THE REVERSED ‘IKEA EFFECT’:
THE FIRM’S REWARD FOR R&D EFFORT
Alwin Korthof
*,
BSc University of GroningenJanuary, 2015
Master thesis
Student number: 10127240
Executive Master Management Studies, Strategy track
Amsterdam Business School, University of Amsterdam
Supervisor: B. Kuijken, MSc
Version: Thesis_Alwin Korthof_Final_v1.1
Statement of Originality
This document is written by Student Alwin Korthof who declares to
take full responsibility for the contents of this document. I declare
that the text and the work presented in this document is original and
that no sources other than those mentioned in the text and its
references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the
supervision of completion of the work, not for the contents.
Signature ____Alwin
Korthof_______________________________________
Table of Contents
ABSTRACT
5
INTRODUCTION
6
2. LITERATURE REVIEW
8
2.1 TIME AND CONSUMER BEHAVIOUR 8
2.2 COSTS AND CONSUMER BEHAVIOUR 9
2.3 EFFORT & CONSUMER BEHAVIOUR 10
2.4 SELF-‐GENERATED EFFORT 11
2.4 CONCEPTUAL MODEL 12
3. RESEARCH METHODOLOGY
13
3.1 RESEARCH DESIGN 13
3.2 PROCEDURE & PARTICIPANTS 14
3.2.1 PRODUCTS 15
3.2.2 AUCTIONS 15
3.3 OPERATIONALIZATION OF THE HYPOTHESES 16
3.3.1 QUESTIONNAIRE 16
3.4 TESTING THE HYPOTHESES 17
4. RESULTS
19
4.1 DESCRIPTIVE STATISTICS 19
4.2 CUE RESULTS 19
4.3 TESTING FOR NORMALITY 21
4.4 TESTING HYPOTHESES 23
4.4.1 TEST OF H1: R&D TIME EFFECT 23
4.4.2 TEST OF H2: R&D COSTS EFFECT 23
4.4.3 TEST OF H3: R&D EFFORT EFFECT 24
4.4.4 TEST OF H4: CONSUMER EFFORT EFFECT 24
4.5 CONTROL VARIABLES 25
4.6 BOOTSTRAP METHOD 26
4.6.1 BOOTSTRAP RESULTS 27
4.6.2 TESTING OF HYPTHESES EXPERIMENT 1 27
4.6.3 TESTING OF HYPOTHESES EXPERIMENT 2 28
5. DISCUSSION
30
5.1 DISCUSSION 30 5.2 LIMITATIONS 32 5.3 IMPLICATIONS 33
6. CONCLUSION
35
REFERENCES
36
APPENDIX
I
A: THE AUCTIONED PRODUCTS AND TREATMENTS I
B: ONE-‐WAY ANOVA TEST IV
C: T-‐TEST VI
D: BOOTSTRAP TEST XIV
Abstract
The effort in R&D -‐ in order to develop a product -‐ has been linked in literature to investments and time (Vernon, 2005). However, how the firm’s effort in R&D impacts consumer behavior has been neglected. With this study the communication of exerted effort, influencing the willingness to pay of the consumer for a product, is examined. Two experiments, in which a new to the market product was auctioned, were conducted, by conducting a sealed-‐bid second-‐price (Vickrey) auction among N=942 participants. Three independent variables were measured: R&D time, R&D costs (together R&D effort) and consumer effort. The three variables were first measured in the treatments (online advertisements in the auction) and subsequently the three first variables were measured again in the questionnaire. The independent variable of this study was willingness to pay, measured in the currency euros. The two experiments were subsequently replicated by systematically and randomly resampling the available sample many times, in order to draw a conclusion without having to make making assumptions regarding the shape of the sampling distribution. The bootstrap test showed that the results of the experiments were not a result of random chance.
Concluding this research it was shown that when a producer exerts and
communicates effort in R&D, consumers reward the producer by increasing their willingness to pay. Even if the produced product is not improved and remains exactly the same,
significant increased willingness to pay was observed for R&D effort. The effect of
communicating R&D effort and the effect of exertion of R&D effort can be rather valuable for firms that are new to the market and / or want to launch a new product. Important decisions regarding time and costs spent in R&D can be made with this knowledge.
