After several revisions (Hong 2004, 2005, 2007), Hong published in July 2011 a working paper entitled “Measuring the Effect of Napster on Recorded Music Sales”, in which he tried to measure the effect of file sharing on recorded music sales. Since he did not directly observe file sharing activity, the author compared a treatment group of Internet users with a control group of non-Internet users before and after the advent of Napster in 1999 and attempted to eliminate the time effect and isolate the so-called “Napster-effect”.
The primary sources of data used by Hong were interviews of the Consumer Expenditure Survey (CEX) by the U.S. Bureau of Labor Statistics. This dataset also included various expenditures on recorded music and Internet service fees. Therefore, Hong defined recorded music expenditures as the sum of expenses pertaining to CDs, tapes, and vinyl LPs purchased in the time span from 1996 to 2002. Furthermore, the Internet user group was defined as households that either spent positive amounts on computer information services or lived in college dormitories, since most of the students there had already broadband access in the late 1990s.
Based on this data framework, Hong estimated in a regression model how recorded music expenditure differed between households with and without Internet access before and after the advent of Napster in May/June 1999.
What sounds like a simple question turned out to be a very complicated task, since the treatment group and the control groups were not the same before and after the emergence of file sharing. I.e. early adopters of the Internet had very different characteristics than Internet users after the emergence of Napster. Beyond that, post-Napster Internet users might have adopted the Internet just to download free music. This might exacerbate a potential negative effect bias due to compositional changes. Since Hong was aware of this problem, he had to apply very complex statistical methods to overcome this bias, which cannot be discussed here in full length.
In short, instead of using one-dimensional propensity scores, he had to apply a two-dimensional method for identification under compositional changes. Then he developed a nonparametric difference-in-difference matching method (DDM) in order to estimate probit models of Internet access separatly for the pre-Napster period and for the post-Napster period. Based on the two propensity scores, he matched each post-Napster Internet user with Internet non-users and pre-Napster Internet users to construct the counterfactual.
The results from this estimate indicated “(…) that the average Internet user during the Napster period would have spent $1.45 more per quarter on recorded music in the absence of Napster” (Hong 2009: 21). Hong calculated on this basis that the “(…) decrease in total record sales from the pre-Napster period to the post-Napster periods amounts to $832.24 million. This suggests that 39.6% of sales decline could be attributable to the presence of Napster” (Hong 2009: 21). However, in a detailed analysis of different age groups of the CEX sample, only the estimations for households with children aged 6-17 delivered a significant result. Thus, the negative Napster-effect for this age group was $3.26 per household and per quarter, which can be translated into a loss of US$ 196 million accounting for about 20% of total record sales decline during the Napster period. The households for the age group 15-34 would account for another 20% of total sales, but with a large standard error of 3.01, which makes the result not reliable enough. Beyond that, the estimation results for the age groups 35-49 and older than 49 are statistically indistiguishable from zero, i.e. they are not significant.
To sum up, despite the very complex statistical methodology the meager result of the study is that households with 6-17 years old children accounted for 20% or US$ 196 million of the decline in record sales from June 1999 until June 2001.
Hong, Seung-Hyun, 2004, The Effect of Napster on Recorded Music Sales: Evidence from the Consumer Expenditure Survey. SIEPR Discussion Paper No. 03-18. Stanford University for Economic Policy Research.
Hong, Seung-Hyun, 2005, The Effect of Digital Technology on the Sales of Copyrighted Goods: Evidence from Napster. Working paper, Department of Economics at the University of Illinois (September 2005).
Hong, Seung-Hyun, 2007, Measuring the Effect of Digital Technology on the Sales of Copyrighted Goods: Evidence from Napster. Working paper, Department of Economics at the University of Illinois (January 2007).
In the next part the paper of Tatsuo Tanaka will be reviewed, in which he “(…) did not find any negative effect of file sharing on CD sales”