Subprime Attention Crisis

by Tim Hwang

  Buyers run into their own problems. Buyers might never have met our hypothetical chicken farmer. They may be far away from our island and unable to get there to evaluate the quality of the chickens. Unlike the local butcher or sandwich shop or chicken fancier, whose proximity to the farmer lets them purchase chickens as needed, the new buyers may want to buy a large number of chickens at once from a single seller. The buyer may not be able to efficiently acquire that number through a piecemeal process of buying a few chickens each from a large number of different farmers.

  Standardization—the first step toward commodification—allows the market to accommodate a larger number of players. Arguably, the greatest friction in this marketplace is that buyers and sellers cannot easily assess what they are buying. It is challenging to quickly determine at a distance whether a chicken is high quality or low quality, particularly where buyers and sellers are strangers and where there is low trust between the various players in the marketplace.

  Say that we bring together the buyers and sellers of chickens, and they agree on common definitions about what is traded in the marketplace. Rather than selling different types of chickens in nonstandard quantities, they specify something that we’ll call a Standard Chicken Lot (SCL). The SCL is a fixed unit of chickens having certain qualities: the chickens might need to be above a certain weight, to not have any diseases, and so on. They hire some inspectors to go around to places where chickens are sold to make sure that people are conforming to this standard. On their own, chickens are a heterogeneous thing. Some are tall and some are short. Some are thin and some are heavy. Some are of one breed and others are of another. The SCL ensures that the chickens traded in our market are of a commonly understood quality and quantity.

  This may not sound like a whole lot, but it changes everything. Suddenly our chicken marketplace is capable of enabling large numbers of buyers and sellers to quickly transact chickens at arm’s length by simply knowing how many SCLs are on offer and for how much.

  Standardization does not just allow the market to expand effectively. It also produces the second step in the process of commodification—abstraction. SCLs are abstract—disconnected from the process of production and the unique identity of the producer. Because the units of chicken being bought and sold are now standardized, a prospective buyer no longer needs to know the specific characteristics of a farmer to move ahead on a transaction. SCLs are interchangeable: an SCL produced by our island chicken farmer is the same as another SCL produced in a different place.

  Abstraction enables arrangements that might never otherwise emerge in our original island chicken marketplace. This kind of abstraction facilitates speculation. Groups with little interest in raising chickens or consuming chickens might choose to engage in buying and selling SCLs to make money from fluctuations in price. One could buy the rights to an SCL in the marketplace and sell the rights at some later date, without ever seeing or moving the actual chickens. The trade in SCLs—which are abstract assets—can take place far away from where the chickens are hatched and where they end up.

  This is more than a just-so story; the dynamics of commodification are very real across many different kinds of markets. Perhaps one of the most dramatic historical examples, as documented by William Cronon in his book Nature’s Metropolis, is the emergence of large-scale grain markets in Chicago during the middle of the nineteenth century.30

  Before 1850, grain was bought and sold in large, open-air marketplaces near the waterfront of Chicago.31 These marketplaces transacted grain in very much the same way as our island chicken farmer sells chickens. Grain was sent by the sackful from a farm to a merchant, who would haggle face-to-face with buyers in an effort to obtain the best price. The merchant acted as a middleman for the farmer, who retained ownership over his grain and paid the merchant a commission for each sale.

  As Cronon explains, the rise of the railroads transformed this mode of exchange and “transmute[d] wheat and corn into monetary abstractions.”32 Railroads allowed crops to be efficiently transported from outlying farms into Chicago, rapidly increasing the amount of grain that entered the city’s market. When it became clear that bulk grain was more efficiently sold at market, traditional grain sacks were abandoned and farmers pooled their crops into freight cars.33 But combining grain from different farms raised the question of how to deal with the ownership of the grain that each farmer contributed to a given carload.

  A private industry consortium, the Chicago Board of Trade (CBOT), eventually solved the problem through standardization. The CBOT designated three categories of grain and four levels of quality (“Club,” “No. 1,” “No. 2,” and “Rejected”).34 Farmers putting grain into a train car received a receipt indicating a quantity of grain and a quality level. The receipt was redeemable for an equal quantity of the same quality grain—not the same grain, but its functional equivalent.

  Once standardized, grain became abstracted into a commodity. The receipts could be bought and sold without regard to the specific identity of the farmer who originally produced the grain. People with no interest in grain production could make a profit by buying and selling receipts. Famously, the CBOT also facilitated the rise of a vigorous trade in “futures,” speculative contracts betting on the future price of grain. According to Cronon, grain came to be seen “as a commodity, not as a living organism planted and harvested by farmers as a crop for people to mill into flour, bake into bread, and eat.”35

  Another outcome of grain commodification was an explosion in the size and importance of the market itself. Chicago commodified faster than other grain markets, outstripping its longtime rival St. Louis in 1855.36 Before commodification, grain produced in the northern United States was sold to markets in southern states. After commodification, the trade reversed direction. “No place was more important than Chicago to this redirection of agricultural trade,” Cronon writes.37 The CBOT remained the world’s dominant commodities market and futures exchange well into the twentieth century.38

  The economics of nineteenth-century grain may seem far removed from the online advertising markets of the twenty-first century. But the huge distances of time mask deep parallels. Attention—a varied, heterogeneous, hard-to-define thing—underwent a similar process of commodification over the course of the 2000s, culminating in the programmatic marketplace.

