Understanding the Economics of Net Neutrality

Saturday, February 6th, 2016

Whether you are new to net neutrality and want to better understand the concept or a seasoned researcher who wants an update regarding open questions, I encourage you to read a recent working paper entitled, “Net Neutrality: A Fast Lane to Understanding the Trade-Offs,” by Shane Greenstein, Martin Peitz, and Tommaso Valletti, a group of economists with a track record researching and writing about Internet economics. Although the article is rather recent, I believe it presents a very good starting point for those interested in taking a deeper dive into both specific theoretical and general empirical issues revolving net neutrality.

In this blog post, I attempt to outline the article for perspective readers and provide a few potentially useful links. Although I abstract completely from the math and intuition behind the results, the article is extremely straightforward with this regard.

A good starting point for a discussion of net neutrality begins with an understanding of the uses of the Internet. As the authors see it, there are four relevant categories of use for the Internet:

  1. Static web browsing and e-mail (low bandwidth; can tolerate delay). Data flows are largely symmetric across users.
  2. Video downloading (high bandwidth; can tolerate delay).
  3. Voice-over IP, video-talk, video streaming and multi-player gaming (high bandwidth; quality declines with delay). Data flows are mostly unidirectional from content providers to users.
  4. Peer-to-peer applications (high bandwidth; can tolerate delay; can impose delay on others).

Although much economic research tends to abstract from the technical issues revolving use of the Internet, many studies of net neutrality implicitly model the third variant above and the authors follow suit. This makes up the bulk of modern Internet traffic: for instance, together, Netflix, Youtube, and Amazon Prime have consistently made up approximately 50 percent of all North American Internet traffic as of late.

There are three common arrangements for moving data from content providers to users:

  1. Move data over “backbone lines” (e.g., Level3) and then to local broadband data carriers (e.g., ISPs) where the user is located. This may entail relying on an ISP to get to the backbone line.
  2. Move traffic to servers located geographically close to users: CDNs (e.g., Akamai).
  3. “Collocate” servers inside the network of an ISP. Payment for collocation was at the heart of negotiations between Netflix and Comcast that put net neutrality in the limelight (see also, John Oliver’s response to Tom Wheeler and my tangential reference inspired by Oliver and T-Mobile CEO John Legere).

The authors focus on two definitions of net neutrality: (1) prohibition of payment from content providers to Internet service providers (referred to as one-sided pricing whereby ISPs can only charge consumers) and (2) prohibition of prioritization of traffic with or without compensation.  As Johannes Bauer and Jonathan Obar point out, these are not the only alternatives for governing the Internet (see Bauer and Obar 2014).  In a simple world with no competition and homogeneous users, the authors suggest that net neutrality does not affect profits or consumer surplus. A number of real world considerations are taken into account, and the potential ramification of imposing net neutrality are suggested as follows.

  1. Users and content providers are heterogeneous. In this case, pressure on one side of the market (between ISPs and content providers) can lead to a corresponding change in prices on the other side of the market (between ISPs and users).
    • For instance, when content providers are identical but consumers are heterogeneous, allowing ISPs to charge termination fees to content providers can induce them to lower prices to consumers.
    • On the other hand, when content providers are heterogeneous but consumers are identical, allowing ISPs to charge termination fees can induce inefficient content provider exit.
  2. Some content providers get money from advertising (e.g., Facebook and Google), others charge users directly (e.g., Netflix).
    • The latter situation can complicate the analysis because ISP termination fees may directly impact downstream content prices.
    • The situation is further complicated if content providers can endogenize their mix of advertising and direct revenue (e.g., Pandora).
  3. Competition differs across markets, with multiple ISPs in some markets and this is relevant for studying net neutrality (see Bourreau et al. 2015). I discuss data that could be used to gauge competition in broadband provision at the end of a prior blog post.
  4. Congestion, quality of service, and network and content investment can be impacted by regulation.
    • Long term trade-offs depend on the competitive setting (e.g., horizontal competition, vertical integration).
    • Peak (termination) pricing that might be forbidden under certain forms of net neutrality could lead to welfare-enhancing congestion reducing investment.
    • Prioritization can lead to both, desirable or undesirable outcomes, and this depends on both ISP and content provider investment in congestion reduction (for instance, see Choi et al. 2014).

