1 15-07-2017
The stability of consumer preferences
An empirical analysis of voting behavior for the “Top 2000”
2 Statement of Originality
This document is written by Student Maria Frederieke Toppen, 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.
3 1. Introduction
In many applications consumer preferences are assumed to remain stable over time (Oppewal et al, 2010). This means that having the same choice at different points in time will express the same preferences. The stability of consumer preferences is important for understanding the validity of fundamental concepts of microeconomic theory (Alevy et al, 2006). These fundamental concepts are often used in public policy analysis, but also in applied fields as marketing science.
Hoeffler and Ariely (1999) mention that most researchers believe that consumers construct their preferences when they are new to a product category, and develop more stable preferences at the end through experience with all the different products available in this product category. However, there are different situations where consumer preferences might not be stable. For example, preferences could shift when consumers are exposed to new product information (Oppewal et al, 2010), or the choices of other consumers may influence our preferences (San et al, 2002). However, these chancing preferences can become stable again in the long-term. But, there are also situations where the stability in the long run may also be challenged. According to San et al (2002), our behavior depends on past experience and that such behavior is governed by adaptive economizing procedures. Under these conditions, both the stability of utility functions and the demand functions in the long-run may be challenged.
Relevance
Many choices that are made in public policy analysis are based on the assumption that preferences remain stable (Alevy et al, 2006). As mentioned above, it is not sure whether these consumer preferences remain indeed stable over time. Unstable consumer preferences can lead to a change in demand. In industrial organizations for example, when analyzing a merger, it is assumed that demand is stable. If the demand is not stable over time, the merger decision based on this analysis can then be misguided. Because of this, it is important to know how demand and preferences changes over time.
Consumer preferences are also depended on price. But in this paper we would like to study preferences which are independent from price. For this reason we have chosen to analyze the music preferences, which preferences are not related to price. The music preferences that will be used, are taken from the annual ranking of the 2000 best pop songs of all time. The aim of this paper is to examine if these consumer preferences changes over time by answering the following research question: are consumer preferences stable over time? The reason why the data of the “Top 2000” is chosen, is explained in more detail in the methodology section. In
4 this paper we will be looking at the overlap between the different rankings of the “Top 2000” from year to year by using a similarity measure for indefinite rankings. Besides using a metric model, the data of the “Top 2000” will be analyzed for answering several more specific questions: to what extent is it possible for a new song to get immediately a high ranking?, to what extent can temporary shocks influence the ranking?, or are consumers more likely to vote for multiple songs of the same artist?
Most studies that have been done about music or pop charts are related to psychology. This study will relate the preferences for the songs in the “Top 2000” more to economy instead of psychology.
The second section of this thesis is the literature review which starts with a brief explanation of the theory about utility and preferences. Next several studies about consumer preferences will be discussed. Studies about the preferences of consumers on music will be looked at. In the third section, the methodology, it will be explained how the research question will be answered and the model that will be used will be explained. In the fourth section the data that will be used is explained. In the fifth section the data of the “Top 2000” will be explored in more detail by using descriptive statistics. In the next section, the analysis, we will use statistical tests to research how much movement there is in the ranking of the “Top 2000”. The final section will conclude the research.
5 2. Literature review
In the utility theory, the assumption is made that individuals are rational. A rational individual is an individual who chooses the option that gives him or her the highest utility, so the option which he prefers the most, everything else being constant. To put products in order, according to someone’s preferences, it needs to satisfy certain assumptions. Preferences should be transitive. This means that if you prefer product a over product b, product b over product c, then you should also prefer product a over product c. Preferences should also be complete. This means that when a consumer has to choose between two products, he can always rank them or be indifferent between the two products. Completeness of preferences rules out the possibility that a consumer cannot choose which product he prefers. The last assumptions is that preferences are reflexive, which means that a product is always as good as itself. (Varian, 1992).
As already mentioned in the introduction, there are different opinions about the stability of consumer preferences. Attitude researchers mention that an individual’s preferences are stable when this individual provides the same preferences at different times and/or in different contexts (Schwarz, 2007). The opinions about consumer preference stability differs especially between the different fields in science. Economists and psychologists have different opinions about utility and preferences (Ariely & Norton, 2008). Economists mostly assume preferences to be stationary (Andersen et al, 2008), which means that preferences do not change. And if consumer preferences are assumed to be non-stable, there are different ways to deal with this. Some economists for example even argue that changing preferences should not be thought about in the discipline of economics (San Miguel, Ryan & Scott, 2002). According to these economists, there are influences on preferences which may have little to do with the variables economist are concerned about (von Weizsacker, 1971). These influences are called exogenous variables. Fundamental concepts of microeconomic theory rest on the assumption of fixed preferences. A change in one of these exogenous variables should not influence the fixed preferences. Other influences could depend more directly on some economic variables, and are called endogenous influences. If these influences are not taken into account, it can lead to false predictions (von Weizsacker, 1971).
Rigby, Burton and Pluske (2016) studied the stability of preferences and the level of choice consistency of respondents within discrete choice experiments by looking at the willingness to pay. Choice consistency is when you are facing the same choice sets, you are making the same decision each time. They found high levels of intertemporal preference stability and choice consistency. For choice consistency, two factors were significant, namely the measure of choice task’s complexity and a measure of respondents’ cognitive capability.
6 Ariely and Hoeffler (1999) studied the development of stable preferences by examining three key dimensions: effort, choice and experience. These three factors should help to stabilize preferences. By looking at both objective and subjective measures of preference stability, they came to the conclusion that to stabilize preferences, it helps to make difficult trade-offs. They believe that there are two different types of processes for learning preferences. With the first type, preferences converge over time toward ideal trade-offs among the different properties. In the other type, preferences change over time as taste matures, such that the ideal trade-offs among the different properties change over time.
Simonson (2008) mentions in his paper to that preferences are constructive and for a large part determined by the choice context, task characteristics and the description of the different options. But he also mentions that even researchers who believe strongly in the construction of preferences, are likely to believe that more stable attitudes do exist and may influence specific choices. These stable preferences which are not determined by context, task or frame, often play a key role in shaping revealed object preferences. Examples of these stable preference components are the taste of liquorish or the experience created by a motion-sensitive videogame remote.
Although consumers cannot have explicit preferences for something where they have not been exposed to yet, consumers do have explicit or implicit preferences for ingredients of potential objects/experiences. Such preferences can change over time, because of changes in lifestyle or because of new information a consumer gets (Simonson, 2008).
Andreasen (1984) tested the effect of consumer life status on brand preferences and overall satisfaction with product and service purchases. The conclusion of his research was that respondents who were undergoing a change in their life status were more likely to change their brand preferences spontaneously. Also some households were showing increased dissatisfaction with their product and service purchases when not adapting to this change in their lifestyles.
According to Simonson (1989) individuals are uncertain about their own preferences because they find it difficult to determine the precise utilities of other options. His paper proposes that choice behavior under preference uncertainty is maybe easier to explain by assuming that individuals select the alternative supported by the best reasons. He explains this with the attraction and compromise effect. The attraction effect occurs when the introduction of an inferior good influences the relative attractiveness of other alternatives in a choice set1.
