“Advertising and Content Differentiation: Evidence from YouTube”, 2019-09-04 (; similar):
Does advertising revenue increase or diminish content differentiation in media markets? This paper shows that an increase in the technically feasible number of ad breaks per video leads to an increase in content differentiation between several thousand YouTube channels. I exploit two institutional features of YouTube’s monetization policy to identify the causal effect of advertising on the YouTubers’ content choice. The analysis of around one million YouTube videos shows that advertising leads to a twenty percentage point reduction in the YouTubers’ probability to duplicate popular content, ie. content in high demand by the audience. I also provide evidence of the economic mechanism behind the result: popular content is covered by many competing YouTubers; hence, viewers who perceive advertising as a nuisance could easily switch to a competitor if a YouTuber increased her number of ad-breaks per video. This is less likely, however, when the YouTuber differentiates her content from her competitors.
[Keywords: advertising, content differentiation, economics of digitization, horizontal product differentiation, long tail, media diversity, user-generated content, YouTube]
…The analysis of around one million YouTube videos shows that an increase in the feasible number of ad breaks per video leads to a twenty percentage point reduction in the YouTubers’ probability to duplicate popular content. The effect size is considerable: it corresponds to around 40% of a standard deviation in the dependent variable and to around 50% of its baseline value.
The large sample size allows me to conduct several sub-group analyses to study effect heterogeneity. I find that the positive effect of advertising on content differentiation is driven by the YouTubers who have at least 1,000 subscribers, ie. the YouTubers whose additional ad revenue is likely to exceed the costs from adapt-ing their videos’ content. In addition, I find heterogeneity along video categories: some categories are more flexible in terms of their typical video duration than others, hence, exploiting the ten minutes trick is more easy (eg. a music clip is typically between three and five minutes long and cannot be easily extended). A battery of robustness checks confirms these results.
…Moreover, I show that ad revenue does not necessarily improve the YouTubers’ video quality. Although the number of views goes up when a video has more ad breaks, the relative number of likes decreases…Table 5 shows the results. The size of the estimates for δ′′(columns 1 to 3), though statistically-significant at the 1%-level, is negligible: a one second increase in video duration corresponds to a 0.0001 percentage point increase in the fraction of likes. The estimates for δ′′′ in columns 4 to 6, though, are relatively large and statistically-significant at the 1%-level, too. According to these estimates, one further second in video duration leads on average to about 1.5% more views. These estimates may reflect the algorithmic drift discussed in §9.2. YouTube wants to keep its viewers as long as possible on the platform to show as many ads as possible to them. As a result, longer videos get higher rankings and are watched more often.