The authors caution against broad policy prescriptions, and rightly so, given the present ambiguity surrounding the impacts of net neutrality.  Along the way, the authors inspire a number of open empirical questions that might help policy makers.

  1. How much would allowing or eliminating termination fees affect the price charged to subscribers?
  2. Which net neutrality regulations (when in place) have been binding in practice?
  3. How do net neutrality regulations impact investment in congestion reduction?
  4. Does competition alter the need for net neutrality regulation?

I suspect that the first two questions are fairly difficult to answer from an economics perspective because in large part they depend on significant insider knowledge about contracting among market participants. The Quello staff and I are presently contemplating how to rigorously answer questions (3) and (4). We are very interested in your feedback.

Tags: , , , , ,

1 Comment

Price-Matching Advertisements and Consumer Search

Tuesday, January 12th, 2016

The only programs I watch on TV with any regularity are cooking competitions like Chopped and American Football, so I find it somewhat odd that I have seen the same Toys “R” Us commercial advertising the toy retailer’s price-matching guarantee as many times as I have.  Perhaps I have been watching too much Chopped Junior.  In the commercial, a creepy children’s toy informs Optimus Prime that if potential consumers find him for a lower price at a competing retailer, Toys “R” Us will match the lower listed price (see it here).  Optimus Prime is impressed not by the seemingly great deal, but by the existential realization that as a toy, he is not unique in our world.

When I first saw that commercial, my first though was, “I have a publication coming out about the practice of price-matching guarantees, AWWWESOME!!!” (this sentence is funnier after watching the commercial).  My next though was, “it would benefit consumers more if retailers competed by lowering prices than pretended to compete by using price-matching guarantees.” I explain below.

As my co-author, Brady Vaughan, and I show in our paper, “Price-Match Announcements in a Consumer Search Duopoly,” forthcoming in the Southern Economic Journal, although advertisements emphasizing retailers’ price-matching guarantees appear to be pro-competitive, price-matching guarantees actually tend to lessen firms’ incentives to lower price.  The intuition behind our main result is as follows.  Suppose that consumers vary in their propensity to shop around for price.  We might for instance think of individuals as valuing their scarce time differently, leading some to feel that their incremental cost of uncovering an additional sample price (i.e., by visiting an additional store), is higher than that of others.  In such a setup, firms face two competing forces when setting their prices: they are inclined to lower prices to attract consumers who tend to shop around for the lowest price, and to raise them in an attempt to take advantage of consumers who find the activity of price comparison too time consuming to bother with (see Varian 1980, Stahl 1989 for the mathematical details behind this outcome).  Price dispersion ensues: firms run sales of different magnitudes in an effort to maximize profits.

Now consider what happens when firms offer price-matching guarantees.  Suppose that those consumers who find it worthwhile to shop around literally go store to store in search of price (this is not the only way to explain our results, but we find it to be one that aides intuition).  Some of these consumers will end their search at a store that does not list the lowest price.  Without a price-matching guarantee, if these consumers wish to procure the good at the lowest possible price observed, they will have to go back to the store with the best offer.  But if the last store they visit offers a price-matching guarantee, they can get the lowest price there instead.  Knowing this, firms realize that they won’t be able to win over as many price conscious consumers with deep discounts, so their incentive to run sales diminishes, leading to higher average prices.

Sound like a roundabout explanation?  Brady and I are far from the first to suggest that price-matching guarantees can diminish competition and some of the earliest explanations seem downright obvious: price-matching guarantees keep firms from lowering prices because rival firms immediately match price-cuts (see Hay 1982, Salop 1986, Doyle 1988).  However, I think that Brady and I have made a fairly cogent argument that takes into consideration how consumers behave and also accounts for the myriad advertisements firms undertake to inform consumers about their policies (here is one from Walmart, another from Toys “R” Us, and one from Staples).

