JOURNALW
Economic-
lidkg&itiOn
Journal of Economic Behavior & Organization Vol. 901 (1997) 185-205
Economics and the architecture of popular music1 W. Mark Crain, Robert D. Tollison* Center for Study of Public Choice, George Mason University, Fairjkx, VA 22030, USA Received 27 April 1995; received in revised form 12 December
1995
Abstract A simple supply and demand framework is developed to study the time-series pattern of music. Changes in the internal structure of successful songs, it will be argued, are tied to market forces. An extensive data set has been developed to enable the investigation of a large number of issues in this spirit. The analysis proceeds by first asking a basic question: Has the structure of music changed in
these 50 years? We discover major regime changes in the mid 1950s and again in the mid 1960s that generally conform to intuition. The analysis then turns to specify a system of demand and supply equations to explain the patterns in popular music over the sample period. JEL classification: Keywords:
L82
Industry studies; Services
1. Introduction It is commonplace today to see applications of economic analysis to an expanded domain of topics not previously touched by the dismal science. These include the economic analysis of the family, law, politics, crime, religion, sports, sociobiology, and a wide array of other topics. Economic methodology has not supplanted competing paradigms in other disciplines, but it has gained a strong following. This paper brings economic analysis to bear on yet another subject - popular culture - where popular culture is construed to be the institutions and fabric of the everyday life of ordinary people. Obviously, economics has traditionally addressed elements of popular culture
* Corresponding author. ‘We are grateful to Robert McCormick and William F. Shughart Rollandini for able research assistance. The usual caveat applies. 0167-2681/97/$17.00 0 1997 Elsevier Science B.V. All rights reserved PUSO167-2681(96)00901-8
II for helpful
comments
and to Mark
186
U?M. Grain, R.D. Tollison/J.
of Economic Behavior & Org. 32 (1997) 185-205
including such topics as art, fashion, substance abuse and addiction, and other behavior which reflects broader lifestyle trends. Such analysis is the bread and butter of modem microeconomic analysis and the laws of demand and supply. In this same spirit we present an economic analysis of popular music. Electronic technology has driven down the cost of distributing music to such a low level that in modern societies few products or services are consumed more than songs. The supply of songs in supermarkets, elevators, and holding on the telephone is highly elastic, which naturally has increased the quantity of songs consumed. The omnipresence of songs make popular music a prime indicator of aggregate preferences and trends in modern culture. Successful songs should have common elements, and it is these elements which form the internal structure of songs that we seek to explain with standard economic principles. A simple supply and demand framework is developed to study the time-series pattern of music. Changes in the internal structure of successful songs, it will be argued, are tied to market forces. The methodological perspective of the paper is clear - economic forces dictate the form and content of music - that is, the economy drives culture. The interplay between culture and economics is obviously more complex, and the direction of causality runs both ways. But our perspective in this paper is on the impact of economic factors on the cultural phenomenon of music. However, music can influence culture in a variety of ways. Music may lead young people to rebel, or it may teach them the value of freedom and competitive markets. If music has a social marginal product, it is the sum of its positive and negative effects that must be ascertained. But the causal chain from music to behavior is not the issue tackled in this paper. First and foremost, we address the impact of economic behavior on the market for music. An extensive data set has been developed to enable the investigation of a large number of issues in this spirit. The sample consists of all songs which attained the status of number one in Billboard Magazine’s ratings over the years 1940 through 1988. There are 921 songs in the data set. The observations on these songs include: the length of its tenure as the number one song; the meter (pace) of the songs (i.e., beats-per-minute); key; length (running time in seconds); and the market share of dominant artists and songs. This information was collected from Billboard Magazine by a trained musician who obtained copies of the sheet music and listened to recordings of the songs with a metronome. The music data are paired with contemporaneous data on the economy and data on technological changes in the music/recording/broadcast business to analyze the pattern of popular music over a period in which American culture underwent considerable flux, including three wars, economic and birth rate booms and busts, and technological evolutions, to cite only a few examples. The analysis proceeds by first asking a basic question: Has the structure of music changed in these 50 years? As subjectively obvious as the answer may seem, we seek to locate structural shifts in the data, and thereby identify musical epochs in a more rigorous manner. We discover major regime changes in the mid 1950s and again in the mid 1960s that generally conform to intuition, In addition to identifying musical epochs, the initial time-series pass at the data reveals considerable variation even within the periods where clear epochal breaks in the data are seen. The analysis then turns to specify a system of
U?M. Grain. R.D. Tollison/J.
