not) and practical restrictions (e.g. time and budget available).
In certain situations, when there is no market (or not yet) for example, managers may have no alternative but to use a contingent valuation or conjoint-based method. As a follow-up to the work of Ding et al. (2005) and Voelckner (2006), it would be interesting to redesign these methods so that the gap between hypothetical and real settings is reduced (e.g. by integrating a purchase obligation). However, the case of high-priced products is difficult to deal with because of the high costs of complying with all purchase obligations and respondents’ monetary constraints that are likely to reduce WTP in real settings. Therefore, two research questions were formulated to guide the second step of this study:
Which method should be used to measure consumers’ WTP for high-priced products? Are there differences between the WTP measured by contingent valuation and conjoint
analysis?
All in all, these results seem to indicate that managers should favor CBC analysis over open-ended CV to measure WTP for optimal price fixing, although CV has some practical advantages like cost and/or time savings. This stands in contrast with some existing studies. For example, Miller et al. (2011) find that open-ended CV can outperform CBC analysis in estimating mean WTP and WTP distribution for an inexpensive, frequently purchased, nondurable product category. This leads to the assumption that the results depend on the specific characteristics of the study. Consequently, it should be asked which factors may have a significant influence on the suitability of the applied methods. This question seems to be the key issue for future research.
I already indicated that the analyzed product category may have an influence on the results. For example, as conjoint analysis supposes a more extensive decision-making process, a digital camera (generally with a price beyond €100) may be a more suitable product for applying conjoint analysis than a cleaning product (with a price less than €10), as in the study of Miller et al. (2011). However, these claims should be further validated in future research where the survey conditions can be systematically varied and controlled. This would require one study in which both a relatively low-priced, frequently purchased product and a relatively high-priced, infrequently purchased product are analyzed.
In a second phase, knowledge of external determinants could be deepened. The question of the product’s nature (products or services, physical or virtual, public or private, new or familiar) offers many avenues of investigation. Despite initial results concerning store brands obtained by Nies & Natter (2007), the question of brand influence on WTP also remains unexplored. WTP for discount brands or, at the opposite end of the spectrum, premium brands raises numerous unsolved questions. Research on these topics is necessary so that managers can choose the optimal research design in a particular context, after evaluating the costs and benefits of different methods for measuring WTP. An even better approach may be to use more than one of the methods mentioned in this paper to measure WTP, then use an objective statistical approach to combine results across methods by choosing appropriate weights for each method (e.g. Jedidi et al., 2003).
REFERECES
Allenby, G.M. & Rossi, P.E. (1999) “Marketing models of consumer heterogeneity”, Journal
of Econometrics, 89, 57-78
Anderson, J.C., Jain, D. & Chintagunta, P.K. (1993) “Understanding customer value in business markets: Methods of customer value assessment”, Journal of Business-to-Business
Marketing, 1(1), 3-30
Andrews, R., Ainslie, A. & Currim, I.S. (2002) “An empirical comparison of logit models with discrete versus continuous representations of heterogeneity”, Journal of Marketing
Research, 39(4), 479-487
Ariely, D., Loewenstein, G. & Prelec, D. (2003) “Coherent arbitrariness: stable demand curves without stable preferences”, The Quarterly Journal of Economics, 118(1), 73-105
Backhaus, K., Wilken, R., Voeth, M. & Sichtmann, C. (2005) “An empirical comparison of methods to measure willingness to pay by examining the hypothetical bias”, International
Journal of Market Research, 47(5), 543-562
Becker, G.M., DeGroot, M.H., & Marschak, J. (1964), “Measuring utility by a single-response sequential method”, Behavioral Science, 9(3), 226-232
Ben-Akiva, M., Bradley, M., Morikawa, T., Benjamin, J., Novak, T., Oppewal, H. & Rao, V. (1994) “Combining revealed and stated preference data”, Marketing Letters, 5(4), 335-350
Berman, B. & Evans, J.R. (2001) “Retail management: A strategic approach”, Upper Saddle River, NJ: Prentice Hall
Botelho, A. & Pinto, L.C. (2002) “Hypothetical, real, and predicted real willingness to pay in open-ended surveys: Experimental results”, Applied Economics Letters, 9, 993-996
Breidert, C., Hahsler, M. & Reutterer, T. (2006) “A review of methods for measuring willingness-to-pay”, Innovative Marketing, 2(4), 8-32
Brown, T.C., Champ, P.A., Bishop, R.C. & McCollum, D.W. (1996) “Which response format reveals the truth about donations to a public good”, Land Economics, 72(2), 152-166
Cameron, T.A. & James, M.D. (1987) “Estimating willingness to pay from survey data: An alternative to pretest-market evaluation procedure”, Journal of Marketing Research, 26, 389-395
Carlsson, F. & Martinsson, P. (2001) “Do hypothetical and actual marginal willingness to pay differ in choice experiments?”, Journal of Environmental Economics and Management, 41, 179-192