Introduction
Effort spent in the research and development (R&D) of a product is a very important factor for a company that intends to launch a new product. However, how this effort is perceived and valuated by the consumer is often not measured or is very hard to measure. A company makes important decisions regarding time and money spent in R&D before any consumer product valuation takes place. This can make knowledge regarding the
consumers’ perception of R&D effort, extremely valuable for firms.
Many studies have been conducted concerning how important factors in our life – time, energy and money – relate to our daily life and influence consumer behaviour. The perception of time and money has also been studied extensively. These studies show time and money are clearly impacting consumer behavior. There has been proof that time is considered a cost or effort for the customer. Although current available literature mentions the effort of producing a product (Jacoby, 1976; Heide & Olsen, 2011; Norton et al, 2011), relatively few research studies have been conducted on how effort exerted and
communicated by firms, is perceived by consumers.
One study examined consumer responses to extra effort exerted by firms in a general manner (Morales, 2004). Morales conducted three laboratory experiments, which showed that extra effort exerted by firms in making or displaying their products led to an increased willingness of consumers to pay for these products. According to Morales (2005), the few studies examining the influence of effort in a consumer context, have studied cases where the effort improves the perceived quality of the products and services or when actual quality is ambiguous, e.g. Kruger’s paper in 2003 (Morales, 2005). Instead, Morales looked at cases where the effort has no impact on product quality and is not personally directed towards an individual consumer.
The effort in R&D -‐ in order to develop a product -‐ has been linked in literature to investments and time (Vernon, 2005). However, how the firm’s effort in R&D impacts consumer behavior has been neglected. There has been little research conducted with respect to how the effort in R&D of the producer can affect the willingness to pay of the consumer. Many studies have been conducted with respect to willingness to pay (Noussair et al, 2004; Morales, 2005). However there are only a few previous studies that conducted experiments using second-‐price sealed-‐bid auctions (Vickrey, 1961) to determine the willingness to pay of the consumer. This leads to the following research question:
‘How does mentioning the amount of time and money spent on the R&D of a product influence consumer’s willingness to pay for that product?’
With this study the communication of exerted effort, influencing the willingness to pay of the consumer for a product, is examined. This research will add to the existing research whether the communication of exerted effort in R&D of the producer can affect the willingness to pay of the consumer. This in turn adds to the research pool on consumer behaviour. The communication of exerted R&D effort can be regarded as a signalling effect towards the consumer if the willingness to pay is influenced by R&D effort. With this knowledge firms can make decisions regarding their exertion or communication in R&D effort.
This paper is constructed as follows: the paper starts by reviewing previous relevant literature with respect to time and money and then effort in relationship to consumer behaviour. It proceeds by providing several hypotheses with respect to effort in relation to willingness to pay. The method for the research that is used will be discussed subsequently, followed by the results of the research. The paper continues with analysing and discussing the results in order to come to a conclusion regarding the stated hypotheses. Lastly, the implications and limitations of this study will be discussed.
2. Literature review
The relationship between time, costs and effort and consumer behaviour will be reviewed, which will lead to the formulation of relevant hypotheses and a conceptual model of this research.
2.1 Time and consumer behaviour
Jacoby et al. (1976) claims to be one of the first to write down systematic thoughts and perform empirically research on the relationship between time and consumer
behaviour. Jacoby defines consumer behaviour as the acquisition and use of goods and services by consumers with the use and expenditure of time integrally involved. Jacoby writes about how aspects of time act as variables affecting consumer decision-‐making. He provides examples from Wright (1974), Winter (1975) and Chestnut (1975) of time as a variable influencing the decision-‐making task and examples of time as measure of the decision-‐making task. Jacoby argues that the consumer owns five subjective states regarding consumer behaviour and time: 1. The perception of urgency: having short time to satisfy the need of a product. 2. Perceived newness: whether a product is new to the market. 3.
Anticipated frequency, duration and extension. The usage of the considered product will be taken into account together with an assumption of its lifespan and future value. 4. Individual differences. Consumers value time more or different than others. 5. Timing, strategy and scheduling. Consumer perceptions about when to buy the product, when to use the product, how long to use the product, et cetera. Jacoby ends his paper by discussing the trade-‐off relationships between time, money and effort. In elaboration of the fourth subjective state as described by Jacoby (1976) Graham (1981) reviewed three different perceptions of time and focused on behaviour of people who hold different perceptions. According to Graham, time can be consumed like a consumer good. This makes it possible to value time in terms of money. Graham argues that the perception of time seems so obvious that it is hard to understand that others have different perceptions of time. According to him, especially European-‐Americans share this perception. They believe time can be saved, spent and wasted like money. But time can also be bought with money. Graham continues that time is a commodity that will be allocated to maximise utility and that the allocation process follows the same process for decision-‐making as consumer products do. Graham came up with three perception models of time: the linear-‐separable model, the circular-‐ traditional model and the procedural-‐traditional model. The first model concerns the common perception about time as described above and is traditionally educated in European-‐American societies.
