title:: Why Google Dominates Advertising Markets url:: author: [[@Dina Srinivasan]] output: tags:: topic: # Notes * Ad markets - increased concentration + distorted growth, trading costs of 30-50% of the trade, opaque with conflicts of interest * ## Concentration * C1: Google and Facebook get majority ad revenue ## Distorted Growth ## Trading costs * T1: ## Opaque * O1: Advertisers do not know what price an ad spot was purchased for. Scope of arbitrage between this value and what they bid on it. * O2: Google may also be a counter party. ## Conflict of interest * Co1: Google's properties share of revenues from Google ad exchanges increased from 64% in 2007 to 85% in 2020. * Co2: Google and Facebook operate ad exchanges and sell their own ad space on it * Co3: From O2 - (as a counter party through its exchange it can drive up costs?) * Co4: Intermediaries can direct customers to transactions which they gain from or in a generally discriminatory manner. ## Policy Intervention 1. Fair access to information and speed. 2. Identfying and Managing Intermediary Conflicts of Interest 1. Structural separations: 2. Conduct and Disclosure Rules 3. Transparency and Disclosure (pushback on privacy, regulators in other industries manage this. Though, scale of data might be different) Publisher -> Intermediary/Ad Server -> Exchange/Trading Venue <- Intermediary/Buying Tools <- Adverstiser - *summary of diagram from p 77* ## Exchanges * An exchange can distort a market by giving some participants an information or speed advantahe * Information: Identity, etc. * Speed: colocation to reduce transit times. ## Trading Intermediaries * Trading intermediaries on both sides can choose which exchanges to route bids or responses to. Ones with significant market shares can distort competition among exchanges. * Information issue when an intermediary handles data for others and also participates in the market for its own financial interests. ## Priniciples from Equities - Fair access to the marketplace including data and speed. - Identification and management of intermediary conflicts of interest. - structural: a company running an exchange cannot also trade. If they are, then rules about disclosures and conflicts of interest kick in - Rules regulate who can use what information while trading. Property rights over customer information is with the customer, and cannot be used by intermediary for its own gain. > * Collectively, these principles—those prohibiting intermediaries from using certain information advantages when trading, those prohibiting in- termediaries from preferentially routing order flow, and those requiring exchanges to provide traders with fair access to information and speed— help to protect competition in the electronically traded equities market.* ## Google engaging in conduct prevented in other electronic trading markets ### Information and Speed advantages - DoubleClick acquisition: Information about publisher's users (cookie set by the exchange). DC was the largest exchange. - Restricting the ability to publishers and adverstisers to access ad server IDs for privacy reasons while allowing its exchange and buying tools to access them by default, which gave an advantage to Google's exchanges. Since only it had the ability to identify users with the same ID. (hashes were used to ensure different publishers did not get the same user id) - This position of advantage for its exchange also may have helped leverage it for its intermediaries (servers and buying tools), since its intermediaried access to these IDs, but other did not. - To work around this, non google intermediaries need to sync cookies with google's exchange, which added latency and did not always match. Without identifying information, bidders tend to bid less or not at all. - Privacy as reason for information asymmetry: but google does it claims to prevent others from doing. Prevents companies from combining google data contractually. - Speed advantage: By virtue of colocation - less time to travel and thus also, more time to query, analyse. ### Discriminatory Routing and Speed - Intermediaries can route orders in a discriminatory manner. #### steering and tying to google properties - New account had to buy google search ads before other publishers (steering). - Advertisers need to use a Google owned intermediary to be able to bid for spaces on YouTube or Google properties. Youtube ad space is withheld from rivals, unsure about Search. (tying) - (steering and tying) Scrambling IDs creates a disincentive for using multiple bidding tools since they could end up bidding against one another (thus pushing advertisers to use one tool). Only 15% of advertiser spends on Google when to non Google properties as of Q1 2020. In 2019, it was 16% of USD 134B (does that correlate to traffic. Ext. src. in 2020, 64% of Google searches ended without a click through, based on SimilarWeb data. Link: https://sparktoro.com/blog/in-2020-two-thirds-of-google-searches-ended-without-a-click/) #### steering and trying to google's exchange - Intermediaries steer buyers and sellers to google's exchange. Until 2016, Google ads only routed to Google Exchange. Apparently, so did its enterprise tool DV 360. Now, though they work with other exchanges, information and speed advantages mean they may still route to Google Exchange. On Publisher side, intermediary used to prevent them from using multiple servers for their inventory at the same time - it was sequential (maybe that was just the technology?) - Exchange competition if publishers put their inventory on google first. This relied on static floor prices when much of market had moved to dynamic bidding - Potential welfare loss as it could have resulted in publishers making less revenue as lower priced bids through Google's intermediary may have won out. - To circumvent sequential placement of inventory, the market came up with header-bidding (javascript code in the header of an html page which allowerd publishers to put their inventory on multiple exchanges simultaneously). Also allowed them to allocate more time for exchange since the request went out first (could have impacted perf though). But ad revenue reportedly went up once this was implemented. Again, surplus angle. - AMP, for faster loading, but limited header bidding since it restricted the use for js. AMP was framed as a way to maximise consumer welfare. But if revenues dropped, they would ultimately have less to invest into content. Google pushed AMP on Google News, visual indicators on Google search. Publishers had to trade-off. - Server-side bidding was a workaround to this. However, this meant auctions moved out of browsers to exchange servers. This re-introduced information and speed asymmetries through more restrictions and cookie-sync issues leading to less identificationa and lower bids. - Publishers had to pay extra transaction costs to use Google ad servers to transact on non-google exchange. 'Unified Pricing' prevented publishers from listing inventory on different exchanges at different floor prices. The use of higher floor prices was because google networks could identify users and therefore received higher bids. - In 2018, Google started penalising slower mobile web pages in search and steered them to AMP - another impediment to header bidding. ### Information abuses - Since Google did not participate in header bidding, publishers needed to run header-bidding and google's ad server sequentially. The paper asserts that the winning bid was passed to Google's ad server and then to Google's exchange, and then its own bidders - who could outbid by a penny. This happened between 2015-2019 when it had a second-price auction, meaning it would get the 'last look'. - When it did stop passing price to beat, around the same time it start rounding off timestamps to hours (from microseconds). - (Changing Ts and Cs) In 2012, GOogle amended its terms and conditions to merge data from DoubleClick and other business units. In 2016, went further to include all profiling data they may have. In 2018, also using data from Chrome. - This combination is viewed as a privacy harm but it also competition impliciations since it means that it can sell ads more effectively on its own properties - using data gleaned from other websites. - Unfair + violation of property rights. ## Debates * Do network effects and economies of scale lead to concentration in big-tech markets. p 66 * Move beyond scope of antitrust endforcement. p 66 * Structural separations - no player and referee. p 66 # Further References