of Economic Behavior & Org. 32 (1997) 185-205
187
demand and supply equations to explain the patterns in popular music over the sample period. On the supply side, for example, market concentration, the value of radio air time, and the availability of substitutes (television) are key determinants of the length and meter of successful songs. The model also examines the interdependent, endogenous relationship between song structure and the market concentration of number one songs. On the demand side the music market responds to changes in subjective rates of time preference, fluctuations in the business cycle, and demographic shifts represented by changes in the teenage share of the population. We are aware, of course, that there is a scholarly literature on the economics of popular music and that this paper is related to and derivative from this work. Some of this research is striclty theoretical in nature, stressing, for example, the industrial organization of the record industry (Baker, 1991) or the aesthetic and economic aspects of the consumption of music (Shanahan, 1978). Other papers are similar to this paper in that they are essentially empirical investigations of the popular music industry. For example, Peterson and Berger (1975) use data on the popular music single record market by year to test hypotheses from the sociological literature on whether there are cycles in culture forms. Other empirical papers estimate the demand for record albums (Belinfante and Davis, 1978), attempt to assess the attributes of different music types (Alpert, 1983), and examine concentration and pricing in the record music industry (Belinfante and Johnson, 1982, Anderson et al., 1980). We acknowledge these earlier contributions. We also suggest that our paper introduces new issues and techniques into this discussion, including the use of time-series analysis, more detailed data on the music itself (e.g., song length and meter), and an effort to place the evolution of popular music into a fairly well specified economic model. For these reasons we think that our paper adds to this literature.
2. Epochs in popular music 2.1. What is a song? A song is a composite of characteristics in the sense described by Becker (1971), and the characteristics that constitute a song can be thought of as Beckerian z-goods. There is some experimental evidence to support the conceptual approach that consumer preferences are tied to these fundamental elements of music. Clyne (1982), a neuroscientist, has monitored human subject brain waves to identify musical forms that elicit statistically similar neurological responses. Out of this research Clyne identifies what he labels seven essentric forms in music: Anger, Hate, Joy, Sex, Love, Grief, and Reverence. Clyne’s work broadly suggests that the consumption of songs can be decomposed into preferences for separable characteristics of more basic elements. In this spirit we first examine two distinct dimensions of songs which lend themselves to objective measurement. We focus primarily on length (running time) and meter (the number of beats-per-minute) as two indicators of the style of music which succeeds in the market in particular years. Of course, the analysis could be applied to additional characteristics of songs.
WM. Grain, R.D. Tollison/J.
188
IS0
of Economic Behavior & Org. 32 (1997) 1X5-205
.
320 :I!
:
65
Year
Fig. 1. Length and beats per minute of no.1 songs: 1940-1988.
2.2.
Time-series analysis
Fig. 1 illustrates the length in seconds and beats-per-minute (BPM) of number one songs over the sample period. The considerable variation in these patterns suggests that the style of popular music was changing. A rigorous test for significant breaks in the data is possible using time-series regression analysis. Table I presents regression results for BPM and length as a function of time. Two specifications are estimated on each characteristic, one which allows for intercept changes and the other for slope and intercept changes. The results show that there are significant breaks in the pattern of both time-series which are helpful in identifying distinct musical epochs. With respect to length, there are three statistically distinct periods in the data: 19401955, 19561964, and 1965-1988. These breaks are displayed graphically in Fig. 2. At the first break in the data, between 1955 and 1956, songs became significantly shorter. At the second break, between 1964 and 1965, songs became significantly longer. These two points in time represent empirically identifiable regime changes in the style of music which dominated popular culture. The averages for song length in each period are shown in Fig. 3. BPM presents a similar story. The breaks in the data are the same, as shown in Fig. 4, although the direction of the change in the BPM data between the periods is the opposite from that in the length data. In 1955-1964, BPM increased 29 percent above the rate in 1940-1955, as shown in Fig. 5. In 1965-1988, BPM dropped by 9 percent.