Basically time is money and vice versa. The second model covers the idea that actions are not regulated by time. Because of the emphasis on the present, there is little connection between time and money according to the circular-‐traditional model. The third model, the procedural-‐traditional model, states that time is irrelevant. Activities will take places according to procedures and what matters is whether the procedures are followed rather than performed on time. What Graham tries to convey with his paper is that there are different perceptions of time and that these perceptions are carried as part of a culture, and that these perceptions have in turn an impact on consumer behavior. His paper supports that understanding the perception of time of consumers, means understanding the behavior of consumers. Following this reasoning, the perception of time of consumers can influence consumer behaviour. Time spent is according to Graham’s linear-‐separable model equal to money spent and can therefore be valuated as money. How time will be valuated will be researched in this study. Both time spent by the consumer and time spent by the producer will be researched. I expect a positive relationship between the consumer’s perception of the time spent by the producer and the consumer’s behaviour. In order to test this, the first hypothesis is formulated as follows:
H1: Mentioning the time spent in R&D increases the willingness to pay of the consumer for a product.
2.2 Costs and consumer behaviour
Like Jacoby (1976) Okada and Hoch (2004) begin their study regarding time versus money by quoting Benjamin Franklin in his ‘Advice to a Young Tradesman’ (1748):
"Remember that time is money." Implying that time is just as valuable as money. They express time in a monetary term as opportunity cost, firstly introduced by von Wieser in 1914. Although Graham (1981) already argued that time does not necessarily relate to money, as discussed in the paragraph above, Okada & Hoch believe that time does relate to money. However, they suggest that consumers are not treating time and money in the same manner. Research by Neumann and Friedman (1980) and Hoskin (1983) implies that because of the under appreciation of the opportunity cost of time, time expressed in money is not always the same. Besides, the perceived valuation of time cannot be precisely expressed in money. In their paper, Okada & Hoch examine the ambiguity of the value of time. By conducting experiments they provide evidence that the value of time is ambiguous. Especially when expressed in terms of money. They explain the difference in time and money valuation in the estimation of their opportunity cost. Since time is not as liquid and
storable as money, estimations for opportunity costs are easier computed for money than for time. The results of the experiments show that people rather pay in time than in money. The experiments could not show whether money or time was spent more wisely. However, it did show that the valuation of time in terms of money depends on the circumstances when time is valuated. Why time is not like money is subject of Soman’s study of 2001 as well. Contrary to what was presumed in the previous paragraphs, Soman empirically proves that time is not like money. However, as for time, I expect a positive valuation for costs as exerted effort. How costs made and communicated by the producer and how consumers valuate these costs, will be researched. In order to test this and to test whether time or money is regarded more important to the consumer, the second hypothesis is consequently:
H2: Mentioning the fixed R&D costs increases the willingness to pay of the consumer for a product.
2.3 Effort & consumer behaviour
The more effort the producer spend developing on a product, the more convenient the product is perceived by the costumer, this was researched by Kruger et al. (2003) and Morales (2005). Kruger et al. argue that effort is used as a heuristic for quality. They
researched that more time and effort spent on producing a product led to higher ratings for that product in terms of quality, value, and liking of the product. They demonstrated, by conducting three experiments, that for three different objects the valuation for the objects increased together with the effort exerted in the objects. The valuation of the (improved) quality of the objects was a moderating factor in the perception of the increased valuation. All three experiments were set up in a similar manner: in the experiments the cue was each time a difference in time spent on the product (high or low).