  The Commodification of Attention

  In many respects, “attention” is completely different from grain and chickens. After all, chickens and grain are real, tangible objects. Attention is bafflingly abstract. It exists for a moment and then drifts somewhere else. Chickens have certain clear, measurable attributes. They can be old or young, sick or healthy, egg producing or not egg producing. Attention is a bit harder to quantify and measure.

  But despite these differences, the maturation of the online marketplace for attention closely mirrors the development of our hypothetical chicken marketplace. In a predigital advertising economy, local newspapers sold space on their pages to generate revenue. Most of the time, that space was purchased by businesses wishing to advertise goods and services. Like the island chicken farmer, the local newspaper sold its space to local businesses and organizations, and negotiated these deals on a bespoke basis.

  But, as with chickens, this artisanal process for buying and selling attention is cumbersome at scale. Facebook serves an astronomically larger number of advertisements than a mid-twentieth-century local newspaper. It also has a vastly broader pool of potential advertisers, from massive companies promoting their latest product to a community group fundraising for the local marching band. Scale introduces the same pressures to the attention market that we saw in our hypothetical chicken market. In order to buy and sell attention across a vast number of players with very different needs, it becomes necessary to standardize what is being sold.

  We need a Standard Chicken Lot for our attention market. The search for this standard rapidly reaches into the realm of the philosophical. Can we measure attention in a simple a
nd reliable way? Can we break human attention down into discrete units that can be exchanged in a marketplace?

  The advertising industry is deeply invested in answering these questions in the affirmative. In 1996, industry leaders formed the Internet Advertising Bureau (later renamed the Interactive Advertising Bureau, or IAB), a private consortium dedicated to developing the standards that enable the online advertising market to function.39 In 2004, the IAB published its “Ad Impression Measurement Guidelines,” an initial set of foundational standards by which an ad is considered “delivered” to a consumer by the industry.40

  The battles over these standard definitions have been fierce, because they define the terms under which advertisers pay for online ads and how they measure their success. For instance, we might say an ad has been “delivered” when it is successfully loaded up on a website. That definition has the advantage of being extremely easy to measure: an ad server can log that it sent an ad, and the website can confirm that the ad has been displayed. But assuming this definition as the bar for success might massively overreport the amount of attention that an ad is capturing. What if the ad is loaded, but is buried in a place on the website where no one ever sees it? Tons of ads could be “successfully delivered” under this metric without reaching anyone.

  We might try to raise the bar by only counting an ad as “successfully delivered” when a user pauses on the ad and actually looks at it for a period of time. But how long do users need to pause on the ad? How can we verify that they are looking at it? Setting the threshold for success too high creates new problems—by defining the attention asset so narrowly, we risk overly constraining the market. Publishers might end up with real attention they are unable to sell because the standards for measuring and verifying it are too stringent.

  Much of the IAB’s work focuses on building consensus in answering these fundamental line-drawing questions: the organization has brought together numerous working groups that create and update guidelines specifying what constitutes a measurable blob of attention.

  Like the standards for soybeans or corn, attention standards can be impressively specific. In 2014, the group released an extension of its 2004 work that standardized the concept of a “viewable impression.” To achieve a viewable impression, more than 50 percent of the pixels in an advertisement must occupy the viewable space of a browser page for greater than or equal to one continuous second after the advertising renders.41 The IAB specifies that “satisfying the minimum pixel requirement should precede the measurement of the time duration; for example, the clock starts on determining whether the ad meets the one continuous second time requirement only when the ad is determined to have met the 50% pixel threshold.”42 Numerous other standards define other aspects of the ad delivery process, from guidelines on the placement of an ad on the page to a commonly recognized set of formats for displaying an ad.43

  The implications of this standardization effort are profound. Commodification enables a fluid marketplace. The amorphous, shapeless concept of attention has been transformed into discrete, comparable pieces that can be captured, priced, and sold. Buyers and sellers can quickly evaluate opportunities and transact in attention at massive scale, without individually evaluating each opportunity.

  Standardization has made attention an abstract, economic asset as well. It is now possible to purchase attention in the marketplaces without knowing where and how that attention was produced. The idea is to enable the capture of a desired amount of attention at a minimum cost. As with standard units like “No. 1 Spring Corn” and the SCL, the modern programmatic advertising marketplace makes attention an interchangeable asset.