U!M. Grain, R.D. Tollison/J. of Economic Behavior & Org. 32 (1997) 185-205 Table 1 Regime changes:
1940-1988 Dependent BPM
Independent
variables
Dummy for years After 1956 Dummy for years After 1964 Interaction term (Time * D>1956) Interaction term (Time * D>1964) Time Constant R-square (adjusted) F-statistic
189
(1) Estimated 27.55 8.57 -11.34 -3.76
variables BPM (2) coefficients 1.32 9.35 -0.89 -7.06 -
94.65 49.07 0.62 40.79
94.7 1 53.78 0.67 49.36
Length (3) (T-ratios) -20.20 -2.02 89.34 9.55
165.87 27.71 0.73 64.41
Length
Weeks
(4)
(5)
-
5.22 2.37 -7.44 -3.18 -0.43 -3.86 -0.31 -2.98 0.14 3.17 4.07 9.8 0.72 25.18
-0.90 -2.22 2.79 7.69 164.75 32.52 0.79 92.35
Fig. 2. Epochs in popular music average song length, 1940-1988
190
WM. Grain, R.D. TollisodJ.
of Economic Behavior & Org. 32 (1997) 185-205
1956-1964
1965-1988
Pericd Fig. 3. Average length of no.1 songs.
The statistical history of popular music can thus be summarized quite succinctly with respect to meter and running time. In the 1940s popular music was stable and static with respect to its pace and length. In 1955 the rock and roll revolution came of age. Not only does one observe significant intercept shifts as described above, but one also observes significant trend changes in both length and BPM. For example, the interaction term between time and the dummy variable for 1956 in Table 1 indicates that BPM was increasing more rapidly than in the other two periods. At the same time length was declining more rapidly than in the other periods (see equations (2) and (4) in Table 1). The slope change model for length is plotted in Fig. 2 and for BPM in Fig. 4. The time-series characteristics of popular music offer an interesting pattern to be explained. Clearly, rock and roll in the mid 1950s was a revolution in the structure of music. Major innovators in this era include the obvious, like Elvis Presley (Don’t Be Cruel), the Everly Brothers (Wake Up Little Susie), and the Coasters (Yukety Yak), as well as lesser known artists such as Danny and the Juniors (At the Hop), Jimmy Rodgers (Honeycomb), and Buddy Knox (Party Doll). And, indeed, conventional wisdom has it that Bill Haley and His Comets debuted rock and roll music with Rock Around the Clock in the movie Blackboard Jungle in 1955. But there is no doubt that rock and roll music is a statistical event worthy of further explanation. But this is not the end of the story. While the rock and roll revolution is clearly more than a stochastic fluctuation in the data, only a portion of this regime change was permanent. In the mid 1960s the structure of music shifted again, in this case incorporating longer and slower features of music. As will be seen later, technology to some extent altered the characteristics of music over all these
WM. Crain, R.D. Tollison/J. of Economic Behavior & Org. 32 (1997) 185-205
Fig. 4. Epochs in popular music average BPM, 1940-1988.
1956-1964
’
Period Fig. 5. Average BPM of no.1 songs.
191
192
WM. C&n.
R.D. TollisodJ.
--
of Economic Behavior & Org. 32 (1997) 18.5-205
ibsrs
-
90.
sonqr
Fig. 6. Average weeks at no.1 and total no.1 songs.
periods, but especially over the period that began in the late 1960s. Both within periods and across periods of music the task remains to explain what forces led to the changes observed in these data. Obviously, changes in the characteristics of songs are going to be reflected in the overall results of the competition to be the number one song. In fact, the characteristics of outcomes of this broader competition will reflect both demand and supply factors at work in the relevant market. The outcomes in this market for the 92 1 songs are shown in Fig. 6, where the total number of number one songs in each year and the average number of weeks a song stayed at number one are plotted. Following the same procedure used for BPM’s and length, time-series regressions are used to test for regime changes. In column live of Table 1, the regression results for average weeks as the number one song are shown. Plots of the results are given in Fig. 7, where statistically significant intercept and slope shifts are shown in 1955 and 1964. One way of interpreting the eventful year of 1955 in these data is that the early rock and roll innovators achieved a transitory dominance of pop music, which quickly broke down under the pressure of entry and imitation.
