Subprime Attention Crisis, Tim Hwang, 2020

Tim Hwang in Subprime Attention Crisis exposes the markets that underlie the internet. These attention markets, where our attention is traded at hyperfast speed, highly parallel financial markets. This is no accident, as the tech and finance industries have been tightly connected since the beginning of the commercialization of the internet. And indeed, many executives in tech came from the finance world.

The market for attention highly resembles more traditional commodity markets such as for oil, timber, or water. Hwang gives a very useful historical overview of the development of the market for wheat, where once farmers travelled to sell their individual yields at marketplaces in the city. Later, wheat was standardized by lot, quality, season, and other characteristics, and then turned into indistinguishable abstracted commodities. Eventually, futures and other financial derivatives developed on top of the underlying asset and became enormous markets. 

Subprime Attention Crisis lays out the argument that board swathes of the internet and wider economy are linked to and reliant on huge inflows of advertising cash to platforms that have set up programmatic markets for buying and selling the ad-placements. These ads are presented as a type of financial derivative whose underlying value comes from the attention paid to the ad by users on the internet. Both the buy-side (advertisers such as Coca Cola) and the sell-side (such as youtubers or bloggers) use apps (often served by dominant firms like Google) that provide programmatic options for setting prices and bids. Algorithms actually conduct real-time bidding at the point of loading for every website and video on the internet with ads. The computational infrastructure to execute at this volume, speed, and reliability is incredible.

But the book warns that the current system is in jeopardy due to a rising reliance on “subprime attention” undergirding this large ad marketplace.  

Hwang draws a comparison between the subprime mortgage crisis and this subprime attention crisis. To oversimplify, in the mortgage crisis, derivatives (such as MBS or CDOs) built on top of risky subprime mortgages became unmoored with the underlying value of the mortgages. Lenders were incentivized to increase volumes of the valuable derivatives and so expanded their willingness to create more and more marginal mortgage loans to satisfy the demand for derivatives. When the shoe dropped and the underlying mortgages stopped performing, the huge derivatives market crashed and took down many financial institutions. 

In Subprime Attention Crisis, we replace mortgages with internet users’ attention and replace MBS/CDO derivatives with algorithmically traded ads. Hwang proposes that the increasing demand for digital advertising creates a need to generate more of the underlying commodity, internet users’ attention. But this attention is increasingly low-value compared to the prices of the ads. More users are turning to adblock and never see the purchased ads. Younger cohorts when studied appear to pay less attention to ads online than older ones and in fact click them less. Video and other more interactive online media although growing in terms of user time spent are not yet as effective in terms of advertising revenues. And click-farms and other ad fraud is rampant, meaning much of the money spent and much of the performance statistics is incorrect or wasted. As such, Hwang suggests that more subprime attention is sold under “AAA” ad space, creating the conditions for the type of crash that happened to mortgages in 2008. 

Hwang speculates a bit on what an internet post-ad crash might look like. Most of the free services like search, email, or maps would retreat behind paywalls, locking the poor out of basic internet services. Activities such as machine learning or autonomous vehicle research that are conducted by entities within conglomerates like Alphabet, all of which are highly subsidized by the advertising revenues of these conglomerates' main-line ad business, would be severely cut or discontinued. Much of the media production, both from amateurs as on YouTube or from professionals such as journalists at the New York Times, rely on advertising revenue and may see their business models further deteriorated to insolvency.

Hwang also points out the level to which the choice of advertising as the primary revenue engine for the internet has in fact thoroughly determined the form and function of the internet. In other words, the realities and norms of the internet today are those that make our attention most easily commodified and sold. In a feedback loop, the success of advertising as a revenue model for the internet continues to tailor the internet to become better suited to supplying advertising. In the book, this is called the “legibility.” 

The concept of “surveillance capitalism” reported in the media, encompassing the tracking of user’s identities, interests, locations, and more, was instituted as tools for the marketplaces like Google to offer more personalized advertising. Social media features such as likes, impressions, or views were introduced to make the measurement of advertising effectiveness easier and to complement the tracking just mentioned. Even the basic idea of user profiles exists so that users can more easily be tracked across time and both serve “better” ads and have their responses to those ads be better measured.

This isn’t the first case where such cycles have presented themselves. One poignant example; governments historically introduced formal identities, including introducing surnames onto populations that had never before featured that cultural norm, to better facilitate taxation. 

Hwang’s argument on its own is strong but makes use of a click-baity and stretch analogy of the attention markets to the mortgage-backed security markets and warns of a 2008-style market crash. There are some surface-level similarities, but the choice of analogy is clearly made to present a familiar hook to readers, many of whom are likely not aware of much in the financial world aside from a general sense of what occurred in the 2008 market crash. Staying away from the constant return to the mortgage crisis probably would make this book better.

The author also does not adequately go into the social harms of baked-in incentives in this system to addict internet users. While Hwang discusses how subprime attention is peddled more and more, he doesn’t discuss how attention is manufactured by the platforms by addicting users so that they spend more time scrolling on Instagram or Tik Tok. Another possible source of crash could be a mental health based rejection of these habits. I for one see it as a possibility that scenes on 2021 where people walk around outside staring at their phones will look as unhealthy and foreign to future viewers as scenes of everyone chain-smoking cigarettes looks to us now.