In the past few years several studies on the impact of P2P music file sharing on recorded music sales were published. They came to very different and even conflicting results, as I highlighted in a 25 part blog series. A recently published study now shifts the focus from file sharing to music video online streaming. R. Scott Hiller of Fairfield University and Jin-Hyuk Kim of University of Colorado Boulder analysed the sales displacement effect of YouTube in a paper entitled “Online Music, Sales Displacement, and Internet Search: Evidence from YouTube“. They concluded that Warner Music Group sold significantly more units of its Billboard 200 albums, when the Warner content was removed from YouTube due to a conflict on licensing fees. In addition, they found no evidence that the blackout had a negative promotional effect for Warner artists.
You can read more about this study and my assessment of the results here:
When Google bought YouTube for US$ 1.65bn at the end of 2006, the new owner entered into a revenue sharing partnership with the then four music conglomerates. However, the record majors were not satisfied with the modest payment from every video streamed and the share in advertising revenue. Thus, they demanded a renewal of the licensing agreement. At the end of 2008, all majors except Warner Music came to an acceptable deal with Google/YouTube. In the following, Warner Music Group (WMG) withdraw not only its original music videos from YouTube but also amateur material that included Warner content. This resulted in a nine-month blackout for Warner music videos on YouTube. On September 29, Google and WMG announced that they eventually reached a new licensing deal.
Data and research method
The authors considered the Warner blackout on YouTube as a natural shock that enables a comparison of album unit sales with Warner’s competitors. The sample is based on Nielsen SoundScan data to compile the weekly U.S.-Billboard 200 album charts, which include physical retail and digital sales. Since Hiller & Kim did not use a representative random sample the results cannot be generalized as they point out on page 9. In total, they obtained a sample of 2.261 albums from 1.663 artists for the period of January to September 2009. They dropped about 10 to 20 non-musical and compilation albums from every week to come to the final sample.
In a next step, the number of albums sold in a given week was defined as depend on:
(1) the “Warner” blackout-effect
(2) the cumulative number of weeks an album stayed in the charts
(3) the number of albums sold in the first week after the initial album release
(4) an indicator for an artist’s debut album
(5) an indicator for an artist’s previous albums
(6) the length of an album stayed in the Billboard 200 charts
(7) the total sales for the last album an artist placed on the Top 200 to make spillover-effects of earlier albums visible
All albums were matched with genre and label information with the Discogs.com database. Eventually the sales information was merged with the weekly USA Airplay Top 200 charts to assume possible promotion effects of radio airplay on record sales. The authors also introduced and tested as second depend variable a complex index of search intensity on Google.com for Warner artists to assess the impact of YouTube’s promotional effect for Warner content.
In the following, all weekly unit sales of Warner albums in the Top-200 were compared with those of other labels before, during and after the blackout period. The authors varied, therefore, the reference periods with 1 year, 9 months and 3 months before and after the blackout respectively.
Hiller & Kim tested the hypothesis whether the Warner blackout on YouTube had either a positive or negative effect on Warner’s album sales. If the Warner artists could sell relatively more album units in the blackout-period than the competitors this would indicate a sales displacement. In addition, they also tested the impact of the blackout on the search intensity for Warner artists.
The regressions run for four periods: 1 year, 9 months, 6 months and 3 months before and after the blackout. The tables 2-5 highlight the regression results for the impact of the “Warner effect” on sales and tables 6-9 on search intensity respectively. The results show “(…) that the Warner artists who had Billboard top 200 albums during the nine-month blackout sold on average larger quantities of albums, as opposed to the non-Warner artists during the same period. (…) [T]he increase in sales (in unit) ranges from 5,718 per week using one-year (pre and post) window to nearly 10,000 per week using a three-month window” (p. 14). By eliminating the top-10 albums in each week, the “Warner effects” weakens to a weekly increase of 2,458 to 3,678 per album. It becomes even smaller by eliminating the top-25 albums – plus 1,419 to 1,870 per album – and disappears if the top-50 albums were eliminated from the data sample. The authors concluded that the blackout had a very positive impact on album sales of Warner’s superstar artists, whereas non-superstars did not benefit. This does not mean, however, that YouTube has a positive sampling-effect on non-superstars, since the regression results are neither negative nor significant.
