Which App Would You Buy?
Analysis of the asymmetric dominance effect in mobile application market
Andis Arnicans
Student no. 10004450
BSc Economics and Business
University of Amsterdam
June 2, 2014
Supervisor: Jindi Zheng
Introduction
We live in a time where innovation is happening at a pace never seen before, allowing many new markets to emerge. Examples of the new markets can be seen both in financial sector (e.g. Bitcoin markets) and in nonfinancial sectors. Due to these fast changes it is very important to investigate if and how the old theories can be applied to these new markets. One of the most important fields of market research is the study consumer choice. Understanding how people make their choices of buying one product instead of another allows learning more about how markets function.
A particular market that is growing at a fast pace is the mobile application market. This is caused by newer technology becoming more accessible to greater amount of people around the globe. Smartphones are now a very big part of people’s lives - many have their smartphones in an arm’s reach 24 hours a day. A smartphone is a phone that allows performing many operations of computer and installing new
software (applications) unlike a regular headset.
One of the most important parts of consumer choice models is to learn how an introduction of a new product to the market affects the choice between the products. While the first models argued that the introduction of a new product would steal away market shares of the existing products, newer researches show that, by carefully designing the new product, the opposite effect can be achieved. To achieve this, marketers, who have to introduce a new product, often take advantage consumers not being able to make a rational choice.
Consequently, this paper will focus on discovering weather these strategies used by marketers in other markets will work in the mobile application market. This will be attained by carrying out an experiment. In particular, the participants of the experiment will have to make theoretical choices between different mobile applications. As a result, this experiment will show if the consumer behavior in the mobile application market can be manipulated by the introduction of new products.
The following chapters will introduce the reader to the previous research done in this field, give an overview of the experiment that was conducted and, finally, discuss the results and their importance.
Motivation
The reason for this study is the lack of research devoted specifically to consumer behavior in mobile application markets such as Apple App store And Google Play. The importance of these markets is pointed out by the estimations done by Gartner. Beginning of 2014 Gartner released a report where they estimate that the mobile application market will grow to $77 billion by 2017. (Gartner, 2014) This study
takes the asymmetric dominance effect and applies it to the mobile application market to see if it exists in this environment as well.
This research adds to the research done on mobile application markets by showing weather in some ways, consumers behave similarly to how they behave in offline markets and other online markets. The
contribution will be both on the business side and the consumer side. Firstly, this paper can help mobile application developers to make better decisions when deciding on the pricing of their new products, so they could increase their revenue. On the other hand, consumers who are shopping in mobile application markets will be better at recognizing the ADE and make better decisions for themselves.
Literature review and previous research Asymmetric dominance effect
Discovering how an introduction of a new product affects the market and its existing products had always been a concern of marketing specialists, therefore this subject has been studied very extensively. There are multiple theories of choice and one of the earlier perspectives suggested that, if a new product is introduced, it would take away market shares from existing products proportionately to what they had before the new product entered (Luce, 1959). Later this assumption of proportionality was challenged by McFadden (1974) and Debreu (1960) as, through real life observation, it was relatively easy for them to show cases where this was not true.
In later years the similarity hypothesis was introduced by Tversky (1972). The hypothesis reflected the popular managerial belief that when introducing a new product, it should be as dissimilar to the firm’s existing products as possible. Tversky argued that the new product takes away a bigger market share from the products that it is the most similar to, so making it very different would minimalize the
cannibalization effect and the company would have a larger total market share. The cannibalization effect is when the company introduces a new product and it moves customers away from an existing product from the company’s product range. (Copulsky, 1976)
Furthermore, the choice models that included this similarity hypothesis (Tversky, 1972), (Hausman & Wise, 1978), (Batsell, 1980), (McFadden, 1980) as well the model by Luce (1959), that didn’t include it, shared the belief that an introduction of a new product cannot increase the probability a consumer buying an existing product. This is called regularity and is a basic assumption of all probabilistic choice models. However, seeing how these choice models don’t hold in many empirical studies, Huber, Payne, and Puto (1982) conducted an empirical study and showed that a new product can increase the probability of an existing product being bought. The new product is then called the asymmetrically dominated alternative.
To better understand the asymmetrical dominance effect, imagine that there is a choice between two products. Each product has some unique good qualities and maybe even some bad ones, making it difficult for a person to choose between these two products. Choosing one of them would involve a trade-off (Payne, Bettman, Coupey, & Johnson, 1992), such as quality over price trade-trade-off. This is where the third, asymmetrically dominated option, is introduced. It has to be constructed in a way that it is dominated by one of the existing options, but not the other one. This option would never be chosen, because it is inferior to at least one of the existing options. However, empirically it is proven that it can increase the probability that someone purchases the option that dominates it, violating both the regularity and similarity hypothesis. The new alternative product is often called a “decoy” (Ariely & Wallsten, 1995).