3. Market forces and music The pattern of song characteristics shown in Figs. 6 and 7 are the result of a market process. The elements of this process and how demand and supply considerations affect
WM. Gain. 0
R.D. Tol1isod.J. of Economic Behavior & Org. 32 (1997) 18.5-205
193
_
0
eyl
--
'DC..-::
,
Fig. 7. Epochs in popular music; average weeks at no.1, 194&1988.
the characteristics procedure. 3.1. Empirical
of pop
music
can be analyzed
with a generalized
least squares
model
We specify a system of equations designed to explain the internal structure of songs as well as three characteristics of the pop music industry. Variable definitions and sources are listed below:
AVERAGE METER OF NUMBER ONE SONGS=
a weighted average of the beats-per-minute for number one songs, where the weights are the share of weeks a song was number one (Billboard, various issues);
AVERAGE LENGTH ONE SONGS=
a weighted average running time for number one songs, where the weights are the share of weeks a song was number one (Whitbum, 1986);
OF NUMBER
194
WM. Grain. R.D. TollisodJ.
of Economic Behavior & Org. 32 (1997) 185-205
AVERAGE WEEKS AT NUMBER ONE =
average number of weeks songs held the number one rating (Billboard, various issues);
SHARE OF TOP 4 SONGS IN CURRENT YEAR=
the share of weeks the top four songs held the number one rating (Billboard, various issues); and
SHARE OF TOP 4 ARTISTS PRIOR YEAR=
the share of weeks the top four artists produced songs that held the number one rating (Whitburn, 1986).
IN
A number of independent variables are used to specify the various equations musical characteristics. The definitions and sources are as follows:
the number of operating FM stations in a year (Statistical Abstract, various editions);
FM RADIO STATIONS=
RADIO ADVERTISING EXPENDITURES=
for these
radio advertising expenditures in constant 1982 dollars (Statistical Abstract, various editions);
REAL
REAL PRIME INTEREST
RATE
ZZ
the prime rate minus the inflation rate (calculated as the growth rate of the CPI) (Economic Report, 1992);
GROWTH RATE IN MILITARY DEATHS=
annual growth rate in battle deaths (Levy, 1986);
REAL WEEKLY EARNINGS=
average weekly earnings in private agricultural industries in constant dollars (Economic Report, 1992);
GROWTH RATE IN REAL PERSONAL INCOME=
the growth rate of personal income in constant 1982 dollars (Economic Report, 1992);
non1982
UW. Cruin, R.D. To&on/J.
of Economic Behavior & Org. 32 (1997) 185-205
195
GROWTH RATE IN TEENAGE POPULATION SHARE=
the annual growth rate in the population between the ages of 13-19 (U.S. Census, various issues);
TEENAGE SHARE OF THE POPULATION=
the percentage of the total population 19 (U.S. Census, various issues);
MISERY INDEX=
unemployment rate plus the inflation rate (growth rate of the CPI) (Economic Report, 1992); and
TV SHARE=
operating broadcast television stations divided by the total number of operating broadcast radio and TV stations in a year (Statistical Abstract, various editions).
age 13-
Some of these right hand side variables in the song and industry characteristics equations are endogenous to the system and are estimated instrumentally. The predicted values are then used in the regressions reported in Table 2. These separate estimations for the endogenous, right hand side variables are provided in Table 3.
3.2. Supply-side factors and song structure We begin with factors affecting the supply side of the music market and market concentration in particular. Fig. 8 illustrates the concentration of the industry as measured by the share of the weeks held by the top four artists in each year. In 1940, for example, four artists (Glenn Miller, Bing Crosby, Tommy Dorsey, and Artie Shaw) produced songs which held the number one rating for 90 percent of the year. Over time, artist concentration has generally fallen, reaching a 50-year low in the mid 1970s. In 1975 the artist concentration ratio was 35 percent, and the top four artists were Elton John, Neil Sedaka, Captain and Tennille, and Tony Orlando and Dawn. Artist concentration is an important facet of the market model because, as we shall argue below, successful performers hold market power which allows them to produce longer songs in subsequent years. This strategy gains more radio exposure for the top artists and preempts air time from competing artists. What, then, are the determinants of the market concentration of successful artists? The results of the specification to explain the variation in this market share measure are shown in column 5 of Table 2. Artist concentration is negatively related to the number of FM radio stations and the growth rate in the teenage share of the population. Increases in the number of FM stations provided more avenues to reach consumers and led to more specialized market niches, such as stereo broadcasts. Both factors enhanced the entry of new performers and reduced market concentration. Large numbers of new consumers of
U!!M. Gain, R.D. To&on/J.