7 The compromise effect is when a consumer is more likely to choose for a middle option in a selection rather than the extreme options. The results show that attraction and compromise effects tend to be stronger among subjects who expect to justify their decisions to others.
Warren, McGraw and Van Boven (2011) studied the construction of preferences. According to them literature shows that preferences are sensitive to context and calculated at the time of choice, which led to the view that preferences are constructed. Constructed means that preferences are not stable and are influenced by properties of the decision task. One of the findings of their research was that preferences were indeed sensitive to context. An example they give in their paper is when they let individuals choose between products. Participants had to fill in a survey with an orange or a green pen. It turned out that the participants with the orange pen were more likely to prefer orange products, like Fanta. And participants with a green pen were more likely to prefer green products, like Sprite. Looking at the voting process of the “Top 2000”, this could also mean that the songs you vote for depend on the music you are listening to at that specific moment. If you are listening to Dutch music the moment of voting, you may vote for more Dutch songs than you would normally do.
Another reason for consumer preferences not being stable is that preferences depend on goals and goals change over time and across context. If goals change, this can lead to preference instability (Warren, McGraw & Van Boven, 2011).
Constructed preferences can also mean that preferences are calculated during the choice process. Calculating preferences means the integration of multiple pieces of information into a decision. The extent to which preferences are calculated depends on goals, cognitive constraints and the experience of the decision maker (Warren, McGraw & Van Boven, 2011).
Music preferences
For the purpose of better understanding consumer preferences in general, preferences on music specific will be discussed. People start to listen to music from a young age and spend many hours listening to the radio and their favorite artists (Boyle, Hosterman & Ramsey, 1981). 13 and 14 years old children prefer listening to music the most as indoor activity (Mulder et al, 2010). But also people of an older age consider music as an important aspect of their lives and listen to it frequently (Rentfrow & Gosling, 2003). How the music preferences of each individual evaluates over time differs. There are differences in music preferences between genders, ages and other characteristics. Also there can be a difference between the choice construction of different kind of preference goods.
8 There is a difference between hedonic and utilitarian goods. Utilitarian goods are primarily instrumental and functional. Examples are goods as ovens or computers. These are more “should” preferences. Hedonic goods are more experiential and provide more pleasure and excitement. Examples are sport cars or designer clothes. These are more “want” preferences. (Dhar & Wertenbroch, 2000). Music can also be seen as a hedonic good, which means that the valuation of music is based on the experience it provides to a consumer (Dhar & Wertenbroch, 2000). Dhar and Wertenbroch (2000) wrote a paper about the differences of consumer choices between these two different goods. The main finding was that a hedonic good is relatively preferred over the same utilitarian good in forfeiture choices than in acquisition choices. Forfeiture choices are decisions which of several products you will give up. Acquisition choices are decisions about which items you will acquire. In the context of the choices for the “Top 2000”, the choice for songs in the “Top 2000” are an example of forfeiture choices, as you are only allowed to choose for a limited amount of songs.
Christenson and Peterson (1988) did a research about the differences between genders in the structure of music preferences. They found evidence that males and females respond different to music. According to them, Frith (1981) mentions that these differences in music orientation between genders is a result of differences in culture and life-style expectations of males and females. In general, females tend to like pop music or the mainstream popular music more than males. Males like in general jazz, country/western and the harder forms of popular music more than females. The fact that music preferences differ between males and females is important to keep in mind for my research. If for example the percentage female voters is much higher a specific year, this can lead to a list with more pop and mainstream popular music ranked higher. It then seems that the preferences of the voters changed, although this was only the result of a different population of voters.
Boyle, Hosterman and Ramsey (1981) studied factors that influenced pop music preferences of young people. Their main findings were that characteristics such as the melody, mood, rhythm and lyrics were the most important reasons for preference. Less important were sociocultural variables. Peer influence, danceability and hearing songs on the radio seemed to be less important too. However, a reason for this could also be that the students did not want to admit that these things influenced them.
Although the research of Boyle, Hostermand and Ramsey (1981) was focused on students, from grade 5 up till college, and the “Top 2000” has on average more older voters, according to a study from Mulder et al (2010) about the music taste of Dutch people, someone’s music taste is already well developed in early adolescence and it stabilizes more during late
9 adolescence and early adulthood. Children younger than 15 years old still seem to have an increased preference for music which is most widely accepted by their peer group. Mostly this is chart-based music. At this point they are still developing knowledge about music. But after this age, children start to individualize their music preferences (Mulder et al, 2010). According to Holbrook and Schindler (1989), at an age around 24 someone’s music preferences become stable. Music preferences on younger ages can also be influenced by an individual’s parents, for example because of where they live/came from. Someone that grew up in Texas may prefer different music than someone who grew up in a city such as New York or Los Angeles (Rentfrow & Gosling, 2003).
In the research of Mulder et al (2010), they also asked the respondents to tell their favorite artist. The younger respondents were more likely to mention newer artist, while older respondents were more likely to mention artists who have been popular for a longer time (2010). Mulder et al mentions also the importance of gender, socio-economic status and educational level to music preferences. According to them males like louder, more monotonous-sounding genres better, and females are more into melodic, relatively softer genres.
A research done by Rentfrow and Gosling (2003) also showed the importance of music. The respondents of their research considered music to be as important as other lifestyle and leisure activities. Music can reveal important information about someone’s personality. There are many links between music preferences and personality, self-views and cognitive ability, and these factors can play a role in the formation and maintenance of music preferences (Rentfrow & Gosling, 2003). In their research is also looked at if individuals like to listen to different music each day, depending on their mood, or if their music preference were stable. The result showed that there was a clear underlying structure to music preferences.
To conclude the literature review, there are different views about the stability of consumer preferences. Preferences could be influenced by the properties of the decision task, and be dependent on goals, which are changing over time and across context. A life status change could also influence a consumers preferences. Difficult choices could help stabilize preferences. Looking at music preferences specific, preferences can depend on gender, age or background. The focus of this paper is on the stability of consumer preferences. That will be explored by using the “Top 2000”. Some issues that will be dealt with are related to the voter, such as gender and age. Other issues are related to the song itself, such as the origin, media attention, date of production and the lifespan in the ranking. Besides this more descriptive approach,
10 statistical analysis will be used to decide on the stability of the preferences of the “Top 2000” voters.
11 3. Methodology
The data
A problem with studying the stability of preferences is that it is hard to make a distinction between revealed and true preferences (Ariely et al, 2003). A reason for this is for example the price. Price can be a major influence on the choices that people make. When choosing between a product A and B, you might choose product A because you prefer it more than product B. However, a reason could also be that product A is much cheaper than product B, although you prefer product B (Hinloopen, 2011). Gneezy et al (2014) also gave an example of how prices can influence preferences. According to them, consumers often use the price as a measure of the quality of the product. If the price is higher, the consumers assume that the quality is higher, and consumers like products with a higher quality more. As a result of this, price is positively correlated with the liking of products. Demand may even increase for a product if the price is higher (Gneezy et al, 2014).
If the choices consumers make do not necessarily reflect their true preferences, using observed choices where price matters do not help us to examine if consumer preferences are stable over time.