A very good example of the ADE (asymmetric dominance effect) was given by a popular TV called “Brain Games” (National Geographic). The experiment was conducted at the popcorn stand in a cinema. Firstly the control group was presented with two options for popcorn – small for $3 and large for $7. Most of the people preferred the smaller option. Later the experimenters introduced a third option – medium for $6.50. Now when people were presented with a choice between these three options, most of them went for the large option. One of the reasons for this was because now the large popcorn was perceived as a very good deal – $0.50 gives an upgrade from medium to a large. This third option,
medium popcorn, was a decoy. Businesses use this effect to steer people into choosing the options that are most profitable for the business.
Moreover, another research done by Dan Ariely will be used as a basis for the experiment design used in this research paper. Ariely was very interested by the pricing strategy that the international
newspaper The Economist used. There were three different subscription options available offering different versions of the newspaper:
Internet-only subscription US $59.00 Print-only subscription US $125.00 Print-and-Internet subscription US $125.00
It is clear that nobody would take the print only version as it is priced the same as the print and online version together. It was taken down from the newspapers website, but Ariely wanted to investigate if this strategy could indeed bring higher revenue. He conducted an experiment with MBA students from MIT’s
Sloan School of Management. He gave 100 students a choice between these three options and the results showed that 84 percent chose the print-and-Internet version and 16 percent chose the Internet-only while nobody took the print-only subscription. Ariely then gave the students the choice between the same subscriptions, but took out the one that nobody chose – the print-only version. Now the results showed that 68 students chose the Internet-only. That is a 52 percentage point decrease in the people who chose the more expensive subscription. This proved that The Economist might have chosen this strategy intentionally to increase their revenue.
Trade-off aversion
Another angle how to look at ADE is to analyze what happens in a person’s brain when making a choice which item to purchase. A research done by Luce, Payne, & Bettman (2001) suggests that a trade-off-type choices are emotionally taxing and that people tend to averse to them. To test whether trade-off aversion explains the ADE, Hedgcock and Rao (2009) did a series of brain scans to see if the negative emotions associated with making a decision between two appealing alternatives can be reduced if a third option, the decoy, is introduced. Hedgcock and Rao focused the scans on the areas of the brain that are asociated with negative emotion Their results show, that indeed trade-off choice sets are associated with relatively greater negative emotion.
This research might be used by marketers and businesses when they are confronted by the costumers for “tricking” them, as it is proven that people experience less negative emotions while making these decisions.
Background information Online market
Most of the previous researches focus on purchase decisions made offline, but nowadays so many purchases are made online and it is estimated that in 2014 the online retail market in US alone will be worth 306 billion US Dollars (Centre for Retail Research). For this reason there is necessity to see how and if the consumer behavior changes when making purchases online. The amount of information
available to compare products and the reduced search costs makes the online shopping environment vastly different for offline shopping environments. One of the papers that researched the asymmetric dominance effect in an online environment was conducted by Fasolo, Misuraca, McClelland, & Cardaci (2006). In
their paper they investigated if the ADE is more or less pronounced depending on the animation of the product display. Firstly they checked if there is an ADE at all and later they examined if the effect is stronger if the photos of the products very moving. In total they conducted experiments in two countries – Italy and USA. They developed “dummy” websites, where university students, could choose the laptop that they would like to buy. First experiment that was conducted in Italy employed 240 Italian
undergraduate students. The results showed that without the decoy 59% chose the competitor while with the decoy 63% chose the target. In the other experiment that was conduct in the US there were 76 participants. In the control group 35% chose the target, while in the treatment group this number rose to 68.9%. In both of the experiments the results were significant and confirmed the existence of ADE in online shopping environment.
Mobile application markets
Apple was the first company to introduce their own software market. Later companies like Google and Windows followed. Shopping in a mobile application store happens in the following steps: consumer provides his/her credit card details only once when registering an account and later he/she can purchase applications, just buy clicking “buy” button. The application then is downloaded and installed on the device. Consumer can only shop at one market which provides the applications for his/her device. This purchasing process makes it different from offline or online shopping patterns and given these differences the motivation for this study is given in the following paragraph.
Hypothesis and predictions
The previous researches consistently show that decoy effect exists among all product groups, leading to the hypothesis of this paper, which is that the introduction of a decoy will increase the proportion of people choosing the more expensive messaging app.
The possible reasons for the hypothesis to fail would be that consumers treat applications differently from other product types. In addition this different treatment could be only concerning messaging applications as they are so widely used. Another prediction is that the decoy is too obvious and people can consciously ignore it. However, previous researches such as Dan Ariely’s (2010) The Economist experiment used a very simple design and did not encounter any problems of participants seeing through the experiment.