196 Table 2 Regression
results for characteristics
of Economic Behavior & Org. 32 (1997) 185-205
of pop music a Avg. meter Avg. length number one number one songs (BPM’s) songs (SECS.) (1) (2) Estimated coefftcients
Right hand variables
Avg. weeks at number one
(3) (asymptotic T-ratios)
Avg. length number one songs
PM radio stations Radio advertising real expenditure Real prime interest rate Growth rate in military deaths
2.72 (1.54)
Real weekly earnings Misery index
128.07 (3.29) 0.07 (4.67) -0.02 (-2.37) -721.38 (-5.94) -8.86 (-1.88) 0.25 (1.94)
Share top 4 artists in prior year
(4)
(5)
0.009 (3.86) 0.05 (6.28) 0.23 (2.88)
Avg. weeks at number one Share top four artists in prior year
Share top 4 songs in current year
-0.0006 (-4.14)
-9.12 (-1.37) -0.32 (-1.32)
77.04 (2.28)
Growth rate in teenage pop. share TV share Constant R-square Mean of endog. var. Std. err. of est. a Variable definitions
-2.09 (-3.27) 175.70 (5.97) 84.49 (17.57) 0.46 107.83 9.21
29.67 (0.48) 0.65 196.32 27.14
-12.17 (-2.23) 4.79 (10.46) 0.22 3.54 1.30
-0.03 (-0.39) 0.48 0.43 0.06
0.60 22.80 0.30 0.52 0.11
and sources are provided in the text.
pop music (faster growth in the teenage population) similarly reduced the market concentration of the top artists. This is undoubtedly the result of less brandname loyalty and consumption capital among these young, new consumers. One way to interpret these results is that the monopoly positions of top performers in the 1940s and early 1950s were broken down by demand (a burgeoning teenage population) and new technological forces (the growth of FM stations). This model explains 30 percent of the variation in artist concentration in the data. In the second market concentration regression in column 4 of Table 2, we see that the market share of the top four songs in a given year is positively related to artist concentration in the prior year and to WEEKS and LENGTH. This regression is a direct test of the ‘crowding out’ hypothesis mentioned above. This model explains 68 percent of the variation in the market concentration of top four songs.
WM. Crain, R.D. TollisodJ. Table 3 Instrumental
equations
used in Table 2 a
Independent variables
PM stations
Real radio ad expenditure
Prime rate
Real weekly earnings
Misery index
Growth in Teenage Pop share
-0.78 (-4.38) o.ooo3 (1.96)
Real weekly earnings Teenage share of population
3850.3 (1.55) 44.58 (5.64)
Time trend (1940=1)
0.002 (8.06)
-0.004 (-17.89) -0.08 (-30.54)
4.09 (15.81)
Dummy for years after 1952 Dummy for years after 1956 Dummy for years after 1958
0.008 (10.40)
0.18 (14.78) -1310.1 (-4.56)
Dummy for years after 1965
-0.14 (-4.52)
Dummy for years after 1973
0.17 (33.20)
Dummy for years after 1980
0.98 (10.79)
Interaction
term (Time * Ds1952)
Interaction
term (Time * D>1956)
R-Square Mean of endog. var. Std. err. of est.
TV share
0.15 (1.95)
Real prime rate of interest
Interaction term (Time * D>1958) Interaction term (Time * D>1965) Interaction term (Time * D>1973) Interaction term (Time * D>1980) Constant
197
of Economic Behavior & Org. 32 (1997) 185-205
0.006 (26.49) -0.009 (-10.93) 50.64 13.18 0.005 -0.005 (4.37) (-34.99)
-818.11 (-1.76) 0.99 1624.5 155.87
-0.04 (-5.13) 0.49 0.016 0.03
263.98 (35.05) 0.84 368.25 25.21
a Variable definitions
and sources are provided in the text.
3.3. Superfluous
capacity as an entry barrier
-0.02 (-10.45) -0.01 (-0.26) 0.80 0.10 0.02
0.14 (84.56) 0.97 0.1176 0.0025
-0.04 (-4.08) 0.87 0.09 0.02
The regression results in columns 4 and 5 of Table 3 illustrate an economic theory at work. Successful artists and songwriters exhibit a behavior that is equivalent to a credible scheme of limit pricing - they sing and write longer songs than less successful artists. This strategy is an impediment to entry into the popular music business. If a number one
198
100)
UM
train,
R.D. Tollison/J.
of Economic Behavior & Org. 32 (1997) 185-205
,;
I
60 45
5s
Fig. 8. Ma&t
65
concentration
75
85
top four artists, 1940-1988.