This result begs the question, (i) have price-matching guarantees always been found to be anti-competitive and (ii) if so, can the anti-trust authorities reasonably do anything to prevent them?  The answer to question (i) is a no.  Although the bulk of the literature lends support to our findings, there are some notable explanations suggesting the contrary.  One that I find somewhat convincing when firms are differentiated is that price-matching guarantees can be a signal that a firm generally has lower underlying costs (perhaps it has negotiated better deals with merchants or doesn’t spend as much on its service quality) and consequently sets lower prices (see Moorthy and Winter 2006, Moorthy and Zhang 2006 for the details).  That is a theoretical argument.  The empirical literature is somewhat mixed, but typically strays to answer the question, “do firms that price-match have higher prices than those that do not,” instead of “are prices in general lower or higher when price-matching guarantees are used by some firms in the market?” More empirical research is ne20160109_145051eded to settle the issue.

As for question (ii), the answer may be no as well.  The anti-trust laws, stemming from the Sherman and Clayton Acts are generally focused on restraints of competition between firms, but price-matching guarantees are effectively standing offers by firms to contract with a consumer by referencing another firm’s price (see Edlin 1997).  Contracts that reference rivals are assuredly of concern to anti-trust practitioners (Scott Morton 2013), but when the contract does not impose any restriction on any party except a commitment to lower price by the firm offering it in response to publicly available information, it would seem (without undertaking a very rigorous empirical examination of the case at hand) rather difficult to make a case that competition is being restrained.

All of this comes with a major caveat.  Although I believe that price-matching guarantees have the potential to lead to higher prices in the market as a whole, if as a consumer, you find yourself in a situation where you can use a price-matching guarantee to save money, by all means do!  Unless all consumers can coordinate with all other consumers to bring about a better situation for themselves, they should do what is in their individual best interest.  I recently visited my family for the holidays and we decided to buy a board game to spend the time.  My brother reminded me to put my research to work.  I saved 20 bucks!

Tags: , , , , ,


Aleks Yankelevich’s First Blog Post (Chipotle, Market Definition, and Digital Inequality)

Wednesday, December 2nd, 2015

Growing up, my parents, brother, and I usually avoided restaurants. For my parents, this was initially out of necessity; as Soviet refugees, they did not have the financial means to eat out. However, even having achieved a modicum of success, my parents are not generally in the habit of frequenting restaurants, having perhaps out of a lifetime habit, developed a taste for home cooking. Restaurants are exclusively for special occasions.

Thus, having never eaten at a Chipotle Mexican Grill, they were sufficiently impressed by the restaurant’s façade to wish to eat there, but only when the grand occasion merits such an extravagant excursion. Their two sons were informed as such. Naturally, my brother and I (perhaps spoiled as we are) jumped at the chance to poke fun at our parents for placing Chipotle on a pedestal. This is, after all, a restaurant chain that is victim to some serious defecation humor, not Eleven Madison Park.

For a number of months, my parents were subjected to text messages and Facebook or Instagram posts with visuals of me or my brother outside various Chipotle restaurants, posing next to Chipotle ads, and in one instance, wearing a Chipotle t-shirt (I have no idea how that shirt found its way into my wardrobe). My parents responded, saying things like (and I could not make this up), “I wish someone would take us to that dream place.”

However, recently, my mother sent a group text directing the family to a news report about dozens of confirmed E.Coli cases related to Chipotle (even the FDA got involved) and asking for alternative dining suggestions. The text responses, in order, were as follows:

Me: California Tortilla
My Wife: Taco Bell
My Brother: Sushi
My Mother: Eating In (with picture of latest home cooked meal)
My Brother’s Girlfriend: Bacon

How does a reasonable individual interpret this chain of responses? As an economist with some regulatory and antitrust experience, I found the answer obvious. I sent the following group text (modified for concision): “Has anyone noticed that this text conversation has turned into the classic antitrust debate about appropriate market definition, with each subsequent family member suggesting a broader market?”

Surprisingly, no one else had noticed, but I was asked to unpack my statement a little bit (my mom sent a text that read: “English please.”).

The U.S. Department of Justice and the Federal Trade Commission’s Horizontal Merger Guidelines stipulate that market definition serves two roles in identifying potential competitive concerns. First, market definition helps specify the line of commerce (product) and section of the country (geography) in which a competitive concern arises. Second, market definition allows the Agencies to identify market participants and measure market shares and concentration.