To answer the research question, and taking into account that prices can lead to biased preferences, we need a dataset which is not influenced by price. However, using a dataset with products that do not have a price can be a problem too. If a product has no price, this means it is a free product. Demand for a free product can be unlimited, which can also lead to consumer not showing their true preferences. For this research the dataset of the “Top 2000” will be used. This is a list of consumer votes for best pop songs of all time. This list of 2000 songs is each year aired on Radio 2, starting Christmas morning at 9:00 and it ends just before the start of the new year. Consumers can vote each year on an amount of songs they think are the best of all time. In 2016 people could vote between November 25th and December 2nd. Consumer where allowed to vote for a maximum of 35 songs, which consisted of a maximum of 20 songs from a list made by NPO Radio 2, and 15 free choices of songs.
The reason why this dataset is useful is because it is not depending on prices. Voting for a song does not cost you any money. Besides that, voters only have a limited number of songs they can vote for, so they have to make a choice between songs, which solves the problem of an unlimited demand for free products. Because of this, the only factor that affects the order of the list is the true preferences of consumers, which makes this chart a good representation of the preferences of consumers. A factor we should take into account is that new songs are made each year, and some of them might get into the “Top 2000”, which will change the ranking.
12 Research
The dataset consist of the ranking for the years 1999 up to and including 2016 of the 2000 best songs of all time. A characteristic of this ranking is that it is an incomplete ranking. Not all songs you can vote for end up in the “Top 2000”. And the fact that you can also vote for free songs, that are not in the list of NPO Radio 2 yet, shows that the domain of the “Top 2000” can be seen as infinitive. In the years from 2003 till 2017 NPO Radio 2 also aired after the “Top 2000” songs that did not made it into the “Top 2000”. Another characteristic of the “Top 2000” is that it is top-weighted. This means that the top of the list is more important than the tail. I think that most voters value it more that their song is ranked 10 instead of 110, than 1610 instead of 1710. The “Top 2000” is also indefinite. The decision to make the “Top 2000” a ranking of 2000 songs was to celebrate the year 2000, however they could have chosen any particular number. The NPO could have chosen a ‘Top million’ for example, because the amount of songs that exist can be seen as unlimited. However, to broadcast only a specific amount of songs makes the “Top 2000” special. My parents for example are listening to the Top 2000 each day, but if instead of 2000 songs, it would be a top of 5000 songs, they probably would not follow it each day. The value of the list decreases with its size.
According to Webber, Moffat and Zobel (2010), there does not exist an indefinite rank similarity measure yet that takes all these characteristics (that a ranking can be incomplete, top-weighted and indefinite) into account. An important detail that this measure should handle is the fact that the different rankings are non-conjointness. A song that was in the “Top 2000” in 2015, may not have been in the “Top 2000” in 2016, the data in each “Top 2000” differs yearly. Webber, Moffat and Zobel (2010) propose in their paper the first similarity measure that is appropriate for indefinite rankings, the rank-biased overlap (RBO).
Rank-biased overlap is an overlap-based metric, which bias the proportional overlap at each depth by a convergent series of weights. Because of this, the infinite tail does not dominate the finite head (Webber, Moffat and Zobel, 2010). Given two infinite rankings, S and T, the value of the overlap is given by:
𝑅𝐵𝑂(𝑆, 𝑇, 𝑝) = (1 − 𝑝) ∑ 𝑝𝑑−1 ∞ 𝑑=1
|𝑆1:𝑑 ⋂ 𝑇1:𝑑|
𝑑
The
∞
shows the indefinite and incomplete nature of the ranked list. However, because our dataset consist of a ranking of 2000 songs,∞
will be replaced by a fixed depth of 2000. 𝑆1:𝑑 ⋂ 𝑇1:𝑑 measures the size of the overlap between ranking S and T to depth d. The parameter13 p determines how steep the decline in weights is. The smaller the value of p, the more top-weighted the metric is. If the value of p is zero, this means that only the number one ranked item is considered. The value of the RBO is between 0 and 1. A value of 0 means that the rankings are disjoint, and a value of 1 means that the rankings are completely identical. In this case of a value of p of zero, the value of RBO can only be 0 or 1 (Webber, Moffat and Zobel, 2010). The songs with a ranking from 1 to 150 have less competition, and the distribution of success is highly concentrated here (Prinz, 2016), which assumes that we have to choose a small value for the value of the parameter p.
The expectations for the results of this test is that the 18 different rankings of the “Top 2000” show a significant level of overlap. Wagner et al. (2014) define RBO values above 0.7 as highly stable, values between 0.4 and 0.7 as medium stable and values below 0.4 as not stable. If there is enough overlap (a value of the RBO of at least 0.4) between the 18 rankings, we may assume that consumer preferences are stable over time.
Besides using this metric model, in this research, the data will also be explored in a different way, such as events like the death of an artist, and these findings will be related to economic theory.
Having a brief look at the “Top 2000”, many songs seems to appear each year in the list. Although, in those 18 years of the “Top 2000”, there have been 4259 different songs in the list. But still each year a high percentage of the songs were the previous year also in the list. Especially the songs ranked higher, for example most songs in the top 10 (see table 1) does not seem to change a lot each year. In this thesis we also explore if new songs can change the ranking. To what extent is it possible for a new song to enter the list? And is it able to get a high position too in the list or are the high ranked songs unchangeable? And when a new song enters the list, will it stay in the ranking or will it leave the ranking soon? An example for the first two questions can be the song of Claudia de Breij, ‘Mag ik dan bij jou’, which is from 2011. As can be seen in table 1, this song was able to end up in the top 10. More examples like this will be analyzed in this research and it will be explained why they end up at a specific ranking.
14 Table 1. The ranking of the top 10 in 2016
Artist/title 2016 2015 2014 2013 2012 2011 2010 2009 2008
Queen - Bohemian Rhapsody 1 2 2 1 1 1 2 1 1
Eagles - Hotel California 2 3 1 2 2 2 1 2 2
Led Zeppelin - Stairway To Heaven 3 5 3 3 3 5 5 4 5
Billy Joel - Piano Man 4 6 7 18 36 29 45 45 50
Deep Purple - Child In Time 5 8 4 4 4 3 4 5 4
Boudewijn de Groot - Avond 6 7 5 5 5 4 3 3 3
David Bowie - Heroes 7 60 34 79 119 145 112 123 112 Claudia de Breij - Mag Ik Dan Bij
Jou 8 4 19 472 1565 0 0 0 0
Pink Floyd - Wish You Were Here 9 9 6 6 6 7 6 9 12
Pearl Jam - Black 10 12 12 16 24 40 452 803 0
Artist/title 2007 2006 2005 2004 2003 2002 2001 2000 1999
Queen - Bohemian Rhapsody 1 1 2 1 1 1 1 1 1
Eagles - Hotel California 3 3 3 2 2 3 3 4 2
Led Zeppelin - Stairway To Heaven 5 5 5 4 4 4 4 3 4 Billy Joel - Piano Man 34 54 49 66 60 58 57 83 121
Deep Purple - Child In Time 4 4 4 3 3 2 2 2 3
Boudewijn de Groot - Avond 2 2 1 5 8 25 41 121 428 David Bowie - Heroes 130 125 137 99 71 63 77 62 70 Claudia de Breij - Mag Ik Dan Bij
Jou 0 0 0 0 0 0 0 0 0
Pink Floyd - Wish You Were Here 14 8 16 21 20 29 25 45 0
Pearl Jam - Black 0 0 0 0 0 0 0 0 0
Besides looking at songs/products individually, artists/firms will also be examined. A lot of artists have multiple songs in the list. What does this say about the preferences? Maybe a consumer is likely to vote on more songs of the same artist, not because he/she likes the song, but because he/she is a fan of the artist and thinks he/she should be in the list more often.