song plays 25 times a day on radio and is one minute longer than other songs, it obviously limits the air time available for competing songs. Moreover, the act of writing and singing longer songs is a form of precommitment, which deters the entry of new artists. Thus, even though potential entrants will have access to the same cost conditions as incumbent artists (studios, instruments, and the like), they will not find it as profitable to enter the business because they face the prospects of obtaining less radio exposure for their songs, Longer songs are thus a form of strategic behavior in this market by dominant ‘firms.’ Dominant artists tend to produce longer songs in order to occupy more air time and to preempt lesser known competitors from the airwaves. This behavior is roughly equivalent to superfluous capacity arguments in the industrial organization literature. 3.4. Demand-side
equations
We specify three regressions to capture demand-side market forces that impact on the characteristics of successful songs. In column 1 of Table 2, the average speed of songs (BPM) is positively related to proxies for the rate of time preference among consumers. The higher the probability of death in war and the higher the MISERY INDEX, the faster are songs. TV SHARE, which measures the extent of a substitute for listening to pop music on the radio, enters with a strong positive sign in the SPEED results. All told,
WM. Crain, R.D. Tollison/J.
of Economic
Behavior & Org. 32 (1997) 185-205
199
SPEED is driven by the rate of time preference and the availability of substitutes, and we can explain almost half of the overall variation in SPEED with this model. LENGTH is modelled in a slightly more complex fashion. The PRIME RATE and DEATHS IN WAR measure the rate of time preference. Both enter negatively with respect to LENGTH. The interpretation of these results is familiar. Song length can be thought of as a capital good, and as interest rates rise, the economic life of capital goods recedes. Just as the forest is harvested sooner and the wine is aged for fewer years, the duration of songs declines when interest rates rise. EARNINGS presents some evidence that LENGTH is a normal good, rising with increases in earnings (although not very elastically). LENGTH is negatively related to advertising expenditures, an indicator of the opportunity cost of radio air time. As advertising expenditures increase, the average number one song is shorter. FM STATIONS proxies a technological force. When FM stations came on line in the mid 1960s they occupied a small and not very popular niche in radio markets. Consequently, the phenomenon of FM ‘album’ stations occurred, where FM stations would play whole albums without commercial interruptions. This marketing strategy by FM stations resulted in longer songs. Finally, the SHARE OF TOP FOUR ARTISTS in the prior year is positively related to LENGTH. More concentration begets longer songs for the reasons discussed above. Major artists produce longer songs both to gain extended exposure over the radio airwaves and to preempt the playtime available to lesser known artists. This model explains 65 percent of the variation in LENGTH over the 1940-88 sample period. The AVERAGE WEEKS AT #l regression illustrates a second dimension of the time preference for music. The estimated effects of PRIME RATE and WAR DEATHS, proxies for rates of time preference, are both inversely related to AVERAGE WEEKS. As time preferences shorten, number one songs remain on the charts for fewer weeks. As such, more uncertainty and discounting of the future leads to more volatility in the market for pop music. TV SHARE again enters as a substitute - as TV stations prospered relative to radio stations, the attention span of listeners diminished, and songs remained at number one for fewer weeks. The model explains 22 percent of the variation in the average length of time songs remained at number one.’ 3.5. An evaluation
of the results
Elasticities for the various models are presented in Table 4. The interpretative framework here is the familiar one - other variables held constant, how responsive are the various dependent variables to percentage changes in the explanatory variables? Overall, the elasticity results seem quite reasonable, and most of the estimated elasticities are relatively small. The length of pop music, for example, responds to the prime rate quite inelastically (-0.058), as does average weeks at number one (-0.041). BPM also exhibits a low elasticity with respect to war deaths and the misery index (0.002 and 0.07, respectively). The elasticity results in the concentration regressions are larger in magnitude. For example, a 1 percent increase in length results in a 4.127% increase in *We do not address the issue of relative price in these demand-side regressions for a simple reason - price is constant across songs in each year. Album and tape prices are the same across albums and tapes, so that the relevant competition in this market takes place along other margins.