Morales (2005) also researched the exerted firm effort in relation to the customer’s valuation of the product. However in her study the firm effort was researched independent of the quality of the product. Moreover her research is more generic applicable, since her research was conducted in a general manner and not conducted toward an individual customer like the experiments in Kruger’s et al (2003) paper. In contrast to Kruger’s paper, consumers did not personally benefit from the effort that was exerted by firms in Morales’ paper. For instance, if a firm spends time in marketing, this effort is directed in a general manner and available to a generic public. The exerted effort was not directed toward an individual person. Morales also conducted three experiments. In the first two experiments the effort cue was stated by a difference in displaying products organized and less
organized: experiment one and two displayed products in a different manner (high effort: neatly organized products, low effort: less organized products). In the third experiment, like Kruger’s experiment, a difference in time spent on the product was used as cue for the exerted effort. Kruger et al. conducted their experiments by deviating the exerted effort expressed in the amount of time invested in the product. Morales did this as well, but also conducted two experiments where the cue of time was not explicitly mentioned. Kruger et al. used a quality scale from 1 to 11 to determine the perceived effort; Morales used willingness to pay as a factor to determine the perceived effort. Neither Kruger et al. nor Morales included the cue of money in their research. I believe that this is a gap in the current literature with respect to firm effort in relationship to customer valuation. The cue money, as expression of effort, was therefore included in this study. The third hypothesis combines both the cue of time and money. This combination is also new in effort literature and is expected to have a stronger effect than H1 and H2:
H3a: Mentioning the R&D effort increases the willingness to pay of the consumer for a product
H3b: Mentioning the R&D effort increases the willingness to pay of the consumer for a product more than mentioning either R&D time or R&D costs
2.4 Self-‐generated effort
The three hypotheses above are stated to research not self-‐generated effort. They apply to other-‐generated effort: effort exerted by the producer or firm (Kruger et al., 2003). Next to this, self-‐generated effort will be researched.
Effort can be psychically and mentally expressed. Heide and Olsen (2010) expect that the use of time is a cost or effort for the consumer. For instance, the time used for preparing a meal is considered either a cost or effort for the consumer. Candel (2001) and Scholderer & Grunert (2005) define in their studies the use of time and the physical or mental effort exerted by the consumer in the production and the consumption of food as ‘convenience’. Heide and Olsen point out that studies have shown that consumers do not always prefer convenience solutions. Their study concerns the opposite of consumer convenience: co-‐production. Co-‐production is the active participation of a consumer in the production process. Heide and Olsen showed that by testing the relationship between convenience and co-‐production as well how the combined role of those constructs
influences global product evaluation. Their study proves that satisfaction with co-‐production has a strong and positive effect on the evaluation of the final outcome of the co-‐processed
product. They also showed that convenience has a positive influence on product evaluation. This is in line with earlier studies from Berry (2002) and Jacoby (1976) regarding that time is a cost perspective, as described previously. Additionally they showed that the less effort a consumer has to spend on a product, the more convenient the product is perceived by the consumer.
Norton, Mochon and Ariely (2011; 2012) continued studying self-‐generated effort and showed in their papers that people are willing to pay more for self-‐made products than for identical ones made by others. They call this finding the ‘IKEA effect’, after the build-‐it-‐ yourself furniture sold by IKEA. They state that consumers ‘assembling products fulfills a core psychological need desire to signal to themselves and others that they are competent and that the feelings of competence associated with self-‐made products lead to the increased valuation.’ This is supported by Heide and Olsen (2010) if we look at their finding regarding co-‐production: satisfaction with co-‐production has a strong and positive effect on the evaluation of the final outcome of the co-‐processed product. I reason that not only co-‐ production effort, but also self-‐generated effort after production has a positive influence on the consumer’s valuation of the product. This leads to the following hypothesis:
H4: Mentioning future consumer effort increases the willingness to pay by the consumer for a product
2.4 Conceptual Model
The discussed literature and consequent hypotheses regarding R&D time, costs, effort and consumer effort in relation to willingness to pay are depicted visually in the conceptual model below:
Figure 2.1 Conceptual Model
Fixed R&D costs (€)
Maximum willingness to
pay (€) Time spent in
R&D (years) R&D effort (€ & years) Consumer effort spent in time (minutes)
3. Research methodology
In order to test the previously formulated hypotheses and to answer the research question, an experimental research was conducted. In this chapter, the methodology of this research is discussed. Research design, data collection and subsequently, a plan of analysis are described and elaborated upon.
3.1 Research design
In this research study, an experiment was conducted instead of conducting research through questionnaires. The experiment that was set up showed true consumer behaviour and real product valuation in terms of willingness to pay. The consumers’ willingness to pay was examined by conducting an online second-‐price sealed-‐bid auction. The rationale for researching willingness to pay in relation to experimental auctions is discussed below.
