In the past few years, a number of digital and media platforms, e.g., YouTube and Spotify, have begun introducing premium services which let paid subscribers see fewer or no adverts. Such a practice seems to be a classic case of second-degree price discrimination through ‘versioning’, which allows consumers to self-select different models of the same product based on their preferences. In other words, consumers who value media content highly or who are averse to ads (higher-type consumers) can choose to pay for a premium version with fewer or no ads, whereas others can choose to remain with the basic version which contains more ads (lower-type consumers). However, offering a premium service may, in fact, reduce advertising revenue, resulting in a trade-off for media platforms. If too many consumers opt for the premium version, the loss of ad revenue can be significant. This, added to the practice of many media publishers setting up paywalls, where consumers pay for premium content associated with more ads, makes the situation become even more intriguing.

A recent study by Song Lin aimed to investigate how ad-financed media platforms should price discriminate in two-sided markets. To do so, he integrated the two-sided market model and price discrimination model, and explored how platforms can implement price discrimination on both sides of the market. It is worth noting that media providers, unlike traditional sellers, not only sell directly to customers, they also act as two-sided platforms. By attracting consumers with free or low-priced content, they can sell consumer attention to advertisers. Hence why media platforms must design pricing policies that balance the incentives for both sides.

After undertaking baseline analysis, with the standard assumption that agents on one side of the market only care about the number of agents they interact with on the other, the paper then explored what happens to media platforms’ policies when consumers care about both the quantity and quality of the advertisers with whom they interact. One of the study’s key findings was a self-reinforcing mechanism, whereby price discrimination on one side can strengthen the incentive for discrimination on the other. As the author explains, “Consumers are screened based on their heterogenous product preferences, and advertisers are screened according to their heterogenous preferences for reaching different consumer segments”. For example, by offering the premium version in addition to the base version, Spotify essentially segments the consumers into two groups, with one group seeing less or no ads while the other seeing all the ads. Then advertisers can choose to reach either one or both consumer groups, depending on their marketing needs. This leads to a segmentation of advertisers and thus different ad exposures associated with the two Spotify versions, which in turn forms the basis for separating the consumer groups.

Interestingly, following the model prediction, it is generally optimal for media platforms to keep limited amount of ads instead of completely removing ads from the premium version. This explains a recent consumer outcry in the China market where many video streaming platforms (e.g., iQiyi, Tencent) still show ads to premium subscribers.

In light of this self-reinforcing mechanism, the research showed that ad allocations across different consumer types depend on how much of a nuisance an ad ‘costs’ consumers, relative to the value it brings them. “Interestingly, higher-type consumers may see more ads than lower-type consumers if the nuisance cost is relatively low”, says Song Lin. The study also revealed that the standard downward quality distortion, in general, fails to materialise in a two-sided market and may even be reversed. Specifically, higher-type consumers may find themselves exposed to too few ads, resulting in a lower total quality than the socially-efficient level, whereas lower-type consumers may receive a socially-excessive quality. The study went on to explore a number of important market features that shape media platforms’ product policies and which may weaken the self-reinforcing mechanism; for example, dependency of advertiser preferences on consumer type and the informational friction between advertisers and consumers.

Having developed the two-sided media model in this study to allow for ease of analysis, the author pointed out a few potential avenues of future research. First, it might be interesting to extend the abstract model of the product market to incorporate richer market structures. Second, the study focused on the incentives of media platforms in a monopoly market. Future research could therefore look to include competition among media. Third, the model assumes that a media platform can post no more than two ad rates in equilibrium to permit self-selection by heterogenous advertisers. As such, it might be worth asking what happens if platforms could extract more surplus from advertisers by selling ad space through auctions, ad networks, or agencies. Finally, examining equilibrium advertising policies in an empirical context might provide additional insights into how they are shaped.