200
WM. Grain, R.D. TollisodJ.
of Economic Behavior & Org. 32 (1997) 185-205
Table 4 Elasticities a
Right hand variables
Avg. meter number one songs (BMP’S)
Avg. length number one songs
Avg. weeks at number one
Share top 4 songs in current year
Share top 4 artists in prior year
(1)
(2)
(3)
(4)
(5)
Avg. length number one songs
4.127
Avg. weeks at number one
0.414
Share top four artists in prior year FM radio stations Radio Advertising Real expenditure Real prime interest rate Growth rate in military deaths Real weekly earnings Misery index Growth rate in teenage pop. share TV share a Variable definitions
0.279
0.339
- 1.874
0.579 -0.387
0.002
-0.058
-0.041
-0.004
-0.008
0.469 0.070 -0.021 -0.303 and sources are provided in the text.
the concentration of the top four songs. This result is another way of looking at the crowding out result discussed earlier. Generally, the elasticities fall within a reasonable range.
4. Market forces, innovations,
and entrepreneurship
The system of supply and demand relationships estimated above does a good job of explaining fluctuations in musical styles as reflected in changes in the length and BPM of number one songs. The predicted values from the model are plotted along with the actual values for length in Fig. 9 and for BPM in Fig. 10. (By inspection, it is apparent that the model does a better job of predicting length than it does of predicting BPM.) How well do the external market forces which are accounted for in this model explain the three major musical epochs which were identified in Section 2? A simple way to evaluate this question is to compare the means for the predicted values within each period to the means of the actual values. The results for the length variable are shown in Fig. 11 and for BPM in Fig. 12. For length the mean of the predicted value is 1 percent below the actual mean in the 1941-1955 period and 2 percent below the actual mean in the 1965 1988 period. The difference is 9 percent in the 1956-1964 period. This is a rough indication that musical entrepreneurship (the unexplained portion) was considerably more important in the period that begot rock and roll music than it was in the subsequent
WM. Gain, R.D. Tollison/J. of Economic Behavior & Org. 32 (1997) 185-205 320-
I
201 1
~~~_,------------.---------_----------_-----.._.---_-..-.------------.---.---..-----.--.----.---.-.-~
Fig. 9. Model predictions
for average length.
change in the structure of music in the mid 1960s. The significant transformation of music that marked the end of the rock and roll era is more easily accounted for by externally generated factors. The comparison of the predicted and actual means for the BPM data (Fig. 12) suggests a similar interpretation. The market model provides a weaker account of the rock and roll period leaving a greater residual to be explained by entrepreneurship, than it does for the pre- and post-rock periods. Coase (1979) has argued that the rock and roll revolution in pop music was facilitated by payola, which is an undisclosed payment or other inducement to a radio disc jockey by a record company to obtain more playtime for a song. His argument was that payola represented an efficient means of promoting new songs so that its later regulation by the Federal Communications Commission distorted the supply of new pop music in a manner that was consistent with entrenched interests in the record business. Coase (1979, pp.3 143 15), observes: In the 195Os, particularly from 1955 on, “rock and roll” music became extremely popular. Many new record companies were formed, mainly concentrating on the new music. The effect on the market shares of existing companies was dramatic. In the years 1948 through 1955, four companies (Capitol, Columbia, Decca, and RCAVictor) had, on an average, 78% of the records which were ever on Billboard’s top ten Hit Parade, and the figure was never less than 71% (in 1953). In 1956, the share of the
202
WM. Chin,
of Economic Behavior & Org. 32 (1997) 185-205
R.D. TollisodJ.
150.
~40_......................... ..............'1........................................................~
0) ~~~........................... ....................................................................... I (I
2 C
y.......................................................................... I-
I
Llu’,, 1941
,,,,(,,,,,,,,,,,,,,,,
,,(,,,/
1946
1951
1956
1961
rl
,,,,,,,,,,,,,,,
' 1966
197i
1976
Pericd I
Acrila1 -
Fig. 10. Model predictions
?EalCCea of average BPM.
1956-1964
’
Period
Fig. 11. Predicted and actual average length of no. 1 songs.