Experimental auctions can avoid potential biases with respect to overstating
willingness to pay and are designed to show real differences in willingness to pay (Alfnes and Rickertsen, 2003). According to Alfnes & Rickertsen and Lusk (2012) several studies
presented that stated preference methods show overestimation of willingness to pay. Therefore it was chosen to conduct an experiment and not to use the questionnaire method. Ruckmick already provided criticism on the questionnaire research method in the beginning of the thirties and considered experimental research delivering higher value for science (Ruckmick, 1930). He calls the questionnaire method a prescientific procedure instead of a research method. He sees the method as an antecedent to and dependent on further experimental research. The main disadvantage of the questionnaire method
mentioned by Rucknick regarding this method is the uncontrolled and uncontrollable nature of the replies. These undesirable responses are for a great deal avoided in this experiment.
In contrast to the earlier mentioned stated preferences indication, which can be retrieved by answers out of questionnaires or created by hypothetical markets, bids in experimental auctions show revealed preferences (Lusk, 2010). This makes experimental auctions in contradiction to other research methods not hypothetical. We want to know what will happen in the real world instead of what would have happened in a hypothetical situation. By employing binding commitments instead of non-‐binding responses, people are forced to more carefully think about their answers when participating in a research (Lusk, 2010). According to Barrot et al. (2010) stated preference research, like survey research, is limited because of the lack of incentive to reveal truthful willingness to pay. They state that research has shown that the most sophisticated elicitation techniques still are subject to questionable responses (Barrot et al, 2010). Noussair et al. continue: demand revealing
auctions can, in contrast to field research, measure directly the limit price someone is willing to pay. The commitment of real money creates an incentive to truthful bid and reveals willingness to pay. Particularly for new products, where no market prices yet exist, accurate willingness to pay information is useful (Noussair et al, 2004).
Experiments with second-‐price sealed-‐bid auctions are not new: Vickrey introduced this particular type of auction in 1961. He showed in his paper of 1961 that second-‐price sealed-‐bid auctions led to bidding your reservation price. With bidding the reservation price Vickrey claims bidding the price based on your valuation of the product, e.g. your willingness to pay for the product. The auction procedure works as follows: bidders place bets with the understanding that the auction winner is the highest bidder but only pays the second highest bid (Vickrey, 1961). The bids are sealed. Since bidders have the understanding of the foregoing, the optimal strategy is to bid your willingness to pay. The price bid would be on the margin of indifference as to whether you win or not. Bidding less would decrease your chance to win and could not affect the price if you were the highest bidder; bidding more could lead to paying more than intended. This makes betting your true value for the product auctioned the optimal strategy (Vickrey, 1961).
3.2 Procedure & Participants
The experimental second-‐price sealed-‐bid auction market is discussed above. For this study an existing online second-‐price sealed-‐bid auction was used. The tool, named Veylinx, was set up by University of Amsterdam (UvA) PhD candidates Bram Kuijken and Anouar El Haji for experimental research purposes. By recruiting panel members for their online auctions, both university students created a population that could participate in auctions in order to perform academic research experiments. The population should ideally reflect the population of the Netherlands.
Initially members were recruited via word-‐of-‐mouth marketing but in a later stage members were also acquired via business channels. In addition, the Veylinx founders had their Bachelor and Master graduate students perform experiments for their own theses and asked – in exchange for letting them using their tool – the graduate students to recruit each 100 new members to sign up for the Veylinx tool. The population of Veylinx, containing only Dutch citizens, is comparable to the population of the Netherlands with regard to age. With respect to gender, the population used in the experiments can be regarded as
representative for the population of the Netherlands as well: according to the Dutch central bureau of statistics (CBS), in 2013 49,5% of the Dutch population was men and 50,5% was women. At the time the experiments were conducted Veylinx had a population of 48,6%
men and 51,4% women. In total 4567 Veylinx subscribers were approached to participate in the auctions, the historical average respond rate of the auctions is 40%.
Both experiments were treated differently because of the gender difference. Although the products for both experiments were similar, distinction between men and women was made and the results for the two experiments were not used transposable. The products auctioned and the rationale for choosing these products will be discussed in the next paragraph.
3.2.1 Products
A product, which was new to the market and with a retail price very few people were aware of, was chosen to auction. Only by auctioning a new product willingness to pay could be examined. If the researched audience knew the retail price, the collected answers could have been biased and mitigating actions had to be taken to remove those answers from the research. Perceived newness of a product is also one of Jacoby’s five subjective states regarding consumer behaviour and time (1976).