Last, the influences of temporary shocks for preferences for songs in the “Top 2000” will be studied as well. For example, there were much more songs of Michael Jackson in the “Top 2000” after he died. What does this say about consumer preferences?
15 4. Dataset
As already mentioned before, the dataset which will be used is the “Top 2000”, which is a list of consumer votes for best pop songs of all time. Christmas morning starts with rank 2000, counting down to rank 1, which is aired just before the start of the new year.
The first time the Top 2000 was aired was in 1999. It was supposed to be an onetime thing, in honor of the millennium. The idea was to herald the new millennium, 2000, with the 2000 best pop songs of all time. Because of the success, NPO Radio 2 decided to make the “Top 2000” a yearly event. The first year, in 1999, from January on, consumers could send in a song with their memory of it, which was in their opinion the best pop song of all time. NPO Radio 2 collected 2000 songs, and posted it on the internet and teletext. From November consumers could vote on the order of these 2000 songs. It is not a chart with the most popular and commercial successful songs, but it is composed of thousands of individuals stories, or collective experiences2. In the following years, the way of voting has changed. In 2000, NPO Radio 2 did not use the selection list, but consumer where free to vote for songs, also for songs that were not included in this selection list. This free voting was only used in 2000. And for example, in 2005, before the voting started, consumers could send in their favorite songs which were not yet in the selection list, which were added later on to the selection list. Between 1999 and 2008, voters were allowed to vote for ten songs, after 2008 consumers could vote for 15 songs. Last year, consumers could vote for a maximum of 35 songs. In 2008, the “Top 2000” was 10 years old. To celebrate this, people could not vote for songs this year. All the votes of the 9 years before were counted together and this made a new ranking. According to Sterken (2014), it did not matter for the composition of the ranking that some years voters were free to vote for songs and other years they had to vote from a preselected list.
The “Top 2000” is becoming more popular each year. Not only people from the traditional audience are listening to it more, but especially more and more young people start listening to the “Top 2000”. In 2014, 3,8 million individuals voted, from which a high percentage was between the 15 and 25 years old2. In 2015, 8,1 million people listened to the
“Top 2000”, which were mainly Dutch people. For comparison, the Dutch population counted 16,9 million people in 2015. From the 8,1 million listeners, about 3,0 million were between the 10 and 34 years old3. Although the “Top 2000” is aired only on the Dutch Radio2, people from
2 http://www.radio2.nl/ebu/%3fpage%3dt2k-impact
16 other countries also vote for the “Top 2000”. Most votes from abroad comes from Belgium, but also people from Germany, the US, Norway or even China vote.
The “Top 2000” is still aired each year. The dataset that will be used consist of the ranking of each year, together with the title of the songs, the artists, and the year in which the songs are produced. We have data for a total of 18 years. However, the data of the year 2008 will not be used, because this ranking was not created by the votes of consumers.
The oldest song that have been in the ranking is from 1924, ‘Rhapsody in Blue’ from George Gershwin. The songs in the ranking are on average 27,1 years old.
17 5. Basic characteristics of the “Top 2000”
In this section we will be looking into the dataset in more detail and try to answer some questions such as the ones which were already mentioned in the introduction. The following issues will be discussed: differences because of gender and age, differences because the age of a song, effects because of the death of an artist, effects because of a change in media attention, the chance for new songs to enter the “Top 2000” high, the influence of the lifespan of a song, the origin of the songs, and the number of songs of an artist. By looking at these issues we hope to find an answer for the question if consumer preferences are stable.
People get influenced by things that happened around them, sometimes without noticing it themselves. Looking at the rankings of the “Top 2000”, many interesting changes can be seen. For example, in 2009 Michael Jackson had much more songs in the ranking than the year before. Or the fact that some old songs have never been in the ranking till a specific year. For example the song of Johnny Cash, “I Walk the Line”, from 1957, has never been in the ranking till last year, 2016. It was ranked 435. In this section we will try to find answers for these changes and we will show some main statistics about the “Top 2000”.
Differences between men, women and age
In the literature review is already mentioned that men and women differ in their music preferences. Women seem to prefer the mainstream music more than men, and men prefer jazz and the harder forms of popular music more than women. Are these differences in music preferences also reflected in the “Top 2000”?
The percentage male voters has always been bigger than the percentage female voters. In 2012, 35,6% of the total voters was female. In 2014 this percentage had increased to 40,9%4. Tables 2 and 3 shows the top 10 if only men or only women had voted, together with the actual ranking of the song in the “Top 2000” in 2016.
18 Table 2. Top 10 of men in 20165
Title Artist Year Actual ranking
1 Hotel California Eagles 1977 2
2 Bohemian Rhapsody Queen 1975 1
3 Stairway to Heaven Led Zeppelin 1971 3
4 Child in Time Deep Purple 1972 5
5 Piano Man Billy Joel 1974 4
6 Black Pearl Jam 1991 10
7 Wish You Were Here Pink Floyd 1975 9
8 Comfortably Numb Pink Floyd 1979 16
9 Heroes David Bowie 1977 7
10 Avond Boudewijn de Groot 1997 6
Table 3. Top 10 of women in 20165
Title Artist Year Actual ranking
1 Bohemian Rhapsody Queen 1975 1
2 Mag Ik Dan Bij Jou Claudia de Breij 2011 8
3 Hotel California Eagles 1977 2
4 Piano Man Billie Joel 1974 4
5 Avond Boudewijn de Groot 1997 6
6 Fix You Coldplay 2005 11
7 Imagine John Lennon 1971 12
8 Oceaan Racoon 2012 24
9 Heroes David Bowie 1977 7
10 Stairway to Heaven Led Zeppelin 1971 3
The average age of a song in the top ten for men was 37,2, for women this was 29 years old. Women seem to prefer newer songs, three songs in their top 10 were from later than 2005. The youngest song in the top ten of the men was from 1997. This is in line with the fact that women prefer popular/mainstream more than men do. Women also seem to prefer Dutch songs more.
19 They have three Dutch songs in their top 10, whereas men only have one Dutch song in their top 10.
The NRC (Dam, 2015) analyzed the different music genres in the “Top 2000”. As graph 1 shows, most songs in the “Top 2000” are popsongs. However, this is slightly decreasing with the years, and rock music is becoming more popular. Since the percentage female voters is increasing, we would expect the rock genre to decline, because this music genre is more preferred by men and the pop genre to decrease, as this genre is more preferred by women. So this observation contradicts with was mentioned by Christenson and Peterson (1988) as was referred to in the literature review.
Graph 16. Amount of songs in the Top 2000 that are pop or rock songs.