Method
Product
The research is method is mostly based on the previously mentioned research done by Ariely (Ariely, 2010) , the difference being that the fictional product, a mobile application, that is offered to the subjects can only be bought in a mobile application market. Messaging applications are chosen because they are one of the most downloaded and widely used applications among smartphone users. The applications were made similar to the popular products on the mobile application markets, such as Whatsapp, WeChat etc.
Participants
The sample of this research consists of my friends and acquaintances, with which I have not discussed this research. Subsequently, the sample is randomly split in half – control group and treatment group. More specifically, all the names of the participants are put into an excel spreadsheet and the program randomly assigns either a value “0” or “1” to each of the participants. People who had a “0” receive the link to the control group survey, the rest were redirected to the experiment group. Each person receives a link to an online survey. Since not all of the recipients took part in the experiment and the control group had relatively larger sample size, the help of the thesis supervisor was asked, to distribute the link to the experiment survey to her friends.
Survey design
When opening the survey, the control group is presented with two available mobile application choices. One of the applications has better features, but the other one is considerably less expensive. Contrary, the treatment group receives a link to a different survey in which they have to make a choice between three products – two of them are the same products as in the control group and a third option. The third option or the decoy costs the same as the expensive product in the control group. Using this method will show if introduction of a decoy will increase the proportion of people choosing more expensive application from the control group. The details of the experiment and the survey design can be found in the appendix. Furthermore, to control for other variables, three other questions were introduced. The results will be controlled for gender, age and if the person possesses a smartphone or not.
Results
The results of the experiment are summed up in the following figure:
Control group
When the participants had a choice between only 2 products, the majority chose the competitor application (App 1). Out of 62 subjects 47 (76%) chose App1 while 15 (24%) chose the target - more expensive App2. Most of the people in this group had a smartphone (77%) and were female (63%). Experimental group
The treatment group, which consisted of 52 other respondents, received a survey witch had an additional option App3 – the decoy. Introduction of the decoy shifted the proportions of the people choosing App1 and App2. Now the more expensive version was chosen 25 times (48%), App1 was chosen 27 (52%). Most of the people in this group had a smartphone (95%) and were female (75%).
Asymmetric dominance effect
The results show that the increase in the proportion of people choosing more expensive option is 24 percentage points. This increase proved to be statistically significant when was tested with a two sample proportion test using 95% confidence interval (z=-2.6768, p-value=0.007). This proves the hypothesis that the introduction of a decoy leads to an increase of the proportion of people choosing the target. Furthermore, it is needed to address possible concerns of the comparability of the two samples used in this experiment. As noted before, the proportions of females and males, age and owning a smartphone differ across the two samples and it is important to understand if this can in any way affect the validity of this experiment. Firstly, by conducting a logit regression of the control variables, it is shown that there are no significant biases that affect whether a person chooses one application over other (see appendix). Secondly, since the sample size of this experiment is fairly small, it may be beneficial for further researchers to check for any influence of the control variables by increasing the size of the samples.
24% 48% 76% 52% 0% 10% 20% 30% 40% 50% 60% 70% 80%
Control group Experiment group
Experiment Results
Target App Competitor App
Conclusion
As mentioned before, the mobile application market is growing at a very fast pace, so understanding how consumers make their choices in these markets are very important for people selling as well as people who are buying the applications. Knowing these mechanisms will lead to higher transparency in these markets. To sum up, the results show that the asymmetric dominance effect affects how people make choices, when choosing the right application for their smartphone. This has several useful implications. Firstly, mobile application designers and businesses can increase their revenue, by smartly developing the features of their applications as well as pricing them in the right way. Secondly this raises consumer awareness for the different ways how the businesses could be manipulating their choice. This may lead to better consumer satisfaction in the long term.
This paper also shows the importance of testing old behavioral economics and consumer behavior theories in new environments and in different conditions.
Further research should test these results in conditions that are even closer to the ones in real life. An example of this might be, letting people choose the application while using a mobile device. The device could have a theoretical mobile application market installed. This would increase the validity of the experiment as this environment would be closer to the one where real-life transactions take place.
Bibliography
Ariely, D. (2010). Predictably Irrational, Revised and expanded edition. Harper Perennial.
Ariely, D., & Wallsten, T. S. (1995). Seeking Subjective Dominance in Multidimensional Space: An Explanation of the Asymmetric Dominance Effect. Organizational Behavior and Human Decision Processes, 63(3), 223-232.
Batsell, R. R. (1980). Consumer resource allocation models at the individual level. Journal of Consumer Research, 7(1), 78-87.
Centre for Retail Research. (n.d.). Centre for Retail Research. Retrieved April 13, 2014, from http://www.retailresearch.org/onlineretailing.php
Copulsky, W. (1976). Cannibalism in the Marketplace. Journal of Marketing, 103-105.
Debreu, G. (1960). Individual choice behavior: A theoretical analysis. American Economic Review, 50(1), 186-188.
Fasolo, B., Misuraca, R., McClelland, G. H., & Cardaci, M. (2006). Animation attracts: The attraction effect in an on-line shopping environment. Psychology & Marketing, 23(11), 799-811. Gartner. (2014, January 22). Retrieved May 30, 2014, from Gartner Says by 2017, Mobile Users Will
Provide Personalized Data Streams to More Than 100 Apps and Services Every Day: http://www.gartner.com/newsroom/id/2654115
Hausman, J. A., & Wise, D. A. (1978). A Conditional Probit Model for Qualitative Choice: Discrete Decisions Recognizing Interdependence and Heterogeneous Preferences. Econometrica, 46(2), 403-426.
Hedgcock, W., & Rao, A. R. (2009). Trade-Off Aversion as an Explanation for the Attraction Effect: A Functional Magnetic Resonance Imaging Study. Journal of Marketing Research, 46(1), 1-13.
Huber, J., Payne, J. W., & Puto, C. (1982). Adding Asymmetrically Dominated Alternatives: Violations of Regularity and the Similarity Hypothesis. Journal of Consumer Research, 9(1), 90-98. Luce, M. F., Payne, J. W., & Bettman, J. R. (2001). The Impact of Emotional Tradeoff Difficulty on
Decision Behavior. In J. B. Elke U. Weber, Conflict and Tradeoffs in Decision Making (pp. 86-109). Cambridge: Cambridge University Press.
Luce, R. D. (1959). Individual choice behavior: A theoretical analysis. New York: Wiley. McFadden, D. (1974, November). The measurement of urban travel demand. Journal of Public
Economics, 3(4), 303-328.
McFadden, D. (1980). Econometric Models for Probabilistic Choice Among Products. The Journal of Business, 53(3), 12-29.
National Geographic. (n.d.). National Geographic. Retrieved April 10, 2014, from
http://channel.nationalgeographic.com/channel/brain-games/videos/the-decoy-effect/ Payne, J. W., Bettman, J. R., Coupey, E., & Johnson, E. J. (1992). A constructive process view of
decision making: Multiple strategies in judgment and choice. Acta Psychologica, 80(1-3), 107-141.
Tversky, A. (1972). Elimination by aspects: A theory of choice. Psychological Review, 79(4), 281-299.
Apendix
Experiment design in details
The experiment was conducted using an online form generator TypeForm (www.typeform.com). The opening page described that this is a marketing research on mobile messaging applications. Later, the participants were presented with a multiple choice question containing an image that shows the options. Question text for the control group was as follows:
“Imagine that you just bought a new smartphone and you have to decide which messaging application to install. You have these 2 options available on the mobile application market (e.g. App Store, Google Play, etc.). Which one would you buy?”
Following image displays the two messaging app options:
The treatment group was presented with the same question, but “2” was changed to “3”. Following image displays the three messaging app options:
Results Control group:
Treatment group
STATA outputs
Asymmetric dominance effect:
Control variables in control group:
Control variables in treatment group:
Pr(Z < z) = 0.0037 Pr(|Z| < |z|) = 0.0074 Pr(Z > z) = 0.9963 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Ho: diff = 0
diff = prop(x) - prop(y) z = -2.6768 under Ho: .0896591 -2.68 0.007 diff -.24 .0879883 -.4124538 -.0675462 y .48 .069282 .3442097 .6157903 x .24 .0542396 .1336923 .3463077 Variable Mean Std. Err. z P>|z| [95% Conf. Interval] y: Number of obs = 52 Two-sample test of proportions x: Number of obs = 62
_cons -1.583084 .7831299 -2.02 0.043 -3.11799 -.0481777 Age .2227333 .3533827 0.63 0.529 -.4698841 .9153506 Gender .2385898 .6204461 0.38 0.701 -.9774623 1.454642 Smartphone .1843981 .7360544 0.25 0.802 -1.258242 1.627038 App Coef. Std. Err. z P>|z| [95% Conf. Interval] Log likelihood = -34.027358 Pseudo R2 = 0.0081 Prob > chi2 = 0.9067 LR chi2(3) = 0.55 Logistic regression Number of obs = 62
_cons .4513492 1.513812 0.30 0.766 -2.515668 3.418367 Age_t -1.741734 .8115118 -2.15 0.032 -3.332267 -.1511997 Gender_t .8390351 .7608707 1.10 0.270 -.6522441 2.330314 Smartphone_t -.3387028 1.504474 -0.23 0.822 -3.287417 2.610011 App_t Coef. Std. Err. z P>|z| [95% Conf. Interval] Log likelihood = -32.843002 Pseudo R2 = 0.0878 Prob > chi2 = 0.0969 LR chi2(3) = 6.32 Logistic regression Number of obs = 52