1981
1986
WM. C&n,
R.D. Tollison/.I. of Economic Behavior & Org. 32 (1997) 185-205
Fig. 12. Predicted
203
and actual BPM of no.1 songs.
hit records of these four companies was 66%, in 1957, 40%, in 1958, 36%, and in 1959, 34%. .. . That payola in the late 1950s was used in the main to promote the playing of rock and roll and similar music is true. Indeed, as early as 195 1, it had been reported in Billboard that “By universal agreement in the music trade the payola situation is at its worst among the rhythm and blues spinners.” There can be no doubt that the new companies, which entered the business in the 1950s and succeeded in securing such an important share of the record market relied on payola to obtain “exposure” for their records. Coase’s analysis of payola is broadly consistent with our analysis, but it is useful to see payola as a ceteris paribus effect. Our results suggest that basic economic forces of supply and demand determined the characteristics of the new music and that the market concentration of songs and artists is also explicable by such forces. In a sense, payola facilitated the emergence of faster, shorter music, but such music was invented according to underlying determinants in the economy. We have not analyzed our concentration time series for possible ‘events’ related to payola and its investigation and regulation. Nonetheless, concentration of artists and songs did decrease in the 1950s on average, and it has continued to decline since that time. The latter observation suggests that the regulation of payola may be ineffective, given that concentration in the pop music industry has continued to be on the low side.
204
WM. Gain,
R.D. Tollison/J.
of Economic Behavior & Org. 32 (1997) 185-205
5. Conclusions Overall, there is ample evidence that economic forces influence the characteristics of pop music and the competition to be a successful artist in this market. Market structure and technological factors and the forces of demand clearly influence musical characteristics in predictable ways. Moreover, the order of magnitude of the estimated effects seems reasonable. While we have certainly not explained all the changes in music in our data set, we have explained enough to say that economics and music are intimately related. And, as stressed above, we have been able to identify periods in which entrepreneurship played a relatively greater role in generating changes in the structure of music. We have focused in this paper on the impact of economic forces on characteristics of popular music. Armed with these results, ideas await analysis. In a more conventional spirit, we have data on writers and performers and whether, in fact, they are vertically integrated stages of production (i.e. one and the same person). Incentives matter, so that when performers compose their own material, songs are likely to be longer and remain at number one for a longer time. Probing the boundaries of the analysis further, pop music is often alleged to encourage anti-social behavior and undermine cultural values (Bloom, 1987). With our data set this is clearly a testable proposition. Are, for example, musical characteristics related to social problems such as teenage pregnancy or drug use? Moreover, we think the problem is more complex than this. Perhaps pop music has favorable effects on culture, leading, for example, to more respect for individual freedom, responsibility, and initiative. Thus, it could well be that pop music has something to do with the deregulation of industry and even the fall of centrally planned economies. These ideas represent a future research agenda.
References Alpert, L., 1983, Estimating a multi-attribute model for different music styles, Journal of Cultural Economics, 7(l), 63-81. Anderson, B.. P. Hesbacker, P. Etzkonn and R. Denisoff, 1980, Hit record trends, 1940-1977, Journal of Communications, 30(2), 3 1-43. Baker, A.J., 1991, A mode1 of competition and monopoly in the record industry, Journal of Cultural Economics, 15(l), 29-54. Becker, Gary, 1971, Economic theory, (A.A. Knopf, New York). Belinfante, A. and R.R. Davis, 1978-79, Estimating the demand for record albums, Review of Business and Economic Research, 47-53. Belinfante, A. and R.L. Johnson, 1982, Competition, pricing and concentration in the U.S. recorded music industry, Journal of Cultural Economics, 6(2), 1 l-25. Bloom, Allan David, 1987, The Closing of the American Mind, (Simon and Shuster, New York). Clyne, Manfred, 1982, Music, Mind, and Brain, (Plenum Press, New York). Coase, Ronald, 1979, Payola in radio and television broadcasting, Journal of Law and Economics, 22(2), 269328. Levy, David and Susan Feigenbaum, 1986, Death, debt, and democracy, In: J.M. Buchanan et al.. Eds., Deficit, (Blackwell, Oxford) pp.236-262. Peterson, R.A. and D.G. Berger, 1975, Cycles, in symbol production: The case of popular music, American Sociological Review, 40, 158-173.
U?M. Grain, R.D. To&on/J.
of Economic Behavior & Org. 32 (1997) 185-205
20.5
Shanahan, J.L., 1978, The consumption of music: Integrating aesthetics and economics, Journal of Cultural Economics, 2(2), 13-26. Whitbum, Joel, 1986, Pop memories, 1890-1954, The History of American Popular Music, (Record Research Inc., Menomonee Falls). Billboard Magazine, various issues (BP1 Communications, New York). U.S. Government, 1992, Economic Report of the President, (U.S. Government Printing Office, Washington, DC). U.S. Government, various editions, Statistical Abstract, (US. Government Printing Office, Washington, DC).