Graph 2 is showing the popularity of electronic music, Dutch music and jazz/soul/blues music. Especially electronic music is increasing in popularity. Till 2013 most electronic songs were from artists like The Prodigy, Fatboy Slim and Massive Attack (Dam, 2015). In 2015 there are 34 different electronic artists in the “Top 2000”. The preference for jazz/soul/blues music is declining, as is for Dutch music, which was increasing till 2007, but decreasing since. The declining of the preference for jazz music can be explained by the fact that the percentage of female voters is increasing over time. Jazz music is, according to the study of Christenson and Peterson (1988), more preferred by males. We would expect to have more Dutch songs in the
6
20 ranking now the percentage female voters is increasing, however, this is not the case. Perhaps other factors than gender are the cause for this trend.
Graph 26. Amount of songs in the Top 2000 that are from the genres electronic, jazz/soul/blues or Dutch.
Many people may expect the younger voters to vote for ‘new’ songs, like songs by Justin Bieber. The average age of a song in the ranking has increased, so apparently the older songs are still popular. NPO Radio2 published the top 10 for different age categories, and surprising is that for the younger age categories, most songs were produced many years before their birth, which means that they did not vote for the newest songs. The songs in the top 10 for voters with an age between 10 and 20 years old are on average 29,6 years old (table 4).
21 Table 4. Top 10 in 2016 of children between age 10 and 207
Title Artist Year Actual ranking
1 Bohemian Rhapsody Queen 1975 1
2 Piano Man Billie Joel 1974 4
3 Hotel California Eagles 1977 2
4 Stairway to Heaven Led Zeppelin 1971 3
5 Fix You Coldplay 2005 11
6 Africa Toto 1982 30
7 Viva La Vida Coldplay 2009 33
8 Smells Like Teen Spirit Nirvana 1991 32
9 Nothing Else Matters Metallica 1992 15
10 Sweet Child O’ Mine Guns N’ Roses 1988 51
Children’s music preferences can be influenced by their background or their parents. I think this is one of the reasons why songs like ‘Bohemian Rhapsody’ or ‘Stairway to Heaven’ are in the top 10 of the age category 10-20. If I hear those songs, it always reminds me of Christmas and new year’s eve, because that is the time when the “Top 2000” is played. I never have voted for the “Top 2000”, but if I would vote I would also definitely vote for songs like ‘Piano Man’ or ‘Bohemian Rhapsody’.
Also for the years 2014 and 2015, NPO2 published an article online about the top 3 for each different age-category (table 5). The age-categories for these two years are smaller than the age categories of 2016, which makes it hard to compare.
Table 5. Top 3 of 2014 and 2015 for different age categories8/9
Age category 2014 2015
6-10 1. Pharell – Happy 1. not given
2. Katy Perry – Roar 2. not given
3. Stromae – Papaoutai 3. not given
11-15 1.Queen – Bohemian Rhapsody 1. Queen – Bohemian Rhapsody
2. Eagles – Hotel California 2. Eagles – Hotel California
7 http://www.nporadio2.nl/nieuws/13526/populairste-top-2000-nummers-per-leeftijdscategorie 8http://www.nporadio2.nl/nieuws/6105/top-3-van-alle-leeftijden /
22 3. Coldplay – Viva La Vida 3. John Lennon - Imagine
16-20 1. Queen – Bohemian Rhapsody 1. Queen – Bohemian Rhapsody
2. Eagles – Hotel California 2. Billy Joel – Piano Man 3. Billy Joel – Piano Man 3. John Lennon – Imagine
21-25 1. Queen – Bohemian Rhapsody 1. Queen – Bohemian Rhapsody
2. Eagles – Hotel California 2. John Lennon – Imagine 3. Billy Joel – Piano Man 3. Billy Joel – Piano Man 26-30 1. Queen – Bohemian Rhapsody 1. John Lennon – Imagine
2. Billy Joel – Piano Man 2. Queen – Bohemian Rhapsody 3. Eagles – Hotel California 3. Billy Joel – Piano Man
31-35 1. Pearl Jam – Black 1. John Lennon – Imagine
2. Metallica – One 2. Pearl Jam - Black
3. Queen – Bohemian Rhapsody 3. Queen – Bohemian Rhapsody
36-40 1. Pearl Jam – Black 1. John Lennon – Imagine
2. Metallica – One 2. Pearl Jam – Black
3. Eagles – Hotel California 3. Claudia de Breij – Mag Ik Dan Bij Jou
41-45 1. Queen – Bohemian Rhapsody 1. John Lennon – Imagine
2. Eagles – Hotel California 2. Claudia de Breij – Mag Ik Dan Bij Jou
3. Pearl Jam – Black 3. Pearl Jam – Black
46-50 1. Queen – Bohemian Rhapsody 1. John Lennon – Imagine 2. Eagles – Hotel California 2. Queen – Bohemian Rhapsody 3. Led Zeppelin – Stairway To Heaven 3. Claudia de Breij – Mag Ik Dan
Bij Jou
51-55 1. Eagles – Hotel California 1.John Lennon – Imagine
2. Queen – Bohemian Rhapsody 2. Queen – Bohemian Rhapsody 3. Led Zeppelin – Stairway To Heaven 3. Eagles – Hotel California 56-60 1. Eagles – Hotel California 1. John Lennon – Imagine
2. Led Zeppelin – Stairway To Heaven 2. Claudia de Breij – Mag Ik Dan Bij Jou
3. Deep Purple – Child in Time 3. Led Zeppelin – Stairway to Heaven
23 61-65 1. Eagles – Hotel California 1. John Lennon – Imagine
2. Led Zeppelin – Stairway To Heaven 2. Claudia de Breij – Mag Ik Dan Bij Jou
3. Boudewijn de Groot – Avond 3. Eagles – Hotel California 66-70 1. Eagles – Hotel California 1. John Lennon – Imagine
2. Queen – Bohemian Rhapsody 2. Claudia de Breij – Mag Ik Dan Bij Jou
3. Boudewijn de Groot – Avond 3. Eagles – Hotel California 71-75 1. Queen – Bohemian Rhapsody 1. Claudia de Breij – Mag Ik Dan
Bij Jou
2. Eagles – Hotel California 2. John Lennon – Imagine 3. Booudewijn de Groot – Avond 3. Eagles – Hotel California
76-80 1. not given 1. Claudia de Breij – Mag Ik Dan
Bij Jou
2. not given 2. Queen – Bohemian Rhapsody
3. not given 3. John Lennon - Imagine
The song from Claudia de Breij, ‘Mag Ik Dan Bij Jou’ was already published in 2011, however it was not till 2015 that it became popular enough to end in the top 3 of the voters. The reason why the song ‘Imagine’ from John Lennon is high in the “Top 2000” in 2015 will be explained later in this part, but the main reason is because of the terroristic attacks of IS. Already in the age category 11-15 the song is in the top 3 in 2015. Even though these voters are young, they already got influenced by the media, which shows that preferences does not have to differ per se between different ages categories.
Comparing the top 3 for 2014 and 2015, each top 3 has changed. Each top 3 of 2014 had to make place for ‘Imagine’ of John Lennon. For the age 11-35 this, and the order of the top 3, changed. For the age categories 36-65 (except 51-55) two songs changed in the top 3.
To conclude, as expected, there seems to be a difference between male and female preferences, but not always as expected. The preferences for different age categories seems to differ less as expected. Young voters like the ‘old’ music too.