In addition, the authors could not find a statistically significant effect on the level of Internet user’s search activity. “Therefore, the hypothesis that a greater exposure to free online content increases the user’s level of interests finds no support in our data” (p. 16).
To sum up the results:
(1) The availability of free music videos on YouTube has a sales displacement effect for albums. Thus, YouTube cannibalizes physical and download music sales to a certain degree.
(2) Although the authors limit their findings to albums in the data set, music streaming in general could have a sales displacement effect. Studies, therefore, have to be conducted.
(3) YouTube has neither a sampling nor a promotion effect for superstar and non-superstar artists.
What are the main results of the study? The authors show that the unit sales of Warner albums in the Billboard 200 list were relatively higher on average than that of non-Warner artists in the first three quarters of 2009 compared to 2008 and 2010. They reveal a correlation of two variables and explain by the blackout of Warner albums on YouTube in the same period. This does not constitute, however, a causal relationship between the blackout and increased album sales. Other reasons for the correlation could also be possible, but they were not considered in the paper.
In consulting the annual report of the Warner Music Group for 2009, we can find a different explanation for the temporary sales increase. The then WMG CEO, Edgar J. Bronfman, pointed out: “According to SoundScan data, we reported the highest U.S. album market share growth of all major music companies in total albums sold between calendar 2004 and calendar 2008” (WMG 2010: 3). The growth is, therefore, not explained by the unavailability of Warner music videos on YouTube, but by Warner’s digital music strategy that was fully implemented in 2008. The 2009 figures reflected the success of Warner’s digital strategy. The digital music sales of WMG increased by 10.7 percent in 2009 after a massive growth of more than a third from 2007 to 2008. In 2010, the sales growth dropped to 1.1 percent reflecting an end of the boom. The digital strategy was particularly successful in the U.S. After the share of digital sales in the total U.S.-sales rose to 29.9 percent in 2008, it jumped to 38.8 percent in 2009. The further growth to 44.3 percent in 2010 was comparable modest.
The success of Warner’s digital strategy becomes even more visible in comparison with market leader Universal Music Group (UMG). Although UMG’s digital music sales increased by 30.5 percent from 2007 to 2008, the growth lagged behind WMG with 7.8 percent in 2009. However, in the following year UMG closed the gap to Warner with a massive increase by 13.8 percent. Compared to WMG, UMG’s share of digital music sales was modest with 25.7 percent in 2009 (Vivendi 2010: 150). Warner’s annual report also highlighted that sub-label Atlantic Records crossed the 50 percent mark of digital music sales for the first time in 2009 (WMG 2010: II).
These facts lead to a different conclusion: WMG’s relatively higher albums sales in the first three quarters of 2009 were not caused by the YouTube-blackout, but by the successful implementation of Warner’s digital music strategy in 2008. As the sales figures show, this relative advantage disappeared in 2010, when the other majors followed suit. The weakening “Warner-effect”, when hit albums are eliminated from the sample, can be explained by Warner’s digital strategy too. Since the majors particularly support their superstar acts with a lot of marketing money, superstar albums benefit more from Warner’s digital strategy than non-superstar albums.
Although the authors operate an elaborated statistical apparatus, this simple reasoning better explains Warner relative album sales’ success in 2009. One have to ask, why Warner’s officers, who knew the exact unit sales figures of every album, renewed the licensing agreement with YouTube, despite a massive sales displacement effect that had to be reflected in the data.
YouTube, therefore, is not that bad as suggested by Hiller & Kim’s study. WMG as well as all other majors and most of the indies believe in YouTube’s positive promotion and sampling effects. In contrast, they invest a lot of money to expose their music videos on YouTube. No label can afford nowadays not to be available on YouTube – it is a must. Therefore, the authors of the study should rethink their research design and should consider other possible explanations for the “Warner-effect”.
Hiller, R. Scott and Kim, Jin-Hyuk, 2014, Online Music, Sales Displacement, and Internet Search: Evidence from YouTube. CAPRI Publication 13-2.
Vivendi, 2010, Annual report of Vivendi 2009. New York: Vivendi U.S.