For both experiments two similar products were auctioned. The first experiment was set up for men and the second experiment for women. The products auctioned were two bikes: a new men and women’s bike from Amsterdam based start-‐up company Veloretti (www.veloretti.com). The bikes can be categorised between a smooth urban bicycle and a sporty race bicycle (see Appendix A). Theses bikes were chosen to auction since there was producer, developer and consumer effort involved in creating, producing and assembling the bikes. These features would allow to offer different real treatments regarding the to be examined R&D and consumer effort. New to the bike market was the fact that the consumer had to put effort in assembling the Veloretti bike instead of buying it in a bike shop (the bikes come in parts in a box and are sold online). This feature would make it even more interesting to see the consumer effort effect in relation to willingness to pay since urban city bikes are usually sold through retail shops.
3.2.2 Auctions
Both auctions were held at the same day and time: Tuesday July 3rd 2014. The
subscribers of Veylinx had from 8am until 10pm to participate in the auctions. All subscribers received an email with the request to participate in the auction. The male and female subscribers were randomly distributed over the respective auctions. After participating in both auctions the participants were asked to fill out an identical questionnaire.
3.3 Operationalization of the hypotheses
All data with respect to the production of the product, the R&D time spent, costs, et cetera was available to use, since the company’s founder and owner was involved in this study. The table below shows the different treatments of the experiments and their cues.
Table 3.1: Experiment treatments
Each experiment contained different cues. Experiment 1 was set up with the following cues: time spent in R&D and costs of R&D. Experiment 2 contained 3 cues: time spent in R&D, costs of R&D and consumer effort. The different cues (treatments) represent the indicated hypotheses. All hypotheses were simulated by the different treatments. Each treatment except the baseline treatment contained a cue in order to compare to the baseline treatment. The cues for experiment 1 (men) and experiment 2 (women) were the same. However experiment 2 had one more cue in addition. Each treatment was offered to a group of either men or women in the online auction: the first group of people (experiment 1, group 1) was offered the product without a cue (baseline). The second group of people (experiment 1, group 2) was offered the same product (and saw the same advertisement) only then including the time spent on R&D mentioned in the product advertisement. Another group of men (experiment 1, group 3) was offered at the same time the product with the costs of R&D mentioned in the advertisement. The same applied for experiment 2: the woman’s bicycle was offered in an advertisement including the R&D time and R&D costs mentioned. In addition to experiment 1, in experiment 2 one more cue (treatment) was added: this treatment mentioned the consumer effort required to assemble the bike in the advertisement. The treatments as offered in the auction can be found in appendix A.
3.3.1 Questionnaire
The questionnaire used in this research is depicted in Figure 3.1. All questions were measured on a 5-‐point Likert scale.
Question 1 was asked to see whether the product was indeed new to the market. Only by auctioning new products willingness to pay is worthwhile to measure. By knowing the product up front, biased bids can be expected with respect to the price since the price of the product can be known. The same applies for question 2. Question 2 was asked to get unbiased results for willingness to pay: if the price of the Veloretti bike was known
beforehand, this could have led to not bidding the real ‘willingness to pay’ price. Those bids then had to be omitted from the data. Question 3 to 5 were asked to get additional
confirmation for the treatment results and hypotheses. Question 3 and 5 were included in the treatments of both experiments. Question 4 was only included in experiment 2 (women).
As control variables the questions 1 and 2 ‘Are you familiar with the brand Veloretti?’ and ‘Did you know the price of the bike beforehand?’ were asked in the questionnaire together with the fourth variable ‘branding’, e.g. product awareness. These control variables help to avoid bias in relation to willingness to pay as discussed above.
1. Are you familiar with the brand Veloretti?
Yes No 2. Did you know the price of the bike beforehand?