  The commodification of attention in the 2000s created a fertile environment for market automation, something that was not possible in the Chicago grain markets of the 1850s. The industry definition of an impression is rigid and measurable, making impressions highly amenable to computerization. The existing infrastructure allows anyone to set up an advertiser account on an ad exchange and start buying and selling attention algorithmically with a global pool of willing counterparties. No special expertise or industry connections are required.

  The overall impact of this seamless transacting and automation is obvious. In the same way that the standardization and automation of financial markets allowed those markets to explode in size and scope, so too has this standardizing process enabled digital advertising markets to see unprecedented growth in the last two decades.44

  But commodification does more than allow buyers and sellers to transact more freely. It also gives rise to a set of incentives that can cause such marketplaces—under the right circumstances—to inherit some of the pathologies observed in financial markets over the past few decades. Now that it has adopted the form of these other marketplaces, a key question becomes whether online advertising, too, will face some of the same structural vulnerabilities.

  Crisis on the Way?

  In 2005, the economist Raghuram Rajan presented a paper at a private gathering of central bankers and economists in Jackson Hole, Wyoming. The paper was titled “Has Financial Development Made the World Riskier?”45 Rajan made the prescient argument that certain changes in global finance—while producing a wide range of benefits—had also created perverse incentives that posed systemic risks to the integrity of the financial sector. At the time, this paper was derided by the former U.S. Treasury secretary Lawrence Summers as fundamentally unsound.46 In hindsight, Rajan’s diagnosis correctly identified many of the factors that produced the global financial crisis just a few years later.

  Reading Rajan’s paper more than a decade after the 2008 mortgage crisis, I find striking similarities between the transition that has taken place in the world of online advertising and the transition that took place in the financial markets during the 2000s. For example, Rajan describes a shift from transactions “embedded in a long-term relationship between a client and a financial institution” to transactions “conducted at arm’s length in a market.” He characterizes this shift as a “process of ‘commodification’ of financial transactions” driven by a combination of technological, regulatory, and institutional change.47 For Rajan, this introduces a series of potentially risky incentives into the system, even as it spreads risk and expands participation in the financial markets.

  Online advertising, too, has become increasingly commodified and “at arm’s length” in its design. By and large, long-term relationships do not characterize the transactions that take place in the real-time bidding systems for allocating advertising inventory. Programmatic advertising goes on without either party knowing much about the other or having to interact person-to-person at all. Indeed, the goal of dominant players like Facebook and Google is to make buying attention on their platforms as “self-serve” and automated as possible. As in the financial markets, commodification has led to a massive increase in the size and interconnectedness of advertising markets and has allowed a much broader set of actors to participate.

  These rough parallels between advertising and finance invite deeper exploration. Like Rajan, we might ask a simple question: Has the development of online advertising made the world riskier? Are the unhealthy dynamics produced by commodification in the 2008 financial markets mirrored in online advertising markets? Perhaps most important, can we use the history of the financial markets as a guide to the future development of the internet’s economy?

  To be sure, there are substantive differences between financial markets and advertising. Advertising markets involve bidding over the right to show something to someone. Advertising is consumed at the point it is acquired, and its value is based on whether it shapes behavior in some way. That might be as concrete as persuading someone to make a purchase, or as abstract as improving someone’s opinion of the brand being promoted. In contrast, stocks are bought and sold, and profit is derived from the difference in value between markets or over time. Buyers of advertising inventory generally do not “hold” ad space to sell at a higher value later (tho
ugh attempts are being made to introduce these types of transactions in the marketplace).48

  However, the mechanisms of a market crisis do not depend on these differences. As the economists Carmen Reinhart and Kenneth Rogoff write, “Financial crises follow a rhythm of boom and bust through the ages. Countries, institutions, and financial instruments may change across time, but human nature does not.”49 Reinhart and Rogoff catalog an entire bestiary of different financial crises, from governments defaulting on their debt to faltering promises to maintain an exchange rate. The specific kind of asset does not matter; a market crisis is ultimately a crisis of confidence.

  A number of factors are handmaidens to the emergence of a full-blown market crisis. Market opacity plays a fundamental role. The inability to see what is actually happening within a marketplace allows doubt and panic about the value of an asset to set in. In the 2008 crisis, financial innovation in the form of collateralized debt obligations and complex options pricing algorithms prevented the players from having a clear idea of what was going on.50 Past financial crises in markets around the world have shown that opaque government balance sheets and finances can also trigger doubt that escalates into panic.51

  Opacity allows the value of the thing being bought and sold to deteriorate in secret, without anyone knowing. In the subprime crisis of 2007–2008, packages of shoddy mortgages that were nearly certain to default at unexpectedly high rates were increasingly circulating in the marketplace. Opacity allowed these toxic assets to trade at prices far above what they were actually worth.