24 The age of songs in the “Top 2000”
In the “Top 2000” are a lot of songs from more than 30 years ago. Expected was that these ‘old’ songs would probably be less well known to new/younger voters (however table 4 shows the opposite). This expectation, and the fact that each year new songs are produced, for which to get a place in the “Top 2000”, other songs have to leave the ranking, would suggest that the songs in the “Top 2000” would be less old each year. In this part we will look at how old the songs in the “Top 2000” are and how this have changed over the years.
Graph 3 shows how many songs in the “Top 2000” are produced in which time period.
Graph 3. Amount of songs each year from different time periods.
The “Top 2000” keeps renewing. The “Top 2000” does not only contain old songs, but also songs that are only one year old. The amount of songs in the “Top 2000” from the years 1920-1989 are on average decreasing (graph 3). The graph lines for the twenties, thirties and forties are almost impossible to see in the graph, because there were each year maximum 5 songs in the “Top 2000” from these periods. Does the decreasing lines for most of the categories mean that preferences are changing? As already mentioned, old songs have to make place for new songs. So the fact that the number of songs originating in some periods are decreasing does not mean that the preferences of the voters are changing. And it makes sense that there is an increasing trend in the graph for the period 2000-2009 during these years. After 2009, the
25 preferences for songs from this period seem to stabilize. Comparing the nineties to the periods before, you would expect that the graph for the nineties would also be decreasing. But it is even increasing. When the first “Top 2000” aired, it was the end of the nineties. This means that in all the other years of the “Top 2000”, there were not more songs from the nineties to vote for than the years before. Although, in 1999 people did not had the whole year to send in songs for the “Top 2000”, so in 2000 there have been some new songs to vote for from the nineties. But after 2000, the songs you could vote for from the nineties are the same each year. The increasing line graph for the nineties suggest that music from the nineties is becoming more popular each year.
The decreasing trends for the older periods suggests that the average age of a song in the ranking should be decreasing, or at least stay the same. However, graph 4 shows an increasing trend in the age of the songs in the “Top 2000”. On average the songs in the “Top 2000” are from before the first ranking.
Graph 4. Average age of the songs in the “Top 2000”
As mentioned in the literature review, someone’s music preferences are formed during late adolescence or early adulthood (Holbrook and Schindler (1989). In 2007, around the 42% of the voters was between the age 45-60 (graph 5). In 2007 the average age of a song was 28,1 years old. This is in line with the research of Holbrook and Schindler. The voters seems to vote for songs from their early adulthood, when their music preference stabilized.
20 21 22 23 24 25 26 27 28 29 30 1999 2000 2001 2002 2003 2004 2005 2006 2007 2009 2010 2011 2012 2013 2014 2015 2016
26 Graph 510. The age of the voters in 2007
Scarcity
In 2007, Michael Jackson had six songs in the ranking. In the year after, 2009 (we do not use the data of 2008), he had 27 songs in the ranking (29 if we include Jackson 5). And Prince had 9 songs in the ranking in 2015, and 17 songs in 2016. David Bowie had 18 songs in the “Top 2000” in 2015, which increased to 26 songs in the ranking of 2016. Why this sudden increase in the amount of songs each of these artist had in the “Top 2000”? The similarity between these artists is that they all died in the year in which their amount of songs in the ranking increased. Michael Jackson died in 2009 and both Prince and David Bowie died in 2016. Another example is Ramses Shaffy. He died in 2009. Although the amount of songs in the ranking only increased with one, the ranking of each of the songs increased substantially. Apparently dying has a positive effect on the popularity of the artist in the “Top 2000”. If an artist dies, his products become more scarce. As consumer you know that there will not be any new products/songs produced. Scarcity increases the value of anything that can be possessed, the more scarce a product is, the more desirable the product becomes (Verhallen & Robben, 1994). The research of Verhallen and Robben (1994) showed that revealed preferences of consumers are influenced by scarcity. Zhu and Ratner (2015) mentioned in their paper that scarcity polarizes preferences.
But for how long does this scarcity has an effect on consumer preferences? Michael Jackson had 18 more songs in the “Top 2000” the year he died. Of these 18 songs, 13 songs
27 disappeared from the ranking the next year, 2010. However, 9 songs entered the list again in 2011 and 8 of them stayed in the list up to and including 2016. Because David Bowie and Prince died in 2016, we do not have data yet about how the preferences for their songs evolved over time. Ramses Shaffy’s amount of songs in the “Top 2000” increased with only one after he died, and this song disappeared after 5 years from the ranking. Looking at all the songs of these artists in the “Top 2000”, the ranking of each song increased the year they died. However, on average, the ranking of most of the songs is slowly decreasing the following years.
To conclude this part about the effect of scarcity on consumer preferences, scarcity does have an effect on consumer preferences. Consumers vote much more on an artist that became ‘scarce’. However, this effect fades away. My expectations are that the preferences for these songs return in the long-term back to their earlier ranking, and that the ranking of the songs of Prince and David Bowie that were already in the “Top 2000” before they died, will stay in the ranking, although the ranking of these song will decrease over the years to their ranking before they died. The songs that entered the “Top 2000” the year they died, will each year have a lower ranking till they disappear from the “Top 2000”.
Change in media attention
Nowadays, the media seems to have a substantial influence on our lives. But can it also influence our preferences?
Almost each year the top 2 of the “Top 2000” consisted of the songs ‘Bohemian Rhapsody’ and ‘Hotel California’. Although the song ‘Imagine’ of John Lennon always had a high ranking (it even was in the top 10 a couple of years), it increased in 2015 with 37 places to a number 1 position. In November that year, the IS attacked Paris. 130 people died because of the attacks (CNN, 2016). One of the attacks was at the Bataclan theatre. The day after the attacks, Davide Martello took his piano to the theatre and played the song ‘Imagine’, which is a song about peace. For many people, this song fitted to their feelings at that moment and gave them hope. Because of this, it was already expected that this song could become the new number one of the “Top 2000”, but social media also had an influence. Many people asked on social media to vote for the song, to make a statement (Telegraaf, 2015). And the song ended indeed at ranking 1 that year. A year later, in 2016, the song decreased to place 12.
The context in which an advertisement is placed influence the attitude of consumers towards the brand in the advertisement. A way people defend themselves against the fear of dying or the fear for other harm is by bolstering their cultural worldview. They create positive beliefs of those who preserve their worldview, and negative beliefs of those who do not (Liu &
28 Smeesters, 2010). In my opinion this happened to consumers who listened to the song ‘Imagine’ after the attacks, they created a (more) positive attitude towards the song.
Another song which position in the “Top 2000” is influenced by the media is ‘Gotta Catch ‘Em All’ by Jason Paige, better known as the theme song of the television show ‘Pokémon’. The song is from 1997, but did not enter the “Top 2000” till 2015. Wisse ten Bosch started in October 2015 an action on Facebook to invoke people to vote for the song. It is unknown how many votes a song needs to enter the “Top 2000”, the ‘NPO’ does not want to share this information (Trouw, 2015). Multiple mathematicians or just fans of the “Top 2000” tried to make a model to see how many votes a song needed. According to Peter Meindertsma (2016) a song needed more than 600 votes to get in to the ranking in 2016. He calculated this with a model that took Zipf’s law into account. He also made a model for the year 2014. According to this model, in 2014 a song ‘only’ needed 400 votes to get a place in the “Top 2000” (Meindertsma, 2014). According to a blog from a guy that calls himself Boudewijn55 (2014) the number 2000 of the list of 2014 needed at least 1000 votes. These are all guesses so we cannot say anything definitive about the number of votes a song need. However, we do know that this is apparently each year more, because the popularity of the “Top 2000” is increasing each year, which means more voters each year (NPO, 2012).