Yes No 3. I find it important to know the development time of a product Likert scale: Fully agree 1 – 2 – 3 – 4 – 5 Fully disagree
4. I find it important to spend time in the product myself Likert scale: Fully agree 1 – 2 – 3 – 4 – 5 Fully disagree
5. I find it important to know the development costs of a product Likert scale: Fully agree 1 – 2 – 3 – 4 – 5 Fully disagree
Figure 3.1 Questionnaire
3.4 Testing the hypotheses
In order to answer the research question, the stated hypotheses were explored, tested and either supported or rejected. First the descriptive statistics of the treatments
were analysed. Secondly, by performing an ANOVA test, the means of the different treatments were tested to determine whether there was a significant difference in means between the base treatment and the individual hypothesised cues: R&D time (H1), R&D costs (H2), R&D effort (H3) and consumer effort (H4). The ANOVA test verified whether the mean of the treatment (the test statistic) was statistically significant from the mean of the base treatment. Establishing the p-‐value of the test statistic based on the sampling distribution was done for this purpose. Based on the outcome, the hypotheses were to be accepted or rejected. However, the test relied on the following assumptions: the data set was assumed to be distributed normally, variances were assumed to be equal and the responses for a given group were assumed to be independent and identically distributed normal random variables (Pallant, 2010). In order to comply with these assumptions of the ANOVA test, the test data can be transformed, in order to end up with more accurate results. There are obvious reasons to decide not to transform the source data, such as problems with the interpretation of the results and not finding the correct transformation of the data. In this research the data was therefore chosen not to be transformed, but to perform another (non-‐parametric) test instead. Non-‐parametric tests replace the original values by values based on their order of rank in the data sample and subsequently perform a ‘classical’ statistical test with these new values (Mann Whitney U test, Wilcoxon test,
Kruskal-‐Wallis ANOVA, Friedman ANOVA, Spearman rank correlation) (Pallant, 2010). There are also several, powerful computer intensive, non-‐parametric methods to determine the confidence levels on certain statistics and p-‐values of certain tests, such as the resampling method of the non-‐parametric bootstrap test. By systematically and randomly resampling the single available sample many times, it is possible to approximate the shape of the sampling distribution (and therefore calculate the p-‐value of the test statistic) with the bootstrap test. This method of analysis was also applied to the data. The advantage of this method is that the test is able to exclude assumptions regarding the distribution of the data set, which neglects potential need for data transformation (Efron, 1979; Davison et al., 1986).
4. Results
The result section starts with describing the descriptive statistics of the experiments and subsequently the cue results will be presented and discussed. Hereafter the distribution of the data set will be explored in order to see if the data set is compliant with the test assumptions. Next the hypotheses are tested and the control variables are looked at.
4.1 Descriptive statistics
The total amount of potential participants of the two experiments was N=4567, all the subscribers of the Veylinx website. The total subscribers were divided into two
experiments: experiment 1 for men (N = 2238) and experiment 2 for women (N = 2329). The historical response rate, all the subscribers who actually open their email, see the advertisement and place a bid was at the time of the auction was approximately 40%. In experiment 1, N = 536 respondents participated and in experiment 2, N = 406 respondents participated. This gives a response rate for men of 24% and for women of 17%. The total response rate was 21% (N = 942 out of N = 4567). If we look at the total respondents for men and women together, 56.90% men responded and 43.10% women. This deviates from the approximately 50-‐50 distribution of the Dutch population. With respect to age, for men as well for women the average age was 41, which is one year younger than the average age in the Netherlands and therefore representative (Central Bureau for Statistics, 2013).
Table 4.1: Population age of the Netherlands (source CBS, 2013)
4.2 Cue results
In table (4.2) below the results for experiment 1 and 2 are depicted. At a first glance the difference in means are notable. For men, treatment 2, including the cue for R&D time, led to a higher mean than the baseline treatment 1 (M = 62.86, SD = 77 versus M = 59.49, SD = 71). Treatment 3, the cue for R&D costs, led even to a higher mean than treatment 2 (M = 74.97, SD = 75 versus M = 59.49, SD = 71). Without exploring the data further into more depth, an increase of the mean of 26% is presented. The respondents of experiment 1 valued R&D effort (both time and costs) in a positive sense (M = 68.92 versus M = 59.49).