As the Agencies point out, market definition focuses solely on demand substitution factors, i.e., on customer’s ability and willingness to substitute away from one product to another in response to a price increase or a corresponding non-price change (in the case of Chipotle, an E.Coli outbreak might qualify as a reduction in quality). Customers generally face a range of potential substitutes, some closer than others. Defining a market broadly to include relatively distant substitutes can lead to misleading market shares. As such, the Agencies may seek to define markets to be sufficiently narrow as to capture the relative competitive significance between substitute products. For some precision with this regard, I refer the reader to Section 4.1.1 of the Guidelines.

As for the group texts above, the reader can now infer how market definition was broadened by each subsequent family member. To reiterate:

Me: California Tortilla (Mexican food in a similar quality dining establishment to Chipotle.)
My Wife: Taco Bell (Mexican . . . inspired . . . dining out, generally.)
My Brother: Sushi (Dining out, generally.)
My Mother: Eating In (Dining, generally.)
My Brother’s Girlfriend: Bacon (Eating.)

Why is market definition relevant to the Quello Center at Michigan State University? As the Center’s website suggests, the Center seeks to stimulate and inform debate on media, communication and information policy for our digital age. One area where market definition plays a role with this regard is within the Quello Center’s broad interest in research about digital inequality.

Digital inequality represents a social inequality with regard to access to or use of the Internet, or more broadly, information and communication technologies (ICTs). Digital inequalities can arise as a result of individualistic factors (income, age and other demographics) or contextual ones (competition where a particular consumer is most likely to rely on ICTs). Market definition is most readily observed in the latter.

For instance, consider the market for fixed broadband Internet. An immediate question that arises is the appropriate geographic market definition. If we rule out individuals’ ability to procure fixed broadband Internet at local hotspots (e.g., libraries, coffee shops) from the relevant market definition, then the relevant geographic market appears to be the home. This is unfortunately a major burden for researchers attempting to assess the state of fixed broadband competition and its potential impact on digital inequality because most market level data in use is at a much more aggregated level than the home. The problem is that when an aggregated market, say a zip code, contains multiple competitors, it is unclear how many of these competitors actually compete in the same home.

Thus far, most studies of fixed broadband competition have been hampered by the issue of geographic market definition. For instance, Xiao and Orazem (2011) extend Bresnahan and Reiss’s (1991, 1994) classic studies of entry and competition in the market for fixed broadband, albeit at the zip code level. Wallsten and Mallahan (2010) use tract level FCC Form 477 data to test the effects of competition on speeds, penetration, and prices. However, whereas there are approximately 42,000 zip codes and 73,000 census tracts in the United States, there are approximately 124 million households, which implies a fairly large amount of aggregation that can lead researchers to conclude that competition is stronger than it actually is.

Another question that arises is whether fixed broadband is too narrow a product market and if the appropriate market definition is simply broadband, which would include fixed as well as mobile broadband. Thus far, because of data limitations, most studies of wireline-wireless substitution have focused mainly on voice rather than on Internet use (e.g. Macher, Mayo, Ukhaneva, and Woroch, 2015; Thacker and Wilson, 2015) and so do not assess whether mobile has become a medium that can mitigate digital inequality. Prieger (2013) has made some headway into this issue by showing evidence that as late as 2010, mobile and fixed broadband were generally not complementary, and that mobile only broadband subscription was slightly more prevalent in rural areas. However, because of data limitations, Prieger does not estimate a demand system to determine whether fixed and mobile broadband are substitutes or complements as the voice substitution papers above do.

Luckily, NTIA’s State Broadband Initiative (SBI) and more recently, the FCC, have enhanced researchers’ ability to assess competition at a fairly granular level by providing fixed broadband coverage and speed data at the level of the census block. Similarly, new data on Internet usage from the U.S. Census should allow researchers to better tackle the wireline-wireless substitution issue as well. The FCC has also hopped on the speed test bandwagon by collaborating with SamKnows to measure both fixed and mobile broadband quality. In the former case, the FCC periodically releases the raw data and I am optimistic that at some point, mobile broadband quality data will be released as well (readers please correct me if I am glossing over some already publically available granular data on mobile broadband speed and other characteristics).

The Quello Center staff seeks to combine such data, along with other sources, to study broadband competition and its impact on digital inequality. We welcome your feedback and are presently on the lookout for potential collaborators interested in these issues.


Tags: , , , , , ,