Wisse ten Bosch says the following about why the Pokémon theme song should be in the “Top 2000” (Trouw, 2015):
De Pokémon Theme Song heeft er tot nu toe nooit in gestaan! En dat terwijl dit nummer vanaf de eerste noten bij een hele generatie een gevoel van nostalgie en geluk teweeg brengt, en uit volle borst wordt meegezongen.”
Without social media, this song would not have made it to the list probably. The song entered the list at place 1666. In 2016, the game ‘Pokémon Go’ came out, which increased the popularity of Pokémon. Probably this game helped getting the theme song increase to the place 232 in 2016. My hypothesis is that, when the hype of ‘Pokémon Go’ is decreased, the ranking of the theme song in the “Top 2000” will probably decrease as well.
Maybe the ranking of the “Top 2000” is unfairly influenced by social media. However, in the literature review is mentioned how consumers cannot have explicit preferences for something where they have not been exposed to yet (Simonson, 20008). Social media may
29 introduce you to songs you did not know yet, or to songs you kind of forgot about. If I now have to call 10 songs that are my favorite of all time, I am sure that I forget some songs that I like even more than the 10 I called.
In the beginning years of the “Top 2000” the list was more often influenced by campaign to vote for a particular song (van Gijssel, 2016). But the voting system has changed which made it more difficult to get a song (high) in the “Top 2000” by start a campaign. However, there are still sometimes new, surprising, songs in the list. In 2016 three songs more than before of Johnny Cash were in the ranking. This is surprising because his songs are already made several years ago, so why was Johnny Cash this year suddenly more popular?
High entrance of a new song in the “Top 2000”
Many songs in the “Top 2000” are old time favorite songs for the voters. It can be hard for a new song to compete with these old favorites. A study done by Prinz (2016) showed that there is almost no competition or innovation between the top 100 songs. According to him, about 50 songs are not contestable, which is equal to 2,5%. For example, from 2004 till 2014 the top 5 of the “Top 2000” consisted of the same songs, only the order of these 5 songs have been different in those years (graph 6). The other 1950 songs compete heavily with each other.
30 Graph 6. The changes in the top 10 of the Top 2000 over the years.
Adele is an example of an artist who was able to enter the list high with new songs. I define a high ranking as a ranking higher than 100. In 2009 Adele’s song ‘Make you feel my love’ entered the list on place 28, and in 2011 she even had three songs that entered high in the ranking. ‘Rolling in the deep’ at place 21, ‘Set Fire To The Rain’ at place 26, ‘Someone Like You’ at place 6, and a little bit lower at place 220 ‘Rumour Has It’. And in 2015 her song ‘Hello’ entered the list at place 23. It is clearly possible to enter the “Top 2000” high, however, it is harder to stay at such a high ranking. As graph 7 shows, most of the songs enter high, but are on average slowly moving to a lower ranking.
31 Graph 7. Changes in the ranking of Adele’s songs
There are many other examples of high entries in the “Top 2000”. The song ‘Viva la Vida’ from Coldplay entered at place 11, the song ‘Oceaan’ from Racoon at place 33, ‘Dochters’ form Marco Borsato at place 25, ‘Rood’ from Marco Borsato at place 17, ‘White Flag’ from Dido at place 57, ‘Nine Million Bicycles’ from Katie Melua at place 23 and the song ‘I Follow Rivers’ from Triggerfinger at place 79. Graph 8 shows how the rankings of these songs changed over time. There is a clear decreasing trend visible in the graph, which means that each year the songs get a lower ranking. It seems that it is not per se impossible to start with a high position in the ‘Top 2000’ as new song, but it is hard to keep such a high position. This suggest that your preference may change for a slight moment, but will turn back after a while. Only the song ‘Oceaan’ from Racoon seems to be able to keep his high position. Also ‘Someone like you’ from Adele is still in the top 100, however there was a substantially decrease last year in the ranking of the song. It even increasing in ranking after it entered. Looking at all the songs that have been in the “Top 2000”, there are three songs that entered in the top 100 and are still in there, besides the song ‘Oceaan’. However, these three songs (‘Home’, from Dotan, ‘Hello’ from Adele and ‘Iron Sky’ from Paolo Nutini) are in the list for 2 and 3 years now, so we have to wait a couple of years before we can conclude something about the preferences for these songs.
32 Graph 8
New songs and songs re-entering the “Top 2000”
Table 6 shows us how many new songs entered the “Top 2000” and how many songs entered the “Top 2000” again after not having been in there the year(s) before. Because in 2008 consumers did not vote, we will not take this year’s ranking into account, and for 2009 the changes compared to 2007 will be given. We see a high amount of new songs in 2009, compared to the years before and after. This makes sense because the new songs that are from 2008 could not be voted on in 2008, so 2009 has the new songs of 2008 and of 2009.
Table 6. Amount of new songs and songs that entered the “Top 2000” again for each year
1999 2000 2001 2002 2003 2004 2005 2006 2007 2009 2010 2011 2012 2013 2014 2015 2016
New songs 0 612 114 35 91 30 119 85 135 220 40 97 99 60 173 187 169
Songs re-entering
0 0 340 199 96 96 55 54 45 85 93 77 80 68 16 15 34
Also interesting to note how many songs have been in the “Top 2000” for only one year. There have been 703 songs in the “Top 2000” that only have been in there for 1 year. 665 songs have been in the ranking all years long. In total there have been 4259 different songs in the “Top 2000”. See graph 9 for which percentages of those 4259 songs have been in the ranking for a specific number of years.
33 Graph 9. Number of years each song has been in the “Top 2000”
The categories one year and 17 years are the biggest. A reason why the one year (703 songs) is big is because each year new songs enter the list. In 2016 166 new songs entered the ranking and for most of these songs it is not possible to have been for more years in the ranking because they were made in 2016. Without these new songs in 2016, 537 songs have been in the ranking for only one year. On average it seems that it is harder to stay for many years in the ranking. After a while the song leaves the “Top 2000”. However, a substantial percentage have been in the ranking for all years, and I expect that these songs will show up in the ranking each year. These are songs like ‘Bohemian Rhapsody’ or ‘’Dancing Queen’, which people seem to prefer always, independent of how old the song is.
The influence of the lifespan of a song in the ranking
As graph 9 showed, many songs have only been in the “Top 2000” for a couple of years. Why are there a substantial amount of songs which were only in the ranking for one year? In this part we will look at the reason why some songs stay in the ranking for many years, and other songs disappear really quick.
Sterken (2014) wrote a paper about collective memory and nostalgia in the Radio2 “Top 2000”. His main findings about the factors that influence the lifespan of a song in the ranking are the superstar status of the performing artist, the length of the recorded song, the use of the domestic language and the debut ranking of the song. He expected that a song might decrease in ranking because it gets older which lead to a loss of memory and consumer will have
0 2 4 6 8 10 12 14 16 18 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
34 decreased preferences for it. However, nostalgic feelings can keep the song from decreasing in ranking. With superstar we mean those artists who are at the top of rankings, in some cases for a long time (Prinz, 2016).