Descriptive statistics Men Women Women Men &
Mean 6558.39 5082.22 5922.16 Median 5000 3600 4500 Standard Deviation 7650.62 5894.28 6983.26 Sample Variance 58531932 34742566.49 48765909.04 Range 45000 30000 45000 Minimum 0 0 0 Maximum 45000 30000 45000 Sum 3515295 2063381 5578676 Count 536 406 942 Largest(1) 45000 30000 45000 Smallest(1) 0 0 0 Confidence Level(95,0%) 649.15 575.06 446.52
Table 4.2: descriptive statistics source data in euro cents
The same valuation of R&D effort applied to the respondents of experiment 2 can be found in table 4.3. Looking at the means, treatment 3 and 4 stand out compared to the baseline treatment: R&D time, (M = 59.94, SD = 71 versus M = 36.52, SD = 44) and R&D costs, (M = 63.05, SD = 56 versus M = 36.52, SD = 44). Also the treatment including
consumer effort has a higher mean than the baseline treatment (M = 44.32, SD = 46 versus M = 36.52, SD = 44). The bicycle advertisement including the cue R&D time was valued 63% higher than the advertisement without mentioning the R&D time. The cue R&D costs even led to an increase of willingness to pay of 72%. These initial results are great indicators for accepting the hypotheses of this research and confirming the expected outcome. Yet, whether these results are indeed significant will be explored in the subsequent sections by performing ANOVA tests and using the bootstrap method.
Table 4.3: Descriptive statistics of the treatments experiment 1 & 2
By excluding the 0 bids from the data, a first attempt was made to come closer to a normal distribution of the data. Out of a total of 942 bids, 250 zero bids were excluded. However, the data including 0 bids will be used later on in the bootstrap test in order to be completely statistically correct. Obviously, the mean increased for both experiments excluding the 0 values, for men (M = 88.55, SD = 77 versus M = 65.58, SD = 77) and women (M = 69.95, SD = 59 versus M = 50.82, SD = 59) (see table 4.2 and 4.4).
Table 4.4: Descriptive statistics of the two experiments (excl. 0 bids)
4.3 Testing for normality
In order to see whether the data set was distributed normally, a first attempt was already made by excluding the 0 values from the data set. In table 4.5 the cue results excluding the 0-‐values are depicted for experiment 1 and 2. Logically, the difference in means for the treatments including cues, has increased compared to the baseline treatments. Especially the cue R&D costs led to higher valuation of the product in both experiments looking at the means: men (M = 100.97, SD = 80 versus M = 80.78, SD = 77) women (M = 88.47, SD = 71 versus M = 53.75, SD =37), see also depicted in graphs 4.1 and 4.2.
However, excluding the 0-‐values did not lead to a (log) normal distribution of the data set. The Jarque-‐Bera test was rejected for both experiments and all treatments (p = 0.01).
Graph 4.1: Mean willingness to pay experiment 1
Graph 4.2: Mean willingness to pay experiment 2
€ 0 € 20 € 40 € 60 € 80 € 100 € 120 Treatments
Experiment 1: Men (µ)
Base (control) R&D rme R&D costs € 0 € 10 € 20 € 30 € 40 € 50 € 60 € 70 € 80 € 90 € 100 TreatmentsExperiment 2: Women (µ)
Base (control) R&D rme R&D costs Consumer effort4.4 Testing hypotheses
In order to test the hypotheses, it was tested whether the means of the cues in the treatments for the two experiments were significantly different from the baseline cue. To determine the significance of the mean differences, the one-‐way ANOVA test was
performed. The means of the different cues were compared to the mean of the baseline cue. By setting up a 0-‐hypothesis stating that the means of the two examined groups (treatments / cues) are equal, using a significance limit of 5%, the one-‐way ANOVA test will either accept or reject the hypothesis, implying that the means are equal or not equal. If the 0-‐hyphothesis is accepted, there is evidence that the means differ significantly at the 5% significance level.
4.4.1 Test of H1: R&D time effect
Hypothesis one predicts that mentioning the time spend in R&D has a positive influence on willingness to pay.
Men There was no significant evidence found that mentioning R&D time increased the willingness to pay of the consumer (M = 84.65, SD = 72 versus M = 80.78, SD = 77, p > 0.05). Hypothesis 1 is therefore rejected for experiment 1.
Women There was significant evidence found that mentioning R&D time increased the willingness to pay of the consumer (M = 77.62, SD = 62 versus M = 53.75, SD = 37, p < 0.05). Hypothesis 1 is therefore accepted for experiment 2.
Table 4.6: Hypothesis 1 -‐ test results
4.4.2 Test of H2: R&D costs effect
Hypothesis two predicts that mentioning the fixed R&D costs have a positive influence on willingness to pay of the consumer.
Men There was significant evidence found that mentioning R&D costs increased the
willingness to pay of the consumer (M = 100.97, SD = 80 versus M = 80.78, SD = 77, p < 0.05). Hypothesis 2 is therefore accepted for experiment 1.