Chakraborty, G., Ball, D., Gaeth, G. & Jun, S. (2002) “The ability of ratings and choice conjoint to predict market shares: A Monte Carlo simulation”, Journal of Business Research, 55(3), 237-249
Currim, I.S., Weinberg, C.B. & Wittink, D.R. (1981) “The design of subscription programs for a performing arts series”, Journal of Consumer Research, 8, 67-75
DeSarbo, W.S., Ramaswamy, V. & Cohen, S. (1995) “Market segmentation with choice-based conjoint analysis”, Marketing Letters, 6(2), 137-147
Ding, M., Grewal, R. & Liechty, J. (2005) “Incentive-aligned conjoint analysis”, Journal of
Marketing Research, 42, 67-82
Elrod, T., Louviere, J.J. & Davey, K.S. (1992) “An empirical comparison of rating-based and choice-based conjoint models”, Journal of Marketing Research, 29, 368-377
Frykblom, P. (2000) “Willingness to pay and the choice of question format: Experimental results”, Applied Economics Letter, 7, 665-667
Garda, R.E. & Marn, M.C. (1993) “Price wars”, The McKinsey Quarterly, 3, 87-100
Gensler, S., Hinz, O., Skiera, B. & Theysohn, S. (2012) “Willingness-to-pay estimation with choice-based conjoint analysis: Addressing extreme response behavior with individually adapted designs”, European Journal of Operational Research, 219, 368-378
Gilbride, T., Guiltinan, J. & Urbany, J. (2008) “Framing effects in mixed price bundling”,
Marketing Letters, 19(2), 125-139
Green, P.E. & Krieger, A.M. (1988) “Choice rules and sensitivity analysis in conjoint simulators”, Journal of the Academy of Marketing Science, 16, 114-127
Green, P.E. & Srinivasan, V. (1990) “Conjoint analysis in marketing: new developments with implications for research and practice”, Journal of Marketing, October, 3-19
Gustafson, A., Herrman, A. & Huber, F. (2003) “Conjoint analysis as an instrument of market research practice”, In: Conjoint Measurement, Berlin: Springer, 5-46
Guyon, H. & Petiot, J. (2011) “Market share predictions: A new model with rating-based conjoint analysis”, International Journal of Market Research, 53(6), 831-857
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E. & Tatham, R.L. (2006) “Multivariate Data Analysis”, 6th edition, Pearson Education Inc., Upper Saddle River, New Jersey
Halvorsen, B. & Soelensminde, K. (1998) “Differences between willingness-to-pay estimates from open-ended and discrete-choice contingent valuation methods: the effects of heteroscedasticity”, Land Economics, 74, 262-282
Hoffman, E., Menkhaus, D., Chakravarti, D., Field, R. & Whipple, G. (1993) “Using laboratory experimental auctions in marketing research: A case study of new packaging for fresh beef”, Marketing Science, 12, 318-338
Huber, J. (1997) “What we have learned from 20 years of conjoint research: when to use self-explicated, graded pairs, full profiles, or choice experiments”, In: Sawtooth Software
Conference Proceedings, Sequim, WA: Sawtooth Software
Jedidi, K. & Jagpal, S. (2009) “Willingness to pay: Measurement and managerial implications”, In: Handbook of Pricing Research in Marketing, Cheltenham, UK: Edward Elgar Publishing, 37-60
Jedidi, K., Jagpal, S. & Manchanda, P. (2003) “Measuring heterogeneous reservation prices for product bundles”, Marketing Science, 22(2), 107-130
Jedidi, K., Mela, C.F. & Gupta, S. (1999) “Managing advertising and promotion for long-run profitability”, Marketing Science, 18(1), 1-22
Jedidi, K. & Zhang, Z.J. (2002) “Augmenting conjoint analysis to estimate consumer reservation price”, Management Science, 48(10), 1350-1368
Johnson, R. & Orme, B. (2010) “Including holdout choice tasks in conjoint studies”, Research
Paper Series, Sequim, Washington: Sawtooth Software
Kagel, J.H. (1995) “Auctions: A survey of experimental research”, In: The Handbook of
Experimental Economics, Princeton University Press, 501-585
Kalish, S. & Nelson, P. (1991) “A comparison of ranking, rating, and reservation price measurement in conjoint analysis”, Marketing Letters, 2(4), 327-335
Kohli, R. & Mahajan, V. (1991) “A reservation-price model for optimal pricing of multi-attribute products in conjoint analysis”, Journal of Marketing Research, 28, 347-354
Konus, U., Verhoef, P.C. & Neslin, S.A. (2008) “Multichannel shopper segments and their covariates”, Journal of Retailing, 84(4), 398-413
Le Gal-Ely, M. (2009) “Definition, measurement and determinants of the consumer’s willingness to pay: A critical synthesis and avenues for further research”, Recherche et
Applications en Marketing (English Edition), 24(2), 91-112
Levy, M., Grewal, D., Kopalle, P.K. & Hess, J.D. (2004) “Emerging trends in retail pricing practice: Implications for research”, Journal of Retailing, 80, 13-21
Louviere, J.J. & Woodworth, G. (1983) “Design and analysis of simulated consumer choice or allocation experiments: An approach based on aggregate data”, Journal of Marketing
Research, 20, 350-367
Lusk, J.L. & Schroeder, T.C. (2004) “Are choice experiments incentive compatible? A test with quality differentiated beef steaks”, American Journal of Agricultural Economics, 86(2), 467-482
Marn, M.V., Roegner, E.V. & Zawada, C.C. (2004) “The price advantage”, Hoboken, NJ: Wiley
McAfee, R.P. & McMillan, J. (1987) “Auctions and bidding”, Journal of Economic
Literature, 25, 699-738
Miller, K.M., Hofstetter, R., Krohmer, H. & Zhang, Z.J. (2011) “How should consumers’ willingness to pay be measured? An empirical comparison of state-of-the-art approaches,
Journal of Marketing Research, 48, 172-184