The decreasing in ranking of songs can also be a result of a younger population of voters. There are more young voters of the “Top 2000” than before11. However, this cannot be the main reason
as we have seen in table 4 that young voters also vote for songs before they were even born.
Origin of the songs
Most songs in the “Top 2000” are from England, the Netherlands and the United States. But there are also songs in the list with a different origin, like Germany or France. We will look if preferences for the origin of songs changed over time.
In total the songs were originating from 46 different countries. On average there were each year 21 songs in the “Top 2000” that were sung by multiple artists with a mixed country background. I did not take these songs into account. To show the results for each of the 46 countries may be overwhelming and unnecessary. In the list are many countries, like the Philippines, Iceland or Peru, which had many years only 0, 1 or 2 songs in the list and are not important in this context.
35 Graph 10. The origin of the songs in the “Top 2000”
Graph 10 shows for each origin the amount of songs in the “Top 2000” for three different years. I choose to only include three years because all the years did almost not differ. I choose the first, last and middle year. The bars (graph 10) seem to change only slightly. This means that on average the preferences for songs for a specific origin did not seem to change. Only the bars for England, the Netherlands and the USA seem to change more over time. These were the bigger categories so it makes sense that these categories also show bigger differences.
The countries itself can also be divided into different regions. Looking at the Netherlands, there are some artists that are well known for the province they came from. Guus Meeuwis who was born in Brabant, or Rowwen Hèze, who is from Limburg. Looking at the top 10 of the “Top 2000” for the voters in each province separately, we see that people tend to prefer music from their own province more. In the top 10 for Limburg, two local songs are in the top 10. Also Brabant, Drenthe and Zeeland have each one local song in the top 10. This is in line with the research of Rentfrow and Gosling (2003), who researched that people in different cities tend to like different music.
0 100 200 300 400 500 600 700 800 1999 2007 2016
36 Artists with many songs in the list
There are multiple artist in the “Top 2000” who have many songs in the ranking. These artists can also be called the superstars of the ranking. For this part I will use the data for the 10 artists that had most songs in the “Top 2000” in 2016, which are The Beatles, David Bowie, Queen, Rolling Stones, U2, Coldplay, Michael Jackson, Abba, Bruce Springsteen and Bløf. Duets, for example between Queen and David Bowie, are counted as one song for each artist.
Graph 11 shows the amount of songs each artist had in the “Top 2000” over the years. Each year The Beatles had most songs in the ranking, although their songs in the “Top 2000” dropped from 50 songs to 41 songs between 2013 and 2014. But does this mean that they became less popular?
Graph 11. Number of songs each artist had in the “Top 2000”
To explore if the preferences for these artist changed over time we have to look at the rankings of their songs too, because as artist you are more preferred if you have five songs in the top 10 than if you have ten songs between ranking 1500 and 2000. We will give an artist for a number 1 ranking 2000 points, for a number 2 ranking 1999 points, etc. Adding the points for each song of an artist together gives us graph 12.
37 Graph 12. How popular are the superstars in the “Top 2000”
For both David Bowie as Michael Jackson there is a huge peek. For both artists this was in the year they died. Michael Jackson is clearly also more popular the years after he died than the years before his death. Comparing both graphs, the popularity of The Beatles is closer to the other artists in graph 12 than in graph 11. In graph 11 The Beatles were more preferred than David Bowie in 2016, however, in graph 12, they are almost equally preferred. The Beatles may have had more songs in the ranking, but David Bowie’s songs were higher ranked. Before 2016, the year David Bowie died, his popularity was almost stable. Both graphs show how the preferences for Coldplay increase strongly in 2009. Queen has a slightly increasing pattern, which is surprising because they did not made any new songs after the start of the “Top 2000”. So apparently they became more preferred after a couple of years. That the preferences for Bløf, Coldplay and U2 were increasing over the years is due to the fact that these were still upcoming artists, unlike the Beatles who already made all their songs before the start of the “Top 2000”.
Conclusion
In this section we looked at the dataset in more detail .The following issues were dealt with in this section: differences because of gender and age, differences because the age of a song, effects because of the death of an artist, effects because of a change in media attention, the chance for new songs to enter the “Top 2000” high, the influence of the lifespan of a song, the origin of the songs, and the number of songs of an artist.
One of the findings was that gender has less influence on our preferences as expected in the literature review. Young people were expected to vote for pop chart music, however, they
38 seem to have similar preferences as the older voters. Another finding was that scarcity can change our preferences. When a product becomes scarce, or looking at the “Top 2000”, when an artist dies, it becomes more popular. However, after a while the popularity decreases again. People get also influenced by the media which can change their preferences.
If consumer preferences are stable, this mean it would be difficult for a new song to enter the ranking high. However, it is shown that it is possible for new songs to enter the list, although most of the times they are not capable of keeping their high ranking.
The origin of the songs in the ranking does hardly change, but the year of production of the songs in the ranking are changing, which is of course due to the new songs that appear. The percentage of new songs that enter the list yearly is rather stable.
39 6. Statistical analysis
In section 5 we discussed some basic characteristics of the “Top 2000”, in this section we will focus on the stability issue. There are several ways to measure stability. We will use a rank-biased overlap model, the Euclidean Distance measure and the percentage of overlap between rankings. We will also look at the average increase and decrease of songs in the “Top 2000”.
Overlap in rankings
We will start with a simple measure of the overlap between the ranking of the “Top 2000”. For each year is measured how many songs the year after were also in the ranking. The results are shown in table 7, as percentage of the total amount of songs in the first year (2000 songs). Only the amount of songs that were both years in the ranking are taken into account, not the specific rankings of each song. The average percentage of overlap is 88,7, which means that on average each year 1774 songs were the year before also in the “Top 2000”. This may seem as a lot of overlap, and you may want to conclude that the preferences are stable. However, the ranking of each song may have changed substantially, which may lead to less stable preferences. We also looked at the overlap between the ranking of 1999 and 2016. ‘Only’ 40,5% of the songs were both listed in the 1999 and 2016 edition. In total, there have been 664 songs that have been in the “Top 2000” each year, which is equal to a percentage of 33,2% of the 2000 songs in the ranking.
Table 7. Percentage of songs that was also in the ranking the year before.
Years 1999-2000 2000-2001 2001-2002 2002-2003 2003-2004 2004-2005 2005-2006 2006-2007 Percentage 69.40 77.30 88.30 90.65 93.70 91.25 93.05 91.00 Years 2007-2009 2009-2010 2010-2011 2011-2012 2012-2013 2013-2014 2014-2015 2015-2016 1999-2016 Percentage 84.90 93.35 91.30 91.05 93.60 90.55 89.90 90.00 40.45
These percentages of overlap between the “Top 2000” each year does not give us any information about how stable the ranking of each song was over the years. To compare the rankings of the overlapping songs a Euclidean Distance measure will be used. The Euclidean Distance (ED) measure is a true-order rank measure (Huang & Ling, 2